Cholesterol Treatment: A Review of the Clinical Trials Evidence (Letter
Report, 05/14/96, GAO/PEMD-96-7).

Pursuant to a congressional request, GAO reviewed the evidence from
clinical trials that the National Heart, Lung, and Blood Institute
(NHLBI) used to develop its National Cholesterol Education Program
guidelines.

GAO found that: (1) meta-analyses of trial data consistently show that
cholesterol treated persons, regardless of their medical history, have
significantly fewer non-fatal heart attacks than untreated persons; (2)
treated persons also showed a reduction in the number of fatal heart
attacks compared to the nontreated group, but the difference was not
statistically significant except among those who had a history of
coronary heart disease (CHD); (3) according to one trial, cholesterol
treatment has not led to a reduction in deaths from all causes; (4) the
increase in deaths from other causes shown in the trials occurred
primarily among persons whose risk for CHD was lower, whose cholesterol
was reduced less, or who used certain drugs; (5) the two trials that
used newer cholesterol-lowering drugs confirmed the finding that the
more cholesterol levels were lowered, the fewer coronary events
occurred; (6) previous trials were not representative of the population
at large, since they focused mainly on middle-aged white men at high
risk for CHD; (7) several clinical trials now under way are designed to
provide additional information about treatment outcomes regarding total
fatalities, persons with a moderate short-term risk for a coronary
event, and the longer-term effects of the newer drugs; and (8) these
trials are large and open to a broader range of participants, but
whether they will provide broader information will depend on their
actual enrollments.

--------------------------- Indexing Terms -----------------------------

 REPORTNUM:  PEMD-96-7
     TITLE:  Cholesterol Treatment: A Review of the Clinical Trials 
             Evidence
      DATE:  05/14/96
   SUBJECT:  Medical research
             Drugs
             Cardiovascular diseases
             Nutrition research
             Statistical methods
             Health hazards
             Public health research
             Educational programs
             Pharmacological research
             Women
IDENTIFIER:  NHLBI National Cholesterol Education Program
             
******************************************************************
** This file contains an ASCII representation of the text of a  **
** GAO report.  Delineations within the text indicating chapter **
** titles, headings, and bullets are preserved.  Major          **
** divisions and subdivisions of the text, such as Chapters,    **
** Sections, and Appendixes, are identified by double and       **
** single lines.  The numbers on the right end of these lines   **
** indicate the position of each of the subsections in the      **
** document outline.  These numbers do NOT correspond with the  **
** page numbers of the printed product.                         **
**                                                              **
** No attempt has been made to display graphic images, although **
** figure captions are reproduced.  Tables are included, but    **
** may not resemble those in the printed version.               **
**                                                              **
** Please see the PDF (Portable Document Format) file, when     **
** available, for a complete electronic file of the printed     **
** document's contents.                                         **
**                                                              **
** A printed copy of this report may be obtained from the GAO   **
** Document Distribution Center.  For further details, please   **
** send an e-mail message to:                                   **
**                                                              **
**                                            **
**                                                              **
** with the message 'info' in the body.                         **
******************************************************************


Cover
================================================================ COVER


Report to the Ranking Minority Member, Committee on Science, House of
Representatives

May 1996

CHOLESTEROL TREATMENT - A REVIEW
OF THE CLINICAL TRIALS EVIDENCE

GAO/PEMD-96-7

Cholesterol Treatment

(973389)


Abbreviations
=============================================================== ABBREV

  4S - Scandinavian Simvastatin Survival Study
  ACP - American College of Physicians
  ALLHAT - Anti-hypertensive and Lipid-Lowering Treatment to Prevent
     Heart Attack
  CHD - Coronary heart disease
  HDL - High-density lipoprotein
  LDL - Low-density lipoprotein
  NCEP - National Cholesterol Education Program
  NHLBI - National Heart, Lung, and Blood Institute
  NIH - National Institutes of Health
  WOSCOPS - West of Scotland Coronary Prevention Study

Letter
=============================================================== LETTER


B-266080

May 14, 1996

The Honorable George E.  Brown, Jr.
Ranking Minority Member
Committee on Science
House of Representatives

Dear Mr.  Brown: 

The death rate from coronary heart disease (CHD) declined
dramatically during the past decade, but CHD is still the most
prevalent cause of death in the United States.  Public health efforts
to prevent heart disease have emphasized reducing risk factors such
as smoking, hypertension, and elevated blood cholesterol levels.  The
National Institutes of Health (NIH) established the National
Cholesterol Education Program (NCEP) in 1985 to encourage Americans
to modify their diet and lifestyle and to provide clinical guidelines
to help health care professionals identify and treat persons whose
risk of heart disease is high because their cholesterol levels are
high.  The NCEP guidelines suggest that 52 million American adults
are candidates for dietary treatment and that 9 million to 18 million
of these have cholesterol levels that are sufficiently high to
warrant drug treatment. 

High cholesterol is now widely recognized by the scientific community
and the general public as an important risk factor for heart disease. 
Physicians and the public are generally aware of the need to lower
cholesterol, and treating patients by lowering their cholesterol has
met with some success.  National survey data indicate that the
proportion of adults whose cholesterol is high (greater than or equal
to 240 mg/dl) declined from 26 percent to 20 percent between 1976 and
1991.\1 The NCEP guidelines have contributed to this decline, but key
aspects of them have been debated since the mid-1980s.  A number of
critics have questioned whether the available scientific evidence
clearly demonstrates that the benefits of treatment outweigh any
possible health risks and whether the benefits are broad enough to
support widespread screening and treatment of the population as the
NCEP guidelines advocate. 

Concerned about the translation of medical knowledge into national
policy and about the high costs of cholesterol treatment, the former
Subcommittee on Investigations and Oversight of the House Committee
on Science, Space, and Technology asked us to review the clinical
trials evidence that was used to support the development of the NCEP
guidelines.  The guidelines were developed from a broad set of
research that included not only clinical trials but also
epidemiological, animal, pathologic, and genetic studies.  We were
asked to review the clinical trials because they are uniquely
designed to assess the effectiveness of interventions to reduce the
incidence of a disease. 

The equivocal results of the many individual clinical trials that
have been conducted over the past 30 years, however, have made it
difficult to determine the efficacy of cholesterol-lowering
treatments in preventing CHD.  Several trials have shown clear
reductions in heart disease rates for persons who were treated, but
many trials showed no improvement either because the types of
treatment were not particularly effective in lowering cholesterol or
because the studies were not large enough to adequately measure the
expected benefits. 

Recently completed studies called meta-analyses, which quantitatively
combine the results of many individual trials, provide new insights
into the interpretation of the overall benefits and risks of
cholesterol-lowering treatments.  Therefore, we examined the benefits
and risks of lowering cholesterol reported from meta-analyses of
available clinical trials.  We report these and the results of two
recently completed clinical trials that used newer, more effective
drug treatments that were not included in the published
meta-analyses.  These trials provide important new evidence regarding
cholesterol treatment and the reduction of CHD.  In addition, we
assessed the extent to which the trials provide information on
different population subgroups, and we determined whether the
clinical trials that are currently planned are likely to fill any
data gaps that exist. 


--------------------
\1 C.  T.  Sempos et al., "Prevalence of High Blood Cholesterol Among
U.S.  Adults," Journal of the American Medical Association, 269
(1993), 3009-14. 


   RESULTS IN BRIEF
------------------------------------------------------------ Letter :1

Meta-analyses that combine previous clinical trial findings
consistently show that persons who receive cholesterol-lowering
treatment, regardless of whether or not they have a history of heart
disease, have significantly fewer nonfatal heart attacks than persons
who do not receive treatment.  The evidence on CHD fatalities (deaths
attributable to CHD) also shows a modest treatment benefit, but the
meta-analyses show that it is found mainly in the group of trials
that assessed persons who already had CHD.  A reduction in CHD
fatality rates is also found for persons who had high cholesterol and
no history of CHD; however, the differences between treatment and
nontreatment groups are not statistically significant. 

With respect to total fatalities--that is, deaths from CHD and all
other causes--most meta-analyses show no significant difference and
thus no improvement in overall survival rates in the trials that
included either persons with known CHD or persons without it.  This
is partly because of an increase in non-CHD deaths across the trials. 
This finding, that cholesterol treatment has not lowered the number
of deaths overall, has been worrisome to many researchers and is at
the core of much of the controversy on cholesterol policy. 

Various researchers have attributed the increase in non-CHD deaths to
chance, to the biological means by which cholesterol is lowered, to
the short duration of the trials, or to the side effects of some
treatments.  However, several recent meta-analyses have identified
important factors that appear to help explain it.  They show, for
example, that the overall death rate fell in the trials that included
persons whose risk for CHD was highest and in the trials in which
cholesterol reductions were large.  The overall death rate rose
significantly among persons whose risk of CHD was lower, whose
cholesterol was reduced less, or whose treatment used certain drugs. 

The finding that reductions in coronary events are greater with
greater amounts of cholesterol reduction is confirmed by two trials
completed too recently to be in the meta-analyses we reviewed.  Using
a new class of cholesterol-lowering drugs--the statins, or HMG CoA
reductase inhibitors--they lowered cholesterol 20 to 25 percent. 
Previous trials using other drugs lowered it 10 percent, on the
average.  One of the trials, conducted in Scandinavia on persons who
already had had a heart attack, found that both CHD outcomes and
total deaths fell significantly for those who received treatment. 
The second trial, conducted in Scotland on persons who had no
previous history of heart disease, also found reductions in coronary
events, although the reductions in coronary deaths and total deaths
did not quite achieve statistical significance. 

While the existing cholesterol trials provide important results about
CHD outcomes, they are not representative of the population at large. 
The trials focused predominantly on middle-aged white men considered
to be at high risk of CHD.  They provide very little information on
women, minority men and women, and elderly men and women.  The few
trials that opened enrollment to these subgroups generally did not
have large enough numbers of them to conduct separate analyses.  What
is known about these groups has been obtained from other types of
research that cannot be used to make causal inferences about how
lowering cholesterol affects coronary outcomes. 

Several new clinical trials under way are intended to provide
additional information about treatment outcomes regarding total
fatalities, persons whose short-term risk of a coronary event is
moderate, and the longer-term effects of the newer drugs being used. 
These trials will be large and open to participants from a broader
portion of the population than previously studied.  However, the
numbers actually enrolled in many of these trials have not yet been
determined, so that the extent to which the trials will provide
information on groups other than middle-aged white men is not yet
known. 


   BACKGROUND
------------------------------------------------------------ Letter :2

Nearly half a million persons die from heart disease each year.  As
many women as men die from CHD, although women die about 7 to 10
years later than men on the average.  The American Heart Association
estimates that heart disease costs $55 billion or more each year. 
Taken together, these figures provide a strong rationale for
prevention and treatment. 

Many factors can contribute to the development of CHD.  A heart
attack--a temporary interruption of blood flow to the heart-- stems
partly from the chronic buildup of fatty plaque in the arteries.  For
example, as figure 1 shows, there is a continuous and upward-sloping
association between higher cholesterol levels and the incidence of
CHD.  Some of the best-known risk factors, such as high blood
cholesterol, cigarette smoking, diabetes, obesity, hypertension, and
physical inactivity, can be treated or modified; others, such as age
and family history of heart disease, cannot. 

   Figure 1:  CHD Deaths by
   Cholesterol Level Among Men
   Screened for the MRFIT Trial\a

   (See figure in printed
   edition.)

\a Deaths over 6 years among more than 355,000 men screened in the
early 1970s. 


      THE NATIONAL CHOLESTEROL
      EDUCATION PROGRAM
---------------------------------------------------------- Letter :2.1

National Heart, Lung, and Blood Institute (NHLBI) education campaigns
address several of the modifiable risk factors, including high blood
cholesterol levels, smoking, obesity, and hypertension.  As a result
of an NIH consensus development conference in 1984, NHLBI established
NCEP to inform the public of the risks associated with cholesterol
and to provide guidelines for physicians on how to manage and reduce
those risks.  Accordingly, NCEP both promotes the reduction of
cholesterol levels in the general population through lifestyle
changes and encourages the identification of persons whose
cholesterol is high as candidates for intensive treatment to lower
it. 

In the latter, high-risk approach, the NCEP guidelines recommend that
all adults be tested for cholesterol at least once every 5 years. 
Test results are used to classify each person free of CHD into one of
three groups:  desirable, borderline-high, or high total cholesterol
levels.\2 Taking other risk factors into consideration, physicians
then identify candidates for further testing (to determine
low-density lipoprotein (LDL) cholesterol levels) and
cholesterol-lowering treatment.  The first line of intervention
against high cholesterol levels is diets that emphasize lowering the
consumption of fat (particularly saturated fat) and dietary
cholesterol.  Prescription drugs are advised if cholesterol goals are
not met through diet.  For most patients, treatment is maintained
over the course of a lifetime in order to keep cholesterol levels
down. 

NCEP revised several elements of the guidelines for adults in 1993,
principally by (1) including age with regard to gender in determining
risk status and in making treatment decisions; (2) increasing the
emphasis on screening and treating persons whose risk of CHD is high,
such as elderly persons and those who already have evidence of CHD;
and (3) recognizing the importance of high-density lipoprotein (HDL)
cholesterol as an independent risk factor for CHD.\3


--------------------
\2 Desirable is below 200 mg/dl, borderline high is 200-239 mg/dl,
and high is equal to or above 240 mg/dl.  Approximately 50 percent of
the U.S.  adult population has total cholesterol levels of 200 mg/dl
or higher. 

\3 For example, the guidelines no longer recommend drug treatment for
men younger than 35 and premenopausal women, both of whom tend to
experience lower rates of CHD.  HDL cholesterol, a component of total
cholesterol, is believed to help remove excess cholesterol from the
blood and prevent the buildup of plaque.  A higher level of HDL
cholesterol is associated with a lower risk of heart disease, while a
lower level of HDL cholesterol can increase the risk of heart
disease. 


      THE CONTROVERSY ON
      CHOLESTEROL POLICY
---------------------------------------------------------- Letter :2.2

Debate after the NCEP guidelines were issued focused on the efficacy
of lowering cholesterol, on its possible risks, and on the
advisability of screening for cholesterol and lowering it among the
many population subgroups that had not been included in the trials
the guidelines had been based on.  Additional, more recent data have
brought greater agreement that cholesterol treatment helps prevent
deaths among persons who have a history of CHD.  Controversy remains,
however, on how useful cholesterol-screening is to the broader
population of healthy people.  The American College of Physicians,
professionally representing 85,000 U.S.  internists, recently
published guidelines under which groups for whom
cholesterol-screening has no proven benefit would not be screened. 

While there is considerable agreement about the risks associated with
high cholesterol, concerns have been raised about the extent to which
cholesterol-lowering treatments reduce the rates of death from CHD
and from all causes.  Some researchers argue that the clinical trials
have shown no difference in total fatalities between groups of people
who received treatment and those who did not and that
cholesterol-lowering treatments may increase the number of deaths
from non-CHD causes.  Advocates of the NCEP guidelines argue that the
trials were not designed to be large enough or long enough to assess
overall mortality; they say that the increases in rates of non-CHD
deaths within the treatment groups are likely to have resulted from
chance or specific treatments rather than from any adverse effects of
lowering cholesterol. 

Because the clinical trials focused almost exclusively on men who had
either existing CHD or high cholesterol or both, they provide little
information about anyone whose risk of heart disease is lower.  Some
researchers and health policy analysts therefore suggest that the
U.S.  guidelines should be more limited in scope, as they are in some
other nations.  Advocates of the current guidelines, however, believe
that some trial data and other, nonclinical trial research, such as
epidemiological and laboratory studies, allow the generalization that
if cholesterol- lowering benefits the tested groups, it will benefit
others. 


   OBJECTIVES, SCOPE, AND
   METHODOLOGY
------------------------------------------------------------ Letter :3

As we stated previously, the main objective of our study was to
examine the NCEP guidelines' clinical-trials base of evidence
regarding cholesterol-lowering treatment.\4 We limited our scope to
mainly randomized clinical trials that, by virtue of their design
features, provide the best evidence for assessing the effectiveness
of interventions to reduce the incidence of a disease.  Other
important research was also used in developing the NCEP guidelines. 
For example, many epidemiological studies have compared population
groups across nations, monitored the health status of entire
community populations, and assessed changes in the health outcomes of
recent immigrants.  These studies have increased knowledge about the
link between diet, elevated cholesterol levels, and increased risk of
CHD.  Because we did not review other types of cholesterol-related
research, we did not evaluate or report the overall adequacy or
utility of the scientific support for the guidelines. 

We reviewed the clinical outcomes of cholesterol-lowering trials
through the published results of 15 meta-analyses.\5 The
meta-analyses, which statistically summarize the results of a number
of individual trials, give an overall assessment of treatment
outcomes.  By combining the results of several trials, meta-analyses
can increase statistical power and the precision of results.  Within
certain methodological limits, they also provide a means of comparing
different subgroups of trials and of exploring issues in ways that
individual trials, which may be constrained by their design or
intent, cannot. 

The many clinical trials that have evaluated cholesterol-lowering
treatments represent diverse types of treatment, health
characteristics, overall sample sizes, duration of trial and
follow-up, and outcomes.  The main treatments consisted of different
diets and drugs (alone or in combination).  Some studies focused on
persons who had existing symptoms of CHD (secondary prevention
trials); others concentrated on those who had no evidence of CHD
(primary prevention trials); a few combined both.  Although virtually
all the trials were designed to assess CHD outcomes, not all measured
and reported the same outcome variables. 

The meta-analyses that combined the numerical results of these trials
are also diverse.  They used different rules for deciding what trials
to include and exclude; in fact, all excluded one type of trial or
another.  Most of the meta-analysts, for example, excluded trials
that used more than one intervention to treat persons who had
multiple risk factors, and most also excluded the hormone trials
because the type of hormone treatment they used is no longer
recommended for men.  As a result, the meta-analyses included as few
as 6 to as many as 35 trials. 

To determine the benefits and risks of cholesterol-lowering
treatments, we present the results of the meta-analyses.\6 We
examined the similarities and differences among them, in terms of the
individual trials they included and their analyses.  We then compared
reported odds ratios (a measure of the relative difference in
outcomes between persons who were treated and those who were not
treated in the trials) and determined the extent to which there was
agreement or disagreement in the direction and statistical
significance of the results. 

To address the issue of the coverage and comprehensiveness of the
clinical trials data, we identified the demographic and risk-related
characteristics of participants in trials included by meta-analysts,
the types of treatment they received, and the outcomes.  In the same
vein, to identify the important data gaps that others have found, we
reviewed NHLBI documents, medical literature, and other sets of
guidelines developed elsewhere.  We also identified and reviewed the
designs of ongoing and planned trials in order to determine which
data gaps may be filled in the near future.\7


--------------------
\4 In a clinical trial, participants are screened and randomly
assigned to a group that receives treatment or to a group that does
not, which allows an unbiased comparison of outcomes between the two
groups. 

\5 We selected meta-analyses that aggregated data from
cholesterol-lowering trials across one or more of the following
health outcomes:  nonfatal heart attacks, CHD fatalities, CHD
incidence (nonfatal and fatal events combined), non-CHD fatalities,
and total fatalities.  Not all the meta-analyses, however, reported
on each of these outcomes. 

\6 We did not conduct our own meta-analysis because each of the
meta-analyses that we selected to review responded to the objectives
of our study.  The results of the different meta-analyses are not
statistically independent because many of the larger clinical trials
tend to be included in most of the meta-analyses.  As a result, even
though there is diversity among the meta-analyses, some convergence
of findings is to be expected. 

\7 Appendix I describes in detail our scope and methodology and their
strengths and limitations.  In appendixes II and III, we summarize
our findings from the meta-analyses and discuss the data gaps in the
clinical trials.  Appendix IV describes U.S.  cholesterol policy in
the United States and abroad. 


   PRINCIPAL FINDINGS
------------------------------------------------------------ Letter :4


      THE BENEFITS AND RISKS OF
      LOWERING CHOLESTEROL
---------------------------------------------------------- Letter :4.1


         NONFATAL AND FATAL CHD
         OUTCOMES
-------------------------------------------------------- Letter :4.1.1

The meta-analyses we reviewed consistently reported a statistically
significant reduction in the rates of nonfatal heart attacks for the
trial participants who were treated for high cholesterol compared to
those who did not receive treatment.  As shown in figure 2, this
finding holds true for both individuals without existing evidence of
CHD--that is, in the primary prevention trials--and those with
it--that is, in the secondary prevention trials.\8

   Figure 2:  Nonfatal Heart
   Attacks Reported in
   Meta-Analyses of
   Cholesterol-Lowering Trials\a

   (See figure in printed
   edition.)

\a The horizontal bars represent estimates (at the 95-percent
confidence interval) of the difference in the odds of nonfatal heart
attacks occurring in the treatment and nontreatment groups studied in
meta-analyses.  At 1, the center of the figure, treatment and
nontreatment groups show no difference in the likelihood that
participants in either group will have a nonfatal heart attack.  Bars
to the left of 1 indicate that treatment leads to a decrease in the
rate of nonfatal heart attack; bars to the right, that it leads to an
increase.  Bars that overlap 1 indicate that differences between
treatment and nontreatment groups are not statistically significant. 
The line at the center of each bar represents the common odds ratio,
or the point estimate of that difference (that is, it is the ratio of
the odds of nonfatal heart attacks occurring in the treatment groups
to the odds of nonfatal heart attacks occurring in the nontreatment
groups).  (Appendix II, table II.1, details the numerical estimates
for each meta-analysis.)

The results that the meta-analyses reported for deaths from CHD
present a somewhat different picture, shown in figure 3.  All the
meta-analyses reported a reduction in the rate of coronary death
among the participants who were treated compared to those who were
not but not as great a relative reduction as that in nonfatal
outcomes.  Also, a different pattern is apparent for the death rate
of participants who were treated and had a history of CHD compared to
participants who were treated and did not.  For those who had a
history, most of the meta-analyses found a significant reduction
favoring the treatment groups.  For those who did not, the
meta-analyses that examined CHD deaths point in the direction of a
small reduction but none found it statistically significant.\9

   Figure 3:  CHD Fatalities
   Reported in Meta- Analyses of
   Cholesterol-Lowering Trials\a

   (See figure in printed
   edition.)

\a The horizontal bars represent estimates (at the 95-percent
confidence interval) of the difference in the odds of CHD fatalities
occurring in the treatment and nontreatment groups studied in
meta-analyses.  At 1, the center of the figure, treatment and
nontreatment groups show no difference in the likelihood that
participants in either group will have a fatal heart attack from CHD. 
Bars to the left of 1 indicate that treatment leads to a decrease in
the rate of fatal heart attack; bars to the right, that it leads to
an increase.  Bars that overlap 1 indicate that differences between
treatment and nontreatment groups are not statistically significant. 
The line at the center of each bar represents the common odds ratio,
or the point estimate of that difference (that is, it is the ratio of
the odds of CHD fatalities in the treatment groups to the odds of CHD
fatalities in the nontreatment groups).  (Appendix II, table II.2,
details the numerical estimates for each meta-analysis.)

The lack of a statistically significant reduction in the rate of
death among participants in the primary prevention trials may be
because the trials generally lowered cholesterol levels by only a
modest amount and the number of events attributable to CHD during the
trials was so small that detecting statistical differences between
the treatment and nontreatment groups was difficult.\10 Several of
the recent meta-analyses found that the degree to which cholesterol
is lowered is related to reductions in CHD events among the trials. 
For example, one study reported that when cholesterol was lowered by
10 percent or more, deaths from CHD fell significantly, by 13
percent, and another reported that nonfatal heart attacks and CHD
deaths fell by 18 percent.\11 Another meta-analysis found that
lowering cholesterol by 12 percent or more corresponded to a
27-percent fall in the death rate from CHD.\12

The extent to which the rate of death from CHD falls is also
associated with the level of risk for CHD that participants have at
the start of a trial.  In one meta-analysis, participants in primary
and secondary trials were categorized according to higher, medium,
and lower levels of risk.  The results showed a significant
26-percent fall in the rate of death from CHD for those in the
higher-risk category but no significant benefit for those in the
medium- and lower-risk categories.  The higher-risk category was
dominated by participants who had a history of CHD (secondary
prevention trials), whereas the lower-risk category was dominated by
participants who did not (primary prevention trials).\13


--------------------
\8 Our findings on the benefits and risks of lowering cholesterol
from the meta-analyses are tabulated and discussed further in
appendix II. 

\9 Another useful measure comparing the effects of treatment is the
number of persons who would have to be treated in order to prevent
one adverse event.  One meta-analysis calculated this number for
death from CHD and found a large difference between those who had
existing CHD and those who did not:  to prevent one death from CHD,
38 persons with existing CHD would have to be treated while 675
persons with no evidence of CHD would have to be treated.  This
calculation does not reflect the larger number of nonfatal heart
attacks relative to fatal attacks and the proportionately larger
number of nonfatal attacks prevented by lowering cholesterol, nor
does it reflect treatment with the statin drugs.  See J.  S. 
Silberberg and D.  A.  Henry, "The Benefits of Reducing Cholesterol
Levels:  The Need to Distinguish Primary from Secondary Prevention. 
1.  A Meta-Analysis of Cholesterol Lowering Trials," Medical Journal
of Australia, 155 (1991), 665-70. 

\10 Cholesterol reduction across the individual trials ranged from
around 1 percent to more than 20 percent, with an average of about 10
percent. 

\11 A.  L.  Gould et al., "Cholesterol Reduction Yields Clinical
Benefit," Circulation, Circulation, 91 (1995), 2274-82, and M.  R. 
Law et al., "By How Much and How Quickly Does Reduction in Serum
Cholesterol Concentration Lower Risk of Ischaemic Heart Disease?"
British Medical Journal, 308 (1994), 367-73. 

\12 D.  J.  Gordon, "Cholesterol Lowering and Total Mortality," in B. 
M.  Rifkind (ed.), Lowering Cholesterol in High-Risk Individuals and
Populations (New York:  Marcel Dekker, 1995), pp.  33-48. 

\13 G.  D.  Smith, F.  Song, and T.  A.  Sheldon, "Cholesterol
Lowering and Mortality:  The Importance of Considering Initial Level
of Risk," British Medical Journal, 306 (1993), 1367-73. 


         NON-CHD AND TOTAL
         FATALITIES
-------------------------------------------------------- Letter :4.1.2

The extent to which overall survival rates improve from lowering
cholesterol has been addressed mainly in the meta-analyses.  The
individual trials were mostly too short and had too few participants
to assess whether death from all causes fell.  As shown in figure 4,
the meta-analyses that reported on deaths from non-CHD causes mainly
show a statistically significant increase within the primary
prevention trials for those who received treatment.  The
meta-analyses similarly point toward an increase within the secondary
prevention trials, but only one found it to be statistically
significant. 

   Figure 4:  Non-CHD Fatalities
   Reported in Meta- Analyses of
   Cholesterol-Lowering Trials\a

   (See figure in printed
   edition.)

\a The horizontal bars represent estimates (at the 95-percent
confidence interval) of the difference in the odds of non-CHD
fatalities occurring in the treatment and nontreatment groups studied
in meta-analyses.  At 1, the center of the figure, treatment and
nontreatment groups show no difference in the likelihood of non-CHD
fatalities among participants in either group.  Bars to the left of 1
indicate that treatment leads to a decrease in the rate of non-CHD
fatalities; bars to the right, that it leads to an increase.  Bars
that overlap 1 indicate that differences between treatment and
nontreatment groups are not statistically significant.  The line at
the center of each bar represents common odds ratio, or the the point
estimate of that difference (that is, it is the ratio of the odds of
non-CHD fatalities in the treatment groups to the odds of non-CHD
fatalities in the nontreatment groups).  (Appendix II, table II.2,
details the numerical estimates for each meta-analysis.)

When the trial data are aggregated through the meta-analyses, total
fatality rates increase among treated participants who had no
evidence of CHD (see figure 5).  However, none of the meta-analyses
found these rates to be statistically greater among participants who
were treated than among those who were not.  The meta-analyses found
a small relative decrease in total fatalities for participants who
were treated and had a history of CHD; only one found it to be
statistically significant (see figure 5).\14 That total fatalities
did not fall significantly after cholesterol-lowering treatment may
be because fatality rates from non-CHD causes were high among these
participants.  These high rates were large enough to cancel out the
modest reduction in CHD deaths after cholesterol-lowering treatment. 

   Figure 5:  Total Fatalities
   Reported in Meta- Analyses of
   Cholesterol-Lowering Trials\a

   (See figure in printed
   edition.)

\a The horizontal bars represent estimates (at the 95-percent
confidence interval) of the difference in the odds of total
fatalities occurring in the treatment and nontreatment groups studied
in meta-analyses.  At 1, the center of the figure, treatment and
nontreatment groups show no difference in the likelihood of total
fatalities among participants in either group.  Bars to the left of 1
indicate that treatment leads to a decrease in the rate of total
fatalities; bars to the right, that it leads to an increase.  Bars
that overlap 1 indicate that differences between treatment and
nontreatment groups are not statistically significant.  The line at
the center of each bar represents the common odd ratio, or the point
estimate of that difference (that is, it is the ratio of the odds of
total fatalities in the treatment groups to the odds of total
fatalities in the nontreatment groups).  (Appendix II, table II.2,
details the numerical estimates for each meta-analysis.)

Some analysts explain the increase in non-CHD fatalities as a matter
of chance.  Others speculate that lowering cholesterol itself
produces these results.  Still others point to the side effects of
one or more of the drugs used for treatment.  Although the
meta-analyses have not resolved this issue--because non-CHD deaths
were not as carefully reported in some of the individual trials as
were deaths from CHD--they have shed some light and have identified
areas that warrant further investigation. 

When one meta-analyst differentiated between the trials whose
participants had higher, medium, and lower risk for CHD, deaths from
all causes fell significantly among those who were at higher risk for
CHD and received cholesterol-lowering treatment, whereas deaths
increased significantly for those who were at lower risk. 
Furthermore, a few other meta-analyses have shown that the more that
cholesterol is lowered, the greater is overall survival.\15 Finally,
recent meta-analyses have found that the greater numbers of non- CHD
deaths associated with cholesterol-lowering treatment are more likely
to be seen in trials that used certain drug treatments, particularly
hormones and fibrates. 


--------------------
\14 The one meta-analysis that found a statistically significant
reduction weighted the result for the degree of cholesterol-
lowering. 

\15 One meta-analysis found that when cholesterol was lowered by at
least 10 percent, total mortality fell about 10 percent.  Another
reported that a 12-percent or greater reduction in cholesterol led to
a 20-percent reduction in total mortality. 


      TWO RECENT
      CHOLESTEROL-LOWERING TRIALS
---------------------------------------------------------- Letter :4.2

The Scandinavian Simvastatin Survival Study (4S), a secondary
prevention trial, and the West of Scotland Coronary Prevention Study
(WOSCOPS), a primary prevention trial, treated men whose total
cholesterol averaged more than 260 mg/dl with either a statin drug or
a placebo.\16 The results of both trials are consistent with the
several meta-analyses that found that greater reductions in
cholesterol yielded greater reductions in CHD events. 

The 4S trial is the first to find that lowering cholesterol can
significantly reduce the total fatality rate of cardiac patients who
are at very high risk for heart attack.  The WOSCOPS trial found that
lowering cholesterol lowered CHD fatalities, although this fell short
of statistical significance.  There were no increases in non-CHD
fatality, and total fatalities fell.  The reduction in CHD deaths in
4S means that 29 patients diagnosed with CHD would have to be treated
with simvastatin at the trial dosages for 5.4 years (the median
length of the trial) to prevent one CHD death.  Given the WOSCOPS
trial data, 143 middle-aged men with no evidence of a previous heart
attack would have to be treated for 5 years to prevent one death from
cardiovascular causes. 

Another concern in the controversy is that lowering cholesterol may
simply substitute one cause of death for another.  The meta-analyses
summarizing most past trials found that lowering cholesterol had the
clearest benefit among men who had been diagnosed with CHD, that the
magnitude of coronary benefits is related directly to the degree to
which cholesterol is lowered, and that higher rates of non-CHD
fatality are likely to be associated with specific treatments.  The
4S and WOSCOPS trials, which did not find a greater risk of death
from non-CHD causes, support the meta-analytic findings that it is
particular treatments to lower cholesterol, rather than the lowering
itself, that increase non-CHD fatality rates.  These data may be
encouraging, but it should be noted that these were 5-year studies
and that drug treatment would be expected to continue for longer
periods. 


--------------------
\16 Scandinavian Simvastatin Survival Study Group, "Randomized Trial
of Cholesterol Lowering in 4,444 Patients with Coronary Heart
Disease:  The Scandinavian Simvastatin Study (4S)," Lancet, 344
(1994), 1383-89, and J.  Shepard et al., "Prevention of Coronary
Heart Disease with Pravastatin in Men with Hypercholesterolemia," New
England Journal of Medicine, 333 (1995), 1301-7. 


      THE POPULATION AND TREATMENT
      GAPS IN THE CLINICAL TRIALS
---------------------------------------------------------- Letter :4.3


         THE POPULATION STUDIED
-------------------------------------------------------- Letter :4.3.1

The clinical trials conducted over the past 30 years concentrated on
white, middle-aged men who were considered to have a high risk of
CHD.  Useful trial data are lacking for women, minority men and
women, elderly men and women, and people who have moderate
cholesterol readings and are generally at lower risk.  The trials
that did enroll them did not do so in numbers large enough to allow
for separate analyses of the trial results.  To achieve sufficient
statistical power at the least cost, researchers selected the
participants they felt were most likely to experience a large number
of CHD events within a trial's duration. 

Because the numbers for women and elderly men and women are so small
and the trials did not separate them out for analysis, the
meta-analyses generally did not report them separately either.  Fewer
than half the trials considered in the meta-analyses were open to
women, and women constitute only about 7 percent of the participants
across all the trials.  The average age of participants in the trials
ranged from 45 to 66, with a median of about 52 across all the
trials. 

The trials examined in the meta-analyses provide very little
information about people whose risk of heart disease as defined by
the NCEP guidelines is moderate.  Most trials included either persons
who were known to have evidence of CHD or persons who had no evidence
of CHD but had high cholesterol and other risk factors.  Even primary
prevention trials included high-risk groups.  The median total
cholesterol for all trials was close to 260 mg/dl. 


         TYPES OF TREATMENT
-------------------------------------------------------- Letter :4.3.2

Several of the trials tested various dietary treatments, but most of
these used diets that differ from the ones that are now recommended
and used to treat persons with high cholesterol.  The diet trials
were also conducted mainly in institutions, where the participants'
diet could be strictly controlled.  Therefore, several reviewers of
the trials have questioned the extent to which their findings apply
to people who do not live in institutions. 

The older drug treatments, those predominantly used in the trials we
examined, have some side effects, and a few have seriously negative
effects.  The long-term effects of the statins, now the most widely
used cholesterol-lowering drugs, have not been extensively
investigated.  These drugs were developed and marketed in the 1980s,
after most of the existing trials were designed or conducted. 


      PLANNED CLINICAL TRIALS
---------------------------------------------------------- Letter :4.4

The gaps in what is known about the relationship between lowering
cholesterol and CHD outcomes are generally recognized by most
experts, including the authors of NCEP's reports.  Many are being
addressed by new trials that should help answer some of the questions
about non-CHD deaths and total fatalities, CHD outcomes for persons
whose short-term risk of heart disease is moderate, CHD outcomes for
population groups other than middle-aged white men, and the long-term
effects of the statin drugs. 

We identified 13 large trials in different stages of design and
implementation in the United States and Europe.  They tend to be
large and long and should have greater statistical power than
previous trials to assess total fatalities.\17 Most will use one of
the statin drugs.  Therefore, trial investigators expect larger
improvements in CHD outcomes. 

Many of the trials plan to provide information about different levels
of CHD risk.  At least 3 trials will study participants who have no
history of CHD, and 5 have broadened the range of cholesterol levels
generally studied to include borderline-high cholesterol levels of
200 to 240 mg/dl.  Three will evaluate whether drug treatment can
raise HDL cholesterol levels and thus improve CHD outcomes.  Previous
trials have not generally focused on treating persons whose HDL
cholesterol was low. 

Several of the new trials will enroll older participants.  One is
designed only for elderly participants.  Two of the new trials will
include a high percentage of women, but it may be difficult to
attribute any observed outcome to cholesterol- lowering treatment
because both will use several interventions to target different risk
factors.  Women will number fewer than 20 percent of the participants
in the other new trials.  Depending on how many women in all are
actually enrolled, there may or may not be sufficient information to
assess coronary outcomes with respect to cholesterol-lowering
treatment.  Only one trial will recruit a large percentage of African
Americans. 


--------------------
\17 All but one of these trials will have 2,500 participants or more. 
One trial will have a follow-up period of 3 years; the others will
run for 5 years or more. 


   CONCLUSIONS
------------------------------------------------------------ Letter :5

We have four main conclusions from our evaluation of the
meta-analyses and the clinical trials they studied: 

1.  The meta-analyses have consistently shown that
cholesterol-lowering treatment benefits middle-aged white men who
have high cholesterol levels and a history of heart disease.  This
appears to be so particularly when the treatment is effective in
lowering cholesterol. 

2.  The meta-analyses also show that men with moderate-to-high
cholesterol levels and no history of heart disease have lower rates
of nonfatal heart attacks but no statistically significant reductions
in rates of CHD death or total fatalities as a result of
cholesterol-lowering treatment. 

3.  The trials generally have not evaluated the efficacy of
cholesterol-lowering treatment for several important population
groups, such as women, elderly men and women, and minority men and
women.  Thus, they provide little or no evidence of benefits or
possible risks for these groups. 

4.  Two recent trials using a new drug class, the statins, show
greater reductions in coronary events with their greater reductions
in cholesterol and no increase in non-CHD fatalities.  One trial
studied men and women who had CHD and found a significant reduction
in total fatalities; the other, which studied only men who did not
have CHD, showed encouraging but not statistically significant
reductions in CHD fatalities. 


   RECOMMENDATION
------------------------------------------------------------ Letter :6

While the clinical trials have answered many important questions
about the benefits and risks of various cholesterol-lowering
treatments, they also leave unanswered several questions about likely
coronary outcomes for persons at different levels of CHD risk and for
persons in population subgroups that they have not thoroughly
studied.  These questions will be answered to some extent by several
trials that are under way, but it is likely to be several years
before they are completed.  We recommend that the director of NHLBI
take steps to closely monitor these trials, evaluate their outcomes,
and determine whether additional trials should be planned in order to
fill in data presently lacking on women, elderly men and women,
minority men and women, and persons whose cholesterol levels are
relatively low or who otherwise are members of low-risk groups. 


   AGENCY COMMENTS
------------------------------------------------------------ Letter :7

Officials from NHLBI reviewed a draft of this report and provided the
written comments that are reprinted in appendix VI.  In general, the
officials found our report to be technically accurate and agreed with
our main findings.  They believed, however, that we placed too much
emphasis on the aggregated results of the previous trials and not
enough on the two recently completed trials that NHLBI believes
conclusively demonstrate the benefits of lowering cholesterol for
patients who do and do not have CHD.  We agree that these recent
trials did produce relatively large CHD reductions compared to
previous trial results, and we have tried to reflect this in our
report.  While the results of the new trials are compelling, it is
important to recognize that they were designed to demonstrate the
efficacy and safety of particular statin treatments in only selected
high-risk populations for a period of 5 years. 

NHLBI was also concerned that in our report we did not sufficiently
take into account the other sources of nontrial evidence that it
believes provides a basis for treating various population subgroups. 
We agree that our report is limited by our not having reviewed other
nontrial sources of information about cholesterol treatment.  While
it is important to consider the full range of evidence, we were asked
to review clinical trials because they provide the strongest evidence
for establishing treatment benefits.  In light of this request, our
study findings can help the Congress and other readers understand the
limits of the trial information that was available in the past while
cholesterol policy and treatment efforts were being developed in this
country. 

As we arranged with your office, unless you publicly announce the
report's contents earlier, we plan no further distribution until 30
days after the date of this letter.  At that time, we will send
copies of this report to the National Institutes of Health and other
health-related agencies.  We will also make copies available to
others on request.  If you have any questions or would like
additional information, please call me at (202) 512-3092.  Major
contributors to this report are listed in appendix VII. 

Sincerely yours,

Kwai-Cheung Chan
Director of Program Evaluation
 in Physical Systems Areas


OBJECTIVES, SCOPE, AND METHODOLOGY
=========================================================== Appendix I


   THE OBJECTIVES AND SCOPE OF OUR
   STUDY
--------------------------------------------------------- Appendix I:1

Our work is set within the context of NIH's NCEP guidelines.  We
describe the guidelines and compare them with cholesterol-screening
treatment guidelines developed elsewhere, particularly in Australia,
Canada, and Europe.  (See appendix IV.) We did not evaluate the
process by which the guidelines were developed or the adequacy or
utility of either the NCEP guidelines or others.  We focused on
clinical trials evidence, one portion of the broad base of evidence
that undergirds the NCEP guidelines.  Accordingly, our three
objectives were to

  -- review the benefits and risks associated with the
     cholesterol-lowering treatments in published randomized clinical
     trials,

  -- review the extent to which the information from these trials
     provides information on different population groups, and

  -- identify and review the ongoing and planned cholesterol-lowering
     clinical trials to determine whether identified data gaps are
     likely to be filled. 

Our first objective encompasses 42 individual randomized
cholesterol-lowering clinical trials and 15 meta-analyses of them.\1
The trials were conducted over the past 30 years and were mainly
designed to test whether lowering cholesterol reduces the incidence
of coronary heart disease.  The meta-analyses were published from
1987 through April 1995. 

The scope of our second objective includes the 42 completed
cholesterol-lowering trials within the 15 meta-analyses whose designs
we had examined for key demographic and risk-related characteristics
among the participants.  Our third objective considers the NCEP
documents and medical literature we reviewed as well as the design
characteristics of 13 planned or ongoing trials. 


--------------------
\1 In the words of Glass, "meta-analysis" refers to "the statistical
analysis of a large collection of analysis results from individual
studies for the purpose of integrating the findings" (G.  V.  Glass,
"Primary, Secondary, and Meta-analysis," Educational Researcher, 5
(1976), 3, cited in H.  Cooper and L.  V.  Hedges (eds.), The
Handbook of Research Synthesis (New York:  Russell Sage Foundation,
1994), p.  5).  Meta-analysis is a relatively new analytic procedure,
although its practice has been growing steadily over the past two
decades. 


   OUR METHODOLOGY
--------------------------------------------------------- Appendix I:2


      RATIONALE
------------------------------------------------------- Appendix I:2.1

Our purpose was to assess the evidence regarding the efficacy of
cholesterol-lowering interventions on measured health outcomes.  We
examined the benefits and risks associated with cholesterol-lowering
through available clinical trials evidence.  We selected randomized
clinical trials because they are considered the "gold standard" in
medical research.  People who participate are assigned randomly to
either a treatment group that receives the intervention or a
nontreatment group that does not.  This ensures that a comparison of
the two groups' outcomes will be unbiased because the two groups are
more or less equivalent, differing only in the factor being
tested--in our case, cholesterol-lowering interventions. 

The NCEP guidelines were developed not only from the evidence from
randomized clinical trials but also from a broad range of
observational studies in epidemiology, animal physiology, pathology,
and genetics.  Many observational studies, for example, compared
population groups across nations, monitored the health status of
entire community populations, and assessed changes in the health
outcomes of recent immigrants.  These studies are important because
they enlarged our knowledge of how diet, cholesterol, and the risk of
CHD are linked.  However, only randomized clinical trials are
uniquely designed to attribute potential reductions in CHD
specifically to cholesterol-lowering interventions. 

However, the individual trials differed in the collection and
reporting of outcome variables.  Few of them were designed to respond
to the debate about the ability of cholesterol-lowering to extend the
life span; they often had too few participants or were too short in
duration.  Therefore, we summarized the results of
cholesterol-lowering trials through the meta-analyses to examine the
evidence for the effect that lowering cholesterol has on five health
outcomes:  CHD incidence (nonfatal and fatal heart attacks combined),
nonfatal heart attacks, CHD fatalities, non-CHD fatalities, and total
fatalities.\2

Meta-analyses statistically combine the results of individual
randomized cholesterol-lowering clinical trials to estimate the
extent to which cholesterol-lowering treatments reduce the incidence
of CHD events.  Therefore, we elected to provide a descriptive
synthesis of the meta-analyses whose purpose had been to assess the
benefits and risks of cholesterol-lowering interventions. 

Meta-analysis pools data from trials that have addressed essentially
the same research question in a statistically rigorous manner,
thereby (1) improving statistical power, (2) improving estimates of
effect size, (3) resolving uncertainty where reports disagree, (4)
answering theoretically relevant questions that were not initially
posed or not possible to address within a single study, and (5)
evaluating the conditions under which effects occur as well as
exploring the mediating processes that may account for them. 
Meta-analysis can point toward fruitful directions for future primary
research. 

However, meta-analysis can be limited by the availability and quality
of the studies whose results it aggregates.  Some criticisms of
meta-analysis are that it (1) does not detect publication bias and
thus can lead to spurious conclusions, (2) obscures potentially
relevant differences between combined trials, and (3) biases findings
through its inclusion and exclusion criteria.  Nevertheless, we
examined the meta-analyses we did because we believed they offered
the best available guide to what is generally known and not known
about the effects of lowering cholesterol among U.S.  adults.  Since
our intent was to shed light on the consistencies and inconsistencies
in all the available and relevant trial data, we did not conduct our
own meta-analysis. 

We examined the clinical trials evidence and other policy groups'
interpretation of that evidence.  We also looked at the
characteristics of the cholesterol-lowering clinical trials and their
applicability to the NCEP guidelines.  Specifically, we looked at the
extent to which the available data applied to different population
groups.  Further, we examined ongoing and proposed new trials to
determine if they would fill any of the data gaps that have been
identified. 


--------------------
\2 We present these five outcomes separately since CHD incidence
cannot be derived from summing nonfatal heart attacks and CHD
fatalities.  This is because any one trial participant may have more
than one nonfatal heart attack. 


      OUR DATABASE DEVELOPMENT AND
      ITS CHARACTERISTICS
------------------------------------------------------- Appendix I:2.2


         OUR SELECTION OF
         META-ANALYSES
----------------------------------------------------- Appendix I:2.2.1

We searched MEDLINE, a comprehensive bibliographic database, and
reviewed agency and other documents.  We also surveyed 11 experts for
their opinion on the relevance of the meta-analyses we had located
and asked them for additional references that we had failed to
uncover.  In this way, we identified 42 meta-analyses and other
studies that synthesized information from the available
cholesterol-lowering trials.  We judged 30 of these as inappropriate
to our objectives for the following reasons:  (1) individual trials
were not quantitatively aggregated, (2) aggregated trials were not
primarily cholesterol-lowering interventions, (3) analytic detail was
insufficient (for example, we excluded a meta-analysis when we were
unable to determine the number or identity of the studies it
included), (4) the number of individual trials was insufficient (for
example, we excluded meta-analyses that included no more than two
randomized trials of a year or more in duration), (5) the study was
not published or not published in English, and (6) the clinical CHD
outcomes we focused on had not been examined.  To the 12
meta-analyses that passed this screen, we added 9 that were published
or located after the NCEP guidelines were issued and otherwise passed
our screen.  This resulted in a core set of 21 meta-analyses. 


         OUR SELECTION OF
         INDIVIDUAL CLINICAL
         TRIALS
----------------------------------------------------- Appendix I:2.2.2

When we reviewed the results of this search, we located 85 references
to 54 individual cholesterol-lowering studies.  Most of the
cholesterol-lowering clinical trials of the past 30 years were
conducted in the United States and Europe with participants who were
mainly middle-aged white men who were considered to be at high risk
for CHD (see table III.1).  We asked the 11 experts about our
including these trials.  They differed on whether we should include
multifactor trials, hormone trials, and nonrandomized trials.  One or
more of them also voted to exclude some studies that had been
included in a meta-analysis.  We decided to consider all trials that
had been referenced by a meta-analyst, thus restricting our review to
the mainly randomized trials in the meta-analyses.  (The
meta-analyses and trials are listed in the bibliography.)


      OUR ANALYSIS DECISIONS
------------------------------------------------------- Appendix I:2.3

Since more than one of the 21 meta-analyses were completed by the
same lead author, we set the following criteria to reduce redundancy. 
(1) When more than one meta-analysis had been completed by the lead
author, we selected the more recent one if it represented all
previous trials and if it used similar configurations of trials to
examine the same outcomes.  (2) If the earlier meta-analysis reported
on a specific outcome that was not addressed in the more recent
meta-analysis, we retained the information from the earlier one.  (3)
If the more recent meta-analysis resulted in a substantial
improvement in quality over a previous one, we replaced the earlier
meta-analysis with the later one.  (4) When an estimate was based on
a very select set of studies that made comparisons with other
meta-analyses improper, we dropped the estimate.  When we had
followed these criteria, the meta-analyses totaled 15, as listed in
table I.l.\3



                                                                      Table I.1
                                                       
                                                        Basic Features of 15 Meta-Analyses of
                                                            Cholesterol-Lowering Trials\a


                                                                                                                                    Expressly
Meta-analysis   Clinical focus        Other                 Trial units\b    Participants\c   Gender             Included           excluded\d
--------------  --------------------  --------------------  ---------------  ---------------  -----------------  -----------------  -----------------
Yusuf 1987      CHD incidence and     Identified dose       19               36,050           Men, women         Diet, drug         Hormones,
                CHD and non-CHD       response                                                                                      multifactor
                fatality for diet     relationship between                                                                          trials
                and drug trials       lowering cholesterol
                                      and CHD incidence

Muldoon 1990\a  CHD, non-CHD, and     Summarized results    6                24,847           Men                Diet, drug         Multifactor
                total fatality (all   for cancer and                                                                                trials
                trials); CHD and      nonillness fatality
                total fatality for    (accident, violence,
                diet versus drug      trauma, and suicide)
                trials

Rossouw 1991    Nonfatal, fatal, and  Summarized %          4 diet, 8 drug   1,915 diet,      Men, women         Diet, drug         Hormones,
                all heart attacks;    reduction in heart                     25,295 drug                                            multifactor
                cardiovascular,       attacks for primary                                                                           trials
                noncardiovascular,    and secondary
                and total fatality    prevention trials
                for drug trials       and by treatment
                                      type (diet and
                                      various drugs)

Silberberg      CHD incidence,        Compared CHD          9                26,609           Men, women         Drug               Hormones,
1991            nonfatal heart        benefits of primary                                                                           multifactor
                attacks, and CHD and  and secondary trials                                                                          trials
                total fatality for    and estimated the
                drug trials           number needed to
                                      prevent an event

Ravnskov 1992   Nonfatal heart        Compared CHD          26               115,960          Men, women         Diet, drug,        \e
                attacks and CHD and   benefits by trial                                                          hormones,
                total fatality        type and duration,                                                         multifactor
                                      gender, and diet and                                                       trials
                                      drug treatment type

CTF 1993\a      Nonfatal heart        Summarized results    6                24,847           Men                Diet, drug         Multifactor
                attacks and CHD,      for gallbladder                                                                               trials
                non-CHD, and total    events and death
                fatality              from cancer and
                                      violence

Cucherat 1993   CHD incidence,        Summarized results    19               48,273           Men, women         Diet, drug,        Hormones,
                nonfatal heart        for death from                                                             surgery\f          multifactor
                attacks, and CHD and  cancer and death not                                                                          trials
                total fatality        related to illness;
                                      compared CHD
                                      benefits of primary
                                      and secondary
                                      prevention trials

Holme 1993      CHD incidence and     Summarized dose-      22               125,673          Men, women         Diet, drug,        \e
                total fatality in     response                                                                   multifactor
                different trial       relationship between                                                       trials, surgery\f
                types with different  lowering cholesterol
                diet and drug         and CHD incidence
                treatments            and total fatality

Muldoon 1993    CHD fatality in       Summarized results    8                30,696           Men                Diet, drug,        \g
                different trial       for nonillness-                                                            surgery\f
                types with different  related suicide and
                diet and drug         death from trauma in
                treatments            different trial
                                      types with different
                                      diet and drug
                                      treatments

Smith 1993\h    CHD, non-CHD, and     Summarized results    35               57,124           Men, women         Diet, drug,        \i
                total fatality        for CHD, non-CHD,                                                          hormones,
                stratified by risk    and total fatality                                                         surgery\f
                of CHD death          stratified by risk
                                      of CHD death in drug
                                      and other
                                      interventions

Law I 1994\j    CHD fatality (all     Summarized results    28               52,350           Men, women         Diet, drug,        Hormones,
                trials); non-CHD and  for death from                                                             surgery\f          multifactor
                total fatality in     accident, suicide,                                                                            trials
                different trial       and cancer and other
                types with different  diseases
                diet and drug
                treatments weighted
                by degree of
                cholesterol-
                lowering

Law II 1994\j   CHD incidence         Compared trial and    28               46,254           Men\k              Diet, drug,        Hormones
                weighted by degree    treatment types for                                                        surgery\f
                of cholesterol-       CHD incidence by
                lowering by trial     trial duration
                duration              differences

Gordon 1995     CHD incidence and     Compared results by   22               31,098           Men, women         Diet, drug,        Multifactor
                CHD, non-CHD, and     trial type,                                                                hormones,          trials
                total fatality        treatment type, and                                                        surgery\f
                                      degree to which
                                      cholesterol was
                                      lowered

Gould 1995      CHD, non-CHD, and     Separated effects of  35               71,313           Men, women         Diet, drug,        \i
                total fatality by     lowering cholesterol                                                       hormones,
                trial type weighted   from effects of                                                            multifactor
                by degree of          treatment type                                                             trials, surgery\f
                cholesterol-
                lowering

Rossouw 1995    CHD incidence,        Summarized results    5 primary, 14    21,639 primary,  Men, women         Diet, drug,        Hormones,
                nonfatal heart        for CHD incidence,    secondary, 9     11,636                              surgery\f          multifactor
                attacks, and CHD,     lesion progression,   angiography      secondary,                                             trials
                non-CHD, and total    and regression in                      1,435
                fatality              angiographic trials                    angiography
-----------------------------------------------------------------------------------------------------------------------------------------------------
\a All 15 meta-analyses summarized the published reports of varying
numbers of the 42 clinical trials we reviewed.  Muldoon (1990) and
CTF included only primary prevention trials; all the other
meta-analysts analyzed primary and secondary prevention trials. 
Complete facts of publication are given in the bibliography.  Trial
types are indicated in table I.2. 

\b Some meta-analysts count the individual treatment arms of a single
trial rather than analyzing the whole trial as a unit.  Therefore,
the number of trial units does not always match the totals in table
I.2. 

\c Includes both treatment and nontreatment groups.  Totals differ
for different analyses. 

\d Excludes other trial types also. 

\e Ravnskow (1992) and Holme (1993) specifically exclude angiographic
trials--that is, trials in which measuring coronary arteries allows
an assessment of the progression or regression rate of
atherosclerosis.  Angiographic trials used various
cholesterol-lowering treatments and recorded clinical outcomes that
could be included in a meta-analysis. 

\f Partial ileal bypass surgery. 

\g Muldoon (1993) excluded 7 randomized secondary prevention trials
that failed to report nonillness-related fatalities (trauma and
suicide) in addition to CHD fatalities. 

\h Included unpublished data. 

\i Not reported. 

\j Law I and Law II included the same set of 28 randomized trials,
supplemented by unpublished data. 

\k A small minority of women were included when gender-specific data
were unavailable. 

We did not develop formal criteria for evaluating the quality of the
meta-analyses we selected.\4 The investigators used accepted
approaches for quantitatively combining results, and most of the
meta-analyses appeared in peer- reviewed journals.  Only one
meta-analysis reported having assessed the variability in the
methodological soundness of the clinical trials that it aggregated. 
However, each meta-analysis included mainly randomized trials.  We
gave all the meta-analyses equal weight. 


--------------------
\3 In table I.1, we retained only the Muldoon 1990 analyses that were
not updated in Muldoon 1993.  We did the same for Rossouw 1991 and
1995.  The Law I 1994 meta-analysis (Law, Thompson, and Wald) relies
on the same set of trials as the Law II 1994 meta-analysis (Law,
Wald, and Thompson).  The former includes data on women, and the two
address different questions. 

\4 One difficulty in doing this relates to publication practice.  For
example, the description of a meta-analyst's search strategy is one
important criterion on which to base a quality assessment.  However,
only 5 meta-analysts discussed their literature search strategy. 
This failure could mean that no systematic search had been
undertaken, or it could mean that the search strategy was not
reported because of the journal's space limitations. 


      OUR COMPARISONS OF THE
      META-ANALYSES
------------------------------------------------------- Appendix I:2.4

The meta-analyses we examined included as few as 6 clinical trials up
to as many as 35 trials, for quantitatively aggregated information on
a range of about 1,500 to more than 125,000 participants.  Their
strategies differed by whether they analyzed data only on men; by
whether they considered treatment duration, the effects of specific
drug interventions, and the degree of cholesterol-lowering; and on
how they analyzed trials that included more than one treatment.\5 The
studies also differed by how they classified mixed-risk
interventions-- trials that included participants both with and
without manifest CHD.  The more recent meta-analyses differed from
the earlier ones in being able to include more-recent trials and
additional follow- up data. 

We compared the 15 meta-analyses for the descriptive characteristics
displayed in table I.1.  The meta-analysts' decision rules differed. 
For example, 9 included a trial that used a surgical intervention. 
Six included trials in which elevated cholesterol, treated with diet,
was only one of several risk factors treated; these are multifactor
trials.  All but 4 meta-analyses excluded trials with only hormone
interventions (because they represented cholesterol-lowering
interventions that are no longer recommended as a treatment option
for men).  Nine of the 15 meta-analysts included trials that had an
angiographic component.\6 Some meta- analysts used trial data that
had been reported in published articles; others obtained additional
data directly from the authors of the individual trials.  In all, the
15 meta-analyses included varying numbers of the 42 individual trials
we reviewed. 

Table I.2 shows that none of the meta-analyses examined all 42 of the
individual trials.  Seven contained less than half.  Still, 9 meta-
analysts included 18 of the same trials; this overlap means that the
results of the meta-analyses are not statistically independent of one
another. 



                                                                         Table I.2
                                                          
                                                               Fifteen Meta-Analyses and the
                                                              Cholesterol-Lowering Trials They
                                                                         Studied\a

Trial by                             Muldo
prevention                    Yusuf  on     Rossou      Silberbe  Ravnskov  CTF     Cuchera  Holme   Muldoo  Smith   Law I   Law II  Gordon  Gould   Rossou
type            Treatment     1987   1990   w 1991      rg 1991   1992      1993    t 1993   1993    n 1993  1993    1994    1994    1995    1995    w 1995
----------  --  ------------  -----  -----  ------  --  --------  --------  ------  -------  ------  ------  ------  ------  ------  ------  ------  ------
Primary
-----------------------------------------------------------------------------------------------------------------------------------------------------------
LA VA           Diet                      \b                                           \b            \b                                 \
(1969,
1971)

Oslo DA         Diet, other                                                                                                                
(1981)\c

MRFIT           Diet, other                                                                                                                
(1982)\c

WHO F           Diet, other                                                                 
(1983,
1986)\d

Gothenburg      Diet, other                                                                                                                 
(1986)\c

Minnesota       Diet                                                                                                
(1975,
1989)\e

Retinopath      Clofibrate                                                                                   
y (1969)\f

Begg            Clofibrate                                                                                                  
(1971)

Upjohn          Colestipol                            \g                               \g            \g     \g     \g\                   
(1978)

Finnish         Diet,                                                                                                                      
(1985)\c        clofibrate,
                probucol,
                other

Helsinki        Gemfibrozil                                                                                                             
(1987,
1988)

WHO             Clofibrate                                                                   \h     \h     \h     \h     \h     \h     \h
(1978,
1980,
1984,
1992)

LRC CPPT        Cholestyrami                                                                                                           
(1984a,         ne
1984b,
1992)

Excel           Lovastatin                                                          \i                      \i     \i     \i
(1990,
1991,
1992)


Secondary
-----------------------------------------------------------------------------------------------------------------------------------------------------------
Corn Oil        Diet          \j                                                                                                             \j
(1965)

MRC Low         Diet                                                                                                                       
Fat (1965)

Oslo DH         Diet                                                                                                                       
(1966,
1970)

MRC Soya        Diet          \k                                                                                                    
(1968,
1974)

Sydney          Diet                                                                                                                 
(1978)

DART            Diet                                                                                                                             
(1989)

India           Diet                                                                                         
(1992)

Estrogen        Estrogen                                                                                                                  
(1961)

Long-Term       Estrogen                                          \l                                        
Estrogen
(1962)

Chicago         Estrogen                                                                                    
(1963)

Newcastle       Clofibrate                                                                                                                
(1971)

Scottish        Clofibrate                                                                                                                
(1971,
1972)

Acheson         Clofibrate                                                                                                                
(1972)

St.             Colestipol                                                                                                 
Vincents
(1973)

Veterans        D-            \m                       \m                                                  \m     \m     \m     \m     \m     \m
Cardiology      thyroxine;
(1968,          D-thyroxine
1974)           + estrogen;
                estrogen;
                estrogen +
                nicotinic
                acid;
                nicotinic
                acid

Veterans        Probucol                                                                                     \n     \n     \n
W. Roxbury
(1981)

NHLBI           Cholestyrami               \p                                                                                                 
(1984)          ne, diet\o

CDP (1970,      Clofibrate,   \q           \q         \q       \q               \q      \q     \q     \q     \q     \q     \q     \q     \q
1972,           D-
1973,           thyroxine,
1975,           estrogen 2.5
1986)           mg, estrogen
                5.0 mg,
                niacin

CLAS            Colestipol,                 \p                                                                                                   
(1987)          niacin\o

Stockholm       Clofibrate,                                                                                                               
(1977,          nicotinic
1980,           acid
1988)

FATS            Lovastatin +                \p                                                                                            \r     
(1990)          colestipol;
                niacin +
                colestipol\o

Lifestyle       Diet,                                                                                        \s
(1990)\c        other\o

POSCH           Partial                                                                                                                      
(1990)          ileal bypass
                surgery

SCOR            Colestipol,                                                                                                                         
(1990)\t        lovastatin,
                niacin\o

Restenosis      Lovastatin\o                                                                                                              
(1991,
1992)

STARS           Diet, diet +                                                                                                      \u     \u     
(1992)          cholestyrami
                ne\o

Ancillary       Gemfibrozil                                                                                                                     
Helsinki
(1993)

MARS            Lovastatin\o                                                                                                                         
(1993)

===========================================================================================================================================================
Trials                        17     6      11          9         22        6       18       22      8       35      28      28      18      29      23
(total =
42)\v

% of total                    40     14     26          21        52        14      43       52      19      83      67      67      43      69      55
trials
-----------------------------------------------------------------------------------------------------------------------------------------------------------

\a Complete facts of publication for the 15 meta-analyses and the
reports of the 42 individual trials in this table are given in the
bibliography. 

\b This trial had primary and secondary prevention components.  All
meta-analysts treated it as a primary prevention trial except Holme,
who classified it as secondary.  Roussouw (1991) did not classify
diet trials by prevention type; Smith did not conduct analysis by
trial type. 

\c A multifactor trial, targeting multiple risk factors
simultaneously with more than one intervention. 

\d A multifactor trial that randomized 66 factories (comprising
49,781 men) rather than individuals. 

\e This trial had primary and secondary prevention components;
primary predominated. 

\f Participants were being treated for diabetic retinopathy.  About
40 percent were diagnosed as having peripheral or coronary vascular
disease. 

\g All meta-analysts treated this as a primary prevention trial
except Silberberg, who separated its primary and secondary
components; Holme, who treated the trial as secondary; and Law I and
Law II, who classified it as a "mixed" trial.  Smith did not analyze
it by trial type. 

\h Included the results of the WHO 1992 intention-to-treat analysis. 

\i This trial had primary and secondary prevention components. 
Cucherat treated it as primary, Smith did not analyze it by trial
type, and Law I and Law II classified it as a "mixed" trial. 

\j Yusuf and Rossouw (1995) included both corn oil and olive oil
components. 

\k Treated as a drug trial. 

\l Combined three estrogen treatments:  anvene, lynoral, and
premarin. 

\m Silberberg and Rossouw (1995) included the nicotinic acid trial
arm only; Yusuf, Law I, and Law II included the nicotinic acid arm
and the arm containing estrogen and nicotinic acid; Gordon and Gould
included all trial arms except estrogen only; Smith included all five
treatment arms. 

\n This trial had primary and secondary prevention components.  Smith
did not conduct the analysis by treatment type.  Law I and Law II
classified it as a "mixed" trial. 

\o A trial in which measuring coronary arteries allows an assessment
of the progression or regression rate of atherosclerosis. 
Angiographic trials used various cholesterol-lowering treatments and
recorded clinical outcomes that could be included in a meta-analysis. 

\p Discussed the three angiographic trial results descriptively; they
are not in the trials total. 

\q Estrogen 2.5 mg and 5.0 mg and dextrothyroxine were discontinued
because of toxic side effects.  Yusuf, Roussouw (1991 and 1995),
Silberberg, and Cucherat treated the clofibrate and niacin arms as
two separate trials; Holme, Muldoon (1993), and Law I and Law II
combined them as one.  Gordon and Gould treated the clofibrate,
D-thyroxine, and niacin arms separately.  Ravnskov treated the five
arms as separate trials; Smith combined them. 

\r Separated results for colestipol and lovastatin from results for
colestipol and niacin. 

\s Smith later corrected his original analysis to say that he should
not have included this multifactor trial with the other,
single-intervention trials; the one death recorded for it did not
affect his results. 

\t Participants had heterozygous familial hypercholesterolemia. 
About half of the nontreatment group received a low dose of
colestipol. 

\u Gordon treated diet and drug components separately.  Gould
analyzed cholestyramine and diet together but also analyzed diet
separately. 

\v Totals are only for the number of trials included in each
meta-analysis, not the number of treatment arms totaled in table I.1. 

Eleven of the 15 meta-analysts combined data across all trial types
and interventions to investigate one or more of the health outcomes
we focused on.  This allowed them to see whether the effect of
lowering cholesterol could be discerned for a common risk factor
among trials whose designs varied greatly.  However, clinical and
statistical heterogeneity can make it difficult to interpret findings
if aggregating some trials with others disproportionately affects the
results.  NCEP acknowledges that broad aggregation can yield such
heterogeneity and that it is potentially misleading. 

Eight of these 11 meta-analysts conducted separate analyses by trial
type--primary and secondary prevention trials--which enabled them to
create less heterogeneous trial groupings.\7 The results of 4
meta-analyses were reported only at lower levels of aggregation, such
as those that considered results mainly for primary prevention
trials.  Some meta-analysts also analyzed the aggregated trials to
isolate the effect of dietary treatment from that of drugs. 
Technically, these subgroup analyses are also meta-analyses.  Almost
all the 15 meta-analysts thus conducted multiple meta-analyses, and
we report some of the main results in the tables in appendix II. 


--------------------
\5 This is a particular problem for the large Coronary Drug Project
study, which receives considerable weight in a meta-analysis.  The
trial as designed had five treatments (clofibrate, niacin, estrogen
5.0 mg, estrogen 2.5 mg, and dextrothyroxine).  The last three were
discontinued early in the trial when adverse effects were observed. 
Therefore, an investigator who includes all five treatments may bias
results in a negative direction.  Some analysts pool the data from
the clofibrate and niacin treatments, which can bias results in a
positive direction. 

\6 In angiography, measurement of the coronary arteries allows
investigators to assess the effect of cholesterol-lowering on the
rate of progression or regression of atherosclerosis. 

\7 The distinction between primary and secondary trial participants
is regarded as somewhat artificial from clinical and methodological
standpoints.  From a clinical standpoint, participants in primary
prevention trials may have had underlying coronary atherosclerosis as
evidenced by those who developed clinical symptoms during the trial. 
From a methodological standpoint, investigators have used different
criteria for classifying primary and secondary prevention trials so
that such classifications do not directly reflect a stratification of
the risk of death from CHD. 


      HOW WE REPORT
      CHOLESTEROL-LOWERING EFFECTS
------------------------------------------------------- Appendix I:2.5

The meta-analyses we studied statistically combined individual trial
results to improve the precision of the estimates of the extent to
which their cholesterol-lowering treatments reduced CHD risk with
regard to CHD incidence, nonfatal heart attacks, and CHD, non-CHD,
and total fatalities.  Most of the meta-analyses expressed their
results in terms of a common odds ratio and confidence interval. 

For an individual trial, the odds ratio is the ratio of two ratios. 
It is defined as the odds of events--such as death from coronary
heart disease--to nonevents in the treatment group divided by the
odds of events to nonevents in the nontreatment group.  An odds ratio
lower than 1 (such as 0.90) indicates that the rate of CHD events
within the treatment group fell compared to the nontreatment group. 
An odds ratio higher than 1 (such as 1.12) indicates that the rate of
CHD events within the treatment group rose compared to the
nontreatment group.  For aggregated trials in a meta-analysis, a
common odds ratio is similarly interpreted. 

The imprecision that results from combining individual odds ratios is
expressed by a confidence interval.  In the meta-analyses that
reported a common odds ratio, the confidence interval was set at 95
percent, meaning that if a meta-analysis were replicated 100 times
and the confidence interval were calculated for each odds ratio, 95
of these 100 confidence intervals would contain the true odds ratio,
even if it were not precisely known. 

With one exception, we did not attempt to re-analyze the common odds
ratios of meta-analyses that reported them.\8 However, to present our
findings consistently, we converted the results of meta-analyses that
used a different measure to common odds ratios or obtained confidence
intervals.\9 We forwarded all our changes to the studies' authors for
their review, and we point them out in our tables in appendix II. 

Not all meta-analysts reported on all five health outcomes or on
pertinent groupings of them, such as primary or secondary prevention
trials or specific treatments.  Therefore, the tables do not show 15
meta-analysis entries for each outcome.  Our tables show the
direction of effects and indicate the findings' statistical
significance (that is, whether the findings showed that differences
between the treatment and nontreatment groups were not likely to have
occurred by chance alone). 

In summarizing the meta-analyses' reports of treatment effects, we
used the following general rule.  If the 95-percent confidence
interval for a common odds ratio included 1 (the point of equivalent
odds), then there was no statistically significant difference (at the
0.05 level) between the treatment and nontreatment groups on the
outcome of interest.  In summarizing the statistical significance of
their findings, we recognize that a statistically significant effect
is not necessarily a clinically meaningful one.  Where the meta-
analysts' results were discrepant, we provide potential explanations
for the discrepancies. 

Not all the detail that the meta-analysts reported is summarized in
the tables in appendix II.  Several meta-analysts provided
information on how their results were influenced by degree of
cholesterol reduction, duration of trials, individual risk status,
and interventions in which adverse effects outweighed the benefits of
treatment.  Since analyses that considered these factors are
important in the interpretation of the effects of lowering
cholesterol, we discuss them in the body of appendix II. 


--------------------
\8 We recalculated the odds ratios in Muldoon's 1993 meta-analysis in
order to include his published, updated data. 

\9 We used the Mantel-Haenszel method and the STAT XACT program
produced by Cytel Software of Cambridge, Massachusetts.  This program
uses the Breslow-Day method of testing for homogeneity of odds
ratios.  We also calculated the confidence intervals surrounding the
odds ratios with the STAT XACT program, basing our calculations on
the variance estimation method of Robins, Breslow, and Greenland. 
See N.  Mantel and W.  Haenszel, "Statistical Aspects of the Analysis
of Data from Retrospective Studies of Disease," Journal of the
National Cancer Institute, 22 (1959), 719-48; N.  E.  Breslow and N. 
E.  Day, The Analysis of Case-control Studies (Lyon, France:  IARC
Scientific Publications, 1980); and J.  Robins, N.  Breslow, and S. 
Greenland, "Estimators of the Mantel-Haenszel Variance Consistent in
Both Sparse Data and Large-strata Limiting Models," Biometrics, 42
(1986), 311-23. 


   THE STRENGTHS AND LIMITATIONS
   OF OUR STUDY
--------------------------------------------------------- Appendix I:3

Summarizing and integrating studies through meta-analysis is, like
primary research, a research process in its own right.  Meta-analysis
imposes methodological standards and statistical principles that
require the analyst to adhere to explicit research and reporting
rules.  These in turn allow other researchers to replicate the study. 
Further, meta-analysis helps develop comprehensive knowledge beyond
the limits of individual studies.  It is for these reasons that we
chose to review and summarize meta-analyses rather than perform a
more traditional literature review. 

Meta-analysis techniques capitalize on accumulating evidence by
making estimates more precise and reliable; they also permit the
testing of hypotheses that may not have been addressed in the
individual trials.  Moreover, these techniques enable meta-analysts
to evaluate the circumstances of the effects, such as how different
treatment modalities or the characteristics of different population
groups are related to an effect.  They also allow meta-analysts to
explore what may underlie an effect, such as the particular degree to
which cholesterol has been lowered. 

Because meta-analysis can increase statistical power and the
precision of estimates of effects, it facilitates the examination of
noncoronary fatalities, even where recorded deaths from specific
causes were few in number.  Moreover, meta-analysis allows
investigators to examine whether lowering cholesterol affects the
life span.  Most individual trials were not designed to test the
effect on total deaths of lowering cholesterol:  either they had too
few participants, rendering them low in statistical power, or they
were too short in duration to examine this outcome. 

The meta-analysis results we present were intended to yield a
coherent, general conclusion about the benefits and risks associated
with cholesterol-lowering treatments.  However, the trials'
considerable clinical heterogeneity in terms of participants and
treatments can result in statistical heterogeneity such that the
meaningfulness of any overall estimate can be called into question. 
This problem is not always detected in formal testing.  Although only
a few of the meta-analyses specifically addressed this issue, their
various subgroup analyses did derive from its recognition as they
attempted to provide more homogeneous groups of participants and
treatments. 

Therefore, it is important to note that the overall significance
tests we present in our tables do not imply that a treatment effect
existed in all primary or secondary trial circumstances.  The common
odds ratios that are a guide to overall treatment benefit do not
accurately estimate the treatment benefit or risk achieved in any
particular trial.  The confidence interval that expresses the
statistical precision achieved for a common odds ratio does not allow
for any potential variation in the magnitude of true effect between
trials (or between treatments or types of participants).  The results
of the meta-analyses we report should not be extrapolated beyond the
circumstances from which they were derived. 

Further, since the individual clinical trials did not report the
effects of lowering cholesterol on the incidence of nonfatal diseases
other than CHD, the meta-analysts could not evaluate these effects. 
Finally, the results of the different meta-analyses are not
statistically independent when they include the same trials.  Many of
the larger single-intervention trials that the analysts included
weigh heavily in the results.  This means that some convergence of
findings can be expected despite the diversity of the meta-analyses. 

The meta-analyses considered only total serum cholesterol levels,
because LDL and HDL were not always recorded in the individual
trials.  It is noteworthy that the NCEP guidelines call for treatment
decisions that base distinctions on LDL and HDL levels.  Therefore,
the results may be less applicable to population groups in which the
configuration of total cholesterol subfractions differs from those
found in the trials.  The guidelines also call for basing
cholesterol-lowering treatment on distinctions of age and gender, but
the major participants in the clinical trials were mainly middle-aged
white men with elevated cholesterol levels. 

Only a few of the clinical trials combined in the meta-analyses that
we reviewed used recent, more efficacious treatments such as the
statin drugs.  Some of the interventions for men in large trials that
weigh heavily in the results have been discontinued or had
cholesterol-lowering treatment regimens that have declined in use. 
Nonetheless, it is important to examine the results of the
meta-analyses that aggregated these data because the
cholesterol-lowering trial results are integral to the evidence on
which the NCEP treatment guidelines were based. 


THE META-ANALYSES OF
CHOLESTEROL-LOWERING TRIALS AND
THE HEALTH OUTCOMES THEY REPORT
========================================================== Appendix II

In this appendix, we summarize our major findings from the 15
meta-analyses for the five health outcomes we focused on:  CHD
incidence (fatal and nonfatal heart attacks combined), nonfatal heart
attacks, CHD fatalities, non-CHD fatalities, and total fatalities. 
The results apply mainly to middle-aged white men who had elevated
serum cholesterol levels, because they have had the greatest
representation in the cholesterol-lowering clinical trials over the
past 30 years.  Meta-analysts continue to actively analyze these
data.  We discuss in appendix III the issue of extrapolating data to
other population groups. 


   A SUMMARY OF OUR MAJOR FINDINGS
-------------------------------------------------------- Appendix II:1

The trials included in the meta-analyses we reviewed lowered serum
cholesterol 10 percent on the average (the range in the individual
trials was from less than 1 percent to more than 20 percent). 
Several meta-analysts identified the extent to which cholesterol was
lowered as a mediating factor in the efficacy of cholesterol-lowering
interventions.  Deaths from causes other than CHD were not related to
a reduction in cholesterol level.  In fact, these fatalities were
fewer when cholesterol reduction was greater.  This suggests that
excess non-CHD risk results not from lowering cholesterol itself but
from particular types of cholesterol-lowering treatment. 

With some differences in the magnitude of benefits, the meta-analysis
results concur that participants with and without manifest heart
disease benefit from cholesterol-lowering interventions. 
Statistically significant reductions in the rates of CHD incidence
and nonfatal heart attacks favor treatment groups. 

Statistically significant reductions in CHD fatality rates were found
mainly for high-risk participants in treatment groups and for treated
participants in secondary prevention trials.  For lower-risk
participants who had elevated cholesterol levels and for participants
in primary prevention trials, evidence was absent that
cholesterol-lowering interventions reduced CHD fatality rates more in
treatment than in nontreatment groups. 

Lower-risk participants who had elevated cholesterol levels and
treatment groups in the primary prevention trials showed
statistically significant increases in non-CHD fatalities.  In
secondary prevention trials, lowered rates of CHD death appear to
offset the increased rates of non-CHD death.  As a result, the
overall death rates for treatment groups were mainly favorable but
differences from nontreatment groups were not statistically
significant. 

For drug trials, the meta-analyses show statistically significant
reductions in rates of CHD incidence and nonfatal heart attacks
favoring treated participants.  However, increases in rates of death
from non-CHD causes among treatment groups compared to nontreatment
groups were also statistically significant.  There were no
statistical differences between drug treatment and nontreatment
groups in CHD fatalities or total fatalities.  CHD death rates
favored treatment groups, whereas overall fatality rates were close
to 1 (the point of equivalent odds between treatment and nontreatment
groups).  Two recent meta-analyses identified two broad classes of
cholesterol-lowering interventions--fibrates and hormone treatment in
men--as accounting for statistically significant increases in non-CHD
fatality rates.  The beneficial effects of treatment outweighed the
adverse effects of resins and niacin. 

Generally, overall survival rates did not improve for
cholesterol-lowering treatment groups compared to nontreatment
groups.  Survival rates were shown to improve when analyses
specifically considered the participants who were at the highest risk
or accounted for the extent of cholesterol reduction.  Greater
reductions in serum cholesterol were associated with a reduction in
deaths from all causes. 

The diet trials included in the meta-analyses used dietary
interventions that are not currently recommended.  The meta-analyses
mainly show no distinctions between dietary treatment groups and
nontreatment groups with regard to the five health outcomes. 


   THREE RECENT RESEARCH TRENDS
-------------------------------------------------------- Appendix II:2

Three recent trends in cholesterol-lowering research are notable
before we present the details of our findings in the text and tables
in this appendix.  The first trend is that angiographic trials have
been included in meta-analyses of cholesterol-lowering trials more
frequently since 1993 than before.  Angiographic trials are secondary
prevention trials because their participants are at very high risk or
had existing symptoms of heart disease.  Although the trials'
principal end point is the arteriographic measurement of coronary
arteries, their clinical outcomes in cholesterol-lowering trials are
available for meta-analysis.  Collectively, they have shown that
intensive cholesterol-lowering slows the progression of
atherosclerotic lesions; in some patients, it even causes the lesions
to regress.  Interventions as diverse as dietary therapy, drugs, and
partial ileal bypass surgery yield similar results. 

The second trend relates to ongoing trials of the statin drugs, a
relatively new drug class that appears to lower LDL cholesterol
levels 20 to 40 percent, although their long-term safety is still
being evaluated.  Two recently completed statin trials found clinical
benefits favoring treatment groups and no differences in adverse
effects between treatment and nontreatment groups.  In 4S, the total
cholesterol of cardiac patients fell 25 percent.  Among other
clinical benefits, the study demonstrated statistically significant
reductions in the frequency of nonfatal heart attacks, fatal heart
attacks, and total fatalities favoring participants treated with
simvastatin within its 5-year span.\1 The study's authors cautioned
against extrapolating their results to other secondary prevention
trials using other statin agents, and they also advised caution in
the extrapolation of their results to primary prevention trials. 

WOSCOPS, a study of men who had never had a heart attack, lowered
total cholesterol levels by 20 percent with the drug pravastatin. 
The results, which favored the treatment group, showed statistically
significant reductions in fatal and nonfatal heart attacks combined,
all cardiovascular deaths, and death from definite and suspected CHD
(but not definite CHD deaths considered independently).  The results
almost reached statistical significance for reductions in the rates
of death from all causes. 

The 4S and WOSCOPS trials were too recent to be included in the
meta-analyses we reviewed.  However, their results--which showed no
statistical difference in deaths from noncardiovascular causes
between the treatment and nontreatment groups--tend to support
meta-analysis reports of earlier trials in which increases in non-CHD
fatality rates were associated not with lowering cholesterol but with
particular cholesterol-lowering treatments. 

The third research trend is that investigators are more frequently
trying to analyze specific interventions, independently of how well
they lower cholesterol, in order to determine their specific effects
on clinical outcomes.  Treatments with hormones and fibrates have
shown statistically significant increases in non-CHD fatalities and
total fatalities.  Hormone treatments at doses used in several
individual trials examined by meta-analysts are not currently
recommended for men.  The fibric acid derivatives clofibrate and
gemfibrozil are effective primarily in lowering triglycerides, but
the use of clofibrate declined after the World Health Organization
trial results were published.\2 Gemfibrozil is currently recommended
only for the treatment of primary prevention patients who have a
combination of high triglycerides, elevated LDL, and low HDL. 


--------------------
\1 Other beneficial results included statistically significant
reductions in major coronary events for women and for patients
younger and older than 60.  Fewer coronary bypass surgery and
angioplasty procedures were conducted for patients taking
simvastatin; the differences between treatment and nontreatment
groups were statistically significant. 

\2 The Food and Drug Administration approves clofibrate for patients
who have very high triglycerides and who are at risk for pancreatitis
and for patients with familial dysbetalipoproteinemia. 


   PRIMARY AND SECONDARY
   PREVENTION TRIAL RESULTS
   COMPARED
-------------------------------------------------------- Appendix II:3

As we discussed in appendix I, the individual trials represent
considerable clinical heterogeneity in treatments and patients.  The
meta-analyses show the overall average effect of lowering cholesterol
across these trials and introduce both clinical and statistical
heterogeneity that can make results difficult to interpret and less
reliable.  For greater homogeneity, investigators classify trials
broadly by prevention type (primary or secondary) or by treatment
type (diet or drugs) or both.  We summarized the meta-analysts'
findings for diet and drug treatment trials earlier in this appendix. 
As for prevention, the relative risk of death for participants who
already show evidence of CHD (secondary prevention trials) is about
five to seven times greater than that for participants who do not
(primary prevention trials).  We summarize these results below. 


      CHD INCIDENCE AND NONFATAL
      HEART ATTACKS
------------------------------------------------------ Appendix II:3.1

Table II.1 shows the results of the meta-analysts who examined CHD
incidence and nonfatal CHD.  It separates the results of the
meta-analysts who combined the primary prevention trials from the
results of those who combined the secondary prevention trials.  The
predominant finding is statistically significant reductions in the
rates of CHD incidence and nonfatal heart attacks for participants in
cholesterol-lowering treatment groups compared to nontreatment
groups.\3 Across primary prevention trials, the common odds ratios
range from 0.85 to 0.77 for CHD incidence and from 0.83 to 0.74 for
nonfatal CHD.\4

Similar benefits are reported across secondary prevention trials. 
The odds ratios range from 0.86 to 0.76 for CHD incidence and from
0.96 to 0.76 for nonfatal CHD.\5



                                     Table II.1
                      
                      Meta-Analysis Results for CHD Incidence
                                 and Nonfatal CHD\a



Me
ta
-
an
al
ys      Odds                Odds                Odds                Odds
is     ratio  Interval     ratio  Interval     ratio  Interval     ratio  Interval
--  --------  --------  --------  --------  --------  --------  --------  --------
Yu      0.85     0.76-                          0.81     0.74-
 s                0.96                                    0.88
 u
 f
 1
 9
 8
 7
 \
 b
Si      0.77     0.67-      0.75     0.64-      0.79     0.71-      0.78     0.67-
 l                0.89                0.87                0.88                0.90
 b
 e
 r
 b
 e
 r
 g
 1
 9
 9
 1
Ra                          0.83     0.75-                          0.96     0.89-
 v                                    0.92                                    1.04
 n
 s
 k
 o
 v
 1
 9
 9
 2
CT                          0.74     0.64-
 F                                    0.85
 1
 9
 9
 3
Cu      0.82     0.74-      0.74     0.65-      0.76     0.70-      0.76     0.68-
 c                0.92                0.86                0.83                0.86
 h
 e
 r
 a
 t
 1
 9
 9
 3
Ho    0.78\c     0.68-                        0.80\c     0.71-
 l              0.89\c                                  0.89\c
 m
 e
 1
 9
 9
 3
Go      0.77     0.67-                          0.86     0.79-
 r                0.87                                    0.93
 d
 o
 n
 1
 9
 9
 5
 \
 d
Ro      0.80     0.71-      0.75     0.65-      0.83     0.76-      0.77     0.68-
 s                0.89                0.86                0.90                0.87
 s
 o
 u
 w
 1
 9
 9
 5
 \
 e
----------------------------------------------------------------------------------
\a Complete facts of publication are given in the bibliography.  CHD
incidence is fatal and nonfatal heart attacks combined.  Intervals
are 95-percent confidence intervals for the common odds ratios.  The
meta-analyses did not report data for the empty cells. 

\b For consistency, we recalculated the results from risk ratios and
combined diet and drug trials for primary and secondary prevention
trials.  Using the Breslow-Day (B-D) method, we tested for
homogeneity of the odds ratios.  The B-D statistic was 8.40 (p = .08)
for primary prevention trials and 36.55 (p = .05) for secondary
prevention trials. 

\c These data are for drug trials.  Separately reported results for
diet trials were primary prevention CHD incidence 1.02, 0.94-1.10;
secondary prevention CHD incidence 0.78, 0.63-0.96. 

\d For consistency, we recalculated the results from percentage
change in risk for primary and secondary prevention trials.  We
recalculated results for primary and secondary trials combined in
order to obtain confidence intervals.  Using the Breslow-Day (B-D)
method, we tested for homogeneity of the odds ratios.  The B-D
statistic was 1.30 (p = .73) for primary prevention trials and 33.6
(p = .05) for secondary prevention trials. 

\e Roussouw's separate analysis for secondary prevention angiography
trials showed a statistically significant reduction in CHD events for
the treatment groups.  The 0.55 odds ratio had a wide 95-percent
confidence interval of 0.30-0.99. 


--------------------
\3 Nonfatal heart attacks predominate in the CHD incidence category,
which represents the combination of fatal and nonfatal heart attacks. 

\4 This range does not include the separate estimate for diet trials
indicated for Holme (1993) in table II.1.  In this meta-analysis,
dietary treatment lowered cholesterol 3.2 percent, but the dietary
trials were mainly multifactor trials, which taken as a group lowered
total cholesterol only 0.7 percent. 

\5 The difference between treatment and nontreatment groups in
Ravnskov (1992) lacked statistical significance.  Among meta-
analysts examining nonfatal CHD, only Ravnskov included hormone
trials, but the analysis has been strongly criticized. 


      CHD FATALITIES
------------------------------------------------------ Appendix II:3.2

Table II.2 shows that in the primary prevention trials, CHD fatality
rates fell with treatment, but because all the confidence intervals
exceed 1, the possibility cannot be ruled out that the CHD death rate
increases in treatment groups compared to nontreatment groups.  Gould
(1995) suggests that differences between treatment and nontreatment
groups may not be apparent because of the greater variability of the
estimate when primary prevention trials are aggregated, the small
number of univariate primary prevention trials, and the narrow range
of cholesterol reductions in these trials, all of which decrease the
precision necessary to find statistical significance. 



                                                                                      Table II.2
                                                                       
                                                                       Meta-Analysis Results for CHD, Non-CHD,
                                                                                 and Total Fatality\a



Meta-analysis     Odds ratio    Interval  Odds ratio    Interval  Odds ratio     Interval     Odds ratio       Interval     Odds ratio       Interval      Odds ratio       Interval
----------------  ----------  ----------  ----------  ----------  ----------  -----------  -------------  -------------  -------------  -------------  --------------  -------------
Yusuf 1987\b            0.93   0.76-1.14        1.19   1.04-1.37                                    0.89      0.80-0.98           1.14      0.97-1.36
Muldoon 1990                                                            1.07    0.94-1.21
Silberberg 1991         0.85   0.64-1.14                                1.14    0.96-1.36           0.84      0.75-0.95                                          0.90         0.80-1
Ravnskov 1992           0.92   0.83-1.02                                1.02    0.95-1.08           0.96      0.88-1.04                                          1.02      0.95-1.10
CTF 1993                0.90   0.71-1.14        1.19   1.03-1.39        1.07    0.94-1.22
Cucherat 1993           0.90   0.75-1.09                                1.07    0.96-1.19           0.87      0.78-0.96                                          0.93      0.85-1.03
Holme 1993                                                            1.20\c     1-1.45\c                                                                      0.93\c    0.83-1.04\c
Muldoon 1993\d          0.93   0.77-1.12
Law I 1994\e                                    1.11   0.99-1.24        1.06    0.97-1.17                                         0.99      0.83-1.18            0.90      0.84-0.97
Gordon 1995\f           0.93   0.75-1.14        1.26   1.06-1.49        1.13    0.98-1.29         0.90\g    0.82-0.99\g           1.22      1.02-1.46            0.97      0.89-1.05
Gould 1995              0.88   0.72-1.07        1.21   1.02-1.43        1.09    0.95-1.24           0.94      0.86-1.02           1.17      0.98-1.40            0.99      0.91-1.07
Rossouw 1995            0.91   0.75-1.11        1.25   1.06-1.47        1.08    0.95-1.24           0.89      0.81-0.99           1.07      0.87-1.34            0.94      0.85-1.04
------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
\a Complete facts of publication are given in the bibliography. 
Intervals are 95-percent confidence intervals for the common odds
ratios.  The meta-analyses did not report data for the empty cells. 

\b For consistency, we recalculated the results from risk ratios and
combined diet and drug treatments for primary and secondary
prevention trials.  Using the Breslow-Day (B-D) method, we tested for
homogeneity of the odds ratios.  The B-D statistic for primary
prevention trials was 4.34 (p = .30) for CHD fatality and 6.35 (p =
.17) for non-CHD fatality; for secondary prevention trials, it was
19.42 (p = .11) for CHD fatality and 4.68 (p = .46) for non-CHD
fatality. 

\c These data are for drug trials.  Separately reported results for
diet trials were primary prevention total fatality 1.07, 1-1.5;
secondary prevention total fatality 0.98, 0.81-1.19. 

\d These data are the result of our recalculating Muldoon's data
after we included the WHO 1992 intention-to-treat analysis furnished
by Muldoon.  Using the Breslow-Day (B-D) method, we tested for
homogeneity of the odds ratios.  The B-D statistic was 9.30 (p =
.10).  Muldoon's results excluding this analysis were 0.85,
0.69-1.05. 

\e Odds ratios are relative odds per 0.6 mmol/l (10 percent)
cholesterol reduction. 

\f For consistency, we recalculated the results from percentage
change in risk for primary and secondary prevention trials and
recalculated the odds ratios given for primary and secondary trials
combined in order to obtain confidence intervals.  Using the
Breslow-Day (B-D) method, we tested for homogeneity of the odds
ratios.  The B-D statistic for primary prevention trials was 4.46 (p
= .22) for CHD fatality, 2.04 (p = .56) for non-CHD fatality, and
4.50 (p = .21) for total fatality.  For secondary prevention trials,
it was 29.68 (p = .05) for total fatality; we used Gordon's
calculations for CHD and non-CHD fatalities. 

\g Gordon reported these data as statistically significant with a
0.996 upper bound of the confidence interval.  Tests for homogeneity
of odds ratios were not reported. 

Looking at the ranges of the odds ratios shows that
cholesterol-lowering treatment has nearly the same effect size in
primary prevention trials (0.93 to 0.85) as in secondary prevention
trials (0.96 to 0.84).  Reductions in CHD death rates in the
secondary prevention trials favored the treatment groups; moreover,
differences between treatment and nontreatment groups were
statistically significant, except in Ravnskov (1992) and Gould
(1995).\6

Putting these results in context requires looking at patients' risk
levels.  For example, Smith and his colleagues demonstrated in 1993
that the magnitude of net benefit depends on the level of CHD risk. 
They found statistically significant reductions in CHD death rates
only for treatment groups with the highest risk.  The common odds
ratios for participants treated for elevated cholesterol who were at
lower risk showed that their rate of CHD death increased.\7

Although odds ratios are similar for primary and secondary trials in
terms of absolute risk, lowering cholesterol has the clearest benefit
for participants in secondary prevention trials who are at greater
risk and who have higher fatality rates.  For example, an analysis
directly comparing drug interventions in primary and secondary
prevention trials estimates that one death from CHD could be
prevented by treating 675 participants in primary prevention trials
but only 38 participants in secondary prevention trials (Silberberg,
1991).\8

In reporting an additional analysis that considered how much
cholesterol was reduced, Gould (1995) found that for every 10
percentage points of cholesterol-lowering, CHD death rates fell 13
percent (p < .002). 


--------------------
\6 Ravnskov (1992), Gordon (1995), and Gould (1995) were the only
meta-analysts examining CHD fatalities in secondary prevention trials
who included hormone trials.  Gordon and Gould included the
dextrothyroxine arm of the large Coronary Drug Project.  Ravnskov
included all three hormone arms--estrogen 2.5 mg, estrogen 5.0 mg,
and dextrothyroxine--but all three were discontinued early in this
trial because of their toxic side effects. 

\7 This analysis was stratified by the degree of risk of CHD death
that is represented by the CHD fatalities in the nontreatment groups. 
The odds ratios (and confidence intervals) were 0.74 (0.60-0.91) for
participants at high risk of CHD fatality (more than 50 deaths per
1,000 person years); 0.92 (0.77-1.09) for participants at medium risk
(10-50 deaths per 1,000 person years); 1.15 (0.80-1.64) for
participants at lower risk (fewer than 10 deaths per 1,000 person
years).  See G.  D.  Smith, F.  Song, and T.  A.  Sheldon,
"Cholesterol Lowering and Mortality:  The Importance of Considering
Initial Level of Risk," British Medical Journal, 306 (1993), 1367-73. 

\8 Silberberg's meta-analysis did not include treatments with the
recent statin drugs.  These drugs' greater ability to lower
cholesterol may affect CHD fatality and, in turn, change these
numbers.  For example, in 4S, which used simvastatin, the estimated
number of those who had to be treated to prevent one CHD death was
29. 


      NON-CHD FATALITIES
------------------------------------------------------ Appendix II:3.3

The investigators who examined deaths from noncoronary causes as an
outcome in primary and secondary prevention trials also focused on
the potentially adverse effects of specific treatments used in
lowering cholesterol and the concern that lowering it may simply
exchange one cause of death for another.  Table II.2 shows that
decreases in CHD fatality rates in primary prevention trials were not
sufficient to offset non-CHD fatality rates.  As a result, total
fatality rates were unfavorable for treatment groups, but differences
from nontreatment groups were not statistically significant.  The
rates of death from non-CHD causes were higher for participants who
received cholesterol- lowering treatment; the higher rate was
statistically significant in all but one meta-analysis.  The common
odds ratios range from 1.11 to 1.26. 

Similarly, the non-CHD death rate was mostly higher for treatment
groups than for nontreatment groups in secondary prevention trials. 
The common odds ratios range from 0.99 to 1.22.  Except in Gordon
(1995), the differences between treatment and nontreatment groups
were not statistically significant.  In the secondary prevention
trials, unlike the primary, reductions in rates of CHD fatalities do
appear to have offset the higher non-CHD fatalities but not to the
extent of achieving a statistically significant reduction in overall
deaths. 

Smith and his colleagues in 1993 reported a favorable but not
statistically significant reduction in non-CHD fatality rates for
higher-risk treatment groups but unfavorable non-CHD fatality rates
for medium-risk and lower-risk treatment groups.  The odds ratios
show that compared to nontreatment groups, non-CHD fatality rates
increased.  For the lower-risk participants, the increase was
statistically significant.\9 Drug treatment trials, but not nondrug
trials, showed higher death rates from causes other than CHD. 

One problem in interpreting these data is that in some trials, deaths
from causes other than heart attack were not reported.  The earlier
meta-analyses either attributed the increases in non-CHD death in
cholesterol-lowering intervention trials to chance or considered the
increases reasonable given that CHD deaths decreased and that death
from other causes could be expected.  Because non-CHD deaths recorded
in individual trials were few in number, the greater statistical
power of meta-analysis could potentially find any real differences
between treatment and nontreatment groups in noncoronary fatality
outcomes.  However, large individual trials with adverse effects that
pertain to specific treatments can dominate the findings of a
meta-analysis that combines such trials with smaller ones. 

The results from several sizable individual trials have alerted the
medical community to the risks associated with fibrates,
dextrothyroxine, and high estrogen dosage in men.  The biological
means through which these drugs lower cholesterol plausibly account
for the higher non-CHD death rates.\10 Less readily explained are the
higher, statistically significant rates of fatality not related to
illness (as from accident, violence, and suicide) among participants
in treatment groups compared to nontreatment participants in primary
prevention trials.  Muldoon (1993) addressed these findings.\11 The
increases in non-CHD death rates have led recent investigators to
examine whether lowering cholesterol is itself harmful or whether the
treatment that lowers it is responsible for the adverse effects. 


--------------------
\9 The odds ratios (and confidence intervals) were 0.95 (0.65-1.40)
for participants at high risk, 1.07 (0.94-1.21) for participants at
medium risk, and 1.33 (1.09-1.63) for participants at lower risk. 

\10 For example, non-CHD fatality was higher and statistically
significant in the World Health Organization primary prevention trial
that used clofibrate.  These deaths were caused by stroke, cancers
(mainly of the liver and gastrointestinal tract), gallbladder
disease, and CHD that may have been related to drug toxicity.  The
Coronary Drug Prevention secondary prevention trial also weighs
heavily in a meta-analysis; as noted above, the three hormone
interventions were discontinued early because of their adverse
effects. 

\11 He found statistically significant increases in fatality rates
not related to illness in treatment groups for primary but not for
secondary prevention trials and in drug but not diet interventions. 
The results of the analysis suggest an association between greater
nonillness-related fatalities and cholesterol- lowering but are not
regarded as definitive. 


      TOTAL FATALITIES
------------------------------------------------------ Appendix II:3.4

Table II.2 reveals greater increases in rates of death from any cause
in treatment groups in primary prevention trials.  The odds ratios
range from 1.02 to 1.20.  Differences between treatment and
nontreatment groups were not statistically significant in any
meta-analysis.  The confidence intervals extend on either side of 1. 
Holme (1993), who aggregated diet and drug primary prevention trials
separately, is an exception. 

The odds ratios range from 0.90 to 1.02 for total fatalities in
secondary prevention trials.  They all fall in a favorable direction
except in Ravnskov (1992), although only Law I (1994) shows
statistically significant reductions in the rate of total deaths in
cholesterol-lowering treatment groups.\12 In the remaining
meta-analyses, the confidence intervals extend on either side of 1. 
Therefore, the possibility cannot be ruled out that overall death
rates in secondary prevention trials increase in cholesterol-lowering
treatment groups; the same is true for primary prevention trials. 

Smith and his colleagues found a net benefit in terms of overall
survival but only among trial participants whose initial CHD risk was
highest.\13 Lower-risk participants in treatment groups had higher,
statistically significant death rates compared to those in
nontreatment groups.\14

The failure of cholesterol-lowering treatment to extend the life span
has become an active topic of investigation and is debated among
meta-analysts and others.  Some investigators have shown that total
fatality results may be mediated by the degree to which cholesterol
is lowered.  Holme, for example, found in 1992 that when cholesterol
was not lowered at all, the risk of death rose about 10 percent.  He
also noted that cholesterol had to fall at least 8 to 9 percent to
outweigh treatment's adverse side-effects on overall survival
rates.\15 In his 1993 meta-analysis, Holme noted that greater
reductions in cholesterol were related to greater reductions in total
fatalities.  In contrast, Ravnskov (1992) found no relationship
between the net mean cholesterol reduction in each trial and total
mortality.\16

However, Gould (1995) demonstrated that for every 10 percentage
points that cholesterol fell, total fatality rates also fell 10
percent (p < .03).  When Gordon (1995) regrouped the trials he had
analyzed by whether cholesterol reductions were greater or less than
the overall median of 12 percent, he found 11 trials that exceeded
this median and that reduced total fatality by 20 percent (p <
.002).\17 These findings are consistent with those of the 4S
secondary prevention trial, in which total cholesterol fell by 25
percent and total fatalities fell by 30 percent. 


--------------------
\12 Law, Thompson, and Wald weighted the odds ratio from each trial
by a 0.6 mmol/l (about 10 percent) reduction is serum cholesterol
concentration. 

\13 This net benefit can be expected in nontreatment groups in which
the rate of CHD death is greater than 3 percent a year.  Conversely,
total mortality can be expected to increase when CHD death in
untreated participants is less than 3 percent. 

\14 The odds ratios (and confidence intervals) were 0.74 (0.60-0.92)
for participants at high risk, 0.96 (0.90-1.02) for participants at
medium risk, and 1.22 (1.06-1.42) for participants at lower risk. 

\15 I.  Holme, "Meta-Analysis of Cholesterol Reduction Trials: 
Coronary Disease and Mortality," Primary Cardiology, 18:7 (1992),
63-70. 

\16 Law I (1994) criticized Ravnskov's analysis because he did not
specify his methods and, therefore, his analysis cannot be
replicated. 

\17 Gordon's regrouped analyses are not reflected in table II.3,
which reports only common odds ratios and confidence intervals for
CHD, non-CHD, and total fatalities. 


   A FINAL NOTE
-------------------------------------------------------- Appendix II:4

The results from the meta-analyses point to important issues
warranting further investigation into the potentially adverse effects
of cholesterol-lowering interventions.  Collaborative meta-analyses
that have been planned will include results from recently completed
and ongoing statin trials that may help resolve questions about the
effects of lowering cholesterol.  Because such analyses will be
prospective, they will have the advantage of specifying in advance
the relationships to be tested. 


DATA GAPS IN RANDOMIZED CLINICAL
TRIALS
========================================================= Appendix III

A formal accounting of the data gaps in the clinical trials shows
that much data are still needed from direct clinical trial evidence,
or from complementary nontrial data, if patients, doctors, and
policymakers are to be well-informed about the benefits and risks of
lowering cholesterol.  NCEP's evolution has been toward recommending
less treatment for less-researched groups; however, NCEP's policy
would continue intensive, physician-directed cholesterol-lowering for
several of these groups. 

In this appendix, we report the kinds of data we did not find when we
examined the individual trials aggregated in the meta-analyses we
studied.  We also report on whether the published commentary of
NHLBI, other cholesterol policy groups, and researchers proposing new
cholesterol-lowering trials confirms the gaps we found.\1 We also
discuss recent and proposed clinical trials that may fill the data
gaps by 2000. 

Useful randomized clinical trials data are generally lacking for
women, minority men and women, elderly men and women, young men, the
majority of Americans with the most common total cholesterol readings
(those between 200 and 231 mg/dl), and groups whose risk of heart
disease is moderate.  Clinical trial data on the most common modern
interventions--low-fat diets and widely prescribed
cholesterol-lowering drugs--are minimal.  The effect of lowering
cholesterol on non-CHD and total fatality rates has been poorly
explored. 


--------------------
\1 Virtually all the clinical trials we studied were examined for the
NCEP guidelines. 


   COMPLETED CLINICAL TRIALS
------------------------------------------------------- Appendix III:1

The clinical trials that have evaluated cholesterol-lowering
treatments represent diverse treatments, risk categories, overall
sample sizes, and trial durations.\2 Most looked at lowering
cholesterol through a single intervention, the main ones being diet
and drugs, but several studied other interventions.  Most studies
focused on people who had existing symptoms of CHD (secondary
prevention trials) while others concentrated on those who had no
evidence of CHD (primary prevention trials) and a few combined both. 
Many trials were very small, half being constituted of fewer than 500
subjects, while the largest included more than 10,000.  Although
virtually all trials intended to assess CHD outcomes, not all trials
reported the same outcome variables.  Table III.1 summarizes the
individual clinical trials we examined. 



                                                                                      Table III.1
                                                                        
                                                                           Basic Features of 42 Cholesterol-
                                                                                   Lowering Trials\a



                                                               Duration            No. of
Trial by prevention type\b       Treatment                      (years)      participants  Gender               Absolute        Average         Ethnic group            Cholesterol\c
-------------------------------  ----------------------  --------------  ----------------  ---------------  --  --------------  --------------  ----------------------  --------------
Primary

LA VA (1969, 1971)\d             Diet                                ï¿½8               846  Men                                  66 W            hite 90%; 2 nonwhite    33
                                                                                                                                                10%

Oslo DA (1981)\e                 Diet, other                      6-7.5             1,232  Men                  40-49           45              3                       25

MRFIT (1982)\e                   Diet, other                        6-8            12,866  Men                  35-57           46              2                       53

WHO F (1983, 1986)\e             Diet, other                        5-6          49,784\f  Men                  40-59           48              2                       16

Gothenburg (1986)\e              Diet, other                         10            20,015  Men                  47-55                                                   250

Minnesota (1975, 1989)\d         Diet                                 1             9,057  Men 49%; women       30-70                                                   207
                                                                                           51%

Retinopathy (1969)\g             Clofibrate                           1                63  Men and women        40-59                                                   250+

Begg (1971)\h                    Clofibrate                           5               155

Upjohn (1978)\d                  Colestipol                           2             2,278  Men 48%; women                       Men 51; women   Men: white 86%,         Men 308; women
                                                                                           52%                                  57              nonwhite 14%; women:    321
                                                                                                                                                76%, nonwhite 24%

Finnish (1985)\e                 Diet, clofibrate,                    5             1,222  Men                  40-55           48                                      275
                                 probucol, other

Helsinki (1987, 1988)            Gemfibrozil                          5             4,081  Men                  40-55                                                   270

WHO (1978, 1980, 1984, 1992)     Clofibrate                           5            10,627  Men                  30-59                                                   248

LRC CPPT (1984a, 1984b, 1992)    Cholestyramine                       7             3,806  Men                  35-59           48              White                   288

Excel (1990, 1991, 1992)\d       Lovastatin                         0.9             8,245  Men 59%; women       18-70           Men 54; women   White 92%; nonwhite 8%  266
                                                                                           41%                                  58


Secondary
--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
Corn Oil (1965)                  Diet                                 2                80                                       55                                      259

MRC Low Fat (1965)               Diet                                 3               264  Men                  Up to 65                                                263

Oslo DH (1966, 1970)             Diet                                 5               412  Men                  30-64           56                                      296

MRC Soya (1968, 1974)            Diet                               2-7               393  Men                  Up to 60                                                270

Sydney (1978)                    Diet                               2-7               458  Men                  30-59                                                   282

DART (1989)                      Diet                                 2             2,033  Men                  30-69           57                                      250

India (1992)                     Diet                                 1               406  Men 90%; women                       51                                      227
                                                                                           10%

Estrogen (1961)                  Estrogen                             5               100  Men                  35-64                                                   236

Long-Term Estrogen (1962)        Estrogen                             5               432  Men                  50-70                           White 74%; Jewish 11%;
                                                                                                                                                black 10%; Mexican 5%

Chicago (1963)                   Estrogen                             5               275  Men                  Up to 50                                                248

Newcastle (1971)                 Clofibrate                         3.5               497  Men 80%; women                       Men 52; women                           Men 246; women
                                                                                           20%                                  54                                      265

Scottish (1971, 1972)            Clofibrate                           3               717  Men 83%; women       40-69           Men 52; women                           Men 269; women
                                                                                           17%                                  54                                      290

Acheson (1972)                   Clofibrate                           7                95  Men 68%; women                                                               288
                                                                                           32%

St. Vincents (1973)              Colestipol                         1-3                52  Men 29%; women       36-80           Men 55; women                           310
                                                                                           71%                                  58

Veterans Cardiology (1968,       D-thyroxine; D-                      5               570  Men                  28-75           51              White 92%; nonwhite 8%  241
1974)                            thyroxine + estrogen;
                                 estrogen; estrogen +
                                 nicotinic acid;
                                 nicotinic acid

Veterans W. Roxbury (1981)\d     Probucol                             1               118  Men                                  50              White 98%; black 2%     305

NHLBI (1984)                     Cholestyramine\i                     5               143  Men 81%; women       21-55                                                   325
                                                                                           19%

CDP (1970, 1972, 1973, 1975,     Clofibrate, D-                       6             8,341  Men                  30-64           7               White 93%; nonwhite     251
1986)                            thyroxine, estrogen                                                                                            %\j
                                 2.5 mg, estrogen 5 mg,
                                 niacin

CLAS (1987)                      Colestipol, niacin\i                 2               188  Men                  40-59                                                   244

Stockholm (1977, 1980, 1988)     Clofibrate, nicotinic                5               555  Men 80%; women                       Men 59; women                           248
                                 acid                                                      20%                                  63

FATS (1990)                      Lovastatin +                       2.5               146  Men                  Up to 62                                                271
                                 colestipol; niacin +
                                 colestipol\i

Lifestyle (1990)\e               Diet, other\j                        1                48  Men and women        35-75                           2                       36

POSCH (1990)                     Partial ileal bypass               9.7               838  Men 90%; women       30-64           51              2                       51
                                 surgery                                                   10%

SCOR (1990)\k                    Colestipol,                          2                97  Men 43%; women       19-72                           3                       72
                                 lovastatin, niacin\i                                      57%

Restenosis (1991, 1992)          Lovastatin\i                         2               157  Men and women        6               0               20                      9

STARS (1992)                     Diet, diet +                         3                90  Men                  Up to 66                        2                       79
                                 cholestyramine\i

Ancillary Helsinki (1993)        Gemfibrizol                          5               628  Men                  4               9               27                      1

MARS (1993)                      Lovastatin\i                         2               270  Men 91%; women       35-67           58              2                       30
                                                                                           9%
--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
\a Data were not available for empty cells. 

\b Complete facts of publication for the reports of the individual
trials are given in the bibliography. 

\c Mean baseline cholesterol in mg/dl. 

\d This trial had primary and secondary prevention components;
primary predominated. 

\e A multifactor trial--that is, a trial that targets multiple risk
factors simultaneously with more than one intervention. 

\f This trial randomized 66 factories comprising 49,784 men. 

\g Participants were being treated for diabetic retinopathy.  About
40 percent were diagnosed as having peripheral or coronary vascular
disease. 

\h Little information from this report is available in English. 

\i A trial in which angiography measured changes in cardiovascular
lesions effected by lowering lipid levels.  While angiography was the
main measured result, clinical outcomes were also recorded. 

\j These percentages are from the clofibrate, nicotinic acid, and
nontreatment arms. 

\k Participants had heterozygous familial hypercholesterolemia. 


--------------------
\2 The trials were conducted mainly in the United States and Europe
and were funded by government agencies or the private sector.  NIH
has funded a relatively small number of the trials but some of these
have been among the largest, such as MRFIT and LRC. 


      TREATMENTS
----------------------------------------------------- Appendix III:1.1

Many of the trials that used diet to lower cholesterol were published
before 1985, but most of them were designed in the 1960s and 1970s
and do not reflect current judgments about the need to reduce
saturated fat and overall fat.  They represent participants in
institutions whose experience may not typify that of individuals who
are outside institutions and concerned about their cholesterol. 
Similarly, the completed drug trials we examined hardly included
tests of today's most commonly prescribed preventive drugs--the
statin drugs and estrogen replacement therapy for women.  Few trials
have studied these interventions, although some trials of statin
drugs have recently been completed. 


      POPULATION GROUPS
----------------------------------------------------- Appendix III:1.2


         WOMEN
--------------------------------------------------- Appendix III:1.2.1

More than half of the trials we examined either did not include women
as participants or did not report that they did.  The remainder
included in total more than 8,500 women.  Women usually constituted
less than 25 percent of the participants in a trial; in some, their
numbers were too small to be analyzed.\3 Where women numbered more
than 25 percent of a trial's participants, the trial either had very
few participants or was too brief to yield clinical differences in
coronary outcomes.  In some cases, the major publication reporting
trial results did not fully include the women's data. 

The three trials that included at least 500 women reveal why so
little concrete evidence on the efficacy of lowering women's
cholesterol resulted from them.  The Minnesota primary prevention
dietary trial in mental hospitals has limited value because the diets
are no longer recommended, the study design suffered from the
shifting of the institutional population, and it is questionable
whether an institutional diet is applicable to the general
population.  The data on women in the Upjohn primary prevention trial
were not included in the original analysis.  The EXCEL trial was too
brief (0.9 year) to show a clinical benefit.\4


--------------------
\3 Because women's rates of coronary events are much lower than men's
until after the age of 70, a small group of women in a trial is not
likely to be analyzed separately. 

\4 When Walsh and Grady pooled the primary prevention data on women
who had no history of heart disease, they found no significant
difference in CHD fatality in the group whose cholesterol was
lowered.  When they pooled the secondary prevention data on women who
had a history of heart disease, they did find a significant benefit
in terms of CHD fatality.  See J.  M.  E.  Walsh and D.  Grady,
"Treatment of Hyperlipidemia in Women," Journal of the American
Medical Association, 274:14 (1995), 1152-58. 


         THE OLD AND THE YOUNG
--------------------------------------------------- Appendix III:1.2.2

Most clinical trials have focused on middle-aged men, although
participants' ages are not consistently reported.  Many of the trials
set upper age limits; when average ages are given, they ranged from
45 to 66.  The young and the old are therefore underrepresented in
the clinical trials data. 


         MINORITY MEN AND WOMEN
--------------------------------------------------- Appendix III:1.2.3

Almost no clinical trials data exist on groups other than white men. 
No clinical trials focus on the largest U.S.  minority groups--blacks
and Hispanics.  Therefore, we have little understanding of the
efficacy of lowering cholesterol for the genetic diversity
represented in the United States. 


      RISK PROFILES
----------------------------------------------------- Appendix III:1.3

Most clinical trials focused on middle-aged white men who had already
had a heart attack or were otherwise at very high risk in order to
make the trials practicable:  these were the people who would
experience the largest number of CHD events in the trials'
2-to-5-year periods.  When the researchers did recruit non-CHD
patients, even these were middle-aged white men who had multiple risk
factors.  Very few trials other than those in institutions included
participants whose risk was low or moderate.  Thus, even the primary
prevention trials typically included very-high-risk groups. 

The participants of all but 14 of the 42 trials either had had a
heart attack or had been diagnosed with CHD.  It is not known how the
results from the high-risk primary prevention trials or from the
secondary prevention trials apply to non-CHD participants whose
cholesterol levels are only moderate and who have few risk factors. 

The trial participants commonly had high cholesterol levels, as shown
in table III.1.  According to the NCEP guidelines, total cholesterol
between 200 mg/dl and 239 mg/dl is borderline high and beyond the
desirable, but there are few clinical trials data on persons within
this range.  When we compared average baseline cholesterol levels in
the trials to the most recent U.S.  cholesterol distribution, we
found that most of the trials chose participants with total
cholesterol levels of 240 mg/dl or higher, while most U.S.  adults
had total cholesterol levels below that.\5 (See figure III.1.)

   Figure III.1:  Cholesterol
   Levels in the Clinical Trials
   Compared to U.S.  Adult Total
   Cholesterol Distribution

   (See figure in printed
   edition.)

Thus, there is a gap in what we know about the most common
cholesterol levels.  The NCEP guidelines' recommendations for
lowering common cholesterol levels have largely been generalized from
trials whose participants had higher cholesterol levels and from
observational and other studies.  In contrast to the clinical trials,
the observational studies did not lower cholesterol; instead, they
compared disease rates at different cholesterol levels and found
better CHD outcomes at lower total cholesterol levels. 


--------------------
\5 In the most recent U.S.  national sample, 75 percent of adult
cholesterol had fallen to below 231 mg/dl, average adult cholesterol
to 204 mg/dl. 


   TRIALS PROPOSED AND UNDER WAY
------------------------------------------------------- Appendix III:2

Table III.2 summarizes the data gaps that NHLBI, other policy groups,
and the designs for proposed trials have identified.  It demonstrates
the considerable agreement among these sources and supports our
conclusions.  It also indicates that newer studies are attempting to
fill some of the gaps by the end of the century.\6



                              Table III.2
                
                Data Gaps in Cholesterol-Lowering Trials
                 and 13 New Studies Attempting to Fill
                                 Them\a


                                                                No. of
                                           Other        New        new
Data gap                        NHLBI   policies     trials  studies\b
--------------------------  ---------  ---------  ---------  ---------
Treatment
----------------------------------------------------------------------
Diet                                -          -          -          1
Statin drugs                        -          -          -         11
Estrogen\c                          -          -          -          1
Other                               -          -          -          4

Population
----------------------------------------------------------------------
Women                               -          -          -         11
Younger than 40, older              -          -          -          9
 than 60
Non-Caucasian                       -         \d         \d          3

Risk group
----------------------------------------------------------------------
Low-risk non-CHD                    -          -          -          3
Borderline-high                     -          -          -          5
 cholesterol
Low HDL cholesterol                 -          -          -          3
Genetic variation                   -          -         \d          1

Outcome
----------------------------------------------------------------------
Non-CHD events                      -          -          -          5
Fatal CHD                                      -          -          6
Total fatality                      -          -          -          4
Quality of life                     -          -          -          5
Statistical power\d                 -          -          -         12
----------------------------------------------------------------------
\a Sources include published and unpublished information available
from the U.S.  General Accounting Office, Program Evaluation and
Methodology Division.  "Other policies" include non-NHLBI national
policies and other sources such as interviews and selected published
meta-analyses. 

\b Includes only trials with 1,000 or more participants. 

\c We did not review the several estrogen replacement studies as
cholesterol-lowering trials.  We did, however, include the Women's
Health Initiative because it combines estrogen therapy with low-fat
diet. 

\d The source did not identify a gap. 


--------------------
\6 New studies include trials too recent to be included in the
meta-analyses of the 42 trials we describe.  Some new studies are
planned; others are under way; several have been completed. 


      TREATMENTS
----------------------------------------------------- Appendix III:2.1

Most of the new trials will use the widely prescribed and highly
effective cholesterol-lowering statin drugs.  Estrogen replacement
therapy is under study in preventing cardiac disease in
postmenopausal women, but it is not a cholesterol-lowering treatment. 


      POPULATION GROUPS
----------------------------------------------------- Appendix III:2.2

Women will be included in 11 of 13 new trials, but because their
numbers will generally be restricted to 20 percent or less, these
trials may or may not allow analysis of the women's data.\7

The Women's Health Initiative will incorporate 48,000 older women in
a randomized trial of a fat-restricted diet and 25,000 women in an
estrogen-replacement trial, but it reportedly has design flaws. 
Critics of this trial have reservations about its similarities to
MRFIT, another large multiple-intervention trial, in which the
results were difficult to attribute to specific interventions. 
Several commentators have also expressed concern that ensuring that a
low-fat diet is actually maintained for 10 years seems infeasible. 

As for ethnic minorities, only ALLHAT will include a large number of
African Americans.  Indeed, it appears to be bearing the burden of
past studies' limitations regarding women, black men and women, and
elderly persons.  Since it studies more than one intervention, trying
to lower both cholesterol and blood pressure, it may be difficult to
attribute differences in results, if any, to particular sources. 


--------------------
\7 Two exceptions are the Women's Health Initiative, with only women
participants, and the Anti-hypertensive and Lipid-Lowering Treatment
to Prevent Heart Attack (ALLHAT) study, with fewer than 50 percent
women. 


      RISK PROFILES
----------------------------------------------------- Appendix III:2.3

Three new trials will focus on persons who have never had a heart
attack; 5 others will include mainly persons with borderline-high
cholesterol levels (200 to 239 mg/dl).  These trials will go a long
way toward filling critical gaps in what is known about these
moderate-risk groups. 


      OUTCOMES
----------------------------------------------------- Appendix III:2.4

Four of the new trials are planned to detect changes in total
fatalities.  Not even the largest of the previous trials found
significant reductions in total fatalities.  Although the planned
trials will mostly study more than 3,000 participants, their designs
may still not yield data that will fill the gaps.  Investigators are
counting on substantial lowering of cholesterol from the new drugs to
reduce total fatalities, within the limits of the participant
populations and trial durations. 


      STATISTICAL POWER
----------------------------------------------------- Appendix III:2.5

Most of the new studies will increase statistical power over that of
the 42 existing studies that we reviewed by recruiting large numbers
of participants, primarily participants whose risk for coronary
events is very high, or by running their trials longer.  There is a
prospective plan to combine the analyses of several of the trials in
a meta-analysis that may yield more information than individual
trials.  All this suggests that some of the constraints on the
statistical power of past trials may be overcome, providing answers
to several outstanding questions about how to reduce the rates of CHD
and non-CHD fatality and how to lengthen the survival of persons who
have certain attributes.  However, pressure to recruit high-risk
participants continues; thus, answers to several remaining questions
will be found at the expense of finding out whether lowering
cholesterol benefits lower-risk groups. 


CHOLESTEROL POLICY IN THE UNITED
STATES AND ABROAD
========================================================== Appendix IV

Cholesterol policies to prevent coronary heart disease by lowering
cholesterol in the United States and other nations typically follow
two complementary and simultaneous strategies:  (1) a
population-based strategy that educates the general public about
dietary change and (2) a high-risk strategy that identifies persons
whose high cholesterol levels warrant physician-directed measures to
reduce CHD risk.  The policies differ considerably, though, in the
breadth of cholesterol-screening they recommend and in their
definition of "high risk." NCEP recommends testing all adult blood
cholesterol levels and suggests various intensities of
cholesterol-lowering treatment depending on individual risk.  The
guidelines indicate that 29 percent of U.S.  adults need a
physician's assistance to lower their cholesterol.  Several other
U.S.  and foreign policies recommend screening and intensive
cholesterol-lowering for fewer adults. 

Coronary heart disease, a major cause of death and disability in most
developed nations, is believed to be partly the result of high-fat
diets, little exercise, obesity, and smoking and, thus, appears to be
preventable.  Formulating policy, however, is complicated by the
interpretation of the underlying data.  A large number of clinical
trials have tested cholesterol-lowering treatments.  Although
clinical trials are the most rigorous type of evidence for
establishing the efficacy of a medical treatment, they do not include
the range of patients, treatments, and outcomes in medical practice,
limiting the ability to predict risks and benefits for many groups. 

Further, while small average reductions in a population's cholesterol
may result in dramatic reductions in CHD fatality rates, this may
provide only a negligible chance of improvement for any particular
person.  This has been termed the paradox of public health
prevention.\1 While CHD rates rise with rising cholesterol levels,
the ability to predict outcomes for individuals is poor.  Many who
have died from heart disease did not have elevated cholesterol
levels. 

The large 1982 MRFIT trial demonstrates the paradox.  Men whose total
cholesterol measured 291 mg/dl or higher constituted only 2 percent
of the total study group and 2 percent of total CHD deaths (figure
IV.1), but they represent the highest rate of CHD death (figure 1). 
Meanwhile, men whose total cholesterol was between 213 and 271 mg/dl
experienced 63 percent of the CHD deaths, by far the majority.  Thus,
any who would make national cholesterol policy must determine whether
to focus on (1) physician- directed cholesterol-lowering efforts at
the highest end of the cholesterol range, where men have the greatest
individual likelihood of having a heart attack, or (2) preventive
activity in the middle range, where most heart attacks occur but
individual risk is much lower.  The latter choice would apply the
high-risk procedures to many people with very low individual risk
because the ability to predict individual outcomes is poor. 

   Figure IV.1:  Cholesterol and
   CHD Death Distributions in the
   MRFIT Trial\a

   (See figure in printed
   edition.)

\a More than 355,000 men screened for the MRFIT trial were followed
for 6 years.  Cholesterol levels were lower than 155 mg/dl in about 9
percent of the CHD deaths.

Source:  Adapted from Geoffrey Rose, The Strategy of Preventive
Medicine (New York:  Oxford University Press, 1992), p.  23. 

The various cholesterol policymaking groups here and abroad have
handled differently the data limitations and the paradox of
prevention.  For example, several North American policy groups
concluded that applying an aggressive cholesterol-lowering policy to
women, young men, and elderly persons does not seem warranted because
of the absence of trial data and the lack of support from nontrial
studies.  Similarly, several foreign policies have set drug
cholesterol-lowering targets closer to those included in the trials
and, therefore, considerably higher than NCEP's targets.  Overall,
the policies cover a wide spectrum, from screening and treating
narrowly to NCEP's more comprehensive policy. 


--------------------
\1 Geoffrey Rose, The Strategy of Preventive Medicine (New York: 
Oxford University Press, 1992). 


   THE NATIONAL CHOLESTEROL
   EDUCATION PROGRAM
-------------------------------------------------------- Appendix IV:1

The 1984 NIH consensus development conference concluded from the
accumulated evidence of clinical, epidemiologic, metabolic, and
animal research that CHD risk is related to serum cholesterol levels. 
A recently completed NHLBI-funded study of coronary heart disease
among high-risk men who had never had a heart attack had found a
significant reduction in combined CHD in the group whose cholesterol
was lowered by means of drugs.  The conference therefore proposed
that Americans whose total blood cholesterol was above the 75th
percentile of the population distribution be given advice and
treatment to lower it.  The conference also recommended the
development of NCEP.  Table IV.1 summarizes the past 12 years of NCEP
policy. 



                               Table IV.1
                
                 National Cholesterol Education Program
                         Milestones Since 1984

Year      Action              Background
--------  ------------------  ----------------------------------------
1984      NIH consensus       CHD associated with high total
          conference held     cholesterol

                              Population distributions of cholesterol
                              recommended for treatment: 75th
                              percentile for diet, 90th for drugs

                              Diet and exercise recommended as
                              cholesterol-lowering treatments

1987      First adult         Highest desirable cholesterol level set
          treatment           at 50th percentile
          guidelines issued

                              Adult cholesterol measurement set at 5-
                              year intervals

                              Treatments established: diet first,
                              followed by drugs

1988      Laboratory
          standards issued

1990      Population
          strategy issued

1991      Children's
          strategy issued

1993      Adult guidelines    High risk redefined for treatment
          revised              Drug treatment reserved for groups
                              with high CHD rates
                               HDL declared beneficial
                               Drug treatment delay recommended for
                              men younger than 45 and women younger
                              than 55
----------------------------------------------------------------------
The guidelines of 1987 focused on identifying and treating high-risk
adults but also recommended regular universal screening, defined
various cholesterol levels as alerts and goals, and established a
method of classifying a person's CHD risk.  Recommendations were made
in 1990-91 for cholesterol-screening for both the general population
and children and adolescents.  The adult treatment guidelines were
revised in 1993, although they continue recommendations for universal
adult screening and cholesterol levels.  The current adult guidelines
refine the risk assessment leading to treatment and differentiate
cholesterol goals for different groups. 


      NCEP'S DEFINITION OF CHD
      RISK
------------------------------------------------------ Appendix IV:1.1

The NCEP guidelines recommend physician-directed interventions for
high-risk adults and diet and exercise changes for the entire
population.  NCEP's definition of desirable, borderline, and high
total blood cholesterol levels is uniform for men and women of all
adult ages: 

  -- desirable = 200 mg/dl or less

  -- borderline high = 200-239 mg/dl

  -- high = 240 mg/dl or higher.\2

Risk factors that NCEP believes are associated with CHD and that can
be modified are high blood cholesterol, obesity, physical inactivity,
hypertension, diabetes, and cigarette smoking.  Unmodifiable factors
that also influence the probability of CHD and are, thus, included in
the assessment of patient risk and decisions to treat include
increasing age, being male, and having a family history of CHD. 

The total cholesterol of about half of all adult Americans is higher
than 200 mg/dl; the policy requires medical consideration of their
total coronary risk profile, including risk factors.  For this group,
NCEP recommends intensive treatment to lower cholesterol for persons
who have coronary heart disease or for those with no evidence of CHD
but with two or more risk factors such as smoking or having a family
history of coronary heart disease.  Individuals from groups with
moderate to low rates of CHD can also be recommended for vigorous
cholesterol-lowering by NCEP's high-risk policy if they have two risk
factors.  For example, a man whose LDL cholesterol was 190 mg/dl or
more would be a candidate for aggressive cholesterol-lowering if he
were older than 45. 


--------------------
\2 According to the NCEP Adult Treatment Panel, these numbers were
based partly on the arbitrary 50- and 75-percent levels for the
population distribution of cholesterol and partly on the observation
of the increasing rates of CHD incidence among persons with more than
200 mg/dl total cholesterol in the MRFIT study.  See NIH, Expert
Panel on Detection, Evaluation, and Treatment of High Blood
Cholesterol in Adults (Bethesda, Md.:  1989), p.  67. 


      CLASSIFYING A PATIENT'S CHD
      RISK
------------------------------------------------------ Appendix IV:1.2

Under current U.S.  policy, all adults are to have their cholesterol
tested every 5 years in order to classify them by their need for
cholesterol-lowering.  Initial screening that reveals their total
blood cholesterol and HDL levels informs the preliminary risk
classification.\3 Subsequent risk classification may include more
cholesterol tests along with consideration of other equally weighted
risk factors. 

Figure IV.2, adapted from the latest NCEP guidelines, shows the
schematic that they recommend that physicians follow in classifying a
patient for treatment and advice.  Starting with the original
cholesterol screening (at the left of the figure), the physician can
arrive at one of six recommendations for advice and treatment
intensity.  For example, adults whose total cholesterol measure is
"desirable," or below 200 mg/dl, but whose HDL measures 35 mg/dl or
more should be given general advice about CHD risk.  (Approximately
51 percent of U.S.  adults fall into this category.) Adults whose
total cholesterol is between 200 mg/dl and 239 mg/dl and who have HDL
lower than 35 mg/dl or two or more risk factors (the remaining 49
percent) should be given a second test, from which LDL cholesterol is
calculated, and their risk should be reevaluated.\4


   Figure IV.2:  Recommended
   Advice and Treatment for U.S. 
   Adults by Cholesterol Level\a

   (See figure in printed
   edition.)



   (See figure in printed
   edition.)

\a Shaded areas are NCEP's alternative recommendations for advice and
treatment intensity.

Source:  Adapted from National Cholesterol Education Program, Expert
Panel on Detection, Evaluation, and Treatment of High Blood
Cholesterol in Adults, 2nd report (Bethesda, Md.:  1993). 

After determining whether or not the patient has evidence of CHD, the
physician places the patient in an LDL category that then leads to
one of the remaining recommendations, depending on the number of
other risk factors.  For example, according to figure IV.2, after
performing a clinical evaluation, a physician should recommend a
reduced-fat diet for a patient who has no evidence of CHD, LDL of at
least 130 mg/dl, and two or more other risk factors.  If dietary
therapy does not successfully lower LDL to target levels, the
physician may then consider cholesterol-lowering drugs, which are
usually taken indefinitely. 

Beyond this schematic, the guidelines advise physicians generally for
certain subgroups.  They urge more-vigorous efforts to lower the
cholesterol of CHD patients and elderly persons, in both of whom CHD
fatality rates are high.  For groups with fewer CHD fatalities, such
as premenopausal women (who are protected by higher HDL levels than
men) and men younger than 45, the guidelines urge that drugs be
delayed in favor of dietary intervention.  They leave considerable
discretion to the physician regarding how to treat several groups who
register unsatisfactory cholesterol readings and have no other risk
factor, such as men older than 45 and postmenopausal women with
borderline-high cholesterol.  These decisions hinge on the accuracy
of the risk classification laid out in figure IV.2.\5


--------------------
\3 As we noted earlier, LDL, usually the largest component of total
serum cholesterol, leads to greater risk of CHD the higher it is.  It
is therefore the main target of cholesterol reduction.  HDL is a much
smaller component of total cholesterol and is beneficial:  low
amounts are considered to increase the risk of heart disease. 

\4 In Cholesterol Measurement:  Test Accuracy and Factors That
Influence Cholesterol Levels, GAO/PEMD-95-8 (Washington, D.C.: 
1994), we explored the considerable imprecision in typical
cholesterol testing, concluding that decisions to classify patients
should be based on the average of multiple measurements, as
recommended by NCEP's guidelines. 

\5 A recent assessment of NCEP's guidelines by Grover and colleagues
found that they predicted fatal heart attacks at a rate 24-percent
better than random chance.  Their model, based on several large
databases, predicted CHD fatality better than NCEP.  That the NCEP
guidelines quantify cholesterol only, weighting all other risk
factors equally, may explain the lesser ability of the guidelines to
predict risk.  See S.  A.  Grover et al., "Identifying Adults at
Increased Risk of Coronary Disease," Journal of the American Medical
Association, 274 (1995), 10. 


      DIETARY AND DRUG TREATMENTS
------------------------------------------------------ Appendix IV:1.3

Reducing dietary fat is central to advice recommended for the general
population; physician-directed diet therapy and subsequent drug
therapy are reserved for high-risk persons.  Dietary therapy follows
one of two diets that limit total fat to 30 percent of calories.  The
Step I diet limits saturated fat to 8 to 10 percent of total
calories, cholesterol consumption to 300 mg a day.  If the Step I
diet fails to lower cholesterol to appropriate levels, the Step II
diet further restricts saturated fat to 7 percent of calories and
cholesterol to less than 200 mg a day.\6


Although NCEP's guidelines estimate that the Step I diet can reduce
serum cholesterol from 3 to 14 percent, Ramsay, Yeo, and Jackson have
concluded that cholesterol-reduction in populations who live outside
institutions averages only up to about 4 percent.\7 These authors
call for a more realistic assessment of dietary response to guide
treatment practice.  Thus, NCEP-recommended screening plus risk
assessment can lead many patients through diet therapy to drug
therapy. 

Table IV.2 lists cholesterol-lowering drugs available in the United
States and their major effects.  Gastrointestinal distress and skin
flushing are generally side effects of the older drugs, although a
few have more serious negative effects.  The newer statins generally
produce few side effects, but their long-term effects are not known. 
Occasionally, they cause mild liver toxicity and muscle pathology.



                               Table IV.2
                
                  U.S. Cholesterol-Lowering Drugs and
                              Their Effect

                                                           Lowers
                                       Raises    Lowers    triglycerid
Class                    Drug          HDL       LDL       es
-----------------------  ------------  --------  --------  -----------
Bile acid-binding resin  Cholestyrami            
                         ne

                         Colestipol              

Estrogen                 Multiple               
                         preparations

Fibrate                  Clofibrate                       

                         Gemfibrozil                      

HMG-CoA reductase        Fluvastatin             
inhibitor

                         Lovastatin              

                         Pravastatin             

                         Simavastatin            

Nicotinic acid           Water-                          
                         soluble
                         vitamin

Probucol                 Probucol      \a        
----------------------------------------------------------------------
\a Lowers HDL.  Lower amounts of HDL increase the risk of heart
disease. 

Source:  A.  Chait, "The High Risk Strategy for Adults," in B.  M. 
Rifkind (ed.), Lowering Cholesterol in High Risk Individuals and
Populations (New York:  Marcel Dekker, 1995), pp.  1-32. 


--------------------
\6 The 30-percent fat limit demanded by these two diets is close to
the diet consumed, on the average, in the United States.  Nutrition
experts have suggested that a lower limit might yield more dramatic
results. 

\7 L.  E.  Ramsay, W.  W.  Yeo, and P.  R.  Jackson, "Dietary
Reduction of Serum Cholesterol Concentration:  Time to Think Again,"
British Medical Journal, 303 (1991), 953. 


   OTHER CHOLESTEROL POLICIES
   COMPARED TO NCEP
-------------------------------------------------------- Appendix IV:2

After reviewing NCEP, we compared it to six recent cholesterol policy
statements from the United States and other nations with similarly
high CHD rates.  The four other nations share not only CHD problems
of similar magnitude but also the published clinical trials, yet
their policies differ in the breadth of screening they recommend and
in the definition of the high-risk group they would treat
intensively.\8 Several groups recommending the less-active policies
are in the United States.  Differences in their willingness to
generalize to untested groups from the results of a limited trial
database are a possible source of policy variation. 

In addition to reviewing the U.S.  and foreign policies, we reviewed
statements from two medical societies in the United States.  Although
the U.S.  and foreign policies and statements differ considerably in
their aims and contents, they can be compared in terms of the
comprehensiveness of the groups to be screened for cholesterol and
the use of cholesterol levels and other risk factors to trigger
intensive treatment.  One policy is similar to NCEP's recommendation
of universal screening and lower cholesterol levels for intensive
treatment.  Others are directed toward a less ambitious program of
screening fewer adults or propose different risk classifications. 
(See table IV.3.)




                                    Table IV.3
                     
                       Cholesterol-Lowering Policies in the
                      United States, Australia, Canada, and
                                     Europe\a



                                         Total
                     Middle              cholesterol\                   Treatmen
Policy     Young     age       Old       d             Plus others      t
---------  --------  --------  --------  ------------  ---------------  --------
Universal
--------------------------------------------------------------------------------
European   Yes       Yes       Yes       250           No other         Diet
Specialty
Associati
on 1992

                                         250           Risk greater     Diet,
                                                       than 2%          drugs

                                         300           No other         Diet,
                                                                        drugs

NCEP 1993  Yes       Yes       Yes       Up to 200     CHD patient and  Diet,
                                                       LDL 100+         drugs

                                         200-239       Non-CHD patient  Diet,
                                                       and LDL 180+     drugs

                                   240+          No other         Drugs


Selective
--------------------------------------------------------------------------------
Canadian   No        Yes\e     No        Up to 265     Male             None
Health
Examinati
on 1993

                                         265+          Male             Diet,
                                                                        drugs

American   No        Yes       No        \f            \f               \f
College
of
Physician
s 1995

     No        Yes       No        \f            \f               \f
U.S.
Preventiv
e
Services
Task
Force
1995


Unspecified
--------------------------------------------------------------------------------
Australia                                250           Cardiovascular   Drugs\g
n 1992                                                 disease

                                         290           No               Drugs\g
                                                       cardiovascular
                                                       disease

British                                  200           CHD patient      Drugs
1993

                                         250                            Drugs
--------------------------------------------------------------------------------
\a Complete facts of publication are in the bibliography.  Data were
not available for empty cells. 

\b Young = men 20-34, women 20-44; middle age = men 35-65, women
45-65; old = men and women 65 and older. 

\c These are protocols that we selected for illustration from the
many available in NCEP. 

\d Mean baseline cholesterol in mg/dl. 

\e Men 30-59. 

\f Not reported.  Several policies reported only screening or only
treatment advice. 

\g The Australian policy declines to specify "a threshold above which
drug therapy should be given and below which it should not because
the long-term net clinical benefit (total mortality and morbidity)
has not been established for these patients." The policy considers
HDL cholesterol of 58 mg/dl and higher to be protective. 

For instance, while the European Specialty Association and NCEP
policies recommend universal adult screening, two others in the
United States and one in Canada recommend screening selectively.\9
The Canadian Health Examination, the American College of Physicians,
and the U.S.  Preventive Services Task Force recommend screening only
segments of the adult population for whom clinical trials have
clearly shown that lowering cholesterol is beneficial--that is,
mainly high-risk middle-aged white men. 

The policies in table IV.3 also differ in their use of high
cholesterol and other risk factors to determine whether to apply diet
and drug treatments.  As we noted in figure IV.2, NCEP's separation
of desirable cholesterol (or up to 200 mg/dl) from borderline-high
cholesterol (200-239 mg/dl) places many persons who are at moderate
risk on a route to physician-directed treatment.  However, NCEP also
refines the cholesterol levels requiring dietary treatment or drugs,
as do several other policies:  they assign more-intensive action to
groups whose CHD rate is high and less-intensive action to groups
whose CHD rate is low.\10 This focuses the most intensive preventive
interventions on the patients who are most likely to benefit. 

Another point of comparison between the policies is their willingness
to generalize from the data on the populations studied in the
randomized clinical trials.  Most of the policies, including NCEP's,
are based not only on the clinical trials but also on observational
research that did not test an intervention or use controlled designs. 
This research included autopsy reports, cross-country comparisons,
community studies, migration studies, and the like.  Although the
clinical trials evidence is mainly restricted to high-risk
middle-aged white men, NCEP has generalized the data to try to cover
some of the gaps.\11 Two of the four policies that recommend
selective rather than universal screening are based primarily on the
clinical trials evidence. 

Several policies are less willing to generalize.  For example, the
European Specialty Association guidelines express the concern that
"no randomized controlled trials have specifically addressed
hyperlipidemia in women.  It is not known whether the results of
existing drug trials can be extrapolated to women."\12 The Canadian
Health Examination and the U.S.  Preventive Services Task Force
require at least one well-conducted clinical trial per population
group before recommending screening or intensive cholesterol-lowering
for that group. 

To sum up, policies differ widely within North America.  At one end,
NCEP applies its physician-directed high-risk options most broadly: 
several untested groups, including elderly men and women, middle-aged
men and women whose cholesterol levels are moderate, and others whose
CHD risk is moderate, may be advised to pursue intensive
physician-directed cholesterol-lowering therapy.  At the other end,
the Canadian Health Examination screens only middle-aged white men
and treats no one, other than CHD patients, whose total cholesterol
is lower than 265 mg/dl.  The U.S.  Preventive Services Task Force
screens only middle-aged persons.  Beyond North America, some
policies are similar to NCEP's in scope while others, such as the
Australian, treat CHD patients whose total cholesterol is 250 mg/dl
or higher but not others unless their cholesterol exceeds 290 mg/dl. 


--------------------
\8 In a recent comparative study of CHD rates, the United States
shared the higher rates found in Australia, Britain, and Canada
rather than the much lower rates in France and Japan.  See Office of
Technology Assessment, International Health Statistics:  What the
Numbers Mean for the United States (Washington, D.C.:  1993), p. 
118. 

\9 On the European Specialty Association, see the task force of the
European Society of Cardiology, European Atherosclerosis Society, and
European Society of Hypertension, in Atherosclerosis, 110 (1994),
121-61.  The Australian and British policies do not specify breadth
of screening.  Several policies refer only to general- population
screening and do not comment on treatment or prevention.  NCEP's
alert levels for total cholesterol are the best known, but NCEP
encourages physicians to incorporate LDL and HDL levels into their
treatment decisions.  LDL is the largest part of total cholesterol,
calculated from total cholesterol, and, within a range, predictable
from it. 

\10 NCEP recommends different intensities of action for different
risk groups, although these have not been widely publicized.  For
example, "if after an adequate trial of diet therapy, LDL cholesterol
remains >190 mg/dl in the absence of risk factors, >160 in their
presence, or >130 in subjects with established atherosclerotic
disease, drug therapy is recommended." (See A.  Chait, "The High-Risk
Strategy for Adults," in Basil M.  Rifkind (ed.), Lowering
Cholesterol in High- Risk Individuals and Populations (New York: 
Marcel Dekker, 1995), p.  4.) The level for initiating
cholesterol-lowering interventions in women otherwise free of
coronary heart disease and with fewer than two other risk factors is
270 mg/dl (Rifkind, pp.  4-31). 

\11 "[E]vidence from RCT [randomized control trials] is strongest in
middle-aged men with high initial cholesterol levels.  However, the
complete set of evidence, including the epidemiologic observations
and animal experiments, strongly supports the generalization that
reducing total and LDL cholesterol levels is also likely to reduce
CHD in younger and older men, in women and in individuals with more
moderate elevations of cholesterol." (NIH, Report of the Expert Panel
on Detection, Evaluation, and Treatment of High Blood Cholesterol in
Adults (Bethesda, Md.:  1989), p.  16.) Similarly, the updated NCEP
policy document states that "Lack of clinical trial data proving that
cholesterol-lowering therapy reduces age-adjusted mortality in
individuals with moderately high blood cholesterol and without other
CHD factors, however, does not preclude efforts to reduce cholesterol
levels in this group." (National Cholesterol Education Program,
Expert Panel on Detection, Evaluation, and Treatment of High Blood
Cholesterol in Adults, 2nd report (Bethesda, Md.:  1993), p.  I-7.)

\12 K.  Pyorala et al., "Prevention of Coronary Heart Disease in
Clinical Practice:  Recommendations of the Task Force of the European
Society of Cardiology," Atherosclerosis, 110 (1993), 151. 


EXPERTS WE CONSULTED
=========================================================== Appendix V

Alexia Antczak-Boukoms, D.Sc.
Decision Sciences and Outcomes Research
Bristol-Myers Squibb Pharmaceutical Research Institute
Wallingford, Connecticut

Trudy Bush, M.D.
School of Medicine
University of Maryland
Baltimore, Maryland

Thomas Chalmers, M.D.  (deceased)
Meta Works
Boston, Massachusetts

Charles S.  Davis, Ph.D.
Department of Preventive Medicine and Environmental Health
University of Iowa
Iowa City, Iowa

Alan M.  Garber, M.D., Ph.D.
School of Medicine
Stanford University
Stanford, California

Michael Glade, M.D.
Chicago, Illinois

David J.  Gordon, M.D., Ph.D.
Lipid Metabolism Atherogenesis Branch
National Heart, Lung, and Blood Institute
Bethesda, Maryland

Michael Gurr, Ph.D.
Maypole Scientific Services
Isles of Scilly, United Kingdom

Stephen Hulley, M.P.H., M.D.
Department of Epidemiology and Biostatistics
Prevention Sciences Group
University of California
San Francisco, California

Telba Irony, Ph.D.
Department of Operations Research
George Washington University
Washington, D.C. 

Kathleen Jablonski, Ph.D.
Medlantic Research Institute
Washington, D.C. 

Mark Lipsey, Ph.D.
Institute for Public Policy Studies
Vanderbilt University
Nashville, Tennessee

Donald Mattison, M.D.
School of Public Health
University of Pittsburgh
Pittsburgh, Pennsylvania

Robert Murray, M.D.
College of Medicine
Howard University
Washington, D.C. 

Henry S.  Sacks, M.D.
Clinical Trials Unit
Mt.  Sinai School of Medicine
New York, New York

Ernst J.  Schaefer, M.D.
Jean Mayer USDA Human Nutrition Research Center on Aging
Tufts University
Boston, Massachusetts

George Davey Smith, M.D.
Department of Clinical Epidemiology
University of Bristol
Bristol, England

Meier Stampfer, M.D.
Harvard Medical School
Boston, Massachusetts

Simon G.  Thompson, M.Sc.
Medical Statistics Unit
London School of Hygiene and Tropical Medicine
London, England

Janet Wittes, Ph.D.
Statistics Collaborative
Washington, D.C. 

Paul Wortman, Ph.D.
Department of Psychology
State University of New York at Stony Brook
Stony Brook, New York




(See figure in printed edition.)Appendix VI
COMMENTS FROM THE U.S.  DEPARTMENT
OF HEALTH AND HUMAN SERVICES
=========================================================== Appendix V



(See figure in printed edition.)

See comment 1. 

See comment 2. 

See comment 3. 

See comment 4. 



(See figure in printed edition.)



(See figure in printed edition.)


The following are GAO's comments on the HHS December 13, 1995,
letter. 

GAO COMMENTS

1.  Since the 1980s, applying meta-analysis to health and medical
research has been viewed as a research process in its own right. 
Meta-analysis offers a statistically rigorous way of pooling the
results of clinical trials, and it has the advantage of increasing
the statistical power of individual trials by combining them.  In the
absence of a constant bias pervading a whole set of trials, the bias
of any one analyst within the set is minimized. 

NIH has paid considerable attention to the use of meta-analysis in
health and medical research, as evidenced by its 1986 Workshop on
Methodological Issues in Overviews of Randomized Clinical Trials.\1
In June 1994, NHLBI sponsored a workshop specifically on
meta-analyses of cholesterol-lowering trials.  NCEP included meta
-analyses in the evidentiary sources it cited in support of the Adult
Treatment Panel II guidelines.  The federal legislation that
established the Agency for Health Care Policy Research required that
guidelines on clinical practice be based on systematic reviews of
research evidence. 

Our task for this report was to provide an overview of the clinical
trials evidence that had been used, among other evidence, to support
the NCEP guidelines.  Given that, and in view of the conflicting
results among individual trials, we believe that our decision to
systematically display and discuss the results of the meta-analyses
that synthesized results across trials on the five health outcomes we
were interested in was an appropriate and efficient way to summarize
this large body of data. 

Although the meta-analysts we studied differed in what they included
and excluded, most of them excluded outmoded treatments such as
hormone treatment in men.  We have stated in the report that clinical
trials using the more recently prescribed statin drugs are not well
represented in the meta-analyses we reviewed.  They do not include,
for example, the results of 4S or WOSCOPS.  However, they do include,
to a greater or lesser degree, all the cholesterol- lowering
randomized clinical trials that were available when the Adult
Treatment Panel II guidelines were published. 

2.  Table II.2 displays the odds ratios for CHD, non-CHD, and total
fatalities.  Across the meta-analyses that examined primary
prevention trials, we found no statistically significant differences
in CHD fatality rates between treatment and nontreatment groups. 
Further, a statistically significant increase in the non-CHD fatality
rate for treated participants was in fact obtained in 5 of the 6
meta-analyses that examined this outcome.  The failure to achieve
statistically significant reductions in fatality rates from all
causes in primary prevention trials should be viewed in the light of
Gould's 1995 meta-analysis, which shows an adverse effect for
treatment with fibrates.  This drug class predominates in primary
prevention trials using drug interventions. 

3.  Terge Pedersen and other 4S investigators have cautioned that the
study's results should not be extrapolated to studies using other
statin drugs, given differences in the cholesterol-lowering efficacy
of various drugs.  However, we note that both the 4S and WOSCOPS
trials--which showed no statistical differences between treatment and
nontreatment groups in deaths from noncardiovascular causes--support
meta- analysis findings that increased non-CHD fatality rates in
treatment groups were associated not with lowering cholesterol but
with particular cholesterol-lowering treatments.  We agree that
safety and efficacy have been demonstrated for simvastatin and
pravastatin for the types of primary and secondary patients who
participated in these trials.  It is also important to recognize that
patients may be expected to continue with drug treatment (especially
for primary prevention) for a period considerably longer than the
5-year duration of these trials, even while information is not yet
complete about the long-term safety of the statin drugs. 

4.  A consideration of the cost-effectiveness of cholesterol-
lowering treatment must also account for costs associated with
prevention.  For example, the 3.5-percent absolute CHD mortality risk
reduction in 4S means that to prevent the death of one patient
diagnosed with CHD, an estimated 29 patients would have to be treated
with simvastatin at trial dosages.  The WOSCOPS authors estimate that
1,000 middle-aged men with hypercholesterolemia and no prior evidence
of myocardial infarction would have to be treated for 5 years to
prevent 7 deaths from cardiovascular causes.  They note that such
findings compare favorably with treatment for mild hypertension in
middle-aged persons.  Nevertheless, any cost- benefit ratio will be
less in a primary prevention setting than in a secondary prevention
setting, because patients are treated whose risk levels are lower. 

5.  Clinical trials do constitute the most important type of evidence
for the benefit of a medical intervention, because less- rigorously
designed cross-country comparisons and other descriptive studies may
be confounded by many factors.  However, the largest portions of the
U.S.  population have not been well-represented in
cholesterol-lowering trials, including among others women, women and
men minorities, young men, elderly persons, and persons with
cholesterol levels lower than 250 mg/dl.  Past cholesterol-lowering
trials have usually been conducted with middle-aged, high-risk, white
men and are thus limited in how well they can comment on the benefits
of lowering cholesterol for others.  When others have been included
in trials, the data sets they provide are usually too small to
analyze, leaving in question the applicability of
cholesterol-lowering from one group to another. 

The contrast that has been best studied is that between people who
have already had a heart attack and those who have not.  Generally,
this can be understood as a contrast between very high-risk persons
and those whose risk is lower, although even the primary trials tend
to focus on individuals with multiple risks for CHD.  The
meta-analyses we reviewed consistently found a pattern of greater CHD
benefit in secondary than in primary trials. 

6.  NHLBI cites selected research other than clinical trials in order
to support the broad application of the benefits found in the trial
results with high-risk men to other groups.  Yet different treatments
are recommended for men and for women in the guidelines.  In 1993,
NHLBI revised the NCEP guidelines to "delay" the application of
cholesterol-lowering drugs to beyond age 55 for women.  Similarly,
NHLBI invited doctors to consider estrogen replacement rather than
cholesterol-lowering drugs beyond menopause.  Many of the women who
display signs of high risk for CHD are older; indeed, elderly persons
as a group are disproportionately women.  A recent review found no
direct evidence of benefit in lowering cholesterol among elderly
persons and mixed evidence that higher cholesterol levels are
associated with CHD.\2


--------------------
\1 T.  Colton, L.  S.  Freedman, and A.  L.  Johnson (eds.),
"Proceedings of the Workshop on Methodologic Issues in Overviews of
Randomized Clinical Trials, May 1986," Statistics in Medicine, 6:3
(1987), 217-410. 

\2 Walter H.  Ettinger and William R.  Hazzard, "Dyslipotroteinemia
in Older People," in Basil M.  Rifkind (ed.), Lowering Cholesterol in
High-Risk Individuals and Populations (New York:  Marcel Dekker,
1995). 


MAJOR CONTRIBUTORS TO THIS REPORT
========================================================= Appendix VII

PROGRAM EVALUATION AND METHODOLOGY
DIVISION

John E.  Oppenheim, Assistant Director
Barbara Chapman, Project Manager
Valerie J.  Caracelli, Senior Social Science Analyst
Harry Conley, Chief Statistician
Penny Pickett, Communications Analyst
Venkareddy Chennareddy, Referencer

ACKNOWLEDGMENT

Douglas M.  Sloane, Senior Social Science Analyst, General Government
Division, also contributed important statistical advice. 


BIBLIOGRAPHY
=========================================================== Appendix 0


   META-ANALYSES WE CITED
--------------------------------------------------------- Appendix 0:1

The meta-analyses preceded by an author's last name (in capital
letters) and a date--the way in which we cite them in the report--are
those whose data we used in our tables and figures. 

CTF 1993:  Canadian Task Force on the Periodic Health Examination. 
"Periodic Health Examination, 1993 Update.  2.  Lowering the Blood
Total Cholesterol Level to Prevent Coronary Heart Disease." Canadian
Medical Association Journal, 148:4 (1993), 521-38. 

CUCHERAT 1993:  Cucherat, M., and J.-P.  Boissel.  "Meta-Analysis of
Results from Clinical Trials on Prevention of Coronary Heart Disease
by Lipid-lowering Interventions." Clinical Trials and Meta-Analyses,
28 (1993), 109-29. 

GORDON 1995:  Gordon, D.J.  "Cholesterol Lowering and Total
Mortality." In B.M.  Rifkind (ed.).  Lowering Cholesterol in
High-Risk Individuals and Populations.  New York:  Marcel Dekker,
1995.  Pp.  33-48. 

GOULD 1995:  Gould, A.L., et al.  "Cholesterol Reduction Yields
Clinical Benefit:  A New Look at Old Data." Circulation, 91:8 (1995),
2274-82. 

Holme, I.  "An Analysis of Randomized Trials Evaluating the Effect of
Cholesterol Reduction on Total Mortality and Coronary Heart Disease
Incidence." Circulation, 82:6 (1990), 1916-24. 

Holme, I.  "Meta-Analysis of Cholesterol Reduction Trials:  Coronary
Disease and Mortality." Primary Cardiology, 18:7 (1992), 63-70. 

HOLME 1993:  Holme, I.  "Relation of Coronary Heart Disease Incidence
and Total Mortality to Plasma Cholesterol Reduction in Randomized
Trials:  Use of Meta-Analysis." British Heart Journal, 69 (1993),
S42-S47. 

LAW I 1994:  Law, M.R., S.G.  Thompson, and N.J.  Wald.  "Assessing
Possible Hazards of Reducing Serum Cholesterol." British Medical
Journal, 308 (1994), 373-79. 

LAW II 1994:  Law, M.R., N.J.  Wald, and S.G.  Thompson.  "By How
Much and How Quickly Does Reduction in Serum Cholesterol
Concentration Lower Risk of Ischaemic Heart Disease?" British Medical
Journal, 308 (1994), 367-73. 

MULDOON 1990:  Muldoon, M.F., S.B.  Manuck, and K.A.  Mathews. 
"Lowering Cholesterol Concentrations and Mortality:  A Quantitative
Review of Primary Prevention Trials." British Medical Journal, 301
(1990), 309-14. 

MULDOON 1993:  Muldoon, M., et al.  "Low or Lowered Cholesterol and
Risk of Death from Suicide and Trauma." Metabolism, 42:9, Supp.  1
(1993), 45-56. 

RAVNSKOV 1992:  Ravnskov, U.  "Cholesterol Lowering Trials in
Coronary Heart Disease:  Frequency of Citation and Outcome." British
Medical Journal, 305 (1992), 15-19. 

ROSSOUW 1991:  Rossouw, J.E.  "Clinical Trials of Lipid Lowering
Drugs." In B.M.  Rifkind (ed.).  Drug Treatment of Hyperlipidemia. 
New York:  Marcel Dekker, 1991.  Pp.  67-88. 

Rossouw, J.E.  "The Effects of Lowering Serum Cholesterol on Coronary
Heart Disease Risk." Lipid Disorders, 78:1 (1994), 181-95. 

ROSSOUW 1995:  Rossouw, J.E.  "Secondary Prevention of Coronary Heart
Disease." In B.M.  Rifkind (ed.).  Lowering Cholesterol in High-risk
Individuals and Populations.  New York:  Marcel Dekker, 1995.  Pp. 
49-67. 

Rossouw, J.E., B.  Lewis, and B.M.  Rifkind.  "The Value of Lowering
Cholesterol After Myocardial Infarction." New England Journal of
Medicine, 323:16 (1990), 1112-19. 

SILBERBERG 1991:  Silberberg, J.S., and D.A.  Henry.  "The Benefits
of Reducing Cholesterol Levels:  The Need to Distinguish Primary from
Secondary Prevention.  1.  A Meta-Analysis of Cholesterol Lowering
Trials." Medical Journal of Australia, 155 (1991), 665-70. 

Smith, G.D., and J.  Pekkanen.  "Should There Be a Moratorium on the
Use of Cholesterol Lowering Drugs?" British Medical Journal, 304
(1992), 431-34. 

SMITH 1993:  Smith, G.D., F.  Song, and T.A.  Sheldon.  "Cholesterol
Lowering and Mortality:  The Importance of Considering Initial Level
of Risk." British Medical Journal, 306 (1993), 1367-73. 

Thompson, S.G.  "Controversies in Meta-Analysis:  The Case of the
Trials of Serum Cholesterol Reduction." Statistical Methods in
Medical Research, 2 (1993), 173-92. 

YUSUF 1987:  Yusuf, S., and J.  Cutler.  "Single Factor Trials:  Drug
Studies," and S.  Yusuf and C.D.  Furberg, "Single Factor Trials: 
Control Through Life Style Changes." In A.G.  Olsson et al.  (eds.). 
Atherosclerosis.  New York:  Churchill Livingstone, 1987.  Pp. 
389-97. 


   CLINICAL TRIALS CITED IN THE
   META-ANALYSES
--------------------------------------------------------- Appendix 0:2

The headings below are the names we have given in our text and tables
to the clinical trials cited by the meta-analyses in our study.  The
entries within each heading are the data sources for that clinical
trial.\1


--------------------
\1 Two other data sources that some meta-analyses cite but that were
not identified with any one trial are F.  Ederer et al., "Cancer
Among Men on Cholesterol Lowering Diets:  Experience from Five
Clinical Trials," Lancet, 2 (1971), 203-06, and D.  K.  Wysowski and
T.  P.  Gross, "Deaths Due to Accidents and Violence in Two Recent
Trials of Cholesterol-lowing Drugs," Archives of Internal Medicine,
150 (1990), 2169-72. 


      ACHESON
------------------------------------------------------- Appendix 0:2.1

Acheson, J., and E.C.  Hutchinson.  "Controlled Trial of Clofibrate
in Cerebral Vascular Disease." Atherosclerosis, 15 (1972), 177-83. 


      ANCILLARY HELSINKI
------------------------------------------------------- Appendix 0:2.2

Frick, M.H., et al.  "Efficacy of Gemfibrozil in Dyslipidaemic
Subjects with Suspected Heart Disease:  An Ancillary Study in the
Helsinki Heart Study Frame Population." Annals of Medicine, 25
(1993), 41-45. 


      BEGG
------------------------------------------------------- Appendix 0:2.3

Begg, T.B., and B.M.  Rifkind.  "Valutazione Della Terapia con
Clofibrate Nelle Arteriopatie Periferiche." Minerva Medica, 62
(1971), 3469-75. 


      CDP
------------------------------------------------------- Appendix 0:2.4

Coronary Drug Project Research Group.  "The Coronary Drug Project: 
Initial Findings Leading to Modification of Its Research Protocol."
Journal of the American Medical Association, 214:7 (1970), 1303-13. 

Coronary Drug Project Research Group.  "The Coronary Drug Project: 
Findings Leading to Further Modifications of Its Protocol with
Respect to Dextrothyroxine." Journal of the American Medical
Association, 220:7 (1972), 996-1008. 

Coronary Drug Project Research Group.  "The Coronary Drug Project: 
Findings Leading to the Discontinuation of the 2.5 mg/day Estrogen
Group." Journal of the American Medical Association, 226:6 (1973),
652-57. 

Coronary Drug Project Research Group.  "The Coronary Drug Project: 
Clofibrate and Niacin in Coronary Heart Disease." Journal of the
American Medical Association, 231:4 (1975), 360-81. 

Canner, P.L., et al.  "Fifteen Year Mortality in Coronary Drug
Project Patients:  Long-term Benefit with Niacin." Journal of the
American College of Cardiology, 8 (1986), 1245-55. 


      CHICAGO
------------------------------------------------------- Appendix 0:2.5

Stamler, J., et al.  "Effectiveness of Estrogens for Therapy of
Myocardial Infarction in Middle-age Men." Journal of the American
Medical Association, 183:8 (1963), 632-38. 


      CLAS
------------------------------------------------------- Appendix 0:2.6

Blankenhorn, D.H., et al.  "Beneficial Effects of Combined
Colestipol-Niacin Therapy on Coronary Atherosclerosis and Coronary
Venous Bypass Grafts." Journal of the American Medical Association,
257:23 (1987), 3233-40. 


      CORN OIL
------------------------------------------------------- Appendix 0:2.7

Rose, G.A., W.  B.  Thomson, and R.T.  Williams.  "Corn Oil Treatment
of Ischaemic Heart Disease." British Medical Journal, 1 (1965),
1531-33. 


      DART
------------------------------------------------------- Appendix 0:2.8

Burr, M.L., et al.  "Effects of Changes in Fat, Fish, and Fibre
Intakes on Death and Myocardial Reinfarction:  Diet and Reinfarction
Trial (DART)." Lancet, 2 (1989), 757-61. 


      ESTROGEN
------------------------------------------------------- Appendix 0:2.9

Oliver, M.F., and G.S.  Boyd.  "Influence of Reduction of Serum
Lipids on Prognosis of Coronary Heart-Disease:  A Five-Year Study
Using Oestrogen." Lancet, 2 (1961), 499-505. 


      EXCEL
------------------------------------------------------ Appendix 0:2.10

Bradford, R.H., et al.  "Expanded Clinical Evaluation of Lovastatin
(EXCEL) Study:  Design and Patient Characteristics of a Double-Blind,
Placebo-Controlled Study in Patients with Moderate
Hypercholesterolemia." American Journal of Cardiology, 66 (1990),
44b-55b. 

Bradford, R.H., et al.  "Expanded Clinical Evaluation of Lovastatin
(EXCEL) Study Results.  I.  Efficacy in Modifying Plasma Lipoproteins
and Adverse Event Profile in 8245 Patients with Moderate
Hypercholesterolemia." Archives of Internal Medicine, 151 (1991),
43-49. 

Tobert, J.A.  "The Cholesterol Controversy." Correspondence.  British
Medical Journal, 304 (1992), 713. 


      FATS
------------------------------------------------------ Appendix 0:2.11

Brown, G., et al.  "Regression of Coronary Artery Disease as a Result
of Intensive Lipid-Lowering Therapy in Men with High Levels of
Apolipoprotein B." New England Journal of Medicine, 323:19 (1990),
1289-98. 


      FINNISH
------------------------------------------------------ Appendix 0:2.12

Miettinen, T.A., et al.  "Multifactorial Primary Prevention of
Cardiovascular Diseases in Middle-Aged Men:  Risk Factor Changes,
Incidence, and Mortality." Journal of the American Medical
Association, 254:15 (1985), 2097-2102. 


      GOTHENBURG
------------------------------------------------------ Appendix 0:2.13

Wilhelmsen, L., et al.  "The Multifactor Primary Prevention Trial in
Gï¿½teborg, Sweden." European Heart Journal, 7 (1986), 279-88. 


      HELSINKI
------------------------------------------------------ Appendix 0:2.14

Frick, M.H., et al.  "Helsinki Heart Study:  Primary Prevention Trial
with Gemfibrozil in Middle-Aged Men with Dyslipidemia:  Safety of
Treatment, Changes in Risk Factors, and Incidence of Coronary Heart
Disease." The New England Journal of Medicine, 317:20 (1987),
1237-45. 

Manninen, V.  "Lipid Alterations and Decline in the Incidence of
Coronary Heart Disease in the Helsinki Heart Study." Journal of the
American Medical Association, 260:5 (1988), 641-51. 


      INDIA
------------------------------------------------------ Appendix 0:2.15

Singh, R.B., et al.  "Randomized Controlled Trial of Cardioprotective
Diet in Patients with Recent Acute Myocardial Infarction:  Results of
One Year Follow Up." British Medical Journal, 304 (1992), 1015-19. 


      LA VA
------------------------------------------------------ Appendix 0:2.16

Dayton, S., et al.  "A Controlled Clinical Trial of a Diet High in
Unsaturated Fat in Preventing Complications of Atherosclerosis."
Circulation, 40:1, Supp.  2 (1969), 1-63. 

Pearce, M.L., and S.  Dayton.  "Incidence of Cancer in Men on a Diet
High in Polyunsaturated Fat." Lancet, (1971), 464-67. 


      LIFESTYLE
------------------------------------------------------ Appendix 0:2.17

Ornish, D., et al.  "Can Lifestyle Changes Reverse Coronary Heart
Disease?  The Lifestyle Heart Trial." Lancet, 336 (1990), 129-33. 


      LONG-TERM ESTROGEN
------------------------------------------------------ Appendix 0:2.18

Marmorston, J., et al.  "Clinical Studies of Long-term Estrogen
Therapy in Men with Myocardial Infarction.  (27531)." Proceedings of
the Society of Experimental Biology and Medicine, 110 (1962), 400-8. 


      LRC CPPT
------------------------------------------------------ Appendix 0:2.19

Lipid Research Clinics Investigators.  "The Lipid Research Clinics
Coronary Primary Prevention Trial:  Results of 6 Years of Post-Trial
Follow-up." Archives of Internal Medicine, 152 (1992), 1399-410. 

Lipid Research Clinics Program.  "The Lipid Research Clinics Coronary
Primary Prevention Trial Results.  I.  Reduction in Incidence of
Coronary Heart Disease." Journal of the American Medical Association,
251:3 (1984a), 351-64. 

Lipid Research Clinics Program.  "The Lipid Research Clinics Coronary
Primary Prevention Trial Results.  II.  The Relationship of Reduction
in Incidence of Coronary Heart Disease to Cholesterol Lowering."
Journal of the American Medical Association, 251:3 (1984b), 365-74. 


      MARS
------------------------------------------------------ Appendix 0:2.20

Blankenhorn, D.H., et al.  "Coronary Angiographic Changes with
Lovastatin Therapy:  The Monitored Atherosclerosis Regression Study
(MARS)." Annals of Internal Medicine, 119 (1993), 969-76. 


      MINNESOTA
------------------------------------------------------ Appendix 0:2.21

Brewer, E.R., P.L.  Ashman, and K.  Kuba.  "The Minnesota Coronary
Survey:  Composition of the Diets, Adherence, and Serum Lipid
Response." Abstract.  Circulation, 51, Supp.  2 (1975), II-2. 

Frantz, I.D., et al.  "The Minnesota Coronary Survey:  Effect of Diet
on Cardiovascular Events and Deaths." Abstract.  Circulation, 51 and
52, Supp.  2 (1975), II-4. 

Frantz, I.D., et al.  "Test of Effect of Lipid Lowering by Diet on
Cardiovascular Risk:  The Minnesota Coronary Survey."
Arteriosclerosis, 9 (1989), 129-35. 


      MRC LOW FAT
------------------------------------------------------ Appendix 0:2.22

Research Committee.  "Low-fat Diet in Myocardial Infarction:  A
Controlled Trial." Lancet, 2 (1965), 501-4. 


      MRC SOYA
------------------------------------------------------ Appendix 0:2.23

Medical Research Council, Research Committee.  "Controlled Trial of
Soya-bean Oil in Myocardial Infarction." Lancet, 2 (1968), 693-700. 

Heady, J.A.  "Are PUFA Harmful?" Correspondence.  British Medical
Journal, 1 (1974) 115-16. 


      MRFIT
------------------------------------------------------ Appendix 0:2.24

Multiple Risk Factor Intervention Trial Research Group.  "Multiple
Risk Factor Intervention Trial:  Risk Factor Changes and Mortality
Results." Journal of the American Medical Association, 248:12 (1982),
1465-77. 


      NEWCASTLE
------------------------------------------------------ Appendix 0:2.25

Group of Physicians of the Newcastle upon Tyne Region.  "Trial of
Clofibrate in the Treatment of Ischaemic Heart Disease." British
Medical Journal, 4 (1971), 767-75. 


      NHLBI
------------------------------------------------------ Appendix 0:2.26

Brensike, J.F., et al.  "Effects of Therapy with Cholestyramine on
Progression of Coronary Arteriosclerosis:  Results of the NHLBI Type
II Coronary Intervention Study." Circulation, 69:2 (1984), 313-24. 


      OSLO DA
------------------------------------------------------ Appendix 0:2.27

Hjermann, I., et al.  "Effect of Diet and Smoking Intervention on the
Incidence of Coronary Heart Disease:  Report from the Oslo Study
Group of a Randomized Trial in Healthy Men." Lancet, 2 (1981),
1303-10. 


      OSLO DH
------------------------------------------------------ Appendix 0:2.28

Leren, P.  "The Effect of Plasma Cholesterol Lowering Diets in Male
Survivors of Myocardial Infarction:  A Controlled Clinical Trial."
Acta Medica Scandinavica, 466 (1966), 1-92. 

Leren, P.  "The Oslo Diet-Heart Study:  Eleven-year Report."
Circulation, 42 (1970), 935-42. 


      POSCH
------------------------------------------------------ Appendix 0:2.29

Buchwald, H., et al.  "Effect of Partial Ileal Bypass Surgery on
Mortality and Morbidity from Coronary Heart Disease in Patients with
Hypercholesterolemia:  Report of the Program on the Surgical Control
of Hyperlipidemias (POSCH)." New England Journal of Medicine, 323
(1990), 946-55. 


      RESTENOSIS
------------------------------------------------------ Appendix 0:2.30

Fail, P.S., et al.  "The Long-term Clinical Efficacy of Lovastatin
Therapy Following Successful Coronary Angioplasty." Abstract. 
Clinical Research, 40:2 (1992), 400A. 

Sahni, R.S., et al.  "Prevention of Restenosis by Lovastatin after
Successful Coronary Angioplasty." American Heart Journal, 121 (1991),
1600-8. 


      RETINOPATHY
------------------------------------------------------ Appendix 0:2.31

Harrold, B.P., V.J.  Marmion, and K.R.  Gough.  "A Double-Blind
Controlled Trial of Clofibrate in the Treatment of Diabetic
Retinopathy." Diabetes, 18:5 (1969), 285-91. 


      ST.  VINCENTS
------------------------------------------------------ Appendix 0:2.32

Gross, L., and R.  Figueredo.  "Long-Term Cholesterol-Lowering Effect
of Colestipol Resin in Humans." Journal of the American Geriatrics
Society, 21:12 (1973), 552-56. 


      SCOR
------------------------------------------------------ Appendix 0:2.33

Kane, J.P., et al.  "Regression of Coronary Atherosclerosis During
Treatment of Familial Hypercholesterolemia with Combined Drug
Regimens." Journal of the American Medical Association, 264:23
(1990), 3007-12. 


      SCOTTISH
------------------------------------------------------ Appendix 0:2.34

Research Committee of the Scottish Society of Physicians.  "Ischaemic
Heart Disease:  A Secondary Prevention Trial Using Clofibrate."
British Medical Journal, 4 (1971), 775-84. 

Dewar, H.A., and M.F.  Oliver, "Trial of Clofibrate." Correspondence. 
British Medical Journal, 1 (1972), 506. 


      STARS
------------------------------------------------------ Appendix 0:2.35

Watts, G.F., et al.  "Effects on Coronary Artery Disease of
Lipid-Lowering Diet, or Diet Plus Cholestyramine, in the St.  Thomas'
Atherosclerosis Regression Study (STARS)." Lancet, 339 (1992),
563-69. 


      STOCKHOLM
------------------------------------------------------ Appendix 0:2.36

Carlson, L.A., et al.  "Reduction of Myocardial Reinfarction by the
Combined Treatment with Clofibrate and Nicotinic Acid."
Atherosclerosis, 28 (1977), 81-86. 

Carlson, L.A., and G.  Rosenhamer.  "Reduction of Mortality in the
Stockholm Ischaemic Heart Disease Secondary Prevention Study by
Combined Treatment with Clofibrate and Nicotinic Acid." Acta Medica
Scandinavica, 223 (1988), 405-18. 

Rosenhamer, G., and L.A.  Carlson.  "Effect of Combined Clofibrate-
Nicotinic Acid Treatment in Ischaemic Heart Disease:  An Interim
Report." Atherosclerosis, 37 (1980), 129-38. 


      SYDNEY
------------------------------------------------------ Appendix 0:2.37

Woodhill, J.M., et al.  "Low Fat, Low Cholesterol Diet in Secondary
Prevention of Coronary Heart Disease." Advances in Experimental
Medicine and Biology, 109 (1978), 317-30. 


      UPJOHN
------------------------------------------------------ Appendix 0:2.38

Dorr, A., et al.  "Colestipol Hydrochloride in Hypercholesterolemic
Patients--Effect on Serum Cholesterol and Mortality." Journal of
Chronic Diseases, 31 (1978), 5-14. 


      VETERANS CARDIOLOGY
------------------------------------------------------ Appendix 0:2.39

Schoch, H.K.  "The U.S.  Veterans Administration Cardiology
Drug-Lipid Study:  An Interim Report." Advances in Experimental
Medicine and Biology, 4 (1968), 405-20. 

Detre, K.M., and L.  Shaw.  "Long-Term Changes of Serum Cholesterol
with Cholesterol-Altering Drugs in Patients with Coronary Heart
Disease:  Veterans Administration Drug-Lipid Cooperative Study."
Circulation, 50 (1974), 998-1005. 


      VETERANS W.  ROXBURY
------------------------------------------------------ Appendix 0:2.40

McCaughan, D.  "The Long-Term Effects of Probucol on Serum Lipid
Levels." Archives of Internal Medicine, 141 (1981), 1428-32. 


      WHO
------------------------------------------------------ Appendix 0:2.41

Committee of Principal Investigators.  "A Co-operative Trial in the
Primary Prevention of Ischaemic Heart Disease Using Clofibrate."
British Heart Journal, 40 (1978), 1069-118. 

Committee of Principal Investigators.  "W.H.O.  Cooperative Trial on
Primary Prevention of Ischaemic Heart Disease Using Clofibrate to
Lower Serum Cholesterol:  Mortality Follow-up." Lancet, 2 (1980),
379-85. 

Committee of Principal Investigators.  "WHO Cooperative Trial on
Primary Prevention of Ischaemic Heart Disease with Clofibrate to
Lower Serum Cholesterol:  Final Mortality Follow-up." Lancet, 3
(1984), 600-4. 

Heady, J.A., J.N.  Morris, and M.F.  Oliver.  "WHO
Clofibrate/Cholesterol Trial:  Clarifications." Lancet, 340 (1992),
1405-6. 


      WHO F
------------------------------------------------------ Appendix 0:2.42

World Health Organization European Collaborative Group. 
"Multifactorial Trial in the Prevention of Coronary Heart Disease. 
3.  Incidence and Mortality Results." European Heart Journal, 4
(1983), 141-47. 

World Health Organization European Collaborative Group.  "European
Collaborative Trial of Multifactorial Prevention of Coronary Heart
Disease:  Final Report on the 6-year Results." Lancet, 2 (1986),
869-72. 


   META-ANALYSES WE OMITTED
--------------------------------------------------------- Appendix 0:3

We omitted the meta-analyses listed in this section for one or more
of the following reasons.  The appropriate reasons are indicated by
the numbers in parentheses at the ends of the entries.  (1) The data
and procedures were insufficiently reported.  (2) The meta-analysis
was not concerned with the clinical outcomes we studied.  (3) The
meta-analysis was not based primarily on more than two randomized
trials of a year or more in duration.  (4) The meta-analysts did not
mathematically cumulate results across studies.  (5) The
meta-analysis was not published in English.  (6) The meta-analysis
was not published in any language.  (7) The meta-analysis was not
concerned with cholesterol-lowering treatment. 

Antman, E.M., et al.  "A Comparison of Results of Meta-Analyses of
Randomized Control Trials and Recommendations of Clinical Experts."
Journal of the American Medical Association, 268:2 (July 1992),
240-48.  (1)

Atkins, D., et al.  "Cholesterol Reduction and the Risk for Stroke in
Men:  A Meta-Analysis of Randomized, Controlled Trials." Annals of
Internal Medicine, 119:2 (1993), 136-45.  (2)

Brown, B.G., et al.  "Lipid Lowering and Plaque Regression:  New
Insights into Prevention of Plaque Disruption and Clinical Events in
Coronary Disease." Circulation, 87:6 (June 1993), 1781-91.  (3, 4)

Criqui, M.H.  "Cholesterol, Primary and Secondary Prevention, and
All-Cause Mortality." Annals of Internal Medicine, 115:12 (1991),
973-76.  (3)

Criqui, M.H., et al.  "Plasma Triglyceride Level and Mortality from
Coronary Heart Disease." New England Journal of Medicine, 328:17
(April 1993), 1220-25.  (4)

Dallongeville, J., et al.  "Modulation of Plasma Triglyceride Levels
by APOE Phenotype:  A Meta-Analysis." Journal of Lipid Research, 33
(1992), 447-54.  (3)

Dattilo, A.M., and P.M.  Kris-Etherton.  "Effects of Weight Reduction
on Blood Lipids and Lipoproteins:  A Meta-Analysis." American Journal
of Clinical Nutrition, 56 (1992), 320-28.  (2, 3)

Ederer, F., et al.  "Cancer Among Men on Cholesterol-Lowering Diets: 
Experience from Five Clinical Trials." Lancet, 2 (1971), 203-6.  (2)

Faergeman, O.  "Total Mortality in Cardiovascular Risk Factor
Intervention Trials." Scandinavian Journal of Clinical Laboratory
Investigation, 50, Supp.  199 (1990), 7-13.  (4)

Gordon, D.J., et al.  "High-Density Lipoprotein Cholesterol and
Cardiovascular Disease:  Four Prospective American Studies."
Circulation, 79:1 (January 1989), 8-15.  (3)

Lechleitner, M., J.R.  Patsch, and H.  Braunsteiner.  "Risikofaktoren
fï¿½r die Koronare Herzerkrankung." Wiener Klinische Wochenschrift,
103:17 (1991), 513-18.  (5)

MacMahon, S.  "Lowering Cholesterol:  Effects on Trauma Death, Cancer
Death and Total Mortality." Australia and New Zealand Journal of
Medicine, 22 (1992), 580-82.  (1)

Malenka, D.J., and J.A.  Baron.  "Cholesterol and Coronary Heart
Disease:  The Attributable Risk Reduction of Diet and Drugs."
Archives of Internal Medicine, 149 (September 1989), 1981-85.  (4)

Mann, J.I., and J.W.  Marr.  "Coronary Heart Disease Prevention: 
Trials of Diets to Control Hyperlipidaemia." In N.E.  Miller and B. 
Lewis (eds.).  Lipoproteins, Atherosclerosis and Coronary Heart
Disease.  Amsterdam:  Elsevier/North-Holland Biomedical Press, 1981. 
(3, 4, 6)

Manson, J.E., et al.  "The Primary Prevention of Myocardial
Infarction." New England Journal of Medicine, 326:21 (May 1992),
1406-16.  (1)

Mellies, M.J., et al.  "Pravastatin Experience in Elderly and
Non-elderly Patients." Atherosclerosis, 101 (1993), 97-110.  (3)

Mensink, R.P., and M.B.  Katan.  "Effect of Dietary Fatty Acids on
Serum Lipids and Lipoproteins:  A Meta-Analysis of 27 Trials."
Arteriosclerosis and Thrombosis, 12:8 (August 1992), 911-19.  (2)

Oliver, M.F.  "Coronary Heart Disease Prevention:  Trials Using Drugs
to Control Hyperlipidaemia." In N.E.  Miller and B.  Lewis (eds.). 
Lipoproteins, Atherosclerosis and Coronary Heart Disease.  Amsterdam: 
Elsevier/North-Holland Biomedical Press, 1981.  (4)

Peto, R.  Hearings, tapes 1-4.  Watergate Hotel, Washington, D.C.,
March 11, 1989.  (6)

Ramsay, L.E., et al.  "Dietary Reduction of Serum Cholesterol
Concentration:  Time to Think Again." British Medical Journal, 303
(October 1991), 953-57.  (1)

Ramsay, L.E., et al.  "Dietary Treatment of Hyperlipidaemia." British
Medical Journal, 308:6933 (1994), 917-18.  (2)

Ripsin, C.M., et al.  "Oat Products and Lipid Lowering:  A
Meta-Analysis." Journal of the American Medical Association, 267:24
(1992), 3317-25.  (7)

Rossouw, J.E.  "Cholesterol and Heart Disease in Older Persons and in
Women," abstract.  National Heart, Lung, and Blood Institute,
National Institutes of Health, Bethesda, Maryland, June 18-19, 1990. 
(6)

Rossouw, J.E.  "Non-CHD Mortality in Cholesterol Lowering Trials."
National Institutes of Health, Bethesda, Maryland.  (6)

Rossouw, J.E.  "Stabilization and Regression of Coronary
Atherosclerosis," abstract.  National Heart, Lung, and Blood
Institute, National Institutes of Health, Bethesda, Maryland.  (6)

Rossouw, J.E., and S.B.  Hulley.  "Deaths from Injury, Violence, and
Suicide in Secondary Prevention Trials of Cholesterol Lowering." New
England Journal of Medicine, 325:25 (1991), 1813.  (1)

Sarkkinen, E.S., et al.  "Long-term Effects of Three Fat-Modified
Diets in Hypercholesterolemic Subjects." Atherosclerosis, 105 (1994),
9-23.  (3)

Sleight, P.  "Cholesterol and Coronary Heart Disease Mortality."
Australia and New Zealand Journal of Medicine, 22 (1992), 576-79. 
(1, 4)

Stampfer, M.J., and G.A.  Colditz.  "Estrogen Replacement Therapy and
Coronary Heart Disease:  A Quantitative Assessment of the
Epidemiologic Evidence." Preventive Medicine, 20 (1991), 47-63.  (3)

Yusuf, S., J.  Wittes, and L.  Friedman.  "Overview of Results of
Randomized Clinical Trials in Heart Disease.  II.  Unstable Angina,
Heart Failure, Primary Prevention with Aspirin, and Risk Factor
Modification." Journal of the American Medical Association, 260:15
(1988), 2259-63.  (1)

*** End of document. ***