Gulf War Veterans: Limitations of Available Data for Accurately
Determining the Incidence of Tumors (Testimony, 05/14/98,
GAO/T-NSIAD-98-186).

GAO discussed the results of its recent work on the: (1) reliability and
utility of existing data for determining the incidence of tumors among
Gulf War veterans; and (2) uses of the data by the Department of Defense
(DOD) and the Department of Veterans Affairs (VA) to monitor tumors in
the Gulf War veteran population.

GAO noted that: (1) based on GAO's completed study, none of the data
sources providing information on the health of Gulf War veterans can be
used to reliably estimate the incidence of tumors or other illnesses;
(2) existing government data systems are generally limited by poor
coverage of the Gulf War veteran population and problems of accuracy and
completeness; (3) as a result, it is not known how many Gulf War
veterans have tumors or whether they have a higher incidence of them
than other veterans; (4) DOD and VA have begun efforts to improve these
data systems but have not developed the capability to effectively link
information from different sources to assess tumors or other illnesses
among Gulf War veterans; (5) DOD and VA also recently funded a few
research studies that should provide additional information on tumor
cases in the future; (6) however, these studies are not the product of a
systematic effort to study the incidence of tumors, and limitations to
the studies will prevent them from providing reliable and valid
estimates of Gulf War veterans' tumors; (7) following completion of
GAO's report, it conducted additional work to assess the feasibility of
linking data from different sources and determining the number of tumors
among Gulf War veterans; (8) based on GAO's preliminary analysis of only
VA health and disability benefits databases, it identified about 14,700
Gulf War veterans with tumors; (9) although GAO has not conducted
appropriate analyses to determine whether the total it computed
translates into a higher or lower rate of cases as compared to other
groups of veterans, it is a larger number of cases than that reported in
other studies of Gulf War veterans; (10) these data reflect cases that
are reported in VA data systems and do not include Gulf War veterans who
are still on active duty or those who use non-VA medical facilities; and
(11) while GAO was able to merge VA data at an aggregate level to
estimate the total number of tumors, detailed analysis involving the
combination of cases by type of tumor will be more difficult to conduct
because of different diagnostic coding systems used in the different
data systems.

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

 REPORTNUM:  T-NSIAD-98-186
     TITLE:  Gulf War Veterans: Limitations of Available Data for 
             Accurately Determining the Incidence of Tumors
      DATE:  05/14/98
   SUBJECT:  Cancer
             Military personnel
             Disease detection or diagnosis
             Health surveys
             Statistical data
             Medical records
             Data collection
             Medical information systems
             Data integrity
IDENTIFIER:  Persian Gulf War
             VA Beneficiary Identification and Records Locator System
             VA Persian Gulf War Health Registry
             DOD Comprehensive Clinical Evaluation Program
             NCI Surveillance, Epidemiology, and End Results Program
             Gulf War Syndrome
             VA Patient Treatment File
             
******************************************************************
** 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


Before the Subcommittee on Human Resources, Committee on Government
Reform and Oversight, House of Representatives

For Release on Delivery
Expected at
10:00 a.m., EDT
Thursday,
May 14, 1998

GULF WAR VETERANS - LIMITATIONS OF
AVAILABLE DATA FOR ACCURATELY
DETERMINING THE INCIDENCE OF
TUMORS

Statement of Kwai-Cheung Chan, Director of Special Studies and
Evaluations, National Security and International Affairs Division

GAO/T-NSIAD-98-186

GAO/NSIAD-98-186T

Gulf War Veterans

(713027)


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

  BIRLS - Beneficiary Identification and Records Locator Subsystem
  DOD - Department of Defense
  SEER - Surveillance, Epidemiology, and End Results
  VA - Veterans Affairs

============================================================ Chapter 0

Mr.  Chairman and Members of the Subcommittee: 

I am pleased to be here today to discuss the results of our recently
completed report, at the Subcommittee's request, on the (1)
reliability and utility of existing data for determining the
incidence of tumors among Gulf War veterans and (2) uses of the data
by the Departments of Defense (DOD) and Veterans Affairs (VA) to
monitor tumors in the Gulf War veteran population.\1 Our review
focused on five types of data sources:  mortality data, DOD and VA
hospital and outpatient medical records, DOD and VA Gulf War health
registries, survey results, and national and state cancer registries. 
It is important to point out that most of the data sources we
reviewed were not designed for medical research purposes but rather
for other uses, such as the administration of records or veteran
outreach.  Nevertheless, researchers have used some of these data
sources to assess the nature and extent of Gulf War illnesses.  Our
intent was to see if the data could be used to determine the
frequency of tumors and to examine the appropriateness of using the
data in other applications. 

I will first summarize our study results.  Then, I will present
observations on subsequent data analysis, requested by the
Subcommittee, to assess the numbers and types of tumors that are
reported for Gulf War veterans in available VA health and benefits
disability databases.  Following this, I will provide more details on
our findings. 


--------------------
\1 Gulf War Veterans:  Incidence of Tumors Cannot Be Reliably
Determined From Available Data (GAO/NSIAD-98-89, Mar.  3, 1998). 


   RESULTS IN BRIEF
---------------------------------------------------------- Chapter 0:1

Based on our completed study, we found that none of the data sources
providing information on the health of Gulf War veterans can be used
to reliably estimate the incidence of tumors or other illnesses. 
Existing government data systems are generally limited by poor
coverage of the Gulf War veteran population and problems of accuracy
and completeness.  As a result, it is not known how many Gulf War
veterans have tumors or whether they have a higher incidence of them
than other veterans.  DOD and VA have begun efforts to improve these
data systems but have not developed the capability to effectively
link information from different sources to assess tumors or other
illnesses among Gulf War veterans.  DOD and VA also recently funded a
few research studies that should provide additional information on
tumor cases in the future.  However, these studies are not the
product of a systematic effort to study the incidence of tumors, and
limitations to the studies will prevent them from providing reliable
and valid estimates of Gulf War veterans' tumors. 

Following completion of our report, we conducted additional work to
assess the feasibility of linking data from different sources and
determining the number of tumors among Gulf War veterans.  Based on
our analysis of only VA health and disability benefits databases, we
identified about 14,700 Gulf War veterans with tumors.  Although we
did not conduct appropriate analyses to determine whether the total
we computed translates into a higher or lower rate of cases as
compared to other groups of veterans, it is a larger number of cases
than that reported in other studies of Gulf War veterans.  These data
reflect cases that are reported in VA data systems and do not include
Gulf War veterans who are still on active duty or those who use
non-VA medical facilities.  While we were able to merge VA data at an
aggregate level to estimate the total number of tumors, detailed
analysis involving the combination of cases by type of tumor will be
more difficult to conduct because of different diagnostic coding
systems used in the different data systems. 


   GAO'S REVIEW OF EXISTING DATA
   SOURCES TO DETERMINE INCIDENCE
   OF TUMORS AMONG GULF WAR
   VETERANS
---------------------------------------------------------- Chapter 0:2


      BACKGROUND
-------------------------------------------------------- Chapter 0:2.1

Although casualties were relatively light during the Gulf War
deployment, thousands of veterans have come forward complaining of
various illnesses, including cancer, in the years following the
conflict.  During the Gulf War, American troops may have been exposed
to several potentially hazardous substances.  These include chemical
warfare agents, depleted uranium from munitions, smoke from oil-well
fires, infectious diseases, pesticides, petroleum fuels, and
vaccines.  Some of these substances have previously been associated
with different types of cancer through animal laboratory studies and
other epidemiological research investigations.  For example,
combustion products from petroleum include polyaromatic hydrocarbons,
benzene, and carbon disulfide, some of which are known to cause lung
cancer when inhaled.  Exposure to certain pesticides has also been
linked to lymphatic and lung cancers.  In addition, exposure to
radioactive particles has been tied to higher rates of respiratory
and other cancers.  Information on exposures that took place during
the Gulf War, however, has been either incomplete or nonexistent due
to the lack of record-keeping and measurement before, during, and
after the deployment of troops; loss of key records; poor recall by
veterans; and other factors. 

The development of cancer is usually characterized by a latency
period of many years from initial exposure to a harmful agent to a
definitive medical diagnosis.  Depending on the nature and extent of
the exposure, type of cancer, and characteristics of different
individuals, the latency period may be as long as 30 years or more. 
The most common types of cancers have a latency period of 15 years or
more, but in certain situations cancer can develop more quickly
(e.g., in cases where the immune system is compromised). 

Given that there is a lengthy latency period for most tumors, it may
be too soon to detect any increase in tumors among Gulf War veterans. 
Also, since cancer is a relatively rare event,\2 large population
groups may need to be observed over several years to assess incidence
and determine whether it has changed over time.  Furthermore, without
credible exposure information, it is hard to form specific hypotheses
about what kinds of tumors might grow in what individuals.  Although
such constraints exist, it is nonetheless important to begin
monitoring Gulf War veterans to assess whether they are suffering
from an increase in tumors so that appropriate health care and
treatment can be provided where needed.  With many types of tumors,
early detection is important to more effective treatment outcomes. 


--------------------
\2 The average age-adjusted incidence rate for all types of cancer
combined was 400 cases per 100,000 population in the United States in
1990-91 (National Cancer Institute, Cancer Rates and Risks, 1996,
p.  17).  Among the age group that served in the Gulf War, incidence
is lower (fewer than 100 cases per 100,000 population for the ages
15-44). 


      MORTALITY DATA
-------------------------------------------------------- Chapter 0:2.2

One source for estimating the incidence of cancer among Gulf War
veterans uses mortality as an indicator.  The VA maintains a large
administrative database, the Beneficiary Identification and Records
Locator Subsystem (BIRLS), which can be used to track the mortality
of veterans.  The system's strength is its broad coverage of the
veteran population (estimated to be 80 to 90 percent of the deceased
veteran population, according to VA and other researchers) and its
cost-effectiveness for ascertaining vital status and causes of death. 
However, mortality is only a proxy measure for incidence because not
all cancers result in death, and those that do may take several years
to show up.  Mortality data provide good estimates of incidence for
cancers that have a high mortality rate (such as lung and liver
cancers), but they are less useful for cancers with lower rates of
mortality (such as prostate and breast cancers).  Because of these
limitations, mortality data will systematically underreport overall
cancer incidence. 

In one published study using data from BIRLS, the VA assessed whether
mortality from a range of diseases (including cancer) was different
for Gulf War veterans compared with a sample of veterans who were not
deployed to the Persian Gulf.\3 The study, which covered deaths
occurring in a 2-year period after the war, found that there was a
small but significant excess of deaths among Gulf War veterans
compared with nondeployed veterans and that the excess was due mainly
to accidents and not disease.  Of the 1,765 Gulf War veterans who
died during the study period, 119 died from cancer, showing no
statistically significant difference compared with the cancer death
rate among nondeployed veterans.  One key limitation of this portion
of the study is the relatively short time period for assessing
veterans that died from cancer. 

Another limitation of the study, which the authors acknowledged, is
whether the study's comparison groups were appropriately matched. 
Military personnel who were ill or recovering from an illness would
not have been deployed to the Gulf War area.  However, these
personnel were included in the comparison group of nondeployed
veterans.  This meant that the comparison group may have been less
healthy than the deployed veterans group.  The extent to which a
higher rate of prior illnesses among nondeployed veterans resulted in
a different rate of mortality (or cancer mortality in particular) and
thus biased the study findings is unknown. 

VA is now updating the study, extending the period to be studied
through 1995, and the results should be published later this year. 


--------------------
\3 Kang, H.  K., and Bullman, T.  A., "Mortality Among U.S.  Veterans
of the Persian Gulf War," New England Journal of Medicine, vol.  335
(1996), pp.  1498-1504. 


      DOD AND VA MEDICAL RECORDS
-------------------------------------------------------- Chapter 0:2.3

Both DOD and VA maintain an automated database containing detailed
medical and demographic information on patients discharged from their
respective hospitals.  Both data systems contain millions of records,
but they do not represent the entire active duty and veteran
population.  DOD's data include most hospitalizations of active duty
personnel, in large part because DOD medical care is free and readily
available to active duty personnel; however, some veterans' groups
have reported that Gulf War veterans are seeking medical care outside
DOD.  VA also has an extensive network of medical centers across the
country, but the overwhelming majority of veterans who have separated
from the military use non-VA hospitals and medical facilities.  A
survey conducted by VA in the late 1980s, for example, estimated that
only about 20 percent of veterans had ever used a VA hospital. 

Another weakness of these data systems has been the lack of coverage
of outpatient medical care.  Coverage of outpatient care is important
because more patients have been diagnosed and treated for many types
of tumors in outpatient facilities in recent years.  DOD currently
has no centralized reporting system for its outpatient facilities,
although an automated system is under development.  In October 1996,
VA established an automated system that includes diagnostic
information, but consistent and reliable outpatient reporting may not
be available for several years. 

In addition to limitations in terms of population coverage, there are
also issues regarding the accuracy and completeness of
hospitalization data reporting.  Miscoding of diagnoses has been
problematic in the past, as shown by VA researchers in previous
assessments of certain types of cancer among Vietnam veterans.  For
example, in one case-control study of over 400 Vietnam veterans
identified in VA's hospitalization database as having malignant
tumors of connective and other soft tissue, close to 40 percent of
the records were found to be miscoded or misclassified when hospital
pathology reports were subsequently collected and independently
reviewed by an expert pathologist.\4

In a large DOD-funded study published in 1997, the hospitalization
experiences of all active duty Gulf War veterans during the period
1991-93 were compared with a sample of other active duty military
personnel who were not deployed to the Gulf region.\5 Overall, the
authors found there was no excess in hospitalizations among Gulf War
veterans compared with other military personnel.  However, Gulf War
veterans had higher rates of hospitalizations in certain years for
mental disorders, diseases of the blood, and diseases of the
genitourinary system.  Hospitalization rates for tumor cases were
also higher for Gulf War veterans, but the differences were not
statistically significant and most involved benign conditions. 

A major strength of this study is its large size and statistical
power to detect differences in rates of hospitalizations between
deployed and nondeployed military personnel.  However, a key
limitation of the study, which influences the interpretation of the
results, is that it excluded hospitalizations of Gulf War veterans
who separated from the services and hospitalizations of active duty
personnel who used non-DOD hospitals.  Another important limitation
of the study is that the time frame was far too short for detecting
any diseases resulting from possible exposures during the war, such
as tumors, which have lengthy latency periods. 

The same researchers have underway a related follow-up study to
examine hospitalizations of Gulf War veterans in military and
nonmilitary hospitals in California.  Although the study results
cannot be generalized to the entire Gulf War population, it is large
and one of the first to systematically combine military and
nonmilitary hospitalizations.  The study period is longer (1991-95);
however, the problem of detecting diseases with a lengthy latency
period is still an issue, and outpatient data will be excluded. 


--------------------
\4 Kang, H.K.  et al, "Soft Tissue Sarcomas and Military Service in
Vietnam:  A Case Comparison Group Analysis of Hospital Patients,"
Journal of Occupational Medicine, vol.  28 (1986), pp.  1215-1218. 

\5 Gray, G.C.  et al, "The Postwar Hospitalization Experience of U.S. 
Veterans of the Persian Gulf War," New England Journal of Medicine,
vol.  335 (1996) pp.  1505-13. 


      GULF WAR HEALTH REGISTRIES
-------------------------------------------------------- Chapter 0:2.4

Both DOD and VA have established separate programs that provide
medical examinations and diagnostic services, free of charge, to Gulf
War veterans.  The VA began its Persian Gulf Health Registry
Examination Program in 1992, and DOD started its Comprehensive
Clinical Evaluation Program in 1994.  An existing health problem is
not necessary for participation in the programs; any Gulf War veteran
with health questions or concerns is eligible to enroll on a
voluntary basis.  Close to 100,000 veterans have completed either the
VA or DOD registry examination.  Of that number, less than 1 percent
of veterans have received a primary diagnosis of a malignant or
benign tumor. 

The suitability of the registries for assessing cancer incidence is
extremely limited.  As designed, the registries are not intended to
be used to determine the frequency and causes of illnesses among the
general Gulf War veteran population, but rather to diagnose and treat
voluntary participants' symptoms.  Because the participants were not
selected based on a random sample, there is no way to know whether
their health problems are similar to those of the general population
of Gulf War veterans.  In addition, because there is no ready
comparison or control group for the registry participants, the
significance of the data reported cannot be determined.  A further
limitation of the registry data is that they capture information
about the health of veterans only at one point in time.  Thus, if a
veteran develops cancer or another illness later on, the registry
data will not reflect this. 

Data quality concerns also have been raised in a previous review of
the VA registry by the Institute of Medicine.  The Institute found,
for example, that there was a considerable delay between the
collection of the examination data and their entry into the registry
database.\6 We also found that VA medical facilities have not
reported registry examination information consistently.  It appears
that a large number of case records submitted for input into the
registry database have been returned to the medical facilities due to
coding errors.  At the same time, effective quality assurance
procedures have not been in place to ensure that rejected records are
corrected and reentered into the database.  Thus, data coverage even
for those who participate in the registries may be incomplete. 


--------------------
\6 Institute of Medicine, Health Consequences of Service During the
Persian Gulf War:  Initial Findings and Recommendations for Immediate
Action, 1995. 


      SURVEY DATA
-------------------------------------------------------- Chapter 0:2.5

Another data approach involves developing information about the
incidence of tumors using survey methods such as a questionnaire
administered to a sample of veterans.  Significant advantages to
using the survey approach include the ability to draw a random sample
of Gulf War veterans and an appropriate comparison group.  A survey
also permits researchers to gather other information, such as
information about exposures and family history, that might shed light
on the causes of a disease.  Limitations of this approach include the
possibility of response bias (individuals who complete the survey not
being representative of the sample as a whole) and the subjectivity
of self-assessments.  The extent to which response bias is a factor,
however, can be estimated through a special survey of nonrespondents,
which may be conducted by telephone or personal interviews.  The
results of the nonrespondent survey can then be compared to the
results of the principal survey to gauge the degree to which
respondents are typical of the overall sample.  Subjectivity of the
assessments of cancer or other illnesses can also be gauged to a
degree through an independent medical review of a subsample of
respondents.  A further concern in implementing large population
surveys is that they tend to be much more costly than the other
approaches.  In addition, the type and number of questions must be
restricted, or the response rate will be low. 

VA is currently using a survey approach to study the general health
status of Gulf War veterans.  The National Health Survey of Persian
Gulf War Era Veterans was mailed to a random sample of 15,000 Gulf
War and 15,000 nondeployed veterans.  The questionnaire includes a
checklist of illnesses, including skin cancer and "any other cancer,"
and a checklist of symptoms such as "coughing" and "skin rashes." In
addition to questions about current health status, respondents were
also asked to report about their exposure to a list of agents,
including nerve gas, depleted uranium, and smoke from oil-well fires,
while they were in the Gulf War region.  The overall response rate to
the survey has been relatively low (57 percent).  VA is conducting a
survey of nonrespondents in order to evaluate nonresponse bias.  VA
is also addressing the limitation imposed by subjective assessments
through an independent review of medical records and the
"comprehensive physical examination" of a subsample of 2,000
respondents (1,000 in each of the Gulf War veteran and nondeployed
veterans groups).  The sample size of the VA survey, however, may be
too small to identify an elevated incidence of most cancers.  VA has
acknowledged this possible limitation. 

A population-based survey to assess the prevalence of self-reported
symptoms and illnesses among Gulf War veterans was also conducted in
Iowa from September 1995 through May 1996.\7 By telephone, a random
sample of about 3,700 Gulf War and non-Gulf War veterans from Iowa
were surveyed.  Overall, the study found that Gulf War veterans
reported a significantly higher prevalence of a wide range of medical
and psychiatric conditions compared with military personnel who were
not deployed to the Gulf War.  The primary conditions on which
differences were reported included depression, posttraumatic stress
syndrome, chronic fatigue, cognitive dysfunction, and respiratory
diseases.  The rate of cancer reported among these Gulf War veterans
was generally low (an estimated rate of about 1 per 100 subjects),
but it was slightly higher than that of the comparison group. 


--------------------
\7 The Iowa Persian Gulf Study Group, "Self-Reported Illness and
Health Status Among Gulf War Veterans," Journal of the American
Medical Association, 277 (1997), pp.  238-245. 


      CANCER REGISTRIES
-------------------------------------------------------- Chapter 0:2.6

Another source for estimating the incidence of cancer is
population-based cancer registries that compile standardized reports
of cancer cases directly from medical facilities (typically
hospitals) on an ongoing basis.  A national cancer registry (the
Surveillance, Epidemiology, and End Results (SEER)) established by
the National Cancer Institute reports incidence rates for the general
population and key subgroups but cannot be used to identify the Gulf
War population.  Many available state registries could be used to
identify Gulf War veterans, but the registries vary in terms of data
quality and reporting consistency and coverage.  For example, many
states require only hospitals to report on cancers and do not capture
cases diagnosed by private physicians, laboratories, and health
maintenance organizations. 

The VA has provided initial funding for a study to assess cancer
incidence among Gulf War veterans in New England, based on cases
reported in the state cancer registries in the region.  Cases will be
identified by matching the registries against a roster of all
veterans who were deployed to the Gulf War conflict.  The first phase
of the study has been funded to develop a framework for merging data
from the individual state registries.  The next phase of the study,
to begin by 1999, will involve an assessment of cancer incidence and
mortality.  Although this study is several years away from
completion, it should provide a useful means for obtaining
information about cancer incidence.  Some of the strengths of the
study are that it will use existing data systems, identify and assess
a large cohort of Gulf War veterans, and can be readily updated over
time.  One key limitation of the study, however, is that the results
cannot be generalized to the entire Gulf War population, since only
the New England states will be included.  Also, there is likely to be
an underreporting of cases in the state registries, particularly
cases diagnosed outside of the hospital setting and cases from border
areas that may be reported in other state registries outside the New
England area. 


      CONCLUSIONS AND
      RECOMMENDATION
-------------------------------------------------------- Chapter 0:2.7

No direct link has been established between potential exposures that
occurred during the Gulf War and the development of tumors among
veterans.  Yet, concerns have been raised because many of the
exposure agents in question have previously been associated with
certain cancers.  This has led to interest in determining whether the
cancer incidence rate among Gulf War veterans is higher than the
rates within other appropriate comparison groups.  If there is a
higher rate that indicates an emerging health problem, then outreach
efforts could be conducted to target appropriate diagnosis and
treatment to those potentially at risk.  The existing data and
research applications we reviewed, however, provide very limited
information about the incidence of tumors or other illnesses. 

To more effectively evaluate the incidence of tumors and other Gulf
War illnesses over time, we recommended in our report that the
Secretaries of Defense and Veterans Affairs improve existing
monitoring capabilities.  Attention should be directed toward
strengthening the utility of existing data systems and particularly
in developing cost-effective ways to make data systems more
compatible with one another so that information from different
sources can be linked.  In addition, steps should be taken to address
the data quality concerns we identified.  While we believe such
improvements can lead to more effective monitoring capabilities, the
existing data systems are likely to be insufficient to answer the
question about cancer incidence or other illnesses among Gulf War
veterans.  Therefore, further research efforts will be needed to
supplement the available data systems.  For example, little is known
about the health status of veterans who receive medical care from
sources other than DOD and VA facilities.  Practical approaches
should be developed to determine whether health problems among these
veterans may be emerging. 

In response to our report, DOD and VA concurred with our overall
findings regarding the inadequacies of existing data systems for
assessing the incidence of tumors among Gulf War veterans and our
recommendation to improve monitoring capabilities.  They emphasized,
however, that they have several initiatives underway to strengthen
the reporting of health information and the linkage of data from
different sources.  While we recognize that these efforts will lead
to some improvements, we are concerned that the available data
sources will continue to be insufficient to assess Gulf War illnesses
such as tumors. 


   GAO'S ANALYSIS OF INFORMATION
   IN VA DATABASES ON TUMORS AMONG
   GULF WAR VETERANS
---------------------------------------------------------- Chapter 0:3

Following completion of our report, we conducted further work, at the
Subcommittee's request, to determine the nature and extent of tumors
among Gulf War veterans that are included in DOD and VA data systems
and demonstrate the feasibility of linking data from different
sources.  As a first step, we analyzed data from VA's disability
benefits and health care reporting systems.  Specifically, we
requested information on Gulf War veterans who from July 31, 1991 to
March 31, 1998, had filed a disability claim based on the presence of
one or more malignant or benign tumors or were diagnosed with a tumor
at a VA hospital or outpatient medical facility.  We obtained cases
from five databases maintained by VA:  (1) Persian Gulf Health
Registry, (2) Patient Treatment File, (3) National Patient Care
Database (outpatient), (4) Compensation and Pension Masterfile, and
(5) Beneficiary Identification and Records Locator Subsystem.\8 As
shown in table 1 in appendix I, these databases represent different
subpopulations of Gulf War veterans, periods of coverage, and levels
of diagnostic information.  For example, the Compensation and Pension
Masterfile includes up to 6 different diagnoses for an individual,
whereas, the Outpatient database records as many as 15 diagnoses. 

The data from these reporting systems do not cover Gulf War veterans
who remain on active military duty or who have separated from the
services and receive non-VA disability benefits or health care.  As a
result, a significant portion of the Gulf War population will not be
included in these data.\9 In addition, VA omitted a number of codes
for tumors, including those for leukemia, Hodgkin's disease,
non-Hodgkin's lymphoma, and benign tumors of the gynecological system
and mammary glands, from the disability data files they provided us. 
Consequently, there is some degree of underreporting in the VA data. 
Nonetheless, these VA data have not been previously combined to show
a larger picture of health conditions, such as tumors, among Gulf War
veterans. 

We completed analyses of the VA data at an aggregate level, the
results of which are presented in tables 2-5 (see appendix I).  A
more detailed analysis, which involves combining cases with similar
tumor types, would be more difficult to complete because different
diagnostic coding systems are used for the health and disability
benefits data.\10

VA has not developed a linkage between these different coding
systems. 

As shown in table 2, a total of 16,226 cases of tumors were reported
for Gulf War veterans across all five databases; the vast majority of
the tumors were benign.  When we removed duplicate cases by matching
social security numbers, the total is reduced to 14,676 veterans with
tumors.  The fact that a relatively small amount of records (about 10
percent) are duplicated reinforces the need to use multiple data
sources to obtain greater coverage of the veteran population.  Use of
only selective data may lead to an undercount of cases and incomplete
information about the health condition of veterans. 

Although we do not know whether the total we computed is higher or
lower than other appropriate comparison groups, it is a larger number
of cases than that reported in other studies of Gulf War veterans. 
In May 1996, for example, VA provided some information, at the
request of the Subcommittee, on the number and types of tumors among
Gulf War veterans.  In estimating the total number of veterans with
tumors, VA used information from three databases:  Persian Gulf
Health Registry, Patient Treatment File, and Compensation and Pension
Masterfile.  The total number of tumor cases reported by VA was 6,397
as compared to the 14,676 cases we identified.  Our total is much
higher because of 2 additional years of data and the inclusion of the
BIRLS and outpatient data. 

As shown in table 3, a large majority of the Gulf War veterans with
tumors also tend to have other diagnosed illnesses as well.  This is
in line with other published studies of Gulf War illnesses that have
noted multiple symptoms and illnesses among the Gulf War veteran
population.  Further details on the different types of tumors
reported from each database are also presented in tables 4-5.  Here,
it is clear that a broad range of different tumors has been reported
for these veterans. 

Mr.  Chairman, this concludes our prepared remarks.  We will be happy
to answer any questions you or other members of the Subcommittee may
have on our completed and ongoing work. 


--------------------
\8 The BIRLS data cover denied and inactive disability claims.  The
data we obtained from VA, however, do not indicate the reasons claims
were denied (e.g., due to lack of service connection, insufficient
medical support).  As a result, these cases should be viewed as
potential tumor cases. 

\9 According to VA, of the 697,000 servicemembers who participated in
the Gulf War conflict, about 568,000 separated from the services,
while 128,000 remain on active duty.  The majority of veterans
separated from the services do not use VA hospitals and medical
facilities. 

\10 The International Classification of Diseases, Version 9 (ICD-9),
which is used in VA's Persian Gulf Health Registry, Patient Treatment
File, and Outpatient data systems, is ordered by groupings of
diseases; VA's disability coding, which is used in the Compensation
and Pension and BIRLS data systems, is structured generally by the
part of the body affected by the disability or disease. 


=========================================================== Appendix I



                                     Table 1
                     
                     Characteristics of VA Databases Used to
                     Assess Information on Tumors Among Gulf
                                   War Veterans

                                        Relevant time       Diagnostic
VA databases        Description         period covered      information
------------------  ------------------  ------------------  --------------------
Persian Gulf        Gulf War veterans   Late 1992 to        Up to 10 diagnoses,
Health Registry     completing          present             coded by ICD-9\a
(PGR)               voluntary medical
                    examinations

Patient Treatment   Veterans receiving  Gulf War to         Up to 10 diagnoses,
File (PTF)          inpatient           present             coded by ICD-9\b
                    treatment at VA
                    hospitals

Outpatient          Veterans receiving  1996 to present     Up to 15 diagnoses,
Database            outpatient                              coded by ICD-9\\\
                    treatment at VA
                    medical facilities

Compensation and    Veterans with 1 or  Gulf War to         Up to 6 diagnoses,
Pension (C&P)       more service-       present             coded by VA
Masterfile          connected                               disability
                    disabilities\c                          categories

Beneficiary         Veterans denied     Gulf War to         Up to 9 diagnoses,
Identification and  and/or no longer    present             coded by VA
Records Locator     receiving                               disability
Subsystem (BIRLS)   disability                              categories
                    compensation\d
--------------------------------------------------------------------------------
\a International Classification of Diseases, Version 9.  Up to 1994,
only three diagnoses were recorded. 

\b Individuals may have multiple records in the PTF and outpatient
files as a result of multiple visits to a VA medical facility.  Thus,
the number of diagnoses reported for an individual may be larger when
multiple records are merged. 

\c Includes veterans whose illnesses are service-connected but who
are receiving no compensation as well as veterans receiving pensions. 
Does not include pending claims. 

\d For veterans whose claims were denied, the database has no
information to indicate reasons for denial (e.g., due to lack of
service connection, insufficient medical documentation).  Diagnostic
information reflects what veterans submitted on claim applications. 
Therefore, these should be viewed as potential cases of veterans with
tumors. 



                                Table 2
                
                Number of Gulf War Veterans with Benign
                 and Malignant Tumors, by VA Database\

                                                                 Total
                                            Benign  Malign    veterans
                                            tumors     ant        with
Database                                        \a  tumors      tumors
------------------------------------------  ------  ------  ----------
PGR                                            837     280       1,117
PTF                                            681     606       1,287
Outpatient                                   2,696   1,902       4,598
C&P                                          4,938     709       5,647
BIRLS                                        3,179     398       3,577
======================================================================
Total tumors                                12,331   3,895      16,226
======================================================================
Total number of unique cases\b              11,550   3,126      14,676
----------------------------------------------------------------------
Note:  A number of codes for tumors, including those for leukemia,
Hodgkin's disease, non-Hodgkin's lymphoma, and benign tumors of the
gynecological system and mammary glands, were omitted by VA when it
created the C&P and BIRLS databases for our analysis. 

\a Veterans with one or more tumors diagnosed as malignant within all
possible tumor diagnoses were placed in the "malignant" category. 
Veterans with only benign tumors were categorized as "benign."

\b Given that the BIRLS data may reflect potential rather than
diagnosed cases of tumors, there is a total number of 11,333 unique
cases when the other 4 databases are combined and the BIRLS records
are excluded from the analysis (8,502 benign and 2,831 malignant). 



                                     Table 3
                     
                     Number of Gulf War Veterans with Tumors
                               and Other Diagnoses

                                    Tumor cases
                ----------------------------------------------------
                                               Multiple     Multiple       Total
                 One tumor,    One tumor     tumors, no  tumors plus    veterans
                   no other   plus other          other        other        with
Database          diagnoses    diagnoses      diagnoses    diagnoses      tumors
--------------  -----------  -----------  -------------  -----------  ----------
PGR                     260          830             10           13       1,113
PTF                     233          775             20          259       1,287
Outpatient            1,671        2,016            212          697       4,596
C&P                     136        5,276              7          196       5,615
BIRLS                   464        2,982             20          110       3,576
--------------------------------------------------------------------------------
Note:  Table does not reflect 39 missing cases of unspecified tumors. 
As a result, the numbers in the column for total veterans with tumors
will differ somewhat from the numbers in the same column in table 2. 



                                     Table 4
                     
                      Number of Tumors in Gulf War Veterans,
                      by VA Health Administration Databases

                                                                 Outpatien
Diagnostic code             Type of tumor                   PTF          t   PGR
                            Malignant
140 -149                    Lip, oral cavity and             91        206     8
                             pharynx
150 -159                    Digestive organs and            341        814    17
                             peritoneum
160 -165                    Respiratory and                 357        821    13
                             intrathoracic organs
170 -176                    Bone, connective tissue,        323      1,531   112
                             skin and breast
179 -189                    Genitourinary organs            389      1,159    45
190 -199                    Other unspecified sites       2,645      1,170    36
200 -208                    Lymphatic and hematopoietic     702      1,848    58
                             tissue
                            Benign
210                         Lip, oral cavity and             23        122    11
                             pharynx
211                         Other parts of digestive        137        753    56
                             system
212                         Respiratory and                  31         60     5
                             intrathoracic organs
213                         Bones and articular              32         42    22
                             cartilage
214                         Lipoma                          176        901   296
215                         Other connective and other       31        128    26
                             soft tissues
216                         Skin                             68      1,439   206
217                         Breast                           16         57     3
218                         Uterine leiomyoma               136        342    30
219                         Other, uterus                    10         15
220                         Ovary                             8          5     1
221                         Other female genital organs       3         15
222                         Male genital organs               2         69     9
223                         Kidney or other urinary           3         20     3
                             organs
224                         Eye                               2         63     1
225                         Brain and other parts of         30         94     6
                             nervous system
226                         Thyroid gland                    15         23     2
227                         Other endocrine glands and       40         87     4
                             related structures
228                         Hemangioma and                   43         80    38
                             lymphangioma, any site
229                         Other and unspecified sites       9        107    13
                            Carcinoma in situ
230                         Digestive organs                  5         28
231                         Respiratory system                2         24     1
232                         Skin                              1         40     3
233                         Breast and genitourinary         19         91     4
                             system
234                         Other and unspecified sites       1         28     1
                            Neoplasms of uncertain
                             behavior
235                         Digestive and respiratory        33         35     2
                             organs
236                         Genitourinary                    13         57    10
237                         Endocrine glands and             33         76    21
                             nervous system
238                         Other and unspecified sites      63        432    32
                             and tissues
239                         Unspecified nature               46        387    46
================================================================================
Total                                                     5,879     13,199  1,14
                                                                               1

                            Number of other non-            444        459   269
                             tumorous diseases
                            Number of other non-          8,621      8,677  1,74
                             tumorous diagnoses                                5
--------------------------------------------------------------------------------


                                     Table 5
                     
                         Frequency of Tumors in Gulf War
                            Veterans, by VA's Benefits
                             Administration Databases

                                                                      BIRL
Diagnostic code                    Type of tumor                         S   C&P
5012                               Bones, new growth, malignant         30    34
5015                               Bones, new growth, benign           229   657
5327                               Muscle, new growth, malignant -       4     4
                                    excludes soft tissue sarcoma
5328                               Muscle, new growth, benign           26    45
5329                               Soft tissue sarcoma                   1    12
6014                               Eyeball, new growth, malignant        2     3
6015                               Eyeball, new growth, benign          43    82
6208                               Ear, new growth, malignant            1
6209                               Ear, new growth, benign              12    46
6819                               Respiratory system, new growth,      57    85
                                    malignant
6820                               Respiratory system, new growth,     145   237
                                    benign
7343                               Digestive system, any specified      67    57
                                    part, new growth, malignant
7344                               Digestive system, any specified      80   187
                                    part, new growth, benign
7528                               Genitourinary system, new growth,    51   148
                                    malignant
7529                               Genitourinary system, new growth,   194   297
                                    benign
7627                               Gynecological system, new growth,    34    65
                                    malignant
7628                               Gynecological system, new growth,     9    18
                                    benign
7818                               Skin, new growth, malignant         156   248
7819                               Skin, new growth, benign           2,74  3,74
                                                                         0     2
7914                               Endocrine system, new growth,        11    52
                                    malignant
7915                               Endocrine system, new growth,        20    60
                                    benign
8002                               Brain, new growth, malignant         36    53
8003                               Brain, new growth, benign            20    75
8021                               Spinal cord, new growth malignant     7     9
8022                               Spinal cord, new growth, benign       3    20
================================================================================
Total                                                                 3,97  6,23
                                                                         8     6

                                   Number of other non-tumorous        448   530
                                    diseases
                                   Number of other non-tumorous       12,6  22,5
                                    diagnoses                           20    76
--------------------------------------------------------------------------------

*** End of document. ***