herbicides in Vietnam, while none was observed amongchildren of control veterans [Wolfe et al., 1995]. Further-more, when the CDC birth defects study was reanalyzedusing the exposure opportunity index based upon interviewdata, the risk of spina bi®da was signi®cantly associatedwith the highest estimated level of Agent Orange exposure[Erickson et al., 1984]. Based on these data and others, anInstitute of Medicine panel suggested an associationbetween herbicide exposure in Vietnam and an increasedrisk of spina bi®da in children [IOM, 1996].Although results of studies of male Vietnam veteransare potentially useful in assessing health consequences of Vietnam service for women, a further study of gender-speci®c health outcomes for women was desired. Maternalexposures to toxicants are more directly associated withadverse pregnancy outcomes, while evidence for the effectsof paternal exposure on pregnancy outcomes is very limitedand indirect [Of®ce of Technology Assessment, 1985;Moore and Persaud, 1998].Many potential risk factors for abnormal reproductiveoutcomes existed in Vietnam for women veterans, includingpsychological stresses of war, various infections, substanceabuse, and Agent Orange contaminated with dioxin [Bakeret al., 1989; Bates et al., 1990; Jordan et al., 1991]. Otherpotential risk factors associated with military hospitalnursing conditions in Vietnam included physical stress,and exposure to waste anesthetic gases and ethyleneoxide[Custis, 1990; Rowland et al., 1996; Biovin, 1997].We conducted a historical cohort study of 8,280 womenveterans, comparing the pregnancy outcomes of 4,140women Vietnam veterans with those of 4,140 non-Vietnamveterans in the military during the same era. Due todif®culty in identifying a large number of women exposedto any particular aspect of Vietnam service with any degreeof certainty, the study was a ``Vietnam experience'' type of study rather than of speci®c exposures.
MATERIAL AND METHODSStudy Subjects
The U.S. Army and Joint Services EnvironmentalSupport Group compiled manually a list of Army womenveterans who served in Vietnam. The Air Force provided acomputer listing of all Air Force women known to haveserved in Vietnam, as did the Navy and Marine Corps. Atotal of 5,230 names were compiled as potential Vietnamveteran study subjects. Military personnel records of thesewomen were retrieved from various locations and theirmilitary service data were extracted. Women Vietnamveterans for this study were de®ned as those women in theU.S. Military whose permanent tour of duty includedservice in Vietnam during the period from July 4, 1965through March 28, 1973, a period of signi®cant U.S.military involvement in Vietnam. After the record review,4,643 women met the eligibility criteria; of these 4,390women were found alive as of January 1, 1992.Women non-Vietnam veterans for the study werede®ned as those women assigned to a military unit in theU.S. during the Vietnam War and whose tour of duty did notinclude service in Vietnam. Potential control subjects fromeach branch of service were selected using the sameprocedures as for Vietnam veterans. A pool of 6,657 womenwere eligible for controls; of these 4,390 women wererandomly selected among the living members of this pool asof January 1, 1992 [SAS, 1990].To determine the feasibility of the present study, weconducted a pilot study on 500 of these women (250Vietnam veterans and 250 non-Vietnam controls), leaving8,280 women (4,140 Vietnam and 4,140 non-Vietnam)available for the study.
Survey Methods
Instrument
A structured health questionnaire was administeredusing a computer-assisted telephone interview softwarepackage to obtain information concerning demographicbackground, general health, lifestyle, menstrual history,pregnancy history, pregnancy outcomes, military experienceincluding nursing occupation and combat exposure. Foreach pregnancy, information such as smoking, drinking,complications, infections, medications, exposure to x-ray,occupational history, exposure to anesthetic gases, ethyle-neoxide, herbicides and pesticides was collected.
Location Strategies
Names and social security numbers of the 8,280veterans were ®rst passed through the Internal RevenueService records and VA's Bene®ciary Identi®cation RecordsLocator Subsystem. Those not located were also searchedin the Defense Manpower Data Center records. Severalnationwide commercial databases such as Telematch, TRW/ Experian, and Equifax proprietary databases, and nationaltelephone directory CD-ROMs were searched for newaddresses and telephone numbers. State nursing boardsand nursing associations were also contacted for newaddresses because most women veterans serving duringthe Vietnam era were nurses.
Outcome Assessment
To satisfy the basic statistical requirement of indepen-dence of observations, i.e., one pregnancy per woman, anindex pregnancy was identi®ed for each woman. For theVietnam veterans, it was de®ned as the ®rst pregnancy after
448 Kang et al.