Item ID Number °1599

Author

Corporate Author

Roport/Artldo HUB Supporting Statement with Attachments

Journal/Book Title

Year 000°

Month/Day

Color ' '

Number of Images ^

DOSGriptOn NotOS 'tem includes: Supporting Statement which discussesaspects of the Ranch Hand II study, Letter fromThomas B. Bruton, Judge Advocate General dated 22October 1984 regarding confidentiality issues with theRanch Hand II study, Fact Sheet sent to Ranch Hand Iparticipants, and pages 17-55 of unknown sourcediscussing the Ranch Hand II study

Wednesday, May 23, 2001 Page 1600 of 1608

SUPPORTING STATEMENT

A. Justification:

1. In 1981, the Assistant to the President for Domestic Affairs and Policy,directed the A1r Force to undertake the Ranch Hand epidemiology study, as rec-ommended by the Interagency Work Group to Study the Possible Long-Term HealthEffects of Phenoxy Herbicides and Contaminants. The Initial portion of thisongoing study was begun 1n October 1981. The follow-up phase of this studywill begin in 1985.

This series of questionnaires will be used as a part of the follow-upphases of the Ranch Hand study. A contractor (National Opinion Research Center[NORC]) experienced 1n Interviewing techniques will conduct the questioning ofstudy participants. The data derived from the questionnaire will be utilizedby Science Applications International Corporation (SAIC) and the USAF School ofAerospace Medicine to statistically assess the existence and/or the developmentof various disease states 1n present and former A1r Force personnel who par-ticipated in herbicide spraying (Operation Ranch Hand) during the years 1962 to1971. A control group of nonexposed individuals will be additionally ques-tioned. No previous data exists which can be used to assess the current healthstatus of these groups.

2. Four data collection instruments will be used, all nearly identical topreviously approved questionnaires (OMB 10701-0033). They will be administeredby personnel of NORC, Chicago, Illinois. Two hundred participants who arenewly entered into this ongoing study will be given an in-home interview,lasting approximately 1.5 hours, and their current and former wives will beinterviewed as well (approximately 200 wives for 45 minutes). A 30-minuteself-administered interval fertility history questionnaire will be mailed tothe 2400 wives who participated in the 1981/82 baseline interviews. Addition-ally, 7863 previously identified but uncontacted Air Force veterans will becontacted by telephone and asked six questions relating to current state ofhealth and socioeconomic status. These data will permit proper matching forthese key demographic variables. All collected data will be used to determinethe presence or absence of health effects related to wartime exposure to AgentOrange. If these data are not collected, resolution of major scientific andsocial issues will not be achieved.

3. Improved data collection techniques and program modifications have resultedin a projected 1021.5 hour decrease in interviewing time.

4. Duplication has been avoided through the action of the Cabinet CouncilAgent Orange Working Group, its Science Panel and Advisory Committee. Thesegroups have coordinated Agent Orange and related research throughout the fed-eral sector to prevent unnecessary duplication and its associated informationcollection burden. Thestudy protocol and the members of the Agent OrangeWorking Group have set a requirement to conduct this study over a 20-yearperiod.

5. Similar health information from an identified population with quantifiableexposure to Agent Orange is unavailable from any other source.

6. Information will be collected from Individual volunteers, and no smallbusinesses or other entitles will be Involved.

7. Baseline data were collected 1n 1981-82 from this predominantly middle-agedpopulation. Follow-up data are required 1n 1985 and 1987 so Increasing diseaseIncidence 1s detected promptly. The average age of this population places themembers at a time when diseases associated with the aging process are mostcommon. After 1987, data collection will only occur at 5-year Intervals.

8. This data collection effort 1s being conducted 1n full compliance with 5CFR 1320.6.

9. The White House Office of Science and Technology Policy, the Cabinet Coun-cil Agent Orange Working Group, Its Science Panel, and Advisory Committee havemaintained a continuing review of this project and approved the data collectionmethods, frequency of collection, Instructions and record keeping procedures,and report formats. Additionally, periodic contact (at least annually) ismaintained with key members of the study population.

a. Name Address Telephone Contact

Mr Jack Spey Ranch Hand Association (904) 243-5696 1983,1984850 Tarpon DriveFort Walton Beach FL 32548

Mr. Ralph Dresser 3410 Stonehaven Drive (512) 699-0761 1983,1984San Antonio TX 78230

Mr. Charles Hubbs 6002 Summerhill Road (301) 899-6716 1983,1984Temple Hills MD 20748

Mr. Delmar Spivey 2524 Jennifer Terrace (813) 784-9681 1938,1984Palm Harbor FL 33563

b. There have been no unresolved problems during the consultations.

c. Other public contacts have included briefings to House and Senate com-mittees, press releases, press conferences, and speeches to public scientificand nonscientific audiences.

10. Confidentiality is granted in accordance with direction of the U.S. AirForce Judge Advocate General and will be strictly maintained. Since "the pro-tective privilege ends where the public peril begins," appropriate civilagencies would be notified in the event serious medical problems which impactpublic health and safety are identified. The Fact Sheet sent to all potentialparticipants and the Judge Advocate General's letter of 22 October 1984 areattached (attachments 1 and 2, respectively).

11. Sensitive Questions:

a. The questionnaire contains questions on the following sensitive topicareas:

Marital and nonmarltal relationsReproductive historiesBirth defectsAlcohol historyUse of marijuana

b. Responses relating to sexual activity and outcome are necessary forassessing the Incidence of birth defects and reproductive effects alleged by anumber of Vietnam veterans. Additionally, since Vietnam was a complex environ-ment, other factors which could bear upon -observed disease Incidences must beevaluated. It 1s essential to the study that these questions be Included.

12. There 1s no cost to the respondents. The data collection for this studywill cost the A1r Force $670,000 In FY85, $218,000 1n FY86, $677,000 1n FY87,and $221,000 1n FY88. These costs are based on a firm, fixed-price contractbetween the A1r Force and SAIC, McLean VA. These are actual costs and the AirForce will Incur no additional costs for data collection using these question-naires.

13. Each of the 7,863 previously uncontacted subjects will be contacted bytelephone and given the 6-mlnute short questionnaire (pages 275-276 of theattached USAFSAM-TR 82-42 [attachment 3]). Two hundred of these Individualswill be selected as replacements for other noncompHant individuals and giventhe additional baseline questionnaire (pages 1-153 of USAFSAM-TR 82-42). Theirwives will also receive the baseline spouse questionnaire (pages 159-193,USAFSAM-TR 82-42). The 2600 spouses of subjects who took part in the study in1981-82 will be sent a self-administered fertility questionnaire, slightlymodified from that given at baseline (pages 159-193, USAFSAM-TR 82-42). Theinstrument will be identical in format but will only collect interval datasince the baseline collection effort in 1981-82. Except for the 200 replace-ment subjects, individuals will only be contacted once.

Consultation with respondents has occurred (see paragraph 9a above). Spe-cific time burdens for each of these questionnaire instruments are shown below.As noted, a decrease of more than 1,000 hours has been realized relative tothe baseline effort in 1981-82. The time estimates are based on the mean timerequired for administration of the same instruments to the same or similarpopulations at baseline. Adequate time has been included in the InformationCollection Budget for these tasks.

Table 1

BURDEN ESTIMATES FY85/86

Instrument

Source Number of(USAFSAM- Respondents Duration TotalTR 82-42) 1981/82 1985 (Hours) 1981/82 1985

Study Subject pp 1-153(Baseline)

Spouse(Baseline)

1532

pp 159-193 1500

200

Spouse Pol low-Up Modified 0 2600(Interval from pp 159-History) 193

Telephone Modified 69 7863Questionnaire from pp 275-

276

Next of Kin Not beingused

1.5

200 0.75

0.5

0.1

Difference1981/82 vs

1985

2298

1125

0 1300 +1300

300 -1998

150 -975

6.9 786

85 0 1.5 127.5 0

Net Difference (in hours)

+779

-127.5

-1021.5

14. The above decrease in burden is due to programquirements of the baseline data collection phase.

modification from the re-

15. The results of the requested data collection will not be published as rawdata; rather, they will be statistically analyzed by SAIC under contract tothe U.S. Air Force prior to publication . All statistical methods and proce-dures are in accordance with the protocol of the study which underwentextensive peer review and was approved by the Agent Orange Working Group. Afterthe data have been analyzed, a report will be prepared for public release inearly 1987, just prior to the start of the next follow-up phase. These samequestionnaires will be readmfnistered in 1987, and the results of the analysisof the new data released in 1989. Data collection using these instruments willbegin as soon as they are approved and continue until the end of the firstfollow-up phase in February 1986. The second follow-up phase will begin in May1987 and continue until March 1988. Additional follow-up data collection ef-forts will take place in 1992, 1997, and 2002, in accordance with thepreviously approved study plan.

B. Collection of Information Employing Statistical Methods

1. All 1278 of the Air Force personnel who were directly involved in the aer-ial spraying of herbicides in South Vietnam from 1962-1971 have been identifiedand all 1256 who survived their tour of duty were included in the study. Nosubsampling was used in the "exposed" group. Approximately 20,00 personnelfrom C-130 equipped units in Southeast Asia were selected as potential compari-son subjects. The "nearest-neighbor" matching technique of Kupper was used to

select an 8:1 matched set of comparisons for each "exposed" Individual (matchedon age, sex, rank, and military occupational category). After the order of thematched set was randomized, the first willing comparison and the "exposed"Individual were Invited to participate 1n the study. Thus, all "exposed" Indi-viduals and a random selection of the comparison group were Involved 1n thebaseline and follow-up data collection phases of the study. Details are pro-vided below:

Table 2

Respondent Selection

Exposed Comparison

Population 1256 20000Selected for Potential Use 1256 10048Selected for Data Collection 1n 1981-82 1256 1256Selected for Data Collection in 1985 1202* 1202

*54 have died

Table 3

Response Rates

Group 1981-82

Exposed Group 97%

Comparison Group 93%

2. The statistical methods used for sample selection were described in para-graph Bl above and are more fully discussed in attachment 4 which is taken fromthe Study Protocol (USAFSAM-TR 82-44). After the comparison "universe" wasselected, "nearest-neighbor" matching resulted in a selection of an eight-mem-ber set of comparisons for each exposed individual (matched for sex, race, age,and military occupational category). A single comparison was then randomlyselected for participation in the data collection effort. No estimation wasrequired and no stratification techniques were used. Accuracy of data col-lected in the baseline was achieved through reviews of medical records andbirth/death certificates. Similar procedures will be used during the follow-upphases. No unusual problems requiring more specialized sampling proceduresoccurred. After the second follow-up phase (1987), further data collectionefforts will only occur at 5-year intervals.

3. Response rates in the baseline collection phase were quite high and areexpected to remain so throughout the follow-up phases of the effort. Ninety-five percent of the individuals who participated in the baseline expressed astrong desire to remain in the study. Nonresponse in the exposed group can not

be corrected through the Inclusion of additional Individuals since the entire"universe" was Included 1n the study. Extensive statistical approaches to theassessment of bias arising from differential participation and other factorsare Included 1n the requirements of the contract with SAIC. All of these Is-sues of cohort selection, sample size, statistical analysis, frequency of datacollection and appropriateness of the study have been evaluated and approved bythe Agent Orange Working Group.

4f Pretesting of the data collection Instruments for the follow-up phase ofthe study will not be necessary since they are Identical to the questionnairesused at baseline. The only modifications occur 1n changing dates ("1981" or"1982" to 1985) and the collection of Interval fertility data (1982-1985)rather than life-long data. Similar modifications of specific questions willbe needed 1n the 1987 questionnaire to reflect 1985-1987 Interval data.

5. Consultants for Statistical Aspects of the Study

Peer Review Groups

Reviewing Agency Date

1. University of Texas, School of Public Health Jun 19792. A1r Force Scientific Advisory Board Aug 19793. Armed Forces Epidem1olog1cal Board Aug 19794. National Research Council, National Academy of Dec 1979

Sciences

Advisory Committee of the Agent Orange Working GroupRobert Miller, M.D., ChairmanChief, Clinical Epidemiology BranchNCI, NIH (Room 8C41)7910 Woodmont AvenueBethesda MD 20814Telephone: (301) 496-5785

Science Applications International Corporation (Contractor)James Striegel, Ph.D.8400 Westpark DriveMcLean VA 22102Telephone: (703) 821-4600

National Opinion Research Center (Subcontractor)Mary C. BurichUniversity of Chicago6030 South EllisChicago IL 60637Telephone: (312) 962-1033

20324

JA

Reevaluation of Assurance of Confidentiality, Project Ranch Hand II

AF/SG

1. We have reviewed the issue of granting full confidentialityto the study pwrticipants in the Ranch Hand II study. Therehas been no significant change in the legal issues regardingthe granting of full confidentiality since our initial recom-mendations to you in 1980 and 1981. We recommend that therebe no change to the current grant of strict confidentialitywith the two exceptions,

2. This recommendation is based upon a review of the currentcase law, statutory and regulatory law, and the public policy •issues governing the protection of the public health andsafety. Even though there is no federal statute mandatingthe disclosure of information obtained during the course ofthe Ranch Hand II study* we still remain subject to comply-ing with a valid court order for production as well asadhering to the court supported proposition that "the pro-.tective privilege ends where 'the public peril begins".(Tarasoff v. The .Regents of the University of California ,551 PZd 334 [~<5al 1976J).

3. The statutory authority relied upon by the CDC/VA ingranting full confidentiality to their Agent Orange studyparticipants is applicable only to CDC epidemiologicaland other specified research projects. It has no applica-fMnn Tn fln Air Frvrr* «rnriy. Fiirfhermnra , if -I c our opinionthat this statute does not grant any protection to the CDCin resisting a valid court order for production, nor doesit grant immunity to the CDC in complying with the prin-ciples ennunciated in Tarasoff when the public health andsafety are involved. The CDC determined that it was notnecessary to state these possible, though remote, exceptionsco the promise of full confidentiality. We believe thatthese exceptions should be stated in any notice of con-fidentiality.

4. If you desire further discussion on this matter, pleaselet us know.

THOMAS 3. ERUTCMMajor Gsnsrd, US..'...?The Ju;i"'r> .'Iciv;^^ "snared

FACT SHEET

INTRODUCTION

The USAF School of Aerospace Medicine, Brooks AFB, Texas, is conductingthe study.

You are being invited to participate in this study because of your spe-cific duties and period of assignment in Southeast Asia.

PURPOSE

To determine whether there is a causal relationship between your duty inSoutheast Asia and adverse health experience.

METHODS

An in-depth health questionnaire will again be administered to you by amember of the health evaluation team from the National Opinion Research Center(NORC).

A complete profile of your current health will be obtained by physicalexamination which will be conducted by the Scripps Clinic and Research Founda-tion in La Jolla, California.

An additional questionnaire and physical examination will be given bythese same organizations in 1987. Other examinations are planned for 1992,1997 and 2002.

Your travel expenses (including food and lodging) for the physical exami-nation will be paid by the Air Force. Lodging will also be provided foraccompanying family members, but their food and travel expenses are not coveredby the Air Force.

A stipend of $100 per day will be paid to participants who are not onactive duty, Government employees, or otherwise precluded by law from receivingsuch a stipend.

Confidentiality is to be strictly maintained. However, legal confidenti-ality of medical information is always based on the proposition that "theprotective privilege ends where the public peril begins." Therefore, seriousmedical problems which impact public health and safety or a judicial order torelease medical data will be the only circumstances in which information on aparticular study participant would be released. The Air Force and the JusticeDepartment were successful in preventing the release of health data from thefirst examination during a major court case in 1984.

If for example, a participant were found to have typhoid fever or tubercu-losis and his illness directly impacted the health and safety of others, a

committee composed of a physician (who is a specialist in the illness in ques-tion), a physician of the participant's choosing, a flight surgeon, an AirForce lawyer, and a representative from the participant's occupation win beformed to review the information. Before any information is disclosed, thecommittee must determine that the findings Jeopardize the public health andsafety.

These same confidentiality procedures were provided during the first ex-amination in 1982, and no problems arose. No one was grounded or had his jobadversely affected by his participation in the study. There is no reason tosuspect that this examination phase of the study will be any different.

BENEFITS TO YOU

You will receive a comprehensive health review and physical examination oftop level, executive caliber at no cost to yourself.

You will be completely informed of all examination procedures, you candecline to take any specific portion of the examination that you wish, and youwill be completely informed of all examination results.

The information from this study will be provided to your personal physi-cian if you so request.

Questions concerning the study may be referred in writing to the USAFSchool of Aerospace Medicine, Epidemiology Division, Brooks AFB, Texas 78235-5000, or if more convenient, the information may be provided by phoning AUTOVON240-2H14, or within the state of Texas call collect at (512) 536-3309, or fromoutside the state of Texas call toll-free 1-800-531-7231.

Please complete the attached locator card and advise the USAF School ofAerospace Medicine, at the above address and phone number, of any address orphone number changes.

V. Epidem1olbg1c Study Design; Matched Cohort

A. Design Considerations

The goal of this study clearly mandates a comprehensive ep1dem1olog1capproach, Incorporating mortality, .and historical, current, and followup mor-bidity studies* Exposure to herbicides during the 1962-1971 time period mayhave Initiated, long-term health effects that may or may not be progressive.If such effects are detectable! by a review of the subject's past medicalhistory, and can be verified*,',direct links to compensation Issues can bemade.\ turhent health status, a:S mirrored by a large number of recent VAclaims* and Inquiries, 1s of major Interest, because such claims and Inquiriesmay Indicate medical conditions that night be confirmed by a comprehensivephysical examination. If analyses of both mortality and morbidity data yieldonly Indetermlnant or weakly suggestive findings, It may be that sufficienttime has not yet passed for substantial emergence of longterm health effects.This dictates a requirement for a follow-up element to the study.

Methodological shortcomings are Inherent In each element of this com-prehensive study. To some extent, the classical deficiencies of each particu-lar ep1dem1olog1c approach are compensated by the concurrent use of the otherelements. For example, the low chance of identifying a relatively uncommondisease solely by the use of a mortality study is offset by the inclusion of acurrent morbidity study. The relatively quick feedback that can be attainedfrom current morbidity and mortality studies will serve to better define thefollow-up study, and will help to alleviate problems that arise as a result ofchanges 1n diagnostic criteria and methods over time. Nevertheless, problemsthat can affect ascertainment of disease In all phases of the study willremain. Inaccurate patient recall of antecedent events, the distortion ofinformation by knowledge of anticipated symptomatology, and participant orobserver knowledge of their exposure status can only be corrected to a limitedextent by review of records for symptom validation and "blind" assessment pro-tocols. In addition, fundamental problems dealing with adequate selection of .a control group and limiting loss to study can influence any comprehensiveepidemiologlc investigation. These and other pitfalls in study design will bediscussed in more detail in Section VIII.

The management of this project will be conducted through standard AirForce Research and Development procedures, including program monitors at AirForce Headquarters and Air Force Systems Command, and a Program Managementoffice at Brooks AFB, Texas. Contract monitors will insure that all contrac-tual efforts are conducted according to strict quality assurance procedures,and an on-s1te monitor will insure that the physical examinations are con-ducted in strict accordance with the study protocol.

Since the study has three elements and confronts a health issue withincompletely specified or uncertain endpoints, strong potential bias, andsevere time contraints, the following design represents the best overallframework for achieving validity. The design process is complex and in itselfMine dependent.

17

B. Selection and Ascertainment of the Populations for Study

(1) The Exposed Military Groups

(a) Operation RANCH HAND Personnel

Operation RANCH HAND personnel flew C-123 aircraft in RVNduring 1962-1971. Data from hand-compiled lists obtained through the RANCHHAND Association (a reunion organization), Air Force personnel records, unithistorical records, and actual C-123 flight orders, place the herbicideexposed population at approximately 1264 Individuals. Of those personnelconfirmed as RANCH HAND participants, 25% are still on active or reserve duty,with the remainder being composed of retired, separated, or deceased persons.To identify all RANCH HAND participants, an-indepth search was conducted ofall organizational records stored at the Military Records Division, NationalPersonnel Records Center (NPRC), St. Louis, Missouri.

Introductory letters will be sent to the last known addressof all Identified persons, and nonresponse will be pursued by cross-locatorsystems available within the government (e.g., Social Security Administration,VA, Internal Revenue Service). Significant efforts will be made to accountfor at least 99% of the total population (see Figure A-2, Section XV).Because of the limited number of RANCH HAND personnel, no subsampllng of theexposed group 1s planned in any phase of the study. All members will bestrongly encouraged to participate In all phases of the Investigation.

All RANCH HAND personnel are males currently ranging in agefrom 30-69 years (mean » 42.4 years). The normal C-123 crew composition wasone pilot, one copilot/navigator (both officers), and one spray equipmentconsole operator (enlisted) 1n the rear of the aircraft. The aircrew officer-enlisted ratio is 2.2:1; however, the Inclusion of RANCH HAND support person-nel (predominantly enlisted) 1n the stMdy will make the overall officer-enlisted ratio 1:1.7. Approximately 98% of the officers and 92% of theenlisted men were Caucasian. Attempts have been made to Identify all mainte-nance personnel assigned to the RANCH HAND units. Maintenance of the RANCHHAND aircraft was performed within a step-wise organizational structure.Routine daily maintenance (primary) was conducted by flight line supportpersonnel who were often dedicated exclusively to RANCH HAND operations. Moreextensive maintenance (secondary) was carried out by consolidated supportunits at the base level, which were also responsible for non-RANCH HAND C-123sas well. Major aircraft overhauls and modification were conducted by mainte-nance units at Clark Air Base, Philippines. The maintenance personnel Inthese centralized units were not directly assigned to RANCH HAND, and theirexposures to RANCH HAND C-123 aircraft and herbicide cannot be 'validated.From 1962 through 1964, the primary flight line maintenance teams were dedi-cated to RANCH HAND aircraft, and these Individuals have been Identified bythe mechanisms described above. In 1965, flight line maintenance was per-formed by personnel of the centralized maintenance organization (secondary),and It is not feasible to adequately Identify all of these Individuals fromavailable records. After 1966, the RANCH HAND organization transferred their

18

base of operations to a new location, and primary maintenance was once againperformed by personnel assigned specifically to RANCH HAND. These individualshave been readily identified. Thus, maintenance personnel directly assignedto RANCH HAND will be included in the study. These complexities are summa-rized in Table 4.

Table 4

FEASIBILITY OF IDENTIFYING AIRCRAFT MAINTENANCEPERSONNEL (TOTAL POPULATION) EXPOSED TO HERBICIDE ORANGE

Primary SecondaryTime Maint Personnel Maint Personnel2

Jan 1962-Jul 1964 Yes No

Aug 1964-Dec 1966 Yes/No3 No

Jan 1967-Oct 1971 Yes No

Hndividual assigned to RH; total number (denominator) knownindividual not assigned specifically to RH, although may have serviced theaircraft; denominator not ascertainable

3other documents permit ascertainment of a portion of this group

Because of the significant combat hazard associated with low, slow flyingmissions, some early RANCH HAND crewmembers were elite volunteers (see Risk-Taking Bias, Section VIII, C). In fact, RANCH HAND crewmembers comprised oneof the most highly decorated units during the RVN Conflict. Anecdotal storiesreveal that most crew members were, on occasion, heavily exposed to HerbicideOrange due to normal or combat induced equipment malfunctions within the air-craft. Many former RANCH HAND personnel are expected to be currently employedin the aerospace industry as commercial airline pilots, airline managers, andflight mechanics. RANCH HAND personnel still on active duty are expected tobe found in senior management positions.

(b) Alternate Exposed Populations

(I) Introduction

The principal investigators, members of all of the peerreview committees, and independent consultants have clearly recognized thatthe statistical power of this RANCH HAND study 1s suboptimal for the detectionof specific uncommon conditions or diseases* This limitation is inherentbecause the size of the RANCH HAND population is fixed at approximately 1200individuals, and 1t cannot be Increased.

19

A brief review of alternate military populations is inorder to highlight the significant advantages of the RANCH HAND population.The desire to achieve more optimal statistical power by merely increasing thesize of the population under study must be balanced with a careful analyticprocess which assesses the exposure level of alternate populations, and cate-gorizes them as either additive or nonadditive to the RANCH HAND study popula-tion.

(2) U.S. Army Ground Personnel

Some U.S. Army personnel were undoubtedly exposed toherbicides during their duty in Vietnam; however, the objective ascertainmentof exposed individuals is not possible. Any attempts to identify individualsassigned to combat units which may have been exposed would result in anunacceptable degree of misclassification since U.S. Army personnel records donot exist which would allow the accurate identification of soldiers below thebattalion level. This lack of demoninator data, and the high degree of mis-classification in determining the exposure status of Army troops makes thispopulation unsuitable for inclusion in the framework of the RANCH HAND Study.

(1) Ancillary Air Force Groups (Non-RANCH HAND Personnel)

Air Force handlers of herbicide drums in RVN wereexposed to herbicides because of drum leakage. As the drum handlers were adlib participants, no personnel designator was assigned to these individuals,thus prohibiting computer tracking and identification. The size of this popu-lation is unknown, but it is expected to be small (less than 200), as themajority of drum handlers are known to have been Vietnamese. Additionalgroups such as U.S. Army helicopter crews, casual observers (both Army and AirForce), and experimental fighter-bomber personnel who may have occasionallyconducted spray operations were also potentially exposed. However,population-at-risk determinations for all of these groups cannot be made, andany identification of individuals exposed in these situations must rely onself-selection or incomplete ascertainment. Also, the selection of suitablecontrol groups for a study of these individuals is difficult If not impos-sible.

(£) U.S. Marine Corps Troops

On 16 November 1979, the GAO released a report whichsuggested that a herbicide-exposed population of nearly 22,000 U.S. MarineCorps troops could be Identified, and that this identified group would beappropriate to study. Records exist which locate Marine Corps battalion head-quarters near the C-123 spray paths. The GAO. made several Improper assump-tions to conclude that all of the identified marines were in fact exposed.Specifically, all battalion troops were assumed to be located at the battalionheadquarters. Further, the effect of prevailing winds on the direction ofspray drift, and the photodegradatlon of the chemicals were not considered bythe GAO. The National Research Council panel considered the GAO analysis, andproposed a study of 5900 marines who were "near" spray paths on the same day

20

as the spraying. The '"exposed" group was to be contrasted with the mortalityexperience of 212,000 presumably unexposed controls (also marines). The RANCHHAND study described in this protocol consists of approximately 1200 exposedindividuals and 6000 controls for the mortality study phase. Despite the factthat the RANCH HAND Study involves a smaller sample size than the proposedMarine effort, the RANCH HAND Study is more powerful statistically. Specifi-cally, lower exposure to herbicide by a conservative factor of from 1/10 to1/1000 and misclassification in Marine exposure groups renders the MarineStudy far less powerful than the RANCH HAND effort. As described in SectionVI, misclassification and decreased exposure are seen to be independentfactors additively decrementing Marine Study statistical power. Even when all21,900 marines within the herbicide spray paths up to 28 days following thespray operations are considered exposed, the RA1O HAND Study is noted to besignificantly superior.

Conclusions

The Operation RANCH HAND participants are the mostsuitable of the military populations to study in evaluating the longtermeffects of herbiclde/dloxin exposures. The RANCH HAND group had a much higherlevel of exposure which was sustained over a prolonged period of time. ThisIncreased level of exposure implies that RANCH HAND personnel would be morelikely to develop more acute and chronic symptoms from the exposure, and wouldmanifest them sooner than the other exposed military personnel. The additionof significantly less exposed and/or misclassified groups to the RANCH HANDpopulation for the attractive purpose of Increasing statistical power wouldconstitute an egregious dilutional error.

(2) Control Group (Not exposed to Herbicide Orange)

A review of all specialized flight units present in SoutheastAsia during the RVN conflict, reveals clearly that there is no absolutelyideal control group for the RANCH HAND population. C-130 aircrew members andsupport personnel were selected because of sufficient population size, similartraining profiles, and psychologic similarities to the RANCH HAND group.

Total ascertainment of the C-130 population is being conducted bycomputer and hand selection for specific military flying organizations, andforeign country service, during the Interval from 1962 thru 1970. Over 2.3million personnel records have been reviewed, and the approximate C-130 popu-lation size Is 23,978 Individuals. Aircrew members who flew C-130 aircraft inSoutheast Asia during 1962-1970 were selected as controls for the RANCH HAND

• aircrew population. The C-130 flight line maintenance population were ascer-tained from personnel records by similar mechanisms, and served as the spe-

'•: £lf1c control population for the RANCH HAND support personnel. The propor-j; jtlons on active duty, and non-active duty status are expected to parallel thef patterns In the exposed group.

* • Another possible control group, the non-RANCH HAND C-123 popula-: tlon, is known to be too small (approximately 3000) to provide adequate samp-

ling flexibility and replacement -under the proposed matched variable concept

21

(see below and Section VI). Also, many of the RANCH HAND aircraft were recon-figured for transport and Insecticide missions and thus, the non-RANCH HANDC-123 crews responsible for these other missions may have been exposed tosignificant Herbicide Orange residue In these aircraft. Therefore, this groupmay not have been truly unexposed to herbicides, and was discarded as anappropriate control population. Crewmembers of C-7 transport aircraft werealso considered as a potential control group; however, because of small samplesize (1000-2000) and the fact that they served 1n RVN only during the post1967 era, they were also dropped from consideration. '

The normal crew composition of a C-130 1s three officers and twoenlisted personnel. The control group Is considered to- be "pure" from thestandpoint of lack of occupational exposure to herbicide. The entire controlgroup will be considered MnonvolunteerM-w1th respect to abnormally high combatrisk. While 1n general they will possess lifestyle characteristics and socio-economic backgrounds similar to the exposed group, their overall combatmorbidity/mortality and the resultant stress influences upon general healthmay be slightly less than in the exposed group. For those separated andretired C-130 controls, similar proportions to the exposed group are expectedto be employed in the aerospace industry. Known and estimated factors of thecontrol and exposed populations are summarized in Table 5.

(3) Matching Procedures and Rationale

Each member of the exposed group has been computer matched to aset of C-130 controls comprised of approximately 10 individuals using threevariables. Since the L,;O groups are highly selected and inherently similarwith respect to many variables, very close matches are feasible. Thisepidemiologic design incorporates a matched concept because: (1) a matchedcohort design will provide maximum test power throughout the entire study, and(2) statistical intergroup comparisons may be made without normalization bythree key variables known to effect symptom frequencies of interest, thusproviding greater power for complex statistical .testing. It is apparent thatfollowing the match, both exposed and control populations will be very nearlyidentical with respect to the three influencing variables so that a replace-ment concept is feasible (see F below). In the event that frequent matchbreaks occur, stratification techniques can be used.

The selection of the control group produces an inherent match forequivalent SEA experience, and additional matching has been conducted for (1)age, by year of birth and closest month possible, (2) Air Force SpecialityCode (AFSC) as an absolute match, and (3) race (Caucasian versus non-Caucasian) as an absolute match. Specific rationale for these variables is asfollows: (1) the age match controls for the many clinical symptoms and signsassociated with advancing age, (2) AFSC controls for officer-enlisted status(as well as crewmember-noncrewmember status), a variable strongly linked toeducational background, current socio-economic status, and moderately linkedto age (5 year median difference) and socio-economic background, and (3) racecontrols for differences in chronic disease development, socio-economic back-ground, etc.

22

fNJ

KNOWN FACTORS

POPULATION SIZE

OFFICER/ENLISTED RATIO

AIRCRAFT FUEL (AV-GAS)

OCCUPATIONAL HERBICIDEEXPOSURE

ESTIMATED FACTORS

OCCUPATIONAL INSECTICIDEEXPOSURE

COMBAT HAZARD

RVN-IN COUNTRY ASSIGNMENT

••>• ••• •• ="T

* '<*' ' sw- ,-u:"-!-.- .-•*.:: ---.---,"•:.. ;•. , ' •. '

'> Tabl* 5..

>ARISON OF THE STUDY GROUP TO POSSIBLE CONTROL GROUPS BY

KNOWN

STUDY GROUP

RANCH HAND C-123

1264

1:1.7

YES (+JP-4)*

YES

2+

4+

4+

AND ESTIMATED FACTORS

POSSIBLE

Non-RANCH HAND C-123

3000

1:2

YES (+JP-4)*

YES/NO**

0

3+

4+

'

CONTROL GROUPS

C-7 C-130

1200 23,978

1:2 1:2

YES NO (JP-4 only)

NO NO

0 0

3+ 2+

4+ 2+

• rv**"i*r

*In 1968, aircraft were modified with a JP-4 booster."Contaminated aircraft reconfigured for transport may have resulted in exposure to non-RANCH HAND personnel

The Inherent match fop SEA experience controls for combat-Induced physiol-ogic, psyehophys1o1og1c, and other related morbidity and mortality disorders.Additionally, this Inherent match may reflect the effects of alcohol consump-tion, the use of chemoprophylactic and/or Illicit drugs, and the acquisitionof tropical diseases associated with life In SEA. The comparisons of theexposed (RANCH HAND) subjects and their selected sets of controls are detailedIn Appendix Table A-4. Only 4 of the ten categorical AFSC/case strata hadless than ten controls for each exposed subject. The group of Caucasianpilots had a mean of only 9.5 controls per exposed subject, due to the extremeages of several Individuals, and the strata of Black pilots and other Blackofficers had means of 2.7 and 5.0 controls respectively. However, since therewere only seven black officers 1n the exposed group and only thirty controls,high numbers of tight matches could not be achieved. Black enlisted aircrew-members had a mean of 9.8 controls each.

(4) Computer Science and Statistical Details of the Matching Process

As described above, the matching for this project has been performed usingthree variables: occupational category, race and age. Five occupationalcategories (officer/pilot, officer/navigator, officer/other, enlisted/flightengineer, and enlisted/other) have been used to reflect socloeconomlc statusand aeronautical rating. The variable of race has been dichotomized Intoblack and non-black. Ten matched controls have been selected for each exposedsubject, regardless of current vital status. The computer method applied toselect the control subjects Is an adaptation of a procedure studied by Raynorand Kupper (Nearest Neighbor Matching on a Continuous Variable, TechnicalReport, Department of Blostatlstlcs, University of North Carolina, 1979). Asthe first step, the RANci! HAND and control groups were partitioned Into tenstrata using the categorical occupational and race variables. The Raynor andKupper matching procedure was then applied Heratively within each of thestrata to match for the continuous variable of age, given In months. TheRaynor-Kupper procedure Involves the following steps:

STEP #1. The RANCH HAND cohort in a given strata is randomly permuted.

STEP #2. The first RANCH HAND subject in the permuted set is selected formatching.

STEP #3. The closest available control is assigned to the selected RANCHHAND subject using the absolute value of the difference between the months ofbirth of the RANCH HAND and the control subjects. If the closest availablecontrol is further than 60 months from the selected RANCH HAND subject, ablank is assigned. Tied assignments are broken randomly.

STEP #4. Step #3 is repeated for all RANCH HAND subjects in the strataproceeding down through the permuted set, until the entire RANCH HAND cohortis exhausted.

STEP #5. Steps #1 through #4 are repeated ten times for each RANCH HANDsubject to construct a 1:10 study set. At the completion of the matchingactivity, the RANCH HAND - Control study matrices for each of the ten occupa-tion-race strata can be diagrammatically represented as in Figure #1.

24

i • Figure 1. MORTALITY ANALYSIS COHORTS

RANCH HAND COHORT

R2

R3

R,

V,J

C i . ip°2 , 1

C-i ,^3,1

Ci» !

C l , 2

°2,2

^3,2

^»f ,2

Ci

r,^2

r.** 3

r<»

3 V* 1 3 • • • • •1 » J

p,3 ^2,<t

^3,^

10

C l ,10

^2, 10

3,10

^ , 10

1200,1 l200,2 C1200,i» 12QO,10

RANCH HANDCOHORT

Figure 2. MORTALITY MATRIX

CONTROL COHORTS -

ci,2,m'

cj,5,m'

Ri200 C120o,i,m' C120o,2,m' ***

In each row of this matrix the controls are ordered from nearest to far-thest in terms of age of the matched RANCH HAND person. The next operationdefining the control group involved randomization of all of the controls ineach row of each stratum matrix to negate the ordering by age. Then, thefirst five members of each control set for each RANCH HAND person are identi-fied as being subjects in the mortality portion of the study. The resultingoccupation-race strata matrices now have the form shown In Figure 2.

In Figure 2, Cj ' or Cj ^1 may be equivalent to anyto the randomization process.

of Figure 1 due

Table 6 summarizes the results of the matching process, and Appendix TableA-5 provides a more complete statistical description of the process. In thesetables, the age difference between the month of birth of the control and themonth of birth of the RANCH HAND person, (counting months from 1900) and thecumulative number of controls and the cumulative percentage with this differ-ence are shown.

25

Table 6. RESULTS OF THE MATCHING PROCESS (1:10)

CumulativeAge Difference (in Months) Number of Controls Cumulative Percent

0 8612 70.6

1 10287 84.3

2 10749 88.1

3 10984 90.1

4 .11167 91.6

5 11322 92.8

6 11410 93.5

12 11688 95.8

24 11921 97.7

36 12028 98.6

48 12129 ' 99.4

60 12197 100.0

(5) Study Group Selection Procedures

(a) Mortality Analysis

A 50% random sample of each control set will be drawn andused to comprise a 1:5 mortality analysis, as described in section (4). Thevital status of each subject In this sample and of all exposed subjects willbe ascertained at a minimum frequency of every five years for the 20 yearduration of the study. Those individuals dying of combat causes will beexcluded from the mortality analysis as it is assumed that combat death isindependent of herbicide effect. Further, the known differential combat deathrate between the RANCH HAND and control groups can be attributed to thehazardous and unique nature of the RANCH HAND mission. Twenty-two RANCH HANDpersonnel (15 officers and 7 enlisted) died 1n combat. Medical record reviewswill be accomplished to assess the illness experience of these individualsprior to combat mortality.

(b) Historical Morbidity Study

Retrospective or historical health data will be gathered oneach exposed subject and from the first randomly selected mortality controlfrom his set by questionnaire techniques. Living but noncompliant controls in

26

the historical morbidity study will be replaced by a compliant controlselected from the control set. In order to avoid an information gap for dataon deceased individuals, surrogate interviews will be obtained from the firstorder next-of-kin of exposed and control subjects dying of noncombat relatedcauses between the date of their assignment to Southeast Asia and the initia-tion of this study. Since the validity and accuracy of surrogate derived datamay not be equivalent to data obtained directly from living study subjects andtheir spouses, these data will be subsetted for analysis. All available medi-cal records, (military, VA, and civilian) will be reviewed for all subjectsselected for this morbidity analysis.

(c) Prospective Morbidity Study

A baseline physical examination and review of systems willbe conducted, and a prospective or followup approach will be used to assessthe current state of health of study subjects using a series of questionnairesand physical examinations over the next 20 years. Each 1iving exposed subjectand the randomly selected primary control will be included in the question-naire and physical examination phases. In this prospective study of morbid-ity, primary controls who are deceased, unaccountable or unwilling to partici-pate in the followup studies, will be replaced by a willing subject from theremainder of the control set (Figure 3). The selected control for a RANCHHANDER dying of a noncombat cause will be retained throughout the question-naire, physical examination, and followup phases of the study. Since thecontrol's vital status and volunteerism should be independent of the matchingsequence, many primary controls should enter the study. The remaining membersof the control set will be used as replacement candidates for possible uselater in the study (see section F below). All replacement controls will beclearly identified for the purposes of subset analysis so that populationdifferences, if any, between the first randomly assigned selectees(noncompliant) and the replacements (compliant) can be assessed. Specificrules and procedures for study entry are found in Table A-6 and Figure A-3 ofthe Appendix.

(d) The Interrelatedness of the Comparison Groups

It should be clear from the foregoing discussion that thestudy populations of the mortality, historical morbidity, and prospective fol-lowup phases are highly related but different. Once selected, the mortalitycontrol cohorts will remain unchanged throughout the 20 years of observation.The population under study in the historical morbidity phase will initially bea randomly selected subset of the mortality comparison group; however, some ofthese primary controls may be decreased or noncompliant for the voluntaryaspects of this phase of the study. In this phase, noncompliant controls willbe replaced, but deceased controls will not, as surrogate Interviews with therrext-of-k1n will be used to reconstruct morbidity data. The subsetting andreplacement procedures create the difference between the mortality and histor-ical morbidity comparison groups. The population 1n the prospective morbidityphase is the comparison group from the retrospective phase plus additionalreplacements for the deceased controls. Thus, It 1s clear that the comparisongroups are slightly different, but they would be identical if no deathsoccurred since 1962 and all primary controls were compliant.

27

Figure 3.

SELECTION PROCEDURE FOR THE QUESTIONNAIRE,PHYSICAL EXAMINATION, AND FOLLOW-UP STUDY

LIVINGRANCH HANDINDIVIDUAL CONTROL INDIVIDUALS

Randomly SelectedMortality Controls

JL

I**

t1:1

t

t Deceased- Unwilling* Volunteer** Replacement Candidates

C. Overview of Statistical Methodology

The design of the study 1s presented In schematic form In Figure 4.R' refers to RANCH HAND personnel and C" refers to the collection of all pos-sible control Individuals. As defined, r 1 and C" will contain individuals whoare deceased of noncombat causes. Combat deaths are excluded from R' and C".Since C" is approximately 20 times larger than R', a randomized subsample C1

and C" will be obtained. C1 will be constructed from C" by computer selectionof the ten matched controls for each exposed study subject. As previouslynoted, close matches will be mide for the variables of age, AFSC, and race.The matched controls will form tin cohorts, Ci through C10, as shown in Figure1. A 50% random sample from each of the matched control sets of 10 will beselected for inclusion in the mortality assessment so that a group, C' isobtained that consists of 5 matched controls for each exposed subject. Thesecontrols will be designated as initial replacement candidates for the morbid-ity and follow-up studies. The remaining individuals in the control set willbe additional replacement candidates in the event that replacement must occurbeyond the members of the mortality set (see Figure 3). C1 w1H be con-structed without regard to whether the individual is currently living or deadso that an assessment of noncombat mortality can be accomplished.

28

FIGURE 4

rovP

DESIGN SCHEMATICEXPOSED

R

MORTALITY ANALYSIS QUESTIONNAIRE PHYSICAL EXAMINATION

UNUSED

Referring again to Figure 4, R and C indicate living RANCH HAND mem-bers and primary matched controls. If mR1 Is the proportion of R1 found tobe deceased, then

R - (1 - mR'jR1

The questionnaire will provide data concerning specific symptoms andother findings In the R and C groups. Thus, various questionnaire findingrates 1n R, SR, will be calculated and compared with the corresponding rates1n C, so

The questionnaire will allow allocation of RANCH HAND personnel Intothose with symptoms on questionnaire, Indicated by RS, and those without, R?.Similarly, the control J/idlviduals will be placed Into symptomatic, IndicatedCS, and asymptomatic, C*S~ groups.

The physical examination performed on Individuals from R and C willallow estimation and comparison of rates of physical findings in thesegroups. Rates of abnormal physical findings can be symbolically Indicated asfR and fc for RANCH HAND and control groups respectively. Comparison ofthese rates is very Important and details will be discussed below.

Let fRs be the rate of physical findings among RANCH HAND personnelwith findings by questionnaire and let fR$ be the rate of physical findingsamong RANCH HAND people with no findings on their questionnaire. For mostdisease processes it would be expected that fR$ should be a larger rate thanfRS*. If fRs 1s observed to be equal to or less than fRjf, an Interpreta-tion of over-reporting may be warranted, although the possibility of subclinl-cal disease is recognized. Rates fcs and fr,$" will also be estimated, andcomparisons between fR$, fc$» fRS and fcrw-|11 De accomplished.

The eight rates IDR', me', SR, Sc, fRS, fR$, frs»and their refinements fully characterize chls study. As depicted in. Figure"vertical comparisons of these rates provide relative risks mR'/mr,'.SR/SC, fR/fc» fRS/^CS ancl ^RS/^CS" which are of central impor-tance in defining herbicide effects. "Horizontal comparisons" relate fR toSR» fRS to fRS» fC to -C and fCS to fCS"' Specifically, theratio fR/SR is the ratio of physical findings to reported symptoms in theRANCH HAND population. This ratio may be contrasted with the ratio fr/scand if fR/SR is less than fc/$c over-reporting is suggested. Like-wise, if fR$ is less than fR$~, over-reporting is further suggested. Acomparison of fR$/fR$ to ^CS/^CS contrasts the odds of findings givensymptoms in the RANCH HAND population with the odds of findings given symptomsin the control group. If these odds are lower in the RANCH HAND group, over-reporting is again implied. Further discussion of these rates is presented inSection V.G.

During the questionnaire and physical examination phases of thisstudy, only one of the five randomly selected mortality study controls will beused for each RANCH HAND individual. If this control is unwilling to partici-pate, another mortality study control will be used as indicated in Figures 3

30

and 7. These replacements w i l l b e carefully labelled for purposes of statis-tical analysis. A detailed discussion of this replacement concept is found inSection VI.

D. Mortality Study

(1) Introduction

The mortality, retrospective morbidity, and follow-up studies arecomponents of a "non-concurrent" prospective study used in the observation ofa specially exposed group or industrial population starting from some date inthe past. The initial exposures occurred 11-19, years ago and varied inIntensity and duration from one RANCH HAND, member to another. Access toemployment, medical, or other types of records is an obvious requisite forsuch a study. The classical "case-control", retrospective study is not opera-tive 1n this protocol due to the lack of defined clinical endpolnts. Themortality study will be conducted in two phases; a review of past mortality,and a continuing assessment of the death rate in the exposed and controlcohorts over the twenty year duration of the RANCH HAND II project.

Based upon USA vital statistics, 8.6% of the study subjects areexpected to have died between completion of their Vietnam tour and Initiationof this study. Of these deaths 1n the control group, approximately 30% shouldhave been due to cardiac causes, 24% to neoplasia, 13% to accidents, 5% tocirrhosis, and 0.1% to leukemia.

(2) Data Collection Methods

The mortality status of the exposed cohort and the randomly sel-ected controls will be ascertained using multiple techniques including: pay-ments of Veterans Administration Death Benefits, Social Security Administra-tion Records, Air Force Accounting and Finance Center wage and retirementpayments, and interviews with subjects or thetr families. Death certificates,autopsy reports, and medical records will be obtained for each deceased sub-ject. The International Classification of Disease, Ninth Revision, 1978, willbe used for coding. At the time of the first followup examination, all par-ticipants will be asked to allow an autopsy to be performed at governmentexpense at the time of their death, and have the tissues forwarded to theArmed Forces Institute of Pathology, and the results sent to USAFSAM.

(3) Analysis of Mortality Data

(a) Basic analyses

Considering the basic groups R1 and C1 1n Figure 4, individ-uals will be classified into three categories: alive, dead, or unaccounted.If a large number of Individuals 1n each group are unaccounted for, the studycan obviously be severely biased. Thus, significant effort will be expendedto reduce the unaccounted category as much as possible. At most, 1 to 3percent of both groups can be allowed to remain unaccounted, with a 1% rate

31

being preferred. If for example, the mortality rate In C' Is 0.10, then anunaccountabmty rate of 0.01 could alter the mortality rate by as much as10%. Whatever the unaccountablllty rates, the pattern of unaccountabmtymust also be compared between groups R1 and C1. For example, the possibilityof age differences must be examined, particularly 1f the unaccountabmtyrates are high. The following paragraphs discuss the analysis of mortalityunder the assumption that low unaccountablllty rates have rendered the mortal-ity analysis meaningful.

Multiple mortality assessments will be accomplished duringthe course of this study, one at the beginning of the study, using availablemortality data on the basic mortality cohorts In C1 and R1 (5:1 ratio), andothers using mortality data on R 1 and all controls used In the study (both C1

and replacements) as controls accumulate prospect1vely. The proceduresdescribed here will be used In all of these assessments.

Henceforth, within the protocol, the term "mortality data"does not distinguish between that data collected Initially and that data col-lected In the future.

The mortality data will be analyzed using several differentapproaches. Crude age-specific death rates will first be calculated andtabulated. Age will be divided Into k strata, and person-years will beobserved for each strata as will be the number of deaths 1n each strata. Inthis manner a tabular display will be developed as shown In Table 7.

1

2

3

Table 7

STRATIFIED FORMAT OF AGE-SPECIFIC DEATH RATES

Ranch Hand

PersonYears

PH

Pl2

Pl3

Deaths

mu

m12

DeathRate

PersonYears

P22

P23

Controls

Deaths

m21

DeathRate

r22

m23

mik P2k rzk

Since the death rates r^j and r2j are Poisson variables, theycan be contrasted directly. If the relationship of rtj to r2j is found tobe consistent between age strata (within statistical variability), a summarymortality index may be calculated. One summary index that will be calculatedis the Standardized Mortality Ratio (SMR) which is (Armitage, 1971):

32

SMR = M x 100

kP

J-lM =

"Classical" standardized mortality ratios using national mortality data as thereference will not be calculated for RANCH HAND II due to the effects of thehealthy worker phenomenon. The term £ rn'ij is the total number of deathsobserved in the RANCH HAND group while I Py r2j is the number of deathsthat would be expected were the age-specific RANCH HAND death rates the sameas the age-specific control death rates. Thus the concern is for an SMRgreater than 100%. If a crude death rate for controls, dc, is calculated as

dc

I P2jkI

J-l

then the standardized crude rate for the RANCH" HAND group dRu is

dRH s MdC'

An approximate statistical test would regard dgx as a Poisson random vari-able with mean dr,.

An alternative approach to the provision of a proportionate mortality ratio isthat of Breslow and Day (1975). In this treatment, a multiplicative model 1sEmployed, for example:

where Ajjfc is the mortality rate, e-j Is the contribution due to populationdifferences (RANCH HAND versus Control), $j is the contribution due to age

33

group, and 1s the contribution due to length of time 1n RVN, etc. Thestatistical approach here 1s via maximum likelihood.

Logistic models (Walker and Duncan, 1967) have been exten-sively studied at USAFSAM for application 1n cardiovascular disease analysis.These models, 1n the herbicide context would have the form

where

P = probability of deathA • age 1n yearsT = length of time 1n RVNR = Indicator variable for raceE = exposure variable

and where olf 61, 1»1,2,... are coefficients to be estimated from the data.Testing for a group difference can be accomplished by estimating Bi, and theInteraction coefficients such as 05. If all interaction coefficientsinvolving the exposure variable E are zero and E is treated as a 0/1 variable,Cox (1958a, 1958b) has shown that tho most powerful test for non-zero ei,, inthe setting of matched pairs, is McNemar's test. This latter test makes fulluse of the paired design of the study. For McNemar's test, the data are castinto a 2 x 2 table as shown in Table 8. In this table, "a" Is the number ofpairs in which both members have died, "b" is the number of pairs in whichonly the RANCH HAND person has died, etc. Using McNemar's test, the teststatistic

X2

b + c

is calculated and referred to the chi-square distribution with one degree offreedom. Cox (1966) and Meittinen (1969) provided extensions of McNemar'stest for R controls per exposed (R-to-1 matching). Of course the above anal-yses will be accomplished considering all deaths, and deaths by specificcause.

As previously discussed, RANCH HAND personnel may be characterized as risktakers. This risk taking behavior may be associated with increased mortalityfrom a variety of causes. On the other hand, herbicide exposure has causedneuropathy in the RANCH HAND personnel, one could anticipate that this dis-ability would increase the probability of accidental death. Therefore,

34

Table 8

FORMAT OF McNEMAR'S TEST

CONTROLS

RANCH HANDPERSONNEL

DeadAliveTotal

DEAD

ac

a+c

ALIVE

bd

b+d

TOTAL

a+bc+dn

accidental death rates among RANCH HAND participants will be corrected forrisk taking. This can be accomplished by including assessment of risk takingbehavior in the questionnaire, indepth interview, and psychologicalevaluation. Both control and RANCH HAND mortality could be corrected usingthese measures, with the resultant rates being less biased and, therefore, abetter indicator of exposed versus control effect.

(b) Morta1ity analysis without covariates.

The first step in the statistical analysis of survival datais descriptive, i.e., the construction of summary measures which provide abasis for comparing different exposure groups without any allowance for theeffects of possibly confounding variables (e.g., age) except perhaps for somelimited stratification. Since one must expect many "losses to follow-up",only methods which take full cognizance of this complication will be consid-ered. It should be pointed out that all the methods described below assumeindependence between censoring (e.g., loss to follow-up) and death or morbidevent, although some techniques permit different patterns of censoring 1n dif-ferent exposure groups.

The life table method can be adapted to obtain a step-func-tion approximation to survival distributions in the presence of censoring(Chiang, 1968, Gross and Clark, 1975). The failure time distribution is thefunction F°(t) which provides the probability of deajfh at or before time tin the study. The Kaplan-Meier estimator of F°(t) is F°(t) where

- 1 - n [1 - i/R(Ti)]1eD(t)

In this equation, D(t) 1s the "death set" at time t, I.e., the set of allIndices i of Individuals who were observed to fail before time t. R(Tf) isthe number of individuals who were at risk just before time Tj, the time ofdeath (or morbid event) of the 1tn study individual in D(t). A nonparamet-ric approach to testing the equality of survival distributions in a matched

35

pair study has been developed by We1 (1980). His statistic Is a generaliza-tion of the Gehan (1965a) statistic. A second test for homogeneity of survi-val distributions for dlscretlzed failure data 1s the test for marginal homo-geneity In a KxK table due to Stuart (1955). Thirdly, the McCullough Modeland test may be used on the KxK array to test for marginal homogeneity andstochastic ordering.

(c) Mortality analysis with covarlates.

These methods allow adjustment of mortality rates or morbid-ity rates using covarlates such as age, race, length of time 1n RVN, AFSC,risk taking score, etc. For t.he purposes of this discussion 1t will beassumed that the covarlables are categorical, that there are only two suchcovaMables and the covarlables do not Interact In affecting the hazard ofdeath or morbidity. These assumptions can all be relaxed using availablemethods.

The hazard function h^(t) for the 1th Individual 1n thestudy is the function which provides the conditional probability of death ormorbid event in the time interval (t, t+dt) given his survival up to time t.The function H,(t) where

tHl(t) = / »H(T)dT

0

is called the cumulative hazard for the ith individual. It 1s readily shownthat the failure time distribution F?(t) is given by:

F°(t) = 1 - exp(-Hi(t))

From this last equation it follows that hi and F° are transforms of eachother, hence the dependence of F° on covariables may be modeled via h^.This may be accomplished as follows. Let X-j(t) and Y^(t) denote discretevalued stochastic processes pertaining to the ith individual and describingtwo covariates of interest (e.g., one may be an exposure variable and theother may be covariate such as age or crew position). A basic model forhazard is:

hi(t) = exp [5Xi(t) + nYi(t)]

where 5 and n are "log-relative risks". This model may t* extended to allowfor any number of possibly interacting factors. Inference about log-relativerisks may be drawn using either an approach derived fro* 0. R. Cox (1972) byE. Peritz and R. Ray (1978) or using an approach descrirvd by Frank (1977).Another model, termed the proportional hazards model, is c*ven by

36

Mt) • *

The proportional hazards model has been discussed, for the special case thatXi(t) does not change with time, by Cox (1972). A test for the equality ofsurvival distributions in a matched pair study which incorporates theproportional hazard model has been given by Breslow (1975). A test of fit forthe proportional hazards model is given by Schoenfeld (1980).

E. Morbidity Study

(1) Genera1 Cons i derat1ons

A vigorous attempt to determine the morbidity experience of allexposed subjects and their primary controls will be undertaken using question-naires, indepth personal interviews, and physical examinations. A waiver willbe requested from the U.S. Attorney General so that medical information col-lected during the conduct of this study may be exempted from subpoena intoFederal Court. Total confidentiality of medical information will be grantedto subjects who are not on active duty, and partial confidentiality will begiven to active duty subjects with release of information to the DOD only ininstances where there is a public safety or national security risk. The sche-dule and method of contact with the study subjects is depicted in the AppendixTable A-7.

(2) Questionnaire Methods

All living exposed subjects and their primary controls will beoffered a comprehensive personal and family health questionnaire administeredin the subject's home by a civilian contractor.

In addition to subject interviews, a face-to-face interview willbe conducted with the current spouses of the subjects to obtain a more accu-rate and complete assessment of fertility and reproductive function. Repro-ductive information that will be collected includes but is not limited to thenumber of live births, the number of still births, the number of miscarriages,the number of children conceived, the number of abnormal offspring, and thetotal years of marriage. Previous spouses of divorced or remarried subjectswill also be interviewed to obtain similar data. Interviews with the firstorder next-of-kin of deceased subjects will provide morbidity data on the sub-ject prior to his death. Whenever subjects, their spouses or next-of-kin willnot consent to participate in a face-to-face interview, attempts will be madeto elicit the Information by telephone.

The questionnaire 1s an important part of this study because non-compliance rates for the physical examination and its face-to-face medicalinterview are expected to be substantially greater than non-compliance withthe initial questionnaire. The questionnaire serves a four-fold purpose: (1)to capture baseline personal and medical data on subjects who might be noncom-pliant for subsequent physical examinations, (2) to serve as a cross-reference

37

source for objective data obtained at the time of physical examination, (3) toobtain a targeted medical Inventory, Independent of the physical examinationprocess, and (4) to obtain health perception data to serve as a foundation forthe replacement strategy. As depicted In the Appendix, Figure A-2, only anestimated 40% of the RANCH HAND population will participate In the examina-tion, while at least 65% will respond to the questionnaire. The Informationcollected by questionnaire from these additional 309 Individuals and theircontrols will provide valuable morbidity data which would otherwise be lost.The questionnaire (see Section XI) will emphasize Identification data, RVNtour history, dermatologlc conditions, neuropsychlatrlc conditions^ fertilityaberrations, genetic defects 1n offspring, sensory defects, and personalityfactors. A targeted medic?:i ^nv^ntory will be Included In the questionnaire,and will Inventory symptoms prior to, .during, and after duty In RVN as well asthose currently manifested. It will take approximately six months to completeall Initial questionnaires on both groups. The questionnaire will be "field-tested" by the contractor on former Air Force personnel with RVN experience.Specific questions on the questionnaire will be directed to verifiableInformation, wherever possible. Questionnaire development and refinement,Including specific response verification procedures have been pursued throughcivilian contract. Questionnaire data will be cross-linked and Integratedwith medical record Information and physical examination findings. Question-naire data from Individuals not completing all phases of the study will not bediscarded, but will be Incorporated within the entire data base where statis-tically appropriate. Each participant will be asked to sign release forms sothat all civilian health records, including those of dependents, can beobtained and reviewed as necessary. Attempts will be made to obtain patholog-ical reports and specimens following surgical procedures. Federal healthrecords on all family members on file in the NPRC will be retrieved. Forretired members, and separated members with VA privileges, all available VAmedical records will be obtained. All retrieved medical records will bereviewed, scored, compared to questionnaire data for reliability, and then beentered into a repository system. I 'antified participants who are non-responsive to questionnaire will be pursued to determine status, disinterest,moribund state or death, etc. These individuals will be cross-referenced inother federal record systems in an attempt to achieve total ascertainment.Death certificates and autopsy reports will be retrieved on all dead exposedand matched control subjects fo • the mortality analysis. Birth/death certifi-cates will be sought for all offspring.

(3) Physical Examination

A voluntary comprehensive physical examination will be offered toall individuals in both the exposed and primary control groups within one yearof questionnaire administration. The condition for entry into the examinationphase of the study will be the completion of the baseline questionnaire. Inthe event that the primary control does not complete both the questionnaireand the physical examination, a replacement will be selected from the controlset [See Figure 3 and Section F(3)J. Statistical testing will be conducted bya variety of techniques on both questionnaire and examination findings. Atthe time of physical examination, an extensive physical examination, medical

38

history, and review of symptoms will be conducted. A standardized protocolwill be used to insure comparability of data. This will provide cross-reference data to the initial questionnaire and to medical record data, ifretrievable. Specific response verification and bias indicator questions w i l lbe included in this interview as well.

(a) Examination Parameters

A comprehensive physical examination will be conducted onall willing participants. The examination will be structured as outlinedbelow and 1n Section XII and will be performed at the earliest practical timefollowing the completion of the questionnaire. The close sequencing of thesestudy components will limit the development of major symptoms in the intervalbetween the questionnaire and the examination. Examinations will be performedunder contract at a single civilian medical center having dermatologic, neu-rologic and electromyogram/ nerve conduction capabilities. Informed consentforms will be obtained for all procedures. Physicians and technicians willhandle all participants without a knowledge of exposed or control status, andwill conduct the examinations by standardized protocols to minimize vari-ability. Medical students, Interns, and residents will not be allowed toperform these examinations, and specialty trained neurologists and dermatolo-gists will perform the appropriate portions of the examination. An onsitemonitor will Insure that the examination protocol 1s followed. All laboratorytests will be subject to rigid quality control. Laboratory and physicalexamination data will be measured on a continuous scale whenever possible inorder to improve statistical power 1n the analysis.

Under special circumstances, additional testing will be accomp-lished. Karyotyping of the Individual and his family members will be con-sidered if clinical history or physical examination findings are suggestive ofthis need. Most well conducted studies have shown that, when present,chromosomal abnormalities due to TCDD are transient. If on detailed analysisof the baseline examination and questionnaire, reproductive areas are heavilyaffected, routine karyotyplng may be included in the test battery for thefollowup phases of the study. TCDO analysis on blood and urine will beconsidered in the future provided that (1) strong cause and effect relation-ships can be ascribed to Herbicide Orange and (2) high resolution massspectrometry technology achieves 10 femtogram sensitivity with high isomericspecificity. Serum, urine, and semen specimens will be obtained from allparticipants, aliquoted, and preserved at -70°C for possible analysis in thefuture. These serum and/or urine specimens will also be used for analysis ofporphyrin metabolites If analytic techniques make this a feasible diagnositcprocedure. Extensive Immunologlc function analyses will be conducted on arandomly selected group of subjects.

Physical examination and laboratory data will be placed in themember's coded master file for detailed cross-analysis to questionnaire data.Information Identifiable to the subject will not be released without his con-sent in accordance with the Privacy Act. However, 1n accordance with AirForce regulations, active duty flying personnel and active duty air trafficcontrollers found to have conditions which are disqualifying for flying dutywill be temporarily "grounded" pending resolution of the medical condition.

39

Physical Examination Profile

General Physical Examination HemoglobinFBS, 2 Hr Post Prandial HematocrltUrinalysls White Blood Cell CountBUN/Creat1n1ne and DifferentialCholesterol/HDL Platelet CountTrlglycerides RBC IndicesSerum Protein Sedimentation RateElectrophoresls Prothrombln Time

Dermatologlc ExaminationUrine PorphyrinsUrine PorphoblUnogenDelta-aminolevulenlc Add

CPKECGChest X-RayVDRL/FTACortlsol DifferentialThyroid Profile (RIA)Pulmonary Function

StudiesBlood Alcohol

Neuro-Psychiatric ExaminationGeneral Neurologic ExaminationPsychological Battery:

MMPIWAISWRATHalstead-ReitanWechsler Memory Scale SubtestsCornell Index

Reproductive ExaminationLH, FSH, TestosteroneSemen Analysis

Nerve ConductionVelocities

Neoplastlc/Hepatlc ExaminationSGOTSGPTGGTP

Alkaline PhosphataseLDH (Isoenzymes 1f elevated)Hepatitis B Antigens/AntibodiesB1l1rub1n, Total and Direct

Additional Studies (Individuals with abnormal history or exam-ination)KaryotypingHepatitis A Antigens/

AntibodiesAnti-Nuclear AntibodyQuantitative Immunoglobulins

ImmunoelectrophoreslsBilateral profile and full-

face photographsSkin BiopsyAdditional Consultations

as Required

Immunologic studies (conducted on a randomly selected group of subjects)Enumeration of B and T Cells B and T Cell FunctionEnumeration of Monocytes

40

(4) Analysis of Questionnaire and Physical Examination Data

The Questionnaire and Physical Examination will produce data ofthree types: (1) dichotomous, (2) polytomous and (3) continuous.

Dichotomous (e.g., present/absent) rates will be evaluated usingthe tools described above for mortality analysis. For example, the question-naire will provide data concerning the first occurrence of disease states byage, and standardized rates and relative risks may be calculated. The occur-rence of such findings can be related to age, time spent in RVN, exposure, andother variables using logistic models followed by McNemar's test where appro-priate. These tests will examine the presence or absence of group effect andallow assessment of the statistical significance on non-unity relative risks.

Polytomous findings will occur in both questionnaire and physicalexamination responses. As an example consider retinal findings categorizedinto four grades, and studied as a function of age and exposure group as rep-resented in Table 9. In this table the XJJK'S are counts of occurrence.In analyzing tables such as these, techniques as described by Bishop,Fienberg, and Holland (1975) will be used. Specifically, if m^ is theexpected value of x-fj|<, general log-linear models of the form

in mfjk = u + u^i) + u2(j) + u3(k)

+ ul3(1k) + u23(jk) + u123(ijk)

will be used, where u^i) is the effect of RANCH HAND membership alone on cellfrequency, u12(ij) is the effect of an interaction on RANCH HAND membershipwith retinal grade, etc. This model can work with dichotomous as well aspolytomous data. Under appropriate conditions on expected values of entries1n Table 9, the pairing in the study design can be used with the data beingorganized as shown in Table 10. In Table 10, N-JJ is the number of pairssuch that the exposed person has retinal grade i, and the control person hasretinal grade j. Appropriate tests for this setting are indicated by Fleiss(1973) and McCullough (1978).

With regard to continuous variables, the intended method followsCarpenter (1977) who found substantial gains in analysis efficiency by match-ing cases, subsequently employing covar lance analysis to remove non-controlledeffects. The conditional logistic regression model for relative risk,Holford, White and Kelsey (1978), Is also applicable and will be used.

41

Table 9

FORMAT OF CATEGORICAL REPRESENTATION OF RETINAL CHANGES

RANCH HAND PERSONNEL CONTROLS"^x^Age Category

Retinal Category^Xs"x^

1

2

3

4

1 2 3 4

*l'.l X112 X113 X1U

X121 X122 X123- X12»»

X131 X132 X133 X13>»

X V V V

1 tl 1 1 li j? 1 4 3 1l4l4

1 2 3 4

X211 X212 X213 X21"+

X221 X222 X223 X22«f

v y v yA231 A232 A233 A23i*

X V V V2t4l 9l4!? 9^4^ 9l4t4

Table 10

FORMAT OF PAIRING FOR GRADES OF RETINAL FINDINGS

.Control Grade

RANCH HAND Grade"

1

2

3

4

1N U ^ 1 2 "13

N21 N22 N23

N31 N32 N33

(5) Analysis of Fertility/Reproduction^Data. The herbicides underconsideration in this study have been alleged to effect fertility and/orreproductive functioning. An attempt will be made to address these allega-tions by analyzing at least three primary variables: the total number ofconceptions since exposure in RVN, the number of miscarriages in spouses sinceexposure in RVN and, the number of abnormal offspring since exposure in RVN.The interview with current and former spouses will provide much more accurateinformation on fertility and reproductive functioning than if similar datawere obtained from the male subjects themselves. The study questionnaire willprovide the numbers of miscarriages, abnormal offspring and of live births.The sum of the number of miscarriages, still births, and live births willprovide an estimate of the total number of conceptions. If differing divorcerates are found in the RANCH HAND and control groups, this may render theaverage number of years of marriage and the distribution of the years ofmarriage different in the two groups. This will be investigated and adjusted

42

for if need be, either by analyzing total number of conceptions divided by (ornormalized by) the number of years of marriage, or by using a more detailedcovariance analysis. Further, the ratio of the number of miscarriages toadjusted total conceptions will be calculated and compared, as will be theratio of the number of abnormal births and adjusted total conceptions.

In summary, the following statistics relating to fertility willbe calculated and analyzed at the very least:

TOTALCONCEPTIONS = Rive Births + IStill Births + #Miscarriages

NORMALIZEDFERTILITY

INDEX

TOTAL CONCEPTIONS

YEARS OF MARRIAGE

MISCARRIAGEFRACTION

# MISCARRIAGES

TOTAL CONCEPTIONS

ABNORMALITYFRACTION

# ABNORMAL OFFSPRING

TOTAL CONCEPTIONS

F. Follow-up Study

(1) Study Adaptations

Following complete data analysis of the initial mortality andmorbidity studies, adaptive or restrictive health surveys will be developedand administered to all follow-up study subjects three, five, ten, fifteen andtwenty years after the initial questionnaire. Similarly, a condensed physicalexamination profile that will achieve adequate sensitivity and specificity forprospective diagnosis will be developed. The adaptive physical examinationwill be offered to all follow-up participants, and will also be conducted inyears three, five, ten, fifteen, and twenty (see Appendix, Table A-5). Aninterim examination in year three 1s essential in this study because the agegroup under study is approaching that portion of the mortality/Illness inci-dence curve with the steepest slope. A lapse of five years between the firsttwo examinations could easily miss significant development of disease in theintervening years. Ample precedent for Interim examinations can be found inthe Framlngham cardiovascular disease study, and In the follow-up evaluationof West Point graduates being conducted by the Air Force.

43

(2) Entry Criteria

All exposed or control Individuals completing the baselinequestionnaire and physical examination will be entered Into the follow-up;further continuation will depend upon the member's willingness/ability toparticipate 1n additional health surveys and condensed examinations. Specificstudy entry rules are detailed In Table A-6 and Figure A-3 of the Appendix.

(3) Loss to Study; Key Issues

Loss of participants over time adversely affects any epidemiol-ogic study In two ways. Ar the sizes of the study groups decrease, statisti-cal power also declines, and bias 1s Injected Into the study If losses are notrandomly distributed in the study populations. It 1s reasonable to assumethat in this study, losses will be non-random with greater non-complianceamong individuals who perceive their health as "well," since there is lessincentive for this group to continue participation. As shown in Figure 5,such a differential pattern of loss will alter the population, and skew thefrequency distribution curve.

Most previous epidemiologic studies have approached the problemof declining statistical power by beginning the study with multiple controlsper exposed subject, and passively allowing attrition to occur throughout thestudy period. However, this approach does not address the problem of bias.This study will take an active approach to both of these problems by using areplacement concept. As a control is lost to study, a replacement will bechosen from the original set of ten matched controls. The replacement will beselected from the control set, and will have a perception of health similar tothat of the lost control (Figure 6). The replacement strategy will maintainstatistical power and the integrity of the matched design despite loss tostudy in the control group, and will correct anticipated bias while minimizingthe number of required physical examinat'ons.

At the initiation of the follow-up study, loss of an exposed mem-ber will not be cause to cease surveillance of his primary matched control.In the event of a control loss (for reasons other than death), another controlfrom the set will be brought to study (Figure 7), the comprehensive question-naire will be administered, and a baseline physical examination performed.

If a control is noncompliant for one portion of the study and isreplaced by another control, the noncompliant individual will be approached atthe time of subsequent questionnaires and examinations, and encouraged toreenter the study. If he reenters, both he and the replacement will beincluded in the evaluation. Similarly, noncompliant exposed subjects willalso be aggressively recruited for all subsequent study phases.

44

FIGURE 5

en

EFFECT OF NON-RANDOM LOSS TO STUDY IN THE

CONTROL POPULATION

N

WELL *~

HEALTH SPECTRUM

IF CONTROL LOSSES ARE ILL, A SPURIOUS EFFECT!S ATTRIBUTED TO HERBICIDE EXPOSURE.

IF CONTROL LOSSES ARE WELL, A TRUE/VALID HEALTHEFFECT IS DILUTED.

ILL

THE REPLACEMENT STRATEGY

EXPOSEDPRIMARY

CONTROLSLOSSES REPLACEMENTS

MATCHED FOR HEALTH STATUS

(DEAD CONTROLS NOT REPLACED

FIGURE 7 CONTROL REPLACEMENT FOR THE MORBIDITYAND FOLLOW UP STUDIES

EXPOSED CONTROL

YEAR 6 -1

• QUESTIONNAIRE DATA

O RECONSTRUCTED DATA

1000

#• LOSS TO STUDY

PHYSICAL EXAMINATION DATA

For exposed and control individuals who drop out of the study butsubsequently re-enter, medical data for the intervening years will be recon-structed from questionnaire and interview responses. IN ALL CASES OF LOSS-TO-STUDY, INTENSIVE EFFORTS WILL BE MADE TO DETERMINE THE SPECIFIC REASONS FORNON-COMPLIANCE, AND DATA FROM REPLACEMENT CONTROLS WILL BE REVIEWED TO ASSESSCOMPARABILITY WITH THE LOST INDIVIDUALS. Medical record reviews of newentrants will continue throughout the follow-up period.

(4) Study Length

The fpllow-up st-'dy is initially planned for 20 consecutiveyears. Procedures, progress, anJ interim results of the study will be moni-tored by an independent scientific review group, responsible to the Office ofScience and Technology Policy in the White House.

G. Determination of "Disease"

(1) Introduction

Since this study is dealing with an unknown clinical endpolntwith unknown latency, determination of a disease state by statistical methodo-logy is a prime scientific thrust of the investigation. From the literature,chloracne is the only generally accepted chronic disease associated with highexposure to dioxin. The questions of primary interest are: (1) Does a his-tory of chloracne invariably lead to future disease? and (2) In the absence ofchloracne, is there emergence of other attributable diseases? Under a broadconcept of "spectrum of illness", either or both of these conditions are" pos-sible. The clarification of their respective contributions to the naturalhistory of past or of subsequent "disease" is of significant interest.

(2) Discussion

Inferences about a disease state from this study can be derivedfrom several logical approaches. Tho'-.e approaches can be grouped into twocategories: (1) those dealing with symptoms which can be used to construct asymptom complex that may represent disease, and (2) those dealing with physi-cal signs which in themselves rep -esent disease. In the former, one can forma subset of individuals that have symptoms (e.g., infertility) and study themduring the morbidity and follow-up studies. Focusing on the overall patternsof alleged symptoms and categorizing them into a symptom complex may identifythose individuals with a disease syndrome, or those at higher risk of develop-ing disease (e.g., genetic disorders, cancer). In the latter approach, dataon abnormal physical signs (e.g., genetic defects in offspring) and laboratoryresults can be compared between exposed and non-exposed groups in an attemptto again establish the presence or absence of disease. By putting this arrayof data into a logical decision-making scheme, specific relative risks can becalculated in the follow-up study, and specific response patterns can beinferred as shown in Figure 8.

48

FIGUkE >'•

INTERPRETATION OF HORIZONTAL COMPARISONS

OVERT EFFECT

MR * McSR * S C

FR - FC

F R O * " Fpcno v o

F RS* FCs

MORTALITY/SYMPTOM/SIGN REGRESSION ONEXPOSURE

SUBCLINICAL OVER-REPORTING

•i = «c MRS — Q CD •— O f* 0 Dn b n

FR * FC FR

F ^ r rD Q • P C r Dn O l/o It

FRS f c FCS FR

= MC

" sc= FC

S-FC S

S=FCS

SIGN REGRESSION NO REGRESSIONON EXPOSURE ON EXPOSURE SEEN

FR = FRS S R " * " F R S u-SR 1

Again referring to Figure 8, at least three clinical patterns candefined. These patterns are delineated using relative risks (mr/mc, sr/sc,fr/fc etc., between group or "vertical" comparisons, referencing Figure 4) andusing within group ("horizontal" study) comparisons such as regressing symtomsand findings rates against an Index of herbicide exposure, and other compari-sons. Specifically, an overt clinical effect would be marked by: anIncreased mortality rate 1n the RANCH HAND group (mr > me), an Increased rateof symptom formation In the RANCH HAND group (sr > sc), and an Increased rateof objective medical findings 1n the RANCH HAND group as compared to thecontrol group (fr > fc). Further, the occurrence of physical or objectivemedical findings would consistently relate to symptoms 1n the overt case (thatis, frs > fcs and frs > fcs^, anH finally, in the classic Instance, mortality,and symptom and sign formation would- be seen to be increased with increasingherbicide exposure.

A subclinlcal pattern is indicated in the central column ofFigure 8. In this setting, one expects no statistically significant differ-ences in mortality or symptom reporting between the two groups, exposed versuscontrol. However, one expects a consistent predominance of medical signs 1nthe RANCH HAND group with regression of the signs on increasing herbicideexposure.

A pattern strongly suggesting over-reporting 1s presented as theright column of Figure 8. In this setting, there is no difference between thegroups as regards mortality or medical sign incidence; however, more symptomsare reported by the RANCH HAND group. While in this pattern the RANCH HANDsubjects are reporting more symptoms, objective medical finding rates are notconsistent with symptom reporting. When no regression of symptoms on exposurelevel is found, over-reporting is clearly and strongly suggested.

This discussion of response patterns has used regression on anexposure index in a central way. Development of such an index Is discussedbelow. It is noted, however, that a direct index of exposure can be con-founded by other factors such as cellular repair mechanisms or bioaccumulationin adipose tissue with release over time upon weight loss. Use of other fac-tors, such as time since exposure, should help to overcome these confounders.

The strength of any inferences made from these analyses is depen-dent upon the statistical power inherent in the study. In addition, due tothe possibility of latency being a factor in this study, a negative analysisat any time within the study does not categorically imply lack of disease,since sufficient time for emergence may not have passed.

H. Exposure Indices

(1) Exposure Concepts

A major concern in conducting this study is the lack of accurateexposure data. Although most personnel assigned to RANCH HAND squadrons wereundoubtedly exposed to Herbicide Orange and TCDD, the exposures within the

50

group must have varied widely. Exposure to herbicides and TCDO by RANCH HANDpersonnel occurred almost daily. Anecdotal information suggests that many haddirect skin contact which was repetitive over a long period of time (one-yeartour for most individuals). Further, it is also suggested that most RANCHHAND personnel felt that the herbicides employed in the operations were nottoxic to animals and man, and hence, they did not exercise the caution in han-dling these chemicals that is recommended today.

From a historical review of RANCH HAND operations, it appearsmost individuals can be classified into one of three groups based on theirlikely potential for exposure to the herbicides:

(1) Pilots, Co-pilots and Navigators:- low potential(2) Crew Chiefs, Aircraft Mechanic, and

other Support Personnel: moderate potential(3) Console Operators and Flight Engineers: high potential

The "pilot" group received most of their exposure during pre-flight checks as well as during the actual dissemination missions. The crewchief group experienced contact with herbicides during dedrumming and aircraftloading operations, as well as during on-site repair of the aircraft and sprayequipment. The console operator group was exposed while supervising the load-Ing of the aircraft, during ground testing of equipment, and by tank leakageduring dissemination missions.

The available historical records on Operation RANCH HAND indicatethat personnel assigned to the project seldom had a "routine" work schedule orenvironment, thus complicating estimates of the level of herbicide and dioxinexposure. Since actual exposure data (e.g., mg of herbicide/kg body wt) arenot available, an exposure index will be used. The exposure indices will becalculated for each RANCH HAND individual to obtain frequency distribution,and will be calculated by evaluating the known factors that would haveinfluenced exposure. These will include such factors as:

(1) Date of tour with RANCH HAND 1n Vietnam.(2) Number and lengths of tours in Vietnam with RANCH HAND.(3) Number of herbicide dissemination missions (as reflected by

flying hours and air medals).(4) Herbicides employed (records are available that reflect the

amount of each herbicide sprayed each month and year).(5) Crew position.(6) Routes of exposure (the major route of exposure for most

RANCH HAND personnel was probably percutaneous, although exposure through.inhalation may have also been significant).

51

A crude exposure Index which 1s applicable to the entire RANCH HANDcohort is expressed with the following formula:

E, • qj X T,

In this formula, Ej is the calculated exposure for the ilL RANCH HANDmember, q\ is the quantity of TCDD-contain1ng herbicide sprayed from air-craft assigned to the iUl subject's base during his assignment, and T^ 1sthe length of the illl subject's assignment (tour length). However, greatcare must be exercised when applying the above index. For example, the Indexshould be used as an independent regression variable aga'lnst clinical findingsonly within occupational strata, to.'avoid confounding occupational effectswith exposure effects. Different degrees of regression between clinical find-ings and the exposure index can be expected 1n differing occupational groupssince: (a) modes of exposure are likely to be different 1n different occupa-tional categories, (b) sodoeconomic correlates within occupational categorycould confound an herbicide effect, and (c) other exposures which couldsynerglstically or antagonistically interact with TCDO-conta1n1ng herbicidemay be correlated with occupational category.

Another factor which must be considered when applying this crude exposureindex is the problem on confounding a possible herbicide effect with an effectassociated with tour length. Being in a comabt zone is a major psychophysio-logical stress, and time spent in such an area may be significantly associatedwith changes in long term morbidity and/or mortality. This crude exposureindex, when used alone, could result in a positive regression with diseaseincidence or prevalence which is not due to the herbicide exposure. Anapproach that will correct for this potential confounding is to regressobserved medical findings on both EJ and T^ to differentiate the indepen-dent effects of herbicide exposure and combat zone experience.

The values of qi and Tj needed to calculate Ej are generally availablefrom government records. Specifically, tour dates are available from militarypersonnel records, and the quantity of herbicide sprayed is available for theperiod January 1965 through April 1970 from the "HERBS TAPES." These tapesare comprised of computerized data obtained from actual spray missionreports. This material provides the date, base of mission origin, amount andtype of material sprayed (Herbicides Orange, Blue, or White) and location ofthe intended spray target. Estimates of the amount of herbicide sprayed priorto 1965 may be available from procurement records for Herbicides Purple, Pink,and Green, which were sprayed exclusively from Tan San Nhut Air Base from 1962through 1964.

Animal data imply that TCDD is the most toxic component in the herbicides usedin RVN. By using qi, the amount of herbicide sprayed, one is using a vari-able that roughly correlates with TCDD exposure. However, it would be highlydesirable to be able to analyze observed health effects in terms of specificTCDD exposure. The material sprayed from 1965-1970 had significantly lower

52

TCDD contamination then did those herbicides manufactured and purchased priorto 1962 and used from 1962 through 1964, but due to data limitations from ascarcity of Herbicide Purple, Pink, and Green samples, TCDD concentrationprofiles for those chemicals cannot be quantitatively determined. However, itmay be feasible to develop estimates of the degree of contamination based uponthe TCDD concentration from military and manufacturers' data.

As another approach to examining the effect of TCDD itself, one might considerstratifying the exposed cohort by date of assignment 1n Vietnam, expectingthat those assigned earlier were more heavily exposed to TCDD. While it maywell be true that earlier assignees were exposed to higher TCDD concentra-tions, it is unlikely that differences between "early" and "late" assignees,if they occur, can be reliably attributed to TCDD concentration changes, sinceseveral potentially confounding variables exist: '(a) volunteerism among earlyassignees, (b) differing assignment patterns between early and late RANCHHANDers (TOY vs long term pattern) and (c) different RVN living conditions.

It is preferable to use an exposure index which 1s more closely tailored tothe specific individual than the crude index discussed above. While T| issubject specific, qj is a value which refers to all individuals on the baseduring the period of time represented by 1\. A refined index for groundcrew can be expressed as:

Ei = F. X qi X C X T1

•7 :

x where,

f"'. Fj = Average flights per day served by the iitL ground crew member.?v °,1 - Average quantity of herbicide dispensed by flights served by the" . ij-ll ground crew member.

C = Estimated TCDD concentration of the herbicides in use during theiM subject's tour of duty.

TI = Time spent in TVN in days for the ilil ground crew member.

The variable F^ can be estimated by dividing the number of RANCH HANDflights per day by the number of crew chiefs during the time period T-j. Allother variables are estimated as with the crude index.

A refined index is also possible for aircrew members and is expressed as;n fOl1OWS :

E1 = M1 X D1 X q1 X C X P1

* total number of missions flown by the iJill air crew member.= average duration of missions flown by the ilP. air crew member.* average quantity of herbicide dispensed per flight served by

ilil air crew member.= estimated TCt)D concentration of the herbicides 1n use.= a cr-?w position weighting factor.

53

the

As with the refined ground crew Index, this refined aircrew index cannotdirectly calculated in a strictly quantitative sense using available govern,ment records, since records to specifically link missions with particularindividuals are not available to objectively determine MJ and Dj. Heever, reasonably accurate estimates of these parameters may be feasible uslniquestionnaire data. Also air medal awards may allow an Indirect estimate,

The crew position parameter PI must also rely upon estimations. While t|specific crew duties of each subject are known, the differential exposurassociated with the crew positions within the C-123 aircraft were hot detq'/mined during RVN spray missions. The 355th TAS/Spray Branch, RickenbacherOH is presently using the C-123 aircraft, configured with the A/A 45 .Internal Dispenser and attempts to as'sess Pj can be made. Air flow measurement and herbicide simulant deposition studies conducted by Meek are perforduring the course of four C-123 flights. However, difficulties with tfmeasurement equipment limit the validity of the value of the data in an expo-sure index. Further work along these lines could yield a more quantitative,position weighting factor, Pf, for each individual.

Refined ground crew and air crew exposure Indices can be used singly orcombination with the crude exposure Index first presented; however, as wltjthe crude index, confounding must be avoided when the refined indices are us'ejin statistical analyses.

The exposure indices listed above are, of course, only applicable to the RamHand cohort. As mentioned, a positive regression of disease Incidenceprevalence with increasing exposure index will strongly support herb1ci(causation. We do not wish to minimize however the role of RANCH HAND versuscontrol group disease incidence/prevalence differences as Indicators of :)herbicide effect. A major component differentiating the RANCH HANDers frethe controls is the increased residence of RANCH HANDers in the RVN Itself^If within country time does not correlate with disease incidence, RANCH HANpJversus control disease incidence differences may be strongly related tcherbicide. If in-country time is significant as a disease correlate, this Ititself will be valuable information with regard to assessment of the RVflexperience.

54

examina-of some

VI. Special Statistical Considerations

I The previous discussion has outlined, the general statistical approach1 followed by this protocol, and has outlined planned analytical methods and

inferential strategies for the mortality, questionnaire and physical; tion study phases. This section provides an indepth consideratiorj . special statistical study aspects.

$ A. Fa1se Report ing/Mis rep resentat i on

| Since concern for compensation could unconsciously or consciously| influence symptom reporting, and since press reporting itself can stimulate* anxiety-based symptom formation, a discussion of false reporting is iridi-| cated. A data pattern indicating overreporting has already been discussed inf>" Section V. The goal here is to understand the effect of misrepresentation onI estimates of relative risk and the odds ratio. Let S stand for presence of a

symptom, and T denote its absence. This false reporting may be represented as1n Figure 9.

Figure 9

FALSE REPORTING/MISREPRESENTATION

REPORTED STATUS

TRUE STATUS

s r

A

C

B

D

Total

A+B

C+D

A+C B+D

The proportion of correctly classified positives is defined by A/(A+C) and islied the sensitivity of the classification scheme; the proportion of cor-ly classified negatives D/(B+D) 1s called the specificity.

When there 1s non-differential misrepresentation, that is, when theisitivity and the specificity are the same among the exposed and nonexposed,

fehe bias Induced in the estimate of relative risk will be-toward the nulltalue. The situation is summarized by Figure 10.

55