Document 2qXjwoXpox9M1zya40pjDJV5r

Herbicide Exposure and Cancer The case-eontrol study in Kansas reported by Hoar et al' in this issue has iniportonce beyond its population of agricul- ture workers to the highly controversial and ad\wsarial issue of health effects associated with herbicide exposure in Vietnam. The well-designed and carefully executed Kansas study adds substantially to the cumulating body of evidence concerning the followinE question: Does human exposure to phenoxyocetic acid herbicides increase the risk of Mft-tiSUe sarcomos (STS),Hodgkin's disease (HD),and non-Hodgkin's lymphoma (NHL)?The findings among Kansas farmers are in accord with "no" answers to the first two malignancies and a "yes" answer to the third. . Most relevant to the Kansas study are the series of cssecontrol investigations undertaken by Hardell and colleaguesw in m d Sweden in the mid- and lata 19701. Hardell's studies found bighly significant relative risks,of magnitude fivefold and sixfold, for exposure to phenoxydd m,herbicides in Oroocintion with HD.and NHL Among the published epidamiologic studien, many regard the Hardell studws as the strangest in study design and drily an providing the strongest available evidence of an association of herbicide exposure with the OcCuTrence of uncer. However, selection bias, observation bias, and uncontrolled confounding Imm as important points of vulnerability and, in fact, have emerged as the focus of sharp criticisms of the Hardell studies. In addition to critiques in the scienWic literature, Hardell's studies have garnered considerable airing in the harsh give-and-take of the courtroom. Most recently, Hardell provided testimony in Australia to the Royal Commission on the Use and Effect3 of ChemiePl Agents on Awtralian Personnel in Vietnam. The commission's Final Report' stabs: This ahsen- of replication. [I divurr this below) the abwnre of specific outcome (ie. 12 types of mft tissue sarcoma, non-Hndgkin's malignant lymphoma and Hadgkin's Disease). admitted information b i 4 the pruence of rignifiunl confounding futorr. the unreliability of thc exporum dab 8nd the other factors detailed above all indicate that the statistical ~ u t i o r du by Dr. Hardell are suspect. "he Commission cmnot, on the balance of pmhnbility. accept them Y supporting an inference of nusal connection ktween mft tir#rc tormmr. maliiant lymphoma and exposure to phenoxy herbicida With regard to STS, one other case-wntrol study has appeared. namely, that of Smith et al' in New Zealand. Smith's findings conflict with those of Hardell and yield a negative result with an estimated relative rid of 13. Smith's study was designed to have adequate statistical power to detect a relative risk of roughly 3, if such a risk truly existed. What now does the study by Hoar et al add to the evidence? Perhap most n o t e w h y art the Mpative find- ings for STSand HD.As a a u d e tally.this makes a score of two negative (Smith and H o u ) vs one positive (Hardell) study for STS and a tis alt one each negative (Hoar)and positive ( H d l ) study for HD.With regard to NHI, the score stands at tm positiva rhdirr (Hoar .ndHardell) More important and rekvank however, OIC how the Haar and H a d r t u h compprr methodo)oeically. Does the Hoar study batter withstand the critical onslaught that has besieged Hucbll's studits? Of cwm, the Hoar study has undergone only the intensity of J M M t peer review process and has not (at knrt to my knowledge) been subject to the rigors and scrutiny of owrtroom debate. Let me atiempt to probe some methoaolqiul contrasts of the two studies. The Hoar study, like Harddl'r, has the strength of case ascertainment with a popllatioa-hssdtumor mgistry along with independent pathologid review and confirmation of disease. The Hoar study ir mmiderably larger than Har- dell's, with nearly twice the number of of malignant lymphoma and three times thc number of controls Thus, Hoar's study has considembly gruter statistical precision and s t a t i s t i d power to detect Both studies employed popuhtion-based aontmls, using living aontrols for living cases and decead mntmls for d d CWL As such, both studies had to rely in part on next of kin's reporting of expovure status. In contrast to what has surfaced regarding the procedure 1176 Sept 5. 1986-Vd 256.No 9 Erllorob . .- . for rontlwting thc intcrvicws in thr Flardrll e studics, the axAI^ in howover, the llorzr studs docs i n d d PrCtvidc' a 1 `(tar SilJdy involved expcricnced and trained intcnilcwcrs , v l l ~f o ) l ~ c ral dctailed, prcdelcrnrincd intrrvicw pm-1 (Dr Ilcmr. oral communication. April 1966). The intcrviewers knew that the study mnccrned chemical exposure, b u t were urio\v~rethat the study's major hypothesis dealt with much firmer and tighkr body of scirntifie c\-idrnrc than do Lhc Hadcll s t a d i a What supporting or amtradidory evidcnoe exists fmm other epidemiologic studies? None of the several industrial cohort moriality studies of d i o h exposure has resulted in a exposure to phcnoxyacetic acid herbicides. Although the n o d perturbation in the frequency of deaths from thcse Hoar study intended blind intrrviews in regard to the mwers. The one published w h o d study of Agent Orange subject's status as a case or control, infoqption provided exposure in Vietnam, namely, the US Air Forcc Ranch Hand during the conversation often indicated clearly ta the Study,' has to date not de& m y i n n in i d e n c r or interviewer the subject's disease classification. Another mortality from these c a n a m But, the industrial studits advantage of the Hoar study is that the interviewing, generally entail extrewly small m p l e Jzef that almost conducted mainly in 1983, occurred at a time when there was totally lack adoquato s t a t i s t i d poazr ta detect meaningful no particular local media attention to herbicide exposure i n c m in risk Even the b c h H m d Study with its and cancer risk. Although there was considerable national defined cohort of roughly 12M)e x p o d individuals lacks concern at that time with Agent Orange exposure in statistical power to detect an increpssd rid of these Vietnam, the media in Kansas in 1963 did not shine the malignancies. whose incidence r r t i~n the general popula- public spotlight on farmers' herbicide exposure and their tion are low. Thus, the negative fmdings of the published risks of these three malignancies. Thio is very much in cohort studies, though at first blush a m aomborativeof the contrast to the timing of the Hardell studies during a period negative c a s e u ~ n t r o ls t d k , provide little subtantiation of intense media attention in Sweden and Dr Hardell's b u s e of their inherent small sample tizc and limited prominent role in the media with his taking of an ad- statistical power. position. Clearly, the interview methodologyand the climate Severzl proportid mortality studies have appeared in of the Hoar study make information bias a much more "bonus" staten (ic. those states that offend a bonus in terms unlikely explanation for their findings compared with the of a financial incentive to men who enlisted for Vietnam Hardell studies. service). Most notable is the Massachuwtt~study` that Another strength of the Hoar study is the validation of a found nine deaths from SIS among V i veterans simple of self-reportedexposure by a survey of suppliers. No compared with only one urpceteh Int~xestingly,the Massa- differences were found in the corroboration among controls chuactts study lacked N n u e n t numbers 16 investigate and each of the STS. HD.and NHL casts. Of course, not all proportional mortality for NHL (Richard ClnpR MPH,oral subjects had such validation of their exposure statui But, communication. March 1986). In mntrast to m n t r o l what was done suggests the unlikely event of large differ- studies. proportional mortPlity s t u d i e have even W t e r ences between cases and controls in their ururzcy of vulnerabiliw to forces of dsb'on. obsemtioq and con- reporting exposure. founding b i u e s Furthermore.these state 6- have three The Hoar study considered and reasonably ruled out additional rcriw limitotionn: (1) the quality md aaxlracy confounding effects of exposure to other herbicides, inseeti- of the dath aertiTkate dirgwdr of these malignancies (2) cider, fungicides, nonfarming exposures, and both the known the effects ofm;sotiOn resulting from the sizable numbers and alleged risk factors for each of these three rnaligmndea Of Vcb- W h q 6 U b . c r l u c a t their BditUy S C r V k , Another strength of the Hoar study is the consistent e m i p i d frmn thc state md thus do not appear on the 41. . J de-response relationship for NHL pacarding to the nature, strtc's death tape%and (3) tbe fact that mow of these inknsity, and duration of exposure. The specificity of the studies determineddeundents'berbiddQ srp~aucbut dealt findings of an association with one cancer and no umeiptioo only with their lyzyjcc im Viihvm Tbm rhrtever the 9 with the other two is another relative strength of the H findings of the proportiolrpl mortality m v e s t i h they study rompared with the Hardell studies. in which th add Little if m y pubsturtive d y t i c midem to tbe results were an across-the-boards elevated relative rtk for fundamental q d o n d eulor risk as a result of berbicide each of the three evlcera Limitations of the Hoar study include those inherent with exporurr Is mom prrtinent d e n e on these mncsz ri$u in the casccontrol studies that rely on subjects' and next of kin's offing? Ye! To my knowluip two additiolul a u e c b n b l r e d l of exposure skitus. Faulty recall leads to misdassifiu- studies of t h e CMOM arc under my, sreh iovolviag a tion. If this occurred equally among cases and mntrolg the considerably lugw number of osea than my ofthe studies study would underestimate and thus pauibly miss a true presently at hand, including thestudy by Hoar- Both studies increase in risk If i t occurred differentially, this a d d lead in prograr involve herbicide rxpaarrr in Vietnam as to bias. As indicated above, however, bias due to differential a s s s s d by the Environmental Support G m p d the Armed recall of cass and controls seems unlikely in the Hoar Forces with tk use of avahble militpry w ~ r d son h o p study. locations in anjunction with rum& of berbiude spnyiqp Another limitation characteristic of observational epidc- The first of these studies, tentatively achcduld for a m p l s miologic studies is the possibility of unaontrolled confound- tion in the summer of 1986, is a joint effort of tbc Veterans ing, particularly with diseases whose epidemiolm is A d m i n k t r d o n and the Armsd Forca Inrtitaalc of Patbob 1unclear. This, of course, remains a possibility and to some gy (Han K a n g PhD,04ummuniation, M u c h 1986). It extent constitutes the frontiers of our know)edgc of the involves ST!3 only and en- nearly 300 antimad cass etiology of these three malignanciea. Maertu'ned through the Armed Forces Institute of Pathole Also, one u n n o t rule out entirely the possibility of c h w a gy and approximately 600 mntrols. The study was designed as an explanation for the findings. to have gmd statistjd power lo detect a twofold inaease in M A . Sept 5.1986- Vd 256.No 9 fmamb l l n . .- * .-_- b- rr&w risk. The second study is the Seleetcd-Cmcers . - component of Epidemiologic the massive Centers Studies of the Mealth for Disease of Vietnam Control's Veterans. The Ccnters for Disease Control'o selected cancers sludy, scheduled for completion in the fall of 1989. entails enroll- ment of approximately 400 incident cases of STS and 1400 malignant lymphomas (HDand NHL)during the pcriod Dee 1. 1984, through Nov 30. 1988 (Edward Bmnn. MD, oral communication,March 1986).Case ascerbinment is through six population-based cancer registries in t)le United States that, interestingly, includes the registry in Kansas The studies by the Centart for Disease Cantrol have been designed to have good statistical power (namely, 60% or higher) to d e w a true twofold increase, in relative risk for each of these three cancers, as well as for other "selected" cancers (ie, primary liver, nasal and nasophatyngeal). Where does this leave us now in regard to herbicide exposure and cancer risk? The Hoar study in Kansas, in conjunction with the Smith study is New kaland, tends to shift our concern toward a somewhat lesser interest in STS and a greater interest in NHL. I note that in Hardell's study of malignant lymphoma,' only one short paragraph a t the end of the "Results"section indiented that the elevated risk was the same for HD and NHL; otherwise, Hardell's analysis considered the two malignancies together a s "ma- lignant lymphoma" Hoar rt a1 hare now focused attention specifically on h31L risk. I ..- .. .. -E