Document ByrzkqVw08Jpmzy0rwB0Lq8rE

G. M. H.Swam I RreCitcUl;8 May 199s f Acccptad: 27 Scpmibtr 1995 Sir: exposue data, from the desctipion ofthe cases in Table 3. Kecenrly an article was published de.scrihingthe incidence rrf rem1 cell tumors in a cohwf of ctudbaml w0rk.w~ exposed to trichlomerllylene (Henschler et al. 1995). This anicie describes a relrospectivc cohort study of 169 worken exposed to trichlo~nethylene(TRI), 'here m several reharks to be mule about lhis study. &I, the original exposed cohort cOiraiSied of 183 male ~wkers.but 14 subjccts were exclude(! because tbey Mused or w e r e too ill io parlicipate or could not be traced by the closing day of 1be swdy. Subjects who are lost Lo Cdlow-up and whose vital swtus cannot be dewmined at ihe end-date of the follow-up shmdd r t c w be excluded b m the analysis ofa retrospectf\vcohort study, but should rcmain in the cohorl until the dale that they were losf from foliuw-up. A similar niethodolopical emor is indicated by Table I (dexripib of rhc TRLexposed and control co- h s t . In th~st a b it is slated rkal ihe mcdian time of entry furthe exposed groupis 1956,with a ininimum of 1918 and maximum of 1972. This means that subjects were alreedy RgUded as being at risk in 1918, which conflicts wtlh the cohcvl ddiniiion, which demands exposure b h v e place between 1956 and 1975. Probably 1918 is the d m of bitth. However. the pem-years at risk geaerslion does not start at Mnh. but after 1 year of starting from t January 1956. Saondly, no exposure information is available. The s- m ~ spyccukte that the exposures ~ l u shl ave hi&, but m y present no evidence, except some C m ofacute toxicity. Thesc cases occurred during q a i r a maintenance and s t m n~ot bc secn as representwive fa gtmal working coaditions. AMW& there are no only two of the nenal cell cancer patients held jobs asso- cialed with a high degree of txposure (job in board niachine area). Thimtly, far tlw workas who had died the causes o death were obtained fm lioepilal medicxi recod$or frrwn the physician who lasf mu4the patient. This infmation was comprrred LO t~mnlityrates of the general tnak populaiim. The inforrrmrion fnnn the hofpital records and treating physicians may be of a beuerqtmIiiy, which makes II comparison with the general population mortality rates incorrect. Fourthly. the cancer incidence section of the paper is very peculiar. in a normal rcvospective cancer incidence study one expects to find thc usual range of cancer lypes, including some lung cancers. digestive tract cancers, skin cancers eic. However, in thib c w e r incidence study only renal cancers we reponed, From this observation,i t can be deduced that no general inventory of cancer incidence was Pcrformcd. in a cmiSiSren1 rqxducib)e fashion. simlw for both cohom. It is MI clear how the CBBC~of renal cancer were detected. lYie only bgical explanation is Ihw the cases of rem1 cancer had already been observed prior 10 the initiation or the lulual study and even worse. Ihal lhc cases of =MI c u m r trisgerea the conduct of Ihe study. If this is true, then we are merely dealing with a cluster inwstiga- t*m,dressed up with some sdditional data WllectioPl to mimic a poorly conducted retrogpediverollort study. In me care of a prior clusu of renal c a m , foHbwcd by a poorly conduded mdy, it L'OIRESas M) surprlac tbr moly: cases of rand cancer were found ihan expacoed. In this sectin%.it is incvilwbk IOfind an increased incidence of renal cancer. if there wa6 n priorcluster of rend EII(IC~Cin the cardboard workcrs that triggered ltw epidemiological invewigrtion, this should have been clearly described in the report. because it purs ibc ?Its in n completely dirrctent per- spxtw. The omissiondlhils i n f m i i o n In thc article is B .sencm violation of epideinkilogic mvendons. The occurrc~lccof dimme is m evenly distributed llc~urrxik total population. Cham: is an impatant factor. Iticrc will always be cluslers of diseases. If these fill in a 10:25 DUW. UNIU. E D CTR LIBWIRY 006 I28 ccmin exposed group, t h y m y appear to be causally tided with the exposure, but in rcaliiy it is just a coinckknce.. Egidemiologists agree thal rhc investigation of a clusler has a limited significance forthe detection of risk factors for diseases. R o l h (1990) d e r s IO this phenomenon as the Texas shaiphoow p r d u r e . The Texas sharpshooter fim fires some shols pnd then dtpws the &get where the shots have hit thc wdl. Ihe uneven dislributMn of diseases in humaa popula- tions is clearly demonstrated in this study. The only mtistieally significant finding of SMR analysis in this ~ study (Table 5 ) is a nearly tasfdd iwmase of brain tumor monality in the non-exposed group. Tbe three deaths from B brain tumor are about half as likely lo occur as the two d w h s from kidney cancer. Finally, Henschler downplays the mkvaocc of the negJive twospecrive cohort mortality Mwhes with ihe argu- inent chat these srudies may have missed a kidney cancer risk, Lmmse kidney cancer has a good survival rille. Cases of kidney m e r would no1 be picked up in mortdity sludics. lhis argument is DO! meet. Ttre survival rate is indimIIy relevant fw dewbiliry in mortality studies. If a disease has a relaiivcly htgb incidence, despite the good prognosis, changes in ihe moneIiiy rates will be detected in mortality studies. The background mortalig,rate is crucial for the statislical powei ofa marzality study, givcn a certain relathe r i ~ k .The age adjusted niomlity rates for kidoey cancer in the general population are in l e same -1 magnitudeas far bmin tumor mortality or leukemia (I;ich cancer mortality = 5.8 per 1OOM10; brain tumor monalii 5.1 pcr 1OOooO; kukemia = 5.S per loOOO0: CBS 199: Givm a -in relative risk, a @dmtmortality s t q jua as capable of detecting a kidney cancer risk as a br lumoc risk or a ledcemia risk. It i s quite clear that the study presented by Htnscblc~ seriously biased by the a priori presence OF B clusterofre1 cell tumors. By withhddiog information 011 the a pri, p.eseacc of a cluster OC rwtal cell cancer, the auh i a c o m ~ l l yprettnd to haw caniad out B proper cohc stady, which is not the case. It is obvious Lhol this styl cannot be used in suppat of a causal association betam exposure io TRl and renal cell cancer.