Document 4Jkjw1XyoKQ6nqyrdwLmv7xm1

JOEM* Volume 38, Number 3, March 1996 239 uanor offpre fn flip F.Hitor /,-> p;GT!r:"- T-i'5 ivATL'filA.L !v!Af i;<_ tGljKJna in 1988, because 1950 was not .ero^xo me N-Y-66?YR1GHT LAW (title i7u.s. cod^ worker,s flKt asbestos exposure. - Therefore Morgan's second assumption Readers axe invited to submit letters for publication in this department Submit them to: The Editor, Journal of Occupational and Environmental Medicine, PO Box 370, Bryn Mawr, PA 19010. Letters should bd typewritten and double spaced and should be designated "For Publication. " >s unrelated to the PSS model; it is simply an arbitrary conjecture. Morgan's third assumption is that the same values of b and k in the PSS formula apply to all of an individual worker's asbestos exposures, in deter mining the causal probability. Bat Peto MesothBlioma: An Unwarranted Causal Model ticular causal stage, the premise is sim ply wrong. Evidence that this assump tion is wrong exists in the literature.4 et al note that the value of b varies from study to study and may be affected by fiber type and the intensity of worker To the Editor: Robert Morgan has proposed a way to calculate the proba bility that exposure to asbestos in a particular year was the cause of an in dividual's mesothelioma, i He claims his model follows from a formula of Peto, Seidmail, and Selikoff (PSS), who fit observed mesothelioma death rates.2 This work has been subsequently cited uncritically as using the PSS for Second, Morgan takes the probabil ity that any particular exposure caused a worker's mesothelioma / years later to be proportional to the incidence of mesothelioma given by the PSS equation n = bfr in a population an equivalent number Of years after first exposure. To understand the error in this assumption, refer to the example that Morgan cites of a former insulation exposure. Thus, Morgan's third assump tion directly contradicts the best avail able information. According to Peto et al, "After exposures for up to 10 years, the risk [determined by the constant factor b] is likely to be roughly proportional to duration."5 Likewise, k may vary within a range that Peto et al estimate. All three of Morgan's assumptions bias the outcomes in the direction of mula to determine the source of expo worker diagnosed with mesothelioma apportioning causal probability toward sure leading to mesothelioma^ and--as Morgan recommends--has been used in 1988, after being exposed to asbestos each year between 1942 and 1944 and earlier years. For instance, taking b to be constant for each worker over all to try to apportion legal liability. In again in 1950. The PSS formula ap years of exposure (contrary to the best some cases, Morgan's model shifts the responsibility for damages to asbestos- proximates the incidence of mesothe lioma among large samples of workers available information) weights earlier years more heavily. Just to explore the exposed workers away from currently who were first exposed 46 years, 45 consequences, we recalculated the operating companies to bankrupt com years, 44 years, or 38 years earlier. probabilities for the worker in Mor panies, depriving workers of compen The members of this population may or gan's example under two (equally arbi sation for illness. But the proposed may not have been exposed subsequent trary) alternative assumptions--first, causal model is scientifically flawed, to their first exposures. Thus, the PSS we let b change from 1944 to 1950 and does not in fact follow from the formula might validly be used to esti within the range discussed by Peto et work of Peto et al. mate the frequency of observed me al, and second, we let both b and k Three questionable assumptions are sotheliomas in 1988, 1987, 1986, or change within the given ranges. The tacitly made in obtaining Morgan's 1980, given that the first exposure of a probability assigned to the year 1950 is model from the PSS formula. Different, population was in 1942. But there is no .156 under Morgan's assumption, .375 equally arbitrary assumptions yield justification at all for correlating this . under our first alternative, and .801 strikingly different conclusions. First, it data with the probability that a particu under our second alternative. But the is assumed that each worker's mesothe lar year of exposure to asbestos, say alternate models are no more or less ar lioma has exactly one cause that hap 1950, caused the worker's mesothe- bitrary than Morgan's. pened in just one year---ruling out cumulative or interactive causes. Be cause we do not know the exact mecha TABLE 1 nism for the development of cancer, the Causal Probabilities attribution of cause to a single exposure Exposure Morgan's Mode! Alternate Model 1* Alternate Model is open to question. Peto et al (who do not make this assumption} discuss vari ous models of carcinogenesis, includ ing multistage models. If, for instance, two stages of exposure to asbestos con 1942 1943 1944 1950 .303 .281 .260 .156 .225 .208 .192 .375 .071 .066 .062 ' .801 tribute to the development of cancer, or a higher concentration of fibers is necessaiy for a cell to go through a par 'Alternate model, in which just 1 oi 3 arbitrary assumptions is changed. This model sets k = 3.5 and b 1.5B x 10- from 1942 to 1944, and k = 3,5 and b = 5.15 x 10-s In 1950. IThis model sals b = 1.58 x id* and k - 3.0 from 1942 to 1944, and b = 5.15 k 10- and k = 3.5 in 1950. All values are within the range suggested by Peto et al. 240 Letters to the Editor Many different causal models may be consistent with an observed set of empirical data, but in general, causality cannot be determined merely from cor relations or patterns in data, without controlled experimentation. The fact that a specific causal model is consis tent with a particular data set does not demonstrate that it correctly attributes cause, nor exclude other models that also fit the data. But a scientifically vi able causal model normally generates some predictions (under the assumption that it is correct) that can at least be compared with the empirical data to see whether the data are consistent with it or not- But Morgan's causal model, even if assumed to be true, generates no predictions at all that can be compared with the data on mesothelioma death rates or with the PSS formula. This in itself demonstrates that the model is un related to the PSS formula, and has no scientific claim to validity. Acknowledgment We thank C.J- Feltz, Division of Statis tics, Northern Illinois University, for helpful comments on art earlier draft of this letter. . David S. Egilman, MD, MPH South Shore Health Center Braintree, MA 02184 Andrea S. Goldin, BA Department ofBiostatistics Harvard School ofPublic Health Boston, MA 02115 Gerald A. Goldin, PhD Department ofMathematics Rutgers University New Brunswick, NJ 08903 References ). Morgan RW. Whodunit? Liability for mesothelioma cases. J Occup Med. 1991; 33:936-957. 2. Peto J, Seidman M, Selikoff 11. Mesothe lioma mortality in asbestos workers: impli cations for models of carcinogenesis and risk assessment. BtJ Cancer. 1982;45:124-135. 3. Lamphear BP, Buncher CR. Latent period for malignant mesothelioma of occupational origin. J Occup Med. 1992;34:718-721. 4. Mark EJ, Shin DH, Asbestos and histogene sis of lung carcinoma. Semin Diagn Pathol. 1992;9:110-116. 5. Peto J, Seidman M, Selikoff U. Mesothe lioma mortality in asbestos workers: impli cations for models of carcinogenesis and risk assessment. Br J Cancer. 1982;45:132. The Author Replies: Thank you for the opportunity of replying to the letter of Egilman et al, commenting on my paper published over 4 years ago.! De spite their deliberately political and inflammatory prose, 1 will respond to their scientific comments. First, Egilman et al claim that my paper, "...does not in fact follow from the work of Peto et ah" This is nonsense, as the Peto et al paper was the direct stimulus for my paper, and I would have been a plagiarist had I claimed credit for the mathematical model on which my paper was based. Readers familiar with the Peto et al model will immediately recognize the link. 1 agree with the authors that my method assumes that exposure in any 1 year is sufficient to cause mesothe lioma. This assumption is made on the basis of numerous papers that demon strate that exposures of less than 1 year, when of sufficient intensity, are capable of causing mesothelioma. I reject the criticism that my calcu lation of probability is simply arbitrary conjecture; my earlier publication speaks for itself and the "arbitrary conjecture" actually reflects simple probabilities. Although the example I provided was fictional, it is common practice to present fictional sample data to provide a basis for calculations. The h and k values were not arbitrary but were taken from the paper by Peto et al. I agree with the authors that my paper does not direct the reader as to how to weight varying fiber types and doses. This is an instance in which the user of the model may decide to either vary the exponent used, or apply some weighting factor applied from knowl edge of doses or fiber type. In their table, Egilman et ai generate a set of numbers by capriciously chang ing the b and k values for different times, with no data to support these changes. I will leave the readers to judge this "fun with numbers" approach. Finally, to answer their last criti cism, I make no claims that the Mor gan model, as they call it, will predict mesothelioma death rates. 1 do believe that the precursor Peto model has been shown to be an accurate predictor. By the way, I prefer to consider my method as an application of the Peto et al model, not as a new "Morgan model," however flattering that may be. Robert W, Morgan, MD Environmental Health Strategies, Inc. Redwood City, CA 94065 References I. Morgan RW. Whodunit? Liability for mesothelioma cases. J Occup Med 1991; 33:956-957. Students' Perception of Occupational Medicine To the Editor: I have been intrigued by the paper of Graber and colleagues1 that reports the view of academic deans about environmental health curricula in US medical schools. This letter is in tended to contribute to the debate about the current low profile of environmen tal and occupational medicine in med ical schools by presenting some recent observations from amongst medical students in Italy. Until 6 years ago, the course of Oc cupational Medicine (OM) was not compulsory for medical students. In the past, no more than 5% of students had an elective course in OM, and their attendance at the course was primarily because of personal interest in the discipline. The recent introduction of OM in the curriculum as a mandaiory course might have caused uncertainty amongst the students as to the clear and correct identification of the disci pline and its role in the formation of the profession.2 This consideration led me to carry out a survey to investigate students' perception of the discipline in the Med ical School of the University of Mod ena in the academic year 1994-1995. JO! Fift fiftl quf cor per wa. ar aw wc the foi abop sic pr be M in re th tb ai te IT .k K ic d b v .1 s l