Document 3QKER8Vp3aG1b6OByNG1LjZQD

Scand J Work epidemiologic a Norway. II. -1074. aoe of primary >81:6-16. study of vinyl American Journal of Industrial Medicine 24:251-256 (1993) LETTER TO THE EDITOR Diagnostic Bias in Occupational Epidemiologic Studies: An Example Based on the Vinyl Chloride Literature Key words: vinyl chloride, brain cancer, emphysema, chronic obstructive pulmonary disease, diagnostic bias INTRODUCTION Dr. Shah's Letter [1993] raised several issues of potential diagnostic bias on the brain cancer and emphysema findings in our recent paper of vinyl chloride workers [Wongetal., 1991]. These are important and pertinent questions not only in our study of vinyl chloride exposed workers, but also in occupational epidemiologic studies in general. BRAIN CANCER The first issue was the potential diagnostic bias of brain cancer in occupational studies. As Dr. Shah correctly pointed out in his letter, there are reports in the literature documenting the potential bias resulting from the more complete reporting and/or diagnoses of brain tumors in employees of large corporations than in the general population. Greenwald et al. [1981] reported a significantly higher frequency of brain scans (61.1% vs. 30.0%) and pneumoencephalograms (35.2% vs. 17.1%) than in other brain tumor patients in the same state. They concluded that the "diag nostic sensitivity bias" would appear to pertain to conditions that are difficult to diagnose, such as brain tumors, and to industries where workers have such employee benefits as health insurance and high-quality employee medical services with referral and follow-up. Thus, the apparent excess of brain tumors may have resulted from the diagnostic sensitivity bias arising from the more complete medical evaluati n of employees of large corporations. Diagnostic sensitivity bias in brain has also been discussed by other investigators [Wong et al., 1986; Wong and Raabe, 1989]. The participants in our vinyl chloride study were employees of major corpora tions in the chemical industry, and were covered by comprehensive medical care Address reprint requests to Dr. O. Wong, Applied Health Sciences, Inc., 181 Second Avenue, Suite 628, P.O. Box 2078, San Mateo, CA 94401. Accepted for publication December 15, 1992. 1993 WUey-Llss, Inc. R&S 143232 252 Wong and Wtaorton TABLE I. Cohort Studies of Worker* Exposed to Vinyl Chloride* Author* and location Cohort size Brain cancer SMR Byren et al. [1976], Sweden Hagmar et al. [1990], Sweden Smulevich et al. [1988], USSR Cooper [1981], USA Waxweiller et al. [1976], USA Wu et al. [1989], USA 771 2,031 3232 10,173 1294 4,835 2/0.33 = 6.12 s 6/2-60 = 2.29 ns 4/2.61 = 1.53 ns 12/5.90 *= 2.03 t 3/0.90 * 3.29 ns 15/9.20 - 1.62 ns i > significant at the 0.05 level; ns " nonsignificant at the 0.05 level. programs. Therefore, our finding of a brain cancer excess could very well have been subjected to this diagnostic sensitivity bias. In our 1991 paper, although we reported a statistical significant increase of brain cancer, we did not discuss our finding in relation to findings reported by other investigators; nor did we conclude that the observed brain cancer excess was related to vinyl chloride exposure. Although Dr. Shah did not raise the question explicitly in bis letter, it seems to us that the most important question is "does exposure to vinyl chloride increase the risk of brain cancer?" The answer to this question should be based not only on our study but also on all other pertinent studies in the literature. Prior to the publication of our 1991 paper, there were a number of cohort studies of workers exposed to vinyl chloride reporting an increased risk of brain cancer (Table I)' Based on a cursory reading of the literature (as Table I suggests), one may be led to believe that there are consistent reports of an increased risk of brain cancer among vinyl chloride workers in this country. In reality, all the American cohort studies were based on the same population. The Tabershaw and Gaffey [1974], the Cooper [1981], and the Wong et al. [1991] reports were all based on the same cohort study sponsored by the Chemical Manufacturers Association (CMA). The workers studied by Waxweiller et al. [1976] and Wu et al. [1989] were employees at plants which also participated in the CMA study. Thus, those who are not familiar with the history of the CMA study can easily be confused and led to assume that there had been more than one American study reporting a brain cancer excess. Essentially the same excess might be counted more than once. To evaluate the relationship between vinyl chloride and brain cancer in the CMA cohort, we must examine, among other criteria, the strength of association [Hill, 1965]. The overall brain cancer standardized mortality ratio (SMR) reported by Wong et al. [1991] was 1.80 (23/12.76), only a modest increase [e.g., see Monson, 1990]. Furthermore, the excess was only marginally significant at the 0.05 level (lower 95% confidence limit of 1.14). Even if the diagnostic sensitivity bias ac counted for only 15% of the excess, the increase would no longer be statistically significant. By way of comparison, the SMR for cancer of the liver, which has been demonstrated to be associated with vinyl chloride exposure, in our study was 6.41 (lower 95% confidence limit of 4.50). With respect to length of exposure, there was no clear upward trend for brain cancer in our study, especially when compared to the striking upward trend observed for cancer of the liver. Perhaps the most remarkable difference between brain cancel and liver cancer (the latter was clearly related to vinyl chloride) is the analysis basec R&S 143233 R is is is ts 11 have been ease of brain ed by other was related explicitly in -lire to vinyl mn should be literature, studies nain cancer one may be wain cancer ican cohort [1974], the iame cohort "he workers es at plants iar with the re had been ly the same ncer in the association reported by * Monson, 0.05 level ty bias acitatistically h has been / was 6.41 ud wfor brain Adobserved vBrain cancer lysis based Diagnostic Bias 253 on time of first exposure. For liver cancer, the excess was most evident among employees first exposed before 1950, when the exposures were higher than those in later years. For brain cancer, the reverse was true: significant excess was reported only among those fust exposed after 1960. This observation argues against a en4 role of vinyl chloride. As we stated in our 1991 paper, the brain cancer excess in the CMA cohort occurred at primarily two plants. Wu et al. [1989] conducted a detailed analysis f both brain cancer and liver cancer at these two plants, and found that the increased brain cancer mortality was not associated with vinyl chloride exposure. On the other hand, these authors demonstrated a significant association between liver cancer and vinyl chloride exposure. Thus, no causal link between brain cancer and vinyl chloride exposure was found in the CMA cohort. This conclusion is supported by other studies. Recently, collaborators from four European countries (Italy, Norway, Sweden, and the United Kingdom) in a cohort of 12,706 vinyl chloride workers reported a brain cancer SMR of 1.07 based n 14 observed deaths [Simonato et al., 1991]. This collaborative study included the two previous small Swedish studies [Byren et al., 1976; Hagmar et al., 1990], which reported an increase of brain cancer. In addition, this study also included updates of studies which had not reported an excess of brain cancer previously [Jones et al., 1988; Pirastu et al., 1990; Storetvedt Heldaas et al., 1984]. Furthermore, a small Canadian study also reported no excess of brain cancer [Thdriault and Allard, 1981]. Therefore, studies from both Europe and Canada support our interpretation of die CMA study in this country that the observed brain cancer excess was not related to vinyl chloride exposure. Sir Richard Doll [1988] in a recent review combined the data from four major studies from the United States, United Kingdom, Canada, and Italy, and reported a small nonsignificant increased risk of brain cancer (29 observed vs. 19.54 expected; summary SMR = 1.48). With the updated International Agency for Research on Cancer (IARC) collaborative study, which includes both the United Kingdom and die Italian studies that Doll relied on, the summary brain cancer SMR will be even lower. In Doll's combined data, our study was the only one reporting a statistically signif icant brain cancer excess. We conclude that our finding of an excess of brain cancer among U.S. vinyl chloride workers reported earlier was not likely related to the chemical. The most likely explanation is that the excess was due either to chance [Doll, 1988] or to the diagnostic sensitivity bias discussed above [Greenwald et al., 1981]. EMPHYSEMA/CHRONIC OBSTRUCTIVE PULMONARY DISEASE (COPD) \ Dr. Shah's second concern about the excess of emphysema/chronic obstructive pulmonary disease (COPD) brings out important issues in evaluating chronic airway obstructive disorders. In our 1991 paper, we were concerned that the term "emphy sema" was too narrow a diagnostic tom. The diagnostic term more frequently used in the United States is "chronic obstructive pulmonary disease." Emphysema and COPD are two diagnostic terms for chronic airway obstructive disorders and do not represent different conditions. In his review Doll [1988] noted: R&S 143234 254 Wong and Whorton TABLE II. Significant Mortality Exccm of Emphysema/COPD Accompanied by Significant Mortality Deficit of Other Noomalignant Respiratory Disease* Disease category Nonmalignant respiratory disease Emphysema/COPD Other noomalignant respiratory disease Observed 70 41 29 Expected 87.61 22.83 64.78 SMR 0.79 1.79 s 0.45 s *t " significant at the 0.05 level. International comparisons of chronic nonmalignant respiratory diseases are complicated by the usage of different terms to describe what is now agreed is best called chronic obstructive lung (or pulmonary) disease, but which in the past tended to be called emphysema in the U.S. and chronic bronchitis in the United Kingdom. It must, therefore, be presumed that the two categories of "emphysema" and "chronic bronchitis" used respec tively in the two large national studies were meant to describe the same thing. Thus, in accordance with Doll's comments, one must include COPD with either chronic bronchitis or emphysema if one is to include all of the diagnostic terms for apparently the same condition in a review or comparison of world literature on vinyl chloride. The Seventh Revision of die International Classification of Diseases (ICD7), which was used in the original as well as the subsequent updates of the CMA study, does not recognize that these different diagnostic terms for chronic obstructive airway disease describe the same generic condition. In ICD7, the code for chronic bronchitis is 502, whereas the codes for emphysema and COPD are found within 527 ("other diseases of the lung and pleural cavity"). Emphysema is coded 527.1 and COPD is coded 527.2. On the other hand, chronic bronchitis with emphysema is coded 502.0. In an historical mortality study, the codes for various subgroups of nonmalignant respiratory disease are further complicated by the changes in nomenclature over time. In our study we found an overall mortality deficit of nonmalignant respiratory disease (SMR = 0.79), consistent with other studies reviewed by Doll [1988] and the most recent IARC collaborative study [Simonato et al., 1991]. However, in our study, we found a significant mortality excess of emphysema/COPD, accompanied by a significant mortality deficit of other nonmalignant respiratory disease (see Ta ble II). As Doll [1988] pointed out in his comments on the above result in our study, the potential for diagnostic misclassification between emphysema/COPD and other non malignant respiratory disease in some patients was certainly possible. This diagnostic problem of COPD based on death certificates in epidemiologic studies has also been discussed in detail recently by Selikoff [1992], In addition to this potential diagnostic misclassification, we did not observe any upward trend in emphysema/COPD mortality by length of exposure to vinyl chloride; in fact, we observed an inverse relationship. Thus, this lack of a positive doseresponse relationship argues against the interpretation that the excess was solely occupational in origin. R&S 143235 eases s now je, but bionic sat the especsame ) with either ic terms for ore on vinyl (ICD7), ZMA study, ctive airway ic bronchitis 527 ("other ndCOPDis oded 502.0. onmalignant e vertime. t respiratory 988] and the sver, in our ccompanied ase (see Ta- ur study, the d other nonis diagnostic as also been bserve any lyl chloride; sitive dose- was solely Diagnostic Bias 255 In our 1991 paper, we stated that we could not explain the emphysema/COPD excess, due to the limited exposure data and, more importantly, the lade of smoking data. Our primary objective in reporting the emphysema/COPD result was to alert other investigators. Clearly, one must include all of the diagnoses used in that par ticular country to make certain that the comparisons are similar. For example, ifJones et al. [1988] had included ICD 527 with ICD 502 in their U.K. study, they might have found a different result. The same is true for the von Greiser et al. [1982] study in Germany cited by Doll [1988]. In summary, although we agree that the excess of emphysema/COPD reported in our study could have been due in part to diagnostic misclassification, we do not believe that the issue of chronic obstructive airway disease in relation to vinyl chlo ride exposure has been examined in sufficient detail. This issue can be assessed best by a comprehensive review of the world literature of epidemiologic studies of vinyl chloride workers, a review that includes the gamut of terminology for chronic airway disorders. Otto Wong, SCO, FACE Applied Health Sciences San Mateo, California M. Donald Whorton, MD, MPH ENSR Health Sciences Alameda, California REFERENCES Byrcn D, Engholm G, Englund A. Westerholm P (1976): Morality and ctmar motbidity in i group of Swedish VCM and PVC production workers. Environ Health Perspect 17:167-170. Cooper WC (1981): Epidemiologic study of vinyl chloride workers: Mortality through December 31, 1972. Environ Health Perspect 41:101-106. Doll R (1988): Effects of exposure to vinyl chloride; an assessment of the evidence. Scand J Work Environ Health 14:61-78. Greenwald P, Friendlander BR, Lawrence CE, Hearne T, Earie K (1981): Diagnostic sensitivity bias--an epidemiologic explanation for an apparent brain tumor excess. J Occup Med 23:690-694. Hagmar L, Alesson B, Nielsen J, Andersson C, Linden K, Attewell R, Mollcr T (1990): Mortality and cancer morbidity in workers exposed to low levels of vinyl chloride monomer at a polyvinyl chloride processing plant. Am J Ind Med 17:553-565. Hill AB (1965): The environment and health: Association or causation? Proc R Soc Med 58:295-300. Jones RD, Smith DM, Thomas PG (1988): A mortality study of vinyl chloride monomer' workers employed in the United Kingdom in 1940-1974. Scand J Work Environ Health 14:153-160. Monson RR (1990): "Occupational Epidemiology," 2nd edition. Boca Raton, FL: CRC Press, Inc. Pirastu R, Combs P, Reggiani A, Foa V, Masina A, Maltoni C (1990): Mortality from liver disease among Italian vinyl chloride monomer/polyvinyl chloride manufacturers. Am J Ind Med 17:155161. Selikoff U (1992): Use of death certificates in epidemiological studies, including occupational Discordance with clinical and aruopsy findings. Am J Ind Med 22:469-480. Shah HC (1993): Diagnostic bias in occupational epidemiologic studies. Am J Ind Med 24:249-250. Simonato L. Abbe KA, Andersen A, Belli S, Combs P, Engholm G, Ferro G, Hagmar L. Langaid S, Lundberg I, Pirastu R, Thomas P, Winkelmann R, Saracci R (1991): A collaborative study of cancer incidence and mortality among vinyl chloride workers. Scand J Work Environ Health 17:159-169. R&S 143236 256 Wong tnd Whorton Smuievich VB, Fedotov* IV, Filatova VS (1988): Increasing evidence of the rise of cancer in workers exposed to vinyl chloride. Br J Ind Med 45:93-99. Storctvedt Heldaas S, Langard SL, Andersen A (1984): Incidence of cancer among vinyl chloride and polyvinyl chloride workers, Br J Ind Med 41:25-30. Tabenhaw IR, Gaffey WR (1974): Mortality study of workers in the manufacture of vinyl chloride and its polymers. J Occup Med 16:509-518. Theriault G. Allard P (1981): Cancer mortality of a group of Canadian workers exposed to vinyl chloride monomer. J Occup Med 23:671-676. von Greiser E, Rein] W, Weber H (1982): Vinyl-chloride exposition und mortalitat deutscher chemiearbeiter im vergleich zur mortalitat nichtexponierter chemiearbciter und PVC-verarfaeiter. Zentralbl Arbeitsmed Arbeitssch Prophyl Ergonomic 32:44-62. WtxweilerRJ, Stringir W, WagonerJK, Jones J. Falk H, Carter C (1976): Neoplastic risk among workers exposed to vinyl chloride. Ann NY Acad Sci 271:40-48. Wong O, Morgan RW, Bailey WJ, Swencicki RE, Oaxton K. Kheifets L (1986): An epidemiological study of petroleum refinery employees. Br J Ind Med 43:6-17. Wong O, Raabe GK (1989): Critical review of cancer epidemiology in petroleum industry employees, with a meta-analysis by cancer site. Am J Ind Med 15283-310. Wong O, Whorton MD, Foliart DE, Ragland D (1991): An industry-wide epidemiologic study of vinyl chloride workers, 1942-1982. Am J Ind Med 20:317-334. Wu W, Steenland K. Brown D, Wells V, Jones J, Schulte P, Halperin W (1989): Cohort and case-control analyses of workers exposed to vinyl chloride: an update. J Occup Med 31:518-523. R&S 143237