Document re89LGBO35n3VrODkp2p0G0q0

Synergy between Asbestos and Smoking on Lung Cancer Risks Thomas C. Erren, Michael Jacobsen, and Claus Piekarski We have examined data from 12 epidemiologic studies for quantitative evidence of biologic synergy between asbestos and smoking on lung cancer risks. Estimates of the effect associated with joint exposure to the two agents exceeded the sum of their separate effects in each study. We explored the variations in the strength of the synergistic effect across the studies using three indices: the ratio of the combined effects to the sum of the separate effects of smoking and asbestos (S), the relative excess risk due to interaction (RERI), and the attributable proportion of risk due to interaction (AP). The weighted average of S across all studies was 1.64 (95% confidence interval = 1.33-2.03). The attributable proportion associated with this average S was estimated as 33%, that suggests that one-third of cancer cases among smokers who were exposed to asbestos can be attributed to the synergistic behavior of the two carcinogens, as distinct from their separate effects and those attributable to other ("background") factors. (Epidemiology 1999;10:405-411) Keywords: synergy, asbestos, smoking, lung neoplasm, meta-analysis, biologic interaction. Following seminal reports published during the 1950s and 1960s,1,2 epidemiological studies have confirmed that both tobacco smoke and asbestos fibers are potent human carcinogens. In 1968, Selikoff et aP noted that the combined exposure to both of these agents appeared to be associated with more lung cancer cases than ex pected from the sum of estimates of their separate effects. Rothman et al4,5 have suggested that this type of phe nomenon may be interpreted as biologic interaction, or synergy, and that it can be quantified by an index, S, defined as the ratio of the observed effects of joint exposure to the two agents to the sum of estimates of the separate effects of each. Greenland and Rothman6 re view varying interpretations of "interactions" in epide miologic data and distinguish between (1) the purely statistical use of the word, which has to be interpreted in the context of the scale of measurement used to repre sent the response variable; (2) interactions that might be interpreted as implying differences in biologic mecha nisms depending on whether the suspect etiologic fac tors are acting singly or in combination; and (3) the public health implications of any such interactions. Var- Fr>>m the In>titut unJ Pelikhnik fur Arbeit'- unJ So:uiimedi:in Jer L'm\er'it.it :u Koln, Koln, Germanv. Address corropondence to: Thomas C. Erren, In'titut und Polikhnik fur Arbeit'und 5o:ialmedi:in der Lniver'itat :u Koln. bC924 Koln (Lindenthal'. Germanv. T. C. Erren 'tarred tho research during ho postdoctoral work at the Division or Public Health Biolouv/E-pidemmlouv. Lmver'itv of California at Berkelev. which n,i' supported bv a izrant from the German Academic Exchange Service. Submitted September lb. 1997; final version accepted February 199A c i'-W bv Epidemiolocn Re>ource< Inc. ious numerical measures that reflect these different ideas are also described.6 This paper focuses primarily on the biologic signifi cance of apparent departures from a simple additive model of the effects of exposure to asbestos and smoking on lung cancer risks. In this paper, we ask the following: Is the available evidence from different studies regarding such biologic synergy qualitatively consistent? What factors might be responsible for observed variations among studies? Is it possible to estimate, quantitatively, the overall magnitude of such synergy? We chose S as a convenient statistic to explore these questions. We also record two other indi ces, the relative excess risk due to interaction (RERI) and the attributable proportion of risk due to interaction (AP), the latter being helpful for an assessment of the public health implications of the findings. Subjects and Methods Data Sources We searched the MEDLINE reference base (1966-1996) and found three reviews1 ^ and 17 research papers' i:~:' that provided quantitative information about occupa tional exposures to asbestos, smoking habits, and the separate and combined associations of these factors with lung cancer risks. We selected studies for inclusion in the analyses reported here by verifying that sufficient data were available in the reports to estimate relative risks of lung cancer, and the corresponding standard errors, for "smokers" and for "nonsmokers" (as defined below), among both those exposed and those not ex posed or minimally exposed to asbestos (Table 1.). 405 406 ERREN ET AL Epidemiology July 1999, Volume 10 Number 4 TABLE 1. Selected Studies for the Analyses of Lung Cancer Risks in Relation to Occupational Asbestos Exposure and Smoking Design, Location Cohort, England Case-referent, England Case-referent, United States Cohort, United States and Canada Cohort, United States Case-referent, United States Nested casereferent, Canada Case-referent, Italy Cohort, England Case-referent, Norway Nested casereferent, Australia Cohort, China Source Population Employed at an asbestos factory in London; men, since 1933, and women, 1936-1942 Men admitted to a TSC during 1 year Residents of coastal Georgia; catchment in four local hospitals Entire membership of the insulation workers' union (United States and Canada) 933 amosite asbestos workers who began work from 6/1941 through 12/1945 in Paterson, NJ Mortality listing during 1976 from Virginia Birth cohort of ~~30,000 past and present employees of the Quebec chrysotile production industry Male workers in industrialized Lombardy region during 1976 and 1979 Employees at an asbestos factory in London; men, since 1933, and women, 1936-1942 Medical wards of Telemark and Vestfold County Hospital during 1979 and 1983 6,500 persons employed in the Wittenoom asbestos industry between 1943 and 1966 Workers employed in the Tianjin asbestos factory on January 1, 1972 Study Subjects 1,300 men, 480 women; study period, 1960-1970 201 cases, 201 matched referents 458 male cases, 553 referents 12,051 workers with at least 20 years of asbestos exposure; study period, 1967 1976 582 men; follow-up from 1961 through 1977 336 male cases, 492 matched referents ILS male cases, /1 3 matched referents 204 cases, 351 matched* referents 1,250 men, 420 women; follow-up from 1971 through 1980 176 male cases, 176 referents 40 male cases, 1,799 matched referents 662 men and 510 women Comparison Group SMR, using general population death rates adjusted lor residence in Greater London and tor observed smoking habits Admission to ISC 1972-1973 Referents from four regional hospitals and from death certificates External: 73,763 men of similar social class selected from an ACS study External: white men with matching smoking history from the ACS study Deaths in Virginia in 1976 Surviving members of same cohort Men from the study area SMR, using sex-, ageand period-specific death rates for England and Wales Medical wards of Telemark and Vestfold County Hospital Surviving members of the same cohort 3,219 workers not exposed to asbestos, dust, fumes, or vapor SMR = standardized mortality ratio; TSC = thoracic surgical center; ACS = American Cancer Society. First Named Author Berry1'-' Martischnig12 Blot1' Hammond14 SelikofU Blot17 Liddell15 Pastorino20 Berry21 Kjuus25 de Klerk74 Cheng25 From this list of 17 reports, we omitted 2,5,11 because a later paper,14 which is included, describes a larger cohort of insulation workers over a longer, overlapping follow-up period. We also did not include the report on Canadian chrysotile production workers,16 because the nested case-referent study by Liddell and colleagues,18 which is included, is based on that cohort and uses improved, less ambiguous classification of smoking habits. We also excluded the report from Acheson et al,19 because the expected number of lung cancer deaths was reported without correction for smoking. Finally, we excluded the Hilt et al22 report, because there were no lung cancer cases recorded among non smokers. We therefore used 12 reports for our analy ses. Berry et al10,21 described lung cancer mortality in the same cohort in two papers. Data from both are included in the analyses described here because they refer to distinct, nonoverlapping follow-up periods. The later paper,21 published in 1985, includes revised estimates of effects reported earlier.10 We used those revised esti mates in the calculations described here. Epidemiology July 1999, Volume 10 Number 4 SYNERGY OF ASBESTOS AND SMOKING 407 TABLE 2. Summary of Information about Smoking and Exposure to Asbestos in 12 Selected Reports, and Criteria Used for Definition of Exposure Dichotomies in this Review Information on Asbestos Exposure Occupational history extracted from the personnel records at the factory; mixed asbestos Assessment of the occupational history when admitted to the TSC; any asbestos Occupational history assessed through interviews with cases and referents or their next of kin; unspecified asbestos Membership in the International Association of Heat and Frost Insulators and Asbestos workers; chrysotile, amosite Work at the factory during 1941-1954; confirmation of predominant use of amosite in the asbestos factory Occupational history assessed through personal interviews with next of kin of cancer patients and referents; unspecified asbestos Assessment of average dust concentrations at specific jobs based on measurements, approximations and interviews; chrysotile Assessment of occupational history in interviews with cases and referents or their next of kin; any asbestos See Berry et al.10 Asbestos history through personal interview and questionnaire for cases and referents: any asbestos Occupational exposure from employment records; crocidolite Detailed work history from employment card; unspecified asbestos TsC = thoracic 'urgical center. Information on Smoking Habits Questionnaire or interview on smoking habits for those alive in 1971; for deceased: medical records, family doctor, relatives Assessment when admitted to TSC Personal interviews of patients and referents; for deceased, next of kin Questionnaire on lifetime smoking habits (recorded in 1966) Ascertainment of smoking habits 20 years after start of employment (1961-1965) Personal interviews with next of kin of cancer patients and referents Questionnaire responses on smoking habits as in 1970 for those alive or from relatives or friends of those who died after 1950 Interviews with cases and referents or their next of kin Questionnaire or interview on smoking habits in 1971 Personal interview and questionnaire for cases and referents Questionnaire in 1979 Smoker's history was determined by number of cigarettes/day, beginning smoking age, number of years smoked Asbestos/No Asbestos Definition Used Here Severe exposure vs external comparison population Exposed vs not exposed Ever employed in shipbuilding, yes vs no Asbestos workers (when at least 20 years exposed) vs external comparison population Amosite asbestos worker vs external comparison population Employed in shipyards before 1950, yes vs no Exposure to >100 fibers vs <100 fibers Exposed (probably) vs not exposed Severe asbestos exposure vs external comparison population Grade 2-3, moderate or heavy exposure vs grade 0-1, no, uncertain, light/sporadic exposure High exposure vs lowexposure Asbestos exposure, yes vs no Smoker/Nonsmoker Definition Used Here Men, smokers vs never-smoked; women, smoked at some time vs never smoked At least 15 cigarettes/day vs 0-14 cigarettes/day Heavy or moderate smoker vs nonsmoker, light smoker, or stopped smoking History of cigarettes vs never smoked regularly History of cigarette smoking vs never smoked regularly Heavy or moderate smoker vs never smoked, light smoker, or long-term ex smoker At least 1 pack-year vs 0 pack-years At least 10 cigarettes/day vs 0-9 cigarettes/day Men, smokers vs never smoked; women, smokers vs never smoked At least 10 cigarettes/day vs 0-9 cigarettes/day Smokers vs nonsmokers (includes cessation for >10 years) Men and women (>1 cigarette/day for at least 1 year): Yes vs no First Named Author Berry10 Martischnig12 Blot1* Hammond14 Selikoff15 Blot17 Liddell13 Pastorino20 Berry21 Kjuus2' de Klerk74 Cheng-' Exposure Classification Table 2 summarizes the information about exposure to asbestos and smoking habits from the reports and indi cates how we used this information in this review. We defined a "no-exposure" category of smoking behavior to include nonsmokers or all of those in the lowest smoking category described in the paper concerned. We excluded ex-smokers from the analysis where possible, on the grounds that lung cancer risks decrease gradually after cessation of smoking.26 We also excluded those who 408 ERREN ET AL Epidemiology July 1999, Volume 10 Number 4 TABLE 3. Standardized Estimates of Relative Lung Cancer Risks Associated with Separate and Joint Exposures to Smoking and Asbestos, and Three Indices of Synergy (S, RERI, AP)* Lung Cancer Risk Ratio Estimates by: Smoking Asbestos No Yes S 95% Cl RERI AP First Named Author No 1 5.0 2.30 0.94-5.60 19.5 0.55 Berry10'!' Yes 12.1 35.5 No 1 1.1 5.30 0.93-30.15 3.71 0.67 Martischnig12 Yes 1.8 5.6 No 1 1.3 1.65 0.98-2.78 2.60 0.34 Blot15 Yes 4.7 7.6 No 1 5.2 3.73 1.71-8.11 38.22 0.72 Hammond14 Yes 10.9 53.2 No 1 25 1.25 0.58-2.68 7.94 0.20 Selikoff1'' Yes 8.7 40.6 No 1 1.9 1.30 0.68-2.48 0.88 0.18 Blot17 Yes 3.1 4.9 No 1 3 1.22 0.78-1.91 1.29 0.16 Liddell18 Yes 4.9 8.2 No 1 2.8 1.41 0.55-3.63 2.57 0.26 Pastorino20 Yes 5.5 9.9 No 1 11.5 1.45 0.70-3.01 9.57 0.30 Berry2't Yes 11.8 31.9 No 1 2.4 3.24 1.15-9.12 13.04 0.66 Kjuus23 Yes 5.4 19.9 No 1 2.2 2.33 0.75-7.27 4.89 0.51 de Klerk24 Yes 3.4 9.6 No 1 5.4 1.54 0.74-3.151 2.71 0.31 Cheng2:> Yes 1.6 8.7 * S = synergy index; RERI = relative excess risk due to interaction; AP = attributable proportion of risk due to interaction, t Men and women combined. t These confidence limits were not calculable directly from the data reported.26 Regression of the estimated standard errors (SEs) of ln(S) from the other 11 studies on the corresponding values of S provided a convincing fit (variance accounted for = 60.2%). The fitted linear equation was used to estimate SE(lnS), and thus the 95% Cl, for S = 1.54. reported smoking a "pipe and/or cigar only" from the analysis. For occupational exposure to asbestos, we defined a no-exposure category that included all of those in the lowest exposure category from each paper. We classified all others as "exposed." Analysis We estimated the relative risks of lung cancer associated with smoking, with exposure to asbestos, and with si multaneous exposure to both for each of the 12 reports. For each cohort study, we calculated the relative risk as RR,j = (RJRqo), where i and j index respectively the presence (1) and absence (0) of exposure to tobacco smoke and to asbestos, as defined above. Rn was calcu lated as x^/PYRi,, and Ri0 was estimated by yi0/PYRa, where xn refers to number of reported lung cancer cases, yi0 to number of cases expected on the basis of agespecific lung cancer death rates in the external compar ison population involved, and PYRn to number of per son-years at risk, for i = 0,1. For case-referent studies, we estimated the relative risks by odds ratios (OR), as follows: OR,j = (cjrlt)l(c00l r00), where c,j and r,3 refer to cases and referents, respec tively, with exposure combination ij. We then calculated excess relative risks, ERRy, for effects associated with smoking and exposure to asbestos, singly and in combination, by subtracting unity from the estimates of relative risk. This calculation allowed us to derive three indices of synergy for each study: the syn ergy index, S27 [S = ERR,,/(ERR10 + ERR01)]; the relative excess risk due to interaction,28 RERI [RERI = ERRn - (ERR10 + ERR01)]; and the attributable propor tion of risk due to interaction,29 AP {AP = [ERRU (ERR10 + ERR0i)]/(ERRn + 1)}. AP is that fraction of total lung cancer risk among those exposed to both factors in the population concerned (including the back ground risk in the absence of the two factors) that is attributable to the combined (as distinct from the sep arate) effects of the two factors. The standard errors of the natural logarithms of S were estimated from the formulae for cohort and casereferent studies proposed by Rothman2' and were used to approximate to 95% confidence limits for S by exponen tiating the limits for ln(S). Values of S from selected groups of studies were pooled to produce weighted aver ages, SP [and their approximate 95% confidence inter vals (CIs)], where Sp = expj[Swln(S)]/(Si(;)}; the sum mation is over the different studies; and the weighting Epidemiology July 1999, Volume 10 Number 4 SYNERGY OF ASBESTOS AND SMOKING 409 factors, w, are the reciprocals of the estimated variances of the ln(S).,c Results Table 3 shows that for each study, the combined effects of smoking and asbestos (ERRn) were greater than the sums of the estimated separate effects (ERR10 + ERR01). Values of S ranged from 1.2 to 5.3 (coefficient of vari ation, CoV = 57%). The differences between the com bined and the sum of the two separate excess relative risks (RER1) ranged from 0.9 to 38.2 (CoV = 117%). The fraction of total lung cancer risk attributable to the biologic interaction among those exposed to both agents (AP) ranged from 16 to 72% (CoV = 52%). Although the 12 estimates of S and of their confi dence limits vary, there was no suggestion of systematic study design-associated variation in S. Of the two high est values of S, one was from a case-referent study,12 and the other was from a cohort study.14 The other four values of S from cohort studies10,1''21,2' were not conspic uously different from those calculated from most of the case-referent data. In seven of the studies, however, including four of the five cohort studies,10141''21 esti mates of the asbestos-associated risks for smokers were less than those for nonsmokers; for the other cohort study,2' these risks were almost identical. The rank correlation between S and AP was high (Spearman's rs = 0.98). The rank correlation between S and RERI was not as high, although it was still positive (rs = 0.59). This lower correlation reflects the relatively low value of RERI in the Martischnig et al study,12 despite the high values of S and AP, and also the converse pattern in results from workers exposed to amosite asbestos.1' A test of the homogeneity of S from 11 studies that provided information allowing direct estimation of the variances of the corresponding ln(S) did not contrain dicate pooling the data concerned (^-210 = 11.48; P = 0.32). The weighted summary value of S based on those 11 studies (that is, omitting the study by Cheng and Kong2') was 1.66 (95% Cl = 1.33-2.06). This weighted average is not overly dependent on just one or two high values of S, because the latter are associated with rela tively wide CIs. Inclusion of results from the 12 th study' in the weighted average, using an additional approxima tion to the variance of ln(S), also made little difference to the estimate of Sr (it was 1.64). Discussion Results from the 12 reviewed studies consistently indi cated that the joint effect on lung cancer risks of smok ing and occupational exposure to asbestos was greater than the sum of the separate effects. The consistency might he artifactual in part, because choice of studies tor inclusion in meta-analyses is always subject to a possible bias arising from the reluctance of authors (and some editors) to publish negative results. In the present case, such "publication bias" would probably he relatively mild with respect to the main issue, the postulated synergy between asbestos and smoking on lung cancer risks. Absence of evidence supporting the idea of synergy between these two agents is unlikely, on its own, to have disqualified otherwise interesting results from publica tion. It was necessary, for inclusion in this review, that reports contain sufficient detail for estimation of both the separate and the combined effects of asbestos and smoking. This requirement may have influenced the pattern of results obtained. Nevertheless, we believe that it is reasonable to conclude that the findings provide a fairly clear-cut answer to the first of the three questions posed in the introduction to this paper. Epidemiologic evidence accumulated over some 30 years is consistent, in broad qualitative terms, with the observation by Selikoff et al,3 in 1968, of a potentiating biologic interac tion (synergy) between the effects of asbestos and smok ing on lung cancer risks. The other two questions prefacing this paper, about variability between the stud ies and quantitative generalizability, require more de tailed consideration. Variability between Studies Indices of synergy derived from five cohort studies did not differ systematically from those based on case-refer ent data (Table 3). Moreover, results from the three hospital-based case-referent studies12'1,'2, were typical of the ranges recorded from all 12 studies. The overall pattern of variability in results is therefore not explicable simply in terms of study design-related factors. The follow-up periods in the five cohort studies were similar, between 10 and 16 years, but the criteria for inclusion in the study of North American insulation workers14 were unusual in that these criteria required that all of the individuals involved had been exposed to asbestos for at least 20 years. It is conceivable that this could have resulted in a pattern of response different from that recorded in the other four cohort studies, in which less demanding exposure criteria were used to define the study groups. Suppose that the 20-year min imum exposure criterion determined that some of these men had also experienced nontrivial exposures to occu pational carcinogens other than asbestos. (McDonald et aP1 have speculated that exposure to materials other than asbestos may he one possible explanation for the apparently higher asbestos-associated cancer risk among these insulation workers as compared with that found in Canadian chrysotile miners and millers. 1',i 1) Suppose additionally that the carcinogenic potential in the lung of such other material is enhanced in the presence of tobacco fumes. This scenario would not affect the asbes tos-specific effect as estimated tor this review (ERRC1), but it would inflate the effects apparently associated with smoking, nominally "on its own" (ERRi:) and in combination with exposure to asbestos (ERRn). The results calculated from the study by Hammond et al14 tollow this pattern. The ERR;i (4.2) is the fourth highest of the 12 values calculated but close to the arithmetic mean (4.6). The ERRi: (9.9) and ERRU (52.2) are both 410 ERREN ET AL Epidemiology July 1999, Volume 10 Number 4 well above the corresponding means (5.2 and 18.6, re spectively.) Thus, the postulated exposure of the insula tion workers to at least one lung carcinogen other than asbestos that potentiates (or is potentiated by) tobacco smoke could partially explain the relatively high synergy indices calculated from those data.14 Sophistication in assessments of exposures to asbestos varied across the studies, from "ever employed in ship building" (mostly during World War II)13 to estimates of individuals' working-life cumulative exposures to asbes tos.115 We used the dichotomy exposed/not exposed in an attempt to place the available data on an approximately common scale. This simplification implies loss of power to identify a possible asbestos dose-related variation in any synergy with smoking. But if the relatively crude exposure gradient defined by the dichotomy does suggest a synergistic effect (as it does here), then there should be a compensatory increase in confidence that the apparent effect is real rather than artifactual, provided that errors in exposure classifications are random. The limits of that confidence are quantified approximately in Table 3 as 95% Cls, on the assumption of no classification bias. That assumption is not necessarily valid, particularly for hospital-based case-referent studies, but a possible asbes tos exposure classification bias would have had to be substantial to have affected the overall pattern of results. Three of the studies15,18,24 allow us to consider varia tion in synergistic effects relative to the type of asbestos involved. Estimates of the three synergy indices for the Australian workers who were exposed "almost exclu sively" to crocidolite24 are all higher than the corre sponding results from the Canadian chrysotile miners and millers.18 But the lower 95% confidence limit for S based on the Australian data (0.75) is almost the same as that calculated from the Canadian results (0.78). The RERI among factory workers who were described as having been exposed almost exclusively to amosite15 is higher than those found for the other two specifically differentiable fiber types,18,24 but this RERI (7.94) ranks as only the fifth highest among those from all 12 studies, and the corresponding values of S and AP are both low. Thus, the available data do not convince us that there is a difference in the synergistic potentials of different types of asbestos. We also used a dichotomy to summarize the varying representations of exposure to tobacco smoke that are recorded in Table 2. The convention that we adopted was arbitrary and may have further reduced power to detect a real synergistic effect. But again, because a synergistic effect was detectable using this crude classi fication of smoking habits, we think it even more likely that the effect would have been found if it had been possible to quantify exposure to tobacco smoke more precisely, provided also that there was no differential misclassification of smoking habits between those with and those without lung cancer. How Strong Is the Synergistic Effect2 We believe that the robustness of our estimate of Sp, and the homogeneity of the contributing sets of data, justify the following generalizations. The excess lung cancer risk arising from simultaneous exposures to particular levels of asbestos and tobacco smoke is higher than the sum of the two separate excess risks by a factor of about 1.64. The 95% Cl associated with this estimate is ap proximately 1.33-2.03. Public health implications of these findings can be assessed by noting that the 12 AP indices that we calculated for this review are very closely related to the corresponding values of S. A fitted quadratic in S ac counted for 98.7% of the variability across the APs. The estimated value of AP, given Sp = 1.64, was 33%. Estimates of AP corresponding to the approximate 95% confidence limits for Sp are 22% and 45%. The results therefore suggest that, among smokers who are also exposed to asbestos, some 33% of lung cancer cases can be attributed to the synergistic behavior of the two carcinogens, as distinct from their separate effects and from those arising from other ("background") factors. Acknowledgments Allan Smith stimulated the work and provided valuable comments on earlier drafts. We also acknowledge gratefully the many insightful suggestions from an anonymous referee. References 1. Doll R, Hill AB. Smoking and carcinoma of the lung: preliminary report. BMJ 1950;2:739-748. 2. Selikoff IJ, Churg J, Hammond EC. Asbestos exposure and neoplasia. JAMA 1964;188:22-26." 3. Selikoff IJ, Hammond EC, Churg J. Asbestos exposure, smoking, and neo plasia. JAMA 1968;204:106-112. 4- Rothman KJ. Synergy and antagonism in cause-effect relationships. Am J Epidemiol 1974;99:385-388. 5. Rothman KJ, Greenland S, Walker AM. Concepts of interaction. Am J Epidemiol 1980;112:467-470. 6. Greenland S, Rothman KJ. Concepts of interaction. In: Modern Epidemi ology. 2nd ed. 1998;329-342. 7. Vainio H, Boffetta P. Mechanisms of the combined effect of asbestos and smoking in the etiology of lung cancer. Scand J Work Environ Health 1994;20:235-242. 8. Steenland K, Thun M. Interaction between tobacco smoking and occupa tional exposures in the causation of lung cancer. J Occup Med 1986;28:110118. 9. Saracci R. The interactions of tobacco smoking and other agents in cancer etiology. Epidemiol Rev 1987;9:175-193. 10. Berry G, Newhouse ML, Turok M. Combined effect of asbestos exposure and smoking on mortality from lung cancer in factory workers. Lancet 1972;2: 476-479. 11. Selikoff IJ. Hammond EC. Multiple risk factors in environmental cancer. In: Fraumeni J, ed. Persons at High Risk of Cancer. New York: Academic Press, 1975:467-483. 12. Martischnig KM, Newell DJ, Barnsley WC, Cowan WK, Feinman EL, Oliver E. Unsuspected exposure to asbestos and bronchogenic carcinoma. BMJ 1977;1:746-749. 13. Blot WJ, Harrington JM, Toledo A, Hoover R, Heath CW Jr, Fraumeni JF Jr. Lung cancer after employment in shipyards during World War II. New Engl J Med 1978;299:620-624- 14. Hammond EC, Selikoff IJ, Seidman H. Asbestos exposure, cigarette smoking and death rates. Ann N Y Acad Sci 1979;330:473-490. 15. Selikoff IJ, Seidman H, Hammond EC. Mortality effects of cigarette smoking among amosite asbestos factory workers. J Natl Cancer Inst 1980;65:507513. 16. McDonald JC, Liddell FDK, Gibbs GW, Eyssen GE, McDonald AD. Dust exposure and mortality in chrysotile mining, 1910-75. Br J Ind Med 1980; 37:11-24. 17. Blot WJ, Morris LE, Stroube R, Tagnon 1, Fraumeni JF. Lung and laryngeal cancers in relation to shipyard employment in coastal Virginia. J Natl Cancer Inst 1980;65:571-575. 18. Liddell FDK, Thomas DC, Gibbs GW, McDonald JC. Fibre exposure and Epidemiology July 1999, Volume 10 Number 4 SYNERGY OF ASBESTOS AND SMOKING 411 mortality from pneumoconiosis, respiratory, and abdominal malignancies in chrysotile production in Quebec, 1926-1975. Singapore Ann Acad Med 1984;13:340-344. 19. Acheson ED, Gardner MJ, Winter PD, Bennett C. Cancer in a factory using amosite asbestos. Int J Epidemiol 1984;13:3-10. 20. Pastonno U, Berrino F, Gervasio A, Pesenti V, Riboli E, Crosignani P. Proportion of lung cancers due to occupational exposure. Int J Cancer 1984;33:231-237. 21. Berry G, Newhouse ML, Antonis P. Combined effect of asbestos and smoking on mortality from lung cancer and mesothelioma in factory work ers. Br j Ind Med 1985;42:12-18. 22. Hilt B, Langard S, Andersen A, Rosenberg J. Asbestos exposure, smoking habits, and cancer incidence among production and maintenance workers in an electrochemical plant. Am J Ind Med 1985;8:565-577. 23. Kjuus H, Skjaerven R, Langard S, Lien JT, Aamodt T. A case-referent study of lung cancer, occupational exposures and smoking. Scand J Work Environ Health 1986;12:203-209. 24- de Klerk NH, Musk AW, Armstrong BK, Hobbs MST. Smoking, exposure to crocidolite, and the incidence of lung cancer and asbestosis. Br J Ind Med 1991;48:412-417. 25. Cheng W-N, Kong J. A retrospective mortality cohort study of chrysotile asbestos products workers in Tianjin 1972-1987. Environ Res 1992;59:271278. 26. Selikoff IJ, Hammond EC. Asbestos and smoking. JAMA 1979;242:458459. 27. Rothman KJ. The estimation of synergy or antagonism. Am J Epidemiol 1976;103:506-511. 28. Rothman KJ. Modem Epidemiology. 1st ed. Boston: Little, Brown, 1986;323. 29. Walker AM. Proportion of disease attributable to the combined effect of two factors. Int J Epidemiol 1981;10:81-85. 30. Greenland S. Quantitative methods in the review of epidemiologic litera ture. Epidemiol Rev 1987;9:1-30. 31. McDonald JC, McDonald AD, Gibbs GW, Siemiatycki J, Rossiter CE. Mortality in the chrysotile asbestos mines and mills of Quebec. Arch Environ Health 1971;22:677-686.