Is an association between pleural plaques and lung cancer without

Is there an association between pleural plaques and lung cancer without asbestosis? 1994;20:62~. OBJECftVES - A recent review or meta-analysis of epidemiologic studies concluded that persons with asbestos-related pleural plaques do not have an increased risk of lung cancer in the absence of parenchymal asbestosis. The reviewer inferred that this conclusion provided indirect supportive evidence for the proposition that asbestosis is a necessary precursor of asbestos-related lung cancer. The objective of the present communication is to contest these claims. METHODS - Finnish epidemiologic data and population statistics were used to estimate the apparent risk ratio of lung cancer associated with radiographic signs of pleural plaques. Power calculations were applied to compute the needed population sizes to demonstrate that the asso ciation is statistically significant. RESULTS- Unrealistically large population studies would be needed to observe the statistical relation between pleural plaques and lung cancer, quantitated as a risk ratio of 1.1, resulting from relatively low levels of environmental asbestos exposure. In realistic and valid epidemiologic studies on heavily exposed subpopulations, a two- or threefold risk can be identified. CONCLUSIONS - Uninformative studies should not be interpreted as providing suppressive evidence that pleural plaques are a noncausal risk indicator of lung cancer. Even for the null hypothesis, the inference that asbestosis is a necessary causal link between asbestos and lung cancer is illogical.

In a recent review, Weiss ( I) exam ined epidemiologi c studies that dealt with the relation ship between pleural plaques unaccompanied by asbe stos is and lung cancer. He used the method of det ecting pleural di sease to excl ude asbe stosi s. In all, there were 13 in vesti gations, six cohort studies, four lung cancer cas e-referent studies, and three autopsy studies . Wei ss judged that only three studies supported the hypothesis th at lung cancer risk is elevated amo ng per son s with pleural pl aques over the risk of unexposed people and that the se three studies had the most defects in desi gn. Ac cordingly, Wei ss concluded that " .. . persons with asbestos -re lated pleural pl aques do not have an increas ed risk of lung cancer in the ab sence of parench ymal asb est osis [p J 854]." Thi s conclusion is que stionable. Weiss argued that the wei ght of ev idence fav ors the hyp othesis becau se the examined studies either yielded a " nega tive" re sult or were in validated by a deficient design. This argument is speci ous. Whil e Weiss acknowledged that it is difficult (i n prin ciple imp ossible) to pro ve a negativ e pr opositi on in empirical studies, he neglected to consider the important issue of statistical power.

A priori epidemiologic and medical knowledge
Pleural plaques are undoubtedly rel ated to asbestos expo sure . Simultaneously, the incidence of lung cancer increase s with asbestos exposure. The epidemiolo gic results indicate th at there should ex ist an association (statistical correlation) between the occurrence of plaques and the realized risk of lung ca nce r. In other words, the presen ce of pleural plaques is a sign that conveys pr edictive information about the presence or future occurrence of lun g cancer with out a cau sal connotation (ie, a noncausal risk indicator of lun g canc er). But it is established medica l kno wledge that pleural plaqu es are not a precursor of lun g cance r. Most lung cancers are diagn osed with out any sign of pleural plaques, and lung parenchyma and the parietal pleura are an atomicall y and histologically different. On the other hand, whether as bes tosis is an intermediate st age in the cau sal process from asbes tos expo sure to the devel opment of lung ca ncer is a matter of some contention in the literature.

Weiss' claims
We iss ( I) addressed the question of a po ssible mcchanism of asbestos-induced lung cancer. He considered the foll owing two pos sible models: one with asbe sto sis as an int erm ediate stag e in the pathogenesis and the other w ith pleural plaques as a link in the causal pathway leading to lung cancer. To discriminate bet we en the mod els, We iss indirectly inferre d that, if the risk of lung cancer is not increased for pers ons with asbestos-rel ated pleural plaques but " From reference 4. b Propo rti on of asbestosi s is 25 : 228 =11% . Table 1. Cross-cla ssifi cati on of lung cancer and asbestosis as the underlyin g or contributory causes of death on the death cert ifi cates of members of a cohort of male asbestos wo rkers" sho wing th at lung can cer can appear without asbesto sis and vice versa in asbestos-exposed persons ."

Asbestos exposure
With With out Total plaques plaques Probably exposed

382 439
Possibly or unlikely

2671 2835
Total 221 3053 3274 Table 2. Cro ss-classifi cation of a popul ation sample of 3274 men accord in g to the likelihood of asbestos exposure and t he presence of pleu ral plaques. " On the assumption that the risk of lung cancer is twofold among the probably exposed men relative to that among th e poss ibl y or unli kely expos ed men, the apparent relat ive ris k associ ated wit h pleural plaque s ca n be computed simply as t he rat io of prevalent numbers weighted by th e spec ific ris k of th e asbestos exposure cate qory.> Recent eviden ce suggests that lung tumors can be caused by relativel y low levels of asbestos exposure independently of parenchymal fibrosis (5 , 7). Third, there remains the core question of how large a population is needed in a valid epidemiologic study to demon strate the association between the occurrence of pleural plaqu es and an excess risk of lung cancer. We have attempted to answer this question of statistical power by considering an empirical example.
To tal

Lung cancer No lung cancer
Quan titat ion of lung cancer risk and po wer calculations In a representative sample of 3274 adult Finnish men, the proportions definitely or probably, possibly, and unlikel y exposed to asbestos were estimated to be 13.4, 52.9, and 33.7%, respectively (8). In these exposure categ ories , the correspond ing prevalences of radio graph ic pleural plaque s were 13.0, 6.5 , and 4.6%, respectively. In two other studies, the risk of dying from lung cancer among asbestos-exposed men was about twofold (4,7). From these findings we computed the relative risk of lung cancer associated with the presence of pleur al plaques to be as low as 1.1 (table 2). Upon adjustment by smoking cate-without asbestosis, then asbestosis would be a necessary precursor for asbestos -related lung cancer. In this argument, pleural plaque s were used as a marker of asbestos exposure among asbestos-exp osed persons. Even though the estimate of relative risk' would not be statistically significantly ele vated for such persons in comparison with the risk for unexposed persons or the general population, Weiss' hypothesis is not neces sarily correct.

Count erclaims to Weiss ' claims
We have three counterclaims to Weiss' claims. First, the valid ity (ie, sensitivity plus specificity) of pleural plaques as a marker of asbestos exposure is usually poor. The sensitivity of pleural plaques is often low. This insensiti vity stems from the fact that not all persons with asbestos expo sure develop pleural plaques that originate from the exposure, as is evident, for example, from the autopsy studies reviewed by Weis s. Regarding the specifi city issue, not all pleural plaques are asbestos-rela ted ; instead, the radiological signs can be due to, for example, tube rculosis. Hypothetically, if pleural plaques were a perfectl y valid marker of asbestos expos ure, then all asbestos-exposed subje cts would be marked as such, and no asbestos-related plaques would be detected in unexp osed subje cts. Weis s limited the selected studie s to compri se asbestos-exposed subjects only, but he neglected to consider the validit y of pleur al plaque s as a biologi cal marker .
Second, from the reject ion of a model with the pleural plaques link it does not follow that asbestosis is a necessary pre cursor among asbestos-exp osed subjects. If asbestosis were a necessary conditi on, then every lung cancer case attributable to asbestos would also involve asbestosis. Unfortunately, one cannot identi fy the exposure-induced cases without resorting to strong biologi cal assumptions, but one can estim ate the proportionate increa se in case load, the excess fraction, due to exposure (2). If the relative lung canc er risk, R, of an asbestos-exposed popul ation is, for example, 2.0, then the excess cases attributable to exposure form the fraction (3), (R-I)R-' , of all cases occurring in the exposed group, which is 0.5; thus, with biase s (eg, confounding by smoking) absent, the proportion of asbestosis in the exposed population would be 50%. But epidemiologic data contradict this estimate (4,5). (See table I. ) Consequentl y, the claim docs not hold.
In addit ion, when both asbestosis and lung cancer are found together, because of the long latency time of these diseases, it can be difficult to ascertain that asbestosis preceded lung cancer. Moreover, there are alternatives to the two-stage models. A simple model assumes that asbestos exposure is a common etiologic agent causing the pleural plaque s on . 1 Relative risk is used here as a generic term for risk ratio, rate ratio, or hazard ratio regardless of the study design which yielded the estimate. Scand J Work Environ Health 1994, vol 20, no 1 gories, the relative risk was exactly the same , 1.1.
To show that such a relati ve risk is statistically significant, one would need to study a representative population sample of approximately 300000 people, which would render the study prohibitively expensive.
One of the studies reviewed by Weiss (1) was a tube rculosis screening survey conducted in three communities in Finland. It enrolled 7986 adults who were followed for 18 years (9). The communities represent ed different patterns of residential and occupational exposure to asbestos. By no mean s were all of the subjects with pleural plaques exposed to asbestos, as reported by Weiss. (See tables 1 through 3 in reference 1.) Moreover, not all of the pleural abnormalities in the asbe stos-exposed per sons were asbe stos-related pleural plaques, as assumed by Weiss in his inference. The observed number oflung cancer cases among the 604 subjects with localized pleural plaques but no other asbestos-rel ated radiographic sign s was 28, wherea s, among the 604 referen ce pers ons with no signs of pleural abnormalities (some of whom could have been exposed to asbestos), 25 lung cancers were identified . The gender-, age-, and residence-adjusted relative risk was 1.1an estimate similar to the aforementioned one -with a 95% confidence interval of 0.6-1.8. Information on smoking habits was not available. Becau se of the generally low levels of environmental asbestos exposure, it is unlikely that an expanded study base (ie, larger population size or longer follow-up time ) that would be practical would demonstrate a statistically significantly elevated risk of lung canc er among the carriers of pleural plaques.
If, instead, the study had focu sed on a definitely or probably asbestos-exposed subpop ulation, then a sample size calculation with the same statistics and estimates would yield the following result: with a significance level of 5% and a power of 80% the required sizes of the asbestos-e xposed and an equalsized unexposed comparison group would have to be 538 +538 or 175+ 175 to detect an apparent relative risk of 2 or 3, respectively . (See referen ce 10.) These sample sizes are tractable.

Concluding remarks
In closing, we cauti on against drawing hast y conclusion s because some of the results of "negative" stud-64 ies on asbestos-expose d popul ations have turned out to be questionable or based on inadequate population-tim e experience. In realistic epidemiologic studies, selection bias, uncontrolled confounding, misclassification of exposure (and often disease), and sampling variability are es pecially difficult to quantify. Pre sum ably, most of these uncertainties distort the pleural plaque-lung cancer association in the negative direction. (For a discussion, see reference 9.) Finally, before conclusions are drawn , the logic of inference from epidemiologic evidence to the probability of causation should be flawles s.