Mortality of workers compensated for silicosis during the period 1959-1963 in the Veneto region of Italy.

After reports appeared from other countries indicating an excess risk of lung cancer among silicotics, a cohort of workers compensated for silicosis during the period 1959-1963 in the Veneto region of Italy was constructed and followed for mortality through 1984. The results of the study showed a large mortality excess for infectious diseases (180 observed versus 9.5 expected), due to silicotuberculosis, and for diseases of the respiratory system (270 observed versus 33.5 expected) due to silicosis. An elevated standardized mortality ratio of 239 (70 observed versus 29.3 expected) from lung cancer was also detected. An increasing pattern was observed with time since first exposure, while the relationship with employment category and duration of exposure was less clear-cut. The lung cancer excess was also strongly associated with cigarette smoking, there being a dose-response relationship with daily cigarette consumption. The study confirms the results from other epidemiologic studies on silicotics which show this pathological condition to be associated with increased lung cancer mortality.

The question of whether exposure to silica dust may entail a carcinogenic risk for the lung in addition to its well-known effects on pulmonary parenchyma was reopened recently afte r a long period during which the epidemiologic data available have been considered to pro vide rather negati ve evidence for this association. Goldsmith et al (6) reviewed the experimental and human studies and concluded that the available evidence is suggestive of an association between lung cancer and expo sure to silica . A similar review by Heppleston (7) arri ved at opposite conclu sions, exonerating silica itself from a carcinogenic role.
The mortality of workers compensated for silicosis has been recently investigated in Sweden (14), Canada (5), Finland (8), and Italy (9). These studies consistently report an elevated mortality for lung cancer, which is however difficult to interpret in relation to the possible association between exposure to silica dust and lung cancer risk. In 1982 we reported (16) the results of a historical follow-up , through 31 December 1980, of a cohort of workers compensated for silicosis during the period 1959-1963 in the Veneto region of Italy.
The Veneto region (northeast Italy) presents one of the highest rates of silicosis in Italy due to the fact that , part icularly in the past , a significant proportion of the male working population has been employed in mines, tunnel s, and quarries in this and other regions of Italy and also abroad . Con sidering the poor work conditions in the first hal f of the century , these workers are likely to have been exposed to high levels of silica dust.
The present paper improves and extends the followup study to 31 December 1984, and investigates whether workers compensated for silicosis are at higher risk of death than the general population, part icular attention being given to tub ercolosis, silicotuberculosis, nonmalignant respiratory diseases, and lung cancer.

Subjects and methods
The members of the study population were all newly compensated cases of silicosis among residents in the Veneto region in the northeast part of Ital y (population in 1980,4296 545) during the period January 1959 -31 December 1963. The cases were ascert ained through the archi ves of the seven provin cial offic es of the National Institute for Compensation of Occupational Diseases.
Th e two major crit eria for compensation were a radiolo gical diagnosis of silicosis and an occupational history of exposure to silica. Only cases of silicosis have been included in the cohort. Two workers compensated for asbestosis during the same period were excluded.
Included in the study were 157 persons compen sated after working as miners in Belgium. As those who have worked in other European countries are not registered by the Institute for Compensation of Occupat ional Dis-eases , it is not possible to evaluate with precision to what extent the loss of such cases affects the co mpleteness of the cohort, but , according to estimates from the Institute, their number should not exceed 100.
Com plete occupational histories for th e period involving exposure to silica du st were available for most of th e cases (88 0/ 0) from the archi ves of the aforementioned instit ute. When missing, the information was supplemented by the clinical records of the hospital s in which th e silicosis had been diagnosed, mainl y the Institute of Occupational Medicine o f the University of Pad ova and the General Hospital in Belluno.
For 31 (2.4 % ) wo rkers (included in the cohort) the type of exposure remained unknown, and for 8 (0.6 % ) the duration of expo sure to silica was no t reported . Th e occupationa l history was not upd ated after the date of compensation, and thu s we cannot exclude some underestimation of duration of exposure in our da ta.
Out of the I 370 cases receiving compensation during the study period, six females (0.4 %) , as well as 51 indi viduals (3.8 %) for which the inform ation on date of birth, residence or date of first expo sur e was not ava ilable, have been excluded from the follow-up.
The status of all the remaining I 313 memb ers of the co hort as of 31 December 1984 has been ascertained through the mortalit y registr ies of the town s o f residence. A copy of the dea th certificates was obt ained for each decedent. The und erlying cause of death was cod ed accor ding to the eighth revision of the International Classification of Diseases (15). For 551 (62.8 % ) of the 878 deaths, the result s of th e postmortem examination were also available. For all but six of these person s, the diagno sis of silicosis was confirmed at autopsy. Th is finding indicates a high level o f reliability for the criteria used in assign ing co mpensatio n.
Inform ation on smo king habits was a vailable for I 225 wor kers (93.3 % ). Th e number of cigare ttes smo ked da ily was known for 889 of the I 064 smokers or ex-smo kers (83.6 %). Thi s infor mat ion was reported in a direct interview with the worker when he appli ed for compensation . Howeve r, any changes that occurred in the smoking habits after the interview are not kno wn.
Since the first report on the follo w-up of this cohort (16), the files of an additional 69 ind ividuals eligible for inclusion in the stud y have been discovered , and th ey have now been included in the study popul ation. For a few individua ls, incorrect dates of first exposure had to be modi fied, and there will be small differences in the distr ibut ion of the observed and expected deaths in any comparisons with the previou s follow-up.
Lung can cer mortality in Ital y is cha racterized by a wide geogr aphic variability with an increasing gradient from the southern region s to the nor thern ones (3). As lung cancer mortality rates in the Veneto region a re so me 40 % higher than national ones, both nat ional a nd regional populations have been used as reference. Age-, sex-and five-year calendar-specific na-tional mortalit y rate s were available for the per iod 1959-1980, and the y were extended until December 1984 in order to cover th e period of ob servation. Regional mortalit y rates were available onl y for two calendar periods, ie, 1967-1973 and 1974-1980. In order to cover th e entire period of observation , the rates of the first period have been used for the period 1959-1973 , and tho se of the second period were extended up to th e end of the follo w-up.
A mod ified life-table procedure and computing pro gram (4) has been u sed for computing the personyears at risk , expected numbers of deaths, a nd standardized mortalit y ratios (observed/ expected x (00) for the period of observat ion 1959-1984. Ninet y-five percent confidence inter vals for the sta nda rdized mortality rat ios have been computed on the assumption of a Poi sson distribution.
Generalized linear interactive modelling (GUM) (to) has been used for some aspects of the analysis of lung cancer mortality .

Results
Dur ing the ob servat ion period 878 deaths occurred among the I 313 memb ers of the cohort, which contr ibuted, for the same period, 23 493.9 person-year s at risk. Th e dist ribution of person-y ears by time since first exposure (table I) clearly shows that the great majority of person-years occurred in th e very long observation periods due to the criteria of inclusion in the cohort. Table 2 reports the result s of the mortali ty analysis during the overall period of observation by cau se of death . Tot al mortality was significantly increased mainly due to the lar ge excess of deaths fro m silicotub erculo sis and silicosis.
Cancer mort ality was also significantly in excess due to the twofold increa se of lung cancer and to an excess of cancer of the upper respiratory and digest ive tracts.
Six deaths from leukemia (3.0 expected) and one from lymphoma (3.5 expected) occurred. Gastr ointestinal cancer was not found to be in excess.
Among the other causes of death only mortality from diseases of the digestive system present ed a statisticall y significa nt increase. Thirty deaths from cirrho sis of the liver accounted for the entire excess.
In table 3 we present the causes o f death of majo r interest a nalyzed by time since first expo sure to silica  International classification of diseases, eighth revision). b The confidence interval was not computed if the expected num ber was less than 0.5. Table 4. Lung ca ncer mortality 20 or more yea rs since the su bjects' first expos ure by duration of expos ure. The mortality of both the national and regional populations have been used as reference. (E = expected number of deaths , SMR = standardized morta lity ratio) Du ration of exposure (years) Years s 9

10-19
'< 20 since first National rates Regional rates Observed N at ion~rates Regional rates National rates Regional rates " P < 0.05. Table 5. Lung cance r mortal ity 20 or more years si nc e fir st expos ure by typ e of ex posure. The mortali ty of the regional population has been used as referenc e. (0 = observed number of deaths, E = expe cted numbers of death s, SMR = stand ardiz ed mortali ty ratio, 95 % CI = 95 % co nf ide nc e int erval) Type of exposure Tabl e 5 shows th e lung cancer mortality by type of exposure 20 o r more years since first exposure . An elevated standardized mortality ratio is discernibl e in all the listed categ orie s. Silicotics with exposure in quarries showed the highest overall standardized mor tality ratio . All the ratios were ho wever based on small numbers.
We also ana lyzed the lung cancer mortality in relation to the calenda r periods in which the members of the coho rt were first exposed to silica . No clear pattern was discernible apa rt from a decreased standardized mortality ratio (close to 100 when the regional reference po pulation was used) for those enteri ng after 1945. Thi s finding is however difficult to inter pret because o f the shorter follow-up of th is subgro up, and these results have not been presented.
Out of the I 225 members of the cohort with known smoki ng hab its, 975 (79.6 070) were current smokers at th e time of the claim for compensation, 89 (7.3 010) were ex-smo kers , an d on ly 161 (13.1 (1(0) had never smo ked . The resu lts of the an alysis by smoking habits indicate that the excess risk for lung cancer was concentra ted amo ng curr ent smokers (59 observed versu s 22. 1 expected, sta nda rdized mo rtality rat io 268).
Tabl e 6 present s the results of the analysis of the lung cancer mortality by the am ount of smoki ng and the duration of exposure to silica. The ex-smokers were assigned to the cigarette smoking category corr espond -ing to their reported previous con sum ption. Th e exclusion from th e an alysis of 262 individuals with an unknown level of smoking and eight without duration of exposure acco unts for the slight di fferences fro m the previou s tables.
No clear synergistic effect between the smoki ng and silica exposure was di scern ible, whereas an increasing pattern was present for lung cancer mo rtalit y with an increasing level of smo king within the exposure categories, th e magnitude of risk being lar gest in the cat ego ry with the longest duration of exposur e. The effect of the two variables (level of smo king and duration of exposure) have formally been tested with GUM (10). Both variables were independently associated with a n increase in the sta nda rdized mortal ity ratio for lung cancer. The associa tion was stati stically significant (P < 0.01) for smok ing but not for duration of exposure (P = 0.2). A multiplicative model poorly fit the data with a significa nt negat ive inte ract ion term for the highest expos ure and smo king categor ies. The effect of each factor was only evident in the lower exposure categor ies of the other .

Discussion and conclusions
The evidence of a lack of associati on between silicosis and lung cancer has been mainly based on findin gs from an autopsy series in which the frequenc y of lung cancer in silicotics was compared to that in nonsilicotics (13). Recent studies in Sweden, the United State s, Finland, and Italy provide new eviden ce of a positive association between silicosis and lung cancer (5,8, I I, 12, 16). All these studies repor ted an elevated overall mortali ty, mainly due to nonmali gnant respiratory diseases and to tuberculosis or silico tuberculosis.
The results of our study confi rm these earlier findings. Our cohort exhibited a large mort ality excess due to silicot uberculosis and silicosis. More than 50 0J0 of the death s were due to these two ca uses, and there was no relat ion ship with dur ation or typ e of exposure or with smoking habits.
Lung cancer mortalit y was elevated , and the results from autopsies indicated that it could well have been undere stimated . Twenty-two lung cancer cases not reported on the death certi ficates were found by postmortem examination , while onl y six lung can cers reported on the death certificates were not confir med at autopsy.
• The lung cancer excess was only slightly associated with duration of exposure (although it showed a clear relationship amon g non smokers and light smo kers). Work ers with the longest duration of exposure experienced the lowest mortalit y from silicotuberculosis and silicosis and the highest mortality from lung cancer. This finding suggests that the risk for lung cancer is diffi cult to detect when it is comp etin g with the high risk of dying from silicotuberculosis and silicosis. The most severe cases of silicosis, in fact, have o ften been associated with relatively short periods of exposure to very high levels of silica dust.
Occupat ions in mines, tunnels, and foundri es involve potential expo sure to oth er well-known carcinogens, eg, radon daughters and polycyclic aro matic hydrocarbons. The possible contributing role of other occupa tional carcinogens such as these cannot be ruled out in our study .
The anal ysis by date of first exposure to silica was suggestive of a decrease in the standardized mortality ratio for lung cancer among tho se employed dur ing the mor e recent work cond itions. Thi s apparent effect might partl y be due to the shorter follow-up of the late entrants in th at neither of the two major causes of death due to silica exposure showed a downward trend with time of starting expo sure.
Although smoking in ou r study population was associated with the lung cancer excess, the magnitude of the risk was such as not to be readily explain ed by the effect of tobacco alone, even with the use of a regional reference population , which sho uld have had characteristics closer to the population under stud y. Among the member s of the cohort for whom the tob acco consumption was known, 15.4 0,70 had never smoked, 73.8 0J0 were light smokers, and 10.8 0J0 were heavy smokers (at least one pack per day). Similar data were not available for the Veneto region for the entire study period. Ho wever, the distribut ion of smoking hab its 122 in the Veneto population according to the 1980 census (9) were the following: 41.7 0J0 never smokers, 50.9 0J0 light smokers, and 7.4 0J0 heavy smokers. The application o f the model suggested by Axelson (2) to our data shows that the difference in the smoking habits between the study po pulat ion and the reference populat ion should account for a 39 0J0 excess of lung cancer among silicotics, as compare d to the 70.4 0J0 found.
Con sideration should also be given to the fact that the expected values were computed on the basis o f the lung cancer rat es of the general popul ation, which is constituted of bot h smo kers and nonsmokers. According to Saracci (13) the expected values for non smoke rs should be reduced by a factor of at least 4 when the possible effect of oth er exposures is being investigated. After the application of this adju stment , the standardized mortality rat ios for nonsmokers would present a three-to fourfold excess for lung cancer, depending on the reference population used. Although these estimates are based on small numbe rs and might well be affected by misclassification , they are however suggestive of the presence of a risk factor for lung cancer in additi on to smoking.
A possible inte rpretation could be that silicosis itself, being a progressive disease, may pred ispose to the developm ent of the neopl asia. Archer and his colleagues (1), in a follow-up study of uranium miners, found a higher frequency of bronchogenic cancer amon g miners with X-ray findin gs of pneumocon iosis as compa red to those without and suggested that pneumoconi osis may increase lung susceptibility to cancer. Studies on silicotic and nonsili cotic workers previously exposed to silica du st ar e warranted to test this hypothesis.
Only silicot ics were enrolled in our study population and ther efore the result s cannot be interpreted in terms of a causal relationship between lung cancer and exposure to silica dust , although the lung cancer mortality tended to increase slightly with dur ation of exposure.
In conclusi on, the results of this study indicate, in accord ance with other studies, that silicosis is a pathological condition entailing an extremely high mort ality risk, one that is also associat ed with a higher risk of lung tumors.