Silica dust exposure and lung cancer

1994;20:407-16. OBJECTIVE - The study evaluated the possibility of a direct association between silica dust exposure and lung cancer. METHODS - Mortality and morbidity among [026 granite workers was followed in [940-1989. Re gional census data for 1970-1985 and lung cancer incidence data were also linked. The cytotoxicity of different granite fractions and their capacity to induce reactive oxygen species (ROS) in human leukocytes was studied in vitro. RESULTS - Excess lung cancer mortality was found during several follow-up periods. The rate ratios were 1 .6-3.8 for different latency periods. Lung cancer risk increased with the length of exposure and latency. Lung cancer morbidity in 1953-1987 showed an excess for the red and grey granite areas, the rate ratio being 2.5 for ~20-year latency. Only one lung cancer case came from the black granite area. The cancer cases from the grey area had a shorter latency than those from the red area. The mineral composition differed for each area. The strongest ROS-inducing activity of grey and red granite was seen in the quartz-containing fractions. In the cytotoxicity tests the quartz-containing frac tions of the grey and red granite also caused the strongest lactic dehydrogenase (LDH) release. How ever, plagioclase, the main constituent of black granite (60%), had approximately the same ability to induce ROS as the quartz-containing fractions of red and grey granite . CONCLUSIONS - The cancer morbidity and mortality figures of the three different granite areas, com bined with the found differences in biological activity of granite dusts and a hypothesis that there is a cancer-inducing mechani sm for ROS, point to a direct role for quartz in cancer induction. mortality, quartz, reactive oxygen species, silica.


The International Agency for Research on Cancer
(lARC) concluded in June 1986 that there was sufficient evidence of carcinogenicity for crystalline silica in experimental animals, but the evidence was insufficient to show carcinogenicity in humans (I).
In March 1987, IARC collected new epidemiologic information on carcinogenic risk from exposure to silica (2). There were still serious difficulties to interpret epidemiologic results based on different study populations varying in , for example, study design, cohort definition, period of exposure, follow-up, quality of data on exposure history, and so forth. (3) . Neither of the following hypotheses could be verified exclusively: (i) exposure to silica is a direct cause of both silicosis and lung cancer and (ii) exposure to silica is an indirect cause of increased lung cancer risk through the development of silicosis, which would have a promoting effect on the carcinogenic process (2).
The pathogenesis of chronic lung disease induced by granite dust (including lung cancer) is dependent on the alveolar deposition of the dust. The role of silica-induced cytotoxicity, macrophage destruction, and the subsequent proliferative collagenous fibrosis for the possible development of lung cancer is not clear (4,5). The host responses including immunologic factors related to long-term exposure to min-eraI dusts and the possible role of reactive o xygen specie s produced by inflammatory cells exposed to mineral particles have been suggested to playa role in carcinogenesis (I , 6,7).
The aim of the present study was to evaluate the possibility of a direct association between silica dust exposure and lung cancer.
The study included the cancer mortality and morbidity of a cohort of 1026 granite workers from 1940 to 1989. In addition regional census data in 1970-1985 and lung cancer incidence data were linked.
Second, in vitro biological cell tests were performed with mineral fractions of different granites.

Study populations
The epidemiologic part of the study comprised the following two different study populations: (i) a cohort of 1026 Finnish granite workers employed in quarries and processing yards in three main g ra nite Material and methods. The cohort co nsis ted of the workers who had started granite work betwe en 1940 and 197 I and had wo rked for at least three months. The data on the subjec ts and their expos ure histories were collec ted fro m the employers' personnel records and fro m the questionnaire carried out in 1986. The data on diagno sed diseases were obtaine d from the questionnaire and from medical reco rds . The causes of death were ascertaine d from the death certificates.
The mortality of the cohort had been followed earlier in 1972 (9), 1975 (10), 198 1 (II, 12), and 1985 (13). The present study was based on the results of a mortality study carried out in 1985 (13). The main results on cancer mortality in 1940-1985 can be briefly summariz ed as foll ows.
The expec ted numbe rs were based on the national figures for men in the median year of the dea ths in the cohort (1975) ( 14). The age-specific observe d numbers of deaths and the expected numbers of deaths according to nation al figur es were computed for the different causes of death . The use of the median year was base d on separate calculatio ns for the median year of deaths and mortality rates specific for five-year calenda r peri ods (15,16). The difference between these calculat ions was negligible. The lung cancer morta lity rates were also ascertained for this cohort of granite workers by co mputing the expected numbers first in parallel using the mortality rates specific for five-year calendar periods and the median year rate s (12). The differe nce between the expected numbers based on the median-yea r calc ulation and the rates specific for calendar period ranged from 0.0 to 0.8. The observed numbers were tested by means of the Poisson distribution model with the expec ted numbers ( 17).
A reanalysis of the mortality by regions, 1 9 40-1 9 8 5 A nested case-referent study on l ung cancer Lung c ancer mor bidit y by regions, 1 9 5 3 -19 8 7 Morta l!t y by regions, 1 9 4 0 -19 8 9 Cohort mor ta l i ty, 1 9 4 0 -1 9 8 5 Expo sure Dust exposure was measure d in the Finnish granite industry in 1970-1972 (8). The geometric mean of the total dust concentration ranged from 1.7 to 39.8 mg· m J and that of quartz dust from 1.0 to 1.5 mg . m'. The highest concentratio ns were found in drilling, being about ten times the hygien ic standard for quartz .
Complete occ upatio nal histories of the cohor t members were obtai ned from the questionn aire carried out in 1986. Using dust measurement s and lifelong exposure histories, an industria l hygienist estimated both the amount of total dust and quartz dust blindly for the cases and referent s.
Confounding occupationa l exposure during a lifetime , as well as smok ing, was chec ked through the questionnaire data and controlled in the data analysis.
areas of Finland and (ii) census-based target populations from the same three regio ns. The study was carried out stepwise accordi ng to the study design s presented in figure I.

-1985
and cancer mor bid i t y , Nested case-refe rent study on lung cancer Materia l and methods. All of the 31 lung cancer cases observed in the cohort as a primary cause of death were include d in a nested case-referent study. Two referents (one living and one dead ; the date of death had to be later than that of the case) without cancer were selected for each case. Both of the referents had to be of the same age (±2 years) and to have entered granite work at the same time (±4 years) as the case. Th e quality of information on lifetime exposure had to be similar for the case and referent. Lifelong (cumulative) quartz dust exposure (the sum of the length of exposure in each job title mult iplied by the job-specific quartz dust concentrations during the corresponding calendar periods) wa s compared between the cases and referents.
Results. The 3 1 lung cancer ca ses were not more heavily exposed than the ir 62 matched pair s. The mean of the lifel ong concentration of quartz dust was 14.4 mg . m' for the cases and 18.9 mg· m' for the referents. In the pair comp arison only nine cases had a clearly higher exposure category than the indi vidual referents. The mean length of exposure time was 19 years for the cases and 22 years for the referents. However, the cases and referents were distributed asymm etrically according to employer. Of the cases 48% came from workplaces from the Vehm aa region and 52% from the Kuru regio n, and none fro m the Viitasaari region. Of the referents 55% came from workplaces in the Vehmaa reg ion, 40 % from Kuru , and 5% from Viitasaari,

Analysis of cohort mortality by regions
Material and methods. Bec ause of the asymmetry in the regional distribution of lung cancer case s and referents, the whole cohort was reanalyzed by regions so that the potential differences in regional lung cancer mortality co uld be determined. The statistical methods were the same as those used in the analysis of the whole cohort, (See the section Cohort Stud y on Lun g Can cer in . Results. The risk for lung cancer was 1.3-fold (SMR 126 , 95% CI 71-208 ) for the workers in Vehm aa (observed 15, expe cted 11.9), and 2. l -fold (SMR 2 11, 95 % CI 120-342) for the workers in Kuru (observed 16, expected 7.6 ) in comparison with that of the general male population (table I ). No ca ncer cases, even other than lung cancer, were obser ved among the workers in Viit asaari. The agespecific smoking habits were similar in the thr ee reg ion al cohorts. There were no statistically sign ificant differences even in the age-specific numbers of heavy smokers (on an average ;::20 cigarettes a day ).
In Vehmaa 10 of the 15 primary and 4 of the 5 seco nda ry lung can cer cases occurred afte r at least 25 yea rs of follow-up, whereas only 6 of the 17 lun g cancer cases in the Kuru region had such a lon g laten cy (table 2).
The type of granite was Balmoral red in Vehmaa, grey in Kuru, and black in Viitasaari. The amount of quartz in the mineral was 36% in the Balmoral red of Vehmaa, 31 % in the Kuru grey, and 0% in the Viitasaari black (table 3).
Regional case-referent study on lung cance r Material and meth ods. A detailed ana lysis of lifelong granite dust ex pos ure wa s made from the region-based case-referent study . Three in the case group had also been workin g in foundries or in welding ; therefore they were excluded from this analysis because of the confounding exposure. Thu s 28 cases (14 cases from Vehmaa and 14 from Kuru) and one referent (living or dead) matched for age (±4 years) and smoking for each case were selected for the region-based case-referent study .
Lifelong (cumulative) qua rtz dust exposure (calculated as in the section Nested Case-Referent Stud y on Lung Canc er ) was compared between the cases and referent s in the Vehmaa and Kuru regions (a case expose d more than its referent receiving + and a case exposed less than its referent rece iving -). The two regional distribution s of the + and -figures were  Respiratory di sease s 26 10.5"" 14 6.9" 0.5 40 17.8""" " P<0.05, "" P<0.01, """ P<0.001 , Poisson distribution.

409
Scand J Work Environ Health 1994, vol 20, no 6  Results. The analysis of lifelong granite dust exposure showed that the cases from Vehmaa were more exposed than their age-and smoki ng-matched referents, whereas the cases fro m Kuru were less exposed than their referents. The difference between the se regional figure s was statistica lly significant (P < O.05, X 2 test) .

10
Regional case-case analysis Material and methods. Differences in exposure were also analyzed betw een the Vehm aa and Kuru lung cancer cases (10 age-matched case-pairs). Lifelon g quartz dust exposure (calculated in the same way as in the preceding section) was compare d between Kuru cases and Vehm aa cases with the aid of the Wilco xon test for two samples (17).
Results. The comparison of the lifelong quartz dust exposure betw een the Vehmaa and Kuru lung cancer cases (10 age-matched case-p airs) showed that the cases from Kuru were statistica lly significa ntly less exposed than the cases fro m Vehmaa (P<O .O I, according to the Wilco xon test).  Material and methods. Th e regional mortality results of the cohort were updated until the end of 1989. The calc ulations of the exposure and latency year s wer e based on the co mplete occupational history obtained from the questionnaire in 1986. The expected numbers were based on the national figures for men in the median year of the death s in the cohort (1977) (18) . The obs erved numb ers were tested agai nst the expected numb ers by the Poisson distributi on mod el (17).

Follow-up of the cohort mortality by regions in
Results. By the end of 1989 the numb er of personyears was 26 113, and 363 deaths had been observed (332.1 expected) (ie, 67 deaths more than at the end of 1985) (   (9.7 expected, SMR 175,95% CI 102-281), and one from the Viitasaari black granite area (0.7 expected). The risk for lung cancer increased as the years of exposure and latency increased (table 6). In the lower latency categories the rates were higher for the Kuru region than for the Vehmaa region.
Lung cancer morbidity in the coho rt in 1953-1987. Material and methods. The same cohort was followed for the study of cancer morbidity from the Finni sh Cancer Register in the years 1953-1 987. The observed numbers were compared with the corresponding national figures registered in the cancer register during the same period.
For the study of cancer morbidity in the cohort the expected numbers of primary cancers were calculated on the basis of national figure s by five-year calendar periods. The observed age-s pecific and causespecific numbers of cancer cases were tested by the Poisson distribution model (17) against the expected numbers.
Results. By the end of 1987, 43 lung cancers had occurred in the cohort , the expected number being 25.3 (SMR 170, 95% CI 123-229). Twe nty-five cases were from Vehmaa (15.4 expected, SMR 162, 95% CI 105-240), 17 cases from Kuru (9.4 expected, SMR 181,95% CI 105-290), and only one case came from Viitasaari (0.5 expe cted). The lung cancer morbidity rates for Vehmaa and Kuru are shown in table 7. There was no statistically significant difference in the lung cancer incidence rates between the Vehmaa and Kuru regions, but the latency was shorter for Kuru (figure 2).

Cancer mortality and morbidity in the target populations
Material and methods. To confirm the results derived from the cohort study, we studied cancer mortality Scand J Wo rk Enviro n Health 1994, vol 20, no 6  and morbidity in the regio nal target populations. The source material of the Finnish regio nal age-and cause-specific cance r mortality statistics (from Statistics Finland) were analyzed from 1962 to 1985. We studied the morbidity from lung cancer by linking the regional census data (from Statistics Finland) from the years 1970,1975,1980, and 1985 and the lung cancer incidence data (from the Finnish Cancer Registry) for the peri od of 1970-1 985. The incidences were compared between granite workers and workers in other occ upations . The classification of occupations was based on the census data. The differences in the age-specific cancer mortality and morbidity rates between the region al target populations were tested with the Mantel-Haenszel chi-square test for incidence data (19) .  ....................................... .  Results. The Finnish regional mortality statistics were studied from 1962 to 1985. The age-specifi c cancer mortality rates for Viitasaari were similar to those for Kuru and Vehmaa (table 8).
In the regional study on lung cance r morbidity the regio nal census data from the years 1970, 1975, 1980, and 1985 and the lung cancer incidence data for the period of 1970-1985 were linked. The cancer incidence rate for the red-granite workers was statistically significantly higher than that for the remaining male population in that area (figure 3). The same was true for the grey -gra nite workers . The blackgranite workers, however, did not have an excess lung cancer incidence (only two cases among the exposed).

Discussion
Excess mortality fro m lung cancer has been repeatedly observed in this cohort of granite workers exposed to pure silica. Similarly the cancer morbidity study showed an excess risk for lung cancer. The workers had long periods of exposure to granite dust [mean 12 (range 0--49) years). The period of follo w-up  was 50 years (variation 18-50 years). The risk for lung cancer increased with the length of exposure and growing years of latency.  Region Figure 2. Age -standardized lung cancer mo rtality rate s ) and morbid ity rates  in th e cohort of the granit e workers (expre ss ed as incidence rates 1/100 000 ). The exposu re and laten cy ca lculations hav e been based on the emp loye rs' rolls .  Excess lung cancer mortality and morbidity was found among the workers finishing quartz-containing granite (Vehmaa Balmoral red and Kuru grey ). The black granite in Viitasaari did not contain any quartz. However, the subcohort of blackgranite workers was too small to give evidence of a hypothetically lower cancer risk in this quartz-free area .
The granite workers had no remarkable exposure history to potential concomitant carcinogens ( 13). Neither did smoking alone explain the exces s risk for lung cancer. The smoking habits of the granite workers were similar to those of other Finnish male workers. No essential differences were observed in the smoking habits between the regional subcohorts, not even in the number of heav y smokers.
Methodological modifiers (eg , variation in the year s of entry into the cohort) may have affected the occurrence of cancer and other latency-dependent diseases. This possibility has been thoroughly discussed earlier (13). The effect of structural differences in the regional subcohorts was further controlled in the present study.
The cohort was restricted to include only workers who entered granite work in 1962 at the earliest (the year from which the data wer e identically available from all the workplaces). Both the lung cancer morbidity and mortality remained highest for the grey-granite subcohort.
The mean exposure time was 7.3 years for the black-granite workers and about four years longer for the red-and grey -granite workers. On the other hand, 73 % of the black-granite workers had a latency of at least 24 years; 55 % of all the lung cancer cases in the red-and grey-granite subcohorts occurred within 24 years .
Shorter lifelong exposure times, as well as shorter periods of latency for lung cancer, were observed for the grey-granite workers in Kuru . The physico-chemical composition of the dusts (size, particle surface characteristics, adsorption features , etc ) (5,7,20) may explain the regional differences in both the risk of lung canc er and the laten cy .
Because the subcohort of the black-granite area in Viitasaari was very small, all available additional material was obtained, and further internal analyses were made for the cohort.
The study on the Finnish regional mortality statistic s showed that the age-specific cancer mortality rates for Viitasaari were similar to those for Kuru and Vehmaa. Thus our result did not depend on the fewer cancer cases in one of the target populations, namely, in the black-granite area.
The analysis of lung cancer morbidity in the regional census material gave results simil ar to those of the cohort study. Consequently, the black-granite workers who had been excluded from our cohort did not show an increased risk for lung cancer.
Because of the small sample size, a potential lung cancer risk may occur later in the black granite area among the persons not included in our cohort, especially among persons with silicosis. Therefore, the Finnish Register of Occupational Diseases in 1964-1991 was checked for the black granite area . Altogether 13 occupational diseases were found ; only two of them were silicosis.
The prevalences of self-reported diagnosed cardiovascular diseases and diagnosed musculoskeletal diseases, except rheumatoid arthritis, were similar in all of the subcohorts. Of the 45 cases of rheumatoid arthritis observed in the whole cohort, 22 originated from Vehmaa, 13 from Kuru , and none from Viitasaari, The prevalences of diagnosed pneumonia, bronchitis, and bronchial asthma were similar in the subcohorts, The prevalences of diagnosed lung tuberculosis, pleurisy, emphysema, and silicosis were clearly lower in the black-granite subcohort, than in the red-and grey-granite subcohorts, These low prev-alence s of chronic lung diseases and rheumatoid arthritis among the black-granite workers were ascertained from the medical records. The low prevalence of chronic lung diseases and rheumatoid arthritis suggests a different disease-generating potential for the quartz-free black granit e (13. 2 1).
The co-occurrence of lung cancer and pulmonary silicosis was rare . In the mortality data for 1940-1989 four lung cancer cases also had a diagnosis of silicos is mentioned on the death certifi cate or in the disability register. Altogether 37 cases of silicosis were recorded on the death certificate or in the disability register for the whole coho rt of 1026 granite workers; only eight of them had lung cancer. The age-standardized incidenc e rate (SIR) (lO ll 00 000) was similar for the cohort members with diagnosed silicosis (SIR 209) and for those without diagnosed silicosis (SIR 192), and similar even for those with a minimum latency of 20 years (being 258 and 220, respectively). Among the 37 silicotic patients 31 cases of silicosis originated from Vehmaa (31/579 = 5.4 %), eight having lung cancer, and six came from Kuru (6/340 = 1.8%), one having lung cancer.
Increased mortality from gastrointestinal cancers (12,13) and excess morbidit y from certain nonmalignant disease patterns (eg, rheumatoid arthritis) (11, 21) were obser ved in the earlier follo w-ups, Inhalation of organic and inorganic material will also cause a variety of immediate systemic consequences. Deposition of dusts within the respiratory tract, local and systemic immunologic responses, toxic reactions , decreased resistance to infecti ous diseases, or translocation of foreign material may explain such increased nonrespiratory morbidity (22)(23)(24).
The following results of the present study and the earlier follow-ups support the hypothe sis that silica per se influence s the development of lung cancer; (i) exce ss morbidity and mortality from lung cancer among the workers exposed to pure silica, (ii) increased cancer risk for the workers finishing quartzcontaining granite, (iii) long period s of exposure to quartz, (iv) long periods of follow-up and the latency dependency of the disease, (v) no remarkable concomitant carcinogens (confounding exposures and methodologi cal modifier s being controlled), (vi) cooccurrence of lung cancer and silicosis rare, (vii) increased mortality from cancer of extra pulmonary sites (gastrointestinal cancers), and (viii) excess morbidity from rheumatoid arthrit is.

In vitro experiments with fractionated dusts of red, grey, and black granite
The inflammatory and carcinogenic potential of mineral dusts has been linked to the feature of dusts to cause cytotoxicity or to generate reactive oxygen species in their target tissues (6). With two tests we studied the proinflammatory differences between dust samples of red, grey, and black granite in vitro.

Material and methods
Mineral dusts were obtained from the different granite samples by grinding, particle s with various densities were separated by liquid density separation, and the minerals contained in the density fraction s «2.63,2.63-2.75, and >2 .75) were determin ed for the red, grey, and black granite . The Kuru grey granite was 80% feldspar , <2 0 % quartz and some biotite in the density fraction of < 2.63; 70% quartz, 25% feldspar, and <5 % biotite in the 2.63-2.75 density fraction; and >90% biotite and <5 % feldspar in the > 2.75 density fracti on. Of the particle s, 59-76% had a diameter of less than 10 urn . The Vehmaa red granite was <90% feldspar and <10% quartz in the density fraction of <2 .63; >75% quartz, <25 % feldspar, and <2% biotite in the 2.63-2.75 density fraction; and> 85% biotite, < 10% feldspar, and <5% quartz in the >2.75 density fraction . Some 64-74% of the particles had a diameter of less than 10 urn. The Viitasaari black granite contained practically no material with a density fraction of <2 .63; in the density fraction of 2.63-2.75 it was predominantl y plagioclase; and in the density fract ion of >2.75 it was augite, enstatite, and biotite. Of the particl es, 74-82% had a diameter of less than 10 urn.
The mineral fractions were analyzed for their cytotoxic potential, lactic dehydrogenase (LDH) release of rat macrophages in the presence of I mg . ml'" of dust (25), and their capacity to generate reactive oxygen species in human polymorphonuclear leukocytes with a chemiluminescent method (26).
In the cytotoxicity tests the quartz-containing fractions (density 2.63-2.75) of grey and red granite caused the strongest LDH release ( figure 4). The differences between the fractions were, however, small.
In their capacity to induce the production of reactive oxygen species in human polymorph onuclear leukocytes, the strongest activit y of grey and red granite was also seen in the quartz-containing fractions (density 2.63-2.75) (figure 5). However, similar reactive oxygen species production was also seen with the quartz -free fraction of black granite. This activity was caused by the almost pure plagiocla se fraction. Compared with the activity of quartz, the activity of 100 ug . ml' granite-dust fractions corresponded to 25-50 ug . ml' of purified DQ-quartz (figure 5).

Discussion
No consistent pattern of biological responses (ie, cytotoxicity or activity inducing reactive oxygen spe-% lDH release 25,--------------------, 20   10 cies) could be observed for the different granites and their mineral fractio ns. The quartz-containing fractions of granite seemed to be the most active component. This finding may be related to in vivo consequences , and it is in keeping with (or does not contradict) the epidemiologic findings from the quartzcontaining grey and red granite areas. That these in vitro biological responses could be related to the epidemiology of cancer in the quart z-free black granite area must so far be considered purely speculative. Although the quartz-free fractions of the black granite were less cyto toxic than the corresponding fractions of the red and grey granite, the membrane effects, as measured by the generation of reactive oxygen species, were very similar. This activity corresponded to the presence of plagioclase. The in vivo effec ts of plagioclase (calci um-con taining feldspar) are not know n. The inflammato ry and carcinogen ic potential of various inhaled dusts have been related to pulm onary depositi on, the surface characteristics of the particle s, inter action with biol ogical membranes, and possibly with their capacity to adsorb chemical carci nogens. The relationship between the present epidemiologic and experimental results must, naturally, be interpreted with caution, but neverthele ss it indicates new directions for the study of regional respiratory morbidity.