British Journal of Industrial Medicine 1990;47:4-9

Silica exposure, silicosis, and lung cancer: a necropsystudy

P A Hessel, G K Sluis-Cremer, E Hnizdo

AbstractRecent studies of the association between lungcancer and silicosis and silica dust have beeninconclusive; some showing positive associa-tion and some showing none. The present studymatched 231 cases of lung cancer with 318controls by year of birth. Subjects were selec-ted from the necropsy records of the NationalCentre for Occupational Health. Data onintensity and duration of exposure to silicadust were obtained from personnel records.Presence or absence of lung cancer and thepresence and severity of silicosis of the paren-chyma, pleura, and hilar glands weredocumented from necropsy reports. Smokingdata were abstracted from records of routineexaminations. No case-control differenceswere noted for any of the exposure indicatorsincluding cumulative dust exposure, totaldusty shifts, weighted average intensity ofexposure, total underground shifts, and shiftsin high dust. Similarly, no association wasfound between lung cancer and the presence orseverity of silicosis of any site. Stratifiedanalyses showed neither significant nor sug-gestive trends when case-control comparisonsfor silicosis were examined by level of dustexposure or smoking. Reasons for disparitybetween these results and those of some otherstudies may include concomitant exposures toradon daughters, asbestos, diesel emissions,and cigarette smoking; idiosyncracies of thecompensation process; and the possibility of athreshold in the relation(s).

The possibility of an association betweenoccupational exposure to silica and a raised risk oflung cancer has been explored recently using variousstudy designs and worker populations. Similarly, thepossibility of an association between silicosis andlung cancer has been investigated. Despite this level

of interest, the questions of the carcinogenicity ofsilica and an association between silicosis and lungcancer remain unanswered.A recent International Agency for Research on

Cancer monograph evaluated the existingexperimental and human data and concluded thatthere was sufficient evidence for the carcinogenicityof crystalline silica in experimental animals and thatthere was limited evidence for the carcinogenicity ofcrystalline silica in man.' There was inadequateevidence for the carcinogenicity of amorphous silicain either animals or man. Limited evidence in manimplies that a causal interpretation is credible butthat alternative explanations, such as chance, bias, orconfounding, could not adequately be excluded.McDonald, commenting on the interpretation of thehuman studies stated that "the epidemiological find-ings are not consistent, risk estimates are generallylow, exposure response has not been studied, and thepossibilities for confounding by other carcinogens,including tobacco, are many."2The present case-control study follows a previous

study in which no association was found betweenlung cancer and either silicosis or exposure to silicadust in South African gold miners identified from therecords of their pension fund.3 The present studyextends that work, identifying cases and controlsfrom the computerised records of necropsies perfor-med on white gold miners at the National Centre forOccupational Health in Johannesburg. Both the roleof silica as a lung carcinogen and the associationbetween silicosis and lung cancer are explored.

Materials and methodsCurrent legislation in South Africa states that anyattending physician who knows or suspects that hisdead patient was a miner is obliged to remove thecardiorespiratory organs and send them to theMedical Bureau for Occupational Diseases ifpermis-sion is granted by next of kin. Pathologists at theNational Centre for Occupational Health examinethe organs. Although adherence to the law is notcomplete, estimates for the gold mining industryindicate that postmortem examinations are perfor-med on approximately 86% of white gold miners.3Since 1975 the results of these examinations havebeen computerised using the pathology automation

Epidemiology Research Unit, PO Box 4584, Johan-nesburg 2000, South AfricaP A Hessel, G K Sluis-Cremer, E Hnizdo

4

on June 13, 2021 by guest. Protected by copyright.

http://oem.bm

j.com/

Br J Ind M

ed: first published as 10.1136/oem.47.1.4 on 1 January 1990. D

ownloaded from

http://oem.bmj.com/

Silica exposure, silicosis, and lung cancer: a necropsy study

system.4 More than 3000 necropsies a year have beenrecorded in this way.

All white miners exposed to silica-that is, thosewith at least 80% of their mining employment insilica containing mines, usually gold mines, withcarcinoma found in the lungs at necropsies-wereselected as potential cases (n = 571). An equalnumber of white miners who had been exposed tosilica was selected as potential controls, stratified byage at death according to the age distribution of thecases. Miners who died between January 1979 andOctober 1983 (the period covered by the previousstudy3) were excluded to ensure independence of theresults of the two studies.The medical files of the Medical Bureau for

Occupational Diseases which included the records ofperiodical (usually annual) examinations, specialistexaminations for the purpose of assessing eligibilityfor compensation, detailed work histories, and theresults of the necropsies were examined. Cases wereexcluded if it was found that their carcinomas werenot bronchogenic (n = 134). Most were statedspecifically to have been metastatic; however, thoselisted as having a primary tumour ofunknown originwere also excluded. Thirteen cases and 35 controlswere excluded because no smoking information wasavailable in the medical files. Miners who hadworked fewer than 1000 shifts (about four years) insiliceous dust were excluded (48 cases, 63 controls) aswere those with more than 20% of their service innon-silica containing mines (15 cases, nine controls).Miners who worked for more than one year in anasbestos mine (four cases, three controls) or asboilermakers (16 cases, 11 controls) were alsoexcluded. Medical Bureau for Occupational Diseasesfiles could not be found for 11 cases and 17 controlsand the files of two cases and one control containedno information. One case and one control were foundto have died outside the period of interest and fourminers originally selected as controls were found tohave had bronchogenic carcinomas that had beenresected in life and therefore had not been documen-ted at necropsy. These four were reclassified as cases.The remaining cases and controls were matched

randomly by age at death. Nine controls could not bematched and were excluded. The remaining studygroup comprised 231 cases and 318 controls; eachcase having at least one control and none having morethan two.

Cumulative exposure to silica dust was calculatedby multiplying the number of shifts worked in adusty atmosphere by a weighting factor proportionalto the dust concentration of each occupation.5 Theintensity of exposure in different occupations(weighted by the number of shifts at each level ofintensity) was calculated. Occupations were coded asnon-dusty, low, moderate, or high dust and weighted0, 3, 6, and 12, respectively. Shifts at high dust

were also considered separately. The intensity ofexposure was derived from the work of Beadle whoperformed full shift estimates of exposure in arandom sample of 650 men in different occupationsin 20 Witwatersrand gold mines in 1966.5The assessment of silicosis was obtained from the

necropsy report. The diagnosis of silicosis atnecropsy is a medicolegal diagnosis and representsthe opinion of the pathologist who has access to theoccupational history of the worker at the time of thepostmortem examination. The diagnosis of slightsilicosis may be made on the basis of a single nodulein the parenchyma. The assessment of silicosis atnecropsy was made with the knowledge of whetherthe miners had lung cancer or not. This was unlikelyto cause bias as the necropsies were conducted beforethe start of the study and the pathologists were notaware of the objectives of the study. Silicosis of theparenchyma and pleura was graded by the patholog-ists as none, minimal, slight, moderate, or pronoun-ced. Silicosis of the hilar glands was graded as none,slight, moderate, or pronounced.

Cases and controls were classified into five smok-ing categories representing their smoking status atdeath (or as near to death as possible) based oninformation in the Medical Bureau for OccupationalDiseases files as: (1) lifelong non-smokers; (2) lightex-smokers, smoking 15 or fewer cigarettes orequivalents a day assessed 10 years before death ortwo or more years before quitting, whichever wasearlier and having quit at least two years before death;(3) heavy ex-smokers, as for light ex-smokers butsmoking 16 or more cigarettes or equivalents a day;(4) light current smokers, smoking 15 or fewercigarettes or equivalents a day assessed 10 yearsbefore death and not an ex-smoker, and (5) heavycurrent smokers, as for light current smokers butsmoking 16 or more cigarettes or equivalents a day.

In addition, the number of cigarettes orequivalents smoked a day at five, 10, 15, and 20 yearsbefore death (within two years) were recorded whenavailable. One gram of pipe tobacco was consideredequivalent to one cigarette and one cigar equivalentto five cigarettes.

Univariate comparisons were performed using thet test for continuous variables and the chi-squaredtest, and the chi-squared test for trend for categoricalvariables. Multivariate comparisons of silicosis andlung cancer were first performed by stratification,looking for trends in the odds ratios, and by usingconditional multiple logistic regression, controllingfor smoking, cumulative exposure to silica dust, andage. Odds ratios for lung cancer and silicosis werecalculated for each quartile of cumulative dustexposure. These odds ratios were tested for equalityusing the Breslow-Day test for homogeneity. Man-tel-Haenszel summary odds ratios were also cal-culated with 95% confidence intervals.

5

on June 13, 2021 by guest. Protected by copyright.

http://oem.bm

j.com/

Br J Ind M

ed: first published as 10.1136/oem.47.1.4 on 1 January 1990. D

ownloaded from

http://oem.bmj.com/

Hessel, Sluis-Cremer, Hnizdo

Table I Description of the study population

Cases Controls

No Mean SD No Mean SD p

Total subjects 231 318Age at death 231 65-6 7-8 318 65-1 8-0 0 44Average cig/day:

5 years to death 202 12-4 10-8 278 8-2 10-0 < 0-000110 years to death 210 14 7 10-9 291 11-3 10-8

Silica exposure, silicosis, and lung cancer: a necropsy study

Table 4 Comparison of cases and controls for presence ofsilicosis of the parenchyma by quartile of cumulative dustexposure (CDE)

Cases ControlsCDEquartile Silicosis No (Go) No (%) OR

1 Yes 17 (29 8) 15 (18 8) 1.84No 40 (70 2) 65 (81 3)

2 Yes 26 (45-6) 41 (51 3) 0.80No 31 (54-4) 39 (48 8)

3 Yes 42 (68-9) 54 (70 1) 094No 19 (31-2) 23 (29 9)

4 Yes 39 (69-6) 53 (65 4) 1.21No 17 (30 4) 28 (34 6)

Total 231 (100) 318 (100)

ORM-H = 1 1, 950o CI = (0-77-1-58).XB-D = 272, df = 3,p > 020.Quartiles were determined by dividing the distribution ofcumulative dust exposure (for cases and controls combined) inquarters. The first quartile represents the lowest quarter of theobservations.

The association between silicosis and lung cancerwas also examined according to the number ofcigarettes smoked 10 years before death (table 7). Notrends were apparent and the summary odds ratiosdid not significantly differ from one.

DiscussionThe data presented above indicate no associationbetween lung cancer and either silica dust exposureor silicosis. Several characteristics of the studyshould be considered in evaluating these results.

(1) Though the necropsy rate among white SouthAfrican gold miners is approximately 86%,3 the nextof kin of miners who have been fully compensated inlife (implying at least 40% disability) cannot benefitfinancially as a result of the necropsy findings.Workers fully compensated in life are, therefore,underrepresented (but certainly not unrepresented)in the necropsied population. Silicosis is not uncom-monly found in fully compensated workers; however,

Table 5 Comparison of cases and controls for presence ofsilicosis of the pleura by quartile of cunulative dust exposure(CDE)

Cases ControlsCDEquartile Silicosis No (%) No (%) OR

1 Yes 4 (7 0) 11 (13-8) 0.47No 53 (93-0) 69 (86-3)

2 Yes 9 (15-8) 18 (22 5) 0-65No 48 (84-2) 62 (77-5)

3 Yes 22 (36-1) 21 (27-3) 1.50No 39 (63-9) 56 (72-7)

4 Yes 16 (28-6) 33 (40 7) 0-58No 40 (71-4) 48 (59 3)

Total 231 (100) 318 (100)

ORm-H = 0-79,95% CI = (0-52-1-19).X2B D= 4-67, df = 3, p = 0-20.Quartiles: see table 4.

Table 6 Comparison of cases and controls for presence ofsilicosis of the hilar glands by quartile of cumulative dustexposure (CDE)

Cases ControlsCDEquartile Silicosis No (%) No (%) OR

1 Yes 38 (66 8) 44 (55 0) 1.64No 19 (33-3) 36 (45-0)

2 Yes 49 (86-0) 64 (80-0) 1.53No 8 (14-0) 16 (20-0)

3 Yes 54 (88-5) 68 (88 3) 1.02No 7 (11-5) 9 (11-7)

4 Yes 51 (91-1) 76 (93-8) 0 67No 5 (8 9) 5 (6 2)

Total 231 (100) 318 (100)

ORM-H = 1 31,95% CI = (0 83-2 08).Ao-D= 1 78, df = 3, p > 0 20.Quartiles: see table 4.

much of the disability is thought to result fromsmoking. In the absence of silicosis 10-15 years ofdusty work is usually required before compensationis considered. As a result, the sampling frameprobably underrepresents gold miners who havebeen significantly disabled in life, owing in part tocigarette smoking but usually with appreciableexposure to silica. Because cases were selected on thebasis ofhaving lung cancer at necropsy and not on thebasis of smoking, it is unlikely that this affected theresults. In addition, the results obtained in this studyare comparable with those reported in our previousstudy which used the same workforce but in whichthe subjects were identified through pension fundrecords.' That sampling frame obviated anyproblems associated with underrepresentation offully compensated workers which might haveoccurred among the cases and controls in this study.

(2) The method of selection of cases and controlsresulted in the elimination from the control group ofsubjects with metastatic lung cancer. Subjects withany cancer of the lung were selected initially aspotential controls and those with metastatic lungcancers were then excluded. Cancers other than lungwere, therefore, underrepresented among controls.If these cancers were related to silica exposure orsilicosis the odds ratios calculated would have been

Table 7 Unmatched odds ratios for the comparison of lungcancer and silicosis by number of cigarettes a day reported 10years before death

Silicosis

Cigarettes a day Hilar glands Parenchyma Pleura

0 1-12 1 62 1-371-10 044 040 09811-20 0-61 0-84 0 92>21 0-75 0-88 1-17M-H* 0-68 0-81 1-0895% CI 0-40-1-14 0 55-1-21 0-60-1-96

*Mantel-Haenszel summary odds ratio.

7

on June 13, 2021 by guest. Protected by copyright.

http://oem.bm

j.com/

Br J Ind M

ed: first published as 10.1136/oem.47.1.4 on 1 January 1990. D

ownloaded from

http://oem.bmj.com/

Hessel, Sluis-Cremer, Hnizdo

overestimates of the true values. Neither silica norsilicosis have been linked to other cancers. Coronaryheart disease, which may have been slightlyoverrepresented among the controls (owing to theabsence of lung cancers and certain other metastaticlung cancers), had not been shown to be associatedwith silica exposure or silicosis in this population.6

(3) The use of dead controls is only valid if thecauses of death of the controls are not related to theexposure under study. Causes of death were re-viewed for all subjects and analyses were repeated,excluding controls and cases whose cause of deathwas listed as pneumoconiosis (one case, one control),tuberculosis (10 cases, six controls), nephritis (threecases, two controls), or obstructive airways disease.No changes in the results were noted.

(4) The possibility that miners who had beencertified for silicosis in life might change theirsmoking habits after certification (specifically, cutdown) was considered. Changes in smoking habitswere determined for those who were certified eitherfor silicosis or obstructive airways disease betweenfive and 20 years before death; the period for whichsmoking data were available. Changes in smokinghabits during the comparable period for those whowere not certified in life showed that those who werecertified for silicosis were less likely to decrease theirsmoking after certification (33% decreased) thanthose certified for obstructive airways disease (48%decreased), and were about as likely as those notcertified in life (37% decreased) to cut down.

(5) The lack of an association between exposure tosilica dust and lung cancer shown in this study isconsistent with most recently published reports.Nevertheless, the possibility that a threshold mayexist for exposure to silica and risk of lung cancershould be considered. The data presented in severalof the reported mortality studies indicate thatexposure levels were probably high in some studies.Twenty four per cent ofthe deaths among controls inthe population based study of Forastiere et al wereattributed to silicosis.7 Similarly, Zambon et alreported that 236 of 878 (27%) deaths in their studywere due to silicosis with a further 175 (20%)attributed to tuberculosilicosis-that is, nearly halfthe deaths reported were associated withoccupational exposure to silica.8 On the other hand,in a recent mortality study of 3971 white SouthAfrican gold miners neither silicosis nor tuberculosiscould be considered the underlying cause of death inany worker when all available information includingroutine and specialist examinations in life (whichincluded radiographs and lung function tests), post-mortem examinations, death certificates, and hos-pital records was considered.6 Mean exposure was0-2 mg/m' of quartz in that study with a range from0-1 to 0 8 mg/m'. In the present study the cause ofdeath was based on the necropsy results and the death

certificate. Pneumoconiosis was considered to be thecause of death for only one subject (a control) andtuberculosis the cause of death for four cases (1-7%)and six controls (1-9%). Possibly the dust concentra-tions in South African gold mines during the past 20to 30 years have been so low that an exposure diseaserelation cannot be found.

(6) The possibility was considered that the entrycriterion of at least 1000 dusty shifts, or employmentpractices and exposure characteristics, or all these, inthe South African gold mines might have resulted insuch limited variability in the exposure measures thatan effect could not be seen. The standard deviationsof the exposure measures listed in table 1 and theranges noted in the footnote indicate a fairly widespread of values for each exposure measure.Most studies exploring the association between

silicosis and lung cancer have shown positive associa-tions, whereas this study and its predecessor3 showedno such association. Possible explanations for such adifference have been suggested by Heppleston,9Archer et al,10 and McDonald,' and include a lack ofinformation on smoking and concomitant exposuresto radon daughters and polycyclic aromatichydrocarbons. It is worth while to explore these andother factors.

(1) Few data exist to evaluate the possible associa-tion between silicosis and smoking. An analysis oftwo cross sectional studies in white South Africangold miners produced odds ratios of 1-5 and 1-7 forever versus never smoking and the presence ofradiological silicosis (unpublished data). Studiesdesigned specifically to explore this association andincluding information on whether smoking and silicaexposure were simultaneous should be undertaken.If such an association exists the results of follow upstudies of silicotic subjects could be confoundedseriously.

(2) Another possible confounder to be consideredis asbestos. A recent study by Davies et al presented19 mesotheliomas among South African gold minerswho had no documented employment in the mining,milling, or processing of asbestos.1' Five were boiler-makers, three fitters, and three electricians. In all, 15of these 19 had evidence of possible exposure toasbestos, usually as artisans. In the present study 20of what would have been 251 otherwise eligible cases(8 0%) and 14 of what would have been 332 other-wise eligible controls (4-5%, OR = 1.97) wereexcluded because of a history of working in asbestosmines (four cases, and three controls) or as boiler-makers (16 cases, 11 controls). We are planning tocompare fibre levels in samples of lung tissue fromcases and controls in the previous study' to assess thepossible role of asbestos in the aetiology of lungcancer in South African gold miners. Investigatorswithout access to necropsy material might exploreoccupational histories in depth to search for sources

8

on June 13, 2021 by guest. Protected by copyright.

http://oem.bm

j.com/

Br J Ind M

ed: first published as 10.1136/oem.47.1.4 on 1 January 1990. D

ownloaded from

http://oem.bmj.com/

9Silica exposure, silicosis, and lung cancer: a necropsy study

of asbestos exposure, especially in case-controlstudies.

(3) The surveillance systems operating in manydusty industries enhance the likelihood of detectingcancers that might otherwise not be detected andbringing workers under intensive medical treatmentwhich might minimise their chance of dying fromanother cause in the interim. Given the usual even-tual outcome of lung cancer, this is probably not amajor problem as most cancers would be recognisedbefore death. In our previous study, however, severalpotential controls who had no indication of lungcancer on their death certificates were found to havehad bronchogenic carcinomas when the results oftheir necropsies were reviewed.3 It is not knownwhether these were carcinomas that were not diag-nosed in life or errors in completing the deathcertificates.

(4) Most follow up studies of silicotic subjects arebased on compensation records. Compensation isusually based on disability as well as a radiologicaldiagnosis of silicosis. These diagnoses of silicosis areoften not made blindly but rather (as is appropriate inthe medicolegal setting) with full knowledge andoften because of the occupational history. Otherfactors such as socioeconomic circumstances mayenter mito these decisions. Importantly, disabilitymust. often be shown before compensation is award-ed. Whereas exposure to silica has been shown tohave a deleterious effect on lung function,'2 currentlyavailable information indicates that simple silicosisper se does not have an effect [F J Wiles personalcommunication]. It is thought by many of thoseconcerned in the awarding of compensation in SouthAfrica that the disability documented in silicoticsubjects and other miners exposed to silica resultsmore from cigarette smoking than from silicosis. Ifthis is general, compensated silicotic subjects in othercountries may be those miners and other dustexposed workers who have been heavy smokers andhence are at higher risk for developing lung cancer.Considering this, and the possibility of anaetiological association between silicosis and smok-ing, it becomes apparent that smoking may seriouslyconfound studies of silicotic subjects either directlyor indirectly.

Future studies of silica, silicosis, and lung cancershould include individual smoking data that are asspecific as possible; attempt to identify likely orpossible exposures to asbestos, polycyclic aromatichydrocarbons, and other potentially confoundingexposures; quantify exposures to silica as accuratelyas possible; and use sensitive measures of thepresence of silicosis, preferably postmortemevidence.In addition to these studies, an attempt should be

made to determine whether an association exists

between silicosis and smoking. Where postmortemmaterial is available, an effort should be made tocompare fibre levels among cases of lung cancerwithout obvious (based on job titles) exposure toasbestos. The measurement of the effects of radiationexposures in these studies is usually difficult owing tosparse or absent historical data on exposures. Theseinvestigations should, however, be attempted.

It is unlikely that any single bias or error canaccount for reported discrepancies. Possibly,however, several of these sources of bias could beoperating simultaneously to produce apparentassociations. At present, viewing the growing body ofpublications, it is not yet possible to state withassurance that there is or is not an associationbetween lung cancer and either silicosis or silica dust.What is clear is that future studies must concemthemselves with the issues raised above if anyprogress is to be made in resolving this importantissue.

Requests for reprints to: Dr P A Hessel, Departmentof Health Services Administration and CommunityMedicine, 13-103 Clinical Sciences Bldg, Universityof Alberta, Edmonton, Alberta, Canada T6G 2G3.

1 International Agency for Research on Cancer. Monographs on theevaluation of the carcinogenic risk of chemicals to humans. Vol42. Silica and some silicates. Lyon: IARC, 1987.

2 McDonald JC. Silica, silicosis, and lung cancer. Br J Ind Med1989;46:289-91.

3 Hessel PA, Sluis-Cremer GK, Hnizdo E. Case-control study ofsilicosis, silica exposure and lung cancer in white SouthAfrican gold miners. Am J Ind Med 1986;20:57-62.

4 Hessel PA, Davies JCA, Goldstein B, Webster I, Hnizdo E,Landau S. Pathological findings in mineworkers. I: Descrip-tion of the PATHAUT data base. Am J Ind Med 1987;12:71-80.

5 Beadle DG, Harris E, Sluis-Cremer GK. The relationshipbetween the amount of dust breathed and the incidence ofsilicosis. In: Shapiro HA, ed. Pneumoconiosis,proceedings of theinternational conference, Johannesburg, 1969. London: OxfordUniversity Press, 1970:473-77.

6 Wyndham CH, Bezuidenhout BN, Greenacre MJ, Sluis-Cremer GK. Mortality of middle aged white South Africangold miners. Br J Ind Med 1986;43:677-84.

7 Forastiere F, Lagorio S, Michelozzi P, et al. Silica, silicosis andlung cancer among ceramic workers: a case-referent study. AmJ Ind Med 1986;10:363-70.

8 Zambon P, Simonato L, Mastrangelo G, Winkelmann R, Saia B,Crepet M. Mortality of workers compensated for silicosisduring the period 1959-1963 in the Veneto region of Italy.Scand J Work Environ Health 1987;13:118-23.

9 Heppleston AG. Silica, pneumoconiosis, and carcinoma of thelung. Am J Ind Med 1985;7:285-94.

10 Archer VE, Roscoe JR, Brown D. Is silica or radon daughters theimportant factor in the excess lung cancer among undergroundminers? In: Goldsmith DF, Winn DM, Shy CM, eds. Silica,silicosis and cancer: controversy in occupational medicine. (Can-cer research monographs, vol 2.) New York: Praeger,1986:375-84.

11 Davies JCA, Landau SP, Goldsmith C, Langton ME. Meso-thelioma risk among gold miners in South Africa. S Afr J Sci1987;83:184-5.

12 Wiles FJ, Faure HM. Chronic obstructive lung disease in goldminers. In: Walton WH, ed. Inhaled particles IV. London:Pergamon Press, 1977:727-35.

Accepted 10 July 1989

on June 13, 2021 by guest. Protected by copyright.

http://oem.bm

j.com/

Br J Ind M

ed: first published as 10.1136/oem.47.1.4 on 1 January 1990. D

ownloaded from

http://oem.bmj.com/