The potential carcinogenicity of crystalline silica to humans remains a controversial issue. The authors conducted an historical cohort mortality study of 2,342 male workers exposed to crystalline silica, predominantly cristobalite, in a diatomaceous earth mining and processing facility in California. During the years 1942-1994, mortality excesses were detected for nonmalignant respiratory diseases (NMRD) (standardized mortality ratio = 2.01, 95% confidence interval (CI) 1.56-2.55) and lung cancer (standardized mortality ratio = 1.29, 95% CI 1.01-1.61). NMRD mortality rose sharply with cumulative exposure to respirable crystalline silica; allowing for a 15-year latency, the rate ratio for the highest exposure stratum (> or = 5.0 mg/m3-years) was 5.35 (95% CI 2.23-12.8). The rate ratio for lung cancer reached 2.15 (95% CI 1.08-4.28) in the highest exposure category. These associations were unlikely to have been confounded by smoking or asbestos exposure. The findings indicate a strong dose-response relation for crystalline silica and NMRD mortality. The lung cancer results, although less convincing, add further support to an etiologic role for crystalline silica.

Annual decline in lung function determined longitudinally is often compared with predicted decline determined cross-sectionally. To test this comparison, spirometric data were collected 5 times over 5 yr from 52 adult male Caucasians. The age regression coefficient for FEV1 and FVC, determined cross-sectionally at each visit, was more than twice the longitudinal annual change computed from the same data as the mean of the slopes of each subject's regression lines. The discrepancy persisted even when the first visit was deleted to reduce learning effects on longitudinal estimates. This discrepancy may be partly explained by the sensitivity of cross-sectional analyses to past noxious influences, whereas longitudinal analyses are sensitive only to influences that continue to affect annual decline during the study period. We also found historical evidence of an increase in height-specific VC, which would artifactually steepen cross-sectionally determined regression lines. Thus, observed longitudinal changes of study cohorts should be compared with control longitudinal data.

A prospective mortality study of 839 men