Statewide cancer registries offer researchers a source of high quality, detailed information on cancer diagnosis. They also create an opportunity to apply new methods to use this information in the evaluation of cancer risk in occupational cohorts. The aims of this study were to: (1) determine the feasibility of utilizing statewide cancer surveillance systems in the evaluation of cancer incidence within occupational cohorts; (2) compare and contrast the relative merits of standardized incidence ratios with standardized mortality ratios as determined from cancer surveillance incidence data and death certificate mortality data, respectively; and (3) provide recommendations concerning how and under what circumstances statewide cancer registries should be utilized in the evaluation of occupational cohorts. Two occupational cohorts, the Highway Worker cohort (N=3497) and Mineral Board Worker cohort (N=5086), were used to conduct standardized mortality ratio (SMR) and standardized incidence ratio (SIR) analyses for 1988-1996. Cancer mortality data were obtained from state mortality records and the National Death Index. Cancer morbidity data were obtained from the Minnesota Cancer Surveillance System, a statewide cancer registry. Because the cancer registry only collects information for Minnesota residents, it was necessary to conduct record-linkages with a variety of information sources to determine Minnesota residency status for each worker for each year from 1988 to 1996. Three models were tested to determine the effect of different residency status assumptions on the SIR estimates (i.e., follow-up bias). SMR and SIR estimates and their 95 percent confidence intervals were calculated. Over the nine years of follow-up (1988-1996), for male mineral board workers, there were 209 cancer diagnoses and 104 cancer deaths. For male highway workers, there were 297 cancer diagnoses and 153 cancer deaths. For female mineral board workers there were 27 cancer diagnoses and 11 cancer deaths. Difficulties in validating the residency history occurred more frequently among older workers, females, and those with earlier dates of employment. Incomplete information concerning full name, date of birth, and social security number increased the difficulties in determining residency. Of the three models tested, use of out-of-state migration rates to complete unknown residency status was the best method for completing the residency history. For cancers with long-term survival, such as prostate cancer, melanoma of the skin, and urinary bladder cancer the SIR was higher than the SMR. Among male highway workers, however, the SIR, for bladder cancer was much lower than the SMR. For cancers with very short relative survival such as lung, pancreatic and stomach, the SMR and SIR were similar. For male mineral board workers, however, lung cancer SIRs were higher than the SMRs. Differences were noted between the SMR and SIR for which cancers reached the usual statistical significance (p<0.05) including colorectal, prostate, lung, urinary bladder, non-Hodgkin's lymphoma and all cancers. Standardized incidence ration analyses are a potentially useful tool for examining the rate of cancer within an occupational cohort. Cancer mortality and incidence data provide very different perspectives about potential cancer issues within occupational cohorts. The potential problem of follow-up bias is substantial however, and such analyses should only be undertaken after careful consideration. If both SMR and SIR analyses are conducted, differences or similarities between their estimates can be evaluated to determine if a priori study expectations are met regarding cancer diagnosis and mortality.
Debora Boyle, DVM, PhD, Chronic Disease and Environmental Epidemiology, Minnesota Department of Health, Minneapolis, MN 55440-9441
Chronic Disease and Environmental Epidemiology, Minnesota Department of Health, Minneapolis, MN