Why did treatment rates for colorectal cancer in south east England fall between 1982 and 1988? The effect of case ascertainment and registration bias.

BACKGROUND: We had two aims in undertaking this study, as follows: (1) to describe regional and district trends in incidence and treatment for colorectal cancer in South East England from 1982 to 1988; (2) to examine the effect of registration practice and case ascertainment on district variations in incidence and treatment using data on death certificate only (DCO) registrations, mortality and stage.

METHODS: We included all cases registered by the Thames cancer registry diagnosed with colon or rectal cancer between 1982 and 1988 and resident in 28 districts in the two South Thames regions. Indirect standardized incidence ratios were calculated for the districts and a alpha 2 test for trend was carried out.

RESULTS: In the SE England regional analysis, between 1982 and 1988 there was a significant increase in the incidence of cases of colon and rectal cancer in the over-75s, but treatment rates remained unchanged. Treatment rates fell significantly in the under-65s although incidence rates remained unchanged. Age is a strong predictor of nontreatment. Between 1982 and 1988 the relative risk of not receiving treatment increased for all ages over 65 years. DCO registrations accounted for 22 percent and 15 percent of all colon and rectal cancer cases, respectively, between 1982 and 1988. The proportions rose (between 1982 and 1988) from 10 and 8 percent to 25 and 19 percent in colon and rectal cancer, respectively. DCO registration rates increased over time and in all age groups in South East England for both colon and rectal cancer between 1982 and 1988, but the largest increase was in the over-75s. Thirty-two per cent of colon and 25 per cent of rectal cases were unstaged. Although the proportion of unstaged cases remained constant over time, they were increasingly the result of DCO registrations. Errors in the registry staging data rendered those cases which were staged unusable. In the district analysis, there were significant variations in age-standardized incidence, treatment and DCO registration ratios across the 28 districts for men and women with colon and rectal cancer between 1982 and 1988. DCO registrations show a negative correlation with treatment for both colon and rectal cancer (p < 0.05) and with incidence for only rectal cancer.

CONCLUSIONS: We report significant differences in age-standardized incidence and treatment ratios across 28 districts in South East England, some of which, can be accounted for by differences in registration practice. There is a complex relationship between DCO registrations and incidence and treatment for both colon and rectal cancer. DCO registrations are a good proxy for under-ascertainment of incidence in rectal cancer but not colon cancer, and are a good proxy for under-ascertainment of treatment in both colon and rectal cancers. Information from the cancer registry can be used to examine registration and treatment rates across districts. However, if variations are to be adequately explained, meticulous data collection on stage and quality control are essential.