Changing rates of cancer. The hazards of using a single definition

Changing cancer rates, abstracted from tumor registries, are used to make inferences about the effect of carcinogens and cancer treatments on a population-wide basis. We compared the annual age-standardized incidence rates of extremity soft tissue sarcomas from two large tumor registries using different case definitions. We identified all limb soft tissue sarcoma cases diagnosed 1973-1993 in the Ontario Cancer Registry (OCR) and the Surveillance, Epidemiology, and End Results (SEER) databases. Two case definitions for limb soft tissue sarcoma were used based on missing data, incomplete diagnostic methods and ICD-9 codes; an upper limit estimate of the rates which included all possible cases of limb soft tissue sarcoma and a lower limit estimate of the rates which included all definite cases of limb soft tissue sarcoma (with the true rates lying in between). The upper limit OCR rates showed a statistically significant decreasing linear trend (slope = -0.021, P < 0.01). Whereas the slope of the OCR lower limit regression line showed a statistically significant increase in rates (slope = 0.01, P = 0.04). Neither the upper or lower limit SEER rates had a statistically significant linear trend (slope = 0.002, P = 0.60 and slope = 0.001, P = 0.18, respectively). Case definition affects incidence rates and changing rates of cancer. Thus the use of a single case definition along with changing coding practices may alone explain changing cancer rates.