Abstract

Five-year age-specific incidence rates were shown to produce small but systematic errors in the calculation of the expected number of tumours in a hypothetical but realistic study population. Underestimates occurred at younger ages (under 55) and overestimates at older ages, with a small overestimate (0.22%) overall. Larger errors (up to 12%) were obtained when there was a rapid change in the single-year age structure of the study population. Interpolation between five-year rates will normally produce an inaccurate set of one-year rates. It is shown, with the example of a logarithmic interpolation, that these rates tend to produce errors of similar size to the five-year rates but with a small underestimate overall (0.37%). However, the interpolated rates produced the smaller errors (up to 1%) when the study population age structure undergoes rapid change. A method is suggested for partially correcting the error in the interpolated rates.