Cancer survival statistics are needed by health authorities (HAs) to develop local policies and manage resources optimally. Methodological issues can affect the interpretation of these and similar statistics. More reliable indicators of cancer survival are obtained by using regional life tables, rather than a single national life table, to take into account background mortality. As cancer survival varies widely with age, age standardisation is required to improve the comparability of survival rates in cancer patient groups with different age structures. The standards based on cases diagnosed in England and Wales during 1986-1990 provide valuable continuity with earlier published results. Estimates of annual rates of change in survival (based on linear regression) together with their statistical significance, give reliable indications of trends.