BACKGROUND: Cancer survival often has been reported as lower for the poor than the rich, but, to the authors' knowledge, systematic national estimates of deprivation gradients in survival over long periods of time have not been available. METHODS: The authors estimated national population-based survival rates for almost 3 million people who were diagnosed with 1 of 58 types of cancers (47 in adults, 11 in children) in England and Wales during the 20-year period 1971-1990 and followed through December 31, 1995. Cancer patients were assigned by their address at diagnosis to 1 of 5 categories (quintiles of the national distribution) of material deprivation by using a standard index derived from census data on unemployment, car ownership, household overcrowding, and social class that was available for all 109,000 census tracts in Great Britain. The authors used relative survival rates: the ratio of observed survival among the cancer patients to the survival that would have been expected if they had had the same background mortality as the general population. Background mortality differed widely among socioeconomic categories, and the authors constructed life tables from raw national mortality data by gender, single year of age, calendar period of death, and socioeconomic category to adjust for it. The authors used variance-weighted least squares regression to estimate both time trends in age standardized survival and socioeconomic gradients in survival. The number of avoidable deaths was estimated from the observed mortality excess compared with the expected mortality in each group of patients. RESULTS: Survival rose steadily for most cancers over 25 years to 1995 in England and Wales, but inequalities in survival between patients living in rich and poor areas were geographically widespread and persistent over this period of time. These patterns existed for 44 of 47 adult cancers examined but not for 11 childhood cancers. These inequalities in survival represented more than 2500 deaths that would have been avoided each year if all cancer patients had had the same chance of surviving up to 5 years after diagnosis as patients in the most affluent group. CONCLUSIONS: The largest national cancer survival study has provided strong evidence of systematic disadvantage in outcome among patients who lived in poorer districts compared with those who lived in wealthier districts.