Large geographical variations in cancer survival are seen across Europe and the UK. Within Scotland there are also large differences between groups of society defined by material deprivation. The main goal of this thesis is to identify the most important determinants of social variation in cancer survival in Scotland and to assess how these might be addressed in order to reduce inequalities. The underlying purpose is to investigate the usefulness of routine data sources in examining prognosis and patterns of cancer care in the general population. Scodand is ideal for this because of the routine linkage of cancer registry data with all hospital inpatient discharge records. Chapter 1 of the thesis introduces the statistical methodology, and the measures of deprivation and comorbidity that have been used. Chapter 2 covers the data, definitions and quality issues. Chapter 3 reviews cancer survival in Scodand, including trends over time, and age-, sex- and deprivation-related differences. Chapter 4 explores the concept of avoidable deaths, and compares the methods available for computing this statistic. Six cancers, each with strong evidence of a deprivation gradient in survival in Scodand, were investigated further (breast, colon, rectum, bladder and kidney, and melanoma of the skin), using data for patients diagnosed in 1997. The analyses focus on patient and tumour characteristics, and health care system and treatment factors. The main findings are that deprived patients have higher comorbidity at diagnosis and appear to present with more advanced tumours. There are wide differences in the treatment offered to affluent and deprived patients, which will be to some extent appropriate because of differences in stage of disease and general health but appear too large to be equitable. Almost half of the excess cancer deaths occurring from these cancers each year, due to differences in survival between deprivation groups, would appear to be avoidable by changes in policy or practice.