Background Our aim in this study is to provide a systematic assessment of the site-specific cancer survival rates of patients with different educational levels, using population-based census-linked registry data covering the entire population of Lithuania. Material and methods The study is based on the linkage between all records of the 2001 population census and all records from Lithuanian Cancer Registry (cancer incidence) and Statistics Lithuania (deaths) for the period between 6 April 2001 and 31 December 2009. Results For the vast majority of cancer sites we found an inverse gradient in survival, with the worst survival indicators in the lowest educational group. We estimated that 18.6% of the deaths in Lithuanian cancer patients could have potentially been postponed, if all the patients had the same cancer mortality as the patients with the highest educational level. Conclusion Our findings offer a warning that although the survival rates of cancer patients are improving, this progress hides disparities between different groups of patients.