Cancer mortality in elderly people is described to highlight the mechanisms that could potentially explain observed differences with other age groups. Data from 15 French cancer registries were considered in the search for the 5-year outcome of patients diagnosed during the period 1989-1997. Relative survival, excess mortality hazard, and hazard ratio of mortality were estimated to describe patient outcome according to age. Five cancer sites were selected: colon/rectum, prostate, breast, head/neck, and lung. An excess mortality rate was found in patients aged over 75 at the time of diagnosis. This excess mortality rate was mainly seen during the first months after diagnosis, then it decreased gradually with time. An initial phenomenon of patient selection, a greater disease severity at the time of diagnosis, and less-effective treatments given to elderly patients are the most plausible explanations for the increased risk of cancer-related death in the eldest patients.