Proportion cured models applied to 23 cancer sites in Norway.

Statistical cure is reached when a group of patients has the same mortality as cancer-free individuals. Cure models predict the cured proportion and the median survival of fatal cases. Cure models have seldom been applied and tested systematically across all major cancer sites. Incidence and follow-up data on 23 cancer sites recorded at the Cancer Registry of Norway 1963-2007 were obtained. Mixture cure models were fitted to obtain trends and up-to-date estimates (based on period approach) assuming cured and uncured groups exist. The model converged for cancers of the mouth and pharynx, oesophagus, stomach, colon, rectum, liver, gallbladder, pancreas, lung and trachea, ovary, kidney, bladder, CNS, non-Hodgkin lymphoma (only for males) and leukemia. The proportion of cured patients increased 1963-2002 for both sexes, with the largest changes (in percent) seen for leukemia (46.4 and 46.7) and CNS (35.9, 42.0), males given first. Median survival time for the uncured cases increased for colon and rectal cancer, and there was a three- fold increase in median survival time for patients with fatal ovarian cancers. Cancers of bladder and CNS had the highest up-to-date proportion cured (in percent), 67.4 and 64.0, respectively, pancreas and liver were amongst the lowest (5.7 and 9.9, respectively). Cure models are useful when monitoring progress in cancer care, but must be applied and interpreted with caution. The absolute estimates of the cure proportion are speculative and should not be calculated where cure is not medically anticipated.