Cure trends in acute lymphoblastic leukemia: is it time for a revised concept of cure?

I n this issue of Haematologica, Gatta et al. describe cure trends for children and adolescents with acute lymphoblastic leukemia (ALL) who were diagnosed between 1982 and 2002, analyzing data obtained from various cancer registries across Europe. 1 The good news is that the projected proportion of cured children increased significantly over the study period, especially in children in the good prognosis age groups: from 70% to 90% in 1-4 year olds, and from 63% to 86% in 5-9 year olds. The projected outcome data for children below the age of one year and for older children/young adolescents was less satisfactory with increases from 26 to 58% for infants, 52% to 77% for 10-14 year olds, and 44% to 50% for 15-24 year olds. The 'proportion cured' rates are lower than the 5-year survival rates, which is caused by deaths during later follow up, including those due to late disease recurrence, second malignant neo-plasms and other toxicities. This is well recognized in prior studies , for instance, from the Childhood Cancer Survivor Study, showing that 66% of deaths in 5-year survivors of ALL were due to relapsed disease or second primary leukemias. 2 Analysis of cancer outcome data using cure models has a long history 3 and has been used in the past in pediatric cancer. 4 The concept of cure as applied in these models is not always well defined, however. The mixture model of the type used by Gatta et al. divides patients into 2 groups, those who are 'cured' and have mortality identical to the normal population, and those who are 'not cured' who have a fixed level of excess mortality. 3 This implies that one can classify individual patients as belonging to one of these 2 groups. In reality, patients who have survived cancer in the short term comprise a spectrum of patients ranging from those with mortality negligibly in excess of normal to those with mortality much in excess of normal. When follow up is limited and a clear plateau in cancer-related mortality has not yet been achieved, the estimates of the 'cured proportion' are really projections, i.e. extrapolations past the data 5 that depend on the linearity and proportionality of risk assumptions implicit in the statistical model. Nevertheless, this model can appropriately reflect the excess mortality among pediatric ALL survivors, and the 'proportion cured' is a useful index of this excess mortality. In pediatric …