Prognostic study of continuous variables (white blood cell count, peripheral blast cell count, haemoglobin level, platelet count and age) in childhood acute lymphoblastic leukaemia. Analysis of a population of 1545 children treated by the French acute Lymphoblastic Leukaemia Group (FRALLE)

Many cutpoints have been proposed to categorize continuous variables in chi ldhood acute lymphoblastic leukaemia (white blood cell count, peripheral bl ast cell count, haemoglobin level, platelet count and age), and have been u sed to define therapeutic subgroups. This variation in the choice of cutpoi nts leads to a bias called the 'Will Rogers phenomenon'. The aim of this st udy was to analyse variations in the relative risk of relapse or death as a function of continuous prognostic variables in childhood ALL and to discus s the choice of cutpoints. We studied a population of 1545 children with AL L enrolled in three consecutive protocols named FRALLE 83, FRALLE 87 and FR ALLE 89. We estimated the risk of relapse or death associated with differen t values of each continuous prognostic variable by dividing the sample into quintiles of the distribution of the variables. As regards age, a category of children under 1 year of age was distinguished and the rest of the popu lation was divided into quintiles. The floated variance method was used to calculate the confidence interval of each relative risk, including the refe rence category. The relation between the quantitative prognostic factors an d the risk was monotonic for each variable, except for age. for the white b lood cell count (WBC), the relation is log linear. The risk associated with WBC values in the upper quintile was 1.9 times higher than that in the low er quintile. The peripheral blast cell count correlated strongly with WBC ( correlation coefficient: 0.99). The risk increased with the haemoglobin lev el, and the risk in the upper quintile was 1.3 times higher than that in th e lower quintile. The risk decreased as the platelet count increased: the r isk in the lower quintile was 1.2 times higher than that in the upper quint ile. The risk increased gradually with increasing age above one year. The s mall subgroup of patients (2.5% of the population) under 1 year of age at d iagnosis had a risk 2.6 times higher than the reference category of patient s between 3 and 4.3 years of age. When the risk associated with a quantitat ive prognostic factor varies monotonously, the selection of a cutpoint is a rbitrary and represents a loss of information. Despite this loss of informa tion, such arbitrary categorization may be necessary to define therapeutic stratification. In that case, consensus cutpoints must be defined if one wa nts to avoid the Will Rogers phenomenon. The cutpoints proposed by the Rome workshop and the NCI are arbitrary, but may represent an acceptable conven tion. (C) 2000 Cancer Research Campaign.