MULTIVARIATE EVALUATION OF PROGNOSTIC FACTORS IN LOCALIZED NEUROBLASTOMA

Purpose: A multitude of risk factors has been described for patients w ith neuroblastoma. Little is known about the mutual interrelationship of these factors and their impact on patients with localized disease o nly. Patients and Methods: We investigated the possible influence of 3 7 variables univariately on event-free survival (EFS) in 308 consecuti ve patients with neuroblastoma stages I-III using Kaplan-Meier estimat es. The chi2 test was applied to detect nonrandom correlations, and th e Cox's regression model was used for the multivate evaluation of iden tified factors. Results: Seventeen factors appeared to influence EFS i n stage I-III patients (p < 0.05, log-rank >3.84), whereas 10 factors were found in the subgroup of stage III patients with midline crossing tumors (= stage III, n = 128). The majority of univariately identifi ed risk factors showed a nonrandom correlation to several others (p < 0.05). The multivariate analysis according to Cox selected for the pat ients with stages I-III the factors lactate dehydrogenase (LDH) (p = 0 .0011), resectability (p = 0.0167), weight loss (p = 0.0185), tumor ex tension beyond midline (p = 0.0207), and age (p = 0.0233). For stage I II patients the model identified the factors LDH (p = 0.0089), weight loss (p = 0.0135), resectability (p = 0.0408), and age (p = 0.0700). The identification of these independent risk factors permitted the des cription of risk groups with EFS ratios after >6 years between 22% and 96%. Conclusions: Risk estimation of high discriminating power is pos sible for patients with localized neuroblastoma using simple, readily available clinical data.