ALBERT

Description: The International Prognostic Risk Index (IPI) offers the most important prognostic information in patients with aggressive non-Hodgkin Lymphoma (NHL). Risk classification is based on presence or absence of age 〉60 yr, stage III or IV, more than one extranodal site, performance score 2–4, or elevated serum LDH above the upper limit of normal (ULN). Because each factor confers a more or less equal independent risk for treatment outcome, they are summed to generate a final IPI risk score (Shipp model). Dichotomization of the continuous variable age is practical, as treatment for patients 〉60 yr often differs from younger ones. However, as LDH also is a continuous variable we wondered if risk based on actual LDH, especially in patients with highly elevated levels, would have additional prognostic impact. IPI risk factors including actual serum LDH at diagnosis were retrieved from 1286 patients (28% 〉60 yr) with advanced aggressive NHL and treated with curative intent in 6 clinical trials conducted by HOVON (5 trials) and EORTC (1 trial). All LDH ULNs from the participating centers were verified. LDH levels were divided by the ULN of each center to generate normalized ratios. Six % of patients had LDH 〉5 times ULN, 8% 3–5 times, 11% 2–3 times, 34% 1–2 times, 20% 0.75–1 times and 20% had levels below 0.75 of ULN. In a multivariate Cox regression model similar independent hazard ratios (HR) ranging from 1.6 to 2.6 were found for the individual dichotomized risk factors according to the Shipp model except for the number of extranodal sites which turned out to be non-significant. This factor was with a HR of 1.4 also the least predictive factor in the Shipp model. In contrast to the dichotomized LDH variable (normal versus elevated), risk for inferior outcome increased linearly with actual LDH levels. Five year OS estimates were 71% for patients with LDH 5 x ULN. The HR for these groups were respectively 0.54, 1, 1.45, 2.46, 2.54 and 5.15. This analysis using actual LDH values gave a better discrimination as compared to the HR of 2.6 (95% CI 2.1–3.1) for the dichotomized LDH (i.e. normal versus elevated) in multivariate analysis. Interestingly, the 235 patients with very low LDH (