The utility of two prognostic models for predicting time to forst treatment in early chronic lymphocytic leukemia patients: results of a comparative analysis

The use of both traditional and novel prognostic parameters combined in a statistical model for predicting patient clinical outcome has been recently proposed by both MD Anderson Cancer Center (MDACC) and German chronic lymphocytic leukemia (CLL) group. Using time to first treatment (TTFT) as end-point, we performed a comparative external validation of MDACC score versus a modified version of German score, which excluded thymidine kinase measurement, in a prospective, multicenter, community-based cohort consisting of 328 patients who had asymptomatic, early stage CLL. With both models a significant correlation between higher score and shorter TTFT could be found. As a matter of fact, patients with total point score ≥25 according to MDACC model (HR, 3.27; 95% CI, 2.07-5.18; P<0.0001) or ≥2 according to modified German model (HR, 2.02; 95% CI, 1.29-3.16; P=0.002) were more likely to receive therapy. Both models provided similar results in terms of sensitivity (MDACC score, 61.5%; modified German score, 57.7%; P=0.79), whereas specificity was significantly higher for MDACC score (72.1% versus 63%; P=0.02). The prognostic utility of either MDACC or modified German score was assessed by time-dependent Receiver Operating Characteristic (ROC) analysis. Results of this comparative analysis showed that after the 2nd year area under curve (AUC) for TTFT was higher than 0.60 for both models and kept unmodified this trend over the time. Results of this study suggest that in CLL both MDACC and modified German score group should be considered the benchmarking of comparison for any novel prognostic proposal having as endpoint TTFT in CLL and including both traditional and newer prognostic parameters.