Abstract
A comprehensive prognostic index that includes clinical (i.e., age, sex, ECOG performance status), serum (i.e., B2-microglobulin, thymidine kinase) and molecular (i.e., IGVH mutational status, del 17p, del 11q) markers with independent prognostic value and combines them into a single risk score was recently developed by the German CLL Study Group (GCLLSG).
We sought to externally validate this novel score in an independent, community-based cohort consisting of 338 patients with early chronic lymphocytic leukemia (CLL) who were registered on a prospective basis during 2006 to 2010 on an observational Italian database (O-CLL1 protocol; clinicaltrial.
gov identifier NCT00917540).
Because serum thymidine kinase was not available a slightly modified version of the German model based on 6 instead of 7 prognostic variables was used.
According to clinical caracteristics of our patients (i.e., all with Binet stage A disease), a point score higher than 5, necessary for inclusion in the high-risk group, was found only in 2 cases. By German prognostic index, 62.9% of patients were scored as having low-risk CLL (score 0-2) whereas 37.1% had intermediate-risk CLL (score 3-5). This stratification translated into a significant difference in the time to first treatment (TTFT). In other words, the higher the score risk the shorter the TTFT [HR = 4.21; 95% C.I. (2.71-6.53) P<0.0001; (Fig 1)]. Interestingly, the prognostic index category remained a predictor of TTFT also when the analysis was restricted to 262 patients with Rai stage 0 disease [HR= 3.12; 95% C.I. (1.84-5.31);P <0.0001]
The model proposed in 2007 by investigators at MD Anderson Cancer Center (MDACC) and barely based on traditional clinical parameters (i.e., age, gender, Rai substage, absolute lymphocyte count, number of involved nodal groups and B2-microglobulin) worked in predicting the TTFT [HR = 2,73; 95% C.I. (1.79-4.17); P<0.0001] in our patient cohort. In comparison to the MDACC score the German model provided similar results in terms of sensitivity (German score, 67%; MDACC score, 58.2; P = 0.28). However, both specificity (74% vs 65.4%; P=0.05) and accuracy (72.1% vs 63.5%; P=0.02) were significantly better with the German score.
In order to investigate which prognostic model (i.e. German or MDACC score) provided a more accurate prediction of TTFT, a C-statistic analysis, considered a measure of concordance between observed and predicted time-dependent events, was performed. Interestingly, the C-statistic of the German score was 0.71 (range, 0.60-0.82), a value higher than c-statistic observed with MDACC score (c=0.65; range,0.53-0.78).
In conclusion, we demonstrated that the German score works in predicting the TTFT of patients with early CLL also when a modified version which does not include serum thymidine kinase is utilized. These changes affected only partially the predictive value of model because C-statistic remained above 0.70, which is considered a necessary threshold of utility at the individual patient level.
No relevant conflicts of interest to declare.
This icon denotes a clinically relevant abstract
