Identification of Eight Surface Molecules with Survival Predictive Power in B Cell Chronic Lymphocytic Leukemia (B-CLL): A Proposal for a Scoring System.

Studies of gene expression profiling of B-CLL cells revealed a phenotype related to experienced B cells, although only a subset of B-CLLs has IgVH mutations. With the aim to identify the immunophenotypic profile associated with a different prognosis, we investigated by flow cytometry the expression of 36 surface molecules (cell-adhesion molecules, integrins, complement activity regulators, myeloid, T and B markers) in 125 B-CLLs, all characterized for IgVH mutations and survival. To recognize the surface molecules with survival predictive power, univariate Cox proportional-hazards analysis was applied to antigen expression values with overall survival as dependent variable. Once identified the antigens whose expression correlated with a z score of ±2.5 (P<0.005) or greater, the maximally selected log-rank statistics were applied to define the optimal cut-off values yielding the best separation of two subgroups with different survival. According to this approach, the following eight antigens were selected (cut-off values in parenthesis): CD55 (30%), CD62L (30%), CD49c (40%), CD11c (20%), CD54 (50%), CD25 (15%), CD79b (65%), CD38 (30%). The first six antigens had negative z score and therefore were identified as favorable prognosticators, while CD79b and CD38 had positive z score, hence were associated with shorter overall survival (negative prognosticators). To build-up a scoring system, we assigned score “1” to each positive prognosticator when its expression was above the designated cut-off (score “0” if below), and score “0” to each negative prognosticator when its expression was above the cut-off (score “1” if below). A total score ranging from 0 to 8 points was therefore obtained in 102/125 cases in which the expression of all the eight markers was available. Three risk groups were identified: i) high-risk (29 cases), score 0–3; ii) intermediate-risk (38 cases), score 4–6; iii) low-risk (35 cases), score 7–8. These three groups differed greatly for survival probabilities (p=5x10–13 by the log-rank test). All patients belonging to the low-risk group were alive throughout the follow-up duration, whereas mean survivals for intermediate- and high-risk groups were 173 months (p=0.032) and 61 months (p=2.0x10–9), respectively. Several relationship between risk groups and other variables was studied: i) patients included in high- and intermediate-risk groups had the same male to female (M:F) ratio (1.4), while the M:F ratio of patients included in low-risk group (group 3) was lower (0.7); ii) Rai’s stage distribution was comparable in the three groups, with the exception of stage “0”, which was significantly less frequent in the high-risk group (p=0.04); iii) if % IgVH mutations (2% cut-off) was checked, mutated to unmutated (M:UM) ratios were 4.8, 2.6 and 0.8 in low-, intermediate- and high-risk groups, respectively (p=0.006); iv) as compared to high-risk group, low- and intermediate-risk groups were characterized by a higher number of B-CLL cases with a IgVH mutational status consistent with antigen-driven selection (20/24 and 17/26 vs. 7/13). In conclusion, the present study introduces a novel predictive tool based on the expression of eight surface molecules, easily investigable, which can stratifies populations of B-CLL patients in three distinct risk categories.