1118 - Similar discriminatory values of IPS-E, AIPSE, and CR0 scores predicting time to first treatment in a large series of unselected patients with early-stage CLL

In Western countries, most patients with chronic lymphocytic leukemia (CLL) are diagnosed in early phases of the disease. Although overall their prognosis is good, there are patients who progress and require therapy shortly after diagnosis. Identifying these latter at diagnosis may be useful to decide the frequency of clinical controls as well as their setting (i.e., academic center versus general practice) to monitor the pace of the disease and to include patients at high risk of disease progression in clinical trials. To date, several scores including IPS-E, AIPS-E, and CR0 scores, have been designed to predict to first treatment (TTFT) in patients with early-stage CLL.

We conducted a retrospective, and non-interventional study performed in two Spanish centers: Hospital General Universitario Gregorio Marañón (HGUGM, Madrid), and Hospital de la Santa Creu I Sant Pau (HSCSP, Barcelona). The primary objective was to compare the discriminatory value of each score for detecting patients with CLL at an early stage with a high-risk of progression using Harrell’s Comparison Index (C index). As secondary objectives, the discriminatory value was also calculated with Receiver Operating Characteristics (ROC) curves, and the concordance between scores was measured with Cohen’s kappa coefficient test. Patients with early-stage CLL (Binet A and Rai 0), and without indication for treatment at diagnosis were included.

354 patients (Table 1) with a diagnosis of CLL in early stage were included: 66 (18.6%) from HGUGM and 288 patients (81.4%) from HSCSP. The median follow-up was 6.9 years (0.13-29.5). The median age at diagnosis was 67.3 years (28.9-93), and 180 were male (50.8%). As high-risk biomarkers at diagnosis, IGHV genes were unmutated in 37.3% of patients (132/354), del(17p) in 5.9% (21/354), and del(11q) in 8.2% (29/354). 140 patients (39.5%) started first-line treatment, with a median TTFT of 5.7 years (0.1-24.9).

The 3 scores were successfully validated in our cohort (Figure 1) and identified low, medium, and high-risk progression groups. According to the C index, all the scores presented a good-discriminatory value (c=0.71 for IPS-E, c=0.71 for AIPS-E, and c=0.7 for CR0) for detecting high-risk patients, with no significant differences between the scores (p=0.46). The discriminatory value of the scores was also high according to the AUC with ROC (AUC=0.73 for IPS-E, AUC=0.72 for AIPS-E, and AUC=0.71 for CR0). The concordance between the scores for categorizing high-risk patients was analyzed, finding a substantial agreement between AIPS-E and IPS-E scores (k=0.68), and AIPS-E and CR0 scores (k=0.69), but it was moderate for IPS-E and CR0 scores (k=0.41). The percentage of patients classified as high-risk was not significantly different between the three scores (IPS-E 15.5% vs AIPS-E 19.2% vs CR0 21.5%; p=0.64).

According to the data from this study, IPS-E, AIPS-E, and CR0 scores identify a similar group of patients with early stage with low, intermediate- and high-risk of disease progression most likely reflecting the items included in the different scores. The scores had a similar discriminatory value according to the C-index (c=0.71, c=0.71, and c=0.7, respectively). Although the c-value is high, there is an uncertainty level of 30% whichever the score. This is a drawback in selecting patients for clinical trials.