Comparing Risk Scores for Disease Progression in Untreated CLL

A study published in Cancer compared the accuracy of five risk scores to predict disease progression in patients with untreated chronic lymphocytic leukemia (CLL): the International Prognostic Score for Asymptomatic Early-Stage CLL (IPS-E), the CR0 score, the Alternative IPS-E score (AIPS-E), the International Prognostic Index for CLL (CLL-IPI) score, and the Barcelona-Brno score.

“These scores predict time to first treatment (TTFT) in a high proportion of patients, but some patients are still misclassified,” wrote lead author Miguel Arguello-Tomas, MD, of Hospital de la Santa Creu in Barcelona, Spain, and colleagues.

The retrospective study included data from 781 patients (median age, 66.7 years) from three academic centers in Spain. There were no significant differences in clinical and biological characteristics at baseline among patients. As for mutational status, 37.3% of patients had unmutated IGHV, 8.3% had del(11q), and 9.2% had TP53 aberrations. Of the 99 patients (12.7%) who had IGHV stereotypic subsets of special interest, 15 (1.9%) were subset #1, 18 (2.3%) were subset #2, 73 (9.3%) were subset #4, and 26 (3.3%) were subset #8.

Validating Risk Scores

The researchers used the following statistical tests to measure the concordance between risk scores:

• Hosmer-Lemeshow goodness of fit to evaluate calibration
• Harrell’s Comparison Index (C index) to calculate the discriminatory value, defined as the risk scores’ ability to predict TTFT
• Akaike information criterion (AIC) to evaluate the level of parsimony of the scores
• Cohen weighted kappa coefficient test to evaluate the concordance between scores

Calibration with Hosmer-Lemeshow goodness of fit was similar and good for the risk scores (P=.98). The discriminatory values using Harrell’s C Index were high for all risk scores (c=0.78 for IPS-E, c=0.79 for CR0, c=0.77 for AIPS-E, c=0.78 for CLL-IPI, and c=0.79 for Barcelona-Brno). AIC scores were similar among IPS-E (840), CR0 (847), AIPS-E (842), and Barcelona-Brno (844); but CLL-IPI had the lowest and best score (441).

Weighted kappa showed substantial concordance (κ=0.61-0.8) between the IPS-E, CR0, AIPS-E, and Barcelona-Brno scores; substantial concordance between the Barcelona-Brno and CLL-IPI scores; and moderate concordance (κ=0.41-0.6) between the CLL-IPI and IPS-E, CR0, and AIPS-E scores.

A total of 269 patients (34.4%) were categorized in the same risk category across all five scores. “The different risk scores were successfully validated in our series and reproduced the low-, medium-, and high-risk groups for TTFT,” wrote Dr. Arguello-Tomas and colleagues. “The risk groups maintained their prognostic value when calculated as a cumulative incidence with competitive events (e.g., death without the start of therapy).”

The researchers also evaluated each risk score’s prognostic value for overall survival (OS). IPS-E, CR0, CLL-IPI, and Barcelona-Brno showed differences in OS among the three risk groups; however, AIPS-E did not show differences in OS between the high-risk and intermediate-risk groups.

IGHV Mutational Status and TTFT

The median TTFT was 3.9 years for unmutated IGHV and not reached for mutated IGHV (P<.001). The researchers found that IGHV mutational status was “significantly correlated” with TTFT. Those with IGHV subsets #1, #2, and #8 had similar TTFT to unmutated IGHV, and those with subset #4 had similar TTFT to mutated IGHV. Those with subset #2 had similar TTFT regardless of IGHV mutational status (P=.54).

When considering all patients with subset #2 as unmutated IGHV, those with mutated IGHV subset #2 shifted from good prognosis (low-risk group) to worse prognosis (intermediate-risk group). “As a result, we found that these patients showed a similar TTFT from the intermediate-risk group in all risk scores,” wrote Dr. Arguello-Tomas and colleagues.

 

As for study limitations, the researchers noted that some information to determine risk scores was missing due to the retrospective nature of the study. Because the study included patients from academic referral institutions, they also noted that the results may not be applicable to community centers.

“IPS-E, CR0, AIPS-E, CLL-IPI, and Barcelona-Brno are robust risk scores to predict TTFT,” the researchers concluded. “We suggest that the IGHV subset #2 appears to be useful to improve the accuracy of the risk scores. IPS-E, CR0, CLL-IPI, and Barcelona-Brno are also able to predict OS.”

Reference

Arguello-Tomas M, Mozas P, Albiol N, et al. Risk scores predicting disease progression in early-stage chronic lymphocytic leukemia: Comparative analysis and usefulness of IGHV subset #2 to improve their accuracy. Cancer. 2025;e35552. doi:10.1002/cncr.35552