The Burkitt Lymphoma International Prognostic Index (BL-IPI)

Background. BL is a rare, high-grade B-cell lymphoma that is often studied in trials with small sample sizes. Historical definitions of “low-risk BL” vary between studies, use arbitrary cutoffs for lactate dehydrogenase (LDH), and identify a small favorable group, leaving >80-90% of patients (pts) in an undifferentiated “high-risk” category. A validated prognostic index will help compare study cohorts and better define good-prognosis pts for whom reduced treatment would be appropriate vs a poor-prognosis group in need of new approaches. Herein, we constructed and validated a simplified prognostic model for BL applicable to diverse clinical settings across the world.

Methods. We derived the BL-IPI from a large real-world evidence cohort of US adults treated for BL in 2009-2018 (Evens A, Blood 2020). Progression-free survival (PFS) from diagnosis until BL recurrence, progression, death, or censoring was the primary outcome. We first determined the best prognostic cutoffs for age, LDH (normalized to local upper limit normal, ULN), hemoglobin (Hgb), and albumin. Independent risk factors were ascertained by forward stepwise selection into Cox regression from candidate variables: age, sex, HIV+ status, ECOG performance status (PS) ≥2, advanced stage (3/4), involvement of >1 extranodal site, bone marrow, central nervous system (CNS), values of LDH, Hgb, and albumin. Derivation models used multiple imputation to mitigate bias from missing data and reported hazard ratios (HR) with 95% confidence interval (CI). BL-IPI groups, defined by inspection of survival curves, were compared using log-rank test for trend. We validated performance of the BL-IPI in an external retrospective dataset of BL pts treated contemporaneously in centers from the United Kingdom, Scandinavia, Canada, and Australia.

Results. Characteristics of pts in the derivation (N= 633) and validation (N=457) cohorts are shown in the Table. Age ≥40 years (yr), LDH >3xULN, Hgb <11.5 g/dL, and albumin <3.5 g/dL were determined as optimal prognostic cutoffs. Age ≥40 yr, PS ≥2, stage 3/4, involvement of marrow, CNS, LDH >3xULN, low Hgb, and low albumin were associated with inferior PFS in univariate tests. In the multivariable model age ≥40 yr, LDH >3xULN, PS ≥2, and CNS involvement were selected as 4 independent prognostic factors; adding stage did not enhance the model. The model was simplified to 3 groups with 0 (low risk; 18% of pts), 1 (intermediate risk; 36% of pts; HR=3.14; 95%CI, 1.61-6.14), or 2-4 factors (high risk; 46% of pts; HR=6.52; 95%CI, 3.48-12.20; Fig A) with 3 yr PFS of 92%, 72%, and 53%, respectively (P<.001, Fig. B); median PFS was reached only in the high-risk group (46 months, 95%CI, 19-53). BL-IPI was similarly prognostic for overall survival (OS, P<.001; Fig. C).

Among pts with stage III/IV (historically classified as “high-risk” and constituting 78% of all pts in the cohort), the BL-IPI further discriminated subgroups with 3 yr PFS of 87%, 71%, and 52%, respectively (P<.001; Fig. D), and OS of 95%, 75%, and 57%, respectively (P<.001; Fig. E). In addition, BL-IPI was prognostic regardless of HIV status, in the subcohort treated with rituximab (3 yr PFS: 92%, 73%, and 55%, respectively, P<.001), and among pts treated with specific regimens: CODOX-M/IVAC±R (3 yr PFS: 88%, 67%, 61%, respectively, P=.004), DA-EPOCH-R (3 yr PFS, 87%, 73%, 51%, respectively, P<.001), or hyperCVAD/MA±R (3yr PFS: 100%, 80%, 54%, respectively, P<.001).

In the international validation cohort, fewer pts had CNS involvement; most received CODOX-M/IVAC+R; and PFS/OS estimates at 3 yr were higher. BL-IPI categories were of similar size (low-risk 15%, intermediate-risk 35%, high-risk 50%), and provided similar risk discrimination (Harrell’s C=.65 in both datasets). PFS at 3 yr was 96%, 82%, and 63%, respectively (P<.001; Fig. F), and OS was 99%, 85%, and 64%, respectively (P<.001; Fig. G). In the validation cohort, BL-IPI remained prognostic in the subsets receiving rituximab (P<.001) and in advanced stage (P<.001).

Conclusions. BL-IPI is a novel prognostic index specific to BL, which was validated to allow for simplified stratification and comparison of risk distribution in geographically diverse cohorts. The index identified a low-risk group with PFS >90-95%, which could be targeted with future strategies for treatment de-escalation. Conversely, only about 55-60% of pts in the high-risk group achieved cure with currently available immunochemotherapy.