Primary Effusion Lymphoma Prognostic Score (PEL-PS): A Validated International Prognostic Score in HIV-Associated Primary Effusion Lymphoma

Background: Primary effusion lymphoma (PEL) is a rare, B cell non-Hodgkin lymphoma (NHL) caused by Kaposi sarcoma herpesvirus (KSHV), also known as human herpesvirus 8, that is most common in people with HIV. It has a unique natural history among NHL presenting as malignant body cavity effusions, accompanying inflammatory signs and symptoms, and 80% of tumors are also EBV positive. PEL has poor survival compared with other HIV-associated lymphomas with fewer than 50% of patients alive 3 years after diagnosis. The International Prognostic Index (IPI) has not been validated in PEL due to its rarity. ECOG performance status, levels of interleukin (IL)-6, IL-10, and EBV positivity of the tumor have been shown to be prognostic in small cohorts of patients with PEL, but these factors have not been validated. In a large international cohort, we sought to develop and validate a prognostic model in HIV-associated PEL.

Methods: We collected demographic and clinical data from patients with HIV-associated PEL who received first-line anthracycline-containing chemotherapy. The training set used to develop the predictive model was identified from patients treated in the HIV and AIDS Malignancy Branch at the National Cancer Institute (NCI) in the United States and Chelsea and Westminster Hospital (CWH) in the United Kingdom from 2000-2022. We performed univariate Kaplan-Meier analyses to determine prognostic factors of overall survival (OS). Factors with unadjusted p-values <0.10 were candidates for inclusion in the Cox proportional hazards model. We applied backward and stepwise selection to determine independent factors for inclusion in the final Cox model. The factors identified in the Cox model were able to form two groups of patients who were predicted to have good or poor survival. The factors included in the Cox model and the groupings formed were then applied to a validation set identified from patients with HIV-associated PEL treated with first-line anthracycline-containing chemotherapy at the Hôpital Saint-Louis in France over the same period. T cell counts and peripheral blood KSHV viral loads were not measured uniformly across institutions and could not be evaluated in the Cox model.

Results: For the training cohort, we identified 59 patients with PEL treated with first-line anthracycline-containing chemotherapy at NCI and CWH. The median OS was 10.6 years and not statistically different between institutions (P=0.27). In univariate analysis, 5 factors were statistically negative prognostic factors: ECOG 3-4, hemoglobin <8 g/dL, albumin ≤2 g/dL, platelets <135 K/L, and IPI 4-5. Two factors were statistically positive prognostic factors: extracavitary only disease and EBV ⁺ tumors. In the multivariable Cox model, only 2 factors, ECOG ≥3 [P=0.007; hazard ratio (HR)=4.0 (95% CI: 1.5-11.1)] and hemoglobin <8 g/dL [P=0.006; HR=3.8 (95% CI: 1.5-9.7)] were jointly associated with lower survival probability. After forming groups based on the permutations of hemoglobin and ECOG, a scoring system was devised using patients with no negative prognostic factors (score 0: hemoglobin ≥8 g/dL and ECOG ≤2) versus patients with 1-2 negative prognostic factors (score 1: hemoglobin <8 g/dL and/or ECOG ≥3) resulting in a median OS of 10.6 years (95% CI: 10.6-not estimable) versus 0.8 years (95% CI: 0.3-2.8 years), p<0.0001, respectively.

The validation cohort was very similar to the training cohort in terms of patient and disease characteristics. When we applied the multivariable model to the 58 patients from the validation cohort, ECOG and hemoglobin jointly predicted OS, validating the prognostic model [ECOG ≥3: P=0.0005; HR=15.6 (95% CI: 3.3-73.5) and hemoglobin <8 g/dL: P=0.04; HR=2.5 (95% CI: 1.05-5.8)]. Median OS in patients with score 0 was 16.9 years (95% CI:8.7-16.9) versus 0.6 years (95% CI: 0.3-1.0) in those with score 1.

Conclusion: Using the largest international cohort of patients with HIV-associated PEL ever evaluated, we developed and validated the PEL-PS, which is easily implementable in any clinical setting using ECOG and hemoglobin. Low hemoglobin may be a surrogate marker for KSHV-associated inflammation and elevated IL-6 levels, and this association should be evaluated further. Patients with poor prognosis identified by the PEL-PS may benefit from novel therapies currently under investigation in PEL, such as immunomodulatory drugs, anti-CD38 monoclonal antibodies, and anti-PD-1 agents.