Refining Prognostication in Peripheral T Cell Lymphoma (PTCL): A Risk-Adapted Prognostic Model for Patients with Limited-Stage PTCL

INTRODUCTION

Peripheral T-cell lymphomas (PTCL) is a rare and heterogeneous group of diseases in which risk-stratified approaches for managing patients (pts) with limited-stage (LS) PTCL remain unexplored. Although existing prognostic systems, such as the International Prognostic Index and the Prognostic Index for PTCL (PIT), have been used to predict outcomes, they lack impact on clinical decision-making. To refine the management of LS PTCL and improve prognostication, we developed a risk-stratified prognostic model and validated the findings in an independent external cohort.

METHODS

We included pts aged ≥18 years with newly diagnosed stage I-II PTCL. Subtypes were classified as anaplastic large cell lymphoma (ALCL, irrespective of ALK status) and non-ALCL (i.e., PTCL not otherwise specified [NOS], angioimmunoblastic T cell lymphoma and lymphomatous adult T cell leukemia/lymphoma [ATL]). Data for the training cohort was derived from the hospital-based registries of the Grupo de Estudio Latinoamericano de Linfoproliferativos and the Brazilian T-cell Project (median follow-up of 39 months). Data for the validation cohort was obtained from 2 academic centers in the United States (median follow-up of 53 months).

Overall survival (OS) was the primary endpoint of the score. We used Directed Acyclic Graphs to identify candidate variables reflecting patient fitness (ECOG performance status), disease activity (B symptoms, serum albumin and LDH levels and extranodal involvement), and disease biology (ALCL vs non-ALCL). Variables with a p-value <0.1 in an univariable Cox regression were included in a multivariable model. We used Kaplan-Meier and C-index to evaluate the discriminatory performance of the score compared to the PIT score.

RESULTS

The training cohort included 199 pts with LS PTCL who received first-line treatment with curative intent. The median age at diagnosis was 52 years (range 18-86) with a male predominance (59%) and good performance status (ECOG ≤1, 87%). B symptoms and extranodal involvement were present in 39% and 40%, respectively. ALK-negative ALCL (17%) and PTCL NOS (59%) were the most frequent subtypes within the ALCL and non-ALCL groups, respectively. CHOP (41%) was the most frequently used regimen. In the validation cohort, clinical features were similar, except a higher frequency of ALCL subtypes (49% vs 27%, p<0.001), lower frequency of extranodal involvement (16% vs. 40%, p<0.001), and a higher use of brentuximab vedotin-based regimens (35% vs. 3%, p<0.001).

Three factors were associated with OS in the multivariable Cox model, ECOG ≥2 (Hazard ratio [HR]=3.12, 95% confidence interval [CI]=1.61-6.05), B symptoms (HR=1.96 [1.16-3.32]) and non-ALCL subtype (HR=2.79 [1.25-6.23]). We used the HRs to assign points to the selected variables and developed a 2-risk group stratification model identifying pts with favorable (score 0-1) and unfavorable (score 2-6) prognoses. Pts with unfavorable risk had worse outcomes than those with favorable status (3-year OS 89% vs. 54%, p<0.001). The C-index was 0.602.

In the validation cohort (n=91), the score identified worse outcomes in pts with unfavorable features than those with favorable (3-year OS 97% vs. 64%, p=0.002), with better discriminatory performance than in the training cohort (C-index=0.690). Although the C-index of the PIT score was about 0.702 in both cohorts, the low sample sizes in the risk subgroups increased the uncertainty of estimates, and Kaplan-Meier curves overlapped. Sensitivity analysis excluding lymphomatous ATL cases did not affect score predictability on OS (p=0.002, C-index=0.605).

CONCLUSION

We have developed and validated a prognostic score for OS in LS PTCL pts reflecting patient fitness, disease activity, and biology. The 2-risk group scoring approach is practical for clinical use and may allow reproducibility in clinical studies. Notably, some pts with favorable status received less aggressive approaches, such as abbreviated courses of chemotherapy plus radiation therapy. Our model was developed in a setting of different PTCL epidemiology from Western countries, but the score maintained its prognostic value in the validation cohort from the United States. The score should be validated in prospective studies and among other Western and Asian populations. Our score may be used in clinical practice and trial designs to allocate pts with LS PTCL who may benefit from novel therapy approaches.