AUTHORS
Chunyi Zhang, Vaishali Gautam, BS2, Lohitha Dhulipalla, MD3, Janamejey Gaur, MD4, Benjamin Mescher5, James P Long, PhD6, Swaminathan P. Iyer, MD7
AFFILIATIONS
1 MD Anderson Cancer Center, Houston, Texas, United States
2 MD Anderson Cancer Center, Houston, Texas, United States
3 MD Anderson Cancer Center, Houston, Texas, United States
4 MD Anderson Cancer Center, Houston, Texas, United States
5 MD Anderson Cancer Center, Houston, Texas, United States
6 MD Anderson Cancer Center, Houston, Texas, United States
7 MD Anderson Cancer Center, Houston, Texas, United States
ABSTRACT
Introduction: Peripheral T-cell lymphomas (PTCL) are rare, heterogeneous, aggressive lymphomas with historically poor outcomes, for which cyclophosphamide, doxorubicin, vincristine, prednisone (CHOP) has long been the backbone of frontline therapy. The phase 3 ECHELON-2 trial (NCT01777152) established brentuximab vedotin plus cyclophosphamide, doxorubicin, and prednisone (BV-CHP) as a new standard of care. However, real-world comparisons of BV-CHP versus CHOP remain limited, particularly in non-ALCL PTCL, which was underrepresented in ECHELON-2. We therefore developed an ECHELON-2 eligibility-informed, real-world CHOP comparator cohort and used it to contextualize real-world frontline BV-CHP outcomes and support future external-control development.
Methods: We retrospectively identified 1042 adults with newly diagnosed PTCL who received frontline CHOP or BV-CHP at MD Anderson Cancer Center from 1997 through 2025. ECHELON-2-informed eligibility criteria included age ≥18 years, Eastern Cooperative Oncology Group (ECOG) performance status ≤ 2, documented CD30 positivity for non-ALCL disease, and receipt of frontline CHOP or BV-CHP. Clinical, pathology, treatment, and outcome data were integrated using the MD Anderson Context Engine/Syntropy Foundry Platform under institutional review board (PA19-0120), with a waiver of informed consent. Histologic subtype was captured from the final pathology diagnosis and categorized as ALCL or non-ALCL. CD30 positivity was ascertained from pathology reports within 60 days of frontline treatment using large language model-assisted extraction. Progression-free survival (PFS) was measured from treatment initiation to progression or death, and overall survival (OS) from treatment initiation to death. Patients without an event were censored at the last evaluable follow-up. Missing ECOG performance status, disease stage, and lactate dehydrogenase (LDH) status were multiply imputed using 20 datasets. Within each dataset, propensity scores for BV-CHP receipt were estimated using gradient-boosted models that included age, sex, race, time from diagnosis to treatment, histologic group, ECOG, disease stage, and LDH status. Inverse probability of treatment weighting (IPTW) was used to estimate adjusted Cox hazard ratios (HRs) with 95% confidence intervals (CIs). Complete-case analyses were performed as sensitivity analyses.
Results: Among 121 eligible patients, 64 received BV-CHP and 57 received CHOP. Compared with CHOP recipients, BV-CHP recipients were older (median age, 60.5 versus 53.0 years), more frequently had non-ALCL PTCL (46.9% versus 35.1%), and more often had stage III-IV disease (67.2% versus 40.3%). After IPTW, all modeled covariates were balanced, with the maximum absolute standardized mean difference decreasing from 0.393 to < 0.1. In the primary multiple-imputation IPTW-adjusted analysis, BV-CHP was associated with numerically favorable PFS (HR, 0.83; 95% CI, 0.42-1.62) and OS (HR, 0.84; 95% CI, 0.37-1.91) compared with CHOP. Among patients with non-ALCL PTCL, the corresponding HRs were 0.74 for PFS (95% CI, 0.33-1.63) and 0.45 for OS (95% CI, 0.17-1.24). Complete-case IPTW analyses were directionally consistent, suggesting a trend favoring BV-CHP.
Discussion: Using ECHELON-2-informed eligibility criteria, we demonstrated the feasibility of constructing a clinically curated, patient-level real-world cohort of patients treated with frontline CHOP. This cohort provides a well-characterized reference for contextualizing contemporary BV-CHP outcomes and informing future external-control development. Adjudicated survival endpoints and multiple-imputation IPTW enabled prognostically balanced comparisons between treatment groups. Adjusted estimates numerically favored BV-CHP; however, wide confidence intervals, treatment-era differences, unequal follow-up, missing prognostic information, and potential residual confounding limit causal interpretation. Differences between ALCL and non-ALCL PTCL further suggest that pooled CHOP benchmarks may obscure the poorer expected outcomes of patients with non-ALCL PTCL. Future collaborative multi-institutional work will focus on emulating a prespecified target trial, addressing temporal trends and differential follow-up, and harmonizing PFS definitions. These efforts may enable more reliable histology-specific CHOP benchmarking and support the development of synthetic control cohorts in future frontline PTCL trials.
DISCLOSURES
Chunyi Zhang: No relevant conflicts of interest to declare.