Development of a Classifier for Metabolic Subtypes of Nasopharyngeal Carcinoma to Guide Personalized Immunotherapy Strategies: Biomarker Analysis of the Phase III CONTINUUM and DIPPER Trials.

Study question
Can tumor metabolic subtypes predict which patients with locoregionally advanced nasopharyngeal carcinoma (NPC) benefit from adding anti–PD-1 therapy to chemoradiotherapy (CRT)?

Design and methods

• Biomarker analysis: Tumor RNA sequencing from two phase III randomized trials:
  • CONTINUUM: Discovery cohort (n = 407 combined across both trials; exact per-trial split not specified in abstract).
  • DIPPER: Validation cohort.
• Metabolic subtyping: Metabolic gene–based consensus clustering in the discovery cohort to define metabolic subtypes.
• Classifier development: Machine-learning classifier trained in the discovery cohort and applied to the validation cohort to assign subtypes.
• Biologic characterization: Gene set enrichment analysis to characterize tumor-intrinsic and immune features of each subtype.
• Clinical endpoint: Event-free survival (EFS), comparing anti–PD-1 plus CRT vs CRT alone.

Key findings

• Metabolic subtypes: Three subtypes (MS1, MS2, MS3) were identified, with distinct biological and immune profiles.
• Benefit in MS1 only: MS1 showed clear benefit from adding anti–PD-1 to CRT:
  • 3-year EFS: 90.2% with anti–PD-1 + CRT vs 69.6% with CRT alone.
  • Hazard ratio: 0.27 (95% CI, 0.11–0.67).
• No meaningful benefit in MS2 and MS3:
  • MS2: 3-year EFS 94.1% vs 93.8% (no meaningful difference).
  • MS3: 3-year EFS 75.0% vs 75.0% (no difference).
• Validation: These subtype-associated patterns of benefit and prognosis were reproduced in the validation cohort.
• Pooled analysis: Across both trials, there was a statistically significant interaction between metabolic subtype and treatment effect (P = 0.0074), supporting a predictive (not just prognostic) role.

Clinical implications

• MS1 subgroup: Metabolic subtyping of NPC by RNA-based profiling appears to identify a subgroup (MS1) that derives substantial EFS benefit from adding anti–PD-1 to CRT.
• MS2 and MS3: These subtypes may not justify routine addition of immunotherapy based on EFS, pending more data on other endpoints and toxicity.
• Predictive biomarker: This molecular classification represents a promising predictive biomarker to guide personalized use of anti–PD-1 therapy in locoregionally advanced NPC.