Novel Immune Archetypes Stratify Overall Survival and Refine the Prognostic Landscape in Head and Neck Squamous Cell Carcinoma

Background
The PD-L1 combined positive score (CPS) is a widely used but imperfect biomarker in head and neck squamous cell carcinoma (HNSCC). The Immune Profile Score (IPS), Tempus’ AI-developed multi-modal biomarker integrating genomics and transcriptomics, offers a potentially more nuanced alternative. This study uses real-world data to define transcriptomic immune archetypes and clarifies their role in integrating key biomarkers to predict clinical outcomes.

Methods
De-identified records of 2,121 HNSCC cases from the Tempus database with whole-transcriptome RNA-sequencing, NGS-derived HPV16 status, PD-L1 IHC, and real-world overall survival (rwOS) data were analyzed. An unsupervised high-dimensional data reduction using a UMAP embedding of RNA-derived immune cell fractions, followed by k-means clustering, defined immune archetypes. Associations with IPS, HPV16 status, CPS, and rwOS were assessed via univariate and multivariate models.

Results
Two distinct archetypes were identified: an 'Inflamed' (n=243) and a vastly larger 'Immune-Desert' (n=1,878). The Inflamed archetype, defined by high CD4+/CD8+ T-cell and B-cell infiltration, was highly enriched in IPS-High tumors (93% vs 26%), HPV16 positivity (54% vs 26%), and CPS>=1 (94% vs 82%). Patients in the Inflamed archetype had markedly prolonged median rwOS compared to the Immune-Desert group (25.1 vs 16.8 months; p<0.001). Interestingly, multivariate CoxPH analysis confirmed IPS-High status, HPV16-positivity, and CPS≥1 as independent prognostic factors for improved rwOS in the overall cohort, suggesting further clinical or molecular patient stratification could be pursued.

Conclusion
Transcriptomic immune archetyping reveals biologically and clinically distinct HNSCC subgroups with profound differences in rwOS. This framework provides a biological context for prognostic scores like IPS by grounding them in the tumor’s specific immune cell composition, and consolidates independent prognostic factors into a single classification that demonstrates superior prognostic stratification. Crucially, this approach highlights the need for novel T-cell recruitment strategies for the high-risk Immune-Desert population.