124 - Biology-Guided Radiotherapy in Pan-HPV Head and Neck Cancer Using Genomic-Adjusted Radiation Dose

03:30pm - 03:40pm ET

Room 204

Presenter(s)

Drew Bergman, MD - Memorial Sloan Kettering Cancer Center, New York, NY

D. T. Bergman¹, N. Joshi², S. Nellore², J. J. Caudell³, E. C. Dee¹, P. Chen⁴, S. A. Eschrich⁵, N. Y. Lee¹, J. F. Torres-Roca³, and J. G. Scott⁶; ¹Department of Radiation Oncology, Memorial Sloan Kettering Cancer Center, New York, NY, ²Cleveland Clinic Lerner College of Medicine, Cleveland, OH, ³Department of Radiation Oncology, H. Lee Moffitt Cancer Center and Research Institute, Tampa, FL, ⁴Case Western University, Cleveland, OH, ⁵Department of Biostatistics and Bioinformatics, H. Lee Moffitt Cancer Center and Research Institute, Tampa, FL, ⁶Department of Radiation Oncology, Taussig Cancer Institute, Cleveland Clinic Foundation, Cleveland, OH

Purpose/Objective(s):

RT dose in HNSCC is often prescribed uniformly despite biological heterogeneity. The genomic-adjusted radiation dose (GARD) integrates physical RT dose with tumor genomics to quantify biological RT effect. We evaluated whether GARD independently predicts OS in HNSCC and whether its prognostic effect differs by HPV status, treatment intent, or subsite.

Materials/Methods:

We performed a retrospective, pooled, multi-institutional analysis of 531 HNSCC patients treated with definitive or adjuvant RT across 5 independent cohorts. HPV status was determined by standard DNA-based methods. Tumor gene expression was measured via microarray (Affymetrix 47.8%, Illumina 44.1%) or RNA sequencing (8.1%). RSI was derived from a 10-gene expression signature; GARD was calculated by integrating patient-specific a (from RSI) with delivered RT dose using the linear-quadratic model. Patients lacking postoperative dose data were assigned 60 Gy (54.4%). OS was censored at 5 years. Cox models were stratified by cohort and treatment intent with GARD modeled continuously.

Results:

The cohort was 69.3% OPSCC and 51.7% HPV+. RT dosing was uniform (mean EQD2 64.5 Gy, SD 4.9), while GARD and RSI varied widely (mean RSI 0.46, SD 0.14; mean GARD 27.2, SD 12.3). HPV+ tumors were predicted to be more radiosensitive (RSI 0.38 vs. 0.54, p<1e-16).

GARD was independently associated with OS in the full cohort (n=531; HR 0.96/unit, p=0.016), corresponding to a 35% reduction in mortality risk per 10-unit increase. The association was driven by definitive RT patients (n=244, p=0.021) and was not significant in postoperative patients (n=287, p=0.29; Table). HPV+ status was strongly prognostic for 5-year OS (HR 0.43, 95% CI 0.24–0.77, p=0.004). In a joint model (n=530), both HPV and GARD remained associated with OS (p=0.034 and p=0.026, respectively), with improved discrimination compared to either alone (C-index 0.62). Within OPSCC, the GARD-OS association was strong in HPV-agnostic analyses (n=368; HR 0.93/unit, p=0.001) and persisted in HPV+ OPSCC (n=258; HR 0.94, p=0.011), with a similar trend in HPV- OPSCC (n=109; HR 0.91, p=0.066; Table).

Conclusion:

Higher GARD was independently associated with improved 5-year OS after RT, driven by definitive RT. Both GARD and HPV status were independently prognostic with improved discrimination compared to either alone. GARD refined prognostic stratification within OPSCC and identified biologically radiosensitive tumors across HPV strata, supporting its role as a biology-driven metric for RT personalization in HNSCC.

Abstract 124 – Table 1

Population	N	Events	HR/unit GARD (95% CI)	p-value	C-index
Full cohort	531	138	0.96 (0.93–0.99)	0.016*	0.56
Definitive RT	244	32	0.95 (0.90–0.99)	0.021*	0.67
Postoperative RT	287	106	0.97 (0.92–1.02)	0.29	0.53
OPSCC (HPV-agnostic)	368	77	0.93 (0.89–0.97)	0.001**	0.79
HPV+ OPSCC	258	28	0.94 (0.89–0.99)	0.011*	0.69
HPV- OPSCC	109	48	0.91 (0.82–1.01)	0.066	0.57
Joint model:	530	137			0.62
HPV+			0.51 (0.28–0.95)	0.034*
GARD			0.96 (0.93–1.00)	0.026*