Predictors of the effectiveness of intensive treatment for locoregionally advanced HNC are lacking. We developed and validated a predictive model to identify patients most likely to benefit from treatment intensification (altered fractionation (AFX) or chemotherapy (CT)) based on the relative hazard for cancer recurrence vs. competing mortality, with patients at higher risk for recurrence relative to non-cancer mortality classified as “high risk.”
We analyzed 22,398 HNC patients treated on randomized trials testing CT and AFX. For predictive model training, we applied a risk score based on generalized competing event (GCE) regression to the MARCH dataset (11,174 patients, 30 trials) and compared this to a risk score based on a Cox model for progression-free survival (PFS). We externally tested both models on the MACH-NC dataset (11,355 patients, 54 trials). Normalized model inputs included age, performance status (PS), sex, site, T and N category, P16 status, and smoking history. P16 and smoking were available for 18% and 49% of patients in MARCH, respectively, but not in MACH-NC. Risk strata were defined as low, medium, and high based on PFS or GCE score tertiles. We assessed for interactions between risk strata and AFX or CT on overall survival (OS).
Factors associated significantly with GCE risk score were age, PS, tumor site, T and N category, and P16 status. The effect of AFX in MARCH was greater in the medium and high (HR [95% CI]=0.91 [0.84-0.98]; 0.92 [0.85-0.99]) vs. low risk group (HR = 0.97 [0.90-1.05]; p = 0.15 for interaction; p = 0.03 for risk score>85th percentile) defined by GCE model. When applied to MACH-NC, the effect of CT on OS was significantly greater in the medium and high (HR = 0.86 [0.78-0.88]; 0.81 [0.75-0.88]) vs. the low risk group (HR = 0.96 [0.86-1.08]; p = 0.014 for interaction). In contrast, the effects of AFX and CT were similar across PFS model strata, indicating its lack of predictive ability (p > 0.05 for all interactions). 30% of patients older than 70 belonged to a GCE risk group benefitting from CT.
GCE risk score is better than PFS risk score for predicting the benefit of intensive treatment on OS in HNC. This work is on behalf of MARCH/MACH-NC Collaborative Group and of the HNCIG.
Clinical trial identification
Legal entity responsible for the study
Loren Mell. MD.
UCSD Head and Neck Cancer Center of Excellence.
All authors have declared no conflicts of interest.