Early response dynamics predict treatment failure in patients with recurrent and/or metastatic head and neck squamous cell carcinoma treated with cetuximab and nivolumab

Objectives: Recurrent and/or metastatic (R/M) head and neck squamous cell carcinoma (HNSCC) is currently an incurable disease. To improve treatment strategies, combinations of cetuximab plus nivolumab or pembrolizumab were evaluated for efficacy and safety for incurable R/M HNSCC. While some patients had a significant clinical benefit with complete or partial response, most patients had stable or progressive disease (PD). To identify patients with a high likelihood of treatment failure and prevent futile treatments, we developed a mathematical model of early response dynamics as an early biomarker of treatment failure. Materials and Methods: Demographics, RECIST assessment, and outcome were obtained from patients who were treated with combination of cetuximab and nivolumab on a previously published phase I/II clinical trial. We trained a tumor growth inhibition (TGI) ordinary differential equation (ODE) model describing patient-specific pre-treatment growth rate and uniform initial treatment sensitivity and rate of evolution of resistance. In a leaveone-out approach, we forecasted tumor burden and predicted time to progression (TTP) and PD. Results: The TGI model accurately represented tumor burden dynamics (R-2 = 0.98; RMSE = 0.57 cm) and predicted PD with accuracy = 0.71, sensitivity = 1.00, and specificity = 0.69 after three serial response assessment scans. Patient-specific pre-treatment growth rate correlated negatively with TTP (Spearman's rho = -0.67, p = 5.7e - 05). Conclusion: The TGI model can identify patients with high likelihood of PD based on early dynamics. Further studies including prospective validation are warranted.

Automated summary

This study developed a mathematical model to identify patients with a high risk of treatment failure in head and neck squamous cell carcinoma, and predicted the time to disease progression. By analyzing data from previously treated patients, the model accurately predicted tumor burden dynamics and disease progression.