Risk Models for Advanced Melanoma Patients Under Anti-PD-1 Monotherapy—Ad hoc Analyses of Pooled Data From Two Clinical Trials

• Published in: Frontiers in oncology 2021-05, Vol.11, p.639085-639085
• Main Authors: Bai, Xue, Dai, Jie, Li, Caili, Cui, Chuanliang, Mao, Lili, Wei, Xiaoting, Sheng, Xinan, Chi, Zhihong, Yan, Xieqiao, Tang, Bixia, Lian, Bin, Wang, Xuan, Zhou, Li, Li, Siming, Kong, Yan, Qi, Zhonghui, Xu, Huayan, Duan, Rong, Guo, Jun, Si, Lu
• Format: Article
• Language: English
• Subjects: best response; melanoma; Oncology; overall survival; PD-1; progression free survival; risk model
• ISSN: 2234-943X; 2234-943X
• DOI: 10.3389/fonc.2021.639085

Background: The best response and survival outcomes between advanced melanoma patients treated with the anti-PD-1 monotherapy vary greatly, rendering a risk model in need to optimally stratify patients based on their likelihood to benefit from the said treatment. Methods: We performed an ad hoc analysis of 89 advanced melanoma patients treated with the anti-PD-1 monotherapy from two prospective clinical trials at the Peking University Cancer Hospital from April 2016 to May 2018. Clinicodemographical characteristics, baseline and early-on-treatment (median 0.6 months after anti-PD-1 monotherapy initiation) routine laboratory variables, including complete blood count and general chemistry, and best response/survival data were extracted and analyzed in both univariate and multivariate logistic and Cox proportional hazard models. Results: After three rounds of screening, risk factors associated with a poorer PFS included a high pre-treatment neutrophil, derived neutrophil-lymphocyte ratio (dNLR), low pre-treatment hemoglobin, and low early-on-/pre-treatment fold change of eosinophil; those with a poorer OS included a high pre-treatment neutrophil, eosinophil, PLT, early-on/pre-treatment fold change of LDH and neutrophil; and those with a poorer best response included a high pre-treatment NLR and early-on-/pre-treatment LDH fold change. Risk models (scale: low, median-low, median high, and high risk) were established based on these risk factors as dichotomous variables and M stage (with vs. without distant metastasis) for PFS (HR 1.976, 95% CI, 1.507–2.592, P < 0.001), OS (HR 2.348, 95% CI, 1.688–3.266), and non-responder (OR 3.586, 95% CI, 1.668–7.713, P = 0.001), respectively. For patients with low, median-low, median-high, and high risks of developing disease progression (PD), six-month PFS rates were 64.3% (95% CI, 43.5–95.0%), 37.5% (95% CI, 22.4–62.9%), 9.1% (95% CI, 3.1–26.7%), and 0%, respectively. For patients with OS risks of low, median-low, median-high, and high, OS rates at 12 months were 82.5% (95% CI, 63.1–100%), 76.6% (95% CI, 58.4–100%), 42.1% (95% CI, 26.3–67.3%), and 23.9% (95% CI, 11.1–51.3%), respectively. For patients with risks of low, median-low, median-high, and high of being a non-responder, objective response rates were 50.0% (95% CI, 15.7–84.3%), 27.8% (95% CI, 9.7–53.5%), 10.3% (95% CI, 2.9–24.2%), and 0%, respectively. Conclusion: A risk scoring model based on the clinicodemographical characteristics and easily obtainable routinely