免疫治疗长期生存模拟方法:挑战与建议

Long-term survival simulation of immunotherapy:challenges and recommendations

目的 由于免疫治疗独特的作用机制,其生存曲线形状区别于常规治疗生存曲线。为了更加明确了解该类生存曲线的形成原因以及如何对免疫治疗生存曲线的长期模拟进行优化,该研究对此进行简介和实例分析,以期为相关研究人员提供参考。方法 以Checkmate-067临床试验的生存数据为例,应用不同生存模型对基于27个月的生存数据进行外推,将28至60个月的外推结果与实际生存数据进行比较验证。结果 该实例中,总生存期曲线形状符合常规形状,应用不同生存模型与之拟合较好;接受免疫治疗后的无进展生存期曲线形状与接受常规治疗的曲线差别较大,整体拟合结果较差。结论 对免疫治疗长期生存模拟,仅考虑赤池信息准则和贝叶斯信息准则做出模型的选择是不推荐的,还要综合免疫治疗的特点进行优化,尽可能做出相应的拟合图以进行视觉检查,从而对拟合情况做出判断。若存在假性进展和平台期等情况时,如若拟合效果不佳,可尝试不在0时间起点拟合或选择更加灵活的生存模型。

Objectives Due to the unique mechanism of immunotherapy, the shape of its survival curve differs from that of conventional treatment. In order to better understand the causes of this type of survival curve and how to optimize the long-term simulation survival curves of immunotherapy, this study provides an introduction and an example demonstration to serve as a reference for relevant researchers. Method Taking the survival data of Checkmate-067 trial as an example, different survival models are used to extrapolate the survival data based on 0 to 27 months, and the results of extrapolated curve are compared with the actual survival curve of 28 to 60 months. Results In this example, the shape of overall survival curve is consistent with the conventional shape, and it is well fitted with different survival models, but the shape of progression free survival(PFS) curve with immunotherapy is significantly different from that with conventional therapy, and the fit is poor. Conclusion For the long-term survival simulation of immunotherapy, it is not recommended to select the model only considering the Akaike information criteria and Bayesian information criteria. Moreover, it is necessary to optimize the model based on the characteristics of immunotherapy, and making the corresponding fitting curve as far as possible for visual inspection. Also, the curve fitting requires special optimization or selection of a more flexible survival model when encountering the pseudo-progression or plateau.