竞争风险模型在肿瘤治疗临床试验中的应用研究
Application of competing risk model to clinic trials for tumor treatment

目的 将竞争风险模型应用到肿瘤治疗临床试验的分析中。方法 为了评价某药作为化疗药物的增敏药物治疗非小细胞肺癌、乳腺癌的临床增效作用，采用多中心、随机双盲、前瞻性、安慰剂平行对照设计的Ⅲ期临床试验，运用竞争风险模型，估计2组的累积缓解率和累积进展率，并用Gray检验进行组间比较，同时采用部分分布比例风险模型进行多因素分析。结果 对于非小细胞肺癌患者，2组化疗后累积缓解率和累积进展率的差异均有统计学意义（P=0.000，P=0.001），用药、年龄和肿瘤大小是化疗后病情缓解的影响因素，而用药、性别、年龄和肿瘤大小是化疗后病情进展的影响因素；对于乳腺癌患者，2组化疗后累积缓解率的差异有统计学意义，而累积进展率的差异无统计学意义，用药和肿瘤大小是化疗后病情缓解的影响因素，年龄和肿瘤大小是化疗后病情进展的影响因素。结论 在有竞争风险存在的情况下，应该使用累积发生函数及部分分布风险模型进行分析，其结论更符合实际。
： Objective To apply the competing risk model to the analysis of clinical trials for tumor treatment. Methods A multi-center, randomized, double-blind, prospective, placebo parallel control design, phase Ⅲ clinical trial was preformed to evaluate the clinical synergism of a medicine served as chemosensitizer for the treatment of non-small cell lung cancer and breast cancer. The competing risk model was used to estimate the rates of cumulative remission and cumulative progression in both groups and comparison was conducted between groups with Gray test. The sub-distribution hazard model was used to perform the multivariate analysis. Results For non-small cell lung cancer patients, the differences in rates of cumulative remission and cumulative progression in two groups after chemotherapy were statistically significant (P=0.000，P=0.001). Medications, age, and tumor size were factors influencing the remission after chemotherapy, while medications, sex, age, and tumor size were factors influencing the progression after chemotherapy. For breast cancer patients, the difference in cumulative remission rate in two groups after chemotherapy was statistically significant and the difference in cumulative progression rate in two groups after chemotherapy was not statistically significant. Medications and tumor size were factors influencing the remission after chemotherapy, while age and tumor size were factors influencing the progression after chemotherapy. Conclusion Cumulative incidence function and sub-distribution hazard model should be used if the competing risk exists because the results are more in line with the actual situation