Alafchi Behnaz, Tapak Leili, Hamidi Omid, Poorolajal Jalal, Mahjub Hossein
Department of Biostatistics, School of Public Health, Hamadan University of Medical Sciences, Hamadan, Iran.
Modeling of Noncommunicable Diseases Research Center, School of Health, Hamadan University of Medical Sciences, Hamadan, Iran.
Iran J Public Health. 2019 Dec;48(12):2249-2259.
Breast cancer is the first non-cutaneous malignancy in women and the second cause of death due to cancer all over the world. There are situations where researchers are interested in dynamic prediction of survival of patients where traditional models might fail to achieve this goal. We aimed to use a dynamic prediction model in analyzing survival of breast cancer patients.
We used a data set originates from a retrospective cohort (registry-based) study conducted in 2014 in Tehran, Iran, information of 550 patients were available analyzed. A method of landmarking was utilized for dynamic prediction of survival of the patients. The criteria of time-dependent area under the curve and prediction error curve were used to evaluate the performance of the model.
An index of risk score (prognostic index) was calculated according to the available covariates based on Cox proportional hazards. Therefore, hazard of dying for a high-risk patient with breast cancer within the next five years was 2.69 to 3.04 times of that for a low-risk patient. The value of the dynamic C-index was 0.89 using prognostic index as covariate.
Generally, the landmark model showed promising performance in predicting survival or probability of dying for breast cancer patients in this study in a predefined window. Therefore, this model can be used in other studies as a useful model for investigating the survival of breast cancer patients.
乳腺癌是女性中首位非皮肤恶性肿瘤，也是全球癌症死亡的第二大原因。在某些情况下，研究人员对患者生存情况的动态预测感兴趣，而传统模型可能无法实现这一目标。我们旨在使用动态预测模型分析乳腺癌患者的生存情况。
我们使用了一个源自2014年在伊朗德黑兰进行的一项回顾性队列（基于登记）研究的数据集，对550例患者的可用信息进行了分析。采用地标法对患者生存情况进行动态预测。使用时间依赖性曲线下面积和预测误差曲线的标准来评估模型的性能。
根据基于Cox比例风险的可用协变量计算出风险评分指数（预后指数）。因此，高危乳腺癌患者在未来五年内死亡的风险是低危患者的2.69至3.04倍。以预后指数作为协变量时，动态C指数值为0.89。
总体而言，在本研究中，地标模型在预定义窗口内预测乳腺癌患者的生存或死亡概率方面表现出良好的性能。因此，该模型可在其他研究中作为研究乳腺癌患者生存情况的有用模型。
