目的:研究中国乳腺癌病人术后死亡风险时间分布规律。方法:回顾性研究1992年1月至2003年12月复旦大学附属肿瘤医院手术治疗并且资料完整的原发性乳腺癌2 214例临床资料，对其进行生存及死亡风险分析。结果:单因素和多因素生存分析均提示肿瘤大小、腋下淋巴结转移状态和孕激素受体状态是影响总存活率的预后因素（P<0.05）。全组病人年死亡风险曲线呈现双峰型，死亡高峰分别位于术后第2年及术后第9.5～10年，且该死亡模式在肿瘤较大及淋巴结阳性的亚组中更为明显。激素受体状态虽不改变病人的双峰型复发模式，但却呈现出一定的时间依赖性。结论:乳腺癌术后死亡风险时间分布呈现出一定的规律，针对性制定术后随访和辅助治疗的新策略，可能有助于改善乳腺癌的治疗效果。

Objective:To investigate the rule of mortality risk for breast cancer patients undergoing surgery in China. Methods:We performed a retrospective study of 2214 female unilateral breast cancer patients with complete data undergoing surgery in our hospital in order to conduct survival and mortality risk analysis. Results:Both univariate and multivariate survival analysis indicated that tumor size, axillary nodal status and progesterone receptor (PR) were prognostic factors for overall survival (P<0.05).Annual mortality hazard curve for entire population showed doublepeaked pattern.The analysis according to tumor size and axillary nodal status proved that the doublepeaked pattern was almost completely generated by patients with larger tumors and by N+ patients.While hormone receptor did not change the bimodal pattern of mortality risk,the results showed that it was a timedependent variable. Conclusion:According to the time distribution of mortality hazard, we could formulate novel strategies for postoperative treatment and follow-up, which provides the possibility to improve the therapeutic effect and reduce the mortality hazard.