Blackwood J M, Seelig R F, Hutter R V, Rush B F
Surgery. 1977 Oct;82(4):443-7.
It has been stated that breast cancer survival rates follow an exponential distribution. This would mean that the mortality rate is constant. Survival distribution was analyzed by the clinical life table method in one series of 10,752 patients and in another of 656 patients followed up to 8 and 18 years, respectively. Part of the larger series' table is (table: see text). Necessarily, clinical survival data are censored progressively. These kinds of data are analyzed best by examining the hazard function, which is the instantaneous death rate, or force of mortality. If an exponential distribution described survival in breast cancer correctly, the hazard function would be constant. These data clearly are not consistent with an exponential distribution, as the hazard function decreases. The survival distribution calculated from these data shows that the chance of dying of cancer decreases the longer a patient survives. This is more optimistic and consistent with clinical experience than is the exponential distribution.
有人指出，乳腺癌生存率呈指数分布。这意味着死亡率是恒定的。采用临床生命表法对两组患者的生存分布进行了分析，一组为10752例患者，另一组为656例患者，随访时间分别长达8年和18年。较大系列表格的一部分如下（表格：见正文）。临床生存数据必然会逐渐受到截尾影响。通过检查风险函数（即瞬时死亡率或死亡力）能最好地分析这类数据。如果指数分布能正确描述乳腺癌的生存情况，那么风险函数将是恒定的。由于风险函数下降，这些数据显然与指数分布不一致。根据这些数据计算出的生存分布表明，患者存活时间越长，死于癌症的几率越低。这比指数分布更乐观，也与临床经验相符。
