Hoehler F K, Mantel N, Gehan E, Kahana E, Alter M
Stat Med. 1984 Jul-Sep;3(3):225-37. doi: 10.1002/sim.4780030305.
Clinical trials of treatments for rare or fatal diseases must often use historical rather than randomized concurrent controls. Randomized trials may not be possible if (1) the number of patients available is quite small, (2) ethical considerations discourage the assignment of patients to control treatments known to be ineffective or (3) some evidence suggests that the proposed new treatment is substantially more effective than the control. We suggest that, when randomization is not feasible, adequately maintained medical registers can serve as a source of control patients. Investigators must pay careful attention to prognostic factors that might bias the results. However, statistical techniques for controlling such bias exist and, if these are properly employed, registry-derived historical controls can be used to evaluate the effectiveness of treatment.
针对罕见或致命疾病的治疗方法进行临床试验时，往往必须采用历史对照而非随机同期对照。如果出现以下情况，随机试验可能无法进行：（1）可用患者数量极少；（2）伦理考量不鼓励将患者分配至已知无效的对照治疗组；（3）有证据表明拟用的新治疗方法比对照治疗显著更有效。我们建议，在随机化不可行时，妥善维护的医学登记册可作为对照患者的来源。研究人员必须密切关注可能使结果产生偏差的预后因素。然而，存在控制此类偏差的统计技术，若能正确运用这些技术，源自登记册的历史对照可用于评估治疗效果。
