Stewart L A, Parmar M K
MRC Cancer Trials Office, Cambridge, UK.
Lancet. 1993 Feb 13;341(8842):418-22. doi: 10.1016/0140-6736(93)93004-k.
The use of meta-analyses or overviews to combine formally the results of related randomised clinical trials is becoming increasingly common. However the distinction between analyses based on information extracted from the published literature and those based on collecting and reanalysing updated individual patient data is not clear. We have investigated the difference between meta-analysis of the literature (MAL) and meta-analysis of individual patient data (MAP) by comparing the two approaches using randomised trials of cisplatin-based therapy in ovarian cancer. The MAL was based on 788 patients and the MAP on 1329 and estimated median follow-ups were 3.5 and 6.5 years, respectively. The MAL gave a result of greater statistical significance (p = 0.027 vs p = 0.30) and an estimate of absolute treatment effect three times as large as the MAP (7.5% vs 2.5%). Publication bias, patient exclusion, length of follow-up, and method of analysis all contributed to this observed difference. The results of a meta-analysis of the literature alone may be misleading. Whenever possible, a meta-analysis of updated individual patient data should be done because this provides the least biased and most reliable means of addressing questions that have not been satisfactorily resolved by individual clinical trials.
使用荟萃分析或综述来正式合并相关随机临床试验的结果正变得越来越普遍。然而，基于从已发表文献中提取的信息进行的分析与基于收集和重新分析更新后的个体患者数据进行的分析之间的区别并不明确。我们通过使用基于顺铂的卵巢癌治疗随机试验比较这两种方法，研究了文献荟萃分析（MAL）和个体患者数据荟萃分析（MAP）之间的差异。MAL基于788例患者，MAP基于1329例患者，估计的中位随访时间分别为3.5年和6.5年。MAL得出的结果具有更高的统计学显著性（p = 0.027对比p = 0.30），且绝对治疗效果的估计值是MAP的三倍（7.5%对比2.5%）。发表偏倚、患者排除、随访时间长度和分析方法都导致了这种观察到的差异。仅基于文献的荟萃分析结果可能会产生误导。只要有可能，就应该进行更新后的个体患者数据荟萃分析，因为这为解决个体临床试验未能令人满意解决的问题提供了偏差最小且最可靠的方法。
