Hay S C, Kirpalani H, Viner C, Soll R, Dukhovny D, Mao W-Y, Profit J, DeMauro S B, Zupancic J A F
Department of Neonatology, Beth Israel Deaconess Medical Center, Boston, MA, USA.
Division of Newborn Medicine, Boston Children's Hospital, Harvard Medical School, Boston, MA, USA.
J Perinatol. 2017 Nov;37(11):1215-1219. doi: 10.1038/jp.2017.126. Epub 2017 Sep 7.
To assess the impact of the latest randomized controlled trial (RCT) to each systematic review (SR) in Cochrane Neonatal Reviews.
We selected meta-analyses reporting the typical point estimate of the risk ratio for the primary outcome of the latest study (n=130), mortality (n=128) and the mean difference for the primary outcome (n=44). We employed cumulative meta-analysis to determine the typical estimate after each trial was added, and then performed multivariable logistic regression to determine factors predictive of study impact.
For the stated primary outcome, 18% of latest RCTs failed to narrow the confidence interval (CI), and 55% failed to decrease the CI by ⩾20%. Only 8% changed the typical estimate directionality, and 11% caused a change to or from significance. Latest RCTs did not change the typical estimate in 18% of cases, and only 41% changed the typical estimate by at least 10%. The ability to narrow the CI by >20% was negatively associated with the number of previously published RCTs (odds ratio 0.707). Similar results were found in analysis of typical estimates for the outcomes of mortality and mean difference.
Across a broad range of clinical questions, the latest RCT failed to substantially narrow the CI of the typical estimate, to move the effect estimate or to change its statistical significance in a majority of cases. Investigators and grant peer review committees should consider prioritizing less-studied topics or requiring formal consideration of optimal information size based on extant evidence in power calculations.
评估最新的随机对照试验（RCT）对Cochrane新生儿综述中每项系统评价（SR）的影响。
我们选择了报告最新研究主要结局风险比的典型点估计值（n = 130）、死亡率（n = 128）以及主要结局平均差（n = 44）的荟萃分析。我们采用累积荟萃分析来确定每次试验加入后的典型估计值，然后进行多变量逻辑回归以确定预测研究影响的因素。
对于所述的主要结局，18%的最新RCT未能缩小置信区间（CI），55%未能使CI降低≥20%。只有8%改变了典型估计值的方向性，11%导致了有或无统计学显著性的改变。最新RCT在18%的情况下未改变典型估计值，只有41%使典型估计值至少改变了10%。使CI缩小>20%的能力与先前发表的RCT数量呈负相关（比值比0.707）。在对死亡率和平均差结局的典型估计分析中也发现了类似结果。
在广泛的临床问题中，大多数情况下，最新的RCT未能大幅缩小典型估计值的CI，未能改变效应估计值或其统计学显著性。研究者和资助同行评审委员会应考虑优先研究较少的主题，或在功效计算中根据现有证据要求正式考虑最佳信息量。
