Department of Clinical Epidemiology, Biostatistics and Bioinformatics, University of Amsterdam, PO Box 22700, Amsterdam 1100 DE, The Netherlands.
J Clin Epidemiol. 2012 Oct;65(10):1088-97. doi: 10.1016/j.jclinepi.2012.03.006. Epub 2012 Jun 27.
Meta-analysis of predictive values is usually discouraged because these values are directly affected by disease prevalence, but sensitivity and specificity sometimes show substantial heterogeneity as well. We propose a bivariate random-effects logitnormal model for the meta-analysis of the positive predictive value (PPV) and negative predictive value (NPV) of diagnostic tests.
Twenty-three meta-analyses of diagnostic accuracy were reanalyzed. With separate models, we calculated summary estimates of the PPV and NPV and summary estimates of sensitivity and specificity. We compared these summary estimates, the goodness of fit of the two models, and the amount of heterogeneity of both approaches.
There were no substantial differences in the goodness of fit or amount of heterogeneity between both models. The median absolute difference between the projected PPV and NPV from the summary estimates of sensitivity and specificity and the summary estimates of PPV and NPV was 1% point (interquartile range, 0-2% points).
A model for the meta-analysis of predictive values fitted the data from a range of systematic reviews equally well as meta-analysis of sensitivity and specificity. The choice for either model could be guided by considerations of the design used in the primary studies and sources of heterogeneity.
一般不鼓励对预测值进行荟萃分析，因为这些值会受到疾病流行率的直接影响，但敏感性和特异性有时也会表现出显著的异质性。我们提出了一种双变量随机效应对数正态模型，用于诊断测试阳性预测值（PPV）和阴性预测值（NPV）的荟萃分析。
重新分析了 23 项诊断准确性的荟萃分析。我们使用单独的模型，计算了 PPV 和 NPV 的汇总估计值以及敏感性和特异性的汇总估计值。我们比较了这两个汇总估计值、两个模型的拟合优度以及两种方法的异质性程度。
两个模型的拟合优度或异质性程度没有显著差异。敏感性和特异性汇总估计值与 PPV 和 NPV 汇总估计值之间预测的 PPV 和 NPV 的中值绝对差异为 1 个百分点（四分位间距，0-2 个百分点）。
预测值荟萃分析的模型同样可以很好地拟合来自一系列系统评价的数据，与敏感性和特异性的荟萃分析一样。可以根据原始研究的设计考虑因素和异质性来源来选择使用哪种模型。
