无金标准诊断情况下评价诊断试验准确性的统计学方法

Statistical strategies for assessing accuracy of diagnostic test in the absence of gold standard

评价诊断试验的准确性是诊断试验研究的核心,传统的诊断试验研究需要建立在金标准诊断的基础上。但在医学研究中,常存在无金标准或金标准不完善的情况,给诊断试验准确性评价带来一定挑战。近年,方法学专家提出了多种无金标准或金标准诊断不完善情况下有效评价诊断试验准确性的方法。本文详细阐述了缺失数据填补或偏倚校正、潜在类别分析构建参考标准、Bayesian评价方法、差异化验证/分析等统计学方法在无金标准或金标准不完善情况下评价诊断试验准确性的应用。

The main objective of a diagnostic test study is to confirm its diagnostic accuracy with comparison to traditional gold(or reference)standard.Unfortunately,it is not uncommon that there is no available gold standard or its results are incomplete in medical studies,which is a major challenge for accuracy assessment of a diagnostic test.In recent decades,statisticians developed some methodological strategies to assess accuracy of a diagnostic test under the situations that gold standard is absent or its results are incomplete,including missing value imputation or bias-adjustment,latent class analysis to construct reference standard,Bayesian method,differential verification/discrepant analysis.This article introduced these methodological strategies in detail for application purpose.