Meta-analysis of diagnostic tests with imperfect reference standards

Area of Science:

• Biostatistics
• Medical Informatics
• Diagnostic Accuracy Research

Background:

• Combining evidence from multiple diagnostic accuracy studies is crucial for clinical decision-making.
• Existing meta-analysis methods often assume an error-free reference standard, which can bias results.
• Accurate estimation of diagnostic test performance requires robust statistical approaches.

Purpose of the Study:

• To develop and present a novel method for estimating summary receiver operating characteristic (SROC) curves.
• To incorporate the possibility of errors in the reference standard within diagnostic accuracy meta-analyses.
• To provide a more reliable approach for assessing diagnostic test performance across studies.

Main Methods:

• Utilized a latent class model to account for potential errors in the reference standard.
• Estimated study-specific sensitivity, specificity, and case prevalence.
• Employed a meta-analysis regression method (e.g., Moses et al.) to fit the SROC curve using estimated parameters.

Main Results:

• The proposed method allows for imperfect reference standards, unlike traditional approaches.
• Adjusting for reference standard imperfections typically reduces data scatter in SROC plots.
• The method often indicates improved diagnostic test performance compared to analyses that ignore reference standard errors.

Conclusions:

• The latent class model approach provides a more accurate estimation of SROC curves by addressing reference standard imperfections.
• This method enhances the reliability of meta-analyses for diagnostic test evaluation.
• The findings suggest that accounting for reference standard errors can lead to a more favorable assessment of diagnostic test utility.