A Bayesian hierarchical model for network meta-analysis of multiple diagnostic tests.

Area of Science:

• Medical Statistics
• Diagnostic Test Evaluation
• Biostatistics

Background:

• Traditional meta-analysis of diagnostic tests (MA-DT) primarily evaluates single tests against a reference standard.
• Multiple diagnostic tests and diverse study designs (multiple test comparison, randomized, non-comparative) complicate comparative accuracy assessment.
• A need exists for a flexible framework to synthesize evidence from various study designs for simultaneous inference.

Purpose of the Study:

• To develop a Bayesian hierarchical model for network meta-analysis of diagnostic tests (NMA-DT).
• To create a missing data framework enabling the combination of diverse study designs and data types.
• To offer a flexible approach for simultaneous inference on multiple diagnostic tests.

Main Methods:

• Development of a Bayesian hierarchical model for network MA-DT (NMA-DT).
• Implementation of a missing data framework to integrate studies with different designs and gold standard availability.
• Application of the NMA-DT method to a case study involving deep vein thrombosis tests.

Main Results:

• The proposed NMA-DT framework successfully combines studies utilizing multiple test comparison, randomized, and non-comparative designs.
• The method accommodates studies with and without a gold standard, and those with varying sets of candidate tests.
• The framework accounts for heterogeneity across studies and complex correlations among multiple diagnostic tests.

Conclusions:

• The developed NMA-DT framework provides a significant advancement over traditional MA-DT methods.
• This approach enables a more comprehensive and flexible synthesis of evidence for multiple diagnostic tests.
• The method is applicable to real-world scenarios, as demonstrated by the deep vein thrombosis case study.