Outlier detection and influence diagnostics in network meta-analysis

Area of Science:

• Biostatistics
• Evidence Synthesis
• Pharmacological Research

Background:

• Network meta-analysis (NMA) synthesizes multiple treatments but can be skewed by influential outlier studies.
• Identifying and addressing these outliers is critical for unbiased and reliable NMA results.

Purpose of the Study:

• To propose novel frequentist methods for detecting outlying and influential studies within NMA.
• To enhance the accuracy and trustworthiness of evidence synthesis in comparative effectiveness research.

Main Methods:

• Developed four influence measures for NMA, including leave-one-trial-out cross-validation (comparison-specific studentized residual, relative change measures for covariance and heterogeneity matrices).
• Proposed a model-based approach using a likelihood ratio statistic with a mean-shifted outlier detection model.
• Applied methods to an NMA of antihypertensive drugs, including models with missing outcomes and adjusted degrees of freedom.

Main Results:

• Successfully identified three influential trials in the antihypertensive drug NMA, including one retracted due to data falsification.
• Demonstrated that omitting these influential studies significantly altered comparative efficacy estimates and drug rankings.
• Validated the effectiveness of the proposed detection methods in a real-world NMA scenario.

Conclusions:

• The proposed frequentist methods effectively detect outlying and influential studies in network meta-analysis.
• Exclusion of identified influential studies can substantially change NMA outcomes, highlighting the importance of outlier detection.
• These methods improve the reliability of evidence synthesis for clinical decision-making.