Evaluating novel agent effects in multiple-treatments meta-regression.

Area of Science:

• Medical research methodology
• Biostatistics
• Evidence synthesis

Background:

• Multiple-treatments meta-analyses are crucial for comparing competing medical regimens.
• Newer treatments in evolving fields may show exaggerated effectiveness due to novelty bias.
• This bias, stemming from optimism or conflicts of interest, is difficult to detect in pairwise meta-analyses.

Purpose of the Study:

• To develop and validate a statistical model for detecting and adjusting for novelty bias in multiple-treatments meta-analyses.
• To quantify the magnitude of novelty bias in comparative effectiveness research.
• To assess the impact of adjusting for novelty bias on treatment rankings.

Main Methods:

• Developed a Bayesian multiple-treatments meta-regression model.
• Applied the model to analyze chemotherapy and non-hormonal systemic treatments for three malignancies.
• Explored different assumptions regarding the exchangeability of novel agent effects across studies.

Main Results:

• The developed meta-regression model effectively identifies and quantifies novel agent effects.
• In the analyzed cancer treatment networks, novelty bias led to an estimated 6% exaggeration of the hazard ratio (95% credible interval: 2-11%).
• Adjusting for novelty bias altered the relative rankings of different treatment regimens.

Conclusions:

• Multiple-treatments meta-regression is a valuable tool for assessing novelty bias in comparative effectiveness research.
• Novelty bias can systematically overestimate the effectiveness of newer agents.
• Accounting for novelty bias provides a more accurate assessment of treatment efficacy and ranking.