MacAma: Multi-AI Agent as a Co-Scientist for Automated Meta-Analysis
Yilin Yuan1, Pingping Li2, Yang Wang3
1State Key Laboratory of Nuclear Physics and Technology School of Physics Peking University Beijing China.
Smart Medicine
|August 4, 2026
Summary
MacAma, a novel AI framework, streamlines meta-analysis by automating tasks like literature screening, reducing manual effort by over 80%. It ensures human verification for critical data extraction, enhancing evidence-based medicine.
Area of Science:
- Biomedical Informatics
- Evidence Synthesis
- Artificial Intelligence in Medicine
Background:
- Traditional meta-analysis workflows are time-consuming and prone to bias.
- Existing AI agents lack data fidelity for rigorous quantitative synthesis, especially with scientific charts.
Purpose of the Study:
- Introduce MacAma, a semi-automated, multi-agent framework for protocol-constrained and human-verifiable meta-analysis.
- Enhance data fidelity and methodological traceability in AI-assisted evidence synthesis.
Main Methods:
- MacAma operationalizes PRISMA 2020, PICOS, and SYRCLE criteria using structured prompts and decision rules.
- Employs a risk-aware automation strategy, delegating low-risk tasks to AI and reserving high-impact steps for expert verification.
- Integrates audit records for transparent workflow tracking.
Main Results:
- In a preclinical radiotherapy study, MacAma reduced manual screening burden by over 80%.
- Demonstrated competitive screening performance against benchmarks.
- Showcased AI's capability to support transparent screening, data extraction, synthesis, and drafting.
Conclusions:
- MacAma offers a scalable and auditable framework for AI-assisted meta-analysis.
- Highlights the potential for AI to improve efficiency and transparency in evidence synthesis.
- Acknowledges limitations in full-text access, quantitative chart data extraction, and heterogeneity interpretation.