In some of the meta-analyses I've worked on, even after applying a random-effects model, the heterogeneity remains very high (I² > 75%), and the confidence intervals stay wide. In such cases, I’ve considered meta-regression, subgroup analysis, and even sensitivity analysis, but I’m curious about how others approach this challenge—especially when journal reviewers raise concerns.
Do you typically exclude outliers, downgrade the evidence level, or proceed with narrative synthesis in parallel? I’d appreciate hearing your methods, experiences, or tools you find helpful.