# Evidence Synthesis Gone Awry: The Perils of Aggregating Ineffective or Unsafe Doses in Alopecia Areata Reviews

## TL;DR

This paper warns that combining ineffective or unsafe doses in alopecia areata reviews can lead to misleading treatment rankings and offers solutions to prevent this.

## Contribution

The paper introduces methodological safeguards to avoid misleading inferences in evidence syntheses by excluding unapproved doses.

## Key findings

- Including unapproved doses in meta-analyses can distort efficacy and safety estimates.
- Unapproved regimens may be ranked higher than licensed therapies due to flawed aggregation.
- Prespecifying exclusion of unapproved doses and transparently reporting regulatory actions can mitigate these issues.

## Abstract

Network and conventional meta-analyses can increase precision for clinical decision-making but risk producing misleading hierarchies when they pool ineffective, unsafe, or unapproved dosing regimens alongside licensed therapies. Using recent evidence syntheses in alopecia areata, we show how inclusion of small, underpowered dose strata and regimens that never advanced to pivotal trials or were not pursued for regulatory approval (for example, deuruxolitinib 4 mg twice‑daily {BID}, deuruxolitinib 12 mg BID, and ritlecitinib 200 mg loading doses) can distort pooled efficacy and safety estimates, elevate unapproved regimens in rankings, and invite inappropriate causal inferences. We outline key methodological safeguards and offer concrete recommendations as follows: prespecify exclusion or planned sensitivity analyses for unapproved doses, transparently report approval status and relevant regulatory actions (for example, clinical holds) alongside pooled safety estimates, avoid causal attributions from small strata and present uncertainty appropriately, and report sensitivity analyses that omit unapproved doses. Implementing these practices preserves the statistical advantages of meta-analysis while protecting clinicians, guideline panels, and payors from misleading inferences drawn from small or unrepresentative dose groups.