Model estimates (unconditional) and 95% credible intervals for lnRR and lnSD.

Estimates of lnSD were obtained from an arm-based, multilevel regression model (MLMR) of lnSD in infarct volume for both treatment and control groups. Original fixed (drug treatment group and sex) and random effects (study ID, effect size ID, and strain) were fit, in addition to a nested random effect of “Drug treatment group | pairwise ID.” Estimates of lnRR are from the main analysis of mean drug treatment effects and are the same as in S3 Table. Treatment effects (DrugGroup) are ordered from groups that produce, on average, the greatest reduction in infarct volume (i.e., the most effective, as indicated by most negative estimates of lnRR) to groups that are, on average, the least effective. lnRR, log response ratio; lnSD, log standard variation; MLMR, multilevel meta-regression.

(DOCX)