Table 4

Summary of selected univariate meta-regression results based on different estimation techniques using each study quality variable as a moderator to explain the study heterogeneity present in infarct volume data

MethodVariableEstimate
(SE/MC error)
95% ClP valueτ2I2 (%)R2 (%)
REMLNo covariate35.5 (2.35)30.8 to 40.2<0.0001213.982.4
Blinded induction of ischaemia (intercept)39.0 (2.76)33.5 to 44.5<0.0001204.579.84.36
Blinded induction of ischaemia (true)−11.2 (4.88)−21.0 to −1.460.03
Control of temperature (intercept)38.2 (3.19)31.8 to 44.5<0.0001219.181.00
Control of temperature (false)−5.67 (4.70)−15.1 to 3.720.23
DLNo covariate35.7 (2.35)31.0 to 40.4<0.0001235.283.7
Blinded induction of ischaemia (intercept)39.5 (2.76)34.0 to 45.1<0.0001267.683.80
Blinded induction of ischaemia (true)−11.5 (4.88)−21.2 to −1.730.02
Control of temperature (intercept)38.6 (3.19)32.2 to 45.0<0.0001268.483.90
Control of temperature (false)−6.00 (4.70)−15.4 to 3.400.21
BayesianNo covariate35.5 (0.03)30.3 to 40.7<0.0001224.2
Blinded induction of ischaemia (intercept)38.9 (0.06)32.9 to 45.3<0.0001212.9
Blinded induction of ischaemia (true)−11.1 (0.11)−22.3 to −0.120.04
Control of temperature (intercept)38.0 (0.07)31.0 to 45.4<0.0001227.7
Control of temperature (false)−5.58 (0.10)−16.5 to 4.840.31
  • P values are calculated based on Wald-type test statistics for frequentist approaches and using probability of direction for Bayesian approach as detailed in ref 39. Associated full tables can be found in the Appendix (https://osf.io/6usxq/) (Bayesian priors: Embedded Image, Embedded Image).

  • DL, DerSimonian and Laird; MC, Monte Carlo; REML, restricted maximum likelihood.