Objectives: To determine if results of meta-analyses with continuous data are different if final value (treatment end value) only, change value (treatment end value - baseline value) only or a combination of these are used.

Methods: Baseline and final values were simulated by generating bivariate normal random variables ( 1 x , 2 x ) from normal random variables ( 1 z , 2 z ) by using the equations below:

Formulae not included

Baseline and final values for outcomes were simulated for each participant in both the experimental and control groups. Several sets of baseline and final values were generated to simulate different trial data. Change values were calculated from baseline and final values. Meta-analyses were performed with final values or change values or a combination of these.

Results: Data for five trials were simulated. Of the five trials two involved 1400 participants each and three involved 400 participants each. The weighted mean difference (WMD) and 95% confidence interval (CI) obtained by meta-analyses with only final values and only change values were respectively -0.10 (-0.17, -0.02) and -0.17 (-0.28, -0.05). The WMD obtained using final values appeared less than that obtained using change values but this difference was not statistically significant (p = 0.36). For meta-analysis with final values for the three small trials and change values for the two large trials, and for meta-analysis with the reverse case, the WMD (95% CI) were -0.17 (-0.30, -0.04) and -0.10 (-0.16, -0.03) respectively. Neither value was significantly different from results with final or change values alone.

Conclusions: Preliminary results suggest that there is no significant difference in WMD (95% CI) values obtained by meta-analysis with final values, change values or their combination. Further studies are in progress to determine if there are instances when there is a difference.