Six RCTs were included in the review (n=385 patients, range 33 to 90) . Two trials had adequate randomisation sequences and allocation concealment. One trial had intention-to-treat analysis. All the trials were double blind.
The incidence of post-stroke depression was significantly reduced with fluoxetine compared to control (OR 0.25, 95% CI 0.11 to 0.56; three RCTs). Reductions in the rate of post-stroke depression were significantly different when fluoxetine was given within one week after the occurrence of stroke (OR 0.15, 95% CI 0.05 to 0.51; one RCT), but there was no evidence of a difference when fluoxetine was given four weeks or more after stroke (one RCT). There was no evidence of a significant difference in depression scores on the Hamilton Depression Rating Scale between groups (WMD -3.97, 95% CI -9.85 to 1.9; four RCTs); this finding had substantial heterogeneity (p<0.00001; I2=97%).
A statistically significant difference was found in the recovery of neurological impairment between groups (WMD -4.72, 95% CI -8.31 to -1.13; two studies); there was substantial heterogeneity (P=0.02; I2=82%). There was a significant reduction in the Barthel Index scores for activities of daily living with fluoxetine compared with control (WMD -8.04, 95% CI -13.4 to -2.68; two studies) and in the Functional Independence Measure scores (difference not reported; p<0.05; one RCT).
There was no evidence of a difference between groups in drop-out rates (OR 0.88, 95% CI 0.31 to 2.49; three RCTs). There was also no evidence of a difference between groups in the rates of individual adverse events. However, incidence of nausea, insomnia and epileptic seizure were greater than 10% in the fluoxetine group compared with 0% in the control group.
Sensitivity analysis with the recalculation of summary effects in trials with sample sizes greater than 50 patients and trials with unclear randomisation method indicated that results were similar.
There was no evidence of publication bias from the fail-safe number.