Nineteen trials were included in the review (n=3,504 patients, range 21 to 984). All the trials reported randomisation, but only six trials specified method of randomisation. Four trials reported allocation concealment. Two trials reported blinding, but in only one of these was blinding adequate. Fourteen trials reported follow-up. Seven trials had detailed reporting allowing the use of intention-to-treat analysis.
Pneumonia (19 trials): Patients receiving epidural analgesia had statistically significantly reduced odds of contracting pneumonia compared with those receiving systemic analgesia (7.5% of patients in epidural arm; 12.8% of patients in systemic arm; OR 0.54, 95% CI 0.43 to 0.68; NNT=18, 95% CI 14 to 27; p value for heterogeneity <0.10), but heterogeneity was statistically significant.
Prolonged ventilation (seven trials): Epidural analgesia significantly reduced the odds of prolonged ventilation compared with systemic analgesia (OR 0.61, 95% CI 0.40 to 0.93; NNT=30, 95% CI 19 to 167; p value for heterogeneity >0.10).
Re-intubation (seven trials): Epidural analgesia significantly reduced the odds of re-intubation compared with systemic analgesia (OR 0.70, 95% CI 0.55 to 0.88; NNT=21, 95% CI 14 to 62; p value for heterogeneity >0.10).
A wide range of subgroup analyses were performed. One important result was that older trials (published between 1970 and 1980; OR 0.17, 95% CI 0.05 to 0.57) showed a greater difference in the odds of catching pneumonia between epidural and systemic analgesia patients than newer trials (published from 2000 onwards; OR 0.62, 95% CI 0.47 to 0.81).
There was no evidence of publication bias for the primary outcomes.