We did not perform certain sensitivity analyses as planned. First, we did not combine placebo‐ and active‐controlled trials as, on reflection, it did not make clinical sense to combine these data. Therefore, sensitivity analysis was not necessary. Second, there were insufficient data to perform a sensitivity analysis where adults and pediatric patients were looked at separately.

Conversely, we performed three post‐hoc analyses that were not planned in the protocol. First, we looked at whether study design (see Summary of main results) affected our primary outcome (number of patients achieving at least 50% pain relief with IV propacetamol or IV paracetamol versus placebo). Second, we assessed the effect of type of surgery performed on our primary outcome (see Potential biases in the review process) and demonstrated greater efficacy in dental studies versus orthopedic studies. Last, at the request of a referee, we performed a sensitivity analysis using a random‐effects model instead of our original fixed‐effect model. This analysis demonstrated no difference in either the direction or magnitude of effect for any of the primary outcome estimates, but in some of the secondary outcomes estimates that were previously marginally statistically significant became non‐significant.