We thank Li et al. for these perspectives. We agree that nonrandomized studies have greater risk of bias and confounding, and the results may therefore be misleading. This certainly applies to studies using propensity-based methods. We would first like to point out that in their letter Li et al. state we used propensity score matching. In fact, we actually used inverse probability of treatment weighting. These are distinct methods (although both based on propensity scores) and estimate different quantities (effect of treatment overall vs. effect of treatment in the treated).
More importantly however, our comments regarding the risk mitigation associated with antiemetic prophylaxis in patients exposed to nitrous oxide were based not on the secondary analysis referred to by Li et al. but in a preplanned secondary analysis of the original large randomized trial.1 Relevant, expanded details are provided in table 1. The emetogenic effect of nitrous oxide was less apparent in those who received prophylactic antiemetics before the end of surgery compared with those who did not. The interaction P value was 0.001, indicating that there was a statistically significant differential effect between these two subgroups. We acknowledge that use of antiemetic prophylaxis was left to the discretion of the attending anesthesiologist, but such use was more likely in those with more risk factors for postoperative nausea and vomiting (PONV; as we reported in our publication).2 That is, there was a selection bias, but it would underestimate the protective effect of antiemetic prophylaxis because such use was higher in those with greater risk of PONV. We therefore stand by our conclusion that PONV prophylaxis near-eliminates the risk of nitrous oxide-induced severe PONV after major surgery.
The authors declare no competing interests.