To the Editor:
Apfel et al.1 have devised a simple scoring system to quantify the risk of a patient having postdischarge nausea and vomiting (PDNV). I would like to commend the authors for being able to distil and present the multiple factors contributing to postoperative nausea and vomiting (PONV) and PDNV.
However, I have a few issues to raise. First, while the authors’ analysis described five seemingly independent parameters, there may still be residual confounding to be addressed. Given the emetogenic side effects of opioids, opioids given in the postanesthesia care unit and nausea in the postanesthesia care unit are not independent factors. Similarly for female gender and history of PONV, it is plausible that a difference in reporting a past history of PONV exists between the different genders. The authors’ analysis does not clearly resolve these cause-and-effect relationships, and leaves the possibility for residual bias. If the factors were positively correlated to each other, rather than to an independent outcome of PDNV, it would still fit a linear regression model.
Also, certain factors that were significant for PONV in the postanesthesia care unit were not statistically significant for PDNV; however, vomiting in the postanesthesia care unit which incorporates these factors is still statistically significant. Would the authors then consider these factors to be clinically, even if not statistically, significant?
Second, I would like to discuss the clinical relevance of the score. Unlike a similar article published by Apfel et al.,2 in this case the authors do not suggest a cutoff where postdischarge antiemetogenic therapy should be prescribed. I would like to prompt the authors to provide a recommendation on this and demonstrate the clinical impact even with a hypothetical number needed to treat analysis.
In his previous article, Apfel et al.2 has indicated that he would consider a cutoff of 39% risk of PONV to treat (where two risk factors are present). Assuming a similar case where two risk factors are present in the current study for PDNV (30% risk), why not simply treat all laparoscopic patients when there is a 38% incidence of PDNV, whether or not any risk factors are identified? This strategy is unable to be tested, because in some instances the patient characteristics differ quite significantly from their development and validation groups, such as for laparascopic approach (development 14.8%, validation 1.2%), and several other nonstatistically significant factors which differ by a factor of 3–10.
In light of the number needed to treat figures described by Gupta et al.,3 where the number needed to treat of prophylactic combination therapy for PDNV is about 5, would there be any significant benefit with a risk stratification strategy in the overall number needed to treat?
The fundamental issue is that PONV and PDNV are issues with a high incidence rate, and we should work on strategies to mitigate this problem. Apfel et al. have taken a significant step toward addressing this problem in a scientific manner. But would any risk score be able to significantly affect clinical practice?