Postoperative Nausea and Vomiting Database Research: Limitations and Opportunities

Ashraf S. Habib, MBBCh, MSc, FRCA, and Tong J. Gan, MD, MHS, FRCA, FFARCS(I)

From the Department of Anesthesiology, Duke University Medical Center, Durham, North Carolina.

Address correspondence and reprint requests to Ashraf S. Habib, MBBCh, MSc, FRCA, Duke University Medical System, Box 3094, Durham, NC 27710. Address e-mail to

Postoperative nausea and vomiting (PONV) has a multifactorial etiology. Some advocate universal combination antiemetic prophylaxis given the high efficacy, relatively inexpensive drug acquisition cost, and favorable side-effect profile of the most frequently used antiemetics.1 However, because many patients will not experience PONV even without prophylaxis, and there are side effects and added costs, albeit low, associated with current interventions, others recommend a risk-adapted strategy for PONV prophylaxis.2 In this context, identification of risk factors for PONV allows the anesthesiologist to tailor the anesthetic management and prophylactic antiemetic administration to patients' underlying risks, with more aggressive prophylaxis being reserved for high-risk patients. However, our current understanding of risk factors for PONV is far from being complete. Numerous studies have identified several patient, anesthesia, and surgical factors that might increase the risk for PONV. A number of scoring systems incorporating those risk factors have also been published. Given the multifactorial etiology of PONV, and the gaps in our understanding of the risk factors, available scoring systems have only shown poor to moderate accuracy with areas under the receiver operating characteristic curve ranging from 0.56 to 0.785 (0.5 being no better than tossing a coin).3–9

In general, there has been agreement about a number of patient-related risk factors, specifically female gender, history of PONV or motion sickness, and smoking status.3 A number of studies also confirmed numerous anesthesia-related risk factors including longer duration of anesthesia and the use of volatile anesthetics, nitrous oxide, and opioids (both intra- and postoperative).10 There has been inconsistency, however, and controversy about the type of surgery as being an independent risk factor for PONV. Although numerous studies have identified the type of surgery as a risk factor,7,8,11–13 the specific procedures implicated as particularly emetogenic did vary and were inconsistent among studies. Types of procedures that were reported as possible risk factors include abdominal, laparoscopic, major gynecologic, orthopedic, ear, nose, and throat, breast, plastic surgery, neurosurgery, and, in children, hernia repair, adenotonsillectomy, strabismus, penile surgery, and orchiopexy. However, only 3 of 8 published PONV risk scoring systems included the type of surgery as one of the risk factors.5,7,9 Some argue the fact that gynecologic surgery has been reported as a risk factor, for instance, is not due to the surgery itself but due to the high risk for PONV of the female patients undergoing this surgery, and that scores considering the type of surgery do not lead to better prediction of PONV compared with procedure-independent models.14 Therefore, there is a clear need for large well-designed studies to address this controversy.

In this issue of the journal, Ruiz et al.15 specifically investigated the impact of the surgical procedure on the risk of PONV. They used their electronic anesthesia information system to collect data on all patients who were admitted to the postanesthesia care unit (PACU) over a 2-yr period. Known risk factors for PONV were collected and documented in all patients. In addition to the surgical procedure, they included 9 potential risk factors in the general linear model: gender, smoking status, history of PONV or motion sickness, duration of anesthesia, number of prophylactic antiemetics administered, dose of intraoperative opioids, use of ketorolac, use of an epidural for postoperative analgesia, and use of postoperative opioids. To study the impact of the surgical procedure, they arbitrarily categorized all surgical procedures on the basis of the anatomical site and used a group undergoing integumentary, musculoskeletal, and superficial (IMS) surgeries as a reference group. The end point used was the need for antiemetic administration within the first 2 h of PACU admission. They reported that, compared with IMS surgeries, neurological, head/neck, and abdominal surgeries were associated with greater PACU use of rescue antiemetics. Breast/axilla and endoscopic procedures did not differ from the IMS group, whereas thoracic procedures were associated with significantly less PACU antiemetic administration. The authors also found that PACU antiemetic administration was significantly associated with previously known risk factors for PONV, such as female gender, history of PONV or motion sickness, nonsmoking history, anesthesia duration, and intraoperative and postoperative opioid administration.

This study involved a large number of patients (18,109), being the second largest study published to date investigating this question, and the patient risk factors for PONV seem to be carefully documented and recorded for all included patients. However, despite these strengths, it is retrospective and, similar to examination of any observational data, is subject to the influence of unmeasured confounding variables. The dataset also lacks a number of variables, such as the volume of intraoperative fluids given and the dose of postoperative opioids, that could affect the incidence of PONV and therefore the receipt of rescue antiemetics. Furthermore, rather than study individual types of surgery, the authors chose to classify surgical procedures into 7 anatomical groups. However, there seems to be little evidence to support the categorization of the surgical procedures "anatomically" as was done by the authors. There is also little information in this article to show that the receipt of rescue antiemetics (which will affect the results) was homogeneously distributed within the different surgical procedures included in the same group. Within the neurosurgery group, for instance, the authors reported an overall PACU antiemetic administration of 36.2%. However, the group of patients who underwent hypophysectomy (26% of the neurosurgery group) had a PACU antiemetic administration rate of almost 87%, showing the wide variation in the rate of PACU antiemetic administration within the neurosurgery group. Other studies suggest that the incidence of PONV might be different among procedures included in the same anatomical group in this study. For example, the need for rescue antiemetic therapy in the PACU may be lower in patients undergoing robotic versus open prostatectomy.16 Similarly, hysteroscopy was included in the endoscopy group together with bladder and prostate procedures, whereas previous data suggest a higher incidence of PONV with hysteroscopy.7 Furthermore, Sinclair et al.7 reported in a prospective study a wide variation in the incidence of PONV in the PACU among the different procedures within the same surgical specialty such as orthopedic and ophthalmic surgery. Breast surgery also has been reported to be associated with a high risk for PONV.7 In this analysis, however, the breast/axilla group did not receive higher PACU antiemetic administration compared with the reference group. The inclusion of other procedures within this group or the confounding influence of the aggressive antiemetic prophylaxis that was used in this group might explain this unexpected finding. Therefore, this arbitrary classification into anatomical groups risks inaccurate interpretation of the results, giving the impression that the various procedures within the same category are associated with a similar degree of PONV risk, which is not the case. This limits the practical "take home" message from the study.

Another limitation of this study includes the use of rescue antiemetic administration in the PACU as a surrogate for the occurrence of PONV. Although this is a reasonable surrogate, it does not have the same accuracy as the prospective collection of data on the incidence of nausea, vomiting, and retching. Not all vomiting episodes will be treated with rescue antiemetics because patients frequently feel better after an episode of emesis. Mild nausea might not be treated with a rescue antiemetic, and the threshold for treatment might vary among different PACU nurses. The use of this end point also does not take into account the severity of nausea or the frequency of vomiting episodes. Including data from only the first 2 h after surgery limits the conclusions that could be drawn from these results, because some patients might develop PONV for the first time after PACU discharge, and as care providers, we are interested in predicting PONV not only in the PACU but, more importantly, beyond PACU discharge when patients no longer have close monitoring and frequent nursing observations as in the PACU. Most studies have reported that the cumulative incidence of PONV at 24 h is higher than at 2 h.

In the study by Ruiz et al., the authors considered all prophylactic antiemetics as having similar efficacy. Whereas there are good data to support the fact that ondansetron, dexamethasone, and droperidol have comparable efficacy,10 the same cannot be said for other antiemetics included in this analysis. For instance, metoclopramide in a dose of 10 mg is less effective than both ondansetron and droperidol.17 Similarly, other included antiemetics such as promethazine do not have sufficient good quality data in the literature to support their efficacy as prophylactic antiemetics.

Despite these limitations, this analysis did reproduce some of the previously known prospectively studied risk factors, therefore giving validity to the information provided by the authors. Furthermore, the large number of patients included in the analysis lends further strength to the study. To avoid some of the limitations of database research, it is important to determine prospectively the questions that might need to be answered using the database to ensure that all data required to address this issue are prospectively collected. To optimize the use of databases for PONV research for instance, specific end points such as the occurrence of PONV, severity of nausea, and frequency of vomiting and retching need to be documented, in addition to the use of rescue antiemetics. There is also a need to obtain information beyond the PACU period. The use of databases also represents an opportunity for multicenter collaboration on collecting outcome data. For instance, the Society for Ambulatory Anesthesia is establishing an outcome database with the hope that prospectively collected well-defined outcomes using standardized forms by a large number of ambulatory centers will provide useful data. Although this is not a substitute for prospective, randomized controlled trials, such prospective collection of planned standardized data could represent an opportunity to overcome some of the limitations of database research and generate valuable data that could be used for hypothesis generation, filling gaps in our knowledge, and ultimately improving the care of patients.