THE rapid increase in ambulatory surgery procedures performed over the past 10 yr has resulted in greater focus directed toward the control of postoperative nausea and vomiting (PONV), sometimes appropriately termed the “big, little problem.”1The incidence of this complication is variable, 2but PONV has emerged as one of the commonest complications following ambulatory surgery, 3and the one that patients would most like to avoid. 4Specifically, the postdischarge period has been poorly studied but is important from the patient's perspective.
Despite the increasing availability of newer and more expensive drugs for the prevention of PONV, some authors question the routine use of single-drug prophylaxis in low-risk patients. 5The long-term (> 6 h) effect of antiemetics is even more uncertain in the ambulatory setting, because the focus of attention has previously been on the management of early PONV (in the day surgical unit). As many as 35–50% patients continue to have postdischarge nausea and vomiting (PDNV). 6,7It is important to control this symptom after discharge, because resumption of normal activities may be delayed if PONV is prolonged 6and ambulatory surgical patients are not under direct medical supervision after discharge. 7In our experience, few centers routinely provide antiemetics for the control of PDNV at home. Numerous systematic reviews have now been published in the literature on PONV, but only one has focused on PONV after hospital discharge in ambulatory surgical patients. 8
This systematic review was completed to address the question of whether the routine prophylactic use of antiemetics affects the incidence of PDNV following ambulatory surgery. We restricted our analysis to randomized, controlled studies published in the English literature.
Citation Search Strategy
The primary question for this systematic review was: Does prophylactic management of nausea and vomiting affect postdischarge outcome following ambulatory surgery? We searched MEDLINE (1966 through May 2002) via PubMed using the MeSH terms:postoperative, nausea, vomiting , and postoperative nausea and vomiting . The results were limited to studies in humans, the English language, and the adult patient population (> 19 yr old). Publication types were limited to clinical trials, controlled clinical trials, or multicenter studies. After obtaining a list of all the identified articles, a careful hand-search was done from the reference list of all relevant articles. The related articles search strategy of PubMed was also used to identify any missing articles. Finally, review articles or systematic reviews of PONV were searched to identify articles missing from the original list.
The following criteria were used to identify appropriate studies that could be included in the analysis: human studies in adults published in the English literature of patients undergoing ambulatory surgery. The studies also had to be controlled and randomized with the specific primary endpoint of assessing nausea or vomiting after discharge for at least 24 h after surgery.
Studies were excluded if they presented data for and included the first postoperative 24 h (i.e. , were not limited to the postdischarge period). If a study presented a nausea score or a visual analog scale for nausea without providing the number of patients with nausea, the data were not included in the analysis. Those studies in which the authors did not use the terms ambulatory surgery or outpatient surgery but instead used major surgery or the patients were admitted to the ward were also excluded from the analysis. Those studies in which direct comparisons were made between two drugs without a placebo group were also excluded.
Data extraction was done independently by two primary reviewers (A.G., N.E.), who were unaware of each other's results. Data included aspects of patient characteristics, anesthesia, type of surgery, and incidence of PONV at 0–6 h (predischarge) and postdischarge. Extracted data were compared for agreement between the two primary reviewers by a third reviewer (C.L.W.). Any discrepancies were noted and discussed further to come to an agreement between all reviewers. Studies in which combinations of drugs were used for prophylaxis against PONV (> 1 antiemetic drug) were treated and analyzed together, irrespective of the type of antiemetic used (combination treatment).
Data Synthesis and Analysis
Only when data were extractable from more than one study was it subject to a meta-analysis. Dichotomous data extracted from the relevant studies were entered into the program RevMan 4.1 (Review Manager; Cochrane Collaboration, Oxford, U.K.) and analyzed. The relative risk with the corresponding 95% confidence intervals were calculated for each study, and the results were pooled together using the Mantel-Haenszel Method for combining trials. The overall estimate of pooled effect was calculated. Heterogeneity was determined under the assumption (null hypothesis) that there were no differences in treatment effect between trials. For combined data, a fixed effect model was used when there was no significant heterogeneity (P > 0.05). The numbers-needed-to-treat were calculated for those antiemetics in which a significant overall effect was seen (P < 0.05) by the reciprocal of the control event rate (CER) − experimental event rate (EER): 1/CER − EER. The chi-square test was used for calculating differences in the overall incidence of PONV or PDNV. A P value < 0.05 was considered to be statistically significant.
The results from outpatients and ambulatory surgical procedures were combined using the OR function to yield 21,739 articles. The results from nausea and vomiting were combined to yield 38,561 articles. The term postoperative was combined with the above searches using the AND function to reveal 495 articles. As a last step, the LIMITED function was used to limit articles to humans, to the English language, and to the adult population (> 19 yr) and was restricted to clinical trials or controlled clinical trials or multicenter study using the function PUBLICATION TYPE, which yielded 189 articles. The abstracts of these 189 articles were read to see if they addressed the primary question. Forty-eight articles reported a primary aim of comparing drugs in the control of PONV. These 48 articles were further analyzed to see if any data were reported that could be extracted on postdischarge nausea or vomiting (PDNV). A hand-search was also done through the reference lists of these articles to find any “missing” articles. We realized that some articles were missed because authors had not used the term outpatients or ambulatory surgical procedures in their study, and that the MeSH term ambulatory surgical procedure was introduced as late as in 1998. Therefore, a second search was done without these MeSH terms to reveal a total of 1,315 articles, which were searched to see if they addressed the primary question. A final count of 22 articles was found in which the appropriate postdischarge data (nausea or vomiting) could be extracted (table 1). 9–30The reasons for excluding the articles from the final list were: primary aim different from the assessment of PONV, lack of postdischarge data, studies in which data combined the immediate postoperative period and the 24-h postoperative period (not restricted to the postdischarge period) into a single data point, studies in which no comparison was made against a placebo, and studies in which nausea was reported as a visual analog scale or score rather than the number of patients with nausea.
Of the studies that met our inclusion/exclusion criteria, 19 reported the number of patients with postdischarge nausea (PDN), and 21 reported postdischarge vomiting (PDV) during the first 24 h. Primary admission because of persistent PONV was reported in five studies. No patients were readmitted because of PDNV in any study. The need for antiemetics after discharge to home was reported in only one study. 10
The demographic data of the population studied and the type of surgery in the 22 studies included in the meta-analysis are shown in table 1. Of the total 3,629 patients studied, 1,216 were given a placebo (33.5%) and 2,413 were given an active drug (66.5%). The mean age of the patients in the treatment group was 33.2 yr and in the placebo group, 34.1 yr (P > 0.05). The majority of patients studied were women (88.3%) undergoing gynecologic surgery.
The following drugs were compared with placebo: droperidol < 1 mg (four studies) and > 1 mg (four studies); cyclizine 50 mg (one study); metoclopramide 10 mg (four studies); betamethasone 12 mg (one study); scopolamine 1.5 mg (one study); prochlorperazine 6 mg (one study); dexamethasone 4–10 mg (four studies); and ondansetron 1 mg (one study), 4 mg (13 studies), and 8 mg (one study). In addition, four studies compared combination treatment (2 or 3 drugs) against placebo. Meta-analysis was not performed when less than two 2 studies with extractable data could be identified.
A total of 815 patients had PDN (32.6%). The number of patients who had PDN in the placebo and treatment groups was 276 (35.7%) and 539 (31.2%), respectively (P < 0.05). The overall incidence of postoperative nausea in these studies was 26.0% and 40.4% in the treatment and placebo groups, respectively (P < 0.05). When comparing different drugs with placebo in a meta-analysis of all studies with relevant data, ondansetron 4 mg, dexamethasone 4–10 mg, and combination treatment with more than one drug had a significantly lesser relative risk compared to placebo for PDN (P < 0.05) (table 2). The pooled data for the relative risk of PDN in the ondansetron 4 mg and combination treatment groups are shown in figures 1 and 2, respectively. No significant heterogeneity was observed between the studies. The overall incidence of PDN was significantly lower in the ondansetron 4 mg group (39.9%) than in the placebo group (48.1%) (P < 0.05). The number-needed-to-treat was large with all drugs except combination treatment, when single studies were excluded (table 2). In patients treated with ondansetron, the overall incidence of PDN in a subgroup of desflurane-anesthetized patients was 44.6% compared to 38.3% for isoflurane (not significant). The relative risks for desflurane and isoflurane were 0.75 (CI, 0.59–0.96) and 0.66 (CI, 0.47–0.92) compared to placebo.
A total of 497 patients had PDV (14.7%). In the placebo group, 19.6% of patients had PDV, whereas 12.1% of those in the treated group had PDV (P < 0.05). The overall incidence of postoperative vomiting was 14.6% and 26.5% in the treatment and placebo groups, respectively (P < 0.05). The relative risk for ondansetron 4 mg and combination treatment with two or more drugs was significantly lower than with placebo for PDV (P < 0.05). The pooled data for the relative risk of PDV in the ondansetron 4 mg and combination treatment groups are shown in figures 3 and 4, respectively. No significant heterogeneity was observed between studies. The relative risk for PDV in patients treated with dexamethasone was 0.39 (CI, 0.14–1.05) (not significant). No drug had a significantly different relative risk as compared to placebo. The number needed to treat for ondansetron was 13.8 compared to 5.0 for combination therapy (table 3). The incidence of PDV in a subgroup of desflurane or isoflurane anesthetized patients, all of whom received ondansetron, was 17.4% and 22%, respectively (not significant). The relative risks for desflurane and isoflurane were 0.46 (CI, 0.30–0.71) and 0.55 (CI, 0.33–0.93) as compared to placebo.
Numerous studies have examined the effects of various antiemetics on PONV in ambulatory surgical patients before hospital discharge. In our systematic review of all randomized controlled trials, we observed an overall beneficial effect with the use of either combination treatment or ondansetron 4 mg alone for the prevention of postdischarge nausea and vomiting. A similar beneficial effect was also seen with dexamethasone, which prevented nausea but not vomiting in the postdischarge period after ambulatory surgery.
Postdischarge Nausea and Vomiting
The overall incidence of PDN was found to be 32.6%, with significant differences between treatment and placebo groups. This is much higher than that reported earlier in one systematic review published recently (17%) 8but is similar to that in another prospective randomized study in which the authors specifically addressed the issue of PDNV (30%). 6Similarly, the incidence of PDV was 14.7%, with a significant difference between placebo and treatment groups, which is similar to that in the study by Carroll et al. 6but is higher than that reported previously by Wu et al. (8%). 8The authors of the latter study did not analyze randomized, controlled studies, which may explain their reported lower incidence of PDNV.
When studying the different drugs individually, however, no differences were found between placebo and treatment groups for droperidol (< 1 mg and > 1 mg) and metoclopramide (10 mg). However, both prophylactic ondansetron and dexamethasone in a dose of 4 mg can reduce the risk of PDNV. Similarly, combination management of PONV with more than one drug leads to a reduction in the relative risk for PDNV. The other drugs and different doses used with extractable data from less than two studies were not included in the meta-analysis.
Ondansetron has an elimination half-life of approximately 3.5 h in young, healthy volunteers, 31which would not explain the long duration of action after a single intravenous dose administered perioperatively. The incidence of PDNV in the placebo group in the ondansetron trials was much higher than the average incidence of PDNV observed in all trials (48.1%vs. 35.7%, respectively), which is perplexing but may be attributable to a multitude of factors, including the number of women studied, number of patients that underwent laparoscopic gynecologic surgery, history of smoking, and past history of nausea and vomiting or motion sickness, all of which are known to affect the incidence of PONV. Furthermore, because of the overrepresentation of women in these studies (88%) undergoing operative procedures that are known to be highly emetogenic (e.g. , laparoscopic surgery), the use of inhalation agents for maintenance of anesthesia (instead of propofol infusion), and the use of opiates during and after surgery, it is important that the results not be extended to other groups of patients.
Dexamethasone reduced the incidence of PDNV, which was in agreement with the findings of Henzi et al. , 32who showed that dexamethasone prevented both “late” nausea and vomiting. However, these authors included inpatients in their analysis and defined late PONV as that occurring from 0 to 24 h postoperatively (not restricted to the postdischarge period).
One unexpected finding in our analysis was that droperidol did not prove to be effective as a prophylactic against PDNV. In a large study conducted on ambulatory surgery patients, Hill et al. 33found that droperidol 1.25 mg was effective and cheap and provided similar patient satisfaction compared with 4 mg ondansetron. The data in this study was confined to the postoperative period and did not include the postdischarge period. It is possible that droperidol might be effective for the prevention of “early” PONV but might not be equally effective in “late” (postdischarge) PONV.
A confounding factor in all these studies is the role of anesthetics in producing PONV. Although most of the studies in this review used propofol for induction of anesthesia, none used propofol as an infusion during anesthesia (in two studies, intermittent propofol injections were used for the maintenance of short-duration anesthesia). It is well known that propofol when used as an infusion can reduce PONV. 34However, inhalational anesthetics are thought to be emetogenic probably only in the early postoperative period. 35We observed no difference in the rates of PDNV between isoflurane and desflurane when controlling for nitrous oxide use, and the efficacy of ondansetron was similar between the two anesthetics, supporting the conclusion that the choice of anesthetic had minimal effect during this period.
The last and perhaps the most important factor, which has been repeatedly stressed in recent studies, is the role of combination therapy in the control of PONV and, probably, PDNV. 36Because many factors induce PONV, it is increasingly believed that drugs acting at different receptors would reduce the risk for PONV. The major issue in drawing any definite conclusions from our findings is that the number-needed-to-treat for ondansetron 4 mg was high (> 12). In contrast, treatment with a combination of drugs produced a low number-needed-to-treat (5.2 and 5.3 in PDN and PDV, respectively). This would imply that combination management of PDNV has the greatest benefit for patients in the prevention of PONV, a fact supported by some other authors. 26,36
Although we were able to identify as many as 1,000 studies addressing the issue of PONV, a majority of these studies had been in inpatients. Many of the authors did not use the term ambulatory surgery ; therefore, it was sometimes difficult to determine if the setting was ambulatory or not. Observation of patients in the hospital or in specific designated facilities outside the home environment for less than 24 h and subsequent discharge creates some problems, because this may not result in the same risk for PDNV as that associated with ambulatory surgery. Data extraction was not easy, because standard terminology is not often used. In some studies, the authors had provided data for the early postoperative period (0–2 h) and subsequently for the 0–24 h period. We deemed these studies unfit for the research question (PDNV) and consequently excluded them from the final analysis. Another problem was that some investigators provided the results as PONV and not separately as postoperative nausea or postoperative vomiting. We agreed to include these studies, but the data were assumed to refer to postoperative nausea and were excluded from the analysis from postoperative vomiting. When only single studies were available for a given drug or dose, we thought it inappropriate to perform a “meta-analysis”; consequently, these studies were eliminated, which further restricted our findings. A final problem was that the total number of patients studied in the postdischarge period was not specified by all authors. It was not always stated if there were dropouts from the initial number of patients recruited. Unless otherwise stated, we assumed that all patients initially randomized were also included in the postdischarge data (intention-to-treat). Our study also suffers from the same bias as reported in many others, in that we analyzed only published data in the English language. It has been reported that negative-outcome studies tend to be published non-English language journals, 37which would mean that the number-needed-to-treat may be even higher than that found in our study.
The postdischarge period must be defined in a standardized fashion, and future studies should specifically focus on and properly examine this period of patient care. There is a need for prospective, randomized studies with a large number of patients, and a defined group of patients undergoing specific procedures. This could prevent “overtreatment” of low-risk patients, provide optimal management of the patients-at-risk, and, in the long run, prove to be cost-effective. Certain drugs that seem to be effective but inadequately studied (e.g. , betamethasone, ondansetron 8 mg, and prochlorperazine) must be better documented, specifically in the postdischarge period. The optimal time of administration of prophylactic antiemetics must be clearly defined, specifically in the context of ambulatory surgery. The emetogenic role of opioids and inhalation anesthetics should be confirmed in good prospective trials. Finally, the use of combination therapy for the prevention of PDNV should be further examined to confirm the results of the present meta-analysis.
In this systematic review of the literature, we found that prophylactic treatment of PDNV with ondansetron 4 mg or combination treatment with two drugs results in a significant decrease in the risk of PDNV when compared to placebo. The numbers-needed-to-treat was, however, 13 for ondansetron, which questions its routine use, especially in low-risk patients, to prevent PDNV. In contrast, the numbers-needed-to-treat for combination treatment was 5, which would favor this for prophylactic management of PDNV, particularly in high-risk patients.
The authors thank the staff at the Welch Medical Library, Johns Hopkins Medical Institutions, Baltimore, Maryland, for their constant support and help in getting the articles together. The authors also thank the Peri-operative Clinical Research Unit in the Department of Anesthesiology and Critical Care Medicine at Johns Hopkins University, Baltimore, Maryland, for their assistance.