Although prophylactic administration of antiemetics reduces the incidence of postoperative nausea, vomiting, or both (PONV), there is little evidence to suggest this improves patient outcomes. The authors hypothesized that early symptomatic treatment of PONV will result in outcomes, including time to discharge, unanticipated admission, patient satisfaction, and time to return to normal daily activities, that are similar to those achieved with routine prophylaxis.
Men and women (n = 575) scheduled for outpatient surgery during general anesthesia were randomized to receive either 4 mg intravenous ondansetron or placebo before operation and either 1 mg intravenous ondansetron or placebo if postoperative symptomatic treatment of PONV was necessary. Patients were stratified into subgroups by risk factors for PONV.
No differences occurred in the time to discharge, rate of unanticipated admission, or time to return to normal activity between the prophylaxis and treatment groups. The reported level of satisfaction with control of PONV was 93% in the treatment arm and 97% in the prophylaxis arm, which fall within the limits defined a priori as clinically equivalent. Female patients with a history of motion sickness or PONV who were undergoing highly emetogenic procedures had a higher reported level of satisfaction with prophylaxis than with treatment (100% vs. 90%, P = 0.043); however, the level of satisfaction with the overall outpatient surgical experience was not different.
Although PONV is unpleasant, the data indicate little difference in outcomes when routine prophylactic medications are administered versus simply treating PONV should symptoms occur.
This article is accompanied by an Editorial View. Please see: Fisher DM: Surrogate outcomes: Meaningful not! Anesthesiology 1999; 90:355-6.
THE observed rate of postoperative nausea and vomiting (PONV) is influenced by various factors, including gender, age, anesthetic technique, prophylactic use of antiemetics, use of opioids, and type of surgery. For instance, the reported incidence of postoperative emesis occurring during the 24 h after outpatient surgery, when no prophylaxis is administered, may occur in more than half of women undergoing gynecologic surgery. Approximately one third of men undergoing various procedures may experience emesis. In these same groups, the administration of serotonin (5-HT (3)) antagonists prophylactically (i.e., ondansetron) decreased the incidences of emesis to 23% and 20%. Reductions in the incidence of postoperative nausea were also found when ondansetron was administered prophylactically. [1,2]There have been similar findings for other 5-HT (3) antagonists. [3-5]The 5-HT3antagonists have been established as an effective treatment for postoperative emesis when compared with placebo. [6,7]However, data from large multicenter trials have shown that no medications (including 5-HT3antagonists) are universally effective in preventing or treating PONV. [1,6]
Nausea and vomiting rarely harm patients seriously. The rate of unexpected admission occurring because of intractable PONV is also low. Nevertheless, the symptoms are unpleasant. However, there is no objective evidence to indicate that the routine administration of prophylactic antiemetic medications will result in either better objective or better subjective patient outcomes when compared with prompt treatment of symptoms, if they occur, in the postanesthesia care unit (PACU). Our hypothesis is that timely symptomatic treatment of PONV will result in outcomes (patient satisfaction, time to discharge, rate of unanticipated admission, and time to return to normal daily activity) equivalent to those seen with routine administration of prophylactic anti-emetics.
Five hundred seventy-five men and women, aged 18 to 65 yr, who were classified as American Society of Anesthesiologists physical status 1, 2, or 3 and undergoing outpatient surgery during general anesthesia, were enrolled in this study, which was performed in the Outpatient Surgery Center of Wake Forest University Baptist Medical Center. The study was double blinded, randomized, and placebo controlled. The protocol was approved by the Clinical Research Practices Committee (i.e., our institutional review board). After we obtained informed consent, we stratified patients to eight sub-groups based on the presence or absence of risk factors (gender, history of PONY or motion sickness, and emetogenic or nonemetogenic surgical procedure) for development of PONV. Patients scheduled either as outpatients or for short stay (23 h or less) were eligible for enrollment. Basic demographic information was collected from each patient after enrollment. All patients underwent a baseline nausea assessment 30 min before anesthesia was induced. Nausea was scored using an 11-point linear numeric scale from 0 to 10, with 0 representing no nausea and 10 representing nausea “as bad as it can possibly be.” Patients were randomized before operation, by random-number table, to receive either 4 mg intravenous ondansetron or an equal volume of saline as a placebo at the induction of anesthesia. The anesthetic regimen was not specified by the protocol to more closely approximate the typical standard of care normally used for outpatients in our institution. Patients were again randomized, by random-number table, to receive either 1 mg intravenous ondansetron or saline as a placebo if symptomatic treatment of PONV was necessary after admission to the PACU. A block randomization procedure was used to ensure that similar numbers of patients received active and placebo medications within each subgroup for prophylaxis and treatment (should treatment become necessary while patients were in the PACU). If subsequent treatment (i.e., rescue) for persistent nausea or vomiting was necessary, patients received 1 mg intravenous ondansetron (with the investigator not blinded). Additional treatments, if needed for continued symptoms of PONV, were at the discretion of the attending anesthesiologist.
Anesthetic medications and duration of anesthesia were obtained from the patient record. At admission to the PACU, the patients were monitored by study personnel for effectiveness and outcomes. Nausea assessments were performed at 30-min intervals while patients were in the PACU beginning at arrival and continuing until discharge from the PACU to home or to the Day Hospital (our institution's “short-stay” unit). All patients were admitted to stage I of the PACU for initial evaluation and management. Patients were discharged to a step-down unit called the stage II PACU when their Aldrete Score was 10/10 (or the highest number possible for that patient) and any initial pain, PONV, or both were controlled. Nausea scores were also recorded whenever treatment for PONV was given while patients were in the PACU. Each episode of emesis, defined as expulsion of gastric contents or an unsuccessful attempt to expel gastric contents (e.g., retching), was recorded during the patient's recovery in the PACU. Patients received initial treatment immediately in response to the first emetic episode or patient request because of nausea. Patients received an additional treatment (i.e., rescue) medication consisting of 1 mg intravenous ondansetron (non-blinded) if nausea or vomiting persisted more than 15 min after the initial treatment. Subsequent treatment medications were administered if symptoms persisted for an additional 15 min.
Home-readiness in the stage II PACU was evaluated using the postanesthetic discharge scoring system, with a score of 9 (or the highest possible for the patient) necessary for discharge. No minimum time to discharge was necessary in the stage I or stage II PACU. Those patients scheduled for a short stay were discharged to the Day Hospital when their stage I PACU Aldrete score was 10/10 or the highest possible. Time to discharge to home or to the Day Hospital was recorded for all patients.
Patients were given questionnaires at the time of discharge (see Table 10, Table 11). Subsequently, they were contacted by telephone to obtain the data recorded at 24 h and 5 days after discharge from the PACU. Patients were asked to rate how satisfied they were with the management of their PONV first as a categoric “yes or no.” Then they were asked to rate their level of satisfaction on an 11-point (0-10) linear numeric scale. At 5 days, patients rated their level of satisfaction with their overall outpatient surgical experience using the same 11-point linear numeric scale. Patients also answered a series of questions designed to identify the time necessary to return to normal daily activity after surgery (see Table 10, Table 11).
Two separate power analyses were performed after the first 200 patients were enrolled. Equivalency of patient satisfaction with management of PONV was defined as a difference in satisfaction evaluated by a “yes or no” response of 10% or less. The equivalent time to discharge from the PACU was defined as a discharge time difference of 15 min or less. It was determined that a total enrollment of 575 patients, representing the larger of the two sample sizes calculated from the power analyses, would provide an 80% power ([small alpha, Greek]= 0.05, two-sided) to detect equivalent satisfaction. Statistical analyses were performed using SAS, version 6.11 (SAS Institute, Cary, NC). Analysis of variance was used to compare continuous parameters (e.g., age, weight, intravenous fluids). The duration of anesthesia and times to PACU discharge were not normally distributed. Log transformation normalized the data for analysis of variance; therefore, geometric means were reported rather than arithmetic means. Intergroup comparisons were made within the analysis of variance using pair-wise contrasts. Frequency data (e.g., the number of patients with emetic episodes) were compared using contingency tables with either chi-squared analysis or Fisher's exact test as appropriate. Intragroup pair-wise comparisons were made using 2-by-2 contingency tables. No corrections for the multiple comparisons were made for intragroup comparisons. A log-rank survival analysis was performed to compare the time needed to return to normal activity after surgery using data obtained from the 24-h and 5-day surveys.
We evaluated 575 patients in this study. The frequencies of treatment needed for nausea, vomiting, or both after surgery were determined for patients receiving ondansetron or placebo prophylaxis (Table 1). The incidence of symptomatic PONV in those patients receiving placebo prophylaxis ranged from 16% in subgroup D to 57% in subgroup E, and in those patients receiving ondansetron prophylaxis it ranged from 0% in subgroup A to 45% in subgroup F. Table 2summarizes the patient characteristics by initial randomization (ondansetron and placebo) and anesthetic regimens. There were no significant differences in demographics among the subgroups. The prevalence of histories of motion sickness, PONV, or both were not different. Only four patients did not receive propofol as an induction agent, and most had anesthesia maintained with a combination of nitrous oxide, a potent inhalation agent, and fentanyl. Approximately 90% of patients received a neuromuscular blocking agent. Of those receiving a neuromuscular blocking agent, pharmacologic reversal was administered in approximately 90% of the patients. All patients who did not receive pharmacologic reversal of neuromuscular blockade had received mivacurium.
There was no difference in the baseline nausea scores. Nausea scores at PACU entry also showed no difference (Table 3). Overall, more patients who received prophylaxis required no treatment for symptomatic PONV (71.6%) than did those who received placebo (61.7%)(P = 0.012, Table 3). Of the patients who required symptomatic treatment in the PACU, there was no difference (P = 0.633) in the need for treatment for nausea; however, the need for treatment of vomiting was approximately 2.5 times greater in those receiving placebo compared with those who received ondansetron (P = 0.001, Table 3). There was no difference in nausea scores at the time of discharge.
(Table 4) shows efficacy data. The treatment phase of the protocol began when the patients were admitted to the PACU. Patients in both the placebo and ondansetron prophylaxis groups had three possible pathways, as shown in Table 4. Asymptomatic patients received no further treatment. Symptomatic patients received either 1 mg ondansetron or placebo, followed by unblinded rescue if necessary, as noted in the methods section. Similarly, Table 5shows the data for the outcome measures.
(Table 6) compares the outcomes for prophylaxis with those for treatment. Patients who received ondansetron prophylaxis who subsequently required no treatment or received ondansetron as a treatment for symptomatic treatment of PONV were compared with patients who received placebo prophylaxis and who subsequently required no treatment or received ondansetron as a treatment for symptomatic PONV. There was no difference in discharge times to home or to the Day Hospital. Confidence limits of the difference in times to discharge between active prophylaxis and placebo prophylaxis showed that improvement was no more than 12 min for outpatients and 5 min for patients admitted to the Day Hospital, with 97.5% certainty. There was also no difference in the percentage of patients admitted to the Day Hospital. In addition, no patients were admitted to the Day Hospital because of intractable PONV.
Patients who received prophylaxis had a higher satisfaction score (97%) with management of PONV than did those who received placebo (93%)(Table 6); however, this difference was within the limit (10%) set a priori to represent no clinically significant difference. The confidence limits of the differences in these satisfaction scores show that the improvement in satisfaction with control of PONV in the active prophylaxis (ondansetron) group can be no more than 8.3%, with 97.5% certainty. There was also no difference in 5-day satisfaction with the entire outpatient surgical experience between patients receiving prophylaxis compared with those who received placebo. The confidence limits of the difference in patients reporting complete satisfaction with their outpatient surgical experience showed the difference to be no more than 9.8%, with 97.5% certainty. There was no difference between active prophylaxis and placebo in the times necessary to return to normal activity as assessed by the 24-h and 5-day questionnaire either by individual indices or by survival analysis (Figure 1) of all indices.
The subgroup data are shown in the appendix (Table 7, Table 8, and Table 9). Analysis is limited to the comparison of active prophylaxis with active treatment, as described for the entire patient population. There was a significant difference in the duration of anesthesia between prophylaxis and treatment for subgroup D. Only subgroups A and E had a significant differences in the incidence of treatment necessary for symptomatic PONV. Only subgroup B showed a higher nausea score at discharge in those patients who received placebo initially compared with prophylaxis. Subgroup D showed a longer time to discharge to home in those patients who received placebo (84 min; 95% confidence interval, 76 to 91) compared with those who received prophylaxis (67 min; 95% confidence interval, 60 to 75). There was no difference in the time to discharge to the Day Hospital for any of the subgroups. Only subgroup E showed a higher satisfaction with control of PONV in those receiving prophylaxis (100%) compared with those who received placebo (90%)(P = 0.043); however, at 5 days there was no difference in overall satisfaction with the outpatient surgical experience between patients who received prophylaxis and those who received placebo, except for those in subgroup H, who showed a higher level of reported satisfaction (P = 0.040) with placebo. When individual subgroups were analyzed by survival analysis for return to normal activity, there was no beneficial effect from prophylaxis compared with treatment; however, individual indices showed an earlier return to normal activity in certain groups (i.e., subgroups E, G, and H) in those patients who received prophylaxis.
Antiemetic agents administered prophylactically can decrease the incidence of PONV in various patient populations. [1,3,5]In addition, antiemetics also have been shown to be effective in treating PONV that occurs in the PACU after surgery. [6,7]What has been unclear, however, is whether routine prophylaxis or even prophylaxis in certain presumably high-risk patient populations actually has a beneficial effect on objective measures of outcome when compared with timely treatment of symptoms. Clinical trials that evaluated the efficacy of antiemetic therapy typically have focused on either prophylaxis or treatment. To date no studies have tried to show that prophylaxis is superior to symptomatic treatment when objective measures of outcome are compared. Undeniably, prophylaxis with an appropriate antiemetic agent (for example, a 5-HT3antagonist such as ondansetron) will have superior effectiveness when compared with placebo if only the frequency of emesis and severity of nausea are compared. Nonetheless, it has remained unclear whether this translates into improvement in more objective measures of outcome such as decreased time to discharge, decreased rate of unanticipated admission, improved patient satisfaction, or return to normal daily activities. 
This study has tried to answer this question by evaluating the effect of routine antiemetic prophylaxis versus prompt symptomatic treatment in patients undergoing outpatient surgery. Ondansetron was used as a prototypical antiemetic. It was chosen as the study drug because of its very low incidence of side effects. [1,6]Because ondansetron is nonsedating, it could be administered as the prophylactic or as the treatment medication without introducing the possibility of delayed time to discharge caused by excessive patient sedation. A prophylactic dose of 4 mg ondansetron was chosen based on current product information and Food and Drug Administration approval. Although higher doses may be more effective, standard practice would seem to favor a 4-mg dose for various reasons, including significant differences in cost with little demonstrable improvement in effectiveness. The prophylactic study medication (ondansetron or placebo) was administered at the beginning of the case, before induction of anesthesia. This was standard practice when this study was initiated. More recently published information suggests that ondansetron may be more effective when administered at the end of surgery before the patient emerges from anesthesia ; however, it did not seen appropriate to change the study design midway through the study, nor was it practical to repeat the entire study. Although it is possible that the time of administration of a prophylactic antiemetic may affect outcome and effectiveness, this seems unlikely. We chose a dose of 1 mg ondansetron for treatment. There is no conclusive evidence that higher doses are more effective in treating established PONV; in fact, when data from two large parallel multicenter trials are combined, neither 4 mg nor 8 mg ondansetron were more effective than 1 mg when used to treat PONV.
Patients were also randomized to receive ondansetron or placebo if symptomatic treatment of PONV was necessary after they were admitted to PACU. A placebo treatment group was necessary to ensure that observer bias would not affect the results. Had symptomatic treatment been provided without benefit of the placebo control, tacit assumption of benefit could not have been avoided by study personnel. Of course, this resulted in symptomatic patients receiving a placebo treatment; however, this is no different than what would have occurred had this been a treatment-only trial. Provisions for rescue with active medications were made when symptoms of PONV persisted, as noted in the methods section.
The anesthetic regimen was not controlled during this study so conditions that exist in the “real world” could be replicated more accurately. Typically, antiemetic efficacy studies, particularly large multicenter trials sponsored by pharmaceutical companies, have specified that propofol not be used as part of the anesthetic to avoid the potential antiemetic effects of that drug. However, propofol is an extremely common drug used to induce anesthesia, particularly in outpatient anesthesia. As shown in Table 3, propofol was used as part of the anesthetic technique in nearly all patients. A typical anesthetic consisted of a propofol induction with small doses of midazolam and fentanyl followed by a nondepolarizing neuromuscular blocking agent to facilitate intubation. Anesthesia was maintained with a potent inhalation anesthetic and nitrous oxide. The neuromuscular blocking agent was reversed in most cases. This typical anesthetic resulted in a wide range of symptomatic PONV that necessitated treatment in the PACU (Table 1). Although, overall, the difference in the incidence of PONV that necessitated treatment between patients receiving ondansetron prophylaxis and those who received placebo appears small (Table 3), the incidence of PONV by risk group (16% to 57%) is consistent with the reported ranges of PONV for comparable patient populations. [1,2]
This study evaluated the effect of routine prophylaxis of PONV on five objective measures of outcome: time to discharge, patient satisfaction with treatment of PONV, patient satisfaction with the overall outpatient surgery experience, unanticipated admission, and the time necessary to return to normal daily activity. The outcome measures were analyzed for the entire study population and for each of the eight subgroups defined before operation by risk factors for PONV. There is no evidence to support a hypothesis that routine prophylaxis for PONV provides a clear-cut improvement in any of these measures of outcome. Although patient satisfaction with management of PONV was higher in subgroup E when prophylaxis was administered (100% vs. 90%), no other outcome measure was improved consistently, even in this highest-risk subgroup. Although this difference would not have been significant had there been correction for multiple comparisons, this difference exceeds the limits defined a priori as clinically equivalent. There was also a “statistically significant” difference in satisfaction of PONV management with prophylaxis for the overall study population (97% vs. 93%); however, this difference falls within the limits defined a priori as clinically equivalent. Although this represents a significant difference, it is less apparent whether it represents a clinically important difference. To assess the clinical significance of a statistically significant finding, some method or measure must be identified. One such proposed method is the “number needed to be treated,”which is defined as the reciprocal of the absolute risk reduction. The number needed to treat thus identifies the number of patients who must be treated to prevent one adverse event. For the increased improvement in patient satisfaction with control of PONV achieved with routine prophylaxis, the number needed to treat is 25. This indicates that it would be necessary to routinely give prophylaxis to 25 patients for one patient to have a higher level of satisfaction with prophylaxis than treatment. The wholesale acquisition cost of ondansetron is approximately $16 for 4 mg. Therefore, the direct cost associated with the improvement in satisfaction of 1 of 25 patients is approximately $400.
However, cost as a measure of outcome was not specifically addressed by this study. Although several studies have tried to use retrospective data to identify “cost savings” as a result of routine prophylaxis in certain patient populations, it is often difficult to quantify the actual cost savings that can be realized. Trying to attribute cost savings to outcome measures, such as decreased time to discharge, is difficult, if not impossible. In fact, even if PONV was eliminated completely, it is unlikely that a true cost savings would be realized. 
No one is likely to dispute patient dislike of PONV. However, because of the amount of reduction in PONV that can be achieved with available antiemetic regimens, there is little indication that a clinically and statistically significant improvement in objective measures of outcome results from routine prophylaxis of PONV, even in high-risk groups. Although elimination of PONV is still a worthwhile goal, we do not know whether complete prevention of PONV would result in a difference in outcome measures such as those studied here. Unless complete elimination of PONV becomes a clinical reality, routine prophylaxis for PONV confers little, if any, clinical benefit in objective measures of patient outcomes after outpatient surgery.
The authors thank Wilson Somerville, Ph.D., Medical Editor, and Addie Larimore for their assistance in preparing this manuscript, and Dennis Fisher, M.D., for his many helpful suggestions while this manuscript was being written.