Presented at the Society for Academic Emergency Medicine Annual Meeting, San Francisco, CA, May 2006. This project was funded by the makers of ondansetron. We approached GlaxoSmithKline (GSK) with the idea for the study, and they agreed to provide funding. GSK remained apart from all phases of the project: they did not approve the study design nor participate in data collection, analysis, writing the manuscript, or approval of the final manuscript. ClinicalTrials.gov Registration Identifier NCT 00429832.
Objectives: The authors sought to compare ondansetron and promethazine among emergency department (ED) patients with undifferentiated nausea. The hypothesis was that ondansetron was not inferior to promethazine and that rates of adverse effects were similar.
Methods: This was a randomized double-blind noninferiority clinical trial conducted in an urban academic ED. A convenience sample of nonpregnant adults with at least 40 mm of self-reported nausea measured on a 100-mm visual analog scale (VAS) were enrolled. Patients who had already received more than 1 L of intravenous fluid or an antiemetic agent were excluded. Subjects were block-randomized in groups of 10 to either 4 mg of ondansetron or 25 mg of promethazine delivered intravenously. The primary outcome was change in nausea over 30 minutes. The authors used a 15-mm margin of noninferiority. Secondary endpoints included changes in anxiety, sedation, and other adverse effects. Analyses included t-tests, tests for proportions, and 95% confidence intervals (CIs).
Results: A total of 120 subjects completed the study, 60 in each arm. Baseline nausea, anxiety, and sedation scores were similar. Ondansetron and promethazine reduced nausea similarly (ondansetron −34 mm, promethazine −36 mm; difference −2 mm; 95% CI = −13 to 8 mm). The reduction in anxiety was similar (ondansetron −13 mm, promethazine −14 mm; difference −1 mm; 95% CI = −10 to 10 mm). Promethazine was associated with significantly more sedation than ondansetron (ondansetron 5 mm, promethazine 19 mm; difference 14 mm; 95% CI = 5 to 24 mm). There were no cases of akathisia in the ondansetron group and 2 cases in the promethazine group.
Conclusions: Promethazine and ondansetron have similar efficacy in reducing nausea among ED patients. Change in anxiety was similar, but promethazine was associated with greater sedation.
Nausea is a common complaint in the emergency department (ED), often prompting administration of an antiemetic agent (AA).1 Limited research has been performed in the ED setting to determine which AA is most effective with the fewest adverse effects. Previous adult ED trials have evaluated prochlorperazine alone;2,3 prochlorperazine versus promethazine;4 or prochlorperazine, droperidol, metoclopramide, and placebo,5 all with mixed results. Promethazine is the most widely used AA used in the ED; thus, we selected this agent as the standard for comparison.1 In the setting of chemotherapy and anesthesia, serotonin 5 -hydroxytryptamine 3 (5HT3) receptor antagonists such as ondansetron have become the preferred agents.6–9 Emergency physicians (EPs) have subsequently begun using this more expensive class of medication. Two trials in pediatric ED patients found ondansetron to be effective compared to placebo,10,11 and one trial found it superior to intravenous (IV) hydration with or without dexamethasone.12 To our knowledge, ondansetron has not been previously studied in the adult ED population.
We sought to determine whether ondansetron was noninferior to promethazine in reducing self-reported nausea among a population of ED patients with undifferentiated conditions. We hypothesized that ondansetron was not inferior to promethazine and that rates of adverse effects were similar.
We conducted a randomized, double-blind, noninferiority, clinical trial, comparing the efficacy of ondansetron to promethazine for the treatment of undifferentiated nausea and vomiting in adult ED patients. The study design was reviewed and approved by our institutional review board. Written informed consent was obtained from eligible subjects by the research assistant (RA) or primary investigator.
Study Setting and Population
This study was conducted in an urban, university teaching hospital ED, with approximately 75,000 visits annually. A convenience sample of adult patients (18 years of age or older) with a chief or secondary complaint of nausea or vomiting was identified by physician and nursing staff, direct observation, or review of ED charts and computerized patient tracking system between August 2003 and November 2005. Potential subjects were screened for eligibility and enrolled by trained RAs or one of the primary study investigators. Subjects were enrolled only when any of the RAs or study investigators were in the ED for this expressed purpose. RAs worked sporadic shifts, typically between 8 am and 8 pm, 7 days a week.
Patients were excluded 1) if they were less than 18 or more than 65 years of age, 2) if they were unable to provide informed consent, 3) if they rated their nausea less than 40 mm on a 100-mm visual analog scale (VAS), 4) if they had received any AA in the previous 24 hours, 5) if they were pregnant or possibly pregnant, 6) if they reported allergies to either of the study medications, 7) if they had received more than 1 L of IV fluids, or 8) if their primary EP declined to have the patient participate for clinical reasons such as acuity requiring immediate treatment. As part of determining eligibility, patients were asked to rate their nausea on a 100-mm VAS.
After consent was obtained, patients had a peripheral IV line of normal saline placed, if not already present. IV fluids were set at a to-keep-open rate until the 30-minute study period began. The pharmacy provided blinded identical vials containing either 4 mg of ondansetron or 25 mg of promethazine diluted in 10 mL of normal saline. The study medications were block-randomized, with each block of 10 vials containing five doses ondansetron and five doses of promethazine. The medication was administered by a registered nurse over 2 minutes and a 30-minute timer was started; IV fluids were then set to run wide open until the end of the 30-minute study period. The nurses, RAs, primary investigators, and subjects remained blinded to the AA until final data analysis.
Subjects were asked to rate their self-reported nausea, anxiety, and sedation along a 100-mm VAS. The VAS for nausea but not anxiety or sedation has been validated in adults.13 The VAS was anchored at 0 (no nausea) and 100 (most nausea imaginable). Similar anchors were provided by anxiety and sedation or sleepiness. Subjects were asked to make a vertical mark along this scale. Nausea was screened three times: 1) to determine eligibility, 2) just prior to medication administration, and 3) at 30 minutes after medication administration. Anxiety and sedation were screened twice: just prior to medication administration and at 30 minutes after medication administration.
The total amount of IV fluid administered, the number of emetic episodes during the study period, additional antiemetic administration, adverse effects, and length of stay (LOS) were recorded. The subjects were also asked if they would like further treatment (rescue medication) for their nausea and evaluated for evidence of extrapyramidal reactions. In addition, charts were reviewed after discharge for additional evidence of adverse effects or other antiemetic administration and to determine LOS if the RA left before final disposition. Discharged subjects were contacted, when possible, 24 to 48 hours after treatment to obtain follow-up data including further nausea or vomiting, further treatment, visits to a physician, taking of fluids, and adverse effects.
We compared the reduction in nausea associated with ondansetron treatment to the reduction associated with promethazine treatment. We specified the margin of inferiority at 15 mm. This margin was selected because previous research has shown that the minimum clinically significant VAS change in nausea is 15 mm.13
All measurements were made by investigators masked to the treatment allocation. We used an intention-to-treat analysis. The primary outcome measure was reduction in nausea, as measured by a vertical line on a 100 mm VAS. Additional outcome measures included change in sedation and anxiety as measured on a 100-mm VAS (common side effects of antiemetic medications). Paired nausea measurements in millimeters were made at Time 0 (prior to randomization) and at 30 minutes after treatment. The difference in nausea score (nausea at Time 30 minutes, less nausea at Time 0) were compared between the two treatment arms.
For our principal outcome (change in nausea), we calculated 95% confidence intervals (CIs) about the difference in nausea scores between the two treatment arms. To assess for noninferiority, we compared the upper bound of the 95% CI (thus, technically a one-sided 97.5% CI) for the difference in nausea reduction to our margin of noninferiority. We interpreted an upper-bound CI that excluded the margin of noninferiority as evidence of noninferiority.14 Using the CI approach, a sample size of 120 (60 in each arm) provided adequate power (80%) to detect a 15-mm difference.14
To test for statistical significance between the two treatment arms for sedation and anxiety, we used a t-test and presented the mean absolute differences along with 95% CI. Similar analyses were conducted for the other (secondary) endpoints of anxiety and sedation. SAS software (Version 9.13, SAS Institute, Cary, NC) was used for all analysis.
We also examined the use of rescue medication in each of the two treatment arms. We then compared the proportion of subjects requiring rescue or repeat medication for nausea. To test for statistical significance, we used a test of proportions. Where indicated, we used Fisher’s exact test. We used a two-tailed Type I error rate of 5% to determine significance. Masking of the treatment allocation persisted through the data analysis phase of the project. Masking was removed after analysis of results was complete.
We evaluated 511 potential subjects for study inclusion; 391 of these were excluded, primarily for either nausea scores < 40 mm at baseline (42%) or recent antiemetic drug use (18%) (Table 1). We enrolled 120 subjects; 60 subjects were in each treatment arm (Figure 1).
Table 1. Frequency of Exclusionary Criteria
Data are reported as number (%).
VAS = visual analog scale.
*Percentage sum to more than 100% because more than one exclusion criterion was possible.
Total subjects excluded*
Nausea rated less than 40 mm on VAS
Recent antiemetic use
Patient declined participation
Received more than 1 liter intravenous fluid before enrollment
Non-English speaking patient
Patient left without being seen or unable to locate
Treating physician preference
Unable to establish intravenous line
Patient discharged prior to enrollment
Exceeded age eligibility (<18 or >65 years)
Allergy to study medication(s)
Other and unknown reasons
Study subjects were similar with respect to age and IV fluids administered prior to randomization; however, there was a trend toward more women in the promethazine arm (p = 0.08; Table 2). Subjects had similar baseline nausea VAS scores (promethazine 67 mm, ondansetron 69 mm) as well as anxiety and sedation VAS scores (Table 2).
Table 2. Baseline Characteristics, for Total Sample and by Treatment Group
SD = standard deviation.
Characteristic, n (%)
Intravenous fluid prior to randomization (mL)
Baseline scores in mm, mean (SD)
At 30 minutes after drug administration, subjects had received a similar amount of additional IV fluid (promethazine 497 mL, ondansetron 460 mL; (Table 3). One RA deviated from protocol and verbally administered the outcome measures using a numeric rating scale (NRS; 0 to 10) instead of asking the subjects to mark on the VAS in 11 of 60 (18%) promethazine and 8 of 60 (13%) ondansetron subjects. The NRS was mapped onto the VAS for analysis (e.g., “5” translated to 50 mm).
Table 3. Primary and Secondary Absolute and Relative Endpoints, by Treatment Group
95% CI Difference
CI = confidence interval; SD = standard deviation; VAS = visual analog scale.
Change in VAS score at 30 minutes less baseline score
Nausea, mean (SD)
Anxiety, mean (SD)
Sedation, mean (SD)
Intravenous fluid after randomization (mL)
Wanted more medication for nausea, n (%)
Vomited after medication administration, n (%)
Nausea scores at 30 minutes after drug administration decreased among subjects in both arms (promethazine −36 mm, ondansetron −34 mm). There was no statistical difference in the reduction in nausea score between the two treatment arms (difference −2; 95% CI = −13 to 8; (Table 3; Figure 2). As the upper bound of the difference in nausea reduction for ondansetron excluded the margin of noninferiority, we found evidence that ondansetron was noninferior to promethazine.
Anxiety scores at 30 minutes after drug administration also decreased among subjects in both arms (promethazine −14 mm, ondansetron −13 mm). There was no statistical difference in the reduction in anxiety score between the two treatment arms (difference −1 mm; 95% CI = −10 to 10 mm; Table 3; Figure 2).
Sedation scores at 30 minutes after drug administration increased among subjects in both arms (promethazine 19 mm, ondansetron 5 mm). There was a significantly greater increase in sedation, however, among subjects in the promethazine arm compared to subjects in the ondansetron arm (difference 14 mm; 95% CI = 5 to 24 mm; (Table 3; Figure 2).
A similar percentage of subjects wanted additional antiemetic medication at 30 minutes in both arms (promethazine 18%, ondansetron 25%). A similar percentage of subjects in both arms vomited subsequent to medication administration (promethazine 7%, ondansetron 8%; (Table 3). Two subjects complained of symptoms consistent with akathisia; both were in the promethazine group (3%).
Subjects were given a variety of discharge diagnoses (Table 4); the most common was nonspecific abdominal pain or vomiting. Discharge diagnoses were similar by treatment arm. Admission rates were similar: ondansetron 22% and promethazine 23%.
Table 4. Principal Emergency Department Discharge Diagnoses
Principal Discharge Diagnosis
Total (N = 120)
Numbers in parentheses are percentages.
GI = gastrointestinal; NOS = not otherwise specified; PUD = peptic ulcer disease.
Vomiting/abdominal pain NOS
Other medical conditions
Diabetic ketoacidosis/diabetic-related vomiting
No chart available/unknown
About three-fourths (77%) of subjects completed telephone follow-up at 24 hours (Table 5). Subjects in both study groups reported similar satisfaction with their study medication (promethazine 79%, ondansetron 91%; difference −12%; 95% CI = −26% to 3%) and were able to eat and drink (promethazine 73%, ondansetron 89%; difference −16%; 95% CI = −31% to 1%). Many subjects in both study groups reported continued nausea or vomiting with a nonsignificant trend toward less among the ondansetron group (promethazine 46%, ondansetron 27%; difference 19%; 95% CI = −1% to 36%).
Table 5. Follow-up Characteristics, for Total Sample and by Treatment Group
Total (N = 120)
Promethazine (n = 60)
Ondansetron (n = 60)
Numbers in parentheses are percentages.
Completed follow-up interview
0.517 (proportion test)
Was satisfied with their study medication
0.202 (F test)
Able to tolerate food/drink
0.101 (F test)
Nausea/vomiting since discharge
0.104 (F test)
Saw a physician in follow-up
0.338 (F test)
0.466 (F test)
Continued to take antiemetic medication
0.128 (F test)
Experienced muscle spasms
Experienced medication side effects
Our results indicate that, at the dosages used, promethazine and ondansetron have similar efficacy in reducing nausea among ED patients, but that promethazine is associated with greater sedation and occasional akathisia compared to ondansetron. Sedation may be desirable in some patients and dangerous in others, especially elders or those receiving concomitant narcotics or other sedating medications. Akathisia, on the other hand, is never desirable.
If recognized early and treated appropriately, the 3.3% incidence of akathisia with promethazine may be acceptable to some practitioners unless one is working in a high-risk environment such as air medical transport.15 Our incidence of akathisia with promethazine appears much lower than the 16% to 44% noted in ED patients receiving prochlorperazine16,17 or the up to 72% incidence with droperidol if every patient who reported anxiety or restlessness was considered to have some degree of akathisia.5
In our previous trial, IV droperidol (1.25 mg), but not prochlorperazine (10 mg) or metoclopramide (10 mg), was superior to IV fluids for the control of undifferentiated nausea among ED patients.5 We had planned to compare ondansetron to droperidol until the FDA placed a black-box warning on droperidol.18 We then elected to compare ondansetron to promethazine, as it is the most commonly used ED antiemetic1 and because our previous results invalidated two of the other common agents. Ernst et al.,4 however, have previously found prochlorperazine to be superior to promethazine.
We attempted to select the most common dosages in clinical practice, but clearly, different dosages may have different effects. The selected dose of promethazine is higher than some clinicians use, which may have contributed to the observed higher incidence of sedation, but lower dosages may not have the same degree of antiemetic effect. Ondansetron is commonly used at higher dosages (up to 32 mg) for the treatment of chemotherapy-induced nausea19 and may have been more effective than promethazine at this higher dose. Higher dosages may cause more adverse effects and add additional cost, although the cost difference between the two study medications is shrinking with the availability of generic ondansetron; the current Red Book price for generic IV ondansetron 4 mg is $3.17, while that for IV promethazine 25 mg is $1.80.20
The chosen 30-minute study period may have limited the maximal effect of some medications, but was determined to be the longest period that we and our institutional review board were comfortable withholding a rescue medication. Longer-term effects were evaluated as a secondary outcome; however, 24-hour follow-up was limited to 77% of subjects.
We previously determined that all ED patients with nausea improve over time or with the administration of IV fluids.5 There was no difference in the amount of fluid received by the two groups. We did not include a placebo group, as our previous work found placebo to be less effective than some AAs and therefore could not be justified. We cannot exclude the possibility that the administered medications had no additional effect.
As our population was undifferentiated (i.e., no specific etiologies for nausea were selected nor excluded), there were potential confounding variables, including other medications administered during the study, different etiologies for the nausea, and different relative effectiveness of medications for different disease states. As in our previous work, it was our intention to study AA usage in a way that most closely mimics real clinical practice. In addition, by the time an etiology for nausea and vomiting is determined, an AA has usually been administered. We tried to eliminate any medication with antiemetic properties; however, some medications, such as corticosteroids, may have been inadvertently included. Medications that might stimulate nausea and vomiting were not excluded. Therefore, we believe that we were biased in favor of drug underperformance.
Our assessments for akathisia consisted of self-reported symptoms during ED treatment and telephone follow-up, as well as a structured chart review. We did not employ any objective observational scales to detect akathisia. Previous research has shown remarkably high incidences of akathisia when specific scales are used.21 It is possible that the true incidence of akathisia was underreported. We believe that we captured the most clinically significant cases.
Our study was neither designed nor powered to detect long-term differences between medications. Thus, our ability to draw clinical conclusions from follow-up data is very limited.
Enrollment was intentionally skewed toward the moderately symptomatic patient, as less symptomatic patients were systematically excluded. Clinically, however, less symptomatic patients are often not treated with an AA. The sickest patients, those actively vomiting, or those at increased risk of complications from sedation, may not have been enrolled because of concerns by the nursing and physician staff and study investigators, for delaying care and obtaining informed consent in these patients. These results are not necessarily applicable to all 5HT3 antagonists, as different members of the class have been shown to have different effects.7
One of our RAs verbally administered the outcome measures using a NRS (0 to 10) instead of asking subjects to mark on the VAS. Results from subjects using NRS compared to the VAS were similar, but we cannot measure the potential biases that may have been introduced by this variation. Previous work has demonstrated a high correlation between the NRS and VAS in pain assessment.21 We have no reason to suspect that this method of assessment differed between the two treatment groups in our study.
Promethazine 25 mg and ondansetron 4 mg administered intravenously to ED patients complaining of undifferentiated nausea have similar efficacy at 30 minutes. Ondansetron may be associated with less nausea and vomiting at 24 hours. Change in anxiety was similar but promethazine was associated with greater sedation. The choice of medication will likely be made primarily based on side effect profile. Promethazine may be preferred when sedation is clinically desirable, while ondansetron appears preferable when sedation is undesirable and/or akathisia cannot be tolerated. It is unclear if these results can be generalized to other 5HT3 antagonists.