Two interventions for every adult, and rescue changes class

Why it matters

Postoperative nausea and vomiting is the PACU problem residents believe they understand best, and the one they most often reason about wrongly. Two wrong ideas persist. The first is additive: two antiemetics are twice as good as one, and a patient who vomits despite prophylaxis needs more of the same. The trial that tested prophylaxis factorially says the effects are independent and multiply apfel-2004. The second is that prophylaxis is a decision to be made patient by patient from a risk score. The current guideline has moved: every adult, at any risk, receives prophylaxis with two interventions, because risk factors go undetected, risk-stratified algorithms are unevenly followed, individual risk factors are read ambiguously, and modern antiemetics are safe at perioperative doses; and rescue must change class gan-2026.

The consequence of getting this wrong runs in both directions. A patient is given nothing because a score said low risk, and a patient who vomits in the PACU is given a second dose of the drug that already failed.

Mechanism

Each antiemetic class blocks a different input to the vomiting reflex: serotonin at 5-HT3 receptors, dopamine at D2 receptors, corticosteroid effects that are still incompletely characterized, and the removal of emetogenic anesthetic agents themselves. Because the inputs are distinct, blocking one does not change the fraction of remaining risk that another can remove. That is what independence means pharmacologically, and it is why the arithmetic is multiplicative on the relative-risk scale: each intervention removes its own fraction of whatever risk is left. It is also why a rescue drug must act on an input the prophylaxis did not touch. Repeating a class that has already failed to control the reflex adds little except its side effects.

Evidence

Apfel 1999 identified four predictors, female sex, history of motion sickness or PONV, non-smoking status and postoperative opioids, and cross-validated the score between two centers apfel-1999. With 0, 1, 2, 3 or 4 factors the abstract gives risks of 10%, 21%, 39%, 61% and 79%; the results text gives 78% at four factors, so the paper is internally inconsistent at that step, and the discriminating power was an area under the curve of about 0.75 apfel-1999.

The IMPACT trial randomized 4123 of 5199 enrolled patients across 64 combinations of six interventions apfel-2004. Ondansetron, dexamethasone and droperidol each reduced relative risk by about 26%, propofol by 19%, and avoiding nitrous oxide by 12% apfel-2004. The authors’ own words on combining them: all the interventions acted independently of one another and independently of the patients’ baseline risk, so the relative risks of the combined interventions could be estimated by multiplying the relative risks of each apfel-2004. Absolute risk reduction, they add, is a critical function of baseline risk, and prophylaxis is rarely warranted in low-risk patients apfel-2004.

The Fifth Consensus Guidelines, in print in 2026, supersede the Fourth gan-2026 gan-2020. Guideline 3, verbatim heading: administer PONV prophylaxis using 2 interventions in adults at any risk for PONV gan-2026. The reasoning is set out under Guideline 1, not under Guideline 3, and it names four drivers: undetected risk factors, non-adherence to risk-stratified PONV algorithms, ambiguity in interpreting individual risk factors, and increasing evidence that antiemetics are safe at perioperative doses, supported by the spread of multimodal prophylaxis within enhanced recovery pathways gan-2026. The same passage closes with a qualifier this page had been missing, verbatim: practice shifts towards general, multimodal prophylaxis does not discredit the validity of PONV prediction scores, nor an appropriately implemented risk-adapted PONV protocol gan-2026. The guideline retains the Apfel score for estimating risk, rounding the ladder to approximately 10%, 20%, 40%, 60% and 80% and banding it as 0 factors low, 1 to 2 medium, and 3 or more high, but the score no longer gates whether prophylaxis is given gan-2026.

Guideline 4 is headed: provide antiemetic treatment to patients with PONV who did not receive prophylaxis or when prophylaxis failed gan-2026. The class-change rule this page rests on is body text under that heading rather than the heading itself, verbatim: patients should receive rescue treatment from a different pharmacological class to the prophylactic agents administered gan-2026. If more than 6 hours has elapsed, a second dose of a 5-HT3 antagonist or a butyrophenone may be considered if no alternatives are available; long-acting agents, aprepitant, fosaprepitant and palonosetron, should not be re-dosed in the PACU; and transdermal scopolamine is not an effective rescue agent because of its slow onset gan-2026. For rescue, ondansetron 4 to 8 mg and droperidol 1 to 1.25 mg are effective in antiemetic-naive patients, graded A3; metoclopramide 10 mg has limited efficacy, A2; propofol 20 to 40 mg is effective, A3 gan-2026. For a patient in whom ondansetron prophylaxis has already failed, another 5-HT3 antagonist adds nothing over redosing ondansetron or placebo, A3 and low certainty, while promethazine 6.25 mg, B1 and low certainty, and amisulpride 10 mg, A3, were effective gan-2026.

The guideline’s prophylactic doses, read from its Table 1 and agent sections, are ondansetron 4 mg intravenously by slow injection or 16 mg orally disintegrating 1 hour before induction; dexamethasone 4 to 8 mg intravenously at induction, A1; droperidol 0.625 mg at the end of surgery, A1; haloperidol 0.5 to 2 mg intravenously at induction or at the end of surgery, A1; amisulpride 5 mg intravenously, A2; palonosetron 0.075 mg intravenously, A1; aprepitant 40 to 80 mg by mouth; and a transdermal scopolamine patch the prior evening or 2 hours before surgery, A1 gan-2026. Metoclopramide 10 mg is not recommended unless other dopamine antagonists are unavailable gan-2026. Note the asymmetry that catches residents out: droperidol’s prophylactic dose is 0.625 mg, and the 1 to 1.25 mg figure belongs to rescue gan-2026. On dexamethasone the guideline adds that 8 mg seemed to have the optimal antiemetic and analgesic efficacy, that a Cochrane review of 38 studies and a subsequent large randomized trial found a single dose up to 8 mg did not increase postoperative infection including in diabetes, A1, and that the transient rise in glucose in diabetes is small, about 2 mmol/L or 36 mg/dL, and was not associated with impaired wound healing gan-2026.

For a patient carrying three or more risk factors the guideline does not recommend a third drug. Verbatim: in patients with 3 or more PONV risk factors, there is a paucity of clinical data on the use of 3 or more prophylactic interventions, we therefore recommend maximizing the use of risk mitigation strategies gan-2026. It adds that published evidence on combination therapy with 3 or more agents remains limited, and offers acupoint stimulation, graded A1, as the addition for patients who require 3 or more prophylactic interventions gan-2026.

Two changes from the Fourth guideline are worth carrying. Intravenous aprepitant is added: an injectable lipid emulsion in a 32 mg dose given as a 30-second injection, with 97% receptor occupancy within 5 minutes, and it is the only intravenous formulation with an FDA indication for PONV prophylaxis gan-2026. Casopitant is removed, for lack of widespread global availability gan-2026. The Fifth also carries a sentence that belongs to Day 17 as much as to this day: recent evidence suggests that transdermal scopolamine may increase the risk of postoperative urinary retention, especially following urogynecologic procedures gan-2026.

The Fifth guideline also grades as A1 that low-dose gabapentinoids, below 300 mg of pregabalin or 900 mg of gabapentin per day, reduce PONV risk despite no significant analgesic effect, and as A1 that opioid-free anesthesia reduces PONV compared with opioid-based anesthesia gan-2026.

What this does not show

The multiplicative rule is a statement about relative risk. It does not mean every patient benefits equally: the same interventions remove a large absolute risk from a patient at 80% and a small one from a patient at 10%, which is why the trial’s own conclusion discouraged prophylaxis in low-risk patients apfel-2004. The Fifth guideline knows this and recommends two interventions for every adult anyway, on the grounds set out above gan-2026. That is a change of policy, not of arithmetic; the absolute benefit in a low-risk patient is still small, and a page that quotes the relative risk reductions without the baseline is still quoting half the result. The guideline itself refuses the stronger reading, saying in the same passage that the shift does not discredit the validity of PONV prediction scores nor an appropriately implemented risk-adapted protocol gan-2026.

What the guideline says about patients with three or more risk factors, and the doses it recommends, were read from the full text on 2026-09-06 and now appear in the evidence slot above. What is still not on this page from the Fifth guideline is its pediatric section, its section on discharge and post-discharge nausea and vomiting, and the full evidence tables behind each grade.

The Apfel score is a population instrument with an area under the curve of about 0.75 apfel-1999. It sorts patients into bands; it does not predict an individual. The two versions of its ladder, the 1999 values and the guideline’s rounded ones, disagree by one or two points at every step, and this page uses the guideline’s rounded values because they are current, they are what the guideline algorithm uses, and rounded values do not invite spurious precision gan-2026.

The IMPACT per-intervention incidence table broken down by number of antiemetics has not been extracted from the held paper; only the summary relative risk reductions appear here apfel-2004. The erratum to the Fourth guideline has not been read, so no dose or table from the 2020 document is quoted on this page gan-2020-erratum.

The drug-specific points on the day’s content list now have sources. Dexamethasone timing was tested directly: 10 mg intravenously before induction, at the end of anesthesia, or placebo, in 120 women having abdominal hysterectomy, gave PONV in the first 2 h of 15%, 45% and 53%, rescue antiemetic use of 8%, 30% and 35%, and PONV at 2 to 24 h of 25%, 28% and 55%; giving it before induction was the more effective timing wang-2000. Droperidol carries a boxed warning: cases of QT prolongation and torsade de pointes have been reported at or below recommended doses; it is reserved for patients who fail other adequate treatments, a 12-lead ECG is taken before administration, and it is withheld if the QTc exceeds 440 msec in men or 450 msec in women; the maximum recommended initial dose is 2.5 mg intramuscularly or by slow intravenous injection, with additional 1.25 mg doses to effect fda-droperidol-label-2009. Ondansetron prolongs the QT interval in a dose-dependent manner, with postmarketing torsade, and is avoided in congenital long QT syndrome; its label also states that a second intravenous dose of 4 mg postoperatively does not provide additional control of nausea and vomiting, which is the class-change rule seen from the label’s side fda-ondansetron-inj-label-2025. Its effect on the vestibular component is not in the label. What a search found is narrower than the claim usually made: in 63 highly susceptible subjects rotated at 20 rpm in darkness, ondansetron 24 mg differed from placebo in no measure, neither head movements tolerated nor time rotating nor symptom scores, and the authors concluded that neither ondansetron nor dimenhydrinate prevented motion sickness muth-2007. That shows ondansetron does not work for motion sickness. It is not a comparison showing the drug is weaker against vestibular input than against other emetic inputs, which is the stronger claim residents are usually taught, and this page does not make it. Transdermal scopolamine is applied the evening before surgery and removed 24 h after it; plasma concentrations are detected within 4 h with the peak, on average, at 24 h, which is the onset lag that makes it useless as rescue; and elderly patients had an increased risk of neurological and psychiatric adverse reactions such as hallucinations, confusion, dizziness and drug withdrawal syndrome fda-scopolamine-label-2026.

A correction to this page’s own provenance. Until 2026-09-06 the rationale for Guideline 3 was given here as the sentence that clinicians’ adherence to risk-based adjustments in PONV prophylaxis is suboptimal. That sentence is in the guideline, but it sits in Guideline 5, the pediatric section, and describes pediatric risk models; it is not the stated rationale for the adult recommendation. The adult reasoning quoted above was read from Guideline 1 in the full text on 2026-09-06, along with the Guideline 3 and Guideline 4 headings, the dose table, and the three-or-more-risk-factor recommendation gan-2026.

Two prophylactic agents in common use have been removed from this page rather than relabeled, because what this page carried for them were doses and the doses have no source here: dimenhydrinate, and promethazine used for prophylaxis rather than rescue. Neither appears in the guideline’s dose table, which is cited above. Both, and the figures that were removed, are recorded in the repository’s gap register.

At the bedside

Give every adult two interventions from different classes; that is the guideline’s floor at any risk gan-2026. Count the factors anyway and write the band down, using the rounded ladder, because the band tells you how much absolute benefit the two interventions bought, and because the guideline is explicit that the general-prophylaxis policy does not retire the score gan-2026 apfel-2004. If the patient carries three or more factors, the guideline’s next move is to maximize risk mitigation rather than to add a third drug gan-2026.

One habit is worth naming because the guideline contradicts it. Adding a third drug for a patient with three or more risk factors is widely taught, and it is not what the guideline recommends. Its position is that clinical data on three or more prophylactic interventions are scarce, so it recommends maximizing risk mitigation instead, offering acupoint stimulation for patients who do require three or more interventions gan-2026. Treat a third agent as local practice, not as guideline. When a patient vomits in the PACU, the first question is which classes were already given. The rescue must come from a class the patient has not received gan-2026. Do not re-dose a long-acting agent, do not reach for a scopolamine patch as rescue, and reserve a second dose of the prophylactic class for the situation the guideline describes, more than 6 h elapsed and nothing else available gan-2026.

Local practice differs. the specific rescue agents stocked and the order in which they are used differ between institutions; the class-change rule does not.

Check yourself

Five items. Every option carries an explanation. Progress is not saved.