Secondary Logo

Journal Logo

Original Article

Algorithms for the prevention of postoperative nausea and vomiting: an efficacy and efficiency simulation

Kranke, P.*; Eberhart, L. H.; Gan, T. J.; Roewer, N.*; Tramèr, M. R.

Author Information
European Journal of Anaesthesiology: October 2007 - Volume 24 - Issue 10 - p 856-867
doi: 10.1017/S0265021507000713

Abstract

Introduction

Prevention of postoperative nausea and vomiting (PONV) is a worthwhile clinical outcome and has been extensively investigated. An important step towards an improved control of PONV has been the identification of risk factors for PONV [1-3]. Knowing risk factors may help to predict the baseline risk of patients. Prophylaxis may then be targeted towards those who are at an increased risk for developing PONV. These patients are thought to benefit most from a prophylactic anti-PONV strategy.

A number of algorithms for the prevention of PONV have been proposed. They range from dichotomous approaches [4,5] to stepwise models [6]. A dichotomous approach seeks to identify patients at high risk of PONV; only these patients would then receive antiemetic prophylaxis. A stepwise model aims to identify patients with different levels of baseline risks; according to the estimated baseline risk, a more or less aggressive prophylactic antiemetic strategy would then be chosen.

While some algorithms have been tested in highly selected patient groups [7-9], little is known about the performance of algorithms in unselected patients and their impact on outcome. In this context, performance may be defined as efficacy or efficiency. Efficacy is the ability to produce a desired reduction in the incidence of PONV. The term efficiency describes the state or quality of productive working with a minimum wasted effort or expenses. In this context, effort and expenses may be summarized as acquisition costs of antiemetic drugs and costs from managing drug-related adverse effects.

The aim of our study was to compare the performance of published and hypothetical algorithms for the prevention of PONV. We used simulation models, assuming various baseline risks in different patient cohorts. Pharmacological strategies for preventing PONV that have shown efficacy in systematic reviews and large randomized trials were used alone or in combination to simulate different degrees of antiemetic efficacy. Finally, we developed a decision aid to optimize antiemetic prophylaxis.

Methods

In accordance with the literature, our analyses were based on several assumptions.

First, there were four widely practiced prophylactic antiemetic interventions: total intravenous anesthesia with propofol and air-oxygen [10-12], dexamethasone [11,13], 5-HT3 receptor antagonists (ondansetron [14], granisetron [15], tropisetron [16], dolasetron [17]), and dopamine receptor antagonists (droperidol [18], haloperidol [19]). When used in adequate doses for the prevention of PONV, these interventions had a similar degree of antiemetic efficacy [11]; compared with placebo or no intervention, they achieved a relative reduction in the risk of PONV of about 30% and independent of the baseline risk [20]. For a high-risk population with a baseline risk of 60%, this translated into a number-needed-to-treat of 5 for the prevention of PONV within 24 hours when compared with placebo. These interventions could be combined to achieve additive effects. However, with an increasing number of antiemetic interventions, the additional benefit was decreasing since with each additional antiemetic intervention, the ‘new’ baseline risk was becoming lesser (i.e. more favourable). For instance, in a patient population with a baseline risk of PONV of 40%, the prophylactic administration of one antiemetic intervention (relative risk reduction, 30%) was decreasing the baseline risk to 28%, and accordingly, the number-needed-to-treat was about 8 (1/(40-28%)). Adding a second antiemetic intervention (relative risk reduction, 30%) further decreased the baseline risk from 28% to 20%; accordingly, the number-needed-to-treat of this additional benefit was about 13 only (1/(28-20%)).

Second, female gender, a history of motion sickness or PONV, non-smoking status, and the use of postoperative opioids were regarded as independent risk factors for PONV in adults and they could be combined in risk scores to estimate the baseline risk. For simplicity and for the purpose of our study, we assumed that scoring systems that were based on these four risk factors yielded satisfactory results in terms of predictive values, sensitivity and specificity, and that they were robust across different patient populations. Thus, the expected incidences of PONV would be about 10%, 20%, 40%, 60% and 80% if none, one, two, three or four of the risk factors are present [2].

Finally, we assumed that in a clinical study, there was a close relationship between the reported incidence of PONV in patients who did not receive antiemetic prophylaxis (i.e. the control event rate) and the true underlying (or baseline) risk of that patient population.

Antiemetic algorithms

We searched Medline for reports of PONV algorithms. We combined terms that were designed to retrieve studies on PONV [21] with terms targeting scores and algorithms, i.e. ‘score’ or ‘algorithm’ or ‘risk assessment’. Further, we used the ‘Cited Reference Search’ option provided by Web of Science® to locate relevant articles on risk scores. We selected algorithms that were using female gender, history of motion sickness or PONV, non-smoking status, and use of postoperative opioids as risk factors. The last search was performed in November 2005. During an international PONV consensus meeting that was held in Miami, Florida, in November 2005, we discussed algorithms that may be useful in daily clinical practice (see Acknowledgements section for a list of participants).

Efficacy and efficiency analyses

For each algorithm and each baseline risk, we computed the total number of patients receiving prophylaxis, the total number of antiemetic interventions administered, and the cumulative incidence of PONV at 24 h. We did not take into account the severity of PONV (for instance, the number of vomiting episodes or the degree of nausea). Additionally, for each antiemetic intervention we calculated an Efficiency Index similarly to a number-needed-to-treat but considering that sometimes more than one intervention was administered, i.e. the number of antiemetics applied divided by the absolute risk reduction that was achieved [22]. The index was expected to be low if the efficiency of an antiemetic strategy was high (strong efficiency). Analyses were performed for cohorts of 100 patients. For analyses we used Microsoft Excel® on a PC.

Benchmarking

To facilitate comparisons between algorithms across different patient populations, we calculated average values for efficacy and efficiency parameters. We also arbitrarily defined criteria that would identify a clinically ‘useful’ algorithm based on current knowledge and practice reality. These criteria were as follows: for a cohort of 100 patients, (1) No more than 75% of the patients received prophylaxis; (2) The number of administered antiemetic interventions was <200 (<2 antiemetics per patient); (3) The Efficiency Index was <10; and (4) The achieved, final incidence of PONV was <25%.

Results

Algorithms

We identified three published algorithms that were tested prospectively in clinical practice [7-9]. One of those was excluded since the distribution of risk factors in the studied population was not reported and the authors were unable to provide that information [9]. We added three published algorithms that had been proposed but not validated [4-6,23,24], and made up a further five (Table 1). The latter group included three algorithms without any risk-adaptation (fixed single, double or triple antiemetic prophylaxis for all patients). Finally, one algorithm was not considered by us as it did not take into account the four risk factors that we had selected for the purpose of our analyses [25].

Table 1
Table 1:
Efficacy and efficiency parameters of algorithms for the prevention of PONV. Scenarios are shown for different algorithms and baseline risks.

There were three fixed algorithms; independent of the presumed baseline risk, all patients received one, two, or three prophylactic antiemetic interventions (Table 1A). Two were dichotomous and proposed administration of one antiemetic only (Table 1B). Three used an escalating approach; one to four antiemetic interventions were administered depending on the presumed baseline risk (Table 1C). Finally, two algorithms proposed the identification of high-risk patients and to administer three or four antiemetic interventions exclusively to these high-risk patients (Table 1D).

Patient populations

Two published studies reported on validation of risk scores in real-life surgical populations [7,8] (Table 1). The study by Biedler and colleagues [8] was performed in 162 patients undergoing elective surgery (mainly orthopaedic and gynaecological). Patients were stratified into ‘low-risk’ patients with 0 or 1 risk factor who did not receive any prophylaxis and ‘high-risk’ patients with 2 to 4 risk factors who received one antiemetic intervention. The estimated baseline risk was on average 47%. The study by Pierre and colleagues included 428, mainly female, patients undergoing gynaecological, throat or thyroid surgery [7]. Patients were stratified to receive no antiemetic intervention (‘low-risk’ patients with 0 or 1 risk factor), one intervention (‘moderate-risk’ patients with 2 risk factors), or three interventions (‘high-risk’ patients with 3 or 4 risk factors). The estimated baseline risk was on average 50.7%. In a historical control population, the observed PONV incidence was on average 49.5% [26].

We added unpublished institutional data (Marburg; average baseline risk, 44.6%) from one of the authors (LE), and made up four further, hypothetical patient populations (Table 1). The hypothetical populations represented homogenous groups with mainly low, mainly intermediate, or mainly high-risk patients according to Pierre and colleagues [7], or with an even distribution of the number of risk factors. Depending on the distribution of risk factors, the average calculated incidence of PONV in the hypothetical patient populations varied from 24.2% (mainly with low risk) to 59.8% (mainly with high risk).

Efficacy and efficiency of antiemetic algorithms

Application of all tested algorithms to all populations provided 70 different scenarios (Table 1).

Number of treated patients

The average number of treated patients varied from 41 to 100. The lowest average number (n = 41) was with dichotomous, risk-adapted approaches (No. 4, 9 and 10). Of all the 70 scenarios, the lowest number of exposed patients (n = 10) was when dichotomous, risk-adapted approaches (No. 4, 9 and 10) were applied to a hypothetical population with mainly low baseline risk. The highest average number of treated patients (n = 100) was with fixed algorithms that did not use any risk-adaptation (No. 1 to 3).

Number of administered antiemetic interventions

The average number of administered antiemetic interventions varied from 41 to 300. The lowest average number (N = 41) was with the dichotomous approach that used one single intervention in high-risk patients only (No. 4). Of all 70 scenarios, the lowest number of administered interventions (n = 10) was when the dichotomous, single-intervention algorithm (No. 4) was applied to a mainly low-risk population. The highest average numbers of administered interventions (n = 200 to 300) were with those algorithms without risk-adaptation that used two or three interventions (No. 2 and 3).

Efficiency index

The average efficiency index varied from 5 to 12. The lowest average index (5) was with the dichotomous algorithm that proposed to administer one intervention to high-risk patients only (No. 4). The highest average index (12) was with the algorithm without risk-adaptation that used three interventions (No. 3). Of all 70 scenarios, the highest index (19) was when the algorithm without risk-adaptation that used three interventions in all patients (No. 3) was applied to the mainly low-risk population.

Final cumulative incidence of PONV

The average incidence of PONV varied from 15% to 36%. The lowest incidences (15% and 18%, respectively) were with the dichotomous algorithm with risk-adaptation that proposed the administration of three interventions (No. 3), and with the algorithm with escalating risk-adaptation that included up to four antiemetic interventions (No. 8). Of all 70 scenarios, the lowest PONV incidence (8%) resulted when the algorithm without risk-adaptation and three interventions (No. 3) was applied to a population with mainly low risk. The highest PONV incidence (36%) was observed in the algorithm with risk-adaptation and one single intervention (No. 1).

Benchmarking

Several algorithms (No. 4-7, 9 and 10) satisfied all four arbitrarily defined usefulness criteria (<75% of the patients received prophylaxis, the number of administered antiemetic doses was <200, the efficiency index was <10, and the incidence of PONV was <25%). However, this was true for the two hypothetical populations with low and even risk distributions only. When considering average efficacy and efficiency parameters across all patient populations, one algorithm (No. 10) satisfied all criteria of a useful algorithm, five (No. 4-7, 9) satisfied three criteria, two (No. 1 and 8) satisfied two, and two (No. 2 and 3) satisfied one.

We plotted the number of administered antiemetic interventions with each algorithm and each patient population against the resulting incidence of PONV (Fig. 1). With an increase in the number of administered antiemetic interventions, the incidence of PONV was consistently decreasing, independent of the algorithms. With dichotomous risk-adaptation algorithms, where exclusively high-risk or moderate to high-risk patients received one antiemetic intervention (No. 4 and 5), the number of administered antiemetic interventions was lowest (<100) but the resulting incidences of PONV were highest (>30%) with most patients populations. With the algorithm without risk-adaptation where all patients received three antiemetic interventions (No. 3), the number of administered antiemetic interventions was highest (300) but the resulting incidences of PONV were the lowest (8-21%). All algorithms had a particularly favourable ratio between the number of administered antiemetic interventions and the resulting incidence of PONV when they were applied to the population with mainly low baseline risk.

Figure 1.
Figure 1.:
Relationship between the total number of administered antiemetics and the incidence of PONV. For each algorithm, 7 symbols are plotted that represent one of 7 patient populations with different baseline risks (seeTable 1). The grey shading area covers arbitrarily defined degrees of worthwhile efficacy and efficiency (i.e. cumulative PONV incidence <25%, total number of administered antiemetic interventions <200). The ellipse surrounds algorithm-population combinations that perform particularly well within the pre-specified degrees of worthwhile efficacy and efficiency. Algorithms without risk adaptation (No. 1-3, Table 1): All patients receive prophylactically one (□), two (Symbol) or three (Symbol) antiemetic interventions. Algorithms with dichotomous risk adaptation (No. 4, 5, Table 1): One antiemetic intervention is administered to patients with high risk (○) or moderate to high risk (Symbol). Algorithms with escalating risk adaptation (No. 6-8, Table 1): One to two (△), one to three (Symbol) or one to four (Symbol) antiemetic interventions are administered depending on the presumed baseline risk. Algorithms with dichotomous, risk (No. 9, 10,Table 1): Three (▽) or four (Symbol) antiemetic interventions are administered to high-risk patients.
Symbol
Symbol
Symbol
Symbol
Symbol
Symbol
Symbol
Symbol
Symbol
Symbol
Symbol
Symbol

Determining a ‘tailored’ antiemetic approach

Table 2 describes the various steps to create an individual algorithm, considering institutional policies, resource allocation, and baseline risk. This approach is subject to various sources of uncertainty, such as the distribution of risk factors, the degree of antiemetic efficacy, and the precision of the risk score. A calculator to perform more detailed analyses of specific populations has been made freely accessible on the Internet (http://qm-ains.de/PONV.xls).

Table 2
Table 2:
Decision aid to find a tailored antiemetic approach.

Discussion

This study has three main results. First, a number of algorithms satisfied our arbitrarily defined criteria of usefulness of PONV prophylaxis, but only when applied to specific patient populations; none had consistently satisfactory values for efficacy and efficiency across all patient populations. Second, when certain algorithms were applied to selected patient populations, the PONV incidence could be decreased below 15%; however, none produced consistent PONV incidences below 20% across all populations. Third, the single most important factor that had an impact on the incidence of PONV was the number of antiemetic interventions that was administered to the patients.

The strength of an algorithm for the prevention of PONV may be estimated using very different criteria. In a healthcare system with limited resources, for instance, the total number of administered antiemetic interventions may be of most interest, since this will have most impact on cost. Others may be more interested in keeping the number of exposed patients as low as possible to minimize the risk of adverse drug reactions. Criteria may even vary within one setting. For instance, in patients with wired jaws, PONV may be life threatening, and consequently the threshold to apply antiemetic prophylaxis will be low. It may be argued that individual patients should not inform the general approach; in specific cases, the general approach should be modified based on setting and rational decisions. Such a context-sensitive approach is in accordance with international recommendations [27].

It is interesting to note that with most algorithms a favourable ratio between a (lowest) number of administered antiemetics and a resulting (lowest) incidence of PONV could be achieved (Fig. 1). However, this was often associated with the mainly low-risk patient population. In these patients, prophylaxis may not be useful.

Our analysis has some limitations. First, we limited to four the number of factors that were thought to determine the baseline risk. It has been claimed that these factors accurately predicted the risk of PONV in individual patients [1,2]. This, however, is contentious. One external validation study confirmed the validity of the four risk factors [28]. Another analysis concluded that calibration and discrimination of scores that were based on these four risk factors were unsatisfactory and that additional factors should be considered [29]. This paper also suggested that other risk scores (i.e. the score developed by Koivuranta) might better predict PONV in the investigated population [29]. Alternative PONV risk factors have been described [30]; inclusion of those into our model may further improve efficacy and efficiency variables of the algorithms. However, it is unlikely that prediction of the PONV baseline risk based on clinical predictors only will become more reliable [31]. At least this is true if no customization is performed [29,32].

Second, we limited the number of antiemetic interventions to four. Although there is strong evidence that all these interventions are truly antiemetic and that their combination leads to an additive effect, we cannot exclude that alternative antiemetic interventions would further increase efficacy. Further, we cannot exclude that occasionally observed interactions, for instance between gender and the efficacy of antiemetics, might impair the implications and the transferability of our findings to every antiemetic interventions and every subgroup of patients. Since the necessary data are lacking, we do not know either whether newer antiemetic drugs, such as neurokinin-1 antagonists, would be more efficacious when used alone or in combination. It is, however, unlikely that the combination of more than four antiemetic interventions would lead to a relevant increase in efficacy since with each further intervention that is added to a combination therapy, the additional benefit will diminish. This, however, does not mean that in specific groups and for defined postoperative periods (i.e. stay in the PACU) a multimodal PONV-management programme will not be able to drastically reduce PONV [33]. Also, we assumed that regimens and timing of administration of the antiemetic drugs were optimal. On the other hand, the implications of such research and the results of any PONV trial may be questioned due to the binary consideration of the treatment success. It should not be left unmentioned that an alleviation of symptoms may well imply a benefit for the individual patient.

Third, we tested efficacy and efficiency of a limited number of real-life and theoretical algorithms. Alternative algorithms that were not tested here may indeed achieve better results. However, the calculator that is freely accessible on the Internet may facilitate the development and testing of further algorithms.

Fourth, we restricted the number of patient populations to seven. As a consequence, our scenarios do not represent all potential clinical situations. Again, the calculator may help clinicians to simulate alternative risk constellations that may best correspond to their patient populations. The distribution of risk factors and the expected risk reduction will define what algorithm offers most benefit. In a given patient population, a risk-adapted stepwise algorithm or a risk-adapted dichotomous approach may yield most efficiency compared with a general single prevention without losing much efficacy. Alternatively, in some clinical settings (e.g. Pierre and colleagues), it may be irrelevant with respect to the number of exposed patients and the number of antiemetics spent whether a fixed single prevention (algorithm No. 1, PONV incidence 36%) or a risk-adapted single intervention (algorithm No. 5, PONV incidence 36%) is chosen. As a rule of thumb, this ambivalence happens most often when the population consists mainly of high-risk patients where selective administration of antiemetic interventions does not greatly affect the overall efficiency. The opposite is true for patient populations with a homogenous distribution of risk factors. There, the restrictive use of antiemetics in low-risk patients and the more generous use in high-risk patients makes sense from an economic point of view. For that group of patients it has been demonstrated that prevention is likely to increase patient satisfaction compared with treatment [34]. However, it has never been demonstrated that a distinct algorithm offers any benefit over another in terms of increased patient satisfaction, providing that aggressive treatment of established PONV symptoms is assured. This is clearly an issue for future research.

Finally, we presented point estimates without any quantification of the degree of precision. This could easily be done by computing, for instance, 95% CIs around the point estimates. However, upper and lower limits of 95% CIs depend on sample size. We have arbitrarily set the size of the cohorts at 100 patients. Increasing that size would not have an impact on the point estimates but would simply reduce the 95% CIs. For instance, when choosing the ‘one antiemetic for each patient’ algorithm for the Biedler cohort, the predicted average incidence of PONV was 33%, and the 95% CI went from 24% to 43%. If we chose a cohort of 1000 patients, the average PONV incidence was still 33%, but the 95% CI went from 30% to 36%. We preferred to lose some precision for the sake of clarity, and therefore decided to present point estimates only.

Ideally, an algorithm would reduce the incidence of PONV to a minimum. There is strong evidence from our analyses that the total number of administered antiemetic interventions is the single most important factor that eventually determines that incidence. Thus, the most likely strategy to achieve the lowest incidence of PONV would be to administer all antiemetic interventions to all patients and without any risk stratification. There are obviously arguments against this approach. For instance, our analyses suggest that even with the most successful combinations of algorithms and populations, the incidence PONV will remain 10% to 20%. Also, universal prophylaxis without any risk stratification would unnecessarily increase acquisition costs and the number of exposed patients who actually do not need any prophylaxis since they would not vomit anyway. This is particularly true for low-risk populations; indeed, the consistently least attractive (i.e. highest) efficiency indexes were seen with the fixed triple intervention (algorithm No. 3) when applied to a population with mainly low-risk patients. The other extreme would be to wait until the patient develops PONV symptoms in the postoperative period and then to treat aggressively. The often unsatisfactory values for efficacy and efficiency when the tested algorithms were applied to some of the patient populations would be an argument in favour of this pragmatic ‘wait and see’ approach. The unsatisfactory values regarding the prophylactic efficacy of antiemetics are in accordance with studies that aimed to determine patient satisfaction with treatment as compared to prevention [34]. Scuderi and colleagues, for instance, found that for the average patient population there was little difference in outcomes when routine prophylactic medications were administered versus simply treating PONV symptoms [34].

In conclusion, there is strong evidence that the combination of antiemetic interventions increases antiemetic efficacy. However, there is no gold standard algorithm for the prevention of PONV. It is likely that despite optimized prophylaxis, the incidence of PONV will remain 10% to 20%. While the combination of risk-adapted multimodal approaches and simplified risk scores have been widely advocated, there is evidence that in populations with particular risk distributions, a pragmatic fixed approach or a dichotomous approach may be more successful. Clinicians should know about risk distributions in their patient populations. They need to decide whether all resources should be used to further improve prophylaxis, or, alternatively, whether a part of these resources may be spent for the treatment of established symptoms. An initial observation phase could help in order to better define the local risk level and thus customizing the most suitable institutional approach. Rational management of PONV should remain context-sensitive and should take into account patient preferences, risk factors, and surgical setting.

Future research should not only concentrate on the final incidence of PONV but should also concentrate on patient satisfaction, the risk of PONV-related complications, the risk of adverse drug reactions, costs of antiemetics, and costs incurred by unanticipated admission due to intractable PONV symptoms. Practical issues regarding the implementation of new PONV policies, quality management and adherence to guidelines should be studied. Finally, and perhaps most importantly, treatment of established PONV symptoms should inform future research, considering the inherent limited efficiency of antiemetic prophylaxis. Treatment appears to be fundamentally different from prophylaxis. When treatment is chosen, all exposed patients will have PONV symptoms before they receive the treatment; thus, there will always be a 100% baseline risk. For therapeutic antiemetic interventions, numbers-needed-to-treat to prevent further PONV symptoms compared with placebo are between 4 and 5 [35]. As a consequence, the degree of therapeutic efficacy of antiemetics is likely to be valid for all nauseous or vomiting patients and is not context sensitive. Also, since with the treatment approach only patients who actually need an antiemetic will eventually receive that antiemetic, the risk of drug-related harm in the absence of efficacy will be minimized. The occurrence of PONV in the PACU may also be considered as a kind of tracer indicating that an aggressive treatment and further prevention (secondary prophylaxis) is needed. Valid clinical trials on the treatment of PONV symptoms and the capability of antiemetic drugs, alone or in combination [36], to abolish established emetic symptoms are needed.

Acknowledgements

We would like to thank the participants of the 2005 international PONV consensus meeting who contributed to the project with their lively and valuable discussion. Discussants were: T. Meyer, C. C. Apfel, F. Chung, P. Davis, A. Habib, V. Hooper, A. Kovac, P. Myles, B. Philip, G. Samsa, D. I. Sessler, J. Temo and C. Vander Kolk (apart from the authors P. Kranke, T. J. Gan and M. R. Tramèr). This meeting took place in Miami, Florida, November 4-6, 2005 and was financially supported by a grant from the Society for Ambulatory Anesthesia (SAMBA). Intramural departmental funding only.

Conflict of interest

None.

References

1. Koivuranta M, Läärä E, Snare L, Alahuhta S. A survey of postoperative nausea and vomiting. Anaesthesia 1997; 52: 443-449.
2. Apfel CC, Läärä E, Koivuranta M, Greim CA, Roewer N. A simplified risk score for predicting postoperative nausea and vomiting: Conclusions from cross-validations between two centers. Anesthesiology 1999; 91: 693-700.
3. Eberhart LH, Geldner G, Kranke P et al. The development and validation of a risk score to predict the probability of postoperative vomiting in pediatric patients. Anesth Analg 2004; 99: 1630-1637.
4. Tramèr MR. Rational control of PONV - the rule of three. Can J Anaesth 2004; 51: 283-285.
5. Tramèr MR. Strategies for postoperative nausea and vomiting. Best Pract Res Clin Anaesthesiol 2004; 18: 693-701.
6. Apfel CC, Roewer N. Risk assessment of postoperative nausea and vomiting. Int Anesthesiol Clin 2003; 41: 13-32.
7. Pierre S, Corno G, Benais H, Apfel CC. A risk score-dependent antiemetic approach effectively reduces postoperative nausea and vomiting - a continuous quality improvement initiative. Can J Anaesth 2004; 51: 320-325.
8. Biedler A, Wermelt J, Kunitz O et al. A risk adapted approach reduces the overall institutional incidence of postoperative nausea and vomiting. Can J Anaesth 2004; 51: 13-19.
9. Rüsch D, Eberhart L, Biedler A, Dethling J, Apfel CC. Prospective application of a simplified risk score to prevent postoperative nausea and vomiting. Can J Anaesth 2005; 52: 478-484.
10. Apfel CC, Kranke P, Katz MH et al. Volatile anaesthetics may be the main cause of early but not delayed postoperative vomiting: a randomized controlled trial of factorial design. Br J Anaesth 2002; 88: 659-668.
11. Apfel CC, Korttila K, Abdalla M et al. A factorial trial of six interventions for the prevention of postoperative nausea and vomiting. N Engl J Med 2004; 350: 2441-2451.
12. Tramèr MR, Moore A, McQuay H. Propofol anaesthesia and postoperative nausea and vomiting: quantitative systematic review of randomized controlled studies. Br J Anaesth 1997; 78: 247-255.
13. Henzi I, Walder B, Tramèr MR. Dexamethasone for the prevention of postoperative nausea and vomiting: a quantitative systematic review. Anesth Analg 2000; 90: 186-194.
14. Tramèr MR, Reynolds DJ, Moore RA, McQuay HJ. Efficacy, dose-response, and safety of ondansetron in prevention of postoperative nausea and vomiting: a quantitative systematic review of randomized placebo-controlled trials. Anesthesiology 1997; 87: 1277-1289.
15. Kranke P, Apfel CC, Eberhart LH, Georgieff M, Roewer N. The influence of a dominating centre on a quantitative systematic review of granisetron for preventing postoperative nausea and vomiting. Acta Anaesthesiol Scand 2001; 45: 659-670.
16. Kranke P, Eberhart LH, Apfel CC, Broscheit J, Geldner G, Roewer N. Tropisetron for prevention of postoperative nausea and vomiting: a quantitative systematic review. Der Anaesthesist 2002; 51: 805-814.
17. Kranke P, Eberhart LHJ, Morin AM, Roewer N. Dolasetron for prevention of postoperative nausea and vomiting: A meta-analysis of randomized controlled trials. Anaesthesiol Intensivmed 2002; 43: 413-427.
18. Henzi I, Sonderegger J, Tramèr MR. Efficacy, dose-response, and adverse effects of droperidol for prevention of postoperative nausea and vomiting. Can J Anaesth 2000; 47: 537-551.
19. Büttner M, Walder B, von Elm E, Tramèr MR. Is low-dose haloperidol a useful antiemetic?: a meta-analysis of published and unpublished randomized trials. Anesthesiology 2004; 101: 1454-1463.
20. Kranke P, Eberhart LH. Nausea and vomiting after anesthesia - what is certain in prophylaxis and treatment? Anasthesiol Intensivmed Notfallmed Schmerzther 2005; 40: 549-554.
21. Kranke P, Morin AM, Roewer N, Wulf H, Eberhart LH. The efficacy and safety of transdermal scopolamine for the prevention of postoperative nausea and vomiting: a quantitative systematic review. Anesth Analg 2002; 95: 133-143.
22. Tramèr MR, Walder B. Number needed to treat (or harm). World J Surg 2005; 29: 576-581.
23. Watcha MF. The cost-effective management of postoperative nausea and vomiting. Anesthesiology 2000; 92: 931-933.
24. Gan TJ. Postoperative nausea and vomiting - can it be eliminated? JAMA 2002; 287: 1233-1236.
25. Habib AS, Gan TJ. Evidence-based management of postoperative nausea and vomiting: a review. Can J Anaesth 2004; 51: 326-341.
26. Pierre S, Benais H, Pouymayou J. Apfel's simplified score may favourably predict the risk of postoperative nausea and vomiting. Can J Anaesth 2002; 49: 237-242.
27. Gan TJ, Meyer T, Apfel CC et al. Consensus guidelines for managing postoperative nausea and vomiting. Anesth Analg 2003; 97: 62-71.
28. Apfel CC, Kranke P, Eberhart LH, Roos A, Roewer N. Comparison of predictive models for postoperative nausea and vomiting. Br J Anaesth 2002; 88: 234-240.
29. van den Bosch JE, Kalkman CJ, Vergouwe Y et al. Assessing the applicability of scoring systems for predicting postoperative nausea and vomiting. Anaesthesia 2005; 60: 323-331.
30. van den Bosch JE, Moons KG, Bonsel GJ, Kalkman CJ. Does measurement of preoperative anxiety have added value for predicting postoperative nausea and vomiting? Anesth Analg 2005; 100: 1525-1532.
31. Apfel CC, Kranke P, Greim CA, Roewer N. What can be expected from risk scores for predicting postoperative nausea and vomiting? Br J Anaesth 2001; 86: 822-827.
32. Engel JM, Junger A, Hartmann B et al. Performance and customization of 4 prognostic models for postoperative onset of nausea and vomiting in ear, nose, and throat surgery. J Clin Anesth 2006; 18: 256-263.
33. Scuderi PE, James RL, Harris L, Mims III GR. Multimodal antiemetic management prevents early postoperative vomiting after outpatient laparoscopy. Anesth Analg 2000; 91: 1408-1414.
34. Scuderi PE, James RL, Harris L, Mims III GR. Antiemetic prophylaxis does not improve outcomes after outpatient surgery when compared to symptomatic treatment. Anesthesiology 1999; 90: 360-371.
35. Kazemi-Kjellberg F, Henzi I, Tramèr MR. Treatment of established postoperative nausea and vomiting: a quantitative systematic review. BMC Anesthesiol 2001; 1: 2.
36. Rüsch D, Arndt C, Martin H et al. Dexamethasone enhances antiemetic efficacy of dolasetron and haloperidol for treatment of established PONV. Eur J Anaesthesiol 2006; 23: A-59.

Appendix: Formulas and examples for the calculation of the indicators of efficiency used for the evaluation of the different antiemetic strategies

Example A: Algorithm without risk-adaptation, single intervention (‘one intervention for all’), applied to the population described by Biedler and colleagues ([8], Table 1, A1).

Table
Table

Derived calculations for various outcomes

Absolute risk reduction (ARR):

  • ARR = risk (without prophylaxis) − risk (with prophylaxis)
  • = 47.0% − 32.9% = 14.1% (not explicitly stated in Table 1)

Relative risk reduction (RRR):

  • RRR = ARR/risk (without prophylaxis)
  • =14.1%/47.0% = 30%

Efficiency index (EI):

  • EI = number of antiemetics (for 100 patients)/ARR
  • =100/14.1 = 7.1 (rounded to ‘7’ in Table 1)

Number-needed-to-treat (NNT):

  • NNT = 1/ARR
  • = 1/14% = 7.1

Example B: Algorithm suggesting one to four interventions for low- to high-risk patients, applied to the population described by Biedler and colleagues ([8], Table 1, C8).

Table
Table

Derived calculations for various outcomes

Absolute risk reduction (ARR):

  • ARR = risk (without prophylaxis)−risk (with prophylaxis)
  • = 47.00%−18.2% = 28.8% (not explicitly stated in Table 1)

Relative risk reduction (RRR):

  • RRR = ARR/risk (without prophylaxis)
  • 28.8%/46.7% = 61.2%

Efficiency index (EI):

  • EI = number of antiemetics (for 100 patients)/ARR
  • =233/28.8 = 8.1 (rounded to ‘8’ in Table 1)

Number-needed-to-treat (NNT):

  • NNT = 1/ARR
  • = 1/28.8% = 3.5
Keywords:

Information science, computer simulation; HEALTHCARE EVALUATION MECHANISMS, meta-analysis, logistic regression; EVALUATION STUDIES, programme evaluation; POSTOPERATIVE COMPLICATIONS, postoperative nausea and vomiting; QUALITY ASSURANCE, HEALTHCARE, practice guidelines

© 2007 European Society of Anaesthesiology