To decrease the incidence of postoperative nausea and vomiting (PONV), several antiemetic strategies are available, e.g., use of propofol anesthesia or prophylactic administration of antiemetic drugs. Routine administration of antiemetics, however, increases costs and subjects all patients to potential side effects (1). Reliable preoperative prediction of PONV risk would enable anesthesiologists to selectively apply antiemetic strategies.
Prediction rests on knowledge of the etiology of PONV, which is multifactorial and includes patient-related, anesthetic, and surgical factors (1). Sex, smoking, history of PONV or motion sickness, inhaled anesthesia, surgery type, and postoperative opioid analgesia are reported predictors of PONV (2–8). Several authors have suggested that preoperative anxiety might also be predictive for PONV (1,9). However, data supporting this hypothesis are still lacking.
To assess the added value of anxiety in preoperative prediction of PONV, we first analyzed the (univariate) association of anxiety and PONV. Subsequently, we assessed the predictive accuracy of a model with previously reported predictors only, and finally, we assessed whether preoperative anxiety improves the predictive accuracy of these predictors.
We analyzed data from patients who participated in a randomized trial investigating differences in PONV incidence after IV or inhaled general anesthesia. The design and results of this trial have been reported previously (8). In brief, after approval by the institutional medical ethics committee and written informed consent, inpatients and outpatients scheduled for surgery and aged 18–80 yr were randomized to receive surgery with IV anesthesia with propofol or with inhaled anesthesia using isoflurane and nitrous oxide. All types of surgery were included except for cardiac surgery and intracranial neurosurgery. Exclusion criteria were emergency surgery, (possible) pregnancy, ASA class IV, morbid obesity (weight >120 kg), renal or liver disease precluding use of either anesthetic technique, and use of antiemetic or proemetic medication in the 2 wk before surgery. Patients receiving regional anesthetic techniques were not included in the study, except in the case of upper abdominal surgery, for which epidural analgesic supplementation was permitted. Blinding was maintained at all stages of the trial. This analysis was based on data from the inpatients.
To quantify preoperative anxiety, patients filled in two previously validated questionnaires for assessing preoperative anxiety: the Dutch version of the Spielberger State-Trait Anxiety Index (STAI) (10) and the Amsterdam Preoperative Anxiety and Information Scale (APAIS) (11). The STAI questionnaire consists of 40 items on a 4-point rating scale measuring anxiety state (20 items) and anxiety disposition (20 items). Minimum and maximum scores range from 20 to 80 points. The APAIS questionnaire consists of six questions, each scored on a 5-point scale from 1 (not at all) to 5 (extremely). The APAIS scale was specifically designed to assess patients’ preoperative anxiety (four questions; range 4–20) and information-seeking behavior (two questions; range 2–10) regarding the scheduled surgery and anesthesia.
On the basis of previous studies on PONV prediction (1–4,6,8), we investigated 12 additional potential predictors of PONV. These included age, sex, ethnicity (Caucasian versus non-Caucasian), body mass index (weight [in kilograms] divided by squared height [in meters]), ASA physical status, smoking, history of PONV (PONV after previous anesthesia: yes or no), history of motion sickness, anesthetic technique (IV anesthesia with propofol versus inhaled anesthesia with isoflurane/nitrous oxide), and type of scheduled surgery. Type of surgery was categorized as superficial, laparoscopy, upper abdomen, lower abdomen, strabismus, middle ear, and other (seven categories). In addition, postoperative opioid analgesia and the duration of anesthesia were recorded for each patient (2,4).
The induction of anesthesia was achieved with thiopental in patients randomized to isoflurane/nitrous oxide or with propofol in patients randomized to anesthesia with propofol/air. Anesthesia was maintained with propofol or isoflurane in nitrous oxide according to the randomization. The treatment allocation was concealed until immediately before the induction of anesthesia. Prophylactic antiemetics were not permitted. Intraoperatively, the anesthesiologist was free to choose the types and doses of muscle relaxants and (opioid) analgesics, as well as other drugs for the supplementation of anesthesia (“pragmatic study design”). Postoperative administration of antiemetic therapy was the responsibility of postanesthesia care unit (PACU) nurses. The protocol prescribed IV metoclopramide (0.15 mg/kg) as the first-choice antiemetic therapy, followed by IV droperidol (15 μg/kg) if necessary.
The outcome of this study was PONV within the first 24 h after surgery. The presence of PONV was defined as at least one episode of nausea (any degree, including mild nausea) or vomiting or retching, or any combination of these emetic symptoms within the first 24 h after surgery. Each patient with PONV was counted only once, irrespective of the frequency of emetic symptoms.
A trained, blinded research nurse, who was not responsible for the postoperative care, recorded nausea, retching, and vomiting (scored separately). During the patient’s stay in the PACU, nausea, retching, and vomiting were recorded at four time points: on arrival, 1 h after arrival, at the time of permission for discharge, and at the time of actual discharge. At the ward, symptoms were recorded each hour up to 24 h after surgery. We also documented, at each time point, whether the patient was sufficiently alert to complete a verbal rating scale. Patients who were too drowsy or sleepy received a score of 0 for that time point.
First, we quantified the univariate association between each anxiety variable and the incidence of PONV. Subsequently, the predictive accuracy of combinations of previously reported predictors was quantified by using multivariate logistic regression modeling. The resulting (basic) model, which included the independent predictors of PONV only, was then extended by adding the anxiety (APAIS and STAI) variables, separately and in combination, to estimate their added predictive value. Of particular interest was the difference in incremental value, if any, of the brief APAIS questionnaire compared with the much longer STAI. Differences between models were estimated with the log likelihood ratio test, and a model’s ability to discriminate between patients with and without PONV was quantified by using the area under the receiver operating characteristic curve (ROC area).
A simple prediction rule based on these results was derived for clinical practice. The predictive accuracy of this rule was estimated by its calibration (i.e., agreement between observed and predicted probabilities of PONV), discrimination, and the absolute number of correctly predicted patients with and without PONV across various risk categories. Calibration was tested by using the Hosmer and Lemeshow method (12).
The accuracy of a prediction rule is always too optimistic on the original data set. Hence, we estimated this over-optimism by using bootstrapping techniques (13) (rather than split-sample or cross-validation methods, because the former has proven to be superior) (14), adjusted the rule for this overoptimism, and calculated the ROC area that can be expected in future patients (15).
In total, 667 patients (48%) experienced PONV within 24 h after surgery, of which 414 (62%) developed PONV in the PACU and 253 (38%) developed PONV 2–24 h after surgery. Vomiting occurred in 402 patients (29%).
All anxiety measurements, except the APAIS need for the information score, were significantly associated with PONV (Table 1). A higher anxiety level was associated with a more frequent incidence of PONV. Also, most investigated predictors were associated with PONV, except for ASA physical status, premedication with benzodiazepines, and perioperative and postoperative opioid analgesia (Table 1). The predictors “history of PONV” and “history of motion sickness” showed similar associations with PONV (Table 1). Therefore, they were combined into a single predictor “history of PONV or motion sickness,” which yielded an odds ratio (OR) of 2.41. Likewise, “lower abdominal surgery” and “middle ear surgery” were combined to the index category “high-risk surgery” versus all other types of surgery as the reference category (Table 1). A model including the four established predictors sex, history of PONV or motion sickness, smoking, and postoperative opioid analgesia (2) yielded an ROC area of 0.67 (95% confidence interval [CI], 0.65–0.69). All predictors were significantly associated with PONV (P < 0.001), except for postoperative opioid analgesia (P = 0.20). Excluding postoperative opioid analgesia and adding age, anesthetic technique, and type of surgery (included either as six indicator variables or as a dichotomous predictor) significantly increased the ROC area to 0.72 (95% CI, 0.70–0.74). All six predictors in this basic model were significantly associated with PONV (P < 0.001). None of the other previously reported predictors further improved the predictive accuracy of this basic model. For example, the P value for premedication was 0.55 by multivariate analysis.
When adding the two STAI anxiety predictors to the basic model, the STAI anxiety state was significant (OR, 1.01 per unit change; 95% CI, 1.00–1.02; P = 0.04), but the STAI anxiety trait was not (OR, 1.00 per unit change; 95% CI, 0.99–1.01; P = 0.8). The ROC area remained 0.72. When the 2 APAIS anxiety predictors were added to the basic model, the APAIS anxiety score was significant (OR, 1.04 per unit change; 95% CI, 1.02–1.05; P = 0.02), but the APAIS need for information score was not (OR, 0.97 per unit change; 95% CI, 0.94–1.01; P = 0.22). Again, the ROC area remained 0.72. The combined addition of the four anxiety predictors did not improve the predictive accuracy of the basic model. When nausea at 24 h or vomiting at 24 h was used as an end-point, the same predictors were found, and the addition of preoperative anxiety as a variable again did not improve the predictive accuracy of the model.
Hence, the final prediction rule included sex, age, smoking, history of PONV or motion sickness, type of surgery, and anesthetic technique. Table 2 shows the association of each of these predictors with the outcome after adjustment for over-optimism. Validation of the model with bootstrapping techniques yielded an estimated over-optimism in the ROC area of 0.02, such that the ROC area of the final model that might be expected in future patients is 0.70 (95% CI, 0.68–0.72). The calibration of the final model was good (the P value of the Hosmer and Lemeshow test was 0.45).
With the nomogram (Fig. 1), one can preoperatively estimate for each future patient the probability of developing PONV within 24 h after surgery. Alternatively, one can estimate this probability by using the score table in Appendix 1. Table 3 shows the observed incidence of PONV per risk category of the final model, which ranged from 23% to 76%. Horizontal reading of the table provides estimates of the sensitivity and specificity of the rule at different thresholds. For example, when introducing a threshold at a score of 34 or an estimated probability of 0.55, a score ≥34 (probability ≥0.55) should be considered as “test positive.” When administering preemptive antiemetics to these and only these patients, 72% [(124 + 166 + 190)/667] of the patients who indeed developed PONV would be treated correctly (sensitivity or true-positive rate), whereas in those with a score less than 34, in 57% [(270 + 141)/722], antiemetics would be correctly withheld (specificity or true-negative rate). Using a lower threshold at a score of 29 (probability ≥0.45), the sensitivity would be 88%, and the specificity would be 37% (Fig. 2).
Although several authors have suggested that preoperative anxiety is a predictor of PONV (1, 9), this is the first study that systematically quantified this association while accounting for established predictors of PONV. The results of this study among adult surgical inpatients show that high levels of preoperative anxiety are associated with PONV. However, because this association was relatively weak, the prediction of PONV based on previously reported preoperative predictors was not improved by the addition of preoperative anxiety measures. Although, for several reasons, it may be useful to be informed about the patient’s preoperative anxiety status, routine measurement of anxiety to predict the occurrence of PONV does not seem warranted if the other predictors are already considered.
There is evidence from other disciplines that anxiety disorders are associated with nausea in the general community (16). For a relationship between anxiety and nausea after anesthesia, however, no evidence was available. In a study of 51 unpremedicated children undergoing outpatient surgery and standardized general anesthesia, anxiety was shown to have no predictive value for the occurrence of PONV 24 hours after surgery (17). Also, Shafer et al. (18) showed that midazolam premedication significantly reduced preoperative anxiety levels in outpatients, without a parallel decrease in the incidence of PONV. The only data that support the association between anxiety and PONV come from a survey conducted by Quinn et al. (19), who questioned patients about their experience during the first 24 hours after anesthesia and surgery. This study, however, did not use a validated instrument to measure anxiety, and the anesthetic technique and surgical procedure were not standardized. Therefore, the evidence for a link between preoperative anxiety and PONV from this study and previous reports remains weak.
The results of this study confirm that sex, smoking status, and history of PONV or motion sickness independently contribute to the prediction of PONV. Furthermore, the predictive power of these established predictors can significantly be improved by adding the predictors age, anesthetic technique, and type of surgery. In this data set, the ROC area of a model with the predictors sex, history of PONV or motion sickness, smoking, and postoperative opioid analgesia was 0.67. The addition of age, type of surgery, and anesthetic technique, while at the same time eliminating postoperative opioid use, resulted in a significantly higher ROC area of 0.72. Multivariate analyses on our data did not confirm the independent predictive value of postoperative opioid analgesia and duration of anesthesia found in previous studies (2,4,5,7). One possible explanation is that the predictive information of these two predictors was already provided for by the type of surgery.
Evidence on type of surgery as an independent predictor is somewhat controversial. We found a significant association between lower abdominal (and middle ear) surgery and PONV, whereas this was not found by Cohen et al. (3). We did not find an independent predictive effect of laparoscopic surgery; this was in accordance with some previous studies (3,4) but conflicted with another study (20). For other types of surgery, no independent predictive ability for PONV has been found, and this is supported by our results. It should be noted that the number of subjects in some of our surgical categories was relatively small. A center effect can therefore not be excluded. Hence, the predictive effect of abdominal and middle ear surgery for the occurrence of PONV needs validation in future studies.
Risk scores seem to be the only objective method for a risk-based antiemetic strategy (21,22), because single predictors such as the type of surgery or the patient’s history of PONV alone do not sufficiently predict PONV. Several prediction rules for PONV have therefore been proposed. It should be noted that any prediction rule, including ours, performs better on the data from which it is developed than in another institution (over-optimism). For this reason, we estimated the amount of over-optimism of our rule by using bootstrapping techniques. We adjusted the rule for this over-optimism and calculated the ROC area that can be expected in future patients. After bootstrapping, our rule had a predictive accuracy of 0.70. Hence, 0.70 is the ROC area that can be expected in new patients. Because this was only 0.02 less than before bootstrapping, the rule seems quite robust. External validation studies in new patients may yield additional evidence for the robustness of our prediction rule across clinical settings. Admittedly, our rule, which includes 6 predictors, is slightly more complex than the previously developed 4-item risk score (2). Nevertheless, using the nomogram or score table presented in this article, the probability of PONV in individual patients can easily be estimated, particularly when electronic patient records are used.
In this study, patients received either inhaled anesthesia with isoflurane and nitrous oxide or IV anesthesia with propofol. There is evidence that omission of nitrous oxide decreases the incidence of PONV (23). Furthermore, in some countries, use of isoflurane has decreased in favor of sevoflurane or desflurane, which are frequently used without nitrous oxide. If the PONV incidence would be less frequent with these newer inhaled anesthetic techniques, the estimated OR (Table 2) for use of inhaled anesthesia might also be lower. We realize that changes in the practice of inhaled anesthesia might compromise the generalizability of the present prediction rule. However, several systematic reviews suggest that the PONV incidence with isoflurane does not substantially differ from that with other inhaled anesthetics (24,25). Furthermore, even if the incidence of the outcome were less frequent in future patient populations, this might affect the intercept (constant) and, thus, the calibration of the prediction model, but it would not affect the model’s discriminative ability. Therefore there is no reason to assume that the generalizability of the rule is affected by local preferences for sevoflurane or desflurane over isoflurane, with or without nitrous oxide.
To further appreciate these results, a few issues need to be addressed. First, although outpatients were included in the trial from which the data originated, for these analyses we deliberately restricted the study population to inpatients, because the PONV incidence in outpatients was substantially less frequent (34%) and because different types of surgery were performed (e.g., no abdominal surgery). Accordingly, our results should primarily be generalized to inpatients. It should be noted that, currently, no rules are available that were derived on both inpatients and outpatients. This is still a subject for future research, particularly given the increase of ambulatory surgery. Second, in this study, the PONV incidence was frequent (48%) but in accordance with previous studies (3,4). The frequent incidence found in this study may be partly explained by the frequent assessment of the outcome (4 times at the PACU and each hour at the ward versus at 2 and 24 hours after surgery in other studies) and the wide definition of PONV. However, additional analyses with separate end-points, i.e., nausea only at 24 hours and vomiting only at 24 hours, yielded the same results for anxiety and for the other predictors and had similar predictive accuracy. Finally, we used only outcome data from the early postoperative period (first 24 hours) to conform to previous studies on prediction of PONV. However, additional analyses with the later outcomes (PONV at 72 hours or 14 days) also yielded the same predictors and identical predictive accuracy.
In conclusion, high levels of preoperative anxiety are associated with the occurrence of PONV. However, because this association is weak, routine measurement of preoperative anxiety to aid in the prediction of PONV does not seem warranted, provided that the predictors sex, age, smoking, history of PONV or motion sickness, type of surgery, and anesthetic technique are already considered.
1. Watcha MF, White PF. Postoperative nausea and vomiting: its etiology, treatment and prevention. Anesthesiology 1992;77:162–84.
2. Apfel CC, Läärä E, Koivuranta M, et al. A simplified risk score for predicting postoperative nausea and vomiting. Anesthesiology 1999;91:693–700.
3. Cohen MM, Duncan PG, DeBoer DP, Tweed WA. The postoperative interview: assessing risk factors for nausea and vomiting. Anesth Analg 1994;78:7–16.
4. Koivuranta M, Läärä E, Snåre L, Alahuhta S. A survey of postoperative nausea and vomiting. Anaesthesia 1997;52:443–9.
5. Palazzo M, Evans R. Logistic regression analysis of fixed patient factors for postoperative sickness: a model for risk assessment. Br J Anaesth 1993;70:135–40.
6. Pierre S, Benais H, Pouymayou J. Apfel’s simplified score may favourably predict the risk of postoperative nausea and vomiting. Can J Anaesth 2002;49:237–42.
7. Toner CC, Broomhead CJ, Littlejohn IH, et al. Prediction of postoperative nausea and vomiting using a logistic regression model. Br J Anaesth 1996;76:347–51.
8. Visser K, Hassink EA, Bonsel GJ, et al. Randomized controlled trial of total intravenous anesthesia with propofol versus inhalation anesthesia with isoflurane-nitrous oxide: postoperative nausea and vomiting and economic analysis. Anesthesiology 2001;95:616–26.
9. Lerman J. Surgical and patient factors involved in postoperative nausea and vomiting. Br J Anaesth 1992;69(Suppl 1):S24–32.
10. Spielberger CD, Gorush RL, Lushene RE. State Trait Anxiety Inventory manual. Palo Alto, CA: Consulting Psychologists Press, 1970.
11. Moerman N, Van Dam FSAM, Muller MJ, Oosting H. The Amsterdam Preoperative Anxiety and Information Scale (APAIS). Anesth Analg 1996;82:445–51.
12. Hosmer D, Lemeshow S. Applied logistic regression. New York: Wiley, 1989.
13. Harrell FE, Lee KL, Mark DB. Multivariable prognostic models: issues in developing models, evaluating assumptions and adequacy and measuring and reducing errors. Stat Med 1996;15:361–87.
14. Steyerberg EW, Harrell FE, Borsboom GJJM, et al. Internal validation of predictive models: efficiency of some procedures for logistic regression analysis. J Clin Epidemiol 2001;54:774–81.
15. Efron B, Tibshirani R. An introduction to the bootstrap. New York: Chapman & Hall, 1993.
16. Haug TT, Mykletun A, Dalh AA. The prevalence of nausea in the community: psychological, social and somatic factors. Gen Hosp Psychiatry 2004;24:81–6.
17. Wang S-M, Kain ZN. Preoperative anxiety and postoperative nausea and vomiting in children: is there an association? Anesth Analg 2000;90:571–5.
18. Shafer A, White PF, Urquhart ML, Doze VA. Outpatient premedication: use of midazolam and opioid analgesics. Anesthesiology 1989;71:495–501.
19. Quinn AC, Brown JH, Wallace PG, Asbury AJ. Studies in postoperative sequelae: nausea and vomiting—still a problem. Anaesthesia 1994;49:62–5.
20. Apfel CC, Greim CA, Haubitz I, et al. A risk score to predict the probability of postoperative vomiting in adults. Acta Anaesthesiol Scand 1998;42:495–501.
21. White PF, Watcha MF. Postoperative nausea and vomiting: prophylaxis versus treatment. Anesth Analg 1999;89:1337–9.
22. Watcha MF. The cost-effective management of postoperative nausea and vomiting. Anesthesiology 2000;92:931–3.
23. Hartung J. Twenty-four of twenty-seven studies show a greater incidence of emesis associated with nitrous oxide than with alternative anesthetics. Anesth Analg 2004;83:114–6.
24. Sneyd JR, Carr A, Byrom WD, Bilski AJT. A meta-analysis of nausea and vomiting following maintenance of anaesthesia with propofol or inhalational agents. Eur J Anaesthesiol 2003;15:433–45.
25. Tramer MR, Moore RA, McQuay HJ. Propofol anesthesia and postoperative nausea and vomiting: quantitative systematic review of randomized controlled studies. Br J Anaesth 1997;78:247–55.