Comparison of surgical site and patient s history with a simplified risk score for the prediction of postoperative nausea and vomiting

Transcription

1 Comparison of surgical site and patient s history with a simplified risk score for the prediction of postoperative nausea and vomiting C. C. Apfel, 1,2 P. Kranke 2 and L. H. J. Eberhart 3 1 Department of Anaesthesiology and Perioperative Medicine, and OUTCOMES RESEARCH (TM) Institute, University of Louisville, 501 E Broadway, Suite 210, Louisville, KY, USA 2 Department of Anaesthesiology, Julius-Maximilians-University, Oberduerrbacher Str. 6, D Wuerzburg, Germany 3 Consultant, Department of Anaesthesiology, Philipps-University of Marburg, Baldingerstraße, D Marburg, Germany Summary Although site of surgery and previous occurrence of postoperative nausea and vomiting are often used to decide whether prophylactic anti-emetic drugs are indicated, the value of these predictors is unclear. We compared these two risk factors against a simplified four-factor risk score. We analysed data from 1566 adult inpatients who received balanced anaesthesia without prophylactic antiemetics. Sensitivity, specificity, predictive value and area under the receiver operating characteristic curve were used to quantify predictive properties. Nausea and vomiting occurred in 600 (38.3%) patients within 24 h. Sensitivity and specificity were, respectively, 47% and 59% for surgical site; 47% and 70% for history of postoperative nausea and vomiting; and 58% and 70% for risk score with three or more factors. The area under the curve for surgical site was 0.53 (95% CI ); that for patient s history was 0.58 (95% CI ) while for risk score it was 0.68 (95% CI ; P < 0.001). Prediction using surgical site or patient s history alone was poor while the simplified risk score provided clinically useful sensitivity and specificity. Keywords Postoperative nausea and vomiting.... Correspondence to: Dr C. C. Apfel christian.apfel@louisville.edu Accepted: 25 May 2004 Postoperative nausea and vomiting (PONV) occurs in 25 30% of surgical patients and therefore continues to be a concern for anaesthetists [1,2]. In order to diminish this problem, innumerable trials have investigated prophylactic anti-emetic drugs. Quantitative systematic reviews reveal that single anti-emetics given prophylactically may reduce the incidence by approximately a third [3 5]. Considering this limited efficacy, the wisdom of routine prophylactic administration of a single anti-emetic to all patients is debatable [6]. However, a multimodal approach is much more effective, and such a prophylactic strategy may be justified in patients with high risk for PONV [7,8]. Since the number-needed-to-treat (NNT) to prevent one patient from suffering PONV is unacceptably high when the control event rate is low, the use of a prophylactic anti-emetic strategy should depend on the patient s risk for PONV [6,9]. Since 1993, several risk scores have been developed to predict the probability of PONV based on logistic regression analysis [10 14]. While most scores use a rather complicated formula, two of them apply a simplified method by considering how many of four or five predictors the patient has [11,14]. Simplified risk scores are apparently more robust and have better external validity than the more complicated scores [15,16]. A recent study, though, has questioned the reliability of risk scores since the correlation between them was poor [17]. However, this is not surprising, given that the simplified risk scores appear to have higher external validity [18]. Despite these recent developments, many clinicians still use a single predictor such as the surgical location or the patient s history of PONV to decide whether to give prophylactic anti-emetics. The argument against this approach is the concern that sensitivity and specificity 1078 Ó 2004 Blackwell Publishing Ltd

2 C. C. Apfel et al. Æ Prediction of postoperative nausea and vomiting using a risk score should be better. However, this has never been formally tested. Therefore, we compared the usefulness of two single-risk factor models often used in clinical practice with a well-established simplified risk score [14]. Methods This study was approved by the Human Studies Committee of Ulm. Patients gave informed consent before being recruited into the study. The prospectively collected data used for this analysis were obtained from a survey of 1566 adult inpatients [15]. All patients received balanced anaesthesia with volatile anaesthetics in nitrous oxide and oxygen (2 : 1). None received any prophylactic drugs with potential anti-emetic properties. The most frequently used anaesthesia technique (in 60% of cases) consisted of propofol, fentanyl and vecuronium for induction, followed by desflurane in nitrous oxide and oxygen for maintenance with additional doses of fentanyl as required. Patients who experienced any degree of nausea, an emetic episode, or both within the first 24 h postoperatively were classified as having PONV. Two different single risk factors were investigated as predictors of PONV: surgical location and the patient s previous history of PONV. Ophthalmological, ear, nose and throat (ENT), abdominal, and gynaecological procedures were classified as high-risk operations while orthopaedic and other operations were classified as lowrisk. For each surgical location, the incidence of PONV was calculated and taken as the individual risk of PONV for that surgery. For each risk factor, the incidence of PONV, the relative risk, and the odds ratio (adjusted using multiple logistic regression analysis) and 95% CI were calculated. In addition, for each factor the patients were divided as to whether they had PONV or not, and the sensitivity, specificity, and predictive value were calculated. Receiver operating characteristic (ROC) curves were then drawn for surgical site and previous history of PONV. In brief, ROC curves describe the relationship between sensitivity and specificity for all possible decision criteria. The area under the curve (AUC) gives a global estimate of the model to predict the outcome independent of the decision criterion so that different models can be easily compared. These results were compared against our previously developed simplified risk score (but on data that have never been used for the construction of a risk score [14]). The four strongest independent risk factors in that model were female gender, non-smoking status, history of PONV or motion sickness, and the use of postoperative opioids. The chance that a patient will develop PONV increases with each risk factor: approximately 10% if no risk factor is present; 20% for one risk factor; 40% for two risk factors, 60% for three and 80% for and all four risk factors [14]. A significant difference was defined as P < 0.05 or the comparable values were beyond the 95% CI of the AUC. The data were recorded in Access 2000; statistical analyses were performed using Medcalc 4.20 and the program Confidence Interval Analysis 2.0 from Altman, Machin, Bryant and Gardner. Results Six hundred out of 1566 patients (38.3%) suffered from PONV (Table 1). The incidence of PONV, relative risk, odds ratio and predictive characteristics of surgical site, history of PONV and the simplified risk score are shown in Table 2. Sensitivity and specificity were 47% and 59% for surgical site; 47% and 70% for history of postoperative nausea and vomiting; and 58% and 70% for risk score with three or more factors, respectively. The AUC for the Table 1 Patients characteristics and incidence of postoperative nausea and vomiting (PONV), with univariate relative risk and adjusted odds ratio for PONV (n = 1566). Values are number (proportion) or number (95% CI). Incidence of PONV Relative risk* Odds ratio (adjusted) Female 699 (44.6%) 375 (53.6%) 2.07 ( ) 2.73 ( ) Non-smoker 1098 (70.1%) 475 (43.3%) 1.62 ( ) 1.76 ( ) History of PONV 569 (36.3%) 280 (49.2%) 1.53 ( ) 1.79 ( ) Postoperative opioids 1078 (68.8%) 434 (40.3%) 1.18 ( ) 1.50 ( ) Ophthalmological 75 (4.8%) 31 (41.3%) 1.08 ( ) 1.45 ( ) Ear, nose and throat 93 (5.9%) 37 (39.8%) 1.04 ( ) 1.60 ( ) Abdominal 127 (8.1%) 62 (48.8%) 1.31 ( ) 2.00 ( ) Cholecystectomy 42 (2.7%) 29 (69%) 1.84 ( ) 3.23 ( ) Gynaecological 52 (3.3%) 31 (59.6%) 1.58 ( ) 1.84 ( ) Orthopaedic 885 (56.5%) 317 (35.8%) 0.86 ( ) 1.30 ( ) Others 292 (18.6%) 93 (31.8%) 0.80 ( ) Reference group All patients 1566 (100%) 600 (38.3%) 0.12 (constant) *from univariate analysis; from multivariate analysis; excluding laparoscopy; usually laparoscopic Ó 2004 Blackwell Publishing Ltd 1079

3 C. C. Apfel et al. Æ Prediction of postoperative nausea and vomiting Anaesthesia, 2004, 59, pages Table 2 Incidence of postoperative nausea and vomiting (PONV), relative risk, odds ratio and predictive characteristics of surgical site, history of PONV and the simplified risk score [14] in 1566 patients undergoing anaesthesia. The same measures based purely on random prediction (i.e. 50% chance) are also given, for comparison. Values are number or proportion. Random operations* High-risk History of PONV Risk score = 2 Risk score = 3 Risk score = 4 No. of patients with the predictor and PONV (a) No. of patients with the predictor but without PONV (b) Incidence; a (a + b) 38.3% 41.6% 49.2% 45.6% 54.5% 64.2% No. of patients without the predictor and PONV (c) No. of patients without the predictor but without PONV (d) Incidence; c (c + d) 38.3% 35.8% 32.1% 15.7% 27.2% 34.7% Relative risk; (a (a + b)) (c (c + d)) Odds ratio (unadjusted); (a b) (c d) Sensitivity; a (a + c) 50.0% 47.2% 46.7% 90.0% 58.0% 20.7% Specificity; d (b + d) 50.0% 58.8% 70.1% 33.2% 70.0% 92.9% Positive predictive value; a (a + b) 38.3% 41.6% 49.2% 45.6% 54.5% 64.2% Negative predictive value; d (c + d) 61.7% 64.2% 67.9% 84.3% 72.8% 65.3% Overall predictive value; (a + d) (a + b + c + d) 50.0% 54.3% 61.1% 55.0% 65.4% 65.2% * High risk operations are those listed in Table 1 except orthopaedic and others. Sensitivity simplified four-factor risk score (0.68 [95% CI ]) was significantly greater than those for the single predictors (surgical site: 0.53 [95% CI ]; history of PONV 0.58 (95% CI ; P < 0.001; Fig. 1) Specificity Figure 1 Receiver operating characteristic curves of surgical site (n), previous history of postoperative nausea and vomiting (s) or simplified risk score (d) [14] for predicting PONV. Areas under the curve are 0.53 (95% CI ) for surgical site, 0.58 (95% CI ) for history of PONV and 0.68 (95% CI ) for the simplified risk score (P <0.001 compared with the single predictors) Discussion We found that the site of surgery and the patient s history of PONV were not good predictors of PONV. The simplified risk score had a better predictive value, but even it was only modestly successful. It is well known and generally accepted that certain sites of surgery are associated with an increased incidence of PONV [1, 20]. This observation led to the assumption that this must be caused by certain pathophysiological pathways[ 19,20]. However, as PONV is a multifactorial problem, the relative impact of a single factor can only be quantified by multivariate analysis to account for other confounding factors. This was taken into consideration by calculating the odds ratio using multiple logistic regression analysis (adjusted odds ratio). Further, the only published study to date with sufficient power to detect two factor interactions did not provide any evidence that the efficacy of anti-emetic drugs depended on the type of surgery, casting doubts upon the asserted pathophysiological hypothesis [21]. Very recent other work further suggest that the type of surgery is not a reliable predictor for PONV [22]. Ophthalmological procedures appear to be highly emetogenic since strabismus surgery in children is usually associated with a very high incidence of PONV [23], especially Faden s procedure [24], and van den Berg and colleagues hypothesised that an oculo-emetic reflex is responsible [25]. In contrast, low incidences have been reported in adults undergoing strabismus surgery [26] and multivariate analysis in another trial did not support a clinically relevant independent impact of ophthalmo Ó 2004 Blackwell Publishing Ltd

4 C. C. Apfel et al. Æ Prediction of postoperative nausea and vomiting logical procedures on PONV [27]. It is therefore not surprising that we found no correlation between ophthalmological surgery and PONV in adults. There are only a few reported studies on the incidence of PONV after ENT surgery [12,19]. Both tympanoplasties [12] and adenotonsillectomies in children [28] (but not in adults [12]) may be associated with an increased risk of PONV. However, ENT surgery did not appear to be a strong and reliable predictor for PONV in the current study. In the population studied here, abdominal (non-laparoscopic) surgery, gynaecological and laparoscopic surgery (e.g. cholecystectomy) were associated with an increased risk of PONV, even when corrected for other confounding factors. The literature for these sites of surgery is conflicting with an increased risk [13,29], no relevant impact [11,29], and even a lower risk [27] reported. A similar discrepancy can be observed with orthopaedic surgery. While it is generally assumed to be associated with a low incidence of PONV [19], a normal incidence in the range of 30% was found in a validation study [30] and a fold increased risk has also been described [13]. Taking these discrepancies into consideration, it seems questionable to base a clinical judgement as to whether a patient is likely to suffer PONV on the site of surgery. This is corroborated by the low discriminating power as expressed by the small AUC (Fig. 1). Using the patient s history of PONV led to slightly better prediction. Although this risk factor is frequently used in clinical practice to estimate the patient s risk of PONV, the prediction is far from perfect. The main reason is that the adjusted odds ratio of 1.8 is not very strong (i.e. patients who have not suffered from PONV previously might have it in the future). In addition, patients who have never had anaesthesia were classified as never having had PONV. We have previously analysed whether considering the number of times a patient underwent anaesthesia with or without PONV provides better prediction. Despite including data from over 1000 patients, we did not find a significant improvement (unpublished calculations using data from [12]). The best sensitivity (90%) was achieved by the simplified risk score when patients with at least two risk factors were classified as high risk, but at the price of a low specificity (33%). If at least three risk factors were present, the sensitivity was about 10% higher compared to the single predictors while still having a specificity of about 70%. If only patients with 4 risk factors were taken into account, the sensitivity was low but the specificity was > 90%. Thus, if patients are only to be classified into groups of low or high risk, the best overall sensitivity and specificity will be obtained by selecting patients with at least three risk factors of the simplified risk score. If either a high sensitivity or specificity is aimed for, the simplified score with at least two or four risk factors, respectively, seems reasonable. This is helpful if three or more risk classes are intended [6,9]. For example, no prophylactic anti-emetic drugs may be needed for patients are at very low risk for PONV. If at least two risk factors are present (i.e. the risk for PONV is in the range of 40%), lowering the baseline risk by avoiding the emetogenic impact of volatile anaesthetics and nitrous oxide or applying a single anti-emetic would result in a sensitivity of > 90% [21,31]. If three or four risk factors are present (i.e. the risk for PONV is approximately 60% to 80%) a multimodal approach (lowering the baseline risk in conjunction with prophylactic anti-emetics) may be preferable, and the high specificity of > 90% would justify these efforts [7,8]. It has recently been shown in a cancer centre with gynaecological emphasis that this risk-dependent approach was able to lower the initial institutional incidence of 49% down to 14% [16,32]. This advantage of considering the different risk levels is reflected by the larger AUC of 0.68 of the simplified risk score compared to a single predictor model (Fig. 1). In addition, it allows quantification of the patient s risk for PONV. Acknowledgements The authors would like to thank Dr Daniel Sessler for his constructive and helpful comments and Dr Nancy Alsip for her very efficient editorial assistance, both from the Outcomes ResearchÔ Institute, University of Louisville. The study was performed without any funding from pharmaceutical companies or any other third party sources. References 1 Watcha MF, White PF. Postoperative nausea and vomiting. Its etiology, treatment, and prevention. Anesthesiology 1992; 77: Kovac AL. Prevention and treatment of postoperative nausea and vomiting. Drugs 2000; 59: Tramer 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: Henzi I, Walder B, Tramer MR. Dexamethasone for the prevention of postoperative nausea and vomiting: a quantitative systematic review. Anesthesia and Analgesia 2000; 90: Kranke P, Morin AM, Roewer N, Eberhart LH. Dimenhydrinate for prophylaxis of postoperative nausea and vomiting: a meta-analysis of randomized controlled trials. Acta Anaesthesiologica Scandinavica 2002; 46: Gan TJ, Meyer T, Apfel CC, et al. Consensus guidelines for the management of postoperative nausea and vomiting. Anesthesia and Analgesia 2003; 97: Ó 2004 Blackwell Publishing Ltd 1081

5 C. C. Apfel et al. Æ Prediction of postoperative nausea and vomiting Anaesthesia, 2004, 59, pages Scuderi PE, James RL, Harris L, Mims GR III. Multimodal antiemetic management prevents early postoperative vomiting after outpatient laparoscopy. Anesthesia and Analgesia 2000; 91: Eberhart LHJ, Mauch M, Morin AM, Wulf H, Geldner G. Impact of a multimodal anti-emetic prophylaxis on patient satisfaction in high-risk patients for postoperative nausea and vomiting. Anaesthesia 2002; 57: Watcha MF. The cost-effective management of postoperative nausea and vomiting. Anesthesiology 2000; 92: Palazzo M, Evans R. Logistic regression analysis of fixed patient factors for postoperative sickness: a model for risk assessment. British Journal of Anaesthesia 1993; 70: Koivuranta M, Laara E, Snare L, Alahuhta S. A survey of postoperative nausea and vomiting. Anaesthesia 1997; 52: Apfel CC, Greim CA, Haubitz I, et al. A risk score to predict the probability of postoperative vomiting in adults. Acta Anaesthesiologica Scandinavica 1998; 42: Sinclair DR, Chung F, Mezei G. Can postoperative nausea and vomiting be predicted? Anesthesiology 1999; 91: Apfel CC, Laara 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: Apfel CC, Kranke P, Eberhart LHJ, Roos IA, Roewer N. A comparison of predicting models for postoperative nausea and vomiting. British Journal of Anaesthesia 2002; 88: Pierre S, Benais H, Pouymayou J. Apfel s simplified score may favourably predict the risk of postoperative nausea and vomiting. Canadian Journal of Anesthesia 2002; 49: Thomas R, Jones NA, Strike P. The value of risk scores for predicting postoperative nausea and vomiting when used to compare patient groups in a randomised controlled trial. Anaesthesia 2002; 57: Apfel CC, Koivuranta M, Sweeney B, Thomas R, Jones N, Strike P. Study of postoperative nausea and vomiting: recommending risk models for group comparisons. Anaesthesia 2003; 58: Lerman J. Surgical and patient factors involved in postoperative nausea and vomiting. British Journal of Anaesthesia 1992; 69: 24S 32S. 20 Andrews PLR. Physiology of nausea and vomiting. British Journal of Anaesthesia 1992; 69: 2S 19S. 21 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. British Journal of Anaesthesia 2002; 88: Apfel C, Korttila K, Abdalla M, et al. A factorial trial of six interventions for the prevention of postoperative nausea and vomiting. New England Journal of Medicine 2004; 350(24): Sadhasivam S, Shende D, Madan R. Prophylactic ondansetron in prevention of postoperative nausea and vomiting following pediatric strabismus surgery: a dose response study. Anesthesiology 2000; 92: Rusch D, Happe W, Wulf H. Postoperative Übelkeit und postoperatives Erbrechen nach Strabismuschirurgie bei Kindern. Inhalationsanästhesie mit Sevofluran Lachgas im Vergleich zu intravenöser Anästhesie mit Propofol Remifentanil. Der Anaesthesist 1999; 48: van den Berg AA, Lambourne A, Clyburn PA. The oculoemetic reflex. A rationalisation of postophthalmic anaesthesia vomiting. Anaesthesia 1989; 44: Tramer MR, Fuchs-Buder T, Sansonetti A, Rifat K. Low incidence of the oculocardiac reflex and postoperative nausea and vomiting in adults undergoing strabismus surgery. Canadian Journal of Anaesthesia 1997; 44: Apfel CC, Greim CA, Haubitz I, et al. The discriminating power of a risk score for postoperative vomiting in adults undergoing various types of surgery. Acta Anaesthesiologica Scandinavica 1998; 42: Splinter W, Roberts DJ. Prophylaxis for vomiting by children after tonsillectomy: dexamethasone versus perphenazine. Anesthesia and Analgesia 1997; 85: Cohen MM, Duncan PG, DeBoer DP, Tweed WA. The postoperative interview: assessing risk factors for nausea and vomiting. Anesthesia and Analgesia 1994; 78: Eberhart LH, Hogel J, Seeling W, Staack AM, Geldner G, Georgieff M. Evaluation of three risk scores to predict postoperative nausea and vomiting. Acta Anaesthesiologica Scandinavica 2000; 44: Divatia JV, Vaidya JS, Badwe RA, Hawaldar RW. Omission of nitrous oxide during anesthesia reduces the incidence of postoperative nausea and vomiting. A Meta-Analysis. Anesthesiology 1996; 85: Pierre S, Corno G, Benais H, Apfel CC. A risk scoredependent antiemetic approach effectively reduces postoperative nausea and vomiting a continuous quality improvement initiative. Canadian Journal of Anesthesia 2004; 51: Ó 2004 Blackwell Publishing Ltd