| JOURNAL HOME | CME HOME | THIS MONTH | PAST ISSUES | ETOC | COLLECTIONS |
| AUTHORS | REVIEWERS | EDITORIAL BOARD | FEEDBACK | RSS | HELP |
| ||||||||||||||
| ||||||||||||||
|
|
|
|
||||||||||||
|
||||||||||||||
From the *Department of Anaesthesiology, University Hospitals of Würzburg, Würzburg, Germany; Department of Anaesthesiology and Critical Care, University Hospitals of Marburg and Giessen GmbH, Campus Marburg, Germany; and Department of Anaesthesiology, Evangelischen Krankenhauses Oberhausen, Germany.
Address correspondence and reprint requests to Priv.-Doz. Dr. Peter Kranke, Department of Anaesthesiology, University Hospitals of Würzburg, D-97080 Würzburg, Germany. Address e-mail to peter.kranke{at}t-online.de.
Abstract
BACKGROUND: A score to predict postoperative vomiting (PV) in children (POVOC score) has recently been published but has not yet undergone an external validation.
METHODS: We studied 673 patients (age 0–16 yr) undergoing a variety of surgical procedures (but excluding strabismus surgery, one of the risk factors according to the POVOC score) using standardized anesthesia techniques without administering antiemetics. The patients were prospectively screened for PV in the postoperative period and these incidences were compared with the predicted risk for PV according to the POVOC score. The POVOC score was evaluated with respect to its ease of use, discrimination, and calibration.
RESULTS: Complete data to predict the risk for PV could be obtained in 95% of patients. The actual observed incidences of PV were 3.4, 11.6, 28.2, and 42.3% for the presence of 0, 1, 2, or 3 risk factors, resulting in a regression line with a slope of 0.78 and an offset of 2.37. The area under the receiver operating characteristic curve was 0.72 (95% CI: 0.68–0.76).
CONCLUSIONS: Using the POVOC score, PV in pediatric patients can be predicted with sufficient accuracy comparable to the results in adult patients, even if one of the risk factors is not applicable.
Recent advances regarding the safety of anesthesia and the concomitant reduction in mortality and major morbidity have shifted the focus of interest towards discomfort and dissatisfaction as crucial outcomes for anesthesia. The prevention of postoperative nausea postoperative vomiting and their combination (PONV) can help to achieve excellent patient satisfaction (1). However, this has only been proven for patients at increased risk, e.g., female patients or those who suffered from PONV during previous anesthetics (2). Despite intensive research on prophylactic interventions for PONV, there is no universally effective antiemetic substance that is able to abolish all PONV. Therefore, multimodal approaches have been advocated for both adults and children (3). Since efficiency in low-risk patients is restricted because of a limited absolute risk reduction, it has been proposed that antiemetic prophylaxis should be restricted to high-risk patients (4). Although most of the available antiemetics are generally well tolerated, this approach enables minimizing potential harm due to the side effects of antiemetics. Although there are numerous risk scores available for adults, only one published risk score exclusively focused on a pediatric patient population (Postoperative Vomiting in Children [POVOC] score) (5). This is important because it has been shown that transferring risk scores for adults to a pediatric population yields no meaningful predictive conclusion (6). The POVOC score is a simplified risk score considering the following clinical risk factors: Duration of surgery 
30 min, age 3 yr, strabismus surgery, and a positive history of PV in the child, or history of PONV in parents or siblings. Depending on the presence of none, 1, 2, 3, or 4 risk factors the estimated incidence for PV in the pediatric patient is 9, 10, 30, 55, and 70%, respectively. The POVOC score has not yet undergone an external validation (7). However, an external validation should be performed before any recommendations can be made for routine clinical use (8). Strabismus surgery is not considered a risk factor in many institutions. Additionally, in the last few years, the frequency of strabismus surgery has declined (9). Thus, the aim of this prospective evaluation study was to test the ease of application of the POVOC score as well as the discriminating power and the calibration characteristics of the POVOC score in children undergoing various types of surgical interventions and in the absence of the factor "strabismus surgery."
METHODS
With approval of the hospital and after parental consent, 673 pediatric patients were prospectively studied at a community hospital for the occurrence of PV and associated risk factors during the first 24 postoperative hours.
Biometric data and potential risk factors were obtained from the parents during the preanesthesia assessment.
To enable a generalization of the results and to ensure external validity, a wide range of patients undergoing various types of surgery were considered eligible, as follows: children between 0 and 16 yr of age, with an ASA I to III status, and undergoing an elective or unscheduled procedure, either as outpatients or inpatients. No major restrictions with respect to the standard procedure for general anesthesia were applied except that inhaled anesthetics should be used as maintenance drugs.
Using the preoperatively assessed risk factors and the duration of the surgical procedure for each patient, the individual risk for PV was calculated before the end of anesthesia (5).
After an appropriate fasting period, all patients received a balanced general anesthesia. This consisted of:
7 months If feasible, perioperative analgesia was supplemented using a regional nerve block. The methods applied were dorsal penile nerve block, ilioinguinal nerve block and caudal analgesia, each using ropivacaine (0.375%) or bupivacaine (0.25%).
Postoperative systemic analgesia was started intraoperatively using rectal paracetamol (25 mg/kg) in combination with codeine, diclofenac (1 mg/kg), ibuprofen (10 mg/kg), or IV metamizol (15 mg/kg).
In case of expected moderate to severe pain, IV piritramide, starting before the end of anesthesia, was used to titrate analgesia until a satisfactory pain level was achieved.
PV was assessed in the postanesthetic care unit by a trained investigator. Twenty-four hours later, the children (if feasible) and their parents were asked about the occurrence of PV. In addition, relevant medical records were screened and the nursing staff was queried in order not to miss emetic events. Data from outpatients were obtained using a standardized and structured telephone interview 24 h after tracheal extubation.
Rescue antiemetic therapy consisted of rectal dimenhydrinate (first-line) and IV granisetron (second-line). The following endpoints were analyzed:
Practicability (User-Friendliness and Completeness of Data)
We aimed to descriptively assess the user-friendliness and completeness of data with respect to relevant risk factors. For this purpose, the time to record all relevant risk factors for the POVOC score was assessed. Furthermore, completeness of this preoperative risk-assessment was recorded.
Discrimination Characteristics of the POVOC Score
For each patient, the individual risk for PV according to the POVOC score was calculated and correlated to the true occurrence of PV. Then, a receiver operating characteristic (ROC) curve was constructed. This curve correlates the true- and false-positive rates ("sensitivity" and "1 minus specificity," respectively) for a series of cut-off points for a test. Here the cut-off point is a certain number of risk factors. If a patient exceeds the cut-off point, then this person is classified as a "risk patient." The area under this ROC curve (AUC) represents the probability that a child developing PV has a higher POVOC score than a child without PV (10). Thus, the discrimination properties represent the ability of a scoring system to predict PV in an individual patient. The AUC, including its 95% confidence interval (CI), was calculated using the statistical package JMP 5.1 (SAS Institute Inc., SAS Campus Drive, Cary, NC).
A significant difference, as opposed to a random guess, was assumed if the 95% CI did not include the value 0.5. This is the pre-hoc likelihood for a random guess.
The hypothesis that the discrimination of the POVOC score is as good in the present patient sample lacking one predictor (strabismus surgery) compared with the original publication (5) was rejected if the 95% CI of the AUC for PV did not include the value from the POVOC study (AUC = 0.73).
Calibration Characteristics of the POVOC Score
To test the calibration characteristics, the predicted incidences according to the POVOC score were plotted against the actual observed incidences for PV. The number of cases in each of the risk groups was used to perform a weighted regression and to calculate the underlying linear equation, which represents the calibration characteristics of a given score for a specific population. Thus, the calibration curve represents how well the predicted incidences coincide with the actual observed incidences and whether the occurrence of PV is systematically over- or under-rated. Since the calibration characteristic makes a statement with respect to real incidences, it is the decisive measure of when the threshold of antiemetic prophylaxis should be determined.
RESULTS
Period of Data Collection and Preoperative Assessment
Between September 1, 2004, and March 31, 2005, 673 patients had a surgical procedure at the department of pediatric surgery. Of these patients, 149 had to be excluded from the analysis. Sixty-six patients had to be excluded because of extended intraabdominal surgery requiring the postoperative use of a nasogastric tube and/or postoperative ventilatory support. Preexisting nausea or vomiting (n = 26) and the use of drugs with known antiemetic potential in the perioperative period (n = 15), as well as incomplete patient data (n = 42), were the reasons for exclusion in other patients, so that 524 patients remained for final analysis.
Patient Characteristics and Surgical Procedures
Characteristics of included patients are given in Table 1 and 2. Demographic data, as well as duration of anesthesia and the surgical procedure, were comparable to the initial data set in which the POVOC score was developed (Table 2). Surgical procedures performed are shown in Table 3.
|
|
|
Incidence of PV
The overall incidence of PV (at least one emetic episode) during the 24 h observation period was 20.2% (106 of 524 patients). The first emetic episode predominantly took place during the first 4 h after anesthesia. A first emetic episode during the remaining observation period was only observed in 26 of 524 patients (24.5%).
Practicability
The individual risk factors assessed preoperatively are shown in Table 1. The acquisition of relevant data for the POVOC score was easy to perform, straightforward, not associated with ambiguity nor time-consuming (<1 min per patient). Complete data could be obtained for 95% of patients. Problems with the full assessment of risk factors occurred, for instance, if the history of PONV could not be assessed for the parents, i.e., if one parent was absent during the preanesthesia visit or in the case of foster parents.
Discrimination
The corresponding ROC curve for the POVOC score (Fig. 1) is associated with an AUC of 0.72 (95% CI: 0.68–0.76) and thus is identical with the AUC described in the POVOC-trial (AUC = 0.72; 95% CI: 0.68–0.77).
|
The 95% CI did not include the value of 0.5 and was thus significantly better than a random guess.
Calibration
The predicted incidences of PV as calculated with the POVOC score showed a good correlation with the actual observed incidences in the investigated patient population, being 3.4, 11.6, 28.2, and 42.3% for the presence of 0, 1, 2, or 3 risk factors (Table 4). With the exception of the actual observed incidence of PV associated with the presence of three risk factors, the 95% CIs included the originally reported incidences of the POVOC trial. The calibration curve with the actual observed incidences plotted against the predicted incidences with the POVOC score results in a regression line with a slope of 0.78 and an offset of 2.37 (Fig. 2). The corresponding correlation coefficient R2 with a value of 0.94 showed a pronounced positive correlation.
|
|
DISCUSSION
Although internally validated using a randomly chosen evaluation dataset the POVOC score has not been evaluated in other patient populations. Further, it remained speculative whether the POVOC score could maintain its predictive properties in the absence of the risk factor "strabismus surgery," a patient subgroup that is often not available in a common pediatric surgery setting and may further decline in the future (9).
The presented evaluation indicates that even in the absence of strabismus surgery, the POVOC score allows estimation of the risk of PV in pediatric surgical patients. The discriminative properties as shown by the area under the ROC curve, as well as the calibration characteristics, indicate that it is able to predict PV in children with acceptable accuracy and that these predicted incidences fit well with the actually observed incidences. However, although the POVOC score was better in predicting PV than random guesses, it may not accurately predict PONV in children of all races undergoing different surgery in different settings using different anesthetics. In order to be sure that the POVOC score is valid for all types of surgery, a further evaluation in a large-scale survey with rigorous assessment of emetic symptoms may be helpful.
Two further questions arise from this analysis. First, there are reasons to suppose that, despite the evaluation in 1781 patients and five independent institutions, there is still the chance that in some settings the performance of the POVOC score is inferior compared with the performance indices currently available (AUC of the ROC curve: 0.72 and above).
Such a scenario has been observed in the "career" of risk scores for adults. In this patient population, the scores developed by Koivuranta et al. (11) and Apfel et al. (12) appear to be the most popular because of their simplified calculation. The first evaluation trials uniformly reported that these risk scores developed in one center may be transferred to other settings without losing their predictive properties (13–15). However, recently there have been reports suggesting that customizing risk scores for specific settings may be required to maintain their accuracy (16,17). Thus, in some institutions, the developed scoring tools may fail to predict the risk for PV with acceptable accuracy. On the other hand, not all validation trials use the established and suggested rigor in the assessment of symptoms (18) and some rely on anesthesia management information technology, which may skew the results and hamper conclusions (16).
Considered from another point of view, the fact that scoring systems deliver conflicting values is not so surprising after all. For instance, it has been demonstrated that using an artificial neural network has no substantial added value when compared with conventional risk scores for the prediction of PONV (19), whereas another trial claimed that the opposite is true (20). In addition, simulation studies have indicated that providing that risk factors are associated with comparable odds ratios, they may be interchangeable to a certain degree and an observed difference can be reproduced (21). In the event that an institution intends to implement an antiemetic protocol based on prediction tools, an appropriate assessment under routine conditions needs to be instituted. And, using such an approach, many optimistic statements suggest that scoring systems may help in daily clinical practice (22–25).
Second, the most pertinent question is whether the outcome in terms of the incidence of PV (or PONV) can be improved or efficiency can be enhanced by using risk scores in daily practice. The optimistic assumption would be that the efficiency can be improved if anesthesiologists adhere to the POVOC score when deciding which patient needs prophylaxis and to which extent (single versus multimodal prophylaxis). Immediately after the introduction of risk scores for PONV, it was hoped that the efficacy and efficiency of antiemetic prophylaxis could be improved. This was based on the understanding that focusing the resources on the patients who actually need prophylaxis might save resources. However, there is good evidence that the overall benefit of a risk-adopted antiemetic approach is highly dependent on the distribution of risk factors and thus on the patient population in a specific setting. For adults, using the simplified score developed by Apfel et al. (26), it could be demonstrated that the golden rule "the more antiemetics the lower the institutional incidence" cannot be overruled. However, in some settings, in populations with an equal distribution of risk classes, for instance, a score-based stepwise antiemetic approach may be useful to obtain nearly the same results (i.e., the same low incidence of PV) with considerably fewer resources and fewer patients exposed to potential side effects compared to a general prevention with two antiemetics (26,27).
In summary, there is good evidence that the risk factors included in the validated POVOC score are among the most reliable and relevant (in terms of its prevalence) risk factors for PV in children that are currently available. Therefore, the use of the POVOC score can be recommended to guide antiemetic prophylaxis in a setting where no routine prevention seems to be advisable. The risk factors of the POVOC score are valid across different patient populations and in various settings for patients undergoing a large variety of surgical procedures. Therefore, a more widespread use seems to be justified.
Footnotes
Accepted for publication August 21, 2007.
REFERENCES
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
