Randomized Controlled Trial of Total Intravenous Anesthesia with Propofol versus  Inhalation Anesthesia with Isoflurane–Nitrous Oxide: Postoperative Nausea and Vomiting and Economic Analysis

Click on the links below to access all the ArticlePlus for this article.

Please note that ArticlePlus files may launch a viewer application outside of your web browser.

• https://links.lww.com/ALN/A102

SEVERAL studies have suggested that total intravenous anesthesia (TIVA) with propofol reduces the incidence of postoperative nausea and vomiting (PONV) and results in shorter emergence times. 1–5However, a metaanalysis showed that most studies were smaller, did not have follow-up beyond 6 h postoperatively, and were often sponsored by industry. Results were difficult to combine as a result of heterogeneous definitions of PONV. 6,7 

At present, propofol TIVA is more expensive than inhalational anesthesia with isoflurane and nitrous oxide (N²O). However, some investigators have suggested that TIVA could be cost-effective because the costs of treating PONV and of increased recovery room stay after inhalational anesthesia offset the additional drug acquisition costs of propofol TIVA. 8,9 

We compared the incidence of PONV up to 72 h postoperatively in a large group of unselected elective surgical inpatients and outpatients who were randomized to receive either inhalational anesthesia with isoflurane–N²O or TIVA with propofol–air. The primary hypothesis was that propofol TIVA would reduce the incidence of PONV compared with a conventional inhalation anesthetic technique. In addition, we hypothesized that the results might reveal subgroups with a high baseline risk for PONV, such as females and patients undergoing certain types of surgery, who would benefit more from TIVA than patients with a low baseline risk. This would allow identification of subgroups for whom TIVA could be especially advantageous. In addition, we performed an economic evaluation of the two techniques, testing the hypothesis that propofol TIVA would be an economically viable alternative to isoflurane in both inpatients and outpatients undergoing a broad array of surgical procedures. Because cost–benefit analysis requires that a dollar value be placed on PONV, and these data are not available, TIVA was considered economically viable if the additional drug acquisition cost for TIVA was balanced by the financial gains from reduced recovery time.

The study was conducted at the Academic Medical Center of the University of Amsterdam and was approved by its institutional medical ethics committee. All patients gave written informed consent before being included in the study.

From April 1997 to January 1999, 1,447 inpatients and 563 outpatients scheduled to undergo elective surgery with general anesthesia were enrolled in the study. Exclusion criteria were emergency, cardiac, or intracranial surgery; American Society of Anesthesiologists physical status classification greater than III; age less than 18 yr or more than 80 yr; (possible) pregnancy; renal or liver disease precluding use of either anesthetic tech-nique; use of antiemetic or proemetic medication in the 2 weeks before surgery; body weight greater than 120 kg; previous enrollment in the same study; and insufficient command of the Dutch language. 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.

The two anesthetic regimens consisted of (1) inhalational anesthesia with isoflurane and 60% N²O in oxygen (isoflurane group) for maintenance of anesthesia after induction with thiopental (inpatients) or propofol (outpatients), and (2) TIVA with propofol and air–oxygen for anesthesia induction and maintenance (TIVA group). Propofol dosage and isoflurane concentration were determined by the attending anesthesiologist. Prophylactic antiemetics were not permitted. Anesthesiologists were free to choose the type and dose of muscle relaxants and (opioid) analgesics as well as other drugs for supplementation of anesthesia, as deemed necessary. Postoperative analgesics were given as ordered by the responsible anesthesiologist. Administration of antiemetic therapy was the responsibility of postanesthesia care unit (PACU) and day care unit (DCU) nurses. The protocol prescribed intravenous metoclopramide (0.15 mg/kg) as the primary antiemetic therapy, followed by intravenous droperidol (15 μg/kg) if necessary.

Patients were randomized (unstratified) to either isoflurane or TIVA by an electronic randomization program running on a protected server. Medication for both anesthetic techniques was brought into the operating room on a covered tray, concealing the randomization from the attending anesthesia staff until directly before induction of anesthesia.

Personnel in the PACU and DCU, including the physician on duty, remained blinded to the randomization. Intravenous lines and stopcocks were flushed with saline to remove visible traces of propofol before transportation of patients to the PACU–DCU. Strict measures were taken to rule out exchange of information between blinded and nonblinded personnel. All patients received oxygen by mask for at least 30 min after arrival in the PACU–DCU. This served to camouflage any odor of residual isoflurane in expired air.

Research nurses who made intraoperative and postoperative observations were not involved in the care-giving process. During stay in the PACU, the anesthesia record was kept in a sealed envelope to be opened only in case of emergency by the research nurse who had performed the intraoperative observations. Patients were blinded to the anesthetic technique at all times during the study.

A research nurse recorded baseline characteristics of consenting patients, including information on previous PONV and motion sickness. Patients completed a questionnaire covering general health (Short Form-36). 10,11 

Intraoperatively, the following data were recorded: type and dose of premedication, duration of induction, airway management (endotracheal tube or laryngeal mask airway), type and doses of all medication administered intraoperatively, surgical time, and anesthetic time. Time to awakening (response to verbal command) and time to extubation after discontinuation of anesthesia were also recorded.

The surgical procedures were categorized into superficial, upper abdominal, lower abdominal, laparoscopic, inner ear, and strabismus surgery. The superficial surgery category included all procedures in which body cavities were not opened, e.g. , orthopedic, plastic, oral, and nasal surgery.

Routine monitor-ing was initiated on arrival in the PACU or DCU. Every 15 min, a research nurse recorded PONV, analgesic, and antiemetic medication. Nausea, retching, and vomiting were scored separately. Times to permission for discharge and actual discharge from the PACU or DCU (no step-down unit) were also recorded.

Approximately 24 h after surgery (range, 20–28 h), a research nurse visited inpatients at the ward or telephoned outpatients at home to record occurrence of PONV, use of antiemetics or analgesics, and the presence of possible postoperative complications and side effects subsequent to discharge from the PACU. Patients were asked to quantify discomfort caused by nausea, retching, and vomiting and to rate their anesthetic experience (scale of 0 to 10). Similar information was collected 72 h postoperatively with reference to the previous 48 h. All measurement procedures were tested in a pilot study.

Fourteen days postoperatively, patients received a questionnaire by mail regarding PONV experience from 72 h to 14 days, ratings of anesthesia, and general health as measured by the Short-Form 36. 10 

Primary outcome was the cumulative incidence of PONV at 72 h. PONV was defined as the occurrence of nausea or retching or vomiting and was combined from separate observations of nausea, retching, and vomiting (scored on seven occasions). Secondary outcomes were subjective ratings of the severity of PONV (from 0 = no discomfort from PONV, to 10 = worst discomfort possible), ratings of the anesthetic experience, and use of antiemetic drugs. Subgroup analysis was scheduled for type of surgery and gender.

We performed an evaluation of the two anesthetic techniques from an economic perspective. Costs for both anesthetic strategies were specified. Health outcomes were measured in terms of PONV incidence in the 72 h postoperatively and in terms of general health (Short Form-36) at 14 days postoperatively. As the study was based on data from individual patients, economic evaluation of consequences on the setting level (e.g. , the efficiency of hospital-wide adoption of either of the anesthetic techniques) required extrapolation.

Individual cost registration started at the time of induction. Detailed volume data on intraoperative medication, anesthesia time, and other time-related variables, postoperative medication (antiemetics and analgesics), length of stay in the PACU or DCU, and length of stay in the hospital were recorded in the case report form. For the last 753 patients, time to return to work was recorded at 14 days postoperatively.

Intraoperative use of isoflurane was measured by weighing the vaporizer on an electronic precision scale preoperatively and postoperatively (Vibra HG-10 K, accuracy 0.05 g, Ridderkerk, The Netherlands). The volume of N²O was calculated from the fresh gas flow and the fraction of N²O during steady state anesthesia. Use of antiemetics during stay in the PACU or DCU was recorded by a research nurse.

The variable and fixed-cost components and their corresponding prices are shown in the Web Enhancement (table 1). The costs of drugs and resources were based on the actual acquisition costs to the center and calculated according to the number of opened packages, i.e. , including the “spill” of drugs (any drugs left over in an opened vial or syringe at the end of a procedure were not used for a subsequent patient, but were discarded; this waste was termed “spill”) . For propofol, the costs of spill were calculated separately. Costs of equipment were composed of the depreciation, maintenance, and repair of devices specific for isoflurane and TIVA (vaporizer, oxygen–N²O fail-safe device, gas analyzer, gas evacuation equipment on the ventilator, scavenging equipment in the operating room, and infusion pumps).

Power calculations were performed for inpatients and outpatients separately. For inpatients, the expected baseline event rate in the control group (isoflurane) was estimated at 25%, the weighted average of approximately 25% high-risk procedures (55% PONV; abdominal, laparoscopy, middle ear, strabismus 12–16) and approximately 75% low-risk procedures (15% PONV; superficial). For outpatients, the baseline rate was estimated at 12.5%. Targeting detection of a 7.5% risk reduction for both inpatients and outpatients, an α of 0.05 and a β of 0.10, and anticipating 10% postrandomization loss caused by organizational constraints, sample size was set at 700 inpatients and 300 outpatients per treatment arm.

Missing individual data from the multiple observation points were not substituted except during analysis of primary outcome, which was conducted both with and without substitution of missing data (imputation). Data were imputed drawing at random from the combined PONV distribution of TIVA and isoflurane patients.

All data were analyzed according to the intention-to-treat principle. All comparisons between isoflurane and TIVA were performed for inpatients and outpatients separately, unless stated otherwise. Data are presented as mean ± SD or median with 10th–90th percentiles for nonnormally distributed variables. The number needed to treat (NNT) was calculated as 1/absolute risk reduction. Variables were compared using parametric and nonparametric tests as appropriate. Conventional survival time analysis was used to compare cumulative incidence of PONV with standard test statistics (log rank) for differences according to anesthetic treatment. Subgroup analyses were scheduled for type of surgery and gender. In addition, multivariate logistic regression analysis with 72-h cumulative incidence of PONV as the dependent variable was conducted to determine the relative role of baseline characteristics and to determine interaction, if any, between anesthetic technique and other characteristics. All statistical analyses were performed using SPSS version 9.0 for Windows (SPSS Inc., Chicago, IL).

Direct cost analysis was performed by comparing hypnotic and total anesthetic drug costs, the latter including opioids, muscle relaxants, and antagonists. In addition, direct costs of antiemetics were calculated. Return to work was used as an indicator for indirect nonmedical costs, excluding persons aged 65 yr and older and persons without a paid job.

Economic evaluation was performed using cost-effectiveness analysis, in which the NNT was used to calculate the cost of averting PONV in one patient. Economic consequences for the hospital under study regarding the budget of the anesthesiology department (based on cost reservations for anesthetics and antiemetics), assuming implementation of one technique only, were also calculated. These calculations were based on mean costs. Because no reliable data are available to value the impact of suffering from PONV in monetary terms, we refrained from cost–benefit analysis.

Length of stay in the PACU was recorded to assess the effect of technique on recovery time and PACU utilization. We evaluated the difference in time until readiness for discharge and actual discharge (log-rank statistics). From the study data (including data on variance per patient subgroup) and data on the number of surgical procedures performed per year in the hospital under study, we intended extrapolation of the economic effects on the PACU level, assuming hospital-wide implementation of either of the two anesthesia strategies.

A total of 2,010 patients were included and randomized to receive either isoflurane or TIVA (563 outpatients and 1,447 inpatients). Figure 1shows inclusion and reasons for dropout in both treatment arms. The pattern of withdrawal was similar across TIVA and isoflurane groups. Complete 72-h follow-up was obtained in 1,952 patients (97%). Postoperative questionnaires were returned by 1,803 patients (90%). Information on complications up to 14 days postoperatively was obtained for all 2,010 patients in the study, either directly from patients or from medical records and letters of discharge.

Baseline patient characteristics were similar across allocation groups (table 1). The average age of inpatients and outpatients was 45 and 38 yr, respectively. Slightly more women underwent surgery. A majority of patients reported previous surgery (83% of inpatients, 76% of outpatients). General health (Short Form-36) was lower than healthy population norms. 11 Table 2shows intraoperative data. Outpatients seldom received premedication. Endotracheal tubes and muscle relaxants were used significantly more often for inpatients than for outpatients. Among inpatients, laryngeal mask airways were used more often in the propofol group than in the isoflurane group (12%vs.  8%, P = 0.005). Conversely, more isoflurane inpatients were intubated. All patients received opioids intraoperatively. Average use of propofol for anesthesia maintenance was 8.9 mg · kg⁻¹· h⁻¹(inpatients) and 10.2 mg · kg⁻¹· h⁻¹(outpatients). Isoflurane use was 13.2 ml/h (inpatients) and 14.4 ml/h (outpatients). There were no statistically significant differences in the duration of inpatient anesthesia across techniques and in the time from discontinuation of anesthesia until response to a verbal command in outpatients after TIVA or isoflurane (median, 9 min). For inpatients, the times until response to a verbal command were 11 and 12 min after isoflurane and TIVA, respectively (P = 0.036).

The cumulative 72-h incidence of PONV among inpatients was 61% in the isoflurane group compared with 46% in the TIVA group (relative risk, 1.32; 95% confidence interval [CI], 1.20–1.46; imputed dataset). In outpatients, the cumulative incidence of PONV was 47% after isoflurane and 29% after TIVA (relative risk, 1.61; 95% CI, 1.29–2.02; imputed dataset). Figure 2shows the cumulative incidence PONV up to 72 h as survival curves. After 24 h, the relative risk for PONV in isoflurane was 1.43 (95% CI, 1.28–1.59) for inpatients and 1.83 (95% CI, 1.42–2.34) for outpatients. Table 3shows the development of PONV in the various postoperative time periods. The difference between isoflurane and TIVA was apparent at PACU discharge. Beyond 24 h, PONV development was similar in both treatment arms. Results were identical using the imputed dataset or using the dataset of patients with 21 complete observations for PONV (88%).

Venn diagrams (fig. 3) show the individual contributions of nausea, retching, and vomiting to the combined PONV end point with the two anesthetic techniques. The percentage of patients who had all three emetic symptoms decreased from 26.5% to 14.8% for inpatients (relative decrease, 44%) and from 12.1% to 6.2% for outpatients (relative decrease, 49%).

Antiemetics were administered more often in patients who received isoflurane (inpatients: 36%vs.  18%, P < 0.001, relative risk 2.06, 95% CI 1.71–2.49; outpatients: 20%vs.  8%, P < 0.001, relative risk 2.65, 95% CI 1.65–4.24). Subjective ratings of discomfort from nausea at 24 h were lowest in the TIVA group. The percentage of patients reporting “no discomfort” from nausea after 24 h was 58%versus  70% among inpatients and 70%versus  80% among outpatients for TIVA and isoflurane, respectively. After 24 h, ratings of the anesthetic experience in general were significantly higher for TIVA inpatients than for isoflurane inpatients. After 14 days, ratings of the anesthetic experience were significantly higher for all TIVA patients (data on patient ratings of PONV and anesthetic experience can be found in the Web Enhancement, table 2).

One outpatient in the TIVA group and two in the isoflurane group (incidence, 0.5%) were admitted to the ward because of intractable PONV. In the structured 24-h postoperative interview, seven patients reported awareness for the anesthetic period, an incidence of 0.4% (isoflurane, n = 4; TIVA, n = 3). Two outpatients in the TIVA group had memories of pain and sounds during anesthesia, and one isoflurane inpatient and one isoflurane outpatient remembered sounds and inability to move. The remaining three patients remembered being aware of sounds.

Total intravenous anesthesia reduced the incidence of PONV most in the subgroup of patients undergoing superficial surgical procedures (absolute risk reduction, 20%). TIVA reduced PONV after middle ear surgery but not after abdominal and laparoscopic surgery (table 4). Because of the small number of patients undergoing strabismus surgery (n = 15), no conclusions could be drawn on the effect of TIVA in these patients.

Female patients suffered substantially more PONV than male patients (women at 72 h: isoflurane 65%vs.  TIVA 50%; men: isoflurane 46%vs.  TIVA 31%;P < 0.001). Survival curves depicting PONV development for men and women separately can be found in the Web Enhancment (fig. 1.) Multiple logistic regression confirmed the effect of anesthetic technique (odds ratio for inpatients: 2.03 [95% CI, 1.57–2.59], for outpatients: 2.30 [95% CI, 1.52–3.49]) when corrected for baseline characteristics such as age, gender, and previous PONV.

For inpatients, length of stay in the PACU until readiness for discharge was 135 min with isoflurane versus  115 min with TIVA (P < 0.001). For outpatients, length of stay in the DCU until readiness for discharge home was 160 min with isoflurane and 150 min with TIVA (P = 0.039). Times until actual discharge varied accordingly (data not shown). The difference between isoflurane and TIVA remained the same throughout the entire range of discharge percentiles. (The distribution of time to readiness for discharge with both strategies is shown in fig. 2of the Web Enhancement). Significant differences in times until readiness for discharge between randomization arms were also observed in the subgroup of patients who did not experience PONV in the PACU or DCU. The median (10th–90th percentile) times until readiness for discharge for inpatients without PONV were 125 min (75–239) after isoflurane versus  110 min (65–210) after TIVA (P = 0.001). For outpatients without PONV, these times were 155 min (95–230) after isoflurane versus  145 min (85–219) after TIVA (P = 0.03).

On average, 18 adult inpatients and 12 outpatients per day are eligible for either isoflurane or TIVA in our center. Extrapolation of recovery time data to these patients could lead to efficiency gains along the following lines. In the scenario in which all patients receive TIVA, the reduction in PACU time in the most optimistic situation would be 15 min per inpatient and 10 min per outpatient (based on median time to readiness for discharge after TIVA vs.  isoflurane). Patient time saved in the PACU and DCU in the course of 1 day could theoretically be 4.5 h (18 patients times 15 min) and 2 h (12 patients times 10 min), respectively.

Detailed drug acquisition costs at the time of the study can be found in the Web Enhancement, table 1). Table 5shows the intraoperative volumes of anesthetics. For inpatients (median duration of anesthesia = 2 h) median costs (10th–90th percentile) of induction with thiopental and maintenance with isoflurane were $10.84 (5.67–22.64) versus $39.53 (19.89–75.74) for propofol TIVA. In outpatients (median duration of anesthesia = 1 h), these amounts for induction with propofol and maintenance with isoflurane were $13.10 (8.51–20.18) versus $28.31 (19.89–47.69) for propofol TIVA.

Use of antiemetics was twice as high in the isoflurane group (36%vs.  18%). The total costs of antiemetics comprised less than 2% of total drug costs. No differences in the use of analgesics were observed. The actual costs of all anesthetic drugs in the study groups can be found in the Web Enhancement, table 3).

Median length of hospitalization was 3 days in both randomization arms (10th–90th percentiles: 1–9 days after isoflurane and 1–10 days after TIVA; difference not significant). Sixty percent of the inpatients and 77% of the outpatients had a paid job. There was no statistically significant difference in the number of days to return to work with both strategies (median, 14 days for inpatients and 10 days for outpatients).

After 14 days, Short Form-36 general health scores showed no differences between patients in the isoflurane and TIVA groups. Therefore, as intended, we conducted a cost identification analysis that focused on overall cost differences given equivalent clinical outcome. The additional cost per surgical session for TIVA when compared with isoflurane was $28.98 for inpatients and $14.87 for outpatients.

If we were to consider the transient reduction of PONV as a clinical end point, then cost-effectiveness analysis would be justifiable (with cost per additional patient free of PONV as the outcome measure). To avoid PONV in the first 24 h after surgery in one inpatient who would have suffered from PONV after isoflurane, six patients would have to receive TIVA (absolute risk reduction, 17; NNT = 6), totaling $174 ($29 × 6). Accordingly, to avoid PONV until 24 h in one outpatient, five patients would have to receive TIVA (absolute risk reduction, 20; NNT = 5), amounting to $75.

In 1998, 15,590 patients underwent surgery in the hospital under study. Approximately 7,770 of these patients were older than 18 yr and eligible to receive either isoflurane or TIVA. If all patients would have received TIVA, the additional cost of propofol for 1998 would have been $225,218. The additional costs for TIVA are lower when the decreased use of antiemetics after TIVA is taken into account. When either metoclopramide or droperidol are used as primary antiemetic therapy, savings resulting from decreased use of antiemetics would amount to $2,220 after TIVA, equaling 1% of the additional drug acquisition cost for propofol. If ondansetron, a more expensive antiemetic, would have been the first-choice antiemetic therapy, decreased use of antiemetics after TIVA would save $6,324 in the cost of antiemetics.

The cumulative incidence of PONV was significantly lower after TIVA than after isoflurane. Absolute risk reduction with TIVA was between 15 and 20% (NNT = 7–5) depending on duration of follow-up. Moreover, from the patients’ perspective, TIVA was superior. The PONV reduction in the current study is in agreement with results from two recent metaanalyses that pooled data from several smaller studies comparing propofol with inhalational agents. Tramer et al.  6and Sneyd et al.  7found an NNT with propofol TIVA of 6 and 7, respectively, to prevent one early PONV incident (< 6 h). Our follow-up period was long compared with other PONV studies. The effect of the anesthetic technique was most prominent in the first 24 h after surgery (early PONV), whereas beyond that point the incidence of PONV increased equally in both groups. This suggests that anesthetic-induced PONV is most important in the first 24 h after surgery, whereas PONV resulting from the surgical procedure and postoperative analgesics dominates thereafter.

Power analysis was based on PONV incidences from the literature available at the time of study design. 12–16The higher-than-expected PONV incidence increased the power of the study to detect a difference in PONV between TIVA and isoflurane. Moreover, the large sample size strengthens the results of subgroup analyses and the inference regarding the lack of difference in the incidence of complications between the TIVA and isoflurane groups. 17 

As expected, type of surgery was a major determinant of PONV frequency in both groups, and it modified the effect of the anesthetic technique on PONV. Patients undergoing superficial surgical procedures benefited most from TIVA (absolute risk reduction = 18%; NNT = 6). An unexpected finding was that, in the patients undergoing abdominal procedures, TIVA was unable to suppress the occurrence of PONV, although the number of intraabdominal procedures was relatively low. We cannot exclude that TIVA may suppress early PONV for intraabdominal procedures. For laparoscopic procedures, we were unable to detect a protective effect from TIVA. This finding has not been previously reported and refutes results from previous studies. 5Demographic characteristics also affected the probability of PONV, with female gender and younger age predisposing toward higher incidence in both groups.

Various definitions of PONV have been used in studies on the effects of anesthetics and antiemetics, e.g. , nausea only, nausea and vomiting, or vomiting only. 18,19This has hampered interstudy comparability. 18Because we scored nausea, retching, and vomiting independently, our data allowed for alternative end-point definitions. The Venn diagrams in figure 3show that PONV is primarily determined by the presence of nausea. When vomiting and retching are combined and taken as one end point, the incidence of PONV is lower, but similar differences between isoflurane and TIVA remain. Accordingly, the results of the various possible PONV end points are comparable, provided that nausea is included.

The incidence of PONV in the current study was high. The reported incidence of PONV varies widely in the anesthesia literature, depending on setting and case mix (from < 5% to > 50%). 16,20Diversity in methods of data collection may also account for some of the observed differences. Emetic symptoms can be quantified as retrospective self-report, established through explicit questioning, or observed on site by a third party. As a consequence of the effects of both suggestion and increased detection, repeatedly questioning patients about PONV might result in a higher percentage of patients reporting PONV and receiving antiemetic therapy than would be the case in normal practice. In our study, blinded trial nurses did not communicate PONV findings to PACU nurses caring for patients, and PONV scores in the case report form were unavailable to PACU nurses.

One hypothesis at the outset of the study was that the results might reveal subgroups of patients who would benefit more from TIVA. This would allow identification of subgroups for whom TIVA could be especially advantageous. However, except for abdominal and laparoscopic procedures, TIVA proved beneficial to the same extent for all patient groups. Therefore, the practice of reserving TIVA for high-risk patients only seems unjustified.

Among inpatients, the study anesthetics consisted of induction with either propofol or thiopental, followed by TIVA or isoflurane. All outpatients received propofol for induction. If propofol used for anesthesia induction only were antiemetic, irrespective of the maintenance regimen, the magnitude of the reduction in relative risk among outpatients would have been less than actually occurred among outpatients in the current study. This finding supports results from a metaanalysis by Tramer et al. , 6who showed that propofol for anesthesia induction followed by a nonpropofol maintenance technique did not result in PONV reduction.

The difference in airway management among the inpatients (more laryngeal mask airways in the propofol group) is probably caused by the fact that the ultimate selection of airway management may have been determined or altered after disclosure of the anesthetic allocation (TIVA is recommended in combination with the laryngeal mask airway). A recent study by Joshi et al.  21showed that PONV was similar for patients with endotracheal tubes and laryngeal mask airways when an identical anesthetic technique was used. Therefore, the difference in airway management in the current study is not likely to be a confounding factor.

Many anesthesiologists add N²O to propofol anesthesia because the additive effect allows for lower infusion rates, which reduces cost. 22,23Furthermore, it has been suggested that supplementing TIVA with N²O would reduce the incidence of awareness and recall (equal for both techniques in this study). 24However, N²O may increase the incidence of PONV. 25A recent metaanalysis showed a pooled odds ratio for omitting N²O of 0.63 (0.53–0.75, 95% confidence interval). 26The current trial was not designed to determine to what extent PONV after inhalational anesthesia is caused by N²O or isoflurane. If the treatments had consisted of propofol–N²O versus  isoflurane–N²O, the treatment effect of TIVA might have been smaller.

The shorter length of stay at the PACU or DCU after TIVA (approximately 15 min) has also been reported by other investigators. 9,27,28Patients without PONV were discharged from the PACU–DCU sooner after TIVA than after isoflurane. Our protocol required patients to remain in the PACU for at least 1 h. The observed 15-min shorter PACU times will probably not hold for procedures in which a patient can be discharged to the ward after 30 min. Although it would appear that faster recovery times directly decrease costs, it is difficult to convert these shorter PACU and DCU stays into economic gains. The flow through the operating rooms and PACU–DCU is a chained process; Dexter and Tinker 29stated that “the major determinant of PACU costs is, by far, the distribution of admissions.” Furthermore, a system of flexible staffing would be a necessity. 8,29Therefore, in the setting of this study, it is unlikely that the PACU time with TIVA can be used to recover one additional patient or to reduce the number of PACU nurses by one. In a more homogeneous population with respect to type of surgery, and in case of brief standardized procedures, translation of shorter recovery after TIVA into economic gains is more likely. This could probably be best achieved in an office-based anesthesia setting or an ambulatory surgicenter.

At the time of the study, TIVA with propofol was two to three times more expensive than conventional anesthesia with isoflurane and N²O when considering intraoperative costs only. Using the NNT, the costs of preventing PONV in one additional patient by using TIVA instead of isoflurane were $174 in inpatients and $75 in outpatients. A reduction of propofol acquisition cost by 65% would make TIVA equally expensive as isoflurane–N²O in outpatients. For inpatients, a propofol cost reduction of 75% would make the cost of TIVA similar to that of isoflurane–N²O. The patent on the current propofol emulsion (Diprivan, AstraZeneca Nederland, Zoetermeer, The Netherlands) has recently expired in The Netherlands, resulting in a steep decrease of propofol acquisition cost (up to 75% reduction, depending on the formulation [1%vs.  2%] and type of packaging [prefilled syringes or glass vial]). With this unanticipated large cost decrease, several cost considerations have lost relevance (university setting, case mix, drug waste caused by the forced use of prefilled 50-ml syringes when using target controlled infusion pumps).

Because this study was conducted in a single academic institution, external validity (ability to generalize) is an issue. However, the study population comprised a large heterogeneous group of unselected patients and surgical procedures of varying duration, although very brief procedures were underrepresented. In addition, the design of this trial, apart from random assignment of TIVA or isoflurane and strict blinding procedures, did not interfere with current practice patterns. Therefore, our study patients and surgical procedures are comparable to those in other teaching institutions.

In conclusion, we have shown that propofol TIVA results in a reduction of PONV, particularly in the early postoperative period. TIVA increases patient comfort and patient ratings of anesthesia., while slightly reducing PACU and DCU discharge times. However, anesthesia costs were greater, and no clear economic gains were found.

The authors thank Wilco E. van Genderen, M.D. (Anesthesiologist, Medical Center Alkmaar, The Netherlands), and Lucilla E. Overdijk, M.D. (Anesthesiologist, Onze Lieve Vrouwe Gasthuis, Amsterdam, The Netherlands), for constructive input in the design of this study, the team of research nurses for their enthusiasm and dedication, and the anesthesiologists and anesthesiology residents of the Academic Medical Center, University of Amsterdam, for their support and help with patient inclusion. Many thanks also to John Hartung, Ph.D. (State University of New York Health Science Center, Brooklyn, NY), whose careful editing of the manuscript was invaluable.