Early resumption of feeding after surgery is of concern for both patients and healthcare providers and reflects that the patient is returning to normal physical and psychological functioning. The ability to eat after ambulatory surgery is one of the items of the Post-Anesthetic Discharge Scoring System (PADSS) score.1 Importantly, resumption of feeding may be a treatment in itself. For example, early feeding after major colorectal surgery reduces postoperative infection complications and shortens the time to resumed gastro-intestinal transit.2,3,4 Early postoperative feeding could also be clinically important in situations of acute (e.g., cancer, inflammation) or chronic (e.g., anorexia nervosa) under-nutrition.

The sensation of hunger is finely regulated, involving both neural and peripheral mechanisms. The rise in plasma levels of ghrelin, produced by the stomach, increases hunger,5 while insulin, produced by the pancreas, and leptin, produced by adipocytes, reduce this sensation.6 These peptides in turn modulate the hypothalamic control of feeding behaviour.7 The inflammatory response mediated by surgery can affect hypothalamic activity responsible for postoperative anorexia.8

Anesthetic drugs may alter feeding behaviour and preclinical data suggest that propofol may be orexigenic. For example, we previously reported that, unlike isoflurane, propofol induces feeding behaviour in mice for at least six hours.9 In clinical studies, compared with isoflurane or halothane, propofol has been reported to enhance a return of the patient’s appetite after neurosurgery10 and ophthalmological procedures.11 While it has been claimed that propofol is associated with an intense feeling of hunger in outpatients undergoing colonoscopy, no corroborating data were provided.12

To further our understanding of the influence of anesthetics on postoperative appetite, we compared the effects of propofol with sevoflurane on appetite and plasma levels of peptides involved in food intake. To limit potentially confounding variables related to age, sex, duration of anesthesia, and surgical procedure that may influence postoperative hunger, we focused on a reasonably narrow patient population undergoing a standardized surgical procedure: transvaginal ultrasound-guided oocyte retrieval.

Methods

Trial characteristics

This study was approved by the ethics committee of Rouen University Hospital, France (Institutional Review Board no. 01/013/2014, 22 September, 2014) and by the French national agency for drugs and health products (Agence Nationale de Sécurité des Médicaments et des Produits de Santé, no. 140854A-31). The trial was registered prior to patient enrolment at clinicaltrials.gov (NCT02272166). The manuscript is written in accordance with Consolidated Standards of Reporting Trials (CONSORT) guidelines. Written informed consent was obtained from all participants.

Population

Study subjects (recruited February 2015–July 2016) were adult (18–45 yr old) females undergoing scheduled oocyte retrieval under general anesthesia. All patients had undergone similar ovarian stimulation using gonadotropins, combined with either an agonist or an antagonist of the gonadotropin releasing hormone (GnRH). Exclusion criteria were American Society of Anesthesiologists (ASA) physical status score ≥ III, Apfel score for postoperative nausea/vomiting13 ≥ 2, diabetes mellitus, cognitive impairment, body mass index (BMI) > 35 kg·m−2, diminished nutritional status or chronic eating disorder, or chronic treatment with benzodiazepines or serotonin reuptake inhibitors. Patients anesthetized after 10:00 a.m. were excluded because of possible confounding related to prolonged fasting.

Endpoints

The primary endpoint was the time between the end of anesthesia (EOA), defined as the interruption in anesthetic administration (hypnotic and remifentanil), and the recovery of hunger, defined as ≥ 50/100 mm on a visual analogue scale (VAS).14,15 Secondary endpoints included the time between EOA and ingestion of food, caloric intake at first feeding, incidence of nausea or vomiting, time for readiness to discharge from the post-anesthesia care unit (PACU), a postanesthetic discharge scoring system [PADSS] score ≥ 9/101 (see the Appendix), and plasma levels of ghrelin, leptin, and insulin at ten minutes, one, and two hours after EOA.

Study design

We conducted a randomized, 1:1 parallel-arm, prospective, patient-blinded, single-centre study comparing the effects of propofol vs sevoflurane on postoperative recovery of hunger. Randomization was performed using CSOnline Software v7.5.30 (Clinsight©, Poitiers, France). A single permuted block randomization list was generated with nine blocks of 12 subjects and one block of eight subjects.

Procedures were performed between 08:00 and 10:00 a.m. During the preoperative period, fasting (six hours for solids, two hours for clear liquids) was required and paracetamol 1 g was administered orally before the start of anesthesia. Hydroxyzine 0.5 mg·kg−1 was administered if patients were severely anxious. No preoperative administration of carbohydrates was provided. A catheter inserted in an upper extremity vein was used for administration of anesthesia drugs while a second catheter in a contralateral arm vein was used for blood sampling. Standard monitors were applied (electrocardiogram, pulse oximetry, and non-invasive blood pressure) and end-tidal expiratory gases were measured via the anesthesia machine (Zeus® and Infinity® acute care system, Drägerwerk AG, Lübeck, Germany). Patients were pre-oxygenated with 100% O2 to achieve an expired fraction ≥ 90%. Remifentanil was administered with a target-controlled infusion (TCI, Minto model) of 4 ng·mL−1 to induce anesthesia, followed by a target effect site level of 1–4 ng·mL−1. Intravenous dexamethasone 4 mg and droperidol 0.625 mg were administered to help prevent postoperative nausea. In the propofol group, loss of consciousness was initiated by administering propofol using a 3–8 μg·mL−1 TCI, and maintained with a TCI of 1–10 μg·mL−1. In the sevoflurane group, unconsciousness was induced by an inhaled concentration of sevofurane 6% via a tight-fitting face mask, and was maintained with an end-tidal concentration of 1–4%. The depth of anesthesia in both groups was monitored using the bispectral index (BIS quarto Sensor® and BIS monitor®, Medtronic, USA) with sevoflurane or propofol administration adjusted to achieve a BIS target of 40–60. In both groups, a supraglottic device (i-gel®, Intersurgical Ltd, Wokingham, UK) was inserted once the depth of anesthesia was sufficient (e.g., stable blood pressure and heart rate, easy bag-and-mask ventilation, easy mouth opening). No neuromuscular blocking drug was used. Mechanical ventilation was performed with a tidal volume of 7 mL·kg−1 and a respiratory rate (12–20 breaths·min−1) to achieve an end-tidal carbon dioxide of 30–40 mmHg (air/oxygen mixture of 50% fraction of inspired oxygen, and fresh gas flow was automatically adjusted by the anesthetic machine). Nitrous oxide was not used in either group. Fluid administration was standardized to sodium chloride 0.9% at 5 mL·kg−1·hr−1 during the procedure. At the end of the procedure, all drug administration was terminated and patients were ventilated with high-flow 100% oxygen. Upon awakening, the supraglottic device was removed and patients were transferred to the postanesthesia care unit (PACU) and then to the ambulatory unit. During the postoperative period, tramadol (50 mg iv) was administrated for pain relief followed by morphine (2 mg iv every five minutes), as required. Nausea and/or vomiting were treated with ondansetron 4 mg iv.

Nurses involved in the postoperative care of the patients (PACU and ambulatory unit) were blinded to the group allocation. Nurses evaluated the sensation of hunger every 20 min using the VAS, and as soon as the score was ≥ 50/100 mm, they gave the patient a meal and quantified the caloric intake. Validation of the PADSS score for readiness to discharge home was also assessed by nurses every 20 min.

Plasma measurement

Plasma levels of ghrelin, leptin, and insulin were sampled immediately before anesthesia and at ten minutes, one hour, and two hours after the EOA. At each time-point, two ethylenediaminetetraacetic acid-coated tubes were filled with blood. One of the tubes contained 1 mg·mL−1 of 4-(2-aminoethyl)-benzenesulfonyl fluoride hydrochloride to limit the ex-vivo degradation of ghrelin. After 15-min of centrifugation (3000 G, 4°C), the supernatant was collected and immediately frozen at − 20°C until assayed. Enzyme-linked immunosorbent assay kits were used for leptin and ghrelin analysis using duplicate measurements (ghrelin EZGRT-89K and leptin EZHL-80SK, EMD Millipore, Burlington, MA, USA). Insulin plasma levels were assessed using electrochemiluminescent immunoassay (Elecsys Insulin, Roche Diagnostics, Indianapolis, IN, USA).

Statistical analysis

All analyses were performed as intention-to-treat. Missing quantitative data were simply imputed to the median value of the other group (tendency to favour the null hypothesis). Quantitative data are presented as median [interquartile range (IQR)] and compared using a two-sided Mann–Whitney test. Qualitative data were compared with Fisher’s exact test. A P < 0.05 was considered significant. Eighteen Mann–Whitney tests were performed to compare sevoflurane and propofol groups and the three peptides at three time-points (EOA + ten minutes, EOA + one hour, EOA + two hours) with two methods (crude values and values expressed as % in change, the latter being a post hoc analysis), so we adjusted the P value to the number of comparisons using the family-wise error rate controlling procedure following the single-step resampling method.16 Within-groups variations of peptide levels over time were also tested by Wilcoxon paired samples tests (36 tests for three peptides, three times, two groups, and crude vs %-change analyses) with the same Westfall and Young procedure16 but in a different multiple-testing family since they were unrelated. Since the treatment (sevoflurane vs propofol) was randomized, the baseline values were not statistically compared because we assumed that they were not different. The Hodges–Lehmann estimator was used to calculate two-sided 95% confidence intervals (CI) of differences between groups; it can be interpreted as the median of differences that would be obtained by randomly sampling a patient in each group. The primary analysis was the comparison of the time between EOA and the recovery of hunger between the two groups using a Mann–Whitney test. This test compares the area under a receiver operating curve (AUC) of the quantitative variable (time for recovery of hunger) with the binary variable (group) to the value of 0.50 (no association).

The sample size was calculated assuming an AUC equal to 0.65 (moderate association), a 1:1 randomization, power equal to 80%, a 5% type-I error rate and a two-sided formulation. Assuming a loss of power equivalent to 5% of patients (three patients per group) due to imputation of missing data, the required sample size was determined to be 58 subjects per group. Analyses were performed using SAS® (version 9.4, SAS Institute, Cary, NC, USA) and R statistical software (version 3.5.0, The R Foundation for Statistical Computing, Austria).

Results

Of the 438 screened patients, 116 were enrolled in the study, with 58 patients allocated to each group (Fig. 1). Two patients in each group did not reach the primary endpoint of recovery of hunger (≥ 50/100 mm VAS). According to the intention-to-treat methodology, the median values of the other group were attributed to these patients and so data from 58 patients were analyzed. Patients in each group could not be differentiated based on age, BMI, American Society of Anesthesiology (ASA) physical status score, duration of fasting, baseline cardiorespiratory values, and Apfel score (Table 1).

Fig. 1
figure 1

Study flowchart.

Full size image
Table 1 Preoperative characteristics
Full size table

The median [IQR] times between the EOA and recovery of appetite (hunger) were similar in the propofol vs sevoflurane groups (97 [75–138] vs 97 [80-–140] min, respectively; median difference, 1; 95% CI, − 15 to 14; P = 0.91) (Table 2, Fig. 2). Similarly, the time before the first meal was not different between groups and there was no difference in the caloric intake (Table 2).

Table 2 Intra- and postoperative results
Full size table
Fig. 2
figure 2

Kaplan–Meier curve showing the proportion of hungry patients (VAS ≥ 50/100 mm) vs time (minutes from the end of anesthesia). VAS = visual analogue scale.

Full size image

Patients in either group could not be differentiated based on readiness for discharge from the PACU or length of stay in the PACU (Table 2). A similar number of patients in either group required postoperative analgesics (Table 2). While there was no statistical difference in the occurrence of nausea and vomiting (Table 2), it is noteworthy that no patients in the propofol group experienced this compared with four in the sevoflurane group. Moreover, two patients in the sevoflurane group required anti-nausea medication in the PACU (Table 2) and one of these patients required a 24-hr hospital stay because of prolonged nausea and dizziness.

There was no difference in crude insulin, leptin, or ghrelin plasma levels (planned analysis) between groups at the three blood sampling time-points. When peptide plasma levels were expressed as a % change from baseline (post hoc analysis), there was a higher insulin plasma level one hour after EOA in the sevoflurane group (median difference, 4.9%; 95% CI, − 16.2 to 43.4) compared with the propofol group (median difference, − 21.2%; 95% CI, − 35.7 to 9.1), adjusted P = 0.01 (Table 3). Within each group, compared with baseline, both crude and relative insulin plasma levels were raised two hours after EOA, ghrelin levels were reduced at ten minutes, one hour, and two hours after EOA, and leptin levels were reduced at one hour and two hours after EOA (Table 3).

Table 3 Plasma levels of insulin, ghrelin, and leptin at four time-points
Full size table

Patients in the sevoflurane group experienced a slightly longer median [IQR] duration of anesthesia compared with the propofol group (27 [21–30] vs 23 [18–29] min, respectively; P = 0.04, post hoc analysis) and a higher median [IQR] dose of remifentanil (3.8 [3.1–5.1] vs 3.5 [2.7–4.2] ng·kg−1, respectively; P = 0.02, Table 2).

Discussion

We showed that the choice of hypnotic drug—propofol vs sevoflurane—did not influence the recovery of hunger after minor outpatient surgery. Similarly, the choice of anesthetic drug did not influence crude plasma levels of insulin, ghrelin, and leptin, all of which are related to appetite control. When plasma peptide levels were calculated as % change from baseline values, the sevoflurane group had a small increase in insulin levels at one hour following termination of anesthesia compared with the propofol group. This is the first randomized trial to specifically explore the impact of anesthetic agents on postoperative food intake. Very few studies have examined the impact of anesthesia on food intake behaviour; this is surprising given that early resumption of feeding after surgery is of major concern. Indeed, early feeding after major surgery may help to reduce postoperative complications and enhance recovery after surgery interventions.2,17,18 Importantly, a reduced sensation of hunger and caloric intake following surgery can contribute to a catabolic state and a negative energy balance.19 While early feeding after minor surgery may not have as great an impact, this could still influence overall patient comfort and satisfaction, facilitating an earlier return home and resumption of daily activities.

Our study design avoided several potential sources of confounding errors: type and duration of surgery, variability in inflammation, heterogeneity of patients, and variable fluid management, to name a few. Indeed, animal and human studies suggest that advanced age20,21 or inflammation8,22 can alter hunger and feeding behaviour. Moreover, these factors may differentially affect males and females.8 The results of our study conflict with other findings suggesting that unlike volatile anesthetics, propofol exerts orexigenic properties.9,10,11,12 In part, this may be explained by the different populations that have been studied as well as the type and duration of surgery.

The sensation of hunger is a complex process involving neuronal and systemic factors. The arcuate nucleus of the hypothalamus plays a key role with two specific neuronal populations: neuropeptide-Y (NPY)-producing neurons with orexigenic function23 and pro-opiomelanocortin-producing neurons with anorexigenic function.24 The arcuate nucleus is considered to be at the crossroads of systemic and neuronal pathways because of the particular weakness of the blood–brain barrier in this area.23 We studied three peptides involved in feeding behaviour: ghrelin has orexigenic properties while insulin and leptin show anorexigenic properties. Insulin is secreted from β-cells in the pancreas, in response to the elevation of glycemia after eating, and low insulin plasma levels enhance the production of the orexigenic neuropeptide-Y (NPY) in the hypothalamus, triggering hunger.23 Of note, animal studies have shown that propofol administration is associated with a rise in insulin production compared with halogenated drugs.25,26 This may be explained, in part, by the fact that propofol contains 10% fatty acids, and central sensing of fatty acids by the hypothalamus enhances insulin production.27

Leptin is mainly produced by white adipose tissue and is positively correlated with fat mass and nutrient intake.28 It decreases food intake in mice via activation of central neurons implicated in feeding behaviour.24 Ghrelin is produced in the fundus of the stomach and its production is enhanced by fasting.29 It is currently the only gut hormone with known orexigenic proprieties, and high plasma levels induce NPY production.23 In our study, no between-group differences were observed in leptin- and ghrelin plasma levels and changes in plasma levels of the three hormones within each group showed similar fluctuations for propofol and sevoflurane, with slight to moderate variations over the time. To the best of our knowledge, no other study has explored the impact of anesthesia on these hormones.

Several limitations of our study should be acknowledged. The study focused on procedures performed between 08:00 and 10:00 a.m. (to limit potential confounding related to duration of fasting) and this may limit the generalizability of our data. We also cannot exclude the possibility that the administration of hormones before oocyte puncture may have altered patients’ sensations of hunger, even though the randomization process achieved a homogeneous distribution of patients who received either GnRH agonists or antagonists. We also cannot exclude the possibility that the relatively brief duration of anesthesia our patients were exposed to limited the potential effects of anesthetic drugs on hunger, notwithstanding the anecdotal evidence that propofol enhances hunger in outpatients undergoing brief endoscopic procedures.11,12 A potentially confounding variable was the difference in duration of anesthesia between groups: patients in the sevoflurane group had a somewhat longer duration of anesthesia including a higher cumulative dose of remifentanil, which could have impacted feeding behaviour. As sevoflurane is expected to reduce the sensation of hunger,9,10 we anticipate that this difference in anesthesia time would have (if anything) exaggerated any pro-feeding influence of propofol. Another potentially confounding variable is the effect of other drugs administrated to both groups, including paracetamol, hydroxyzine, dexamethasone, droperidol, and remifentanil. Although opioids do not appear to modify appetite in cancer patients,30 the endogenous opioid-cannabinoid system is implicated in the reward system for food intake,31 and we cannot rule out a confounding effect of remifentanil on hunger. Moreover, opioids can induce nausea and/or vomiting, which may adversely affect hunger.32 Finally, as this study was performed only on female patients, the relevance of these findings to males is questionable.

In summary, compared with sevoflurane, propofol did not accelerate the recovery of hunger after outpatient minor surgery. Moreover, propofol was not distinguished by its effect on other clinical or biological parameters associated with food intake.