Investigators from Bristol described a fentanyl- and diclofenac-based analgesic technique for tonsillectomy with low postoperative nausea and vomiting rates and low pain scores. This study compared the effectiveness of a modified Bristol technique with a codeine-based regimen with respect to PONV and analgesia. Sixty-five children, ASA 1-2, were randomly assigned to either the Bristol group (fentanyl 1–2 μg.kg−1 and diclofenac 1–2 mg.kg−1) or codeine group (codeine 1.5 mg.kg−1). All children received paracetamol 15 mg.kg−1 and dexamethasone 0.1 mg.kg−1. Postoperative nausea and vomiting and pain scores were recorded hourly, and fitness for discharge was assessed at 4 h. The overall incidence of postoperative nausea and vomiting was 21% with no difference between groups (Bristol group 8/30, codeine group 5/32, p = 0.29). Children in the Bristol group required analgesia earlier than those in the codeine group (p < 0.005), but maximum pain scores were not different (Bristol group median (IQR [range) 4.5 (3-5 [0-5]), codeine group 4.0 (2-5 [1-5]), p = 0.15). Twenty-three per cent of children were assessed as not fit for discharge at 4 h. The codeine-based regimen may have a small advantage over the Bristol regimen, but neither technique seems ideally suited for a day-case service without a longer period of observation.
Tonsillectomy is a common operation associated with a high incidence of postoperative nausea and vomiting (PONV) and significant postoperative pain. Postoperative nausea and vomiting occurs in over 70% of children undergoing tonsillectomy, if prophylactic anti-emetics are not used [1, 2]. It is distressing for children and may cause dehydration, electrolyte disturbances, aspiration of gastric contents and an increased risk of postoperative bleeding . Adequate analgesia is important not only on humanitarian grounds but also because pain may impair swallowing, and consequently promote infection . Tonsillectomy is increasingly performed as a day-case procedure. Day-case surgery is cost effective and reduces pressure on in patient beds. It causes less psychological trauma as children and parents prefer to avoid overnight admission where possible [5, 6]. It has a good safety record although unplanned admissions occur in 3–4.7% of cases, most commonly for PONV [7, 8].
The literature considering alternative analgesic strategies for tonsillectomy is extensive; yet a consensus on the best technique remains elusive and practice in the UK shows considerable variation . In 2005, a group from Bristol described a fentanyl and diclofenac analgesic regimen, and reported a postoperative vomiting rate of just 5% in the first 24 h after surgery, without the use of prophylactic anti-emetics . Although the study was observational, a regimen with these benefits is appealing particularly in a day-case setting. The aim of our study was to compare the Bristol regimen with a codeine-based regimen used commonly at our institution. We wished to ascertain if PONV rates may be improved without compromising pain relief, and if fitness for discharge rates differed between the two techniques.
This was a randomised controlled trial with blinded assessment of PONV and pain. Research ethical approval was obtained. Families were provided with age appropriate information by post, one week before admission, and informed consent was obtained on the day of surgery. Children were eligible if they were undergoing tonsillectomy or adenotonsillectomy, aged between 4 and 16 years and of ASA physical status 1-2. Exclusion criteria were bleeding diathesis, renal or hepatic impairment, neurological disorder impairing accurate pain assessment, parental language barrier, involvement in another study, sensitivity to any trial drug, requirement for premedication and other analgesics administered on the day of surgery.
Children were allocated to either the Bristol group or the codeine group; the protocol for each group is shown in Fig. 1. We made two adjustments to the published Bristol regimen for the purpose of this study: paracetamol was administered intravenously rather than orally, and dexamethasone was administered as a prophylactic anti-emetic . Randomisation was by numbered envelopes ordered from a computer-generated random list, and was stratified by age group (4–10 years and 11–16 years), as we expected higher numbers of younger children. The anaesthetist for each case selected the next numbered envelope for the relevant age group. We did not formally control for surgical technique, but all operating surgeons used cold steel dissection with bipolar diathermy for haemostasis.
An investigator blinded to the allocated group made all postoperative observations. Children were assessed for PONV and pain every 10 min in recovery, and every hour on the ward up to 4 h. Postoperative nausea and vomiting was assessed with a four-point scale (see Table 1), and treated at scores of 2 and 3. Pain assessment in recovery used the objective pain score (five criteria, maximum score of 10), and on the ward used the Wong and Baker Faces Pain Scale (score 0–5) , with analgesia administered at scores of 5 and 2 respectively. Fitness for discharge was assessed at 4 h by four criteria: no evidence of bleeding; pain score 0 or 1; PONV score 0; and when oral fluid intake was established.
Table 1. Postoperative nausea and vomiting scoring system.
Nausea or retching
Persistent nausea for > 30 min or vomiting more than once in 30 min
Our primary outcome measure was the incidence of postoperative vomiting, with nausea assessed as a distinct entity and included as a secondary outcome measure. Vomiting alone was used as a primary measure as it is objective, common following tonsillectomy and is the main reason for unplanned admission after day-case surgery. The Bristol study recorded nausea and vomiting separately, but many other studies include only vomiting episodes, and this approach allowed us to compare our results with both. Other secondary outcome measures were time to first postoperative analgesia, maximum pain score and fitness for discharge at 4 h.
Sample size was based on the incidence of postoperative vomiting in the first 4 h after tonsillectomy. The Bristol study had no cases of vomiting in this time period (their overall incidence of postoperative vomiting was 5% and due to two cases of vomiting between 4 and 24 h). The incidence of postoperative vomiting with a codeine-based regimen was estimated from published data. Splinter and Roberts reported a postoperative vomiting rate of 25% and 36% with a technique incorporating intramuscular codeine and dexamethasone, respectively [12, 13]. We assumed the lowest reported incidence of 25% for our codeine group and a zero rate for the Bristol group. Power calculations indicated that 30 children in each group would detect a 25% reduction in incidence of postoperative vomiting in the first 4 h, with 85% power (α 0.05, two-sided). Statistical analysis was performed with Minitab. We used Fisher’s exact test and chi-squared test to compare PONV and fitness for discharge rates between groups, the logrank test for time to first analgesia and Kruskal–Wallis to compare maximum pain scores.
We enrolled a total of 65 children. Three children were excluded from analysis, two due to protocol violations and one for being cancelled after randomisation due to lack of theatre time. Thirty children were allocated to the Bristol group and 32 to the codeine group. Patients’ and surgical characteristics are shown in Table 2.
Table 2. Characteristics of patients receiving peri-operative analgesia with either fentanyl and diclofenac (Bristol group) or codeine (codeine group) for tonsillectomy or adenotonsillectomy. Values are mean (SD) or number.
(n = 30)
(n = 32)
*T; tonsillectomy, TA; adenotonsillectomy.
Operation duration; min
Overall, 13 (21%) children experienced PONV within 4 h, with 10 children (16%) experiencing vomiting and three (5%) nausea; there was no difference between groups (Table 3). There were no cases of retching. Maximum PONV scores recorded during the 4-h observation period are shown in Fig. 2. The three cases of nausea (score 1) resolved without intervention. All ten children who vomited received ondansetron, and this treatment was successful in nine cases, with one child having persistent nausea and vomiting despite treatment (score 3).
Table 3. Incidence of postoperative nausea and vomiting in children receiving intra-operative analgesia with fentanyl and diclofenac (Bristol group) or codeine (codeine group) for tonsillectomy. Values are number (proportion). No significant differences between groups.
(n = 30)
(n = 32)
No symptoms of nausea and vomiting
Time to first analgesic request was shorter in the Bristol group than the codeine group (p < 0.005, Fig. 3). However, there was no difference between groups in the overall number of children requiring at least one dose of postoperative analgesia (Table 4), and maximum pain scores recorded during the observation period were comparable (Bristol group median (IQR [range) 4.5 (3-5 [0-5]), codeine group 4.0 (2-5 [1-5]), p = 0.15). Twenty-five children (40%) recorded the maximum pain score of 5 at some point in the first four postoperative hours.
Table 4. Postoperative analgesic requirements in children receiving intra-operative analgesia with fentanyl and diclofenac (Bristol group) or codeine (codeine group). Values are number (proportion). No significant differences between groups.
(n = 30)
(n = 32)
Recovery to 2 h
2 to 4 h
Fourteen children (23%) were deemed not fit for discharge at 4 h, six from the Bristol group and eight from the codeine group (p = 0.42). For 10 children, this was due to inadequate analgesia, and for two children due to PONV. One child had not had a drink and one child was asleep for all assessments. No child included in this study experienced either primary or secondary haemorrhage.
Postoperative nausea and vomiting remains a significant problem for children undergoing tonsillectomy, and is the commonest reason for unplanned admission after day-case procedures. This study compared two different analgesic techniques, but did not demonstrate a difference in the PONV rates experienced. Our codeine group had a vomiting rate of 13%, which was much lower than the 25–35% vomiting rates at 4 h previously reported by Splinter and Roberts [12, 13]. This may have been due to the high proportion of children induced with propofol  or the absence of neuromuscular blockade and hence neostigmine from our protocol .
For the purpose of this study, we modified the Bristol regimen to include dexamethasone for its anti-emetic and co-analgesic properties. The use of prophylactic anti-emetics represents usual practice at our institution and in the UK . Techniques using codeine have an incidence of PONV of up to 76% [12, 16], and we felt that withholding prophylaxis from the codeine arm of the trial would be unethical, thus included it in both groups. Dexamethasone is well established as an effective anti-emetic for children undergoing tonsillectomy [12, 17, 18], and we anticipated that its use would reduce PONV in the Bristol group, even though the reported rate was extremely low anyway. However, our observed PONV rates were much higher than in the original study. The Bristol investigators had only two cases of nausea in the first 4 h, whereas we observed six cases of vomiting as well as two cases of nausea. It is possible that our more frequent observation in the early postoperative period identified cases of PONV that may have been missed by the Bristol investigators.
Routine administration of ondansetron may have reduced PONV rates further (5), but we reserved it for use as a rescue anti-emetic for several reasons. First, we expected a low incidence of PONV in the Bristol group and secondly, one aim of this study was to determine if modifying analgesic regimens could preclude the need for prophylaxis. Finally, ondansetron has been associated with concealed haemorrhage in children undergoing tonsillectomy and a review in adults suggests that it is more effective at treatment than prevention [1, 19, 20].
In this study, children in the Bristol group required analgesia earlier than those in the codeine group. The clinical significance of this is probably limited, as there was no difference between groups in either maximum pain scores experienced, or the total number of children who eventually required additional pain relief. The original Bristol study reported low postoperative pain scores that we did not replicate in our Bristol group. They found that 16% of children required analgesia within 2 h, while our comparable figure was 86%. A possible explanation for this discrepancy lies in the controlled nature of our trial. Children were assessed regularly by investigators with fixed criteria for intervention; these assessments often prompted analgesic administration. In contrast, the Bristol study was observational with pain scores recorded routinely only every 4 h and when additional analgesia was required, which was at the discretion of nursing staff. We do not feel that our incidence of early analgesia represents overtreatment; pain scores in our study were high with 40% of children recording the maximum score of 5 at some point in the first 4 h postoperatively. Our frequent and detailed observation probably reveals a real representation of pain experience with the techniques used.
All analgesics used in our study have documented efficacy in tonsillectomy patients. We used intravenous paracetamol, avoiding uncertainty about pre-operative administration or oral/rectal absorption. Intra-operative fentanyl has been shown under controlled conditions not only to provide good analgesia in tonsillectomy patients but also to be equivalent to morphine in terms of analgesic effect . Concerns about non-steroidal anti-inflammatory drugs and increased risk of postoperative haemorrhage do not seem borne out by meta-analysis ), and they are effective analgesics for adenotonsillectomy . Despite a reported 1.4% risk of primary and 1.8% risk of secondary haemorrhage , no child in our study experienced any bleeding complications.
Intramuscular codeine has been shown to be equi-analgesic to morphine in a randomised double-blinded study involving children undergoing tonsillectomy . The role of codeine in paediatric practice has been questioned, in part due to concerns over the ability of a proportion of children to metabolise the drug to its active components [25, 26]. However, in a recent pro-con debate concerning the role of codeine in paediatric practice, both sides accepted the value of codeine because of long-term familiarity with its use, and because it may be easily and safely prescribed for use at home .
We used a 4-h postoperative observation period because this timeframe has been identified as desirable by our ear-nose-throat surgeons for day-case tonsillectomy, based on the risks of postoperative haemorrhage, and the need for patients in the afternoon to be discharged at a reasonable time in the evening. Previous studies have monitored children for a minimum of 4 h, but do not report the actual time required to allow discharge [8, 12, 13]. Two studies in the UK with unplanned admission rates of 0% and 3.3% had observation periods of 6–8 h [5, 7]. To achieve this longer timeframe would require significant changes to our theatre and list organisation. We hoped to identify a technique that would permit safe discharge after day-case surgery without requiring such changes.
At 4 h, 23% of children did not meet our discharge criteria. In most cases, this was due to a pain score greater than one, although half of these children were noted to be clinically well enough for discharge (i.e. up and playing) despite pain scores of two. Clearly, flexibility is required in both the design and application of discharge criteria, and so children are not inappropriately judged unfit to go home. However, even allowing for this, 13% of children were not fit to leave at 4 h, which is too high for a day-case service.
In summary, children in the codeine group had a lower level of PONV compared with previous studies and required analgesia later than the Bristol group. As such, our codeine-based regimen may have a small advantage over the fentanyl and diclofenac regimen described by the Bristol investigators. However, at 4 h, a significant minority of children were still in pain or suffering from PONV and deemed unfit for discharge. Thus neither technique seems ideally suited to day-case tonsillectomy without incorporating a longer observation period.
The authors thank Dr Paul Davies, Birmingham Children’s Hospital, for his help with statistical analysis.
No external funding and no competing interests declared.