With ethics committee approval, we recruited and followed up patients between March 2005 and February 2008. Trial registration: ANZCTR.ORG.AU (identifier: ACTRN12606000291583).
This study was conducted at a publicly funded teaching hospital and a private day-surgical clinic in metropolitan New Zealand.
We included adults undergoing extraction of at least one lower wisdom tooth with or without one or more upper wisdom teeth by one of three participating surgeons. We excluded patients if they were under 16 yr old; weighed <50 kg; had taken any NSAID (other than aspirin in a dose of 150 mg daily or less) within 24 h of the operation; had taken acetaminophen or acetaminophen containing medicines within 12 h of the operation; were taking an angiotensin-converting enzyme inhibitor, warfarin, steroid (other than interoperative dexamethasone), or any immunosuppressive drug; were intolerant to any NSAID or acetaminophen; were suffering from a severe local infection; had a history of peptic ulceration, asthma, or severe haemopoetic, renal or hepatic disease; were participating in the investigation of another experimental agent; or if the clinician believed for any other reason that participation in the study might not be in their best interests.
Randomization and blinding
Tablets of identical appearance, packaging, and dosage instructions were provided in each of the following formulations: (i) acetaminophen 500 mg+ibuprofen 150 mg per tablet (Maxigesic®; Sigma Laboratories, Nashik, India which was MHRA approved for manufacturing pharmaceuticals under GMP); (ii) acetaminophen 500 mg per tablet; or (iii) ibuprofen 150 mg per tablet.
Patients were first approached by the surgeon and then by the study nurse. They were given written and verbal information about the study, and invited to participate. If they consented, patients were then randomized into one of the three study groups in a sequential order to receive one of these formulations, in blinded packs. The randomization sequence was computer generated by the study statistician as a 1:1:1 allocation ratio to the three treatments in a sequence of permuted blocks with stratification for anaesthetic type (local or general) and study centre. Stratification by anaesthetic type ensured a balance between treatments in terms of the number of teeth extracted, as most patients having more than two teeth extracted have a general anaesthetic. Only the statistician had access to the schedule of patient numbers by drug allocation. Participants and investigators were blinded and the randomization code was not broken until the final database had been checked and locked.
Participants were asked to take two tablets of the study medication before operation (as close as possible to the start of surgery) and then 4 times a day (as close as possible to 6 hourly) for up to 48 h after surgery. All participants were given bupivacaine local anaesthetic blocks by the surgeons. For those participants undergoing general anaesthesia, this was induced with propofol and maintained with isoflurane and nitrous oxide in oxygen. Monitoring was in accordance with the guidelines of the Australian and New Zealand College of Anaesthetists.16
All extractions were carried out by one of three surgeons, each using his normal technique.
If participants required additional postoperative pain relief while in hospital, a rescue dose of fentanyl 10 µg was given i.v., as required. After discharge to home, codeine was provided (again, as rescue medication) in 30 mg tablets, one to two to be taken as needed up to 4 hourly.
Participants were asked to rate their pain on 100 mm visual analogue scales (VAS), printed one per double page in a booklet that they took home. Ratings were requested at baseline (immediately before administration of the first dose of study medication); after operation (once the participants were sufficiently awake to respond); and 1–2 hourly thereafter, while awake, for 48 h. The study nurse maintained contact with participants by telephone to facilitate compliance with data collection and the return of diaries.
The primary outcome measure was the area under the curve (AUC) of these VAS ratings divided by time, at rest and on activity. The AUC was divided by the period of the completed assessments to adjust for the fact that some patients recorded pain for shorter periods than others. This calculation in effect produces a measure of average pain intensity over the study period.
Secondary efficacy outcome measures were: a categorical global pain rating by the participants, taken at the end of the study period; rescue analgesia consumption over the study period; a categorical global rating of nausea by the participants, taken at the end of study period; the number of episodes of vomiting over the study period; and a rating of sleep disturbance on a 100 mm VAS assessed after each night during the study period. In addition, participants were asked to rate their experiences of participating in the study.
Sample size estimation
We obtained blood samples from the 38 participants undergoing general anaesthesia in order to have evaluable pharmacokinetic data for at least 30 patients. The first sample was obtained 30 min after the first dose of study medication, the second sample at the end of anaesthesia, and additional one or two samples after operation in hospital. The plasma concentration of acetaminophen and ibuprofen were measured by the sponsor and used to form individual time–concentration profiles. The analytical method used an HPLC-DAD (Diode Array Detector) assay for the simultaneous determination of acetaminophen and ibuprofen in plasma. Precision and accuracy for acetaminophen and ibuprofen assay were validated over the concentration range 0.5–50 µg ml−1 for both drugs. The intra- and inter-batch precision of the assays at low, medium, and high concentrations of acetaminophen and ibuprofen varied from theoretical values by <15%. The lower limit of quantification for each drug was 0.5 µg ml−1. The sponsor monitored all data collected during the study and queries and corrections were made when any inaccuracies or inconsistencies were identified.
Sample size estimation
We estimated that 120 participants (40 per group) in the intention-to-treat (ITT) population would provide 80% power to detect differences between the groups of 9 (sd
14) mm in our primary endpoint for resting assessments and 13 (sd
21) for measures during activity,10,17
with a one-sided type I error rate of 5%. These differences equate to ~25%. Differences of this magnitude were considered clinically important and comparable with differences typical of previous published studies.10
The data were analysed using SPSS version 15.0 (SPSS Inc., Chicago, IL, USA). Efficacy analyses were conducted on an ITT basis with the additional provision that there were at least three VAS measurements over at least 12 h available to calculate the primary endpoint. All participants who were randomized into the study were included in the safety evaluations. As the first dose of study medication was taken before operation while under the supervision of the surgeon, all randomized patients took at least a single dose of study medication. A last observation carried forward approach was used for those subjects who left the study prematurely for non-AUC based variables.
We compared the primary endpoint between the combination group and each of the acetaminophen and ibuprofen arms, at rest and on activity, using a general linear model (GLM) which included terms for treatment, the centre, and anaesthetic stratum. Additionally, to confirm the consistency of the treatment effects across strata, the stratum treatment interaction terms were tested and included in the final model. The analysis was also checked with number of teeth extracted as an additional factor. Continuous secondary efficacy endpoints were tested for significance using the same models as used for the primary endpoint.
A one-tailed P≤0.05 was pre-specified to indicate statistical significance. We required a statistically significant result favouring the combination from each of the two planned comparisons with the constituents to define superiority for either rest or on activity measures. We used one-tailed tests as there seemed no theoretical or empirical basis for expecting that combining these analgesics could result in a reduction in efficacy, and because the requirement for each of two comparisons to be significant at P≤0.05 is stringent. Secondary categorical efficacy endpoints were compared between the groups using χ2 tests and Mann–Whitney U-tests as appropriate.
We used non-linear mixed effect models (NONMEM VI, Globomax LLC, Hanover, MD, USA) to estimate population pharmacokinetics, with a Compaq Digital Fortran Version 6.6A compiler on an Intel Celeron 333 MHz CPU (Intel Corp., Santa Clara, CA, USA) under MS Windows XP (Microsoft Corp., Seattle, WA, USA). This model allows assessment of inter-individual variability, covariance between pharmacokinetic parameters and residual error. We judged the quality of fit of the pharmacokinetic model to data using the NONMEM objective function examination of plots of observed vs predicted concentrations and visual predictive checks.