Description of studies
One hundred and seven published articles were identified from the search as potential double blind, RCTs of oral ibuprofen in postoperative pain. Of these, three reports did not state whether they were randomised and/or double blind, so had to be excluded and three were abstracts.
Of the remaining 101 RCTs, 67 were excluded. In 28 studies patients did not have baseline pain of at least moderate severity, this includes the studies which administered the intervention pre-operatively. This is methodologically important as testing the intervention on patients with established pain ensures adequate sensitivity (Lasagna 1962). Nineteen studies did not have a placebo control and in six studies the participants were children. In three studies pain scales other than those described in the selection criteria were used and in two, data was provided for less than four hours. As the method for generating dichotomous data has only been verified for the most commonly used pain scales, applied over four to six hours, other outcome measurements cannot be legitimately used with this technique. Three studies provided multiple dose data without analysing the first dose separately and two studies did not measure pain directly. In two studies patients continued to complete the diaries after re-medicating with rescue analgesics thus invalidating the results and one trial presented the data from a cross-over design as one data set. One study used a controlled release formulation of ibuprofen. Thirty four reports of 35 studies met our inclusion criteria and were included in the analysis. One was author contacted and kindly provided information on the number of patients in each treatment arm (Laska 1986).
Although the search identified nearly 2000 trials the majority were in chronic pain or the intervention was administered before the patient experienced pain. Over 500 reports were found of trials involving rectal, intravenous and intramuscular diclofenac. Predominantly these reports were not placebo controlled, did not use standard pain outcome measures or the intervention was administered before the patient experienced any pain.
Only 47 published articles were identified as potential double blind, RCTs of oral diclofenac in postoperative pain. Of these, one report failed to state whether it was randomised, so had to be excluded and five were abstracts.
Of the remaining 41 RCTs, 35 were excluded. In 13 studies patients did not have baseline pain of at least moderate severity, this includes the studies which administered the intervention pre-operatively. Sixteen studies did not have a placebo control and one study only reported the first hour's data for the placebo group. Five reports identified by the search could not be obtained despite attempts to contact the authors, ordering through the British Library and help from the librarians at Novartis and Knoll pharmaceuticals (Joubert 1977, Vigneron 1977, Carlos 1984, Frezza 1985, Iqbal 1986).
Six reports of six studies met our inclusion criteria and were included in the analysis. One author was contacted for information on the number of patients in each treatment arm (Nelson 1994a); they were unable to provide this information and so an equal split of 50 patients per group was assumed.
Risk of bias in included studies
Each report was independently scored for quality by two of the authors using a three-item scale with a maximum score of five (see below) (Jadad 1996b); all of the authors then met to agree a 'consensus' score for each report.
The quality scores (QS) for individual trials are reported in the notes section of the included studies table. These scores were not used to weight the results in any way.
The scale used is as follows:
Is the study randomised ? If yes - 1 point
Is the randomisation procedure reported and is it appropriate ? If yes add 1 point, if no deduct 1 point
Is the study double blind ? If yes then add 1 point
Is the double blind method reported and is it appropriate ? If yes add 1 point, if no deduct 1 point
Are the reasons for patient withdrawals and dropouts described ? If yes add 1 point
Effects of interventions
Ibuprofen versus placebo
Thirty four reports of 35 trials fulfilled our inclusion criteria; 2214 patients were given ibuprofen and 1377 placebo.
Twenty five trials (71%) investigated oral surgery pain (predominantly third molar extraction with bone removal), five trials investigated postpartum pain (predominantly episiotomy and Caesarean section) and four trials investigated postoperative pain (one tonsillectomy, one inguinal hernia, one orthopaedic surgery and one general surgery). Median quality score was 4 (2 to 5), 70% scoring 4 or 5.
One trial (Parker 1986) used a syrup formulation of ibuprofen, two trials (Seymour 1991(study1), OR CHECK Seymour 1991(study2), Seymour 1996) used soluble ibuprofen and liquid in gelatin capsules, two trials (Nelson 1994a OR CHECK Nelson 1994b; Mehlisch 1995) used ibuprofen lysine and two trials (Laveneziana 1996; Pagnoni 1996) used soluble ibuprofen arginine. When these more readily absorbed formulations were pooled and the results compared with those of the standard tablet formulation, no difference was found in relative benefit or NNT. The NNT for a single dose of ibuprofen 400 mg standard formulation tablets (1356 patients) compared with placebo was 2.8 (2.5 to 3.1) and for ibuprofen 400 mg soluble formulations (250 patients) the NNT was 2.5 (2.1 to 3.1). All formulations were therefore pooled for the overall analysis.
The pooled relative benefits for ibuprofen 100 mg, 200 mg, 400 mg and 600 mg were significantly different from placebo, as were the single data sets for ibuprofen 50 mg and 800 mg. The relative benefit (RB) and number-needed-to-treat (NNT) for at least 50% pain relief over 4 to 6 hours compared with placebo in pain of moderate to severe intensity is given below - numbers in brackets are the 95% confidence intervals:
50 mg RB 30 (1.8 to >480) NNT 3.6(2.5 to 6.1) No. of patients: 108
100 mg RB 18 (2.5 to >135) NNT 5.6 (3.8 to 9.9) No. of patients: 186
200 mg RB 4.7 (3.2 to 6.9) NNT 3.3 (2.8 to 4.0) No. of patients: 707
400 mg RB 3.4 (3.0 to 3.9) NNT 2.7 (2.5 to 3.0) No. of patients: 2817
600 mg RB 2.0 (1.6 to 2.6) NNT 2.4 (1.9 to 3.3) No. of patients: 203
800 mg RB 2.6 (1.8 to 4) NNT 1.6 (1.3 to 2.2) No. of patients: 76
Drug-related study withdrawals occurred rarely. One study (Fricke 1993) had one withdrawal on ibuprofen for vomiting which the authors did not attribute to the medication. One study (Seymour 1996) had three withdrawals on ibuprofen and one on placebo for vomiting soon after ingestion of study drug. Another study (Parker 1986) had one patient who withdrew on placebo. The studies reported a variable incidence of minor adverse events which were all mild and transient, with no difference in incidence between ibuprofen and placebo.
Relative risk estimates were calculated for ibuprofen 400 mg versus placebo for the most commonly reported adverse effects ('drowsiness'/'somnolence', 'dizziness', headache, nausea and vomiting). This pooled data showed no significant difference between ibuprofen and placebo for any adverse effect except drowsiness/somnolence which was significantly different, with an NNH of 19 (12 to 41).
Diclofenac versus placebo
Six trials fulfilled our inclusion criteria (528 patients were given diclofenac and 312 placebo). Five trials (83%) investigated oral surgery pain (third molar extraction with bone removal) and one pain following gynaecological surgery. Doses of diclofenac were 25 mg in one trial, 50 mg in six and 100 mg in three. Three trials used an immediate release diclofenac potassium formulation (Hebertson 1994; Mehlisch 1994; Nelson 1994b) and two used dispersible diclofenac (Ahlstrom 1993, Bakshi 1994). One trial used both the immediate release and enteric coated formulations (Bakshi 1992). To ensure comparability only the data from the immediate release formulation were included. Median quality score was 3 (2 to 4), four trials scoring 3.
Pooled relative benefit for diclofenac versus placebo was significant at all doses. The relative benefit (RB) and NNT for at least 50% pain relief over 4 to 6 hours compared with placebo in pain of moderate to severe intensity is given below - numbers in brackets are the 95% confidence intervals:
25 mg RB 5.8 (2.1 to 15) NNT 2.6 (1.9 to 4.5) No. of patients: 100
50 mg RB 3.4 (2.7 to 4.4) NNT 2.3 (2.1 to 2.7) No. of patients: 636
100 mg RB 7.7 (4.5 to 13) NNT 1.8 (1.5 to 2.1) No. of patients: 308
Drug-related study withdrawals occurred rarely. One study (Hebertson 1994) had one withdrawal on diclofenac 100 mg for nausea and vomiting. The studies reported a variable incidence of minor adverse events none of which were serious and with no difference in incidence between diclofenac and placebo.
Only two studies (Bakshi 1992, Nelson 1994b) provided data which could allow calculation of the relative risk associated with the most commonly reported adverse effects. Diclofenac 50 mg was not significantly different from placebo for 'dizziness' (97 patients), headache (100), nausea (97) or vomiting (97). Neither of the studies reported 'drowsiness' or 'somnolence'.
Diclofenac versus ibuprofen
(Numbers in brackets are the 95% confidence intervals)
There were two direct comparisons of diclofenac 50 mg and ibuprofen 400 mg (Ahlstrom 1993, Bakshi 1994). Both trials were in dental pain (third molar removal); 118 patients received diclofenac and 112 ibuprofen. There was no significant difference between diclofenac 50 mg and ibuprofen 400 mg (relative benefit 1.0 (0.9 - 1.2)).
Sensitivity analyses for trial size and quality score
(Numbers in brackets are the 95% confidence intervals)
There were 34 studies of ibuprofen 400 mg. The NNT for trials with 30 or fewer patients given ibuprofen was 2.4 (2.0 to 3.0; 206 patients), compared with 2.8 (2.5 to 3.1; 1400 patients) for trials with more than thirty patients on ibuprofen (z = 1.1). The NNT for trials with quality score of less than 4 was 2.5 (2.2 to 3.0; 449 patients), compared with 2.8 (2.5 to 3.1; 1157 patients) for trials with quality score of 4 or 5 (z = 1.2).