Participants with mild, moderate, or severe pain following total hip arthroplasty. A sub-study of the PANSAID trial on paracetamol and ibuprofen for postoperative pain treatment

Trial overview and main results

PANSAID was a multi-centre, randomised, controlled, parallel, four-group clinical trial investigating the effects of different doses and combinations of paracetamol, ibuprofen and placebo in patients undergoing total hip arthroplasty (THA). The PANSAID trial was conducted at six Danish hospitals from December 2015 – January 2018. The protocol [13], the statistical analysis [12] and the main results [11] has been published previously. The trial protocol was approved by the Biomedical Research Ethics Committee of Region Zealand (SJ-462) and was conducted in accordance with the Declaration of Helsinki. PANSAID was monitored by Good Clinical Practice Units at Odense and Copenhagen University Hospitals, and the trial was registered at ClinicalTrials.gov (NCT02571361), EudraCT (2015-002239-16) and the Danish Data Protection Agency (REG-33-2015). All participants gave written informed consent prior to enrolment.

A total of 556 participants having primary planned THA where randomised into four intervention groups in a 1:1:1:1 ratio, stratified by site. The intervention groups were: group PCM + IBU (paracetamol 1000 mg + ibuprofen 400 mg), group PCM (paracetamol 1000 mg + matching placebo), group IBU (ibuprofen 400 mg + matching placebo), and group HS-PCM + IBU (paracetamol 500 mg + ibuprofen 200 mg).

The medication was taken orally 1 h before surgery and administered a total of four times (at 6-h intervals) during the first postoperative day. Throughout the trial, participants, caregivers, physicians, investigators, statisticians, and conclusion drawers were blinded to the intervention.

The primary analysis was based on the modified intention-to-treat (mITT) population — defined as participants who were randomised and underwent surgery. The two co-primary outcomes were 24-h morphine consumption using patient-controlled analgesia, and proportion of participants having one or more serious adverse events within 90 days following surgery (SAE were defined as any untoward medical occurrence that results in death; is life threatening; requires hospitalization; or results in significant or persistent disability or incapacity; birth defects; or a medical intervention to prevent one of the before-mentioned outcomes). The outcome of morphine consumption was in pairwise comparisons between the four groups (six comparisons) and the outcome of SAEs was a comparison between participants randomised to ibuprofen (groups PCM + IBU, IBU, and HS-PCM + IBU) and participants randomised to paracetamol (group PCM).

The main findings were that combining paracetamol 1 g and ibuprofen 400 mg resulted in reduced morphine consumption compared with each drug alone on the first postoperative day after total hip arthroplasty (16 mg, 99.6% CI: 6.5–24, p<0.001 compared with paracetamol, and 6 mg, 99.6% CI: −2 to 16, p=0.002 compared with ibuprofen). The reduction for the combination group vs. ibuprofen alone, was however, lower than the predefined minimum clinical important difference of 10 mg (10 mg was pragmatically chosen because it is commonly used in acute pain trials). The combination of lower dosages (group HS-PCM + IBU) resulted in reduced morphine consumption compared with paracetamol 1 g only, and comparable morphine consumption to ibuprofen 400 mg only. Regarding harm, the proportion of participants with one or more SAE was numerically higher in participants randomised to ibuprofen (15%) compared with participants randomised to paracetamol only (11%), although not statistically significant (relative risk 1.44, 97.5% CI: 0.79–2.64, p=0.18).

The present sub-study

The present sub-study was planned during recruitment of participants to the PANSAID trial and before any outcome data were explored. A protocol synopsis including choice of outcomes, statistical analysis plan, and covariates for the prediction model was made publicly available at the trial webpage (www.pansaid.dk) on 29 September 2017 [10], and was also outlined in the protocol and at clinicaltrials.gov (NCT02571361). During the re-analysis of the trial data for this article we further decided to include the composite outcome score (Silverman’s Integrated Approach (SIA)) of opioid use and pain scores and, therefore, all SIA-scores are regarded as post hoc. The sub-study included all participants in the mITT population. The primary outcome was proportion of participants experiencing mild pain 6 h postoperatively during mobilisation in the four randomisation groups. This manuscript adheres to the applicable CONSORT guideline.

Participants’ pain

All participants in the trial had their pain intensity measured on a Visual Analogue Scale (VAS; 0 mm, no pain and 100 mm, worst imaginable pain) 6 and 24 h postoperatively at rest and during 30° active hip flexion on the operated side. Mild pain was defined as VAS scores 0–30 mm, moderate pain as VAS scores 31–60 mm, and severe pain as VAS scores 61–100 mm.

Silverman and colleagues suggested integrating the Visual Analogue Scale score and rescue morphine to a composite score to circumvent statistical issues of mass significance [5]. To calculate the SIA score we calculated the mean VAS at rest and during mobilisation (Mean-VAS_{rest, 0–24 h} and Mean-VAS_{mobilisation, 0–24 h}, respectively). Mean-VAS_{rest, 0–24 h}, Mean-VAS_{mobilisation, 0–24 h}, and morphine requirement for each participant were ranked. A mean rank for all participants (across all randomisation groups) was calculated and the differences between the mean rank and the individual participant’s rank were expressed as percentages. The sum of the percentage (percentage-for-VAS_{rest, 0–24 h} + percentage-for-morphine requirement, and percentage-for-VAS_{mobilisation, 0–24 h} + percentage-for-morphine requirement) represent the SIA score at rest and during mobilisation, respectively. Hence, the range of the SIA score is from −200% to +200%. Negative values indicate lower summed pain scores and morphine requirement compared with the entire population.

To assess the participants’ pain, we investigated the distribution of participants’ pain (i.e. proportions of participants having mild, moderate or severe pain), and the proportions of participants having a SIA score of −200 to −100%, −100 to 100% and 100 to 200%.

Preoperative prediction of postoperative pain

To identify potential risk factors for severe postoperative pain we dichotomised pain during 30° flexion of the hip 24 h postoperatively to severe (VAS 61–100 mm) vs. mild-moderate (VAS 0–60 mm). The potential risk factors were chosen based on prior literature suggesting association with postoperative pain [14], [15], [16]. We chose the following potential risk factors: age (continuous), sex (female vs. male), BMI (continuous), ASA-score (dichotomised to I-II vs. III), use of analgesic medications prior to the surgery (none, non-opioid, weak opioid), type of anaesthesia (GA vs. spinal), type of THA (cemented/hybrid vs. no cement), surgical approach (lateral vs. posterior), and duration of surgery (continuous).

Statistical analysis

We assessed the data for normality by inspecting qq-plots and histograms. Data are presented as numbers and percentages, mean and SD, median and interquartile range as appropriate. Statistical tests were chi-squared test, Kruskal Wallis test, ANOVA, and Mann-Whitney U-test as appropriate.

We calculated number needed to treat (NNT) to achieve mild pain (VAS 0–30 mm) between two randomisation groups. NNT was calculated from the logistic regression model (with site as a covariate), if the pairwise comparison was statistically significant, using an online calculator [17].

For the logistic prediction models, we used both univariate and multiple regression. In the multiple logistic regression model, we used backwards stepwise regression removing one covariate (with the highest p-value) at the time. All models had site and randomisation group as co-variates. As sensitivity analyses, all models were done without dichotomising the dependent variable.

The level of significance was predefined and set to 0.01 (a modified Bonferroni correction) for the pairwise comparisons of participants experiencing mild pain to avoid issues with mass significance. For all other outcomes the level of significance was 0.05. Confidence intervals were 99 and 95% accordingly. The Rstudio version 1.1.383 was used for all analyses.

Participants’ pain

During mobilisation at 6 h postoperatively, 33% (95% CI: 26–42) (Group PCM + IBU), 28% (95% CI: 21–37) (Group PCM), 23% (95% CI: 17–31) (group IBU), and 19% (95% CI: 13–27) (Group HS-PCM + IBU) of participants had mild pain (VAS 0–30 mm). Proportions of participants experiencing mild, moderate, and severe pain at rest and at 24 h postoperatively can be seen in Figure 1 and Table S1.

Figure 1:

The proportion of patients achieving mild pain, moderate pain and severe pain during 30° flexion of the hip and at rest in each group 6 and 24 h postoperatively.

Green, mild pain (VAS 0–30 mm); yellow, moderate pain (VAS 31–60 mm); red, severe pain (VAS 61–100 mm).

In pairwise comparisons of participants experiencing mild pain, there were two statistically significant comparisons (Group PCM + IBU vs. Group HS-PCM + IBU at 6 h postoperatively during mobilisation: OR for having mild pain was 2.22 (95% CI: 1.27–4.00; p=0.0006; NNT=7; 95% CI: 3–26); and Group PCM + IBU vs. Group PCM at 24 h postoperatively at rest: OR for having mild pain was 2.63 (95% CI: 1.43–5.00, p=0.002; NNT=6, 95% CI: 5–14)).

Median SIA scores at rest were: Group PCM + IBU: −36% (IQR: −124 to 12), Group PCM: 41% (IQR: −31 to 99), Group IBU: −2% (IQR: −52 to 70), and Group HS-PCM + IBU: 0% (IQR: −70 to 56) (overall difference: p=0.0001). Median SIA scores during mobilisation were: Group PCM + IBU: −48% (IQR: −112 to 31), Group PCM: 40% (IQR: −31 to 97), Group IBU: −5% (IQR: −57 to 67), and Group HS-PCM + IBU: 6% (IQR: −70 to 74) (overall difference: p=0.0001).

The distribution of SIA scores of participants in the four intervention groups are presented in Figures 2 and 3. In general, more participants in Group PCM + IBU had SIA-scores<−100% than all other groups, both at rest (Figure 2) and during mobilisation (Figure 3).

Figure 2:

The Silverman integrated approach combining pain at rest and morphine consumption.

Red, paracetamol 1 g + ibuprofen 400 mg; green, paracetamol 1 g; blue, ibuprofen 400 mg; purple, paracetamol 0.5 g + ibuprofen 200 mg. Each dot represents a single patient.

Figure 3:

The Silverman integrated approach combining pain during mobilisation and morphine.

Red, paracetamol 1 g + ibuprofen 400 mg; green, paracetamol 1 g; blue, ibuprofen 400 mg; purple, paracetamol 0.5 g + ibuprofen 200 mg. Each dot represents a single patient.

Prediction

One hundred and forty-two of five hundred and thirty-three participants had severe pain 24 h postoperatively during mobilisation. Results from the prediction models are found in Table 2. The surgical approach was found as a statistically significant covariate for predicting severe pain in the univariate analysis (OR 0.39, 95% CI: 0.25–0.62, p<0.001 for lateral approach compared with posterior approach, lateral approach less likely to have severe pain), but because of collinearity and perfect prediction, the surgical approach could not be evaluated in the multivariate analysis (one of the two surgical approaches were used exclusively at one hospital and the other approach was used exclusively at all other hospitals). Use of non-opioid analgesics prior to surgery compared with no use of analgesics resulted in lower risk of severe pain in both univariate and multivariate analyses and was the only statistically significant covariate after backwards elimination (non-opioid before surgery: OR 0.50, 95% CI: 0.29–0.82, weak opioid before surgery 0.56, 95% CI: 0.28–1.16, reference no analgesic before surgery, overall p-value=0.02).