Thirteen randomised trials in 1,176 patients tested perioperative duloxetine as part of multimodal analgesia for total hip and knee arthroplasty.
Pain fell at every timepoint measured. At 24 hours, resting pain was 5.09 mm lower on a 100 mm scale and ambulatory pain 4.55 mm lower. At two weeks the differences were 6.48 mm and 8.93 mm. At three months or more, 2.79 mm and 3.76 mm. Every one of those reached statistical significance, and every one sits below the 10 to 20 mm usually taken as the smallest difference a patient can detect. The authors say so themselves, describing the effects as sub-threshold.
Opioid sparing was found at 48 and 72 hours, but the 48-hour estimate had an I-squared of 100% and an interval from 44 to 1 morphine milligram equivalents, which the authors flag as needing substantial caution. That is not a usable number.
On tolerability, duloxetine reduced nausea and vomiting and increased drowsiness with a risk ratio of 1.88. And a leave-one-out analysis is the detail worth carrying: excluding a single high-risk trial removed the significance of the long-term pain benefit, the nausea reduction and the fatigue reduction. Three of the headline findings rest on one study of questionable quality.
- Do not add duloxetine expecting a pain difference the patient will notice: effects were below the detectable threshold.
- Counsel about drowsiness if it is used — risk ratio 1.88, and the most reliable of the adverse findings.
- Treat the opioid-sparing figure at 48 hours as uninterpretable given I-squared of 100%.
- Note the leave-one-out result: three findings depended on a single high-risk trial.
- The nausea reduction is a genuine possible benefit but shares that fragility.
The statistics, in plain English
This is the clearest example you will see of statistical significance and clinical significance being different things. A 5.09 mm reduction on a 100 mm visual analogue scale has a p value of 0.006, so it is unlikely to be chance — and it is well under the roughly 10 to 20 mm that patients report as a noticeable change. Both statements are true at once, and the second is the one that decides prescribing. I-squared of 100% on the 48-hour opioid figure means the trials disagreed completely; pooling them produces a number with no meaning. And the leave-one-out sensitivity analysis is the most valuable thing in the paper: it asks what happens if you drop each study in turn, and three findings did not survive dropping one.
Read the rest in the app
You have read your two free briefings this month. The app carries all 27 specialties, every morning, free — and this finding is waiting in it.
Scan to keep reading on your phone. No account needed to start.
Tomorrow morning, before your first patient
One edition a day for anaesthesiology, written by the desk, every claim tied to its paper. Six minutes.
Get the app — free