Epidural analgesia: the need for more evidence

We read the editorial ‘Epidural analgesia: First do no harm’ [1], where the authors question the routine use of perioperative epidural analgesia for gastrointestinal surgery, with interest. The evidence from the meta-analysis demonstrating the benefits of epidural analgesia [2, 3] have not been supported by the single large randomised controlled study (MASTER study) [4]. The authors quite correctly state that ‘decisions regarding the use of an epidural in any individual patient, and what we should and should not tell our patients, need to be framed from the best available evidence and not the most available’. Unfortunately, the authors go on to overstate the value of the evidence from the MASTER trial. They state ‘mortality at 30 days was low in both the groups with a non-significant trend to an increase with epidural use (4.3% control vs. 5.1% epidural group)’. This is disingenuous at best, and at worst shows a complete misunderstanding of statistics and chance. The MASTER trial was designed to detect an absolute risk reduction of 10% with a control event rate of 50% and a power of 80% for the combined outcome of mortality and major postoperative morbidity. 57.1% of the patients experienced an adverse event in the epidural group compared to 60.7% in the control group, a difference of 3.6%, which was statistically non–significant. All the individual endpoints measured occurred in less than 50% in both the groups. Hence, the MASTER trial was significantly underpowered to detect clinically significant differences in the postoperative outcomes. Changes in the last two decades in anaesthetic and surgical practices have reduced perioperative morbidity and mortality [5, 6] and may have obscured the differences attributable to epidural anaesthesia/analgesia in these settings. The authors of the original study have suggested that a trial of 6 000 patients at high risk would be needed to give 80% chance of declaring statistically significant an absolute difference of 3.6% in the rate of death or major complications. A randomised controlled trial of this size and complexity in never likely to be achieved. The biggest criticism of the MASTER trial design is the primary use of ‘intention to treat’ analysis, despite a staggering 49.7% of patients in the epidural group not complying with the study protocol [7]. To pursue this statistical technique to its logical end, if none the patients randomised to receive epidural analgesia actually did so, then the intention to treat analysis would find no differences in the outcome between the groups. ‘Per protocol’ analysis would clearly have been more appropriate. The MASTER trial has been well criticised, elsewhere for protocol design, timeliness of research and statistical design [8]. Thus, the conclusions of the MASTER trial must be treated with extreme caution. The editorial [1] states that ‘if we look critically at the available evidence…the use of epidural analgesia has little benefit over standard opioid techniques and retains the potential to cause harm’. This statement implies that the alternatives do not have the potential to cause harm. The epidural debate seldom focuses on the incidence of serious harm caused by parenteral opioid therapy. This incidence may be unidentified, but with hypoxia at the top of serious adverse effects, it is clear that opioid therapy can be harmful. Surely, a more balanced conclusion ought to be ‘if we look critically at the available evidence we find it inadequate to draw a sound scientific conclusion’. Lack of evidence of benefit is not evidence for lack of benefit.