Intravenous ketamine for CRPS: Making too much of too little?

Several publications and reports have recently advocated the use of intravenous ketamine infusions for the treatment of chronic pain. In Pain, Schwartzman and colleagues present the results of a randomised controlled trial where 19 patients with CRPS were treated with ketamine or placebo infusions [9]. The trial was stopped prematurely before half of the planned number of patients was included, reportedly because the non-trial experience with higher doses of ketamine suggested they would obtain greater efficacy in a new trial with higher doses. The paper concludes that intravenous ketamine administered in an outpatient setting resulted in statistically significant reductions in many pain parameters. Some pain clinicians regularly use ketamine infusions in their clinical practice, so it is important that this practice is proven in randomised placebo-controlled trials. The trouble is that randomised trials with methodological shortcomings can mislead us. In clinical research, about 15% of initial results are contradicted by further research; while in 15% much stronger effects are found [4]. When journals can publish articles with the title ‘‘Why most published research findings are false” we have a duty to be cautious, and take new findings with a pinch of salt [5]. The size of the pinch is inversely proportional to the quality of the study. This study was prematurely terminated, leaving only groups of 9 and 10 patients. No primary efficacy variable is clearly specified. The process of randomisation is not described, the trial is not convincingly blinded, and withdrawals are not described, so the analysis is per protocol, rather than intention to treat. The study would score only 2/5 on the Oxford quality scale [6], and 5/16 on Oxford Pain Validity Scale [11]; scores like this have been associated with systematic bias. In figuring out how the evidence stacks up for ketamine infusions, and how we should regard it, it behoves us to consider those influences likely to falsely enhance a study. Research [5] suggests that studies are more likely to be false when: