Reply to Dr Pan: Caution Is Warranted Before Off-label Use of Nerve Block Adjuvants With Marginal Benefit
We thank Pan et al1 for their thoughtful commentary regarding our systematic review and meta-analysis titled, “Perineural versus intravenous dexamethasone as an adjuvant for peripheral nerve blocks.”2 In their editorial letter, they update the main outcomes of our meta-analysis with the newly published trial by Sakae et al.1,3 They conclude that “Perineural dexamethasone can induce longer duration of [a] nerve block and provide better postoperative analgesia than intravenous dexamethasone in patients receiving peripheral nerve blocks.”1 Although their findings are generally consistent with our original conclusions, we would like to consider some limitations in their analysis and emphasize that the current evidence base is not strongly in favor of the widespread adoption of off-label perineural dexamethasone.1,2
To allow direct comparison to our original meta-analysis, we repeated our systematic search and meta-analysis to include any new trials up to August 17, 2017. We also manually searched the proceedings of the most recent American Society of Regional Anesthesia and Pain Medicine, European Society of Regional Anesthesia, Canadian Anesthesiologists' Society, and American Society of Anesthesiologists meetings for additional studies.
Our repeated and updated database search revealed 2 new relevant studies: the trial by Sakae et al3 and a large 280-patient study by Holland and colleagues.4 Data from these trials were used to update the meta-analysis with our previously reported methodology.2 Holland et al4 used a 1:1:1:1 factorial design comparing dexamethasone dose (4 vs 8 mg) and route (intravenous vs perineural). We used their 4 mg intravenous versus perineural dexamethasone data for our updated analysis, assuming equal groups of 70 patients each.4 In addition, we also maintained the same random-effects model for pooling the data, given the known clinical heterogeneity among the original pooled randomized controlled trials (eg, differing doses of dexamethasone and different types of nerve blocks) and statistical heterogeneity (eg, I2 value of 68.3% in our original analysis for duration of sensory block).2 Although Pan et al1 likely used a random-effects model as suggested by the labels in their Figure 1, they did not comment upon the high heterogeneity of the evidence base, which is an important limitation of any meta-analysis. We will focus on commenting on the updated duration of sensory blockade and pain score analyses in relation to the presentation by Pan et al.1
The updated meta-analysis revealed increased duration of sensory block with perineural dexamethasone compared with intravenous administration (3.73 hours; 95% confidence interval [CI], 2.00–5.46 hours; I2 = 69.4%; P < 0.001). This magnitude of effect is nearly identical to that of our original meta-analysis (3.77 hours; 95% CI, 1.87–5.68 hours).2 Notably, the statistical heterogeneity remains high, and therefore the updated results should still be interpreted with caution.
When considering pain scores, one should perform a statistical correction to take into account any repeated statistical tests over time and avoid type I error. In our original meta-analysis, we chose a Bonferroni correction because it is simple and statistically conservative. The Pan et al1 editorial letter does not mention correction for multiple comparisons. Furthermore, in contrast to what Pan et al1 state, our original article did report a difference in pain scores at 24 hours, with slightly reduced pain in the perineural dexamethasone group compared with intravenous (10-point visual analog scale score, 1.09; 95% CI, 0.09–2.08; P = 0.032). This difference was not statistically significant after considering a Bonferroni corrected α of 0.0125. With respect to the updated pain score analysis, patients receiving perineural dexamethasone had lower pain scores at 12 hours (−0.76; 95% CI, −1.38 to −0.14; P = 0.02; I2 = 30.5%; n = 3) and 24 hours (−1.15; 95% CI, −2.03 to −0.27; P = 0.