Authors’ Reply

Abstract

We thank Weinhandl1 for his comments regarding our analyses using the US Renal Data System of the association between gabapentin and pregabalin use and adverse outcomes among patients on hemodialysis.2 We acknowledge that we were limited in our ability to reliably capture benzodiazepine exposure with our available data, because benzodiazepines were not covered by Medicare Part D until 2013.3–5 Using more recent data from 2013, among a comparable cohort of 162,965 adult Medicare-covered patients on chronic hemodialysis with continuous Part D coverage during 2013, we found that the prevalence rates of gabapentin, pregabalin, and benzodiazepine use were 21% (n=33,900), 4% (n=7102), and 23% (n=37,112), respectively. The prevalence of benzodiazepine use is within the range that has been reported in prior studies involving patients on hemodialysis.6 Overall, the prevalence of concomitant use (defined as at least one instance of overlapping prescriptions in 2013) of gabapentin and benzodiazepine was 5.5% (n=8907), with a median overlap duration of 62 days (25th interquartile range, 28–152 days). Twenty-six percent of gabapentin users had concomitant benzodiazepine use, and 22% of gabapentin nonusers had concomitant benzodiazepine use. The prevalence of concomitant use of pregabalin and benzodiazepine was 1.3% (n=2054), with a median overlap duration of 57 days (interquartile range, 26–137 days). Twenty-nine percent of pregabalin users had concomitant benzodiazepine use, and 22% of pregabalin nonusers had concomitant benzodiazepine use. Gabapentin and pregabalin use had virtually no association with benzodiazepine use on the basis of the point-biserial correlation (Table 1). The point-biserial correlation is mathematically equivalent to the Pearson correlation and can be interpreted on the same scale (0–0.5 is weak, 0.5–0.8 is moderate, and 0.8–1 is strong). Although Pearson correlations are not usually used for binary variables, such as medication exposure, they are, in fact, what a regression program uses to adjust one predictor for another one with a correlation of zero, indicating no adjustment. Table 1. - Association between gabapentin and pregabalin use with benzodiazepine use Benzodiazepine Users (n=37,112), n (%) Benzodiazepine Nonusers (n=125,853), n (%) Correlation a Gabapentin users, n=33,900 8907 (26) 24,993 (74) 0.04 Gabapentin nonusers, n=129,065 28,205 (22) 100,860 (78) Pregabalin users, n=7102 2054 (29) 5048 (71) 0.03 Pregabalin nonusers, n=155,863 35,058 (22) 120,805 (78) aPoint-biserial correlation. Correspondingly, the effect of benzodiazepine use on event rates would need to be massive to explain away the gabapentin or pregabalin associations. Thus, confounding by concomitant benzodiazepine use was unlikely to have had a large effect on our results. The letter of Weinhandl1 also mentions the possibility of unmeasured confounding by opioid use, and we would like to note that we adjusted for concomitant use of opioids in our analyses. We appreciate the interest in our research, and thank you for the opportunity to elaborate on our methodology. Disclosures None.