Systematic Review and Meta-Analysis of Pharmacological Therapies for Painful Diabetic Peripheral Neuropathy

Background Painful diabetic peripheral neuropathy (pDPN) is prevalent among persons with diabetes and increases over time. Published guidelines recommend a number of medications to treat this condition providing clinicians with a variety of treatment options. This study provides a comprehensive systematic review and meta‐analysis of published pharmacologic therapies for pDPN. Methods The published literature was systematically searched to identify randomized, controlled trials of all available pharmacologic treatments for pDPN (recommended or nonrecommended) reporting predefined efficacy and safety outcomes. Bayesian fixed‐effect mixed treatment comparison methods were used to assess relative therapeutic efficacy and harms. Results Data from 58 studies including 29 interventions and 11,883 patients were analyzed. Pain reduction over that of placebo on the 11‐point numeric rating scale ranged from −3.29 for sodium valproate (95% credible interval [CrI] = [−4.21, −2.36]) to 1.67 for Sativex (−0.47, 0.60). Estimates for most treatments were clustered between 0 and −1.5 and were associated with more study data and smaller CrIs. Pregabalin (≥ 300 mg/day) was the most effective on the 100‐point visual analog scale (−21.88; [−27.06, −16.68]); topiramate was the least (−3.09; [−3.99, −2.18]). Relative risks (RRs) of 30% pain reduction ranged from 0.78 (Sativex) to 1.84 (lidocaine 5% plaster). Analysis of the RR ratio of these 2 treatments reveals marginal significance for Sativex (3.27; [1.07, 9.81]), indicating the best treatment is only slightly better than the worst. Relative risks of 50% pain reduction ranged from 0.98 (0.56, 1.52) (amitriptyline) to 2.25 (1.51, 3.00) (alpha‐lipoic acid). RR ratio for these treatments was not statistically different (3.39; [0.88, 3.34]). Fluoxetine had the lowest risk of adverse events (0.94; [0.62, 1.23]); oxycodone had the highest (1.55; [1.45, 1.64]). Discontinuation RRs were clustered around 0.8 to 1.5, with those on the extreme having greater uncertainty. Conclusions Selecting an appropriate pDPN therapy is key given the large number of available treatments. Comparative results revealed relative equivalence among many of the studied interventions having the largest overall sample sizes and highlight the importance of standardization of methods to effectively assess pain.