Systematic review and quantitative meta-analysis demonstrates analgesic effect of excitatory motor cortex non-invasive brain stimulation, which is inflated by small study and publication bias

Abstract

This systematic meta-analysis evaluated the analgesic effect of excitatory motor cortex (M1) non-invasive brain stimulation (NIBS) in patients experiencing chronic neuropathic pain (CNP). Considering intervention groups with a median sample of 16.5 patients, we expected small-study and publication bias. We searched randomized controlled trials (RCTs) of high frequency (HF) (≥5 Hz) M1 rTMS and anodal M1 tDCS in patients with CNP, using PubMed and Web of Science, covering 1990 to 2019. Our search yielded 58 eligible studies. We calculated random effects models to obtain confidence and prediction intervals (PIs) of mean differences (MD) of analgesic effect. To assess small-study and publication bias, we created funnel plots, and trim-and-fill and Copas models. Excitatory M1 NIBS RCTs included 1246 patients in intervention and 1202 patients in sham groups across 58 studies. The MD was 0.92 with 95% confidence interval (CI) [0.70;1.14] and PI [−0.59;2.44]. Considering HF M1 rTMS studies, MD was 0.99 with 95%CI of [0.71;1.27] and 95%PI of [-0.61;2.58]. In anodal M1 tDCS, MD was 0.86 with 95%CI [0.48;1.25] and 95%PI of [-0.73;2.46]. In funnel plot analysis, 18/58 excitatory NIBS reports fell outside the funnel bounds. After trim-and-fill model imputation of missing studies, MD was 0.44 with 95%CI of [0.18;0.70]. The Copas model revealed, at a 0.5 probability of largest standard error study publication, MD was 0.535. Probability of 0.1 residual selection bias occurred at 0.52 probability of largest SE study publication. Copas model MD was 0.6 with 95%CI [0.46;0.75]. We found significant analgesia following excitatory M1 NIBS in patients with CNP. PIs encompass zero, indicating significant analgesia is not probable in 0.95 of future studies. Results support small-study and publication bias. Future studies should include larger sample sizes. Emphasis on null result publication and pre-publication registration is warranted. Neurosurgery Pain Research Institute, Postdoctoral Scholarship and other funds (TJM and FAL) NIH - National Institute of Neurological Disorders and Stroke - R01NS107602 (FAL).