A systematic review with Bayesian modelling of the prevalence of pain with neuropathic characteristics

Abstract We performed the first systematic review and meta-analysis of the prevalence of pain with neuropathic characteristics using Bayesian methods to correct prevalence estimates for the use of screening tools with imperfect sensitivity and specificity (CRD42023416845). We searched major databases for national or regional epidemiological studies that reported the prevalence of pain with neuropathic characteristics, as identified by the PainDETECT, S-LANSS, or DN4-interview. Of the 1,251 unique records retrieved, 8 were finally extracted. The uncorrected (apparent) prevalence data were pooled using a random-effects meta-analysis for proportions. The corrected (true) prevalence was estimated using Bayesian models incorporating sensitivity and specificity distributions under non-informative [beta(1,1)] and informative priors [beta(4.389, 29.522); based on apparent prevalence]. Using the mean values from Bayesian credible intervals, a pooled estimate of true prevalence was generated using a random-effects model. The pooled estimate for the apparent prevalence was 10.6% (95% CI: 8.5; 12.9). The pooled estimate for true prevalence was 4.9% (95% CI: 3.8; 6.1) using informative priors, and 2.3% (95% CI: 1.5; 3.2) using non-informative priors. The use of imperfect screening tools may have overestimated the prevalence of neuropathic pain. Perspective The prevalence of neuropathic pain may be lower than previously estimated. A lower prevalence should not be equated with reduced societal or clinical significance, but it may have implications for healthcare resource allocation and research funding policies for neuropathic pain.