Evidence for a time-dependent pain response to platelet-rich plasma versus corticosteroid injection in lateral epicondylitis: a cluster-stratified systematic review and meta-analysis of randomized controlled trials

Background: Corticosteroid and platelet-rich plasma (PRP) injections are both used for lateral elbow tendinopathy, with conflicting evidence on the temporal profile of their comparative efficacy. Objectives: To characterise the time-dependent pain response to PRP versus corticosteroid and to formally test whether this response varies with follow-up time. Eligibility criteria: Randomised controlled trials comparing PRP with corticosteroid injection and reporting Visual Analogue Scale (VAS) pain, with at least ten participants per arm. Information sources: MEDLINE, Embase, CENTRAL, Scopus, and Web of Science, searched to April 2026. Risk of bias: Assessed at the outcome level using RoB 2 for the VAS pain result of each trial. Synthesis methods: Random-effects meta-analysis (REML, Knapp–Hartung correction) within four pre-specified follow-up clusters, followed by a random-effects meta-regression formally testing the treatment-by-time trend. Included studies: Six trials, 338 participants (180 PRP, 158 corticosteroid). Methods: Six randomised controlled trials comparing PRP with corticosteroid and reporting Visual Analogue Scale (VAS) pain were identified through MEDLINE, Embase, CENTRAL, Scopus and Web of Science (to April 2026). Scores were pooled within four pre-specified clusters (≤ 4 weeks; 5–11 weeks; 12–23 weeks; ≥ 24 weeks) using a random-effects model with restricted maximum likelihood (REML) estimation and Knapp–Hartung correction. Risk of bias was assessed at the outcome level using RoB 2; certainty of evidence was graded using GRADE. Results: Across 338 participants (180 PRP, 158 corticosteroid), corticosteroid was favoured at ≤ 4 weeks (MD + 1.88; 95% CI 1.05–2.71; p = 0.006; I2 = 38%), exceeding the 1.5-point MCID. At 5–11 weeks, the estimate remained positive but non-significant (+ 1.62; − 1.69 to 4.93; p = 0.170; I2 = 89%). At 12–23 weeks, the estimate crossed zero (− 0.71; − 1.79 to 0.37; p = 0.152; I2 = 86%). Beyond six months, PRP was superior (− 1.49; − 2.08 to − 0.90; p = 0.001; I2 = 56%), though the magnitude fell just below the 1.5-point MCID. A random-effects meta-regression confirmed a significant time trend across all study-by-cluster estimates (slope − 0.087 per week; p = 0.003). Among five trials with data at both early and late timepoints, all showed the reversal pattern (5/5; p = 0.063), a hypothesis-generating finding that did not reach significance. GRADE certainty was moderate for the early and late clusters and low for the intermediate clusters. Limitations: Repeated within-trial observations across clusters violate independence assumptions, so the meta-regression and temporal trajectory are descriptive/hypothesis-generating rather than a formal between-study interaction test; the within-trial concordance analysis did not reach significance (p = 0.063). Conclusions: These data suggest, and now formally support via meta-regression, a time-dependent reversal in comparative efficacy. Level of Evidence: I.