Background: Accumulation of score distribution towards the high end of the measurement scale is an important source of bias related patient-reported outcome measures (PROM). The aim was to evaluate how PROM score distributions, scale boundaries, and sampling variability influence the likelihood of detecting a minimal clinically important difference (MCID) of 10 points between surgical and non-surgical groups in randomised controlled trials (RCTs) of musculoskeletal disorders.

Methods: We did a systematic review and meta-epidemiological analysis of 129 RCT studies comparing surgical and non-surgical interventions in patients with musculoskeletal complaints using a PROM as an outcome measure (1771 group-level PROM measurements) from PubMed and Scopus published until February 26, 2025. Simulations assessed each comparison's likelihood of detecting a difference of 10 points or more.

Findings: The mean difference between groups was 4.6 (SD 7.1) points favouring surgery, with surgical arms scoring higher in 72% of comparisons. The mean likelihood of detecting at least a 10-point difference was 19%, meaning fewer than one in five of such comparisons would detect a true difference. Detection likelihood peaked (35%) at a mean score of 70, declining toward scale extremes. Comparisons with significant observed differences (>10 points, p < 0.05) had a 54% likelihood versus 17% in non-significant comparisons, strongly linking detection likelihood to observed differences.

Interpretation: The majority of the PROM-based RCTs were unlikely to detect differences due to ceiling effects with a constant underestimation of surgical benefit. PROMs with adequate content coverage, better discrimination, and reduced ceiling susceptibility should be selected for clinical practice. Future research should align outcome selection and follow-up timing with expected treatment effects and ensure that measurement properties do not mask meaningful clinical differences.