Objective: To investigate the intra-rater and inter-rater reliability of pain-free shoulder force testing in people with rotator cuff-related shoulder pain (RCRSP), and to explore its associations with patient-reported clinical and psychosocial measures.

Design: Reliability study of fixed portable dynamometry. Assessment order was randomised.

Methods: Two investigators tested mean and maximal pain-free abduction and external rotation force for both shoulders in 31 people with unilateral RCRSP, using a fixed, portable hand-held dynamometer (HHD). The order in which the investigators assessed each participant was randomised. Participants were blinded to their test scores, and each investigator was blinded to the other investigator’s measurements. Secondary analyses examined the relationship between the relative pain-free force index (as well as three psychosocial outcome measures) and the Shoulder Pain and Disability Index (SPADI).

Results: Absolute measurement error was substantial: the inter-rater limits of agreement (LoA) were wide, the standard error of measurement (SEM) ranged from 8.9 to 19.5 N (minimal detectable change at the 95% confidence level [MDC95] 24.7–53.9 N) and the intra-rater repeatability coefficient reached 34.5 N, indicating the need for caution when interpreting small changes in isometric shoulder force; relative reliability was nonetheless good to excellent (intraclass correlation coefficient [ICC] 0.89–0.97 intra-rater and 0.83–0.89 inter-rater). In secondary exploratory analyses, pain-free force deficits showed weak-to-moderate associations with SPADI scores (|r| = 0.01 to 0.43). Psychosocial measures demonstrated stronger correlations (|r| = 0.44 to 0.71; moderate-to-strong associations), with the Pain Self-Efficacy Questionnaire (PSEQ) showing the strongest association.

Conclusion: Pain-free HHD shoulder force testing in RCRSP showed substantial absolute measurement error, with wide LoA despite good relative reliability. Exploratory associations with patient-reported pain and disability were numerically stronger for psychosocial measures than for the relative pain-free force index. These findings suggest limitations in detecting small differences in force and support assessment beyond force testing alone.