Mirels’ score predicts the likelihood of sustaining pathological fractures using pain, lesion site, size and morphology. Its reproducibility and reliability are questioned in the upper limb given small number of patients with metastasis here. The aim is to investigate its reproducibility, reliability and accuracy in upper limb bony metastatic disease and validate its use in pathological fracture prediction. A retrospective cohort study of patients with upper limb metastases, referred to a UK Orthopaedic Trauma Centre over 6-years (2013-18). Mirels’ score was calculated in 32 patients; plain radiographs at index presentation scored using Mirels’ system by 6 raters. The radiological aspects (lesion size &radiological appearance) were scored twice by each rater (2-weeks apart). Inter- and intra-observer reliability were calculated using Fleiss’ kappa test. Bland-Altman plots compared variances of individual score components and total Mirels’ score. Mirels’ score of ≥9 did not accurately predict lesions which would fracture (11% 5/46 versus 65.2% Mirels’ score ≤8, p Kappa values for interobserver variability were k=0.358 (fair, 0.288-0.429) for lesion size, k=0.107 (poor, 0.02-0.193) for radiological appearance and k=0.274 (fair, 0.229-0.318) for total Mirels’ score. Values for intraobserver variability were k=0.716 (good, 95% CI 0.432-0.999) for lesion size, k=0.427 (moderate, 95% CI 0.195-0.768) for radiological appearance and 0.580 (moderate, 0.395-0.765) for total Mirels’ score. We showed moderate to substantial agreement between and within raters using Mirels’ score on upper limb radiographs. This study demonstrates Mirels’ has poor sensitivity and specificity in predicting upper limb fractures. We recommend the score cut-off for prophylactic surgery in upper limb metastases should be lower than that recommended for lower limb lesions.