Keele Research Repository

Introduction
Frozen shoulder, a condition that can cause prolonged pain and disability, has previously been found to be common amongst people with diabetes. This thesis aimed to improve the understanding of the nature of the relationship between diabetes and frozen shoulder.

Methods
Evidence from existing longitudinal observational studies was summarised to assess whether diabetes is a risk factor for the onset of frozen shoulder, as well as whether it is a prognostic factor for poor outcomes of frozen shoulder. A series of cohort studies based on data from Clinical Practice Research Datalink (CPRD) were subsequently undertaken to establish these relationships more accurately and comprehensively, and to investigate whether the effect of diabetes on the risk of developing frozen shoulder is mediated through metabolic health. The association between newly diagnosed frozen shoulder and a subsequent diabetes diagnosis was also estimated.

Results
Most of the relevant studies identified in the literature search were at high risk of bias. Causal mediation analysis of 87,954 patients from CPRD suggested that diabetes does affect the development of frozen shoulder, but the effect is unlikely to be mediated by metabolic health. Following the 15.8-year follow-up, the probability for the frozen shoulder group to be diagnosed with diabetes was 5% versus 0.28% in those without frozen shoulder. Diabetes was a predictor of surgery in 40,644 patients with frozen shoulder.

Conclusion
Evidence suggests that diabetes is a potential cause of frozen shoulder, although our results did not support the hypothesis that metabolic health mediates the effect of diabetes on the development of frozen shoulder. People with frozen shoulder are more likely to have a subsequent diagnosis of diabetes; future research is required to determine whether testing all patients with frozen shoulder is an effective strategy to detect diabetes early in its course and reduce the likelihood of complications.