Releasing Forces in Adhesive Capsulitis Are Important Indicators of Shoulder Stiffness and Postoperative Function

Background Manipulation under anesthesia is a widely used treatment for frozen shoulder, but the factors that influence patient outcomes after manipulation remain unclear. The degree of shoulder stiffness, a critical feature of frozen shoulder, likely reflects the severity of the condition but currently lacks standardized, objective assessment methods. Questions/purposes (1) What are the releasing forces in patients with frozen shoulder, and do the forces vary across different stages of frozen shoulder? (2) Are there differences in postoperative outcomes of manipulation under anesthesia among patients with frozen shoulder at different stages of the condition? (3) Is a higher releasing force associated with poorer outcomes of manipulation, and what threshold of releasing force is optimal for better outcomes? (4) What clinical factors influence the magnitude of releasing forces? Methods This prospective cohort study included patients with primary unilateral frozen shoulder who underwent manipulation under anesthesia after at least 3 months of unsuccessful nonsurgical treatment, which was defined as progressive worsening ROM, failure to make progress, or residual functional impairment after 3 months of treatment. Between December 1, 2022, and December 31, 2023, we treated 280 patients with unilateral frozen shoulder, all of whom were considered potentially eligible for this study. The inclusion criteria were: a reduction of passive external rotation in the affected shoulder to less than 50% compared with the contralateral side, at least 3 months of unsuccessful nonsurgical treatment, absence of shoulder trauma, radiographs and MRI showing no other pathologic lesions in the shoulder, and no prior medical history in the contralateral shoulder. The exclusion criteria were patients who had previously undergone shoulder surgery, those who had bilateral frozen shoulder, patients with anesthesia intolerance, and those with incomplete preoperative assessments. One hundred fifty-six patients were enrolled in follow-up assessments at 1, 3, and 6 months after manipulation. The mean ± SD age for enrolled patients was 54 ± 8 years, 35% (55 of 156) of all participants were male, and the mean BMI was 23 ± 3 kg/m 2 . Two percent (3 of 156) withdrew consent, and 4% (7 of 156) were lost to follow-up, leaving 94% (146 of 156) for analysis. The contralateral unaffected shoulder was used as a self-control. During the manipulation process, the force-time curves for the affected and unaffected shoulders were sequentially recorded using a handheld dynamometer, following the order of forward flexion, external rotation, and internal rotation. Two key force values, an initial tear value and a peak value, were extracted from the curve for the affected shoulder, while only the peak value was recorded for the unaffected shoulder. Passive ROM, the Oxford shoulder score (OSS), and the VAS were evaluated at the baseline and at 1, 3, and 6 months postoperatively. Patients were categorized into four stages according to the patient-reported duration of pain: Stage 1 (0 to 3 months), Stage 2 (3 to 9 months), Stage 3 (9 to 15 months), and Stage 4 (> 15 months). To address our first and second questions, we used ANOVA for multistage comparisons of continuous variables, followed by a post hoc Tukey test for pairwise comparisons. For the third question, we performed univariate regression to analyze the correlation between factors like age, sex, symptom duration, frozen shoulder stage, preoperative ROM, upper arm circumference, fat-free mass, diabetes, thyroid disease, hyperlipidemia, tear value, peak value, and 6-month postoperative ROM, VAS, and the OSS. Factors with p < 0.05 were included in a multivariate regression. A tear value threshold of poor ROM outcomes was evaluated with a receiver operating characteristic (ROC) curve and the Youden index. For the fourth question, we used similar regression models to examine potential factors associated with the releasing force, focusing on both tear and peak values. Pairwise comparisons in this subgroup analysis were performed using the Student t-test. All p values less than 0.05 were considered significant. Results Tear values of each stage were as follows: 25 ± 13 N in Stage 2, 28 ± 15 N in Stage 3, and 38 ± 18 N in Stage 4. The tear value for patients in Stage 4 was higher compared with both Stage 2 and Stage 3 (Stage 4 versus Stage 2, mean difference 13 [95% CI 6 to 20]; p < 0.001; Stage 4 versus Stage 3, mean difference 10 [95% CI 2 to 19]; p = 0.01). Patients in Stage 4 exhibited an increased peak value relative to the other two stages (Stage 4 versus Stage 2, mean difference 11 [95% CI 2 to 20]; p = 0.02; Stage 4 versus Stage 3, mean difference 8 [95% CI 0 to 16]; p = 0.04). The peak value in the affected shoulder was higher than that in the unaffected (mean difference 40 [95% CI 36 to 44] in forward flexion; p < 0.001). At the 6-month endpoint after manipulation, patients in Stage 2 and 3 showed greater ROM in forward flexion than those in Stage 4 (Stage 2 versus Stage 4, mean difference 12 [95% CI 9 to 14]; p < 0.001; Stage 3 versus Stage 4, mean difference 14 [95% CI 11 to 17]; p < 0.001) and a lower OSS than those in Stage 4 (Stage 2 versus Stage 4, mean difference -8 [95% CI -9 to -7]; p < 0.001; Stage 3 versus Stage 4, mean difference -7 [95% CI -8 to -6]; p < 0.001). Two factors were associated with the OSS at the 6-month endpoint: increased tear value (β = 0.47; p = 0.004) and diabetes (β = 0.28; p = 0.02). The optimal thresholds for predicting a forward flexion at least 164° at 6 months was a tear value of 53 N (area under curve [AUC] 0.79 [95% CI 0.68 to 0.91]). Patients with a tear value of below 53 N demonstrated better postoperative ROM (mean difference 10 [95% CI 3 to 16]; p = 0.004) and OSS (mean difference -4 [95% CI -8 to 0]; p = 0.04). The tear value was associated with male sex (β = 0.36; p = 0.03) and ROM in flexion (β = 0.20; p = 0.049), whereas peak value was associated with male sex (β = 0.45; p = 0.001) and diabetes (β = 0.16; p = 0.048). Conclusion These findings suggest that performing manipulation before reaching Stage 4 may result in more favorable outcomes for patients, and evaluating shoulder stiffness by measuring releasing force proved to be feasible. Level of Evidence Level II, prognostic study.