For the purpose of this study, we defined iHSP by refining Singh et al. [8]'s criteria for refractory shoulder pain and Roosink et al. [9]'s criteria for persistent post-stroke shoulder pain. Eligible patients met all of the following inclusion criteria: (1) hemiplegic shoulder pain was defined as shoulder pain confined to the shoulder and/or C5 dermatome of the affected side during rest and/or active and/or passive motion after stroke onset; (2) non-remitting shoulder pain ≥3 months; (3) pain score ≥4.5 on a 0–10 numerical rating scale; and (4) inadequate or no response to conventional treatment including intra-articular corticosteroid injection (≥3 injections). Exclusion criteria were (1) previous history of BTX injection into the subscapularis muscle; (2) clinical examination finding indicating that acromioclavicular disease or rotator cuff disease was the primary source of the shoulder pain; (3) complex regional pain syndrome diagnosed according to clinical diagnostic criteria proposed by Harden et al. [18]; (4) central post-stroke pain diagnosed according to the grading system proposed by Klit et al. [19]; (5) shoulder joint malignancy, prosthetic shoulder joint, or planned shoulder joint surgery in the next 6 months; (6) known allergy to BTX; and (7) unstable and serious medical condition or psychiatric disorder. Demographic and clinical data including age, sex, interval between stroke and injection, cause of stroke, and Brunnstrom stage of the affected upper arm were recorded. Conventional therapy factors that might affect the iHSP, such as modality, medication, and physical therapy, were recorded. This study was conducted in accordance with the Declaration of Helsinki and approved by National Rehabilitation Center Institutional Review Board.

A retrospective chart review was performed of 6 patients with iHSP treated in a rehabilitation hospital in Korea between September 2013 and August 2015.

BTX (Botox; Allergan Inc., Irvine, KY, USA) injection into the subscapularis muscle was performed as previously described. An ultrasonography-guided lateral approach or conventional medial approach was made according to the patients' anatomical differences [20,21]. The needle was used as a stimulation electrode to detect the motor point at which minimal stimulation induces maximal internal rotation. BTX dosage was selected by various patient characteristics such as body weight and spasticity severity. Thus, 60–80 units of BTX were administered in each injection. Physical therapy (capsular stretching, distraction, passive range of motion [PROM] exercise), modality (hot pack, infrared, interferential current therapy), and neuromuscular electrical stimulation of the shoulder muscles, such as the posterior fibers of the deltoid and supraspinatus muscles, were administered to all participants as before the injection.

Pain severity was measured on an 11-point pain intensity numeric rating scale (PI-NRS), where 0 is no pain and 10 is the worst possible pain. Pain-free PROM of abduction and external rotation were assessed. Spasticity of shoulder internal rotator was measured using the modified Ashworth scale (MAS) rated on a scale of 0 (normal) to 4 (extreme). These measurements were carried out at baseline and at 1, 2, 4, and, if possible, 8 weeks after the intervention.

Statistical analyses were performed using SPSS software ver. 17.0 (SPSS Inc., Chicago, IL, USA). A non-parametric statistical test was used because of the small sample size in the study group. Freidman tests were used to compare measurements at baseline and follow-up after treatment. Correlations between PI-NRS and pain-free PROM or MAS of the affected shoulder were investigated using the Spearman test. The level of statistical significance was set at p<0.05.