We prospectively followed a cohort of 141 patients who underwent surgical repairs of full-thickness rotator cuff tears between April 1997 and July 2000. Participants were recruited from two surgeons’ practices; one surgeon (RB) was a shoulder specialist whereas the other (DO) specialized in sport medicine, including shoulder disorders. Eligible patients had a diagnosis of a full-thickness rotator cuff tear based on history, physical examination, and radiographic investigations, without accompanying Bankart lesions, severe glenohumeral osteoarthritis, or any previous surgery on the affected shoulder. Although most patients underwent MRI, CT, or arthrograms in addition to plain radiography preoperatively, the diagnosis of a full-thickness tear was confirmed during surgery for final determination of study eligibility. We received ethics approval from the regional health ethics board for this study. All participants provided signed informed consent before inclusion.
Patients were assessed preoperatively by a shoulder physical therapist (CLC) who did not treat participants. Demographic factors (age, gender) and patient-specific factors (WCB involvement, smoking status, type of injury onset, chronicity of tear) also were collected preoperatively. Of the 141 patients enrolled, 36 (26%) were WCB recipients. The mean age was 54.0 ± 10.4 years and 102 (72%) were male.
We used the Western Ontario Rotator Cuff (WORC) index and the American Shoulder and Elbow Surgeons’ (ASES) questionnaires to assess shoulder pain and function [24
]. Active ROM included flexion measured with the patient in a standing position and external rotation assessed in a supine position with the arm in an abducted position. Abduction was not included in the ROM analysis owing to alterations in measurement technique midway through the study, whereas internal rotation was measured using vertebral levels, a method that did not allow for inclusion in the global shoulder ROM score.
Shoulder strength was measured using a subjective rating scale that could not be reliably quantified [4
]; thus strength was included as a dichotomous variable where patients either did or did not have antigravity strength in forward flexion and external rotation. Patients were reassessed by the same physical therapist 3 and 6 months postoperatively.
Patients were placed in the beach chair position. Both surgeons started with an arthroscopic evaluation. Superior labral anteroposterior lesions were débrided or repaired [40
]. The biceps disorder was treated by débridement or tenodesis. All patients underwent acromioplasty with coracoacromial ligament transection.
Rotator cuff tear size was based on the ASES classification (small, 0–10 mm; medium, 11–30 mm; large, 31–49 mm; massive, > 50 mm) [9
]. Tear size was measured intraoperatively along the anteroposterior base (length) and the widest mediolateral aspect (retraction) using a flexible ruler or a probe tip with millimeter markings.
The two study surgeons chose repair type based on clinical experience and patient characteristics (eg, tear size, tissue quality). For one surgeon (DO), patients underwent mini-open or open repairs. For the other surgeon (RB), if the tear was mobilized easily and bone and tissue quality were adequate to obtain good purchase of suture anchors, an arthroscopic repair was performed. Otherwise, a mini-open or open repair was performed, again depending on tissue quality and mobility. Health status and comorbidities did not alter choice of surgical approach. Mini-open repairs were performed in 84 (60%), arthroscopic in 24 (17%), and open repairs in 33 (23%) patients. WCB recipients and nonrecipients were distributed similarly between the two surgeons (Table ). Surgical approach also was similar between groups, indicating WCB status did not influence the surgeons’ choice of surgical approach (Table ).
Surgeons and surgical approach by distribution
All patients received standardized preoperative and perioperative care at one hospital. Patients were encouraged to attend postoperative therapy at a rehabilitation unit with access to pool therapy and group strengthening classes. However, as a referral center for a large geographic region, some patients received therapy elsewhere.
Regardless of rehabilitation location, method of surgical repair, or WCB status, all patients were prescribed the same regimen. Patients wore a Velpeau sling for 6 weeks and were referred for physical therapy 2 weeks postoperatively. During the initial 6 weeks, only self-assisted ROM (excluding abduction), pendulum exercises, and isometric scapular stabilization exercises were permitted. Between 6 and 10 weeks after surgery, self-assisted stretching toward end range, active shoulder ROM, and closed chain strengthening exercises were added and scapular stabilization exercises progressed. Ten weeks postsurgery, progressive strengthening started, beginning with isometric and progressing to isotonic exercises. Therapist-assisted joint mobilization also was added to maximize ROM.
For measurement of pain and function, the WORC index and the ASES index were the dependent variables. The WORC index, reliable and valid for patients with rotator cuff disorders, is a disease-specific tool that evaluates physical symptoms (including pain), sports and recreation, work, lifestyle, and emotions [24
]. The ASES index is a reliable and valid joint-specific questionnaire that assesses functional limitations and pain during shoulder activities [26
]. Both indices are measured on a scale of 0 to 100, where 100 indicates no pain or shoulder dysfunction.
Active shoulder ROM, the third dependent variable, was measured using a goniometer as per published standards [7
] for forward flexion in standing and external rotation in abduction. For the purpose of analysis, a global ROM score, calculated by summing movement in these planes, was used to determine the impact of WCB status on ROM in more than one plane of movement.
Six months postoperatively, the WORC and ASES questionnaires were completed by 113 patients (80%), whereas 107 patients (76%) returned for ROM assessment. Patients who missed the 6-month assessment were similar (p > 0.05) in baseline and surgical characteristics to those who completed the visit, with the exception of smokers, who were more likely (p = 0.01) to drop out. Nonrespondents were similarly distributed between WCB recipients and nonrecipients (p = 1.0), as was the smoking distribution between nonrespondents in both groups (p = 0.46).
To determine if the WCB recipients and nonrecipients were different before surgery, we compared the following baseline characteristics between the two patient groups: age, gender, smoking status, duration of shoulder symptoms, injury onset type (traumatic/insidious), preoperative WORC and ASES scores, preoperative ROM and strength, tear size, and associated labral or biceps disorder. Chi square tests were used for categorical data (gender, smoking status, type of onset, chronicity and size of tear, labral and biceps disorder, and strength) and independent t tests were used for continuous variables (age, shoulder ROM, ASES and WORC scores). As we focused on shoulder-specific variables, comorbid conditions were not evaluated. We were interested in preoperative patient characteristics, as this information would assist clinicians in determining postoperative needs and setting appropriate patient expectations before surgery.
Comparison then was made of the 6-month WORC and ASES scores and shoulder ROM using independent t tests between groups to determine if the groups had different outcomes when preoperative differences were not considered.
Multiple linear regression analysis then was used to examine 6-month WORC and ASES outcomes and global ROM (flexion plus external rotation), controlling for any baseline differences seen between groups. Regression allowed us to determine the effect of WCB status while simultaneously controlling for the effect of other variables we found different between groups in the first stage of the analysis. Thus, when we examined the independent effect of WCB status, we compared patients with similar characteristics with the exception of WCB status. Variables were considered only for the regression analysis if p ≤ 0.05 to prevent overfitting the model with unnecessary variables. With the number of subjects available for analysis at 6 months, the model was powered to detect a medium effect size (α = 0.05; β = 0.20) with up to nine variables [14
]. We conducted extensive regression diagnostic tests (residual analysis, leverage statistics, and Cook’s distance) to ensure our model met regression modeling assumptions [25
To further alienate the potential confounding effects of baseline factors on the difference between WCB recipients and nonrecipients in the 6-month WORC and ASES outcomes and global ROM, propensity scores were used in the multiple regression analyses instead of the confounders [37
]. Significant differences observed in the confounders between WCB recipients and nonrecipients disappeared when controlled for by the propensity scores (data not shown).
To meet our final objective, differential responses were explored for age and smoking status between WCB recipients and nonrecipients, as these characteristics were different between groups. Only smoking status seemed to have a differential impact; thus, the other exploratory analysis is not shown. Analysis was performed using SPSS® Version 14.0 (SPSS Inc, Chicago, IL).