Ethical approval for this retrospective study of prospectively collected data was granted by the South Eastern Sydney and Illawara Ethics Committee. Between February 2003 and March 2006, 312 consecutive patients with symptomatic rotator cuff tears underwent rotator cuff repair by one surgeon (GACM) using suture anchors to reattach the torn tendon(s) to bone. Patients with glenohumeral arthritis, fracture, previous shoulder surgery, osteonecrosis, and partial-thickness rotator cuff tears and those who were unable or unwilling to undergo ultrasound examination at 6 months and 2 years postoperatively were excluded. Patients who underwent rotator cuff repair within the first 6 weeks of the surgeon changing to new arthroscopic techniques were excluded to take into consideration the initial learning curve associated with the new technique. Of the 312 consecutive patients undergoing rotator cuff repair, 153 were excluded as per the exclusion criteria, including nine with glenohumeral arthritis (defined radiographically and at arthroscopy), seven with shoulder fractures, 18 with previous shoulder surgery, 35 with isolated partial-thickness tears, 62 who were unable or unwilling to undergo ultrasound examination at 6 months, and 22 who fell into the 6-week period identified before the study as the arthroscopic repair learning curve period, leaving 159 patients with 6 months’ clinical followup. All 159 patients underwent ultrasound examination 6 months postsurgery. At 2 years postoperatively, an additional 76 (48%) patients were excluded because they were unable to undergo ultrasound examination, leaving 87 (55%) patients for clinical and structural evaluations at 2 years of followup.
From February 2003 to March 2004, the surgeon performed open rotator cuff surgery using metallic suture anchors (Mitek RC Quickanchor™; Depuy Mitek Inc, Raynham, MA) (open group, n [6 months] = 49, n [2 years] = 20). Between December 2003 and August 2005, the surgeon changed to arthroscopic repair of rotator cuff tears and patients underwent arthroscopic rotator cuff repair with a screw-in metallic suture anchor (Mitek Fastin®; Depuy Mitek Inc) (arthroscopic knotted group, n [6 months] = 53, n [2 years] = 29). During the transition period from open to arthroscopic repair (December 2003 to March 2004), the decision to perform arthroscopic repair was based on tear size, with large and massive tears repaired using the mini-open technique. In August 2005, the surgeon changed to an arthroscopic knotless technique and patients underwent rotator cuff repair using a metallic knotless suture anchor (ArthroCare Opus Magnum; ArthroCare Corp, Sydney, Australia) (arthroscopic knotless group, n [6 months] = 57, n [2 years] = 38).
At all points up to and including the 6-month review, there were 21 men and 28 women in the open group with a mean age of 58 years (range, 28–87 years), 24 men and 29 women in the arthroscopic knotted group with a mean age of 64 years (range, 40–90 years), and 28 men and 29 women in the arthroscopic knotless group with a mean age of 59 years (range, 34–86 years) (Table ). The mean times from initial injury to surgical repair were 13 months (range, 0.5–81 months), 7 months (range, 0.8–39 months), and 6 months (range, 0.2–31 months) in the open, arthroscopic knotted, and arthroscopic knotless groups, respectively. The mean time from initial injury to surgical repair was longer (p < 0.05) in the open than in the two arthroscopic groups.
Demographics of patients in the groups
At the 2-year assessment, there were 10 men and 10 women in the open group with a mean age of 60 years (range, 48–73 years), 10 men and 19 women in the arthroscopic knotted group with a mean age of 63 years (range, 40–79 years), and 28 men and 29 women in the arthroscopic knotless group with a mean age of 61 years (range, 34–86 years) (Table ). The mean times from initial injury to surgical repair were 15 months (range, 0.7–81 months), 7.2 months (range, 1.0–39 months), and 6.6 months (range, 0.5–31 months) in the open, arthroscopic knotted, and arthroscopic knotless groups, respectively. The mean time from initial injury to surgical repair was longer (p < 0.05) in the open than the two arthroscopic groups.
All procedures were performed under interscalene block with the patient in the upright beach chair position as day case surgery. All patients were given a preoperative dose of the same antibiotic (1 g cefazolin intravenously) and one additional dose of this antibiotic 4 hours after completion of the procedure. All patients underwent initial arthroscopic assessment through a standard three-portal technique. A posterior portal was established for initial inspection of the glenohumeral joint. An anterior portal through the rotator interval was established as a working portal for intraarticular débridement. Another lateral accessory portal was created for inspection and decompression of the subacromial space plus rotator cuff repair in the arthroscopic groups. Operative time was defined as the time in minutes from first skin incision until wound closure.
Our surgical technique for open rotator cuff repair was described previously [6
]. Briefly, it included partial splitting of the deltoid in line with its fibers, detachment and later reattachment of the coracoacromial ligament at the level of the anterior acromion, and anterior acromioplasty and bursectomy. Appropriate soft tissue releases (subacromial and extraarticular adhesion, coracohumeral ligament, rotator cuff interval) were performed before rotator cuff repair.
The suture anchors (Mitek RC Quickanchor™) were impacted directly into the bone of the proximal humerus without predrilling of the bone. When possible (ie, when there was enough excursion of the torn tendon), a two-row anchor technique was used for fixation. The suture material used in the above anchors was either Number 2 braided, nonabsorbable polyester suture or Number 2 absorbable Panacryl™ (Ethicon, Inc, Somerville, NJ) suture. The tendons were grasped with the suture material by a horizontal mattress stitch configuration.
Arthroscopic repair was performed using the standard three-portal technique as described by Gartsman and Hammerman [13
]. After arthroscopic acromioplasty, the rotator cuff tear edge was débrided and the landing site at the greater tuberosity was gently débrided and smoothed with an arthroscopic burr. Mobilization techniques were performed to permit anatomic repair of the tendon to the greater tuberosity and reduce tension on the repair. After the mobilization techniques were performed, all tears, including those with asymmetric retraction, could be mobilized to the lateral-most aspect of the greater tuberosity with the arm at 0° abduction. All patients underwent direct repair of the tendon to the bone without medialization of the repair site or a margin convergence type of repair.
For arthroscopic knotted repair, the rotator cuff was repaired to bone in a single row using two simple sutures per anchor. The 5-mm metal double suture loaded corkscrew anchors (Mitek Fastin®) were inserted through the lateral accessory portal anteriorly to posteriorly in a single row in the rotator cuff footprint.
For arthroscopic knotless repair, the rotator cuff was repaired to bone in a single row using one inverted mattress suture per anchor. The torn rotator cuff was grasped with the Opus SmartStitch® Suture Device (ArthroCare Corp), which delivers a Number 2 polyester mattress suture into the cuff through the lateral portal. A hole was punched at the desired position on the landing site through the lateral portal. Both limbs of the suture were passed through the Opus Magnum Knotless Implant; the implant then was inserted into the prepared bone hole and deployed in the bone. The suture then was wound through the anchor reducing the tendon to the bone before locking the suture in the anchor.
Mean preoperative tear sizes were not different among the 6-month cohorts: 4.1 cm2 (range, 0.5–14 cm2), 3.9 cm2 (range, 1–13.5 cm2), and 3.7 cm2 (range, 1–16.5 cm2) in the open, knotted, and knotless groups, respectively (Tables , ). The open group used double-row fixation techniques in 45 of 49 (92%) cases, with the remaining cases using a standard single-row fixation. The open group had repair with 3 ± 0.4 absorbable sutures in 14 patients and 4 ± 0.6 nonabsorbable sutures in 35 patients. Mean preoperative tear sizes were not different among the groups of the 2-year followup cohort: 4.3 cm2 (range, 1.0–14 cm2), 3.9 cm2 (range, 1–13.5 cm2), and 3.6 cm2 (range, 1–15.75 cm2) in the open, knotted, and knotless groups, respectively (Tables , ). The open group had double-row fixation techniques in 19 of 20 (92%) cases, with the remaining case having a standard single-row fixation. The open group had repair with 3 ± 0.4 absorbable sutures in three patients and 4 ± 0.6 nonabsorbable sutures in 17 patients.
Individual preoperative tear sizes and retear rates
A greater number of anchors (p < 0.001) were used for fixation in the open group (mean, 4.0) than in the knotted group (mean, 2.1) and the knotless group (mean, 2.5) in the 6-month and 2-year cohorts, consistent with a two-row technique in the open surgery group. There also was a greater number of anchors (p < 0.01) used in the knotless group than in the knotted group at 6 months and 2 years.
No difference in operation time was noted between open (mean, 60 minutes) and arthroscopic knotted cuff repairs (mean, 55 minutes). Arthroscopic knotless repair (mean, 41 minutes) was faster than open cuff repair (p < 0.001) and arthroscopic knotted repair (p < 0.01).
After surgery, patients initially wore either a shoulder sling (open group) or an ultrasling (sling with a small abduction pillow) (arthroscopic groups) for 6 weeks. They began a gradually progressive home rehabilitation program as described by Hayes et al. [16
]. They began immediate postoperative ROM pendulum exercises. After the first postoperative visit at Day 8, the patients began passive forward flexion, external rotation, and abduction ROM exercises. Active ROM and simple isometric strengthening exercises were initiated at the 6-week postoperative visit. Active overhead activities and lifting 5 kg or more usually began 3 months postoperatively.
Standardized patient-determined [16
] and examiner-determined outcomes [16
] were obtained preoperatively and at 6 weeks, 3 months, 6 months, and 2 years postoperatively. This included the Shoulder Service Questionnaire, which was based on the Shoulder Rating Questionnaire (L’Insalata et al. [25
]) for assessment of functional capacity and physical symptoms. The information gathered allowed calculation of preoperative and postoperative ASES scores. The ASES score is based on activities of daily living scale and pain score and has a maximum of 100 points and was used as the primary outcome measure.
Intraoperative data were collected on standardized forms. The cross-sectional size of the rotator cuff tear was estimated and recorded on a diagram as described previously [4
]. The number and type of suture anchors used in the rotator cuff repair, and any additional procedures, were recorded.
Clinical testing was performed by a blinded independent observer (BB, JM, XW) and included visual estimation of ROM for forward flexion, abduction, external rotation, and internal rotation as described by Hayes et al. [17
]. Quantitative strength measurements of the shoulder in four orientations were measured using a HFG-45 Hand-Held Force Gauge (Transducer Techniques, Temecula, CA) as described by Hayes et al. [18
]. These measurements were used to calculate the rotator cuff functional index as described by Osbahr and Murrell [29
After the 6-month and 2-year postoperative evaluations, patients in each group had a standardized shoulder ultrasound study as described by Bryant et al. [4
]. Ultrasound accuracy has been validated at our institution for evaluation of cuff tear size preoperatively [4
]. Ultrasonography was performed on either an Acuson 128XP™ or Sequoia™ machine (Acuson Corp, Mountain View, CA) with a 7.5- to 12-MHz linear transducer or a General Electric Logiq®
9 (GE Corp, Fairfield, CT) with a 12-MHz linear transducer. Studies were performed by six experienced musculoskeletal sonographers following the standardized protocol. The ultrasonographic examination was performed as described by Teefey et al. [36
]. Static and dynamic imaging were performed. First, the biceps tendon was examined in the short axis from the level of the acromion inferiorly to the point where the tendon merged with the biceps muscle. The biceps tendon also was examined in the long axis. From the short axis biceps position, the subscapularis tendon was examined by externally rotating the arm. Images of the supraspinatus tendon were made with the shoulder extended, the elbow flexed, and the hand fully supinated. This position allowed maximum exposure of the supraspinatus tendon from under the acromion. The transducer orientation was maintained parallel to the tendon to observe the fibers in a longitudinal plane, and it was moved anteriorly and posteriorly and proximally and distally to observe the entire supraspinatus and infraspinatus tendons. The findings were reported on a form specifically designed to represent in two dimensions any defect in the rotator cuff [4
Results are reported as mean ± standard error of the mean. To answer the first and second questions, comparisons between groups were made with two-way paired Student’s t tests for parametric data, Mann-Whitney U tests for nonparametric data, the chi square test, and Kruskal-Wallis one-way analysis of variance on ranks. Correlations relating to the first and second questions were calculated using Pearson’s coefficient and multiple linear regression analysis. Analysis for the third question used Pearson’s coefficient and multiple logistic and linear regression calculations. Dunn’s test was applied to the regression analysis, which is a method that corrects for multiple comparisons, using the Bonferroni adjustment. To determine if the learning curve of arthroscopic repair had an effect on retear rate, the case number (where 1 was the first arthroscopic rotator cuff repair performed by the surgeon, 2 the second, and so forth) was included in the multiple regression analysis. All data were analyzed using SigmaStat®, Version 3.1 (Systat Software Inc, Richmond, CA).