We included 1,380 patients who underwent either an open or an arthroscopic rotator cuff repair between 2012 and 2016 in Asan Medical Center, Seoul, Korea. The inclusion criteria were (1) full thickness rotator cuff tear, (2) and primary repair, (3) in patients with at least one follow-up magnetic resonance imaging (MRI) scan performed 12 months after surgery, and (4) at least 18 months of clinical follow-up. The exclusion criteria were as follows: (1) incomplete medical data (n=30), (2) previous surgery in the affected shoulder, (3) small (<1 cm), medium (1–3 cm), or massive tear (>5 cm), (4) concurrent subscapularis tear, acromioclavicular arthritis that required concurrent distal clavicle resection, superior labral lesions that require concurrent repair, long head biceps pathology that required tenodesis, severe glenohumeral arthritis, anterior glenohumeral instability, (5) bilateral rotator cuff tears, or (6) worker’s compensation case. Rotator cuff repair was performed by (I.H.J) and (J.M.C).

Under general anesthesia, the patients were positioned in the beach-chair position and were given an interscalene block to reduce postoperative pain. Examinations under anesthesia were performed prior to the surgical procedure to assess passive range of motion (ROM).

A standard posterior portal was created 2 cm inferior to and 1 cm medial to the posterolateral acromion corner. The anterior portal through rotator interval was introduced using an outside-in technique. A standard diagnostic round was performed. The arthroscope was then introduced into the subacromial space to assess the acromion undersurface. An anterolateral acromioplasty was routinely performed in all patients. Afterward, a lateral portal was created under direct visualization with the help of a spinal needle; this portal later served as the main viewing portal. A bursectomy was carried out to expose the rotator cuff tear and shape. The mobility of the rotator cuff was evaluated with a retriever. The edge of the rotator cuff was refreshed and trimmed with an arthroscopic shaver and/or a punch. The size of the tear was measured mediolaterally. Greater tuberosity was then prepared with a burr for attaching the remnant tissue. The number of anchors used was dependent on the size of the rotator cuff tear and the repair configuration (single or double row). In the single-row repair configuration, the rotator cuff was routinely fixed using a bio-composite PEEK (polyetheretherketone) anchor (Helicoil PK 4.5 mm; Smith & Nephew, Andover, MA, USA). In the double-row repair configuration, the rotator cuff was routinely fixed with a bio-composite PEEK anchor in a medial row (Helicoil PK 4.5 mm, Smith & Nephew) and lateral row (Footprint Ultra PK 4.5 mm, Smith & Nephew). An attempt was always made for tensionless repair with the maximum surface coverage of the footprint at the greater tuberosity.

A 5-cm skin incision was made longitudinally starting from the mid-point of one-third of the lateral margin of the acromion to the lateral border of the coracoid process. The deltoid was split longitudinally about 3–4 cm between the anterior and middle deltoid. A curvilinear incision was made to take down a small portion of the anterior deltoid, and the coracoacromial ligament was peeled off from the undersurface of the acromial spur and preserved for later reattachment. An anterolateral acromioplasty was routinely performed with an oscillating saw. Multiple non-absorbable traction No. 2-0 Mersilk (Ethicon, Cincinnati, OH, USA) sutures were placed along the edge of the torn rotator cuff to assist in mobilization of the tendon. Gentle release of adhesion and removal of any bursal hypertrophy were carried out using Mayo scissors with respect of the remnant rotator cuffs. Once adequately mobilized, the margin was converged with multiple tendon-to-tendon sutures when necessary, and the torn edge of the tendon was reattached to the greater tuberosity by No. 2-0 Ethibond (Ethicon, Cincinnati, OH, USA) in a trans-osseous, double mattress fashion (Fig. 1). The deltoid was repaired along with the coracoacromial ligament to the acromion with heavy absorbable No. 1 Vicryl (Ethicon) suture.

All patient arms remained in a sling for 6 weeks postoperatively, and only passive ROM was allowed during this time period. After 6 weeks, gradual full active motion was instituted, progressing to resistive strengthening, which was continued for three to 4 months. Heavy labor activities were restricted until at 6 months after surgery.

An independent nurse practitioner documented the clinical assessment of pre-operative and postoperative parameters for (1) pain score with visual analog scale (VAS) score, (2) functional outcome with age-adjusted Constant score and American Shoulder and Elbow Surgeons (ASES) score, (3) ROM (forward elevation and external rotation) with a hand-held goniometer, and (4) muscle power (abduction and external rotator muscle strength) assessed with a myometer (Mecmesin Co., Nottingham, UK). Any complications that occurred following surgery were also recorded.

All patients underwent radiological assessment with a 3-Tesla (3T) MRI at a minimum of one year following rotator cuff repair. The rotator cuff’s integrity was evaluated using the method described by Sugaya et al. [18] The supraspinatus and infraspinatus were evaluated for any fatty infiltration according to the method of Fuchs et al. [19] The deltoid origin muscle thickness was assessed with MRI according to Gerber et al. [20] for integrity, scarring, and thickness. The deltoid origin thickness was measured in four zones: the anterior acromion and the anterior third, middle third, and posterior third at the inferior surface of the lateral acromion border according to the scapular plane (Fig. 1). Any discontinuity of any part of the deltoid origin on all sequences in sequential MRI slices was defined as a deltoid tear (Fig. 2). Scarring of the deltoid was confirmed as a high signal area on T1-weighted images with preserved integrity (Fig. 3). The postoperative thickness of the deltoid in each region of interest of the lateral acromion border was then compared to its preoperative measurement (Fig. 4). The evaluations were done independently by four shoulder-fellowship trained orthopedic surgeons (EK, JMK, HJK, DJP), and any discrepancies were resolved in a consensus meeting; if disagreement persisted, a senior shoulder surgeon who was not involved in the surgery (KHK) was consulted for the final assessment. All radiologic parameters were recorded both pre- and postoperatively.

Sample size calculation with the a power of 90% and a 0.05 two-sided significance level was performed with the minimum expected clinical importance difference in means of constant shoulder score for 10.4 points [21]. A minimum sample size of 47 patients in each group (including an extra 10% due to the risk of loss to follow-up) was required. The propensity score matching technique was carried out with age, sex, affected shoulder, and body mass index as covariates. A total of 1,380 cases were recruited, but 30 were excluded because of incomplete medical data. Of the remaining 1,350 patients, 543 underwent open rotator cuff repair, while 837 underwent arthroscopic rotator cuff repair. Of the 543 open-repair patients, only 91 completed at least 18 months of clinical follow-up. Of these 91 patients, 56 had completed at least 1-year MRI and follow-up visits. The controls were the arthroscopic group that had been matched for age, sex, BMI, and affected side selected by propensity score matching (1:1 matching).

Tests for normality using the Kolmogorov-Smirnov method were applied to all datasets prior to statistical analysis. A Mann-Whitney U-test was used to compare any datasets with a skewed distribution, while an independent t-test was carried out to compare the datasets with a normal distribution. The significance level was set at P<0.05. Statistical analysis was performed using IBM SPSS ver. 23.0 (IBM Corp., Armonk, NY, USA) under the supervision of a biostatistician.

A total of 112 patients was included for analysis. The characteristics and preoperative baseline data of the open group and arthroscopic group are shown in Tables 1 and 2. There were no significant differences in demographic characteristics between these two matched groups.

At the mean follow-up of 19.1 months, ROM, Constant score, ASES score, muscle power, and VAS were significantly improved following surgery. The arthroscopic group showed significantly better ROM (P=0.006) and VAS score (P<0.001) compared to the open group. In contrast, the open group demonstrated significantly better Constant and ASES scores (P=0.012 and P=0.047, respectively). The muscle powers for abduction and external rotation were superior in the open group, though there was no statistical difference (P=0.068 and P=0.182, respectively). No complications were seen in either group. The postoperative clinical outcomes are shown in Table 3.

Fatty infiltration of the supraspinatus and infraspinatus showed no significant difference between the two groups. The retear rate was higher in the arthroscopic group (21.4%) compared to the open group (17.8%), though the difference was not statistically significant (P=0.300). No patient required revision surgery (even in the case of a cuff tendon re-tear and deltoid injury) at the final follow-up as all were in an asymptomatic state. The deltoid origin thickness was significantly greater in the arthroscopic group when measured at the anterior acromion area (6.2±1.4 vs. 4.9±1.1 mm, P=0.006), anterior third (6.3±1.3 vs. 4.5±0.9 mm, P=0.005), and middle third of the lateral acromion border (6.7±1.3 vs. 4.6±1.0 mm, P =0.005). The arthroscopic group had significantly higher intact deltoid integrity with less scarring (P=0.04). There were no full-thickness deltoid tears in our observations for both open and arthroscopic groups. The postoperative thickness of the deltoid insertion was significantly maintained with less than 50% reduction from its preoperative thickness in the arthroscopy group (80.7%) compared to the open group (64.1%; P=0.04).

The most significant limitation of this study is its retrospective nature. The included population was ideal for a clinical study, but it still may limit extrapolation of our findings to a general population, leading to selective bias in this study. This study also only included large tears, which limited translation of our results to all rotator cuff tear sizes. Despite these limitations, we attempted to minimize bias by excluding small, medium, and massive rotator cuff tears from our study design to provide a straightforward result. This study compared the clinical outcomes at the final follow-up visit as opposed to the postoperative timeline. MRI was used to evaluate deltoid origin status; however, the surgical repair itself may result in muscle scarring and thinning, which can lead to false positives on MRI scans. Nevertheless, despite these limitations, the accuracy provided by a 3T MRI scanner should minimize the incidence of false positives. One additional limitation was that it is difficult to blind the radiologic evaluation process due to the holes drilled in the acromion following open rotator cuff repair. Lastly, we suggest that future studies develop a better methodology to evaluate the deltoid status postoperatively.

The current study showed that there was a change in deltoid origin status following both open and arthroscopic repairs of large rotator cuff tears. Open rotator cuff repair resulted in a thinner deltoid origin, especially in the anterior acromion to the middle third of the lateral border of the acromion at the 1-year postoperative MRI evaluation. Meticulous reattachment of the deltoid origin is as essential as a proper rotator cuff repair for the open procedure.