Three medical centers were queried regarding patients with a rotator cuff tear from January 2008 to January 2013. Fig. 1 shows the flow chart of data source, sampling and analysis. A total of 17,286 patients were identified using search terms. Search terms were comprised of diagnosis as 'rotator cuff tear' and assessment methods as 'shoulder ultrasonography', 'shoulder magnetic resonance imaging (MRI)' or 'shoulder MR arthrography'. Among the patients who were identified using these search terms, the patients who had a high-grade partial-thickness (fiber disruption more than 50%) or small-to-medium-sized full-thickness tear (≤3 cm) were extracted randomly in this study. Patients were excluded if they met any of the following criteria: 1) less than 50 years of age, 2) severe adhesive capsulitis (<120° in forward flexion or <40° in external rotation), 3) paralysis affecting shoulder girdle muscles, 4) low-grade partial-thickness (fiber disruption less than 50%) or large-to-massive-sized full-thickness tear (≥3 cm) preexisted, 5) history of fractures, instability or septic shoulder, 6) open or mini-open repair or 7) loss to follow-up without second visit of outpatient clinic. After the patient list was determined, convenient sampling was conducted for each conservative and arthroscopic repair group. Demographic data for the patients in the final list were then collected in the medical records regarding sex, age, tear on dominant side, pain onset and existence of trauma history. Age and pain onset were determined based on the time at the initial visit of the outpatient clinic. Acute onset of pain was defined as occurring less than three months before the initial visit. This study was approved at three medical centers and an agency to which all the authors belonged.

Rotator cuff tendons were evaluated using shoulder ultrasonography, shoulder MRI or MR arthrography in terms of tear thickness, size and fatty degeneration. A partial thickness tear was classified as 'low grade' if less than 50% fiber was disrupted and 'high grade', if more than 50% was disrupted. A full thickness tear was classified as small (<1 cm), medium (1–3 cm), large (3–5 cm) and massive (>5 cm or more than two tendons involved) [15]. The tear size of rotator cuff tendons was determined by measuring the length of retraction and width in each sagittal and coronal plane seen on ultrasonography, MRI or MR arthrography images (Fig. 2). Fatty degeneration was assessed using shoulder MRI or MR arthrography according to Goutallier classification [16].

Primary outcome measures were a pain assessment score and range of motion (ROM). Pain was evaluated with a numeric rating scale or visual analog scale from 0 to 10. Range of motion was measured passively for forward flexion and internal rotation by an examiner. Internal rotation was measured using a vertebral level at the back. Secondary outcome measures were the rate of tear progression or retear, and the rate of symptom aggravation, after treatment. Aggravation cases were defined as patients with a pain assessment score greater than that of baseline or forward flexion and internal rotation ROM less than that of baseline. The clinical outcomes were measured at initial, 2 to 6-month and 1-year visits of the outpatient clinics and compared between conservative treatment and arthroscopic repair groups. A trained interviewer conducted a telephone interview to assess pain and ROM in patients who were lost to follow-up in a year. During the telephone interview, ROM was measured using several landmarks of the body. In the case of forward flexion, arm elevation next to the ear at the end of arc without neck motion was regarded as full ROM. The buttock, lumbosacral junction, pelvic bone, waist and scapula were used as landmarks to measure internal rotation.

The conservative treatment group included all patients with a rotator cuff tear who were not treated by surgical repair. The treatment comprised of active surveillance, a therapeutic modality, oral pain medication, steroid injection and shoulder exercises. Active surveillance included patient education related to activity modification, control of symptoms or prognosis of the disease, without any medication or intervention [3]. The therapeutic modality included transcutaneous electrical nerve stimulation, heating pad or therapeutic ultrasound. Oral pain medication included non-steroidal anti-inflammatory drugs or opioid analgesic agent such as tramadol/acetaminophen. Intra-articular or subacromial-subdeltoid bursa steroid injection was administered blind or guided using ultrasonography by physiatrists or orthopedic surgeons. The shoulder exercise protocol consisted of active and passive ranges of motion as well as strengthening exercises of the rotator cuff and scapular stabilizing muscles. Surgical treatment was restricted to arthroscopic repair, either with or without acromioplasty, and debridement. Open and mini-open repair were excluded. After surgical treatment, the arthroscopic repair group was provided with postoperative management that included shoulder exercises, pain medication or corticosteroid injection. The shoulder exercise was initiated to prevent stiffness after 5–6 weeks of postoperative immobilization. Pain medication and corticosteroid injection within 3 months after surgical treatment was regarded as postoperative management in this study.

To reach statistical significance with alpha=0.05 and power=0.9 for the pain assessment score on the assumed statistical behavior of other previous studies, the number of patients was calculated and 142 patients were needed in the conservative treatment and arthroscopic repair, respectively [14]. By adding 20% of patients in consideration of the missing data, more than 171 patients were needed in each group.

Demographic factors including sex, age, tear on dominant side, pain onset, and existence of trauma history were analyzed with a chi-square test. Likewise, tear classification and fatty degeneration at the initial visit were analyzed with the same method. To quantify the extent of internal rotation, each vertebral level was scaled to be a numerical value from 1 to 16. The sacral level was represented as 1, L5-L1 level as 2–6 and T12-T3 as 7–16. To investigate the simultaneous effects of time and treatment on clinical outcomes, a comparison of clinical outcomes was performed using analysis of variance with repeated measurements and a general linear model. In this model, covariates were used including age, sex, tear on dominant side, trauma history, symptom onset and tear classification. The rate of tear progression in the conservative treatment group or retear in the arthroscopic repair group was analyzed with the chi-square test. The aggravation of clinical outcomes after both treatments was analyzed using the same method. All statistical analyses were performed using SPSS software ver. 19 (SPSS Inc., Chicago, IL, USA). The significance level was set at p<0.05.

Of the patients available for review, 183 who received conservative treatment and 174 who underwent arthroscopic repair met the study inclusion criteria and were enrolled in this study. A telephone interview was conducted for 51.6% of patients in the conservative treatment group and 50.3% in the arthroscopic repair group, who were lost to follow-up at 1-year (p=0.85). Demographic data of the patients at the baseline shows that there was no significant difference between the two groups for each category (Table 1). The mean age was 64.6±8.7 years in the conservative treatment group and 62.3±6.6 years in the arthroscopic repair group.

The changes of pain assessment score and ROM in the 1-year follow-up period are shown in Fig. 3 and Table 2. The pain assessment score and ROM in degrees of forward flexion improved significantly after treatment in both groups (p<0.001) at follow-up after 1 year, but between-group differences at the baseline and at 1 year were not significant. ROM in internal rotation did not significantly change after both treatments over the year period (p=0.34). The subgroup analysis for high-grade partialthickness and small-to-medium-sized full-thickness tears showed that the pain assessment score as well as forward flexion and internal rotation ROMs were not significantly different after arthroscopic repair or conservative treatment (Tables 3, 4). In the subgroup analysis for the arthroscopic repair group, patients who were prescribed with postoperative pain medication, corticosteroid injection or shoulder exercises did not show any difference according to their pain assessment score and ROM for forward flexion and internal rotation as compared with non-prescribed patients.

Rotator cuff tendons were reevaluated for 125 patients (71.8%) in the arthroscopic repair group and 45 patients (24.6%) in conservative treatment group at follow-up 1 year later. There were 12 cases (9.6%) of retear in the arthroscopic repair group and 3 cases (6.7%) of tear aggravation in the conservative treatment group. Retear occurred in 4 patients with the high-grade partialthickness tears and in 8 with the small-to-medium-sized full-thickness tears (p=1.00). Tear progression occurred in 2 patients with high-grade partial-thickness tears and 1 with a small-to-medium-sized full-thickness tear (p=0.57). Retear and tear progression were not associated significantly with sex, age >60, tear on dominant side, trauma history, symptom onset or tear classification.

Table 5 showed that the proportion of those with aggravation for pain and ROM was not significantly different between the conservative treatment and arthroscopic repair groups. There was only an association with increased risk for pain aggravation in males of the conservative treatment group (p=0.034). Other factors such as age, tear on dominant side, pain onset, trauma history, and tear classification were not significantly associated with aggravation of pain and ROM in both groups. In the subgroup analysis of the arthroscopic repair group, the aggravation of pain assessment score and ROM was not significantly different between the groups with and without postoperative pain medication, corticosteroid injection or shoulder exercise. Retear in the arthroscopic repair group and tear progression in the conservative treatment group were not associated significantly with the aggravation of pain and ROM.