This was a longitudinal cross-sectional study approved by the Institutional Review Board at the Gil Medical Center, Gachon University (No. GBIRB2014-329). Between March 2012 and July 2014, 95 consecutive patients who underwent arthroscopic repair for partial thickness rotator cuff tears, operated by a single senior surgeon, were selected for the study. All patients had undergone totally 6 months conservative treatments. The preoperative magnetic resonance imaging (MRI) was examined for all cases, and arthroscopic evaluation confirmed the size of partial thickness rotator cuff tears. Cases selected for the final study included only those subjects with partial thickness rotator cuff tears of 40% to 70% on the bursal or articular side, based on the size of the footprint of the supraspinatus. Cases less than 40% partial thickness tear on MRI with no subsequent surgery were excluded from the study. Additionally, nearly full thickness tears of 70% or more were excluded, since it was difficult to clearly distinguish partial tear of only bursal or articular region. Since the intratendinous tear or severe tendinopathy on MRI is also difficult to perceive, it is theoretically necessary to perform a ballooning test with saline injection into the tendon to identify the definite intratendinous tear. Hence, subjects with intratendinous partial tears were also excluded, due to the vagueness of diagnosis itself. Lastly, definite traumatic rotator cuff tear and rotator cuff tear with severe osteoarthritis were also excluded. Sample sizes were calculated to detect a significant difference. Prior to this study, the authors conducted a pilot study with 20 cases. Spur was found in 13 cases. Among them, 8 cases were heel spur. The effect size was thus calculated as 0.33. Accordingly, the sample size was determined to be 73, and a total of 80 cases were finally confirmed by adding extra cases, considering the two-sided confidence of 95% (type I error) and the power of 80%. Thus, the entire study group consisted only of bursal or articular side partial thickness tears, and 80 cases were finally enrolled for this study. Of the total 80 cases, 41 cases (51.3%) were male. The mean age was 58.0 ± 8.0 years; among total cases, 51 cases (63.8%) were dominant hands, and 43 cases (53.8%) were right shoulders (Table 1) [13].

The Bigliani’s type of the acromion was diagnosed by plain X-ray. According to the Bigliani’s classification, the type of acromion are classified as flat (type I), curved (type II), and hooked (type III) on the supraspinatus outlet view of shoulder radiographs [13]. We observed 6 cases (7.5%) type I, 55 cases (68.8%) type II, and 19 cases (23.7%) type III (Table 1) [13].

All patients underwent preoperative MRI scans, performed using a 3.0-tesla scans (Skyra; Siemens Medical, Berlin, Germany). Oblique coronal, oblique sagittal, and axial T2-weighted fat-saturated images were acquired for assessment of the rotator cuff tendons. Slice thicknesses on the oblique coronal, oblique sagittal, and axial views were 3 mm, 4 mm, and 3 mm, respectively. The interslice gap on the coronal, sagittal, and axial views was 20%, 40%, and 10%, respectively. There were no cases of enhanced MRI [14].

The subacromial spur was evaluated on plain shoulder X-rays and shoulder MRIs, and classified as heel and traction spur. Heel spur is defined as a heel-like irregular bony protrusion located below the anterolateral side of the acromion. True anteroposterior view and supraspinatus outlet view of plain X-ray, and oblique coronal and sagittal image of MRI revealed thickened bony structures with irregular shapes, which were subsequently confirmed by arthroscopic findings (Fig. 1). Traction spur is defined as an anterior protruding bony structure running along the coracoacromial ligament from the anterolateral corner of the acromion. The short and sharp bony protrusions were confirmed on the true anteroposterior view of plain X-ray and oblique coronal image of MRI. Furthermore, bird-beak like protruding bony structures along the coracoacromial ligament from beneath the acromion were identified on supraspinatus outlet view of plain X-ray and sagittal image of MRI. The final diagnosis was made in arthroscopic findings (Fig. 2). In some cases, it was difficult to accurately classify the shape of the spur by X-ray; hence, final classification was done by considering all findings from X-ray, MRI, and arthroscopy. Bony sclerotic changes and minimal bony irregularity on the X-ray and MRI, with no indication of any prominent bony project, were considered to have no subacromial spur. In the current study enrolling 80 cases of acromion, subacromial spur was observed in 55 cases (68.7%), of which heel spur and traction spur accounted for 38 cases (47.5%) and 17 cases (21.2%), respectively (Table 1) [13].

The location and the size of the rotator cuff tear were classified as articular and bursal side tear by oblique coronal MRIs. These findings were subsequently confirmed by arthroscopic examination. In this study, we confirmed 26 cases (32.5%) having articular side tear, and 54 cases (67.5%) with bursal side tear (Table 1) [13].

For comparing the no spur and spur groups, age was analyzed with student’s t-test, and sex, dominant hand, right or left side, type of tear, and Bigliani’s type of acromion were analyzed using chi-squared test. However, in several variables, more than 20% of the expected counts of each cell on the contingency table were less than 5. For those cases, we applied the Fisher’s exact test. In cases that the data table was larger than 2×2, the Freeman-Halton extension of the Fisher’s exact test was used, instead of the traditional Fisher’s exact test [15]. Univariate binary logistic regression analysis was used to analyze the effect of spur on the bursal side rotator cuff tears. Multivariable logistic regression analysis evaluated the effect of spur on the bursal side rotator cuff tears, according to the all independent variables, using the backward elimination procedure. The condition for the inclusion of variables was 0.05 [14].

For comparing the heel spur and traction spur groups, age was analyzed with Mann–Whitney U-test as a non-parametric test (since size of samples was only 17), whereas sex, dominant hand, right or left side, type of tear, and Bigliani’s type of acromion were analyzed by the chi-squared test. Fisher’s exact test and the Freeman-Halton extension of the Fisher’s exact test were similarly adopted to the variables, as above. Analysis of covariance (ANCOVA) is the most appropriate method if random sampling is not performed, and data with unclear causality is taken into consideration for various covariances that may affect the outcome. In this study, however, logistic regression was performed since data were identified as non-normal and non-equivalent data. Univariate logistic regression analysis was used to analyze the effect of heel spur on the bursal side rotator cuff tears. Multivariable logistic regression analysis was used to evaluate the effect of heel spur on the bursal side rotator cuff tears, similar to the procedure described above.

All radiologic and arthroscopic assessments were performed by two board-certified orthopedic surgeons having more than 5 years’ experience. In case of differing interobserver categorization, the final decision was made by a senior orthopedic surgeon who did not participate in the measurements. Cohen’s kappa was calculated as a measure of agreement. According to Landis and Koch [16], the following ratings for the interpretation of kappa were used: poor (0.20), fair (0.21–0.40), moderate (0.41–0.60), good (0.61–0.80), and excellent (0.81–1.00). However, a kappa value less than 0.5 was usually estimated as lower reliability. In this study, the intra-observer and inter-observer coefficients were 0.86 (CI, 0.82–0.91) and 0.75 (CI, 0.69–0.80), respectively. All statistical tests are 2-tailed, and p<0.05 is considered statistically significant. The SAS software package (ver. 9.4 TS Level 1M3; SAS Institute, Cary, NC, USA) was used as a statistical analyzer.