Sublabral clefts and recesses in the anterior, inferior, and posterior glenoid labrum at MR arthrography
- First Online:
- Cite this article as:
- Tuite, M.J., Currie, J.W., Orwin, J.F. et al. Skeletal Radiol (2013) 42: 353. doi:10.1007/s00256-012-1496-0
- 897 Views
To determine the prevalence of a normal variant cleft/recess at the labral–chondral junction in the anterior, inferior, and posterior portions of the shoulder joint.
Materials and methods
One hundred and three consecutive patients (106 shoulders) who had a direct MR arthrogram followed by arthroscopic surgery were enrolled in this IRB-approved study. Scans were carried out on a 1.5-T scanner with an eight-channel shoulder coil. The glenoid rim was divided into eight segments and the labrum in all but the superior and anterosuperior segments was evaluated by two radiologists for the presence of contrast between the labrum and articular cartilage. We measured the depth of any cleft/recess and correlated the MR findings with surgical results. Generalized estimating equation models were used to correlate patient age and gender with the presence and depth of a cleft/recess, and Cohen’s kappa values were calculated for interobserver variability.
For segments that were normal at surgery, a cleft/recess was present within a segment on MR arthrogram images in as few as 7 % of patients (within the posteroinferior segment by observer 1), and in up to 61 % of patients (within the posterosuperior segment by observer 1). 55–83 % of these were only 1 mm deep. A 2- to 3-mm recess was seen within 0–37 % of the labral segments, most commonly in the anterior, anteroinferior, and posterosuperior segments. Age and gender did not correlate with the presence of a cleft/recess, although there was an association between males and a 2- to 3-mm deep recess (p = 0.03). The interobserver variability for each segment ranged between 0.15 and 0.49, indicating slight to moderate agreement.
One-mm labral–chondral clefts are not uncommon throughout the labrum. A 2- to 3-mm deep smooth, medially curved recess in the anterior, anteroinferior or posterosuperior labrum can rarely be seen, typically as a continuation of a superior recess or anterosuperior labral variant.
KeywordsShoulderMR arthrographyLabrumNormal variants
Multiple articles have been published describing normal labral variants of the shoulder on MR imaging [1–12]. These normal variants are portions of the labrum that are not completely attached to the articular cartilage, and have been described mainly in three regions of the glenoid rim. The first is in the superior portion of the glenoid rim where the caudal aspect of the labrum is partially unattached from the 11-o’clock to 1-o’clock position, and is known as a “superior recess” [3, 5]. The second occurs in the anterosuperior labrum where there are three subtypes. One is the sublabral foramen where the labrum is focally unattached to the glenoid rim . The second is the Buford complex where the anterosuperior portion of the labrum is absent and there is a thick, cord-like middle glenohumeral ligament . The third is an anterosuperior sublabral recess, similar to the superior recess . Finally, there is a study describing a sublabral cleft in the 7- to 8-o’clock region of the posteroinferior labrum .
Other than the regions of the normal variants, the labrum is usually described as firmly adherent to, or a rounded extension of the articular cartilage [15–17]. This feature has been recommended to help correctly diagnose on MR those labral tears that appear as a somewhat smoothly margined labral–chondral detachment [5, 7, 11].
Our purpose was to determine the prevalence of a labral–chondral junction cleft/recess in regions other than the superior and anterosuperior labrum on MR arthrogram images.
Subjects and methods
This study was approved by our Institutional Review Board. Our study plan was to retrospectively review MR arthrogram images and identify areas where we saw a cleft/recess. We would then look back at operative reports to identify areas where no labral tear, synovitis, or fraying was seen at arthroscopy. Finally, we would compare the MR-identified clefts/recesses with these normal areas of labrum to determine how often this normal variant MR finding is seen.
The study group consisted of 103 consecutive patients who had not had prior labral surgery and who underwent shoulder MR arthrography between June 2010 and June 2011 followed by arthroscopic surgery. Three patients had an MR arthrogram and surgery on both shoulders within the study period to make a total of 106 shoulders. The arthroscopic surgeries were performed by one of five Sports Medicine fellowship-trained orthopedic surgeons who have primarily a experience in shoulder surgery. There were 71 men and 32 women, with an average age of 36 years (range, 14–76 years). The primary postoperative diagnosis was impingement/rotator cuff tear in 35 shoulders, Bankart labral tear in 28, SLAP tear in 9, Bankart/SLAP in 8, posterior Bankart in 8, multidirectional instability in 5, a 360° labral tear in 5, internal impingement in 4, Bankart/posterior Bankart in 3, and a chondral lesion in 1 shoulder.
MR imaging technique
MR arthrogram pulse sequences
TR/TE eff (ms)
288 × 256
FSE Intermediate T2-weighted
288 × 256
288 × 256
288 × 256
288 × 224
MR image analysis
Clefts and tears had to be identified and confirmed on more than one image. Depth measurements were performed on the image plane most orthogonal to the particular segment.
The average time between the MR arthrogram and surgery was 75 days (range, 3–281 days). The surgeons were not blinded to the MR images or the original MR report. All the surgeons use a clock face to describe the extent of labral tears in their surgical reports, and these were used as the reference standard. At surgery, at least a portion of the 106 anterior labral segments (2:15–3:45 clock face portion) was abnormal (torn, frayed or affected by synovitis) in 51 shoulders, anteroinferior in 51, inferior in 34, posteroinferior in 28, posterior in 34, and posterosuperior in 62 shoulders. We determined the prevalence of a labral–chondral cleft/recess in the segments that were not found to be torn, frayed or synovitic. For the statistical analysis, the segments with a recess of at least 2 mm were all scored as ≥2 mm.
Because the surgeons may not have been exact in determining the clock face extent of tears, we also analyzed our data excluding the segments adjacent to torn labral segments. For example, 8 of the anterior segments were normal at surgery, but adjacent to a torn segment, so only 47 (106 minus 51 + 8) were considered in our more stringent analysis of normal anterior segments.
We also determined our MR arthrogram sensitivity for a labral tear, using the surgical finding of a tear (not including fraying or synovitis) as the reference standard.
Generalized estimating equation (GEE) models were used to assess whether clefts and ≥2-mm recesses were associated with age and gender using the combined grading of both readers. A logit link with independent correlation working structure was used, and odds ratios and 95 % confidence intervals were obtained based on the robust sandwich variance estimator. Inter-reader agreement was assessed with Cohen’s kappa (unweighted), for which 95 % confidence intervals were obtained . p < 0.05 (two-sided) was the criterion for statistical significance. All statistical graphics and computations were obtained in R 2.12.1 ; the “gee package” was used for GEE models .
Normal segment labral–chondral clefts/recesses
Prevalence of a labral–chondral cleft in all arthroscopically normal segments
26 (47 %)
17 (31 %)
12 (22 %)
24 (44 %)
18 (33 %)
13 (24 %)
32 (58 %)
8 (14 %)
15 (27 %)
25 (45 %)
18 (33 %)
12 (22 %)
64 (89 %)
5 (7 %)
3 (4 %)
43 (60 %)
22 (31 %)
7 (10 %)
71 (91 %)
5 (6 %)
2 (3 %)
64 (82 %)
13 (17 %)
1 (1 %)
54 (75 %)
16 (22 %)
2 (3 %)
51 (71 %)
17 (24 %)
4 (6 %)
17 (39 %)
10 (23 %)
17 (39 %)
25 (57 %)
16 (36 %)
3 (7 %)
Prevalence of the labral–chondral cleft or recess excluding normal segments adjacent to the torn segment
No cleft (%)
1 mm (%)a
2–3 mm (%)
No cleft (%)
1 mm (%)
2–3 mm (%)
Cleft or recess not from a labral folda
Of the 43 torn (out of 51 arthroscopically abnormal) anterior labral segments, Observer 1 graded 39 out of 43 as torn for a sensitivity of 0.91. The sensitivity was 38 out of 47 (0.81) for the anteroinferior segment, 21 out of 32 (0.66) for the inferior segment, 21 out of 24 (0.88) for the posteroinferior segment, 26 out of 28 (0.93) for the posterior segment, and 44 out of 46 (0.96) for the posterosuperior segment. For Observer 2, the sensitivity was 39 out of 43 (0.91) for the anterior labral segment, 43 out of 47 (0.92) for the anteroinferior segment, 24 out of 32 (0.75) for the inferior segment, 15 out of 24 (0.62) for the posteroinferior segment, 18 out of 28 (0.64) for the posterior segment, and 29 out of 46 (0.63) for the posterosuperior segment.
There was no association between age (p = 0.42) or gender (p = 0.25) and the presence of a cleft/recess. There was an association between a 2- to 3-mm-deep recess and male gender (p = 0.03), but not age (p = 0.11). The Interobserver variability for each segment ranged between 0.15 and 0.49, indicating slight to moderate agreement. The Cohen’s kappa values for the six segments were: posterosuperior 0.15, posterior 0.46, posteroinferior 0.35, inferior 0.34, anteroinferior 0.49, anterior 0.47.
Our findings show that the articular surface of the anterior, inferior, and posterior labrum is not always flush and continuous with the surface of the adjacent articular cartilage. The most common variant at the labral–chondral junction is a 1-mm cleft, a finding seen in about 40 % of the arthroscopically normal labral segments that we analyzed, mostly the anterior, anteroinferior, and posterosuperior segments. Our study confirms that these clefts are within the normal range of appearances of the glenoid labrum and do not need to be mentioned in MR reports.
A 2- to 3-mm deep, non-fold recess considered normal at surgery was seen by both observers in all but the inferior (5:15–6:45) and posteroinferior (6:45–8:15) segments of the labrum. A 2- to 3-mm recess was seen in 3–8 % of the labral segments that we analyzed. Our results show that a normal variant labral–chondral recess is rare in the six segments analyzed in our study, but they do occur at these additional sites around the glenoid rim.
Several papers have described the anterosuperior labral variants as occurring between the origins of the middle and inferior glenohumeral ligaments, and therefore extending as far inferior as the 3:30 o’clock position [11, 28], which would lie within our anterior segment (2:15–3:45 portion). We also had several shoulders with a recess in the anterior segment that were mainly due to inferior extension of an anterosuperior labral variant, and would not expect these to be confused with a tear (Fig. 3).
We identified a 2- to 3-mm non-fold recess in the anteroinferior segment (3:45–5:15 portion) in 6–8 % of normal segments in our patients. These may represent a far inferior extent of an anterosuperior labral variant in patients with a particularly caudal origin of the inferior glenohumeral ligament. If this recess is seen in patients without instability or a Hill–Sachs lesion, the attachment of the anterior labrum should be followed up to the anterosuperior portion to confirm an associated anterosuperior labral variant (Fig. 3). Recesses in the anteroinferior segment might be particularly confusing on MR because they can mimic partial labral detachment tears that occur after an anterior dislocation/subluxation. Previous papers have reported a specificity as low as 0.55–0.64 for direct MR arthrography of anterior inferior labral lesions , although most papers have reported specificities in the 0.86–0.98 range [30–34]. A labral–chondral recess may have contributed to some of the false-positive cases in these studies.
We found that 39–92 % of segments did not contain a labral–chondral cleft/recess at least 1 mm deep on MR arthrography. We had limited arthroscopic images of some of these segments, but in most the labrum was flush and continuous with the surface of the adjacent articular cartilage. Some shoulders graded as “MR normal” had a tiny groove on the arthroscopic images, but the groove was small and shallow relative to the 5-mm arthroscopy cannula and might have been obscured by partial averaging artifact on our MR images that had pixel dimensions up to 0.49 by 0.62 mm.
We were concerned that some apparent clefts and recesses could have simply represented a labral–chondral junction “kink” or “fold” if those patients’ labrum was bent into the joint toward the center of the glenoid fossa (e.g., the anterior labrum bent back posteriorly into the joint). We therefore drew a concave line along the surface of the articular cartilage and extended it out past the glenoid rim, and recorded if the labrum extended lateral to this line (Fig. 4). This feature was present in 33–37 % of the clefts/recesses in our study, and may have accentuated the depth of some clefts/recesses. Partial labral tears, such as an anterior labrum periosteal sleeve avulsion (ALPSA) lesion, usually displace medially toward the glenoid neck ; thus, this is another feature that can help distinguish a recess from a partial tear.
There is controversy whether some of the normal labral variants, especially the superior recess, constitute an anatomical variant or a developmental partial detachment. De Palma found that the superior recess becomes more prevalent with age and believed that it was an acquired lesion, possibly due to chronic repetitive traction [35, 36]. Tena-Arregui et al. found that a superior recess was not present in still-born fetal specimens . Lee et al., however, looked at a population with an average age of 45 years, and found that 46 % of the labral–chondral clefts in the posterior half of the labrum were in patients <30 years old . Deutsch et al. reported on CT arthrography that “air or contrast may normally track between the base of the labrum and the cartilage,” but didn’t say if this was only in certain portions of the labrum . We found that a 2- to 3-mm deep recess in the six regions of the labrum that we analyzed was more common in men, but there was no association with increasing age. Articles on multiple arthroscopy have stressed that a superior recess or an anterosuperior labral variant is an asymptomatic incidental finding at surgery and if “repaired” will only worsen symptoms [14, 39, 40]. Our surgeons do not repair or debride smooth, nonhemorrhagic labral–chondral junction recesses, and consider these a normal variant.
We also decided to carry out an analysis after excluding data from segments adjacent to arthroscopically abnormal segments. Although this reduced the total number of normal segments available for analysis, we did not have adequate arthroscopic images of some of our patients to confirm the accuracy of the extent of the labral tears described in the surgical reports. We therefore felt it was prudent to include an analysis of only segments that were clearly separate from abnormal segments.
The lack of arthroscopic images for all our clefts/recesses is one of the main limitations of our study because we could not correlate every one with a groove at arthroscopy, or confirm that there was not a pathological partial labral detachment. We attempted to determine indirectly if we might be overcalling clefts/recesses by calculating our sensitivity for labral tears. Our sensitivity for labral tears ranged from 0.62 to 0.96 for the different segments around the glenoid rim, which is within the range of most MR arthrogram articles that also report high specificity [30, 34, 41]. We also only called clefts or recesses with gadolinium or fluid signal, not just increased signal intensity, on the MR arthrogram images.
There are several additional limitations of our study. The surgeons were not blinded to the clinical history or the original MR arthrogram report so may have been biased in their arthroscopic assessment. Although we graded the labrum as either torn or a cleft/recess, we did not perform an in-depth analysis of the criteria for distinguishing a tear from a recess. We had a relatively small number of subjects with a 2- to 3-mm deep recess; thus, our prevalence may not be accurate. Our imaging criterion for calling a labral fold was only modestly objective. Finally, we had only slight to moderate interobserver agreement for identifying clefts in each segment. This may have been due to variability in grading the clefts, as well as whether a given slice was on one side or the other of the border between adjacent labral segments.
In summary, a cleft or recess at the labral–chondral junction is fairly common in the anterior, anteroinferior, and posterosuperior labrum, seen in 38–61 % of arthroscopically normal segments. 55–83 % of these are only 1 mm deep, and 33–37 % are at least partially due to a fold at the labral–chondral junction. Smooth, medially curved high signal at the labral–chondral junction on MR arthrogram images that appears to be a continuation of a the superior recess or anterosuperior labral variant should not be interpreted as a labral tear, particularly in the absence of a history of instability or additional findings such as a Hills–Sachs lesion.
Conflict of interest
The authors have nothing to disclose.