Magnetic Resonance Imaging Does Not Accurately Estimate the Size of Glenoid Labrum Tears in Patients With Shoulder Instability

Abstract

Purpose

To compare magnetic resonance imaging (MRI)‐based measurements of glenoid labral tears with intraoperative measurements.

Methods

A retrospective review was performed of patients that underwent shoulder capsulorrhaphy between 2019 and 2024 at a single academic medical center. All patients with a single labrum tear, an MRI within 6 months of surgery, and historical tear measurements or measurable intraoperative photographs were included in the study. Glenoid labrum tear size was noted using the clockface convention described in “hours.” When clockface measurements were not reported on MRI reads, measurements were made by a radiologist. When clockface measurements were not reported in operative notes, measurements were made by an orthopaedic surgeon using intraoperative photographs. A paired t test was used to evaluate any difference between the MRI and intraoperative measurements.

Results

Two hundred sixteen glenoid labral tears from 211 patients met inclusion criteria. Average age was 27.2 ± 11.8 years, 68.5% of patients were men, and 56.9% of glenoid labrum tears were right‐sided. On average, MRI measurements exceeded intraoperative measurements (4.15 ± 2.90 hours vs 2.80 ± 1.59 hours) with an average difference of 1.36 ± 3.09 hours (95% confidence interval = 0.94‐1.77, P < .001) and a correlation of 0.154 (P = .023). A greater difference in measurement was observed among men compared with women (1.64 vs 0.74 hours, P = .044).

Conclusions

MRI overestimated glenoid labrum tear size compared with intraoperative measurements and differed substantially between the 2 modalities on a case‐by‐case basis, with men displaying a larger discrepancy in measurement difference than women. These findings indicate that MRI may be more appropriately used as a categorical tool for confirming the presence of glenoid labral tears, rather than as a precise measure of tear size.

Level of Evidence

Level III, retrospective diagnostic case series.

The glenohumeral joint is the most unstable articulation in the human body according to studies examining its anatomy and physiology. ¹ , ² Because of subsequent glenohumeral dislocations, glenoid labral tears are a common pathology. Given the high recurrence rates and functional limitations associated with shoulder instability, primary stabilization is a good option for many younger and more active patients. ³ Although indications for operative fixation of labral tears focus on instability, bone loss, and Hill‐Sachs lesions, ⁴ some studies have associated larger tears with higher rates of instability on presentation ⁵ and high rates of postoperative complication. ⁶ In addition to high‐risk symptoms and postoperative complications, tear size has implications for informed consent, ⁷ shared decision making, ⁸ and surgical planning. Larger tears are associated with more anchors being placed intraoperatively ⁹ , ¹⁰ and have played a role in the debate between lateral decubitus and beach chair positioning. Baron et al. showed that surgeons who operated on patients in the lateral decubitus position reported greater labral tear size, ¹¹ indicating that longer labral tears that extend to the inferior aspect of the labrum were better visualized in this position. Although outcomes have been similar between surgeon‐preferred positions, ⁹ , ¹² , ¹³ the debate continues between experts. ¹⁴ , ¹⁵

Prior to the advent of advanced imaging techniques like arthrograms or magnetic resonance imaging (MRI), clinical examination and arthroscopy were the gold standard of diagnosis and evaluation of labral tears. Because of its noninvasive nature and its lack of ionizing radiation, MRI is now a prominent tool in the workup of labrum pathology. ¹⁶ , ¹⁷ Despite its wide use in orthopaedic surgery, its diagnostic accuracy in characterizing glenoid labrum tears remains unclear. Although Liu et al. showed an identical specificity for labrum pathology between MRI and clinical examination, their study showed a 90% sensitivity by clinical examination and 59% for MRI, ¹⁸ with a more recent meta‐analysis indicating a sensitivity of 76%. ¹⁹ Beyond identifying the presence of a labrum tear, the clockface measurement is conventionally used to quantify glenoid labrum tear size on MRI. ²⁰ In an effort to better understand MRI's diagnostic capabilities, Campbell et al. performed a study comparing MRI results of meniscus tears to intraoperative measurements, ²¹ finding that MRI typically measured smaller meniscus tears than were found intraoperatively.

The purpose of this study was to compare MRI‐based measurements of glenoid labral tears with intraoperative measurements. The authors hypothesized that there would be no statistically significant difference between MRI and intraoperative clockface measurements of glenoid labrum tears.

METHODS

This was a retrospective diagnostic review of patients that underwent shoulder capsulorrhaphy at 1 academic institution between January 2019 and December 2024. This study received approval by the University of Cincinnati's Institutional Review Board under OS20007 IRB 2020‐0116. Using Current Procedural Terminology code 29806, all patients that underwent capsulorrhaphy between January 2019 and December 2024 were isolated. Patients were included in the study if they had a shoulder MRI within 6 months of surgery with a clockface measurement in the respective radiology note or an available MRI from which to make measurements. Patients also required a diagnosed labrum tear intraoperatively, and either archived intraoperative images or a description of the size of a labrum tear documented in their operative note for inclusion.

Patients were excluded from the study if they had metal artifact on MRI, if the MRI was read at an outside institution, if the MRI read was signed by a radiologist that had not completed a musculoskeletal imaging fellowship, if they underwent magnetic resonance arthrogram rather than an MRI, had multiple glenoid labrum tears, if they underwent a prior glenohumeral based procedure, or if their intraoperative images were inadequate to make measurements.

The range of MRI parameters included a field of view of 100 to 200 mm, matrix of 256 to 640, and a pixel size of 0.179 to 0.781 mm. All MRI scans were performed using 1.2, 1.5, or 3.0 T magnets. All patients that did not have a historical measurement of labrum tear size in their radiology note or operative note had one or both measurements made by the authors. MRI measurements were made by a radiologist (A.S.C.) with over 5 years of experience who has completed a musculoskeletal imaging fellowship. The reading radiologist was aware that patients underwent capsulorrhaphy for shoulder instability but was blinded to intraoperative findings. After education and initial oversight by a senior author (B.M.G.) that has completed a sports medicine fellowship, intraoperative measurements were made by a third‐year orthopaedic surgery resident (C.C.) based on intraoperative images found in the electronic medical record (EMR).

Through examination of the EMR, detailed chart review was performed on these patients for basic patient demographics including sex, height, body mass index, and injury laterality; radiographic data including date of imaging, type of imaging (MRI or arthrogram), presence or absence of labral tear, and clockface measurement of labral tear size as seen in Figure 1; and surgery data including date of surgery and clockface measurement of labral tear size, as shown in Figure 2.

FIGURE 1.

Representative example of clockface convention as measured on a sagittal cut of an example shoulder MRI (right). Note: the above‐drawn measurement overlaying the provided MRI slice may not be transferrable to other patients, and the recording of a labrum tear size would require viewing multiple sagittal slices to ensure proper measurement. There is also no guarantee that a patient's glenoid labrum/fossa is perfectly perpendicular to the MRI scanner, which reinforces the need for multiple slices and views to be examined before recording a tear size. (Cl, clavicle; Co, coracoid process; G, glenoid fossa; MRI, magnetic resonance imaging; Sc, scapular spine.)

FIGURE 2.

Representative example of clockface convention as measured intraoperatively, this example a right shoulder glenoid labrum measuring a 2:30‐5:00 tear from the anterolateral portal. Note: the above‐drawn measurement overlaying the provided intraoperative image may not be transferrable to other patients, and the recording of a labrum tear size would require viewing multiple images to ensure proper measurement. (G, glenoid fossa; H, humeral head; L, glenoid labrum.)

Statistical Analysis

Statistics were carried out using IBM SPSS Statistics Version 30.0.0. Mean and standard deviation were calculated for continuous variables. A paired t test was performed comparing MRI measurements to intraoperative measurements. Independent‐sample t tests were performed to evaluate the similarity of the measurements made at the time of MRI or surgery and those made by the authors, to compare the differences in measurement between sexes, and to compare difference in measurement between 1.5 and 3.0 T magnet strength. An analysis of variance was performed to analyze differences in MRI measurement by year. Linear regressions were performed to examine predictor variables with continuous outcomes. An ordinal regression was performed to examine predictor variables with the qualitative comparison of MRI versus intraoperative measurement. A significance value of a = 0.05 was established.

RESULTS

The study included 216 glenoid labrum tears in 211 patients (148 male [68.5%], 123 right‐sided [56.9%], and a mean age 27.2 ± 11.8 years). A Strengthening the Reporting of Observational Studies in Epidemiology diagram is shown in Figure 3 to describe the patients that were included in the study and who made measurements of labrum tear size. Comparing averages, MRI measurements exceeded intraoperative measurements (4.15 ± 2.90 hours vs 2.80 ± 1.59 hours). The average difference between the two measurements was 1.36 ± 3.09 hours (95% confidence interval [CI] = 0.94‐1.77, P < .001) and the correlation was 0.154 (P = .023). Male patients’ MRI measurements overestimated intraoperative measurements (4.61 ± 2.64 hours vs 2.97 ± 1.70 hours) with an average difference of 1.64 ± 3.04 hours (95% CI = 1.15‐2.14, P < .001). Female patients’ MRI measurements were not significantly different from their intraoperative measurements (3.15 ± 3.19 hours vs 2.42 ± 1.28 hours), with an average difference of 0.74 ± 3.08 hours (95% CI = −0.01 to 1.48, P = .053). Male tear measurements were 1.46 hours (P < .001) and 0.55 hours (P = .018) larger than female measurements for MRI and intraoperative measurements, respectively. The male measurement difference was 0.91 hours greater than the difference in measurements for female patients (P = .044). Individual categorical comparisons between MRI and intraoperative measurements are given in Table 1, with MRI overestimating intraoperative measurements in most cases (P < .001). Of note, there were 20 cases (9.3%) in which the MRI read did not indicate a glenoid labrum tear.

FIGURE 3.

STROBE diagram exhibiting included patients, reasons for exclusion, and how measurements were made. (CPT, current procedural terminology; MRI, magnetic resonance imaging; STROBE, Strengthening the Reporting of Observational Studies in Epidemiology.)

TABLE 1.

Categorical Comparisons Between MRI and Intraoperative Clockface Measurements of Glenoid Labrum Tear Size

Size Comparisons	Number of Cases (% of Total)
MRI > Intraoperative	133 (62%)
MRI = Intraoperative	20 (9.3%)
MRI < Intraoperative	63 (29%)

MRI, magnetic resonance imaging.

The measurements made at the time of the original MRI and those made by the author were not significantly different (P = .123), nor were the measurements made at the time of surgery and those made by the author (P = .679). When comparing MRI measurements by year, measurements did not differ (P = .640). Patient MRIs using a 1.5 T magnet (n = 194) and those using a 3.0 T magnet (n = 18) did not show a statistical difference in their overestimation of labrum tear size (1.31 ± 3.10 hours vs 0.90 ± 2.12 hours, respectively [P = .583]). Although there were only 4 patients that underwent MRI by a 1.2 T magnet, MRI overestimated their tear size by an average of 5.50 ± 3.42 hours.

Male sex was associated with an increased labral tear size as measured by MRI (Table 2). No other recorded demographic or injury factors were associated with increases in labral tear size or in differences between the 2 modalities. An ordinal regression was performed to evaluate predictor variables for categorical size comparisons, with no demographic or injury factors being associated with increases in labral tear size or in difference between the 2 modalities (Table 3). Table 4 displays the frequency of historical and retrospective measurements and their respective comparisons.

TABLE 2.

Linear Regressions Comparing Predictor Variables With MRI, Intraoperative, and Difference in Clockface Measurements of Glenoid Labrum Tear Size

	MRI Measurement		Intraoperative Measurement		Difference Between MRI and Intraoperative Measurement
	B Coefficient (95% Confidence Interval)	P Value	B Coefficient (95% Confidence Interval)	P Value	B Coefficient (95% Confidence Interval)	P Value
Height	0.000 (−0.006 to 0.006)	.90	−0.001 (−0.005 to 0.002)	.44	0.000 (−0.006 to 0.006)	.90
Weight	0.007 (−0.046 to 0.061)	.78	0.011 (−0.019 to 0.040)	.48	0.007 (−0.046 to 0.061)	.78
BMI	−0.041 (−0.22 to 0.14)	.66	−0.055 (−0.16 to 0.045)	.28	−0.041 (−0.22 to 0.14)	.66
Sex	1.4 (0.38 to 2.4)	.008	0.51 (−0.060 to 1.1)	.079	1.4 (0.38 to 2.4)	.008
Laterality	0.23 (−0.55 to 1.0)	.57	0.18 (−0.26 to 0.61)	.422	0.23 (−0.55 to 1.0)	.57

Note: Bold indicates P < .05.

BMI, body mass index; MRI, magnetic resonance imaging.

TABLE 3.

Ordinal Regression Evaluating Predictor Variables With the Categorical Comparisons of MRI and Intraoperative Measurements

	B Coefficient (95% Confidence Interval)	P Value
Time to surgery from imaging	−0.001 (−0.008 to 0.005)	.65
Height	0.002 (−0.009 to 0.013)	.67
Weight	−0.001 (−0.040 to 0.039)	.98
BMI	0.017 (−0.12 to 0.15)	.80
Sex	−0.67 (−1.4 to 0.051)	.069
Laterality	0.151 (−0.41 to 0.71)	.59

BMI, body mass index; MRI, magnetic resonance imaging.

TABLE 4.

Frequency of Historical and Retrospective MRI Reads and Intraoperative Measurements and Their Respective Measurement Comparisons

	Number, %	MRI Measurement (Average h ± SD)	Intraoperative Measurement (Average h ± SD)	Difference in Measurement (Average h ± SD, (95% CI))	P Value
Historical MRI read vs historical intraoperative measurement	75 (34.7%)	3.80 ± 1.79	2.95 ± 1.66	0.85 ± 2.22 (0.34‐1.36)	.001
Retrospective MRI read vs historical intraoperative measurement	113 (52.3%)	4.02 ± 3.32	2.72 ± 1.59	1.30 ± 3.40 (0.66‐1.93)	<.001
Historical MRI read vs retrospective intraoperative measurement	0 (0%)	‐	‐	‐	‐
Retrospective MRI read vs retrospective measurement	28 (13.0%)	5.64 ± 3.11	2.68 ± 1.47	2.96 ± 3.33 (1.67‐4.25)	<.001

Note: There were no cases where a historical MRI clockface measurement was made that required a retrospective measurement to be made based on intraoperative photographs.

CI, confidence interval; h, hour; MRI, magnetic resonance imaging; SD, standard deviation.

DISCUSSION

The most important finding of this study was that compared with intraoperative measurements, MRI both quantitatively (P < .001) and categorically (P < .001) overestimates the size of glenoid labrum tears, albeit with large variability between cases (1.36 ± 3.09 hours). Although the magnitude of this difference is meaningful, the degree of variability observed across cases represents a central finding of this study. Specifically, the wide range in measurement differences suggests that MRI‐based estimates of tear size are inconsistent at the individual patient level, limiting their reliability for precise preoperative quantification. Despite this variability, MRI showed a high sensitivity for detecting the presence of labral pathology, identifying tears in over 90% of cases in this surgical cohort. Taken together, these findings suggest that MRI may be most appropriately used as a categorical tool for confirming the presence and general extent of labral injury, rather than as a precise quantitative measure of tear size. This distinction has important implications for surgical planning and patient counseling, where overestimation of tear size may influence both perceived severity and expectations of operative complexity.

The discrepancy between male and female measurements warrants further discussion. Men were more likely to have a larger recorded tear both on MRI (P < .001) and intraoperatively (P = .018) than women and showed a statistically significant difference between their MRI and intraoperative measurements (P < .001). Female patients did not display a statistically significant difference between their MRI and intraoperative measurements (P = .053). Although their MRI measurements tended to overestimate labrum tear size as well, the difference is smaller in women than in men (P = .044). The primary contributing factor for this discrepancy seems to be the sex difference in MRI measurement (1.46 hours), although intraoperative measurements contribute as well (0.55 hours). A recent multicenter orthopaedic outcomes network study found that male sex was associated with larger tears. ²² The multicenter orthopaedic outcomes network study's findings account for the discrepancy in labrum tear size at baseline but not the disparity in overall measurement difference. Osteoarthritis research on sexual differences has shown that women tend to have smaller cartilage measurements, ²³ specifically on MRI. ²⁴ This anatomical variation and imaging difference may act protectively to more accurately depict female patients’ glenoid labrum tears on MRI than their male counterparts.

Interestingly, the current study's finding of MRI's overestimation of labrum tear size is inconsistent with that of Campbell et al. who found that MRI underestimated meniscus tear size. ²¹ Important differences between the current study and Campbell et al. include that their study was performed in a smaller cohort, provided objective metric measurements, and evaluated differences between joint compartments as well. Although the current study opted to use clockface measurements to evaluate the size of labral pathology as is consistent with both radiological and orthopaedic practice, ²⁰ the studies provide comparable information regarding the relation between radiology and orthopaedic surgery. The current study's finding that MRI, on average, overestimates glenoid labrum tear size suggests that there is additional nuance to the use of MRI in imaging musculoskeletal injuries and that care must be taken to understand the unique relation between imaging and operative examination in each anatomical structure.

Beyond the important role that labral tear size plays in associated symptomology, preoperative planning, and postoperative course, diagnostic accuracy regarding tear size can play an important part in counseling patients with shoulder instability and associated labral tears. Although labral tear size has been shown to affect presenting symptomology, ⁵ preoperative planning, ⁹ , ¹⁰ , ¹¹ and postoperative course, ⁶ it is noticeably absent in the instability severity index score, the most prominent tool available to stratify risk in patients with shoulder instability. The instability severity index score ranges from 0 to 10, with a score over 6 generally marking an unacceptable risk of future dislocation. ²⁵ As such, patients with higher scores are more likely to be recommended surgery by their treating orthopaedic surgeon. Even in cases where patients fit into many or all categories that predispose them to further injury, the decision‐making process is centered around a risk‐benefit conversation. ²⁶ , ²⁷

In an era where test results are immediately available through the EMR, the reception of abnormal findings has been shown to cause increased worry in patients. ²⁸ It is likely that many patients with glenoid labrum tears have already seen their MRI reads before arriving in the clinic for a conversation about surgery. In a medicolegal system where informed consent ⁷ and shared decision making ⁸ are crucial to patient autonomy, a proper understanding of the capabilities and limits of diagnostic modalities and their relation to injury prognosis plays a meaningful role when providing patients with their options. Regardless of how the measurement influences a patient's decision to pursue surgical or conservative options, the patient is owed a proper understanding of their injury and its prognosis. ⁷ By giving patients a fuller picture of their injury status and prognosis, having more accurate data regarding MRI's accuracy to measure labral tear size will inform some patients’ shared decision‐making process. ⁸ , ²⁶ , ²⁷ Should a patient elect to pursue surgical options, a better understanding of MRI's capabilities to quantify glenoid labrum tear size may also lead to a more efficient allocation of time and resources to successfully repair the injury.

Although MRI read inter‐rater reliability varies depending on the location of the structure, ²⁹ higher inter‐rater reliability has been shown for arthroscopic evaluation of the glenoid labrum, ranging from 60% in the superior labrum to 90% in the anterior labrum. ³⁰ Using both historical measurements made by several different radiologists and orthopaedic surgeons and retrospective measurements made by single authors (A.S.C., C.C.), the study captured both the variability of different observers and the consistency of a single observer. As shown in Table 4, every available comparison reached the same conclusion that MRI overestimated the size of glenoid labrum tear. Finally, historical and retrospective measurements were not statistically different for either MRI (P = .123) or arthroscopy (P = .679).

This study's retrospective measurements were biased by the fact that these patients underwent surgery and therefore had labrum tears, but the authors hoped to limit this bias by separating chart review into radiological measurements performed by a radiologist and intraoperative measurements performed by an orthopaedic surgeon. Variations in field of view and matrix and software updates are an inherent limitation of the study's retrospective design, but the authors hoped to address this limitation by comparing MRI measurements by year, finding them to be not statistically different (P = .640), Additionally, similar to Campbell et al.'s findings, ²¹ the current study found that when directly comparing 1.5 and 3.0 T magnets, these groups did not differ in their overestimation of labrum tear size (P = .583). Lastly, arthrograms were excluded because of a small sample size (n = 6) compared with the rest of the cohort, and it would have been inappropriate to introduce this variability into the study.

Limitations

This study has limitations that warrant discussion. This study only included patients that underwent capsulorrhaphy for shoulder instability, hence only quantifying labral tear size in this patient population and setting. Similarly, knowing that these patients eventually underwent surgery was an inherent bias for the author quantifying MRI labral tear sizes and may have increased MRI's sensitivity to detect glenoid labrum tears in this population. For both MRI and intraoperative measurements, historical measurements from the EMR were used in the current study. There were nine radiologists responsible for measurements in MRI radiology notes and four sports surgeons responsible for measurements in operative notes, which lends this study to observer bias. Because of the retrospective nature of the study and the large sample size, intra‐ and inter‐reliability analyses were not performed. The magnets used for MRI at this institution varied between 1.2, 1.5, and 3.0 T magnets, and periodic software updates were installed that could not be tracked, contributing a level of variability to the study. The study was also limited to patients with measurements completed at the time of MRI and surgery or available MRIs and adequate intraoperative images. This group of patients is younger compared with the group studied by Davis et al. and older than the cohort studied by Cronin et al. (32.6 and 24.7, respectively) ²² , ³¹ and may not properly characterize a random sample of patients with this pathology. This study placed a 6‐month limitation on the time between MRI and surgery, which would limit the generalizability of these findings in the chronic setting.

CONCLUSIONS

MRI overestimated glenoid labrum tear size compared with intraoperative measurements and differed substantially between the two modalities on a case‐by‐case basis, with men displaying a larger discrepancy in measurement difference than women. These findings indicate that MRI may be more appropriately used as a categorical tool for confirming the presence of glenoid labral tears, rather than as a precise measure of tear size.

DISCLOSURES

The authors (S.K.G., A.S.C., W.A.C., C.C., B.E.J., R.J.M., S.A.M., S.E.B., B.M.G.) declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

ACKNOWLEDGMENTS

The authors thank Kimberly Hasselfeld, Isaac Hale, Logan Lake, and Hunter Cain for their administrative contributions to this project. The authors would also like to thank Lucas Huffman for his work in data curation.