MRI Diagnosis of Tears of the Hip Abductor Tendons (Gluteus Medius and Gluteus Minimus)

Transcription

1 Oliver Cvitanic 1 Gregory Henzie 1 Nicholas Skezas 2 Jack Lyons 2 Jon Minter 3 Received October 2, 2002; accepted after revision July 10, Southwest Oklahoma MRI, N Meridian Ave., Ste. 600A, Oklahoma City, OK Address correspondence to O. Cvitanic (ocvitanic@aol.com). 2 Department of Radiology, St. Joseph Hospital, 2900 N Lake Shore Dr., Chicago, IL Montreal Rd., Ste. 304, Atlanta, GA AJR 2004;182: X/04/ American Roentgen Ray Society MRI Diagnosis of Tears of the Hip Abductor Tendons (Gluteus Medius and Gluteus Minimus) OBJECTIVE. The purpose of our study was to determine the accuracy of MRI for diagnosing tears of the hip abductor tendons (gluteus medius and gluteus minimus) and to evaluate various signs of tendon disruption. MATERIALS AND METHODS. We retrospectively evaluated MRIs of 74 hips (in 45 patients) that were obtained using 35- to 42-cm fields of view and interpreted using primary and secondary signs of tendon disruption. Fifteen hips had surgically proven abductor tendon tears, and 59 hips were either asymptomatic or had surgically confirmed intact tendons. MRI findings were scored by two radiologists through consensus and then again independently by a third radiologist to determine interobserver agreement. RESULTS. The accuracy of MRI for the diagnosis of tears of the abductor tendons was 91%. Statistically significant associations were found between tears of the abductor tendons and areas of high signal intensity superior to the greater trochanter on T2-weighted images (p < ), tendon elongation in the gluteus medius (p = ), tendon discontinuity (p = 0.016), and areas of high signal intensity lateral to the greater trochanter on T2-weighted images (p = ). Interobserver agreement was good to fair. CONCLUSION. MRI showed good accuracy for the diagnosis of tears of the gluteus medius and gluteus minimus tendons. The identification of an area of T2 hyperintensity superior to the greater trochanter had the highest sensitivity and specificity for tears at 73% and 95%, respectively. R heumatologists and orthopedic surgeons are becoming increasingly aware that tears of the abductor tendons, as opposed to trochanteric bursitis, may be the leading cause of greater trochanteric pain syndrome [1]. Although several investigators have recommended MRI for diagnosing tears of the abductor tendons of the hip, the relative lack of surgical correlation and absence of data on the prevalence of MRI findings in asymptomatic subjects have been problematic [1 4]. We used 15 hips with surgically confirmed tears of the abductor tendons and 59 surgically intact or asymptomatic hips as case material for a blinded, controlled study evaluating the accuracy of MRI for diagnosing such tears. In the charted preoperative interpretations of the MRIs of the 15 hips with torn abductor tendons in this study, only two cases (13%) of abductor tears had been recognized despite the fact that abductor tendon tear was the diagnosis of exclusion. Considering the short lengths of the distal abductor tendons of the hip (gluteus medius and gluteus minimus), the use of large (35- to 42-cm) fields of view to allow imaging of both hips simultaneously could have reduced the sensitivity of MRI for the direct visualization of such tears (Fig. 1). Certain secondary signs of tears are less affected by the field of view. Specifically, focal areas of T2 hyperintensity superior or lateral to the greater trochanter, elongation of the gluteus medius tendon, and muscle atrophy are readily identifiable on fields of view of 42 cm. In this blinded, retrospective study, we evaluated MRIs obtained at communitybased imaging centers using conventional protocols with large fields of view for each of these secondary signs of tears as well as for tendon discontinuity. To our knowledge, this is the first study to evaluate the prevalence of MRI findings in a randomized mix of patients including those with surgically proven tears of the abductor tendons, intact tendons, and age-matched asymptomatic subjects. The purpose of this investigation was to determine the accuracy of MRI and various AJR:182, January

2 Cvitanic et al. signs for the diagnosis of tears of the hip abductor tendons using standard protocols for a screening assessment of the hips. Materials and Methods Patient Overview The MRIs of 74 hips in 45 patients were studied. Twenty-one consecutive patients had undergone unilateral hip surgery. Our study group comprised the 15 hips with surgically proven tears. The control group comprised the six hips with surgically confirmed intact tendons, the asymptomatic (not surgically treated) hips of five of the 21 patients who underwent hip surgery (16 hips were excluded: two prosthetic hips and 14 hips in which unilateral sagittal hip imaging could have allowed the interpreter to deduce which hip in the patient was symptomatic), and 48 hips in asymptomatic ambulatory volunteers, all of whom were women 55 years or older with no history of hip disease or surgery. Approval to perform MRI on the asymptomatic volunteers was granted by our institutional review board, and informed consent was obtained from all volunteer subjects. Study Group A surgically proven tear of an abductor tendon was the entrance criterion into the study group. The group comprised 15 hips in 15 patients (13 women, two men; mean age, 70 years; age range, years) with 10 focal full-thickness tendon tears, four diffuse full-thickness tendon tears, and one partial-thickness tendon tear. All the study group patients presented with lateral hip pain, usually accompanied by point tenderness over the greater trochanter, without a history of hip trauma. Generally, the clinical presentation of torn abductor tendon was indistinguishable from that of trochanteric bursitis. Nonsurgical treatment, including corticosteroid injections into the trochanteric bursa, was undertaken in all patients but failed to provide lasting relief of symptoms. All study group patients underwent MRI before surgery. The interval between the steroid injection and MRI examination ranged between 2 and 12 weeks (mean, 4.5 weeks). The mean interval between MRI and surgery was 6.3 weeks. The 15 patients in the study group underwent MRI of both hips simultaneously in the supine position. Various MRI systems were used. Ten of the 15 patients were studied using a 1.5-T field strength, and the remaining five were studied using a 0.2- to 0.3-T field strength. All studies included coronal and axial T1-weighted and either conventional T2- weighted (n = 4 patients), STIR (n = 5 patients), or fat-suppressed T2-weighted or proton density weighted with or without fat suppression (n = 6 patients) images obtained with 4- to 6-mm section thickness, 35- to 42.5-cm field of view, and a matrix. Sagittal images with similar parameters were also obtained unilaterally in 14 hips. Control Group The control group comprised 24 prospectively scanned asymptomatic ambulatory volunteers, all of whom were women (mean age, 65 years; age range, years) with no history of hip disorder, hip sur- Fig. 1. Illustration shows normal abductor muscles and tendons in coronal section as seen through hip at level of mid greater trochanter. a = gluteus minimus muscle, b = gluteus medius muscle, c = subgluteus minimus bursa, d = subgluteus medius bursa, e = subgluteus maximus bursa, f = zona orbicularis of hip joint capsule, g = superior neck recess of hip joint, h = inferior neck recess of hip joint, i = vastus lateralis muscle, k = iliotibial band. 138 AJR:182, January 2004

3 MRI of Hip Abductor Tendon Tears gery, or rheumatoid arthritis; six patients with surgically proven intact abductor tendons (five women and one man; mean age, 68 years; age range, years); and five patients in whom the contralateral hip had undergone surgery (five women; mean age, 70 years; age range, years). MRI of the 24 asymptomatic subjects in the control group was performed on both hips simultaneously using a 1.5-T MR system (Signa or Horizon, General Electric Medical Systems, Milwaukee, WI). Studies included coronal T1-weighted (TR/TE, 760/15) and STIR (3,650/32; inversion time, 160 msec) sequences and axial T1-weighted (760/15) and T2-weighted fast spin-echo (3,416/96; echo-train length, 4) sequences, all obtained with a section thickness of 5 mm, a field of view of 36 cm, a matrix, and 3 excitations. MRI parameters used to image the six hips with surgically confirmed intact abductor tendons and the five asymptomatic hips included in the control group were identical to those parameters used to image the study group patients. Surgical Evaluation The MRIs were available preoperatively for every patient who underwent surgery; however, the decision to perform surgery was based primarily on the clinical rather than the imaging findings. All 21 surgeries were performed specifically to evaluate the abductor tendons. Nineteen of 21 surgeries were performed by an orthopedic surgeon who specialized in the repair of torn abductor tendons in the hip. The remaining two surgeries were performed by orthopedist surgeons experienced in hip surgery and interested in the pathologic conditions of the abductor tendons. Surgeons used a binary end point (positive or negative findings for tear) rather than a grading scale to assess tendon injury. The surgical criterion for a tendon tear was a visible disruption of the tendon fibers, the ability to easily pass a blunt probe completely through the substance of the involved tendon, or a diffuse rupture and retraction of the tendon (the latter giving rise to the so-called bald trochanter [3]). Surgeons were either unable or did not attempt to separate the gluteus medius and gluteus minimus tendons that form a distal cuff [5] in any of the patients. At surgery, the presence or absence of trochanteric (subgluteus maximus) bursitis was recorded for every patient. The surgical criterion for trochanteric bursitis was a thickening of the bursal lining with or without leakage of fluid from the incised trochanteric bursa [6]. The subgluteus medius and subgluteus minimus bursae were not evaluated at surgery. MRI Evaluation The MRIs were examined for five major findings: tendon discontinuity, elongation of the tendon of the gluteus medius, atrophy of gluteus medius and gluteus minimus muscles, the presence of an area of high signal intensity superior to the greater trochanter on T2-weighted images, and the presence of an area of high signal intensity lateral to the greater trochanter on T2-weighted images. Tendon discontinuity was considered to be present if the distal tendon of the gluteus medius or gluteus minimus either was replaced by intermediate or high signal (granulation tissue, fluid, or synovitis) or was absent altogether. Tendon elongation in the gluteus medius was present if the apparent length of the tendon (i.e., the measured distance between the greater trochanter and musculotendinous junction) had increased to more than 2 cm on any single coronal slice and was longer than the tendon of the gluteus medius in the contralateral hip. We chose not to attempt to identify tendon elongation in the gluteus minimus because the typically slender, tapered appearance of a healthy tendon closely resembles an elongated tendon. We considered an abductor muscle to be atrophic if an interpreter subjectively judged its size to have decreased by at least 25% compared with the abductor muscle on the opposite side. High signal intensity, defined as hyperintensity relative to fat on T2-weighted images or isointensity relative to the contents of the urinary bladder on fat-suppressed T2-weighted images, indicated the presence of fluid, synovitis, or bursitis. Once recognized, an area of high signal intensity was placed in one of two categories on the basis of its relationship to the greater trochanter on coronal T2-weighted images. Specifically, any such area that was lateral to a vertical line drawn from the lateral margin of the greater trochanter was considered lateral in location, and any area found superior to a horizontal line drawn from the lateral margin of the greater trochanter and between vertical lines drawn from the medial and lateral margins of the greater trochanter was considered superior in location. In terms of anatomy, an area of T2 hyperintensity seen as lateral to the greater trochanter was considered to be in the main trochanteric bursa, whereas an area of T2 hyperintensity seen as superior to the greater trochanter was either in a tendon or between the tendons in the subgluteus medius or subgluteus minimus bursa. We agreed to adopt three findings, seen either independently or in combination, as diagnostic criteria for abductor tears: tendon discontinuity, an area of T2 hyperintensity superior to the greater trochanter, and tendon elongation in the gluteus medius. Findings that we evaluated for this study but did not use as diagnostic criteria for tears included muscle atrophy and T2 hyperintensity lateral to the greater trochanter. Areas of T2 hyperintensity seen lateral to the greater trochanter were interpreted as trochanteric bursitis. Although the interpreters were asked to evaluate the gluteus medius and gluteus minimus tendons separately, the cases in which either of these two tendons was torn were pooled for statistical purposes. Large and small tears were considered to be equivalent, also for purposes of analysis. Because tendinosis had not been confirmed surgically, no imaging finding specific for this diagnosis was recorded. Interpreters The aim of the study was not to test interpreter performance but to determine whether the MRI findings correlated with the occurrence of tears of the abductor tendons. Therefore, we had two wellqualified radiologists with extensive experience in musculoskeletal MRI interpret the images by consensus. Before assessing the images of the hips in the study, the interpreters were shown two examples (drawn from images of patients outside the study group) of each of the five major MRI findings being assessed and were provided with copies of two recent articles dealing with MRI of tears of the abductor tendons [2, 3]. The interpreters were then given the predetermined criteria for diagnosing tears in this study while remaining unaware of all patient data including name, age, sex, symptoms, which hip was symptomatic, and the surgical diagnosis. For each of the 74 hips, interpreters were asked to determine whether any of the five major criteria was present and then whether the abductor tendons were torn. Axial, coronal, and sagittal images were evaluated conjointly and scored using a standardized form. A third radiologist whose credentials were similar to those of the consensus interpreters and who was given the same preparation as they received independently interpreted the MRIs to determine agreement with the consensus interpretation [7]. Statistics Sensitivities for each of the MRI findings and for the diagnosis of a tear of the abductor tendon were calculated on the basis of the consensus interpretation. Associations among the various MRI findings and the surgical diagnosis were examined using the chi-square test. We regarded p values of less than 0.05 as statistically significant. Kappa statistics were used as a descriptive measure of agreement between the consensus and independent interpretations [8]. Interobserver agreement was considered excellent if the kappa value was between 0.80 and 1.0, good if between 0.60 and 0.79, moderate if between 0.40 and 0.59, fair if between 0.21 and 0.39, and slight if less than or equal to 0.20 [9]. Results Surgical Findings A review of the surgical reports indicated that six of the 10 small full-thickness tears found in the study group patients were lamellar, appearing as narrow slitlike defects in the tendons, whereas four were cystic, presenting as 0.5- to 1.0-cm-diameter holes in the tendon. These six tears were considered small. The remaining four full-thickness tears were more than 1.0 cm in the transverse diameter and considered large. Only one partial-thickness tear was identified. Ten tears originated in the anterior aspect of the hip abductor tendon cuff; the remaining five were central, AJR:182, January

4 Cvitanic et al. posterior, or diffuse. The surgeons did not specify which particular tendon was torn. No tear of a musculotendinous junction was reported. Surgeons found trochanteric bursitis in 10 of the 15 hips with a torn abductor tendon and in three of the six hips with intact tendons. Exposure and evaluation of the subgluteus medius and subgluteus minimus bursae were not part of the surgical protocol. MRI Findings The presence of high signal intensity superior to the greater trochanter on T2-weighted images was seen more frequently than any other major finding in hips with abductor tears and appeared in one of three patterns: pattern 1, a focal area larger than 1 cm in diameter (Fig. 2); pattern 2, a thin sheet that completely enveloped the lateral and superior aspects of the greater trochanter (Fig. 3); and pattern 3, a focal area 1 cm or smaller in diameter (Fig. 4). Pattern 1 was found exclusively in six hips with tears and had either smooth (pocketlike) or irregular margins. Pattern 2 was found in five hips with a tear and one hip without a tear. Pattern 3 was found in three hips without tears and none with a tear. STIR and fat-suppressed T2-weighted coronal images were the most sensitive pulse sequences for detection of areas of high signal intensity superior to the greater trochanter. However, because fat-suppressed sequences such as these tended to level the signal intensities of the tendons and intertendinous fat, it was sometimes difficult to determine whether a pocket of high signal situated superior to the greater trochanter was inside or outside the tendons. Sagittal images were generally useful only in verifying an area of T2 hyperintensity superior to the greater trochanter (Fig. 2B) but did not help in determining which tendon was involved. Axial images, on the other hand, were excellent for localizing involvement to individual abductor tendons (the gluteus minimus and gluteus medius tendons inserting onto the anterior and posterior aspects of the greater trochanter, respectively) as well as for confirming tendon discontinuity (Figs. 3B and 5). Thought to represent trochanteric bursitis, ar- A eas of T2 hyperintensity lateral relative to the greater trochanter were often found on coronal STIR images in hips in the control group, complicating the interpretation of these images. Tendon elongation in the gluteus medius was most easily recognized on coronal T1-weighted images and was sometimes accompanied by tendon thickening (Fig. 6). Atrophy of the abductor muscles was prevalent in both the study and control groups. In terms of the correlation between the size of the tear and MRI findings, atrophy was associated with large tears (p <.0001) but not with small tears. By contrast, T2 hyperintensity superior to the greater trochanter was more frequently found in small tears than in large tears, but no statistically significant correlation could be established. In general, the greater the number of major findings identified on MRI, the greater the probability of a tear, particularly a large tear. At surgery, 10 of 15 tears were found to be primarily anterior in location. On MRI, eight of 15 tears were thought to involve the gluteus minimus tendon (Fig. 5). Among the B Fig year-old woman with surgically proven focal full-thickness tear of abductor tendon cuff. A, Coronal STIR image (TR/TE, 2,000/30; inversion time, 160 msec) of left hip reveals 2-cm pocket of high signal intensity superior to greater trochanter (pattern 1) in subgluteus medius bursa (asterisk) located in intermuscular fat (f) between gluteus medius (me) and gluteus minimus (mi) tendons. Tendon tear itself is not directly visualized. B, Sagittal T2-weighted spin-echo image (1,750/102) of left hip shows pocket of high signal intensity superior to greater trochanter (gt) corresponding to swollen subgluteus medius bursa (asterisk). 140 AJR:182, January 2004

5 MRI of Hip Abductor Tendon Tears A B Fig year-old woman with surgically confirmed focal full-thickness tear in right abductor tendon cuff with trochanteric (subgluteus maximus) bursitis. A, Coronal T2-weighted fast spin-echo image (TR/TE, 3,879/102) of hips (obtained at 0.2 T with postprocessing) reveals gluteus medius (me) and gluteus minimus (mi) muscles, elongated and distally discontinuous right gluteus medius tendon (solid arrow), and sheet of high signal intensity outlining lateral and superior margins of greater trochanter (pattern 2) (open arrow). In general, visualization of pattern 2 T2 hyperintensity superior to greater trochanter alone is sufficient to diagnose tear. B, Axial T2-weighted fast spin-echo image (4,005/102) of hips shows area of high signal intensity corresponding to fluid or synovitis replacing distal right gluteus medius tendon (black arrow). Compare with intact appearance of left distal gluteus medius tendon (white arrow). five hips of the control group that showed false-positive findings for tears, three were thought to be centered in the gluteus minimus, whereas two were thought to involve primarily the gluteus medius tendon. Consensus Interpretations Results of the consensus MRI interpretations, including the frequency, sensitivity, specificity, and accuracy of each of the five major findings, are summarized in Table 1. The depiction of an area of T2 hyperintensity superior to the greater trochanter was strongly associated with a tear of an abductor tendon (p < ), but positive associations were also found with tendon elongation in the gluteus medius (p = ), tendon discontinuity (p = 0.016), and areas of T2 hyperintensity lateral relative to the greater trochanter (p = ). The frequency of muscle atrophy was higher in hips with torn abductor tendons than in the hips in the control group, but the difference was not statistically significant (p = ). Positive associations were found between areas of high signal intensity superior to the greater trochanter and tendon discontinuity (p = 0.023) as well as between tendon elongation and atrophy in the gluteus medius (p < ). The overall sensitivity of MRI in the diagnosis of tears of the abductor tendons was 93%, and the accuracy was 91%. Areas of T2 hyperintensity superior to the greater trochanter had a positive predictive value of 79% for tears of an abductor tendon and a negative predictive value of 95% for such tears. The results of the independent inter- AJR:182, January 2004 pretation reaffirmed these associations between findings and tears. A strong statistical association was found between trochanteric bursitis and tears of the abductor tendons (p < ). However, three (23%) of the 13 cases of surgically confirmed trochanteric bursitis did occur in hips in which no abductor tears were found. Areas of T2 hyperintensity were identified superior to the greater trochanter in two of the three hips, but both areas were smaller than 1 cm in diameter (pattern 3) (Fig. 4). degree of interobserver agreement for tears of the abductor tendons was good (κ = 0.72). In terms of the individual findings, interobserver agreement was good for areas of T2 hyperintensity superior to the greater trochanter (κ = 0.74), good for areas of T2 hyperintensity lateral to the greater trochanter (κ = 0.78), moderate for tendon elongation in the gluteus medius (κ = 0.50), moderate for abductor muscle atrophy (κ = 0.45), and fair for discontinuity of an abductor tendon (κ = 0.32). Interobserver Agreement Discussion When the consensus interpretations and independent interpretations were compared, the The current imaging literature on the abductor tendons of the hip emphasizes the appear- Fig. 4. Coronal STIR image (TR/TE, 1,400/30; inversion time, 110 msec) of hips in 66-year-old woman with left-sided greater trochanteric pain syndrome, surgically proven left-sided trochanteric bursitis, and surgically intact left abductor tendon cuff reveals 0.7-cm area of high signal intensity (arrow) immediately superior to left greater trochanter. Interpretation of this example of pattern 3 T2 hyperintensity superior to greater trochanter was false-positive for tear. 141

6 Cvitanic et al. Fig. 5. Axial fat-suppressed proton density weighted image (TR/TE, 4,100/38) of hips in 72- year-old woman with right-sided greater trochanteric pain syndrome reveals T2 hyperintensity corresponding to fluid or synovitis replacing distal right gluteus minimus tendon (open arrowhead). Compare with intact insertion of left gluteus minimus tendon (solid arrowhead). ance of large tears (diffuse avulsions) of the gluteus medius tendon and gives no figure or description of small (focal) tears or of those involving the gluteus minimus tendon [1 4]. Our study was based on the surgical findings in 21 consecutive patients with greater trochanteric pain syndrome. In these patients, small tears with mild local retraction outnumbered large tears with gross retraction, and the gluteus minimus seemed to be torn with a frequency similar to that of the gluteus medius. The imaging findings in this study support the predominance of small tears found at surgery. Specifically, we found that focal areas of T2 hyperintensity superior to the greater trochanter were 46% more common than tendon TABLE 1 Major MRI Findings in Tears of the Hip Abductor Tendons MRI Findings Consensus Interpretation Results Hips with Torn Abductor Tendon (n = 15) a Finding for atrophy of abductor muscles was the only one in which the difference between hips with tears and hips in the control group was not statistically significant (significance was set at p < 0.05). Fig. 6. Coronal T1-weighted spin-echo image (TR/TE, 560/20) of hips in 73-yearold woman with surgically confirmed tear in left abductor tendon cuff shows unilateral elongation and mild thickening of left gluteus medius tendon (arrow). No areas of hyperintensity were depicted in region of greater trochanter on T2- weighted images. Diagnostic Measures (%) Hips in the Control Group (n = 59) Sensitivity Specificity Accuracy No. % No. % High signal superior to greater trochanter Tendon elongation in gluteus medius muscle Tendon discontinuity High signal lateral to greater trochanter Atrophy of abductor muscles a MRI diagnosis of tendon tear discontinuity in hips with tears of the abductor tendons. This distinction is important because previous investigators have interpreted T2 hyperintensity in the distal abductor tendon cuff as tendinitis [1 4]. As described earlier, we divided the areas of T2 hyperintensity superior to the greater trochanter into one of three patterns depending on the size and configuration of the areas. In pattern 1, which corresponded to a tear in every case, the hyperintense area was larger than 1 cm in diameter and often resembled a pocket. Previous investigators have described similar areas of T2 hyperintensity superior to the greater trochanter as originating in the bursae [10]. Although our study lacks precise surgical identification of these bursae, we believe that the unexpected fluid leakage through the tendon cuff that occurred during the surgical exploration of several of the evaluated tears (similar to those described by Kagan [6] as cystic tears) might represent the inadvertent incision of swollen subgluteus medius and subgluteus minimus bursae. Thus, pattern 1 could indirectly represent a tear in the form of swelling in the subgluteus medius and subgluteus minimus bursae. However, pattern 1 hyperintensity could also more directly reflect the presence of a tear if fluid or synovitis occupies a gap in the tendon, even in cases in which the torn edges of the tendon are not directly seen on the large fields of view. This explanation accounts for the low frequency of tendon discontinuity observed in tears associated with patterns 1 and 2. Pattern 3, on the other hand, was not associated with tears. We speculate that the small areas of T2 hyperintensity corresponding to pattern 3 might reflect calcific tendinitis, but this explanation has not been validated surgically. The finding that we termed tendon elongation in the gluteus medius was present in slightly more than one half of all tendon tears and was the primary MRI finding for two large tendon tears. Unlike the term tendon retraction that is customarily used to describe measurable gaps in torn tendons, tendon elongation in the gluteus medius is a term intended to reflect the apparent lengthening of the tendon. This finding may represent nothing more than atrophy. However, we think that it is an effective marker for tears of the abductor tendons in the hip. We conclude that tendon elongation has value as a criterion 142 AJR:182, January 2004

7 MRI of Hip Abductor Tendon Tears for tendon disruption in the distal gluteus medius because of the ease with which it is recognized on T1-weighted images as well as its positive association with tears. Hip abductor muscle atrophy, defined as 25% gross muscle wasting seen on MRI, was found more frequently in hips with tears of the distal abductor tendons than in hips in the control group, but no statistically significant association was found. Atrophy of the gluteus medius and gluteus minimus muscles is relatively common in the age group at risk for tears and has also been shown to be related to disuse, peripheral neuropathy, and osteoarthritis [11]. Unlike patients with denervation atrophy of the gluteus medius muscle, which is a recognized complication of total hip arthroplasty, patients with tears of the abductor tendons may not have Trendelenburg s sign [1]. Despite the lack of statistically significant correlation between atrophy and tears of the abductor tendons, the presence of atrophy in the gluteus medius and gluteus minimus, particularly if atrophy is more pronounced in one hip than in the other or if atrophy disproportionately affects these muscles, should direct one s attention to the distal tendons for other signs of a tear. Areas of T2 hyperintensity lateral to the greater trochanter were commonly found in surgically confirmed tears of abductor tendons but also in 20% of the hips in the control group. Because the main trochanteric bursa is located in the region, we decided to interpret areas of T2 hyperintensity lateral to the greater trochanter as trochanteric bursitis. However, we had too few cases of surgically confirmed isolated trochanteric bursitis to validate any imaging criteria for diagnosing this entity. Although corticosteroid injections into the lateral hip were used to treat greater trochanteric pain syndrome in the study group patients, the lengthy interval between such injections and MRI examinations probably precluded the injections as a possible cause for the lateral areas of T2 hyperintensity. The correlation found between areas of lateral T2 hyperintensity and tears of the abductor tendons in this study most likely reflects the known association between trochanteric bursitis and abductor tendinopathy [1, 2, 5, 6]. We would not anticipate MR arthrograms to show evidence of contrast material extravasation through tears because the hip cuff tendons do not border the hip joint in the way the cuff tendons border the shoulder joint. The most significant limitation of our study was that three of the six hips in our control group that had surgical confirmation of tendon integrity received false-positive interpretations for tears on MRI using our criteria. On the other hand, all three were diagnosed as tears only on the basis of pattern 3 T2 hyperintensity being seen superior to the greater trochanter. The exclusion of these pattern 3 cases could improve the predictive value of T2 hyperintensity superior to the greater trochanter as a criterion for tears. Ideally, a prospective study in which the entire control and study populations comprised patients with greater trochanteric pain syndrome and surgically explored tendons could help confirm the sensitivity, specificity, and accuracy of MRI in diagnosing tears of the abductor tendons. In summary, MRIs of the hips obtained using large fields of view and interpreted using various secondary signs of tears showed good accuracy as a tool for diagnosing tears of the abductor tendons of the hip. The depiction of an area of T2 hyperintensity superior to the greater trochanter, which may directly or indirectly reflect a tendon tear, had the highest positive and negative predictive values for such tears. Those areas that fit pattern 1 or pattern 2 were the most specific for tears. The finding of tendon elongation in the gluteus medius was useful for identifying torn abductor tendons, especially when the contralateral hip was available for comparison. Areas of T2 hyperintensity lateral to the greater trochanter had a mild association with tears of the abductor tendons but may be more specific for trochanteric (subgluteus maximus) bursitis. Atrophy of the abductor muscles was too prevalent in elderly persons to be of much assistance in identifying tears. Using T2 hyperintensity superior to the greater trochanter as a criterion for tendinitis rather than for tears may be an important source of error in the preoperative assessment of the abductor tendons of the hips. With the tear criteria specified earlier, we found that MRI is sensitive for identification of tears of the hip abductor tendons in patients with intractable lateral hip pain. Conversely, the strong negative predictive value of MRI, particularly when STIR or fatsuppressed T2-weighted coronal sequences are used, can help obviate surgery in patients withgreater trochanteric pain syndrome who have no abductor tear. Acknowledgments We thank Abbott Kagan for providing case material and assistance that made this investigation possible and Willis Owen for reviewing our statistical analysis. References 1. Bird PA, Oakley SP, Shnier R, Kirkham BW. Prospective evaluation of MRI and physical examination findings in patients with greater trochanteric pain syndrome. Arthritis Rheum 2001;44: Chung CD, Robertson JE, Cho GJ, et al. Gluteus medius tendon tears and avulsive injuries in elderly women: imaging findings in six patients. AJR 1999;173: Kingzett-Taylor A, Tirman PF, Fell J, et al. Tendinosis and tears of gluteus medius and minimus muscles as a cause of hip pain: MRI findings. AJR 1999;173: Lonner JH, Van Kleunen JP. Spontaneous rupture of the gluteus medius and minimus tendon. Am J Orthop 2002;31: Bunker TD, Esler CN, Leach WJ. Rotator-cuff tear of the hip. J Bone Joint Surg Br 1997;79: Kagan A 2nd. Rotator cuff tears of the hips. Clin Orthop 1999;368: Farooki S, Yao L, Seeger LL. Anterolateral impingement of the ankle: effectiveness of MRI. Radiology 1998;207: Siegel S, Castellan NJ. Measures of association and their tests of significance. In: Siegel S, Castellan NJ, ed. Nonparametric statistics for the behavioral sciences, 2nd ed. New York, NY: McGraw-Hill, 1988: Landis JR, Koch GG. The measurement of observer agreement for categorical data. Biometrics 1977;33: Wybier M, Gires F, Chevrot A, et al. Apport de l imagerie medicale au diagnostic et au traitement de la bursite trochanterienne. Joint Bone Spine 1988;55: Howell GE, Biggs RE, Bourne RB. Prevalence of abductor mechanism tears of the hips in patients with osteoarthritis. J Arthroplasty 2001;16: AJR:182, January