OBJECTIVE. The objective of our study was to determine the accuracy of MRI in diagnosing meniscal tears in older patients and the frequency with which abnormal MRI findings correlate with degeneration, fraying, and tears at arthroscopy.
MATERIALS AND METHODS. Ninety-two patients 50 years old or older who had undergone knee MRI followed by knee arthroscopy within 6 months were selected. Menisci were graded on a 5-point scale: 1, definitely no tear; 2, probably no tear; 3, indeterminate or equivocal; 4, probably a tear; and 5, definitely a tear. Meniscal signal changes, meniscal surface morphology, and extrameniscal abnormalities were noted. Operative notes were reviewed, and the sensitivity and specificity of MRI results were calculated.
RESULTS. For medial tears, MRI had a sensitivity and specificity of 0.91 and 0.94, respectively, when grade 5 was considered a tear, 0.96 and 0.76 when grades 4 and 5 were considered a tear, and 0.99 and 0.47 when grades 3???5 were considered a tear. For lateral tears, MRI had a sensitivity and specificity of 0.73 and 0.91 when grade 5 was considered a tear, 0.88 and 0.80 when grades 4 and 5 were considered a tear, and 1.0 and 0.61 when grades 3???5 were considered a tear. The positive predictive values (PPVs) of MRI for the medial meniscus were 99%, grade 5; 57%, grade 4; 29%, grade 3; 25%, grade 2; and 0%, grade 1. For the lateral meniscus, the PPVs of MRI were 76%, grade 5; 36%, grade 4; 19%, grade 3; and 0%, grades 1 and 2.
CONCLUSION. The accuracy of MRI for diagnosing meniscal tears in older patients is high and similar to that in younger patients when only definitive findings are considered a tear. The specificity decreases if equivocal or probable findings are considered a tear.
MRI has become the primary noninvasive method for diagnosing meniscal tears of the knee . The MRI criteria for diagnosing this condition are well established; these criteria include deformity of the normal shape of the meniscus or increased intrameniscal signal contacting an articular surface of the meniscus [2, 3]. Stoller et al.  were the first to show that increased signal within the meniscus corresponds to degenerative changes and that signal contacting the articular surface is evidence of a tear. De Smet and Tuite  then found that the positive predictive value (PPV) of MRI for meniscal tears can be improved if a ???two-slice-touch??? rule is applied, in which a tear corresponds to increased signal touching the articular surface on at least two slices.
Most of the studies that have evaluated the accuracy of MRI for the detection of meniscal tears have been performed in young patient populations [2, 6???8]; the accuracy of MRI in diagnosing meniscal tears in older patients is less clear. The presence of degenerative changes in the menisci of older patients can make it especially challenging to diagnose a meniscal tear [9, 10]; in our experience, it is not rare to encounter equivocal cases in which it is unclear if MRI changes are the result of a tear or of age-related meniscal degeneration.
Meniscal tears of the knee are very common in older patients. Bhattacharyya et al.  found meniscal tears on MRI in 91% of patients at least 45 years old who had symptomatic knee osteoarthritis. However, to our knowledge, no study has been published reporting the accuracy of MRI in the detection of meniscal tears in this older population. Therefore in this study we sought to determine the sensitivity and specificity of MRI for diagnosing meniscal tears in patients 50 years old or older. We also sought to determine whether equivocal MRI lesions are correlated with tears and, in such equivocal cases, whether any specific MRI findings can help to distinguish tears from degenerative changes.
Materials and Methods
Our institutional review board approved this study and waived the requirement for informed consent. We searched the billing databases of the radiology and orthopedics departments for the records of patients who underwent knee arthroscopy between January 1, 2006, and December 31, 2006, preceded by knee MRI within 6 months. This query yielded 439 patients. From this group, we selected patients who were 50 years old or older, had not previously undergone meniscal surgery, and had intraoperative photos available in their records. A total of 92 patients met these criteria; these patients comprised the study group.
The patient population included 37 men and 55 women ranging in age from 50 to 79 years old (mean, 60.4 ?? 7.7 [SD] years); 46% of the patients were at least 60 years old. The shortest interval between MRI and arthroscopy was 1 day and the longest was 89 days (mean, 35.81 days).
Review of MRI Examinations
Scans were obtained at multiple sites on 1-T (50 patients) and 1.5-T (42 patients) systems. The standard imaging protocol (73 patients) consisted of double-echo sagittal non???fat-saturated spin-echo (SE) and fast spin-echo (FSE) proton density???weighted and T2-weighted sequences (TR range, 2100???4960 ms; TE range for echo 1, 14???20 ms; TE range for echo 2, 73???93 ms; slice thickness, 3???5 mm; echo-train length, 1???5; bandwidth, 100???150 Hz/pixel; matrix, 256 ?? 208???512 ?? 512; FOV, 12???17 cm) and a double-echo coronal fat-saturated FSE proton density???weighted sequence (TR range, 2000???5960 ms; TE range for echo 1, 13???16 ms; TE range for echo 2, 84???98 ms; slice thickness, 3???5 mm; echo-train length, 3???7; bandwidth, 115???150 Hz/pixel; matrix, 256 ?? 256???512 ?? 512; FOV, 13???17 cm); these scans were reviewed for this study. Additional axial fat-saturated FSE T2-weighted, coronal non???fat-saturated SE T1-weighted, and 3D gradient-echo sequences were also performed as part of the routine protocol, but the images were not reviewed for this study.
Two fellowship-trained musculoskeletal radiologists with 4 and 6 years of experience reviewed the MRI examinations in consensus. Both radiologists were blinded to the operative results. The medial and lateral menisci were individually assessed for the presence of a tear and were graded on a scale ranging from 1 to 5 (1 = definitely not a tear, 2 = probably not a tear, 3 = indeterminate or equivocal, 4 = probably a tear, 5 = definitely a tear). Grade 5 was defined as a clear alteration of the morphology of the meniscus, increased proton density???weighted signal clearly touching a meniscal surface on at least two slices, and no artifacts limiting evaluation. Grade 4 was defined as increased signal on proton density???weighted images clearly extending to an articular surface on only one image, minimal alteration of the meniscal morphology, and grade 5 lesions with artifacts limiting evaluation. Grade 3 was defined as increased signal approaching the articular surface but not clearly extending to the surface on proton density???weighted images; questionable alteration of the meniscal morphology; or any lesion not classifiable as grade 1, 2, 4, or 5. Grade 2 was defined as increased signal not extending to the articular surface on proton density???weighted images, normal morphology, or grade 1 lesions with artifacts limiting evaluation. Grade 1 was defined as no increased signal in the meniscus, normal morphology, and no artifacts limiting evaluation.
For menisci that were identified as grades 2, 3, or 4, specific MRI findings were also noted, including information about intrameniscal signal, surface morphology, and extrameniscal abnormalities. The intrameniscal signal was classified as amorphous or linear, and the presence or absence of increased signal that was less than fluid brightness on T2-weighted images, fluid-bright signal on T2-weighted images, signal approaching the surface but not definitely touching the surface, signal touching the surface on one slice, signal touching the surface on more than one slice, signal discretely touching the surface, and signal broadly touching the surface were noted. The surface morphology was assessed for the presence or absence of surface and free-edge irregularities. Assessment of extrameniscal findings included the presence or absence of a parameniscal cyst and marrow edema in the ipsilateral compartment. Additionally, the cartilage in the ipsilateral compartment was graded on a scale ranging from 1 to 4 (1 = normal, 2 = surface irregularity, 3 = partial-thickness loss, 4 = full-thickness loss).
Review of Operative Findings
Surgery was performed by one of 13 orthopedic surgeons. Each patient???s operative report and arthroscopic images were reviewed by a staff orthopedic surgeon with 5 years of experience and a postgraduate year 4 orthopedic surgery resident; both were blinded to the preoperative MR images and report. The presence or absence of a tear, fraying, and degeneration was noted for both menisci. Degeneration was defined as an irregular appearance of the meniscus such as a roughened surface or yellowish color without the presence of a tear. Fraying was limited to free-edge or inner margin fraying of the meniscus without the presence of a tear. For the purposes of data analysis, each meniscus was classified by the most severe abnormality that was present as follows in decreasing order of severity: tear, fraying, degeneration, or normal.
Consensus Review of MRI and Operative Findings
After this prospective review of MRI scans and operative findings, we performed a retrospective review, in consensus, of all the cases with discordant results between MRI and operative findings and of all cases with indeterminate or equivocal results on MRI.
The sensitivity and specificity of MRI were calculated as the proportion of patients with a positive test result among patients with a tear and the proportion of patients with a negative test result among patients without a tear, respectively. A positive test result was defined in three ways: First, only MRI grade 5 was considered positive; second, MRI grades 4 and 5 were considered positive; and, third, MRI grades 3, 4, and 5 were considered positive. The operative findings were considered the gold standard, with a tear considered a positive result and fraying, degeneration, and normal meniscus considered negative results.
Asymptotic methods were used to construct 95% CIs for sensitivity and specificity. For estimates of 1, a 95% CI was constructed using the rule of threes . Predictive values were calculated for each of the five possible grades. The Youden index was computed for the sensitivity and specificity of each MRI finding  and findings with the highest Youden index were included in a multiple-variable logistic regression model.
The sensitivity and specificity of MRI in the diagnosis of medial and lateral menisci are listed in Table 1. The PPV of each MRI grade for diagnosing a meniscal tear is summarized in Table 2. The operative findings and their corresponding MRI grades are summarized in Table 3.
For meniscal tears, when indeterminate (grade 3) was considered a negative result, MRI had a sensitivity of 0.96 (95% CI, 0.92???1.0) and a specificity of 0.76 (95% CI, 0.56???0.96). When grade 3 was considered a positive result, MRI had a sensitivity of 0.99 (95% CI, 0.97???1.0) and a specificity of 0.47 (95% CI, 0.23???0.71). For lateral tears, when indeterminate cases (grade 3) were considered a negative result, MRI had a sensitivity of 0.88 (95% CI, 0.76???1.0) and a specificity of 0.80 (95% CI, 0.70???0.90). When indeterminate cases (grade 3) were considered a positive result, MRI had a sensitivity of 1.0 (95% CI, 0.88???1.0) and a specificity of 0.61 (95% CI, 0.49???0.73). For the medial meniscus, 74 of 92 cases (80%) were interpreted as ???definitely a tear??? (grade 5) or ???definitely not a tear??? (grade 1) on MRI, and only seven cases (8%) were interpreted as grade 3 on MRI. For the lateral meniscus, only 55 of 92 cases (60%) were interpreted as grade 5 or grade 1 on MRI, and 16 cases (17%) were interpreted as grade 3 on MRI. Overall, 23 of 92 cases (25%) were interpreted as grade 3 on MRI.
For both the medial and lateral menisci, when the MRI interpretation was grade 1, this interpretation was correct in all cases based on the operative findings. When the MRI interpretation was ???probably not a tear??? (grade 2), this interpretation matched the operative findings in 75% of cases for the medial meniscus and in all cases for the lateral meniscus.
When the MRI interpretation was ???probably a tear??? (grade 4), this interpretation was correct based on the operative findings in 57% of cases for the medial meniscus and in 36% of cases for the lateral meniscus. When the MRI interpretation was grade 5, this result matched the operative findings in 99% of cases for the medial meniscus and in 76% of cases for the lateral meniscus.
For the medial meniscus, when the MRI interpretation was grade 3, operative findings revealed a tear in two of seven cases (29%); of the remaining five cases, four were normal and one showed degenerative changes at arthroscopy. The meniscal abnormalities observed on MRI were located in the body of the meniscus in five of the seven cases (Figs. 1A, 1B, 1C, and 1D) and in the posterior horn in two cases. The free edge (inner one third) of the meniscus was involved in only one case, and no cases involved the posterior root region.
For the lateral meniscus, when the MRI interpretation was grade 3, operative findings showed a tear in three of 16 cases (19%); of the remaining 13 cases, five were normal and eight showed fraying at arthroscopy. The meniscal abnormalities observed on MRI were located in the body of the meniscus in 10 of 16 cases, in the posterior horn in four cases, and in the anterior horn in two cases. The free edge of the meniscus was involved in 11 cases (Figs. 2A, 2B, and 2C), and the posterior root region was involved in three cases. Figure 3 summarizes the locations of the MRI grade 3 findings.
None of the specific MRI findings recorded relating to meniscal signal, surface morphology, or extrameniscal abnormalities had a high sensitivity or specificity in diagnosing meniscal tears. Because of the small sample size and selective nature of the cases available for multivariate analysis, the Youden index was highly variable and of limited utility.
Excluding the grade 3 cases, there were five discordant cases in the medial meniscus (four false-positives and one false-negative) and 13 discordant cases in the lateral meniscus (13 false-positives and 0 false-negatives). In the medial meniscus, all of the abnormalities were located in the posterior horn (Figs. 4A, 4B, 4C, and 4D). The free edge of the medial meniscus was involved in two cases and the posterior root region was involved in one case. In the lateral meniscus, the abnormalities were located in the body in eight of the 13 cases, in the posterior horn in three cases, and in the anterior horn in two cases. The free edge of the lateral meniscus was involved in 10 cases (Figs. 5A, 5B, 5C, and 5D), and the posterior root region was involved in one case. Retrospective review of all the discordant cases (n = 18) and equivocal cases (n = 23) with knowledge of the surgical results did not identify any findings that would have improved the accuracy of the MRI diagnosis.
In this study of older patients who had undergone knee MRI and arthroscopy, we found that the sensitivity of MRI for detecting meniscal tears was high, but the specificity varied. Additionally, equivocal MRI findings for a tear were more likely to represent degenerative changes or fraying than a tear at surgery in this population.
The use of MRI to establish or confirm a diagnosis of meniscal tear has become a routine part of clinical practice, and the accuracy of this imaging modality for diagnosing meniscal tears has been extensively reported in the literature . However, most of these studies have been performed in younger patients. Oei et al.  performed a meta-analysis of 29 published articles evaluating the performance of MRI in diagnosing meniscal and cruciate ligament tears. In these studies, which had a pooled weighted sensitivity and specificity of 93% and 88%, respectively, for medial meniscal tears and 79% and 96% for lateral meniscal tears, the mean age of the patient populations ranged from 30 to 40 years.
Few studies have assessed the use of MRI in older adults with meniscal tears. Jerosch et al.  found that asymptomatic patients older than 50 years had a high occurrence (25???40%) of severe degenerative changes or meniscal tears when assessed with standard MRI criteria. Similarly, Boden et al.  reported that 36% of asymptomatic patients older than 45 years had a meniscal tear according to standard MRI criteria. Even with 3-T imaging, von Engelhardt et al.  found that the sensitivity of MRI was poor especially for the lateral meniscus (57%) with a high number of false-negatives.
In our practice, we often encounter patients in this age group who have equivocal findings based on established MRI criteria for diagnosing a meniscal tear, such as a questionable articular surface or free-edge irregularity and increased intrameniscal signal that approaches the articular surface but does not definitely touch the surface. However, it is unclear whether these findings correspond to meniscal tears.
In this study of 92 patients 50 years old or older, we found the sensitivity and specificity of MRI for detecting meniscal tears similar to the sensitivities and specificities for younger populations reported by Oei et al. . MRI had a sensitivity and specificity of 91% and 94%, respectively, for the medial meniscus and 73% and 91% for the lateral meniscus when only cases with definite findings for a tear (grade 5 lesions) were considered tears. If probable findings (grade 4 lesions) were also considered tears, the sensitivity increased to 96% and 88% for the medial and lateral menisci, respectively, but the specificity dropped to 76% and 80%, respectively. De Smet and Tuite  reported similar results in a different patient population with sensitivities of 95% and 77% for the medial and lateral menisci, respectively, and specificities of 85% and 89% when a meniscal tear was defined as signal touching the surface on a single slice . When equivocal (grade 3) lesions were also considered tears, the sensitivity rose to 99% for the medial meniscus and 100% for the lateral meniscus but specificity dropped dramatically to 47% and 61%, respectively.
In this study, 25% of patients had equivocal MRI findings for a meniscal tear, with more cases observed in the lateral meniscus than in the medial meniscus. In most of these cases, a meniscal tear was not found during surgery, with a PPV of only 29% for the medial meniscus and 19% for lateral meniscus. These findings are similar to the findings in younger patient populations in which meniscal signal that does not clearly extend to the surface is unlikely to represent a tear [15, 16]. Furthermore, even a large number of cases that we interpreted as probably a tear on MRI were found in surgery to not be tears. In these equivocal and discordant cases, the most common location of the abnormality was the free edge of the lateral meniscus. No specific MRI findings that were prospectively evaluated had a high sensitivity and high specificity for a meniscal tear, and a retrospective review of the indeterminate and discordant cases with knowledge of the surgical results did not reveal additional useful MRI findings to help with a tear diagnosis.
The design of our study had several limitations. The small number of patients limited our ability to statistically correlate the specific MRI findings to the presence or absence of a tear. There was inherent referral bias because only patients referred for MRI were included in the study; additionally, there was inherent verification bias in that all patients who underwent surgery had undergone MRI and the MRI results likely influenced the decision to perform surgery. Context bias was also possible because the MRI readers were aware that all patients had undergone surgery, which may have increased the likelihood of the readers interpreting an abnormality as a tear. The potential for context bias was reduced by the fact that patients with positive findings for any one of several structures underwent surgery. When surgical findings were retrospectively reviewed, we occasionally found that the operative notes were incomplete or inconsistent in the descriptions of the meniscal abnormalities. This limitation was reduced by including only cases with operative photos available for review.
The classification of meniscal changes as degeneration, fraying, or a tear at arthroscopy in this patient population is subject to a wide range of interpretation: In some of the cases shown in this study that were classified as fraying at surgery and retrospectively using the operative photos (e.g., Figs. 2A, 2B, 2C, 5A, 5B, 5C, and 5D), other surgeons may interpret as tears. Our surgeons believe, however, that regardless of whether these types of lesions are termed ???degeneration,??? ???fraying,??? or ???tear??? at arthroscopy that they are unlikely to be a source of symptoms and do not need surgical treatment. In several of these indeterminate cases in this study, the meniscal ???degeneration??? or ???fraying??? was d??brided to a smooth margin at arthroscopy but these patients often had other indications for surgery such as a contralateral meniscal tear that was the primary reason for performing the surgery. The point of emphasis here is an indeterminate meniscal lesion alone does not necessitate surgery.
In conclusion, the accuracy of MRI for diagnosing meniscal tears in older patients is high and similar to that reported in studies of younger patients when a diagnosis of a tear is made only if there are definitive findings on MRI. The specificity decreases if a tear is diagnosed when there are only equivocal or probable findings on MRI because these findings usually do not represent a tear at surgery. In particular, abnormalities involving the free edge of the body of the lateral meniscus should not be overcalled as tears to prevent unnecessary surgical intervention.
We thank Michael P. Recht, of the Department of Radiology at New York University Langone Medical Center, for providing the idea behind the study and helping with the study design. We also acknowledge the assistance provided by Faysal F. Altahawi, of the Cleveland Clinic Lerner College of Medicine, in the preparation of the images in this manuscript.
This article is available for CME credit.
This is a Web exclusive article.
Rosas HG, De Smet AA. Magnetic resonance imaging of the meniscus. Top Magn Reson Imaging 2009; 20:151???173
Fischer SP, Fox JM, Del Pizzo W, Friedman MJ, Snyder SJ, Ferkel RD. Accuracy of diagnoses from magnetic resonance imaging of the knee: a multi-center analysis of one thousand and fourteen patients. J Bone Joint Surg Am 1991; 73:2???10
von Engelhardt LV, Schmitz A, Pennekamp PH, Schild HH, Wirtz DC, von Falkenhausen F. Diagnostics of degenerative meniscal tears at 3-Tesla MRI compared to arthroscopy as reference standard. Arch Orthop Trauma Surg 2008; 128:451???456
Boden SD, Davis DO, Dina TS, et al. A prospective and blinded investigation of magnetic resonance imaging of the knee: abnormal findings in asymptomatic subjects. Clin Orthop Relat Res 1992; 282:177???185
De Smet AA, Norris MA, Yandow DR, Quintana FA, Graf BK, Keene JS. MR diagnosis of meniscal tears of the knee: importance of high signal in the meniscus that extends to the surface. AJR 1993; 161:101???107