Radiographic Evaluation of Knee Osteoarthritis in Predicting Outcomes After Arthroscopic Partial Meniscectomy

Abstract

Background:

Knee osteoarthritis (OA) severity is a predictor of outcomes after arthroscopic partial meniscectomy (APM). Magnetic resonance imaging (MRI) grading of OA is predictive of postoperative outcomes; this prospective study assessed whether radiographic grading is also predictive of outcomes.

Methods:

Patients who underwent APM between February 2015 and January 2016, underwent radiography and MRI ≤6 months before surgery, and had outcomes from the surgery date and 1 year later were included. Surgical failure was defined as <10-point improvement in the Knee Osteoarthritis Outcome Score pain subscore. Radiographs were evaluated using Kellgren-Lawrence (KL) grading and continuous and ordinal minimum joint space width (mJSW) measurements; cartilage loss on MRI was evaluated using a modified Outerbridge system. Predictive abilities were estimated using area under the receiver operating characteristic curve (AUC) with 95% confidence intervals (CIs).

Results:

The study cohort included 66 knees from 64 patients (32 women; mean age, 57.1 years; range, 45-77). Radiographic grading was not predictive of outcomes (KL, AUC = 0.541 [95% CI: 0.358, 0.724]; continuous mJSW, AUC = 0.482 [95% CI: 0.305, 0.659]; ordinal mJSW, AUC = 0.534 [95% CI: 0.433, 0.634]). Comparison of radiographs showing no joint space narrowing (KL grade 0-2) with corresponding MR images demonstrated that 48% of radiographs missed a clinically significant lesion (modified Outerbridge grade ≥3). MRI grading was predictive of outcomes (AUC = 0.720 [95% CI: 0.581, 0.859]).

Conclusions:

Radiographic grading of OA is not predictive of outcomes after APM; radiographs may miss clinically significant lesions. For outcome prediction, MRI should be used.

1. Introduction

Arthroscopic partial meniscectomy (APM) to treat symptomatic meniscal tears is one of the most common orthopedic procedures in the United States, with more than 465,000 procedures performed annually [1]. However, as shown by the MeTeOR trial, when a meniscus tear occurs in the context of osteoarthritis (OA), using physical therapy as a first-line intervention with an option for surgery can be as effective as immediately treating the tear with APM [2]. In this population, the severity of knee OA is the main risk factor leading to poor clinical outcomes after APM [2-7]. It is therefore essential to identify the degree of knee OA before treatment decisions are made.

Arthroscopy is commonly referred to as the reference standard for grading the severity of OA [8-10], but multiple imaging modalities can also be used for this purpose. Magnetic resonance imaging (MRI) of the knee is considered the most accurate imaging modality for evaluating OA [11-13], and studies have shown that more extensive OA on MRI and arthroscopy is predictive of worse outcomes after APM [3,6,7,14]. In the clinical setting, radiography remains the first-line imaging modality for OA. Opinion is divided, however, regarding the optimal use of radiography to evaluate knee OA in patients undergoing APM [6,15], and few studies have addressed whether radiographic grading of OA is predictive of outcomes and which grading system should be used [1,2,8,14]. Of the multiple systems developed to grade OA on radiographs, the Kellgren-Lawrence (KL) system is the most commonly used. This system uses a severity score ranging from 0 to 4 depending on the presence and severity of osteophytes and joint space narrowing (JSN) [2,8,16-19]. Another grading system involves the measurement of minimum joint space width (mJSW); this system is easily applied clinically and has better inter-reader agreement than the KL system [16]. Although these grading systems are widely used, the ability of radiographs to predict outcomes after APM is still unclear.

The primary aim of this study was to evaluate whether OA severity assessed on radiographs using KL grading and mJSW measurements is predictive of 1-year patient-reported outcomes after APM. We also evaluated whether OA severity assessed using MRI is predictive of 1-year patient-reported outcomes after APM (secondary aim 1). Additionally, we compared radiographic OA evaluation methods to determine which demonstrated the best inter-reader agreement (secondary aim 2). Lastly, we analyzed KL grading of radiographs versus clinical MRI grading to identify the number of clinically significant OA lesions missed on radiography (secondary aim 3).

2. Materials and Methods

This study was approved by our institutional review board with a waiver of informed consent.

Cleveland Clinic’s OrthoMidas Episode of Care (OME) registry [20], which is an institutional review board–approved registry that records pre-, intra-, and postoperative data for all patients undergoing orthopedic hip, knee, or shoulder surgery at this institution, was used for this study. Patients enrolled in this registry complete validated surveys regarding patient-reported outcome measures such as the Knee Injury and Osteoarthritis Outcome Score (KOOS) [21] on their surgery date and at one year after surgery. The OME registry is also used by surgeons to record details about each case, and the registry contains demographic data for all enrolled patients. The demographic factors assessed for each patient in this study included age, sex, body mass index, education level, and smoking status.

The patient cohort for this study was selected from patients who underwent preoperative knee MRI and radiography within 6 months before APM at our institution between February 18, 2015 (when the OME registry was initiated), and January 1, 2016. This population of 418 patients was refined by excluding the 168 patients aged younger than 45, as these younger patients have a lower prevalence of OA. This age cutoff was selected to replicate the cohorts in large multicenter randomized controlled trials that have studied or are studying the treatment of patients with meniscal tears and osteoarthritis, such as the MeTeOR trial [2] and the TeMPO trial [22]. We also excluded 89 patients whose MRI examinations were performed at an outside hospital, as the results of these examinations were not available in our system. Of the remaining 161 patients, 27 (17%) were lost to follow-up, leading to a cohort of 134 from which a subset of 80 cases were selected for inclusion. A subset of 80 cases was used as a feasibility cohort because of the substantial time required for reading the radiographs and MR images. In addition, the use of this subset allowed us to weight the sample so that approximately 50% of the cases were patients who had experienced surgical failure, which was defined as failure to achieve the 10-point minimal clinically important difference in the KOOS pain subscore (KOOS_pain) at 1 year after surgery [21,23-25]. The sample was weighted with this increased prevalence of surgical failure to facilitate construction of the predictive model. Of the 80 cases selected for inclusion, 66 also had weight-bearing radiographs available in our system, and so our final cohort size was 66 cases (Figure 1).

Figure 1.

STROBE diagram for case selection/inclusion. Surgical success is defined as reaching the minimally clinically important difference for the KOOS_pain subscore at 1-year postoperatively, and surgical failure is defined as the failure to do so.

2.1. Radiographic and MRI examinations

For radiographic examinations, an anteroposterior (AP) view in a weight-bearing extended position and a posteroanterior (PA) view in a weight-bearing flexed (45 degrees) position (ie, a standardized version of the PA tunnel view) were obtained with a standard radiographic technique. The PA tunnel view, which is the most sensitive view for the evaluation of OA, was used for radiographic evaluation; when there was no available tunnel view, an AP view was used.

MRI examinations were performed on 3T and 1.5T MR systems with the standard clinical knee protocol, which consisted of coronal and sagittal fat-saturated fast spin echo (FSE) proton density–weighted sequences, a sagittal non-fat-saturated FSE intermediate-weighted sequence, a coronal non-fat-saturated FSE T1-weighted sequence, an axial fat-saturated FSE T2-weighted sequence, and a 3-dimensional fat-saturated FSE or gradient echo sequence.

All MRI scans and radiographs were evaluated on a research PACS system (TeraRecon, Foster City, CA, USA).

2.2. Image interpretation and scoring systems

Radiographic evaluation was performed using the KL and mJSW grading systems (Figure 2) [16,17]. With the KL system, grade 0 is defined as no JSN or osteophytes present; grade 1, only questionable osteophytes present; grade 2, definite presence of osteophytes with no JSN; grade 3, <50% JSN with or without osteophytes; and grade 4, >50% JSN with or without osteophytes [2,16,17]. The higher grade from either the medial or lateral compartment was used for modeling. The other grading system used was originally described by Gossec et al [16]. Using this system, the readers identified what they believed to be the narrowest point in the joint and measured the mJSW at that point. The smallest measurement from either the medial or lateral compartment was used for categorical placement. The categories were defined as follows: grade 0, ≥5 mm; grade 1, 3.5 to 4.9 mm; grade 2, 2 to 3.4 mm; and grade 3, <2 mm [16]. Both continuous measurements and ordinal grading were used to assess mJSW.

Figure 2.

Radiographic grading systems.

After being trained by a musculoskeletal radiologist, 1 radiology fellow and 1 medical student independently interpreted the radiographs using the KL and mJSW systems. A 2-week washout period was used between radiograph reading sessions with the different systems. To optimize the predictive accuracy of grading, the readers reached a consensus for any radiographic finding on which they had disagreed.

To evaluate the MR images, 2 musculoskeletal radiologists used the modified Outerbridge classification system to assess the depth of cartilage loss in 6 cartilage regions (medial and lateral femoral condyles, medial and lateral tibial plateaus, patella, and trochlea) [26-29]. The size of cartilage loss in each of the 6 regions was quantified as follows: normal (no loss), fissuring (<2 mm), small (2-9 mm), moderate (10-19 mm), large (20-29 mm), or diffuse (≥30 mm) [30]. For each region, the depth of the highest grade cartilage defect and the sum of the sizes of any low-grade (grade 1 + grade 2) or high-grade (grade 3 + grade 4) cartilage defects were recorded.

All readers were blinded to patient outcomes.

2.3. Statistical analysis

Both univariate and multiple-variable logistic regression models were used to incorporate demographic and imaging data. Surgical success (defined as an increase in the KOOS_pain subscale of ≥10 points 1 year after APM) was the dependent variable. Demographic, radiographic, and MRI data were assessed as independent variables.

Three radiographic models were constructed: one including the consensus KL grade, one including the consensus mJSW grade, and one including the average of the 2 readers’ measurements of mJSW. To evaluate the discrimination ability of the models, area under the receiver operating characteristic curve (AUC) estimates were calculated nonparametrically and 95% confidence intervals (CIs) were constructed. Wald tests were first used to test the null hypothesis that the AUC = 0.5 for the 3 radiographic models, with the significance level set at 0.05. The radiographic model with the highest accuracy for predicting outcome was then compared with a model based on MRI grade only and with a model based on demographic information only using a paired Wald test.

To evaluate the variability between readers for quantitative grading, the total deviation index at 80% was calculated. For qualitative grading, inter-reader variability was assessed using a kappa statistic.

3. Results

A total of 66 scans from 64 patients (32 women; mean age, 57.1 years; range, 45-77 years) were included in the analysis; 2 patients had bilateral scans (Table 1). A weight-bearing tunnel view of the tibiofemoral joint was available for 64 cases; for the 2 cases without a tunnel view, AP views were read (see Figure 3 for examples of tunnel views with different grades).

Table 1.

Patient demographics (N = 66).^a

Characteristic	Value
Mean age, years (SD) [range]	57.1 (8.4) [45-77]
Mean body mass index (SD) [range]	30.1 (6.4) [19-51]
Mean baseline KOOSpain (SD) [range]	49.7 (20.2) [2-94]
Sex, n
Women	32
Men	34
Race, n
Black	3
White	58
Other/Unknown	5
Smoking status, n
Current smoker	7
Never smoker	42
Quit	17
Education, n
Did not graduate from high school	4
Graduated from high school only	18
Received some college education	44
Patients receiving Medicaid, n	3

KOOS_pain, Knee Injury and Osteoarthritis Outcome Score pain subscore.

^aData were obtained from 64 patients; 2 patients had bilateral scans and were considered as 2 separate cases. Therefore, the representative number is 66.

Figure 3.

Posteroanterior tunnel views of the left knee in 2 different patients. (A) Kellgren-Lawrence (KL) grade 0. (B) KL grade 4.

3.1. Primary aim: prediction of outcomes with radiographic grading

In univariate analysis, baseline KOOS_pain was the only significant demographic predictor of outcomes at 1 year after surgery (P =.017), with an AUC of 1.04 [95% CI: 1.01, 1.07] (Table 2). However, the demographic-only model incorporating baseline KOOS_pain was not predictive of outcomes, with an AUC of 0.672 [95% CI: 0.494, 0.851], as the 95% CI included 0.5 (Table 3).

Table 2.

Univariate analysis of demographic factors as predictors of outcome at 1 year.

Possible predictor	Odds ratio [95% CI]	P value
Age	1.01 [0.95, 1.08]	0.711
Female sex	1.27 [0.42, 3.84]	0.670
Body mass index	1.03 [0.95, 1.12]	0.501
Education
Continuous		0.342
Categorical	0.90 [0.72, 1.12]	0.163
No high school vs grad school	4.80 [0.54, 42.6]
Graduated high school vs grad school	1.20 [0.28, 5.16]
Associates degree vs grad school	4.11 [0.96, 17.6]
Smoking status		0.452
Current smoker vs quit	3.50 [0.50, 24.6]
Never smoker vs quit	1.66 [0.40, 6.88]
KOOSpain at baseline	1.04 [1.01, 1.07]	0.017

KOOS_pain, Knee Injury and Osteoarthritis Outcome Score pain subscore.

Table 3.

AUCs of various models for predicting outcome at 1 year.

Model	Description	Estimated AUC [95% CI]
Demographics only	Baseline KOOSpain	0.672 [0.494, 0.851]
KL grading	Consensus of 2 readers’ grades	0.541 [0.358, 0.724]
JSW measurement	Mean of 2 readers’ measurements	Medial: 0.482 [0.305, 0.659] Lateral: 0.602 [0.436, 0.768] Patellofemoral: 0.496 [0.343, 0.650] Overalla: 0.561 [0.397, 0.726]
JSW grade	Consensus of 2 readers’ grades	Medial: 0.534 [0.433, 0.634] Lateral: 0.499 [0.344, 0.654] Patellofemoral: 0.549 [0.488, 0.611] Overallb: 0.534 [0.395, 0.673]
MRI model	No. of surfaces with grade ≥3 and size ≥10 mm	0.720 [0.581, 0.859]

AUC, area under the receiver operating characteristic curve; KOOS_pain, Knee Injury and Osteoarthritis Outcome Score pain subscore; KL, Kellgren-Lawrence; JSW, joint space width; MRI, magnetic resonance imaging.

^aLowest of medial and lateral JSW measurements.

^bHighest of medial and lateral JSW grades.

When radiographic findings were added to the demographic model in multiple-variable modeling analysis, none of the 3 radiographic models were predictive of outcomes, as the 95% CIs for their AUCs all included 0.5 (Table 3).

3.2. Secondary aim 1: prediction of outcomes with MRI grading

When a multiple-variable model was constructed by adding MRI findings to the demographic model, this model had an AUC of 0.720 [95% CI: 0.581, 0.859] (Table 3). This model was therefore predictive of outcomes.

3.3. Secondary aim 2: inter-reader agreement with KL and mJSW grading

Inter-reader agreement with KL grading was 31.8% (Table 4). In addition, 12.1% of cases received a grade of 0 from one reader and a grade of 3 or 4 from the other reader. The weighted kappa statistic for inter-reader agreement was 0.30 [95% CI: 0.16, 0.43], and the nonweighted statistic was 0.15 [95% CI: 0.02-0.27].

Table 4.

Inter-reader agreement for KL grading on radiographs.

Reader 2	Reader 1
	Grade 0	Grade 1	Grade 2	Grade 3	Grade 4
Grade 0	5	7	6	4	3
Grade 1	1	0	2	1	0
Grade 2	0	0	9	5	4
Grade 3	0	0	3	5	7
Grade 4	1	0	0	1	2

KL, Kellgren-Lawrence.

For quantitative mJSW measurements, the mean difference between readers was ≤0.85 mm for 80% of cases in the tibiofemoral medial compartment and ≤1.56 mm in the tibiofemoral lateral compartment. The maximum difference between readers was larger in the lateral compartment (6.99 mm) than in the medial compartment (3.68 mm).

For mJSW ordinal grading, agreement values for the medial and lateral compartments were 72.7% and 90.9%, respectively (Tables 5 and 6). There was only 1 case (in the lateral compartment) in which the readers were more than 1 category apart in their scoring.

Table 5.

Inter-reader agreement for mJSW grading in the lateral compartment.

Reader 2	Reader 1
	>5 mm	3.5-4.9 mm	2-3.4 mm	<2 mm
>5 mm	53	3	1	0
3.5-4.9 mm	2	5	0	0
2-3.4 mm	0	0	2	0
<2 mm	0	0	0	0

mJSW, minimum joint space width.

Table 6.

Inter-reader agreement for mJSW grading in the medial compartment.

Reader 2	Reader 1
	>5 mm	3.5-4.9 mm	2-3.4 mm	<2 mm
>5 mm	29	7	0	0
3.5-4.9 mm	5	14	3	0
2-3.4 mm	0	1	4	2
<2 mm	0	0	0	1

mJSW, minimum joint space width.

3.4. Secondary aim 3: detection of OA lesions with KL grading and MRI

When the consensus KL score for each knee was compared with the MRI cartilage surface score, differences were noted in the degree of OA identified (Table 7). Comparisons between consensus KL grading and MRI findings of tibiofemoral cartilage loss demonstrated that 22 of the 66 cases (33%) had at least 1 surface in the tibiofemoral joint with grade 3 or 4 articular cartilage defects visible on MRI but with no JSN (KL grade 0, 1, or 2) identified on radiographs (Table 8). Among the 46 cases categorized as KL grade 0, 1, or 2, 22 (48%) had at least 1 grade 3 or 4 defect on MRI.

Table 7.

KL scores versus MRI cartilage surface scores.

KL score	MRI cartilage thickness/depth score (tibiofemoral surfaces)a
	0	1	2	3	4
0	73	2	7	6	16
1	17	2	3	2	0
2	38	0	6	4	8
3	26	1	5	8	24
4	6	0	3	0	7

KL, Kellgren-Lawrence; MRI, magnetic resonance imaging.

^aTibiofemoral surfaces include the medial and lateral femoral and tibial surfaces.

Table 8.

Consensus KL grade versus the number of surfaces with MRI-identified clinically significant osteoarthritis lesion(s).

KL grade	No. of tibiofemoral surfaces with thickness/depth grade ≥3
	0	1	2	3	4
0	14	6	3	2	1
1	5	0	1	0	0
2	5	7	1	1	0
3	1	5	5	3	2
4	1	0	2	1	0

KL, Kellgren-Lawrence; MRI, magnetic resonance imaging.

4. Discussion

This study had several key findings. First and foremost, we found that evaluation of OA on preoperative radiographs, regardless of the grading system used, was not predictive of 1-year patient-reported outcomes after APM, which was the primary aim of the study. In contrast, we did find that evaluation of OA severity on MRI is predictive of outcomes, which was one of our secondary aims (secondary aim 1). Additionally, we found that inter-reader agreement with mJSW categorical grading of OA on radiographs, as described by Gossec et al [16], was significantly higher than inter-reader agreement with KL grading (secondary aim 2). Lastly, we found that nearly half of radiographs demonstrating no JSN (consensus KL grade 0-2) contained at least 1 clinically significant (modified Outerbridge grade ≥3) OA lesion on MRI (secondary aim 3). Taken together, these results suggest that MRI, not radiography, should be the imaging modality of choice for predicting outcomes after APM in this patient population.

OA of the knee is a common problem that causes pain and impairs function in more than 250 million people worldwide [31,32]. Many different modalities, such as radiography, MRI, and arthroscopic visualization, can be used to obtain information about the degree of OA present in the knee [2,10-13,16-19,31,33]; this information is particularly relevant for patients considering treatment with APM.

There is limited and conflicting evidence regarding the ability of radiographs to predict outcomes after APM. The MeTeOR trial reported no difference in functional outcome between patients with a meniscal tear and OA who were randomly assigned to physical therapy compared to those randomly assigned to APM based on preoperative KL grading [2]. A recent systematic review reported only two studies on APM that met the following inclusion criteria: discussed validated patient-reported outcomes, enrolled more than 20 patients, included at least 1 year of follow-up, had a small risk of a variety of biases, and excluded procedures other than APM (including meniscal repair and total meniscectomy) [34]. Both studies found that worse patient-reported outcomes (in terms of the Lysholm knee score [35]) were correlated with more severe OA on preoperative radiography [36,37]. One of the studies included only patients with mild OA (KL grade 0-2) and focused on cases of meniscal root tears [37]. This is important because the pathology of root tear is different from that of other meniscal tears; root tear results in the meniscus having minimal to no function, which is a much more severe result than other tear types. This previous study also did not use predictive modeling in its statistical approach. The other study included in the systematic review included patients with all grades of OA but also did not use a predictive modeling approach [36]. The authors also used Fairbank’s criteria [38] to evaluate radiographs rather than KL or mJSW grading.

In this study, we also assessed inter-reader agreement for both radiographic systems and found that ordinal and continuous mJSW grading had higher inter-reader agreement than KL grading, similar to results reported previously. For KL grading, Klara et al [39] reported kappas ranging from 0.66 to 0.97 for inter-reader agreement among multiple nonclinicians reading radiographs and 0.56 to 0.85 when comparing these nonclinicians to a radiologist. Gossec et al [16] reported a weighted kappa of 0.56 for inter-reader agreement. Our weighted kappa for KL grading was even lower at 0.30, perhaps because our study cohort included a lower prevalence of severe OA cases; more severe cases are easier to identify and grade. The lower prevalence of severe OA in our study was the natural result of including only patients who had undergone APM, as surgeons are reluctant to offer APM as a treatment option to patients with severe OA given the known relationship between severe OA and poor outcomes [3,6,7,14]. The higher inter-reader agreement observed with mJSW grading in our study (agreement values for the medial and lateral compartments of 72.7% and 90.9%, respectively) is also in concordance with results from previous research. Gossec et al [16] reported a weighted kappa for the mJSW categorical system of 0.86, and Gunther et al [40] reported interclass correlation coefficients of 0.62 and 0.47 for the medial and lateral compartments when grading JSN subjectively (without relying on any measurement tools). Ravaud et al [41] similarly found that a joint space measurement system was superior to the KL grading system.

The KL grading system is weaker than other measures not only in terms of inter-reader agreement, but also in terms of sensitivity. In a comparison of KL grading versus arthroscopy, the sensitivity of the KL system in identifying grade 3 or 4 lesions was only 6% in both tibiofemoral compartments of the knee [10]. Similarly, we found that 33% of total patients and 48% of patients with no JSN on KL grading had a grade 3 or 4 defect on MRI in at least 1 tibiofemoral articular cartilage surface [10]. Grade 3 and 4 lesions were defined as the clinically significant lesions in this study because when they are seen on arthroscopy, they are predictors of both patient-reported and functional outcomes [25,26,42-45].

Finally, we found that although radiographic grading of OA was not predictive of APM outcomes in this study, MRI grading of OA was predictive. One potential reason for MRI being more effective than radiography may be the increased sensitivity of MRI in detecting clinically significant cartilage lesions, as discussed above. Another reason may be that the patellofemoral compartment was evaluated on MRI but not on radiographs. Both of the radiographic OA grading systems used in this study, which are the two most widely used systems, evaluate only the tibiofemoral compartment [2,16,17]. This is a substantial limitation of these grading systems and may explain at least in part why radiographic grading of OA was not predictive. Additional studies using radiographic grading systems that include patellofemoral compartment evaluation are needed.

This study had several limitations, including its small sample size of 66 cases. To avoid overfitting, multivariable models constructed in this study included only a few of the many potential predictors of outcome after APM. The greater power resulting from a larger study would allow us to increase the accuracy of the multivariable models both with and without imaging findings. Another limitation is that the radiographic evaluations were not conducted by a fellowship-trained musculoskeletal radiologist. However, steps were taken to minimize and determine the effect, if any, of this limitation. The medical student and radiology fellow were trained by a fellowship-trained musculoskeletal radiologist, and inter-reader agreement was determined, steps similar to those taken in previous studies [39,46]. Although some small differences may have occurred between the readers, these differences are not likely to affect the primary outcome: the predictive ability of the model. Additionally, all APM procedures, although performed by different surgeons, were carried out at the same institution; a multicenter approach would improve generalizability. Future studies in younger patients may also be warranted to evaluate the generalizability of these results; although the cohort of this study was restricted to patients aged 45 years or older, the results may also be applicable to younger patients with concomitant OA. Another limitation is that the imaging findings were not compared against a reference standard. However, such a comparison was not necessary to accomplish the primary objective of our study, which was to measure the predictive value of the imaging findings, not to identify the accuracy of those findings. Lastly, although outcomes beyond 1 year were not assessed, most studies evaluating outcomes after APM have used a follow-up term of 1 year or less, and a recent study of patients in the MeTeOR trial demonstrated that KOOS_pain scores at 1 year after APM were maintained at 5 years after surgery [2,47-49].

5. Conclusion

In summary, radiographic grading of OA severity is not predictive of outcomes after APM, whereas MRI grading of OA is predictive of outcomes. While neither KL grading nor mJSW grading systems are predictive, mJSW grading has significantly better inter-reader agreement than KL grading. Lastly, radiographs miss many clinically significant cartilage lesions in this patient population.