Prevalence of Bone Attrition on Knee Radiographs and MRI in a Community-based Cohort

Abstract

Objectives

Bone attrition probably constitutes remodeling of the bone, resulting in flattening or depression of the articular surfaces. Defining bone attrition is challenging because it is an accentuation of the normal curvature of the tibial plateaus. We aimed to define bone attrition on MRI of the knee using information from both radiographs and MRIs, and to assess whether bone attrition is common prior to end stage disease osteoarthritis (OA) in the tibio-femoral joint.

Methods

All knees of participants in the community-based sample of the Framingham OA Study were evaluated for bone attrition in radiographs and MRIs. Radiographs were scored based on templates designed to outline the normal contours of the tibio-femoral joint. MRIs were analyzed using the semi-quantitative whole-organ scoring method (WORMS). The prevalence of bone attrition was calculated using two different thresholds for MRI scores.

Results

Interobserver agreement for identification of bone attrition was substantial for the radiographs (κ=0.71 (95% CI 0.67–0.81) and moderate for MRI (κ=0.56, 95% CI 0.40–0.72). Of 964 knees, 5.7% of the radiographs showed bone attrition. Of these, 91% of MRIs were also read as showing bone attrition. We selected a conservative threshold for bone attrition on MRI scoring (≥2 on a 0–3 scale) based on agreement with attrition on the radiograph or when bone attrition on MRI co-occurred with cartilage loss on OA. Using this threshold for bone attrition on MRI, bone attrition was common in knees with OA. For example, in knees with mild OA but no joint space narrowing, 13 of 88 MRIs (14.8%) showed bone attrition.

Conclusions

Using MRI we found that many knees with mild OA without joint narrowing on radiographs had bone attrition, even using conservative definitions. The validity of our definition of bone attrition should be evaluated in further studies. Bone attrition may occur in milder OA and at earlier stages of disease than previously thought.

Introduction

Bone remodeling is an integral component of osteoarthritis (OA). Scintigraphic studies from the UK showed that “hot joints” could be seen by bone scan in advance of any radiographic changes of OA and, in fact, predicted subsequent radiographic changes of OA¹. Histological studies of bone in OA (far before “end-stage” disease is reached) show increased resorption and increased formation². In experimental models of OA, changes in subchondral bone occur very early after induction of the disease^{3 4}. These studies all suggest that bone remodeling occurs early in the osteoarthritic disease process.

Bone attrition is likely to represents remodelling of the envelope of subchondral bone in osteoarthritis which leads to a consequent change in bone shape or causes bone loss⁵. In surgical specimens of the knee, subchondral bone changes, including bone attrition, are common in persons with advanced OA⁶. Bone attrition is traditionally seen as an additional radiographic feature of advanced knee OA. Based on radiographs, bone attrition was evident in 62% of films from patients awaiting total knee replacement, compared to 1% of films from a control group⁷. However, bone attrition is not a feature that can be read easily because it may be difficult to detect in the absence of clear defects of cortical integrity and because of superimposed bone structures on radiographs.

Magnetic resonance imaging (MRI) in OA has broadened our understanding of OA. In MRI, the reading of bone attrition is usually performed using the Whole-Organ Magnetic Resonance Imaging Score (WORMS)⁸. Flattening or depression of the articular surfaces is graded based on the subjective degree of deviation from the normal contour on multiple tomographic images. As in radiographs, reading these deviations on MRIs is not straightforward. Whereas in convex surfaces like the femur, a deviation from the normal curvature is easy to detect, it is more difficult in the usually slightly concave surface of the medial tibia, especially if attrition is mild (see Figure1 A and B for different grades of attrition)⁹. These difficulties could lead to an under- or overestimation of bone attrition in MRIs.

Figure 1.

Figure 1A) Medial tibial bone attrition on MRI Grade 1

Sagittal and coronal fat-suppressed T2-weighted MR image of the knee. Attrition Grade 1 is visible on the medial tibial plateau (green dense arrow). Additional findings on the coronal plane: bone marrow edema*; and partial medial meniscal maceration (white thin arrow). The cartilage is thoroughly thinned at the central medial tibial and femoral plateau, best seen on the sagittal image.

Figure 1B) Medial tibial bone attrition on MRI Grade 2

Sagittal and coronal fat-suppressed T2-weighted MR image of the knee. Attrition Grade 2 is visible on the medial tibial plateau and femoral condyle (green dense arrows). Additional findings: bone marrow edema*; partial medial meniscal maceration (white thin arrow); marginal osteophytes at medial and lateral tibiofemoral joints, worse on medial. The central plates of tibiofemoral joint are denuded, and the cartilage is thoroughly irregularly thinned at the lateral tibiofemoral joint. Small subchondral bone cysts are best seen on the sagittal image, surrounded by bone marrow edema (black arrow).

If attrition occurs prior to advanced OA, this would suggest that changes in bone shape occur concurrently with cartilage loss and that treatments targeting cartilage loss alone are unlikely to work, even in knees with OA that are not at an advanced stage. Given the challenges in evaluating attrition, the characterization of attrition needs to be refined so that we can better appreciate bony alterations as part of the OA process. The aims of this study were 1) to find a reasonable definition for bone attrition on MRI of the knee using information from both the radiographs and MRIs 2) to assess whether, using this definition, bone attrition is common prior to end stage disease of OA in the tibio-femoral joint.

Methods

Subjects

The Framingham Osteoarthritis Study consists of 2 separate groups, as follows: members of the Framingham Heart Study Offspring cohort, and a newly recruited cohort from the town of Framingham, Massachusetts, the community sample. Since we obtained knee MRIs in the Heart Study Offspring only in persons with knee pain, thus not being a comprehensive sample, and since we wished in this study to examine attrition in a random sample of knees from the community, we did not use the Heart Study Offspring group for this study.

Participants of the community sample were drawn from a random sample of the Framingham, Massachusetts community. Flyers were hung in public areas to increase awareness of the study, which was focused on health, including bone health, foot health, and arthritis. Participants were recruited using random-digit dialing and census tract data to ensure a representative sample of the Framingham community. To be included, subjects had to be at least 50 years of age and ambulatory (use of assistive devices such as canes and walkers was allowed). Exclusion criteria were the presence of either bilateral total knee replacements or rheumatoid arthritis. A validated survey instrument¹⁰ supplemented by questions about medication use was used to exclude patients with rheumatoid arthritis. Participant selection was not based on the presence or absence of knee OA. Subjects were examined between 2002 and 2005.

Approval for the study was obtained from the Boston University Medical Center Institutional Review Board.

Assessment of MRI

All subjects in the community sample underwent bilateral knee MRI, regardless of whether they had symptoms. Participants in whom MRI was contraindicated did not undergo scanning, and in subjects with 1 total knee replacement, only the native knee was scanned.

All studies were performed with a 1.5T MRI system (Siemens, Mountain View, CA) using a phased array knee coil. A positioning device was used to ensure uniform placement of the knee among patients. Proton density-weighted fat-suppressed images in the sagittal, coronal and axial planes were acquired, using the following pulse sequence parameters: time to recovery 3,610 msec, time to echo 40 msec, slice thickness 3.5 mm, and field of view 14 cm.

MRIs were analyzed with the semi-quantitative whole-organ scoring method (WORMS) which incorporates 14 features⁸. All MRIs were read by two trained musculoskeletal radiologists, each of them assessing about 50% of the MRIs. .For the purposes of this paper, we focus on reading of articular cartilage integrity and subarticular bone attrition. WORMS score for cartilage is scored as follows: 0=normal thickness and signal; 1=normal thickness but increased signal on T2-or PD weighted images; 2=partial-thickness focal defect <1 cm in greatest width; 2.5=full-thickness focal defect < 1cm in greatest width; 3 multiple areas of partial-thickness (grade 2.0) defects intermixed with areas of normal thickness, or a grade 2.0 defect wider than 1 cm but <75% of the region; 4=diffuse (≥75% of the region) partial thickness loss; 5=multiple areas of full thickness loss (grade 2.5) or a grade 2.5 lesion wider than 1cm but <75% of the region; 6=diffuse (≥75% of the region) full-thickness loss⁸. Bone attrition was scored from 0 to 3 based on the subjective degree of deviation from the normal contour: 0=normal, 1=mild, 2=moderate and 3=severe (Figure 2A). Cartilage and bone attrition were evaluated in 14 regions of the knee: anterior, central and posterior segments of the medial and lateral femur and tibia, respectively, and the medial and lateral patella. Four of the 14 regions were related to the patello-femoral joint: the anterior femur (medial and lateral) and the medial and lateral patella. Because our focus was only on attrition in the tibio-femoral joint, we excluded the 4 patellofemoral regions from our analysis. In a random sample of 16 knees, read by two independent radiologists, inter-rater agreement for bone attrition was moderate with a weighted κ value of 0.56 (95% CI 0.40–0.72)¹¹.

Figure 2.

(A) Diagram showing bone attrition score by WORMS. Bone attrition is scored on the basis of the degree of flattening or depression of the articular surface relative to normal. Modified from Peterfy et al.⁸. Reprinted with permission from Elsevier.

(B) Template of the outline of the knee joint. Broken lines indicate the cut-off points for levels of bone loss. The templates were overlaid onto knee radiographs from patients with OA. This film shows severe grade 3 bone attrition on the lateral tibial plateau. Modified from Dieppe et al.⁷. Reprinted with permission of Wiley-Liss, Inc., a subsidiary of John Wiley & Sons, Inc.

Due to limited resources reading of the MRI was limited to the right side only. In case of a knee replacement on the right side, the left side was read.

Assessment of radiographs

In all subjects, plain posteroanterior (PA) weight-bearing radiographs of the knees (bilateral) were obtained using a protocol with fixed flexion¹². In addition to Kellgren/ Lawrence (K/L) grades¹³, radiographs were scored for attrition by one trained reader (SR), based on the original templates designed to outline the normal contours of the knee, which could be overlaid onto the knee radiographs to determine the presence of bone attrition (Figure2B)⁷. Alignment of the normal contours of the femur and tibia allows measurement of the extent of bone attrition separately for the femoral condyles and tibial plateaus. Based on Ahlbäck’s original suggestion¹⁴, we graded attrition on a scale of 0–3 as follows: 0=no attrition, 1=attrition of doubtful significance (<5 mm), 2=definite attrition of a moderate degree (5–10mm), 3=severe attrition (>10mm). The assessors of both MRI and radiographs were blinded to each other’s readings. In a random sample of 30 knees, read by two trained readers (SR, DF), inter-rater agreement for bone attrition was substantial with a weighted κ value of 0.71 (95% CI 0.67–0.81)¹¹.

Defining “true bone attrition” on MRI

Because of the potential overreading of the normal curvature of the tibial plateau as bone attrition, we felt that we could not assume that scoring of bone attrition on MRI represented OA-related change in bone shape. Therefore, we attempted to come up independently with a measure of “true bone attrition” using measures outside of bone shape on the MRI to validate this definition. Our two independent measures were bone attrition seen on the radiograph and clear-cut cartilage loss on the MRI. Of those knees scored as having attrition of some grade on MRI of the tibiofemoral joint, we characterized ‘true attrition’ on MRI if 1) it was noted also on the radiograph; or 2) it was present in a knee with cartilage loss on MRI, even if the radiographs did not show bone attrition. For the latter step, we took the following approach: we compared the presence of two different thresholds of bone attrition on MRI, mild and moderate attrition (WORMS score 1 and 2) with the presence of abnormal cartilage as read in the WORMS score for the tibio-femoral joint. Because MRI findings tend to be very sensitive, and we were interested in a level of cartilage loss that suggested unequivocal OA, we chose our cut-off for the WORMS score so as to represent cartilage loss extending to bone (score ≥5). A score ≥ 5 is the lowest score on the 0–6 scale that includes full thickness cartilage loss in an area. We intentionally chose a conservative cut-off so as to represent a knee with OA as it was not clear that milder degrees of cartilage loss could be defined as osteoarthritic. In a sensitivity analysis we compared different cartilage thresholds to different definitions of bone attrition.

Analysis plan

We calculated the prevalence of bone attrition for both radiographs and MRI by K/L grade for the tibio-femoral joint. We calculated the prevalence of bone attrition separately for each of the medial and lateral femoral condyles and the medial and lateral tibial plateaux. We determined the correlation between bone attrition on radiographs and MRI for all radiographs and MRIs read using Spearman’s rank correlation coefficient. We calculated sensitivity and specificity for the different MRI-thresholds. Based on the two definitions for ‘true attrition’ obtained through the processes above, we recalculated the prevalence of bone attrition and estimated the percentage of female, mean age and mean body mass index for those participants with and without ‘true attrition’.

All analyses were performed with SAS version 9.1 (SAS Institute Inc, Cary, North Carolina).

Results

MRIs were obtained in 995 subjects Thirty one subjects were excluded due to missing reading in all 10 subregions on MRI or due to missing radiographs. Bone attrition was therefore read for one knee per person in 964 knees both on MRI and corresponding radiograph. These subjects had a mean age of 63.3 years and had a mean body mass index of 28.5 kg/m². More than half (57%) were women, and, consistent with the community origin of the sample, most radiographs did not show evidence of OA (77.5% had a K/L =0).

Any bone attrition of the tibio-femoral joint, scored ≥ 1, was found in 228 MRIs (23.6%) and in 55 radiographs (5.7%). We found a moderate to strong correlation between MRIs and radiographs for bone attrition of the tibio-femoral joint (r=0.50, p < 0.001). The sensitivity of reading MRI with bone attrition score ≥ 1 was 0.90, the specificity 0.80; When choosing an MRI score ≥ 2, the sensitivity dropped to 0.70 and the specificity increased to 0.94.

Table 1 presents the distribution of bone attrition according to K/L grade. Bone attrition was seen both on MRIs and radiographs in 100% of films with K/L grade 4. Whereas a high degree of bone attrition (75%) was detected by MRIs in films showing K/L grade 3, this proportion detected by radiographs dropped to 34%. Bone attrition was not seen on radiographs of knees with K/L grade =0, but was read in 14.3% of the corresponding MRIs. In these knees with K/L=0, we found a disproportionately high amount of bone attrition in the medial tibial plateau (11.4%) compared to 3.2% in the medial femoral condyle.

Table 1.

Prevalence of bone attrition for radiographs (score ≥ 1) and MRI (score ≥ 1) in 964 knees of 964 subjects for the tibio-femoral joint and different compartments of the knee tabled by Kellgren/Lawrence score.

	K/L = 0 (n=747)		K/L = 1 (n=44)		K/L = 2 (n=88)		K/L = 3 (n=56)		K/L = 4 (n=29)
	Radiograph	MRI	Radiograph	MRI	Radiograph	MRI	Radiograph	MRI	Radiograph	MRI
Tibio-femoral joint (%)	0.0	14.3	0.0	18.2	8.0	47.7	33.9	75.0	100	100
Medial femoral condyle (%)	0.0	3.2	0.0	9.1	0.0	21.6	3.6	46.4	79.3	79.3
Medial tibial plateau (%)	0.0	11.4	0.0	15.9	8.0	35.2	30.4	55.4	82.8	79.3
Lateral femoral condyle (%)	0.0	0.8	0.0	2.3	0.0	2.3	0.0	19.6	20.7	55.2
Lateral tibial plateau (%)	0.0	2.0	0.0	0.0	0.0	9.1	5.4	19.6	20.7	48.3

Taking into account the difficulties distinguishing attrition from normal curvature of the bone of the bone especially in the medial tibial plateau, we adopted a more conservative threshold of attrition for MRI. Table 2 presents the comparison of different thresholds of bone attrition on MRIs with different levels of cartilage loss. Substantial cartilage loss (WORMS ≥ 5 in any region of the knee) was seen in 21% of knees with MRI bone attrition score=1, and in 73% of knees when the MRI bone attrition score was ≥ 2.

Table 2.

Prevalence of different scores of cartilage loss compared to different scores of bone attrition.

Score for Cartilage loss on WORMS scale	Bone attrition score on MRI
	Score=0 (n=736)		Score=1 (n=136)		Score≥2 (n=90)
	n	%	N	%	n	%
Score 0 + 1	402	54.6	34	25	0	0
Score 2 + 2.5	149	20.2	17	12.5	4	1.1
Score 3	155	21.1	42	30.9	10	11.1
Score 4	17	2.3	14	10.3	10	11.1
Score ≥ 5*	13	1.8	29	21.3	66	73.3

^*For score of 5 or greater, cartilage loss extends to bone in at least one part of the region being scored.

Given these results and the high prevalence of attrition read in the medial tibial plateau of otherwise normal radiographs, we concluded that it was more conservative to define ‘true attrition’ on MRI using a bone attrition score of ≥ 2. Using this new threshold, the amount of attrition in knees with severe OA (K/L grade 4) still was high (100%), but it was extremely uncommon (2.7%) in subjects with radiographs scored K/L 0 (table 3) and these knees with normal radiographs had evidence of focal severe cartilage loss (score for cartilage ≥5), suggesting that the radiographs in this case did not detect the presence of disease. We found ‘true’ bone attrition in nearly 15% of MRIs when radiographs showed osteophytes without definite space joint narrowing (K/L grade 2), and it was even present in 6.8% of MRIs with questionable signs of OA on the radiographs (K/L grade 1).

Table 3.

Prevalence of bone attrition for radiographs (attrition score ≥ 1) in 964 knees of 964 subjects for the tibio-femoral joint and different compartments of the knee by Kellgren and Lawrence score. For MRI, attrition was scored as present if the WORMS attrition score ≥ 2.

	K/L = 0 (n=747)		K/L = 1 (n=44)		K/L = 2 (n=88)		K/L = 3 (n=56)		K/L = 4 (n=29)
	Radiograph	MRI	Radiograph	MRI	Radiograph	MRI	Radiograph	MRI	Radiograph	MRI
Tibio-femoral joint (%)	0.0	2.7	0.0	6.8	8.0	14.8	33.9	44.6	100	100
Medial femoral condyle (%)	0.0	0.4	0.0	2.3	0.0	10.2	3.6	19.6	79.3	58.6
Medial tibial plateau (%)	0.0	1.7	0.0	4.6	8.0	6.8	30.4	25.0	82.8	75.9
Lateral femoral condyle (%)	0.0	0.3	0.0	0.0	0.0	0.0	0.0	3.6	20.7	27.6
Lateral tibial plateau (%)	0.0	0.8	0.0	0.0	0.0	1.1	5.4	8.9	20.7	34.5

Compared to persons without attrition, participants with ‘true bone attrition’ were older (67.6 years versus 62.9 years), had a slightly higher body mass index (29.8 kg/m² versus 28.3 kg/m²) and were more often male (51.1% versus 42.2%).

Discussion

Our results suggest that bone attrition is a common feature of osteoarthritis and that it occurs often at a mild stage of disease. Our findings contradict the current assumption, based on radiographs, that bone attrition is a feature only of advanced OA⁷. Because of overlapping structures, radiographs are much less sensitive to abnormalities than MRI. Bone attrition of the tibio-femoral joint in MRI was present in nearly 15% of subjects with no apparent loss of cartilage loss on radiographs (K/L=2), using a conservative approach for defining bone attrition on MRI. This difference in detecting bone attrition on radiograph versus MRI increases when evaluating radiographs with questionable signs of OA (K/L=1). As some investigators place K/L=1 in the OA group^{15 16}, the proportion of almost 7% bone attrition in this group might reflect true bone remodelling in what is considered to be a very mild stage of OA according to radiographs. Compared to radiographs, the sensitivity of MRI to detect bone attrition is the same in advanced OA, where attrition is found in 100% with both techniques.

We are not aware of prior attempts to evaluate the prevalence of bone attrition based on MRI. Our data are based on a large community-based cohort, which is important, as it provides a generalizable estimate of prevalence. Our findings demonstrated a good correlation of bone attrition between MRI and radiograph.

Reading bone attrition on MRI is not straightforward. This was underlined by our preliminary finding (table 1) in which bone attrition on MRI was found in almost 15% of otherwise normal radiographs. This might reflect an overreading. Indeed, when looking at the different knee compartments, most false positive readings originated in the medial tibial plateau, which is normally slightly concave. Using the more conservative approach of defining attrition, bone attrition was no longer prevalent in nonOA knees in the medial compartment.

This study has several limitations. There is no existing definition of OA based on MRI, and we were therefore not able to compare the prevalence of different bone attrition definitions with MRI defined OA. Instead, we compared bone attrition on MRI with cartilage loss on MRI. Because there is a lack of consensus as to which score of abnormal cartilage constitutes clinically relevant cartilage loss, we tried to define our cut-off conservatively to avoid false positive readings. In doing so, we limited the occurrence of ‘true attrition’ to advanced stages of cartilage loss based on MRI reading, although this is not necessarily reflected by a higher K/L grade.

However, we might also argue that in restricting the femoral condyle attrition score to ≥ 2, we missed subjects with ‘true’ attrition, as the femoral condyle normally has a convex curvature and any alteration in that curvature could be easily and correctly characterized as attrition. A conservative threshold at the femur could lead to an underestimation of the prevalence of attrition, and the true overall prevalence in radiographs without OA might lie somewhere between 4.2% and 10%.

The inter-rater agreement for bone attrition on MRI was only moderate with a weighted κ value of 0.56 (95% CI 0.40–0.72)¹¹. This is in line with the initial description of the WORMS score, where the lowest inter-reader agreement (ICC) was obtained for bone attrition (ICC = 0.61) compared to e.g. cartilage loss (ICC = 0.98) or osteophytes (ICC = 0.93))⁸. One of the reason for the imperfect agreement was the challenge in evaluating whether the curvature of the bone represented attrition and thus is further evidence that reading of bone attrition is not straight forward. The introduction of quantitative measurements of bone attrition would be welcome. Another reason for the imperfect agreement could be the use of different scales for bone attrition scoring on MRI and radiographs. Whereas both were scored from 0 to 3, the amount of bone attrition needed for grade 2 attrition on radiographs was higher than needed on MRI. Apart from reading errors, our results might also be affected by reader bias, as bone attrition was routinely assessed as part of the WORMS score, meaning that in the presence of other signs of OA like osteophytes or cartilage loss on MRI, bone attrition was more frequently read than on MRI that showed no features of OA. On the other hand, the assumption so far that bone attrition is a late stage feature of OA could also have led to an underestimation of the feature absent other signs of severe OA on MRI.

Our study results are limited to the tibio-femoral joint only. There is no validated reading of bone attrition in the femoral-patellar joint in radiographs, and we therefore limited our investigations to the tibio-femoral joint.

Our findings have implications for research and therapy. First, further work could lead to better clarification of thresholds for ‘true attrition’, ideally containing an independent reading of bone attrition and other features of OA. Second, cartilage loss is seen as a cardinal feature in development of OA, and the response of bone has, until now, been evident only later in the process. Our data confirm that involvement of subchondral bone might be earlier in the pathogenesis of OA. This parallels our knowledge of OA in animal models which have documented bone turnover occurring early in the disease process^{17 18}. Subchondral bone might be therefore very appealing for further research both related to mechanism of pain generation¹⁹ as well as for the development of new therapeutic strategies like antiresorptive agents²⁰. Further studies should be therefore undertaken to assess the relationship between bone attrition and bone mineral density. Third, as for therapy, the involvement of subchondral bone at an early point of disease might also help to explain the failure of chondroprotective agents as disease modifying drugs. Targeting cartilage alone when bone shape has already changed as part of OA may be ineffective^{21 22}.

In conclusion, our findings suggest that the prevalence of bone attrition in nonadvanced OA is substantial.