|Home | About | Journals | Submit | Contact Us | Français|
Clinically, radiographic joint space narrowing (JSN) is regarded a surrogate of cartilage loss in osteoarthritis (OA). Using magnetic resonance imaging (MRI), we explored the magnitude and regional distribution of differences in cartilage thickness and subchondral bone area associated with specific Osteoarthritis Research Society International (OARSI) JSN grades.
Seventy-three participants with unilateral medial JSN were selected from the first half (2678 cases) of the OA Initiative cohort (45, 21, and 7 with OARSI JSN grades 1, 2, and 3, respectively, no medial JSN in the contra-lateral knee). Bilateral sagittal baseline DESSwe MRIs were segmented by experienced operators. Intra-person between-knee differences in cartilage thickness and subchondral bone areas were determined in medial femorotibial subregions.
Knees with medial OARSI JSN grades 1, 2, and 3 displayed a 190 μm (5.2%), 630 μm (18%), and 1560 μm (44%) smaller cartilage thickness in weight-bearing medial femorotibial compartments compared to knees without JSN, respectively. The weight-bearing femoral condyle displayed relatively greater differences than the posterior femoral condyle or the medial tibia (MT). The central subregion within the weight-bearing medial femur (cMF) of the femoral condyle (30–75°), and the external and central subregions within the tibia displayed relatively greater JSN-associated differences compared to other medial femorotibial subregions. Knees with higher JSN grades also displayed larger than contralateral femorotibial subchondral bone areas.
This study provides quantitative estimates of JSN-related cartilage loss, with the central part of the weight-bearing femoral condyle being most strongly affected. Knees with higher JSN grades displayed larger subchondral bone areas, suggesting that an increase in subchondral bone area occurs in advanced OA.
From a clinical and research perspective, radiographic joint space narrowing (JSN) is often used as a surrogate of disease status and progression in osteoarthritis (OA). It is assumed that increasing femorotibial JSN reflects cartilage loss in knee OA1-3. Yet, it is known that radiography has technical limitations4,5 and that meniscal pathology, particularly meniscal subluxation, is associated with JSN independently of structural changes in the articular cartilage6-8. While it is well known that higher JSN grades are associated with greater cartilage loss, the actual magnitudes of cartilage thickness differences associated with specific JSN grades, and whether these affect certain femorotibial subregions more markedly than others is not known.
A cross-sectional study using magnetic resonance imaging (MRI) in a sample of 372 participants9 reported that knees with medial JSN (mJSN; predominantly grade 1) displayed 11–13% smaller medial and lateral tibia cartilage volume than those without JSN, but displayed no difference in tibial subchondral bone size. However, since women have thinner cartilage than men10, the uneven distribution of gender among study groups in this sample may have confounded the results. A more recent cross-sectional analysis in women with Kellgren Lawrence (KL) grade 0 (healthy controls), grade 2 (osteophytes and no-JSN), and grade 3 (osteophytes and JSN) reported significantly thinner cartilage in the weight-bearing medial femoral, but not in the tibial cartilage of KL grade 3 compared with healthy participants, before and after adjusting for differences in body height11. The medial femorotibial subchondral bone areas were significantly larger in KL grades 2 and 3 than in healthy control participants. However, there were substantial differences in body mass index (BMI) between the OA and healthy participants, which may have confounded the results because of a positive correlation of cartilage thickness with BMI.
Generally, cross-sectional studies comparing different OA subcohorts suffer from large inter-subject variability of cartilage thickness and bone areas12,13, thus, making it difficult to interpret reported differences between groups of participants and rendering these relatively insensitive. Also, potential confounders related to cartilage thickness and subchondral bone area, such as sex, body height and weight10,14 are difficult to control adequately, further limiting interpretations. A superior approach for studying the association between radiographic JSN grade and cartilage thickness would be to follow participants longitudinally, before and after they have developed JSN, with each knee serving as its own control. Given the slow rate of progression of OA, however, this approach is challenging. An alternative strategy would be to compare OA participants with unilateral JSN such that one knee displays JSN (i.e., JSN knee) while the contra-lateral knee does not (i.e., no-JSN knee). This approach is more powerful than a comparison between participants, since intra-subject variability of knee cartilage thickness in healthy subjects measured by MRI is substantially less than the inter-subject variability15. The current study is the first one to take this particular (within-person, between-contra-lateral knee comparison) approach, with MRI being the ideal method for investigating the potential relationship of radiographic JSN with cartilage thickness and the size of the subchondral bone area, as it has been thoroughly validated for quantitative cartilage measurements in vivo16-20. Tibial bone area expansion (i.e., increases in the size of tAB) has been suggested to be a primary event in OA, to be associated with risk factors of OA progression and cartilage defects, and to represent an important target for the prevention cartilage loss and joint replacement21.
The objective of this exploratory and descriptive study was to estimate the magnitude of cartilage thickness differences and differences in subchondral bone area associated with each of the Osteoarthritis Research Society International (OARSI) JSN grades 1–32,3 by comparing MR images of JSN and no-JSN knees in participants with unilateral medial JSN. Subregional analysis approaches22,23 were used to characterize the regional pattern of JSN-related cartilage loss, i.e., to identify whether any anatomical subregions were more markedly affected than others.
The subsample studied was drawn from the first half (2678 cases) of the OA Initiative (OAI) cohort (public-use data sets 0.2.1 for the clinical data and 0.C.1 for the imaging data, www.oai.ucsf.edu). Participants in the OAI cohort are between 45 and 79 years old, stratified by gender and span across a diversity of ethnic backgrounds. Participants with rheumatoid or other inflammatory arthritis, bilateral end stage knee OA, inability to walk without aids, or 3 Tesla MRI contra-indications were excluded from the OAI cohort. The target population for the current study was a selection of participants fulfilling the following criteria: BMI > 25, pain on most days (or at least occasionally) in the past 12 months; OARSI grades 1, 2 or 32,3 in the medial compartment of one knee (mJSN) and no-mJSN (grade 0) in the contra-lateral knee; no lateral JSN or alternatively a lower grade of lateral JSN than mJSN in any knee. Longitudinal data in this sample were previously reported, with the participants displaying a similar distribution of sex and age as the entire OAI cohort24.
Adjudicated central X-ray readings from fixed flexion radiographs25,26 performed at Boston University on 160 cases from the public-use data set 0.1.127 were used to identify 19 participants who fulfilled the criteria above. An additional 54 cases were selected from the 2678 cases of the first half of the OAI cohort (public-use data sets 0.1.1, excluding the 19 previously selected) based on radiographic readings performed for recruitment purposes at the four clinical OAI sites; confirmation of the selection criteria for this study was performed by an experienced musculoskeletal radiologist (AG). In case of disagreement between the site and the study-specific readings (AG), readings were adjudicated by a rheumatologist (DH) and participants were only included if the inclusion criteria were met based on both the site readings and on the study-specific central readings (performed by AG and DH). Finally, a total of 73 participants (2.7% of the first half of the OAI cohort) met the inclusion criteria.
For the OAI, baseline MRI of both knees were acquired at four imaging sites with 3 Tesla MRI systems (Siemens Magnetom Trio, Erlangen, Germany), quadrature transmit–receive knee coils (USA Instruments, Aurora, OH), and a sagittal double echo at steady state sequence with water excitation (Fig. 1; DESSwe: 160 slices; 0.7 mm slice thickness, in-plane resolution 0.37 mm × 0.46 mm interpolated to 0.37 mm × 0.37 mm; 140 mm field of view; 512 × 512 matrix; repetition time 16.3 ms, echo time 4.7 ms, flip angle 25°)16,24. The sagittal DESS has been shown16 to produce cartilage thickness data consistent with that of previously validated spoiled gradient recalled (SPGR) sequences with water excitation or fat suppression19,20, and to display a similar tester–test precision (reproducibility)16,28,29 as these previously validated imaging protocols. Additional details on the MRI protocol and Quality Assurance (QA) procedures were recently published30,31. The OAI study protocol, amendments, and informed consent documentation were reviewed and approved by local institutional review boards.
Images were shipped from the OAI coordinating center to the image analysis center (Chondrometrics GmbH, Ainring, Germany) using hard drives and were quality controlled for segmentation purposes. Paired images of JSN and no-JSN knees were analyzed by seven readers, each with more than 3 years experience in cartilage segmentation and with previous experience in segmenting sagittal DESS images from the OAI pilot studies16,28,29. The readers were blinded to JSN status of the knees. Manual segmentation of the cartilage surface and of the bone cartilage interface of the medial tibia (MT) and entire medial femoral condyle (MF) was performed using specific segmentation software (Chondrometrics GmbH, Ainring, Germany). All segmentations were quality controlled by a single reader, (Dr. Susanne Maschek).
The mean cartilage thickness over the entire subchondral bone area (ThCtAB) and the total subchondral bone area (tAB) were computed after segmentation, including denuded areas of subchondral bone with zero millimeter cartilage thickness32. To define the appropriate femoral regions of interest and separate the MF from the femoral trochlea, the operators first marked a vector running parallel to the femoral shaft and through the most posterior point of the trochlear notch [Fig. 1(a)]16. In a next step, they marked the most posterior points of the medial (and lateral) femoral condyles [Fig. 1(b)]16; the MF was then separated into a weight-bearing medial femur (cMF) and a pMF by a plane parallel to the femoral shaft, cutting through the condyle at 75% of the distance between the trochlear notch and the posterior ends of both the medial and lateral femoral condyles24 [Figs. 1(b) and 2(a)]. Both parts of the femoral condyle were then divided into central, external, and internal subregions, each occupying 33.3% of the tAB [Fig. 2(a)]22. The MT was divided into five subregions: central (20% of the tAB), external, internal, anterior, and posterior [Fig. 2(b)]22. To obtain a single integral measure for the cMF of the medial femorotibial compartment (MFTC), the cartilage thickness of MT and cMF were added24,28,29. The same was done for the central MFTC (cMFTC) by adding cartilage thickness values from the central subregions of the MT and the medial weight-bearing femoral condyle, respectively.
To achieve a finer spatial resolution in the anterior–posterior (A–P) dimension of the femur, the condyle was divided into nine overlapping A–P subregions [Fig. 2(c)]. Each subregion comprised 30° of the A–P range, with a 15° posterior shift relative to the previous region [0–30°, 15–45°, 30–60°, etc. up to 120–150°; Fig. 2(c)]23.
To estimate the magnitude of cartilage loss for JSN grades 1–3, we computed the mean and standard deviation of the differences in ThCtAB (in μm) between the JSN and no-JSN knees for all subregions of interest. The mean difference in thickness was also expressed in percent by relating the average thickness estimate (in μm) in each JSN group (1–3) to the average estimate in the no-JSN group. Z-scores13 were derived by dividing the average difference in each JSN group by the standard deviation of the no-JSN group. To explore whether observed differences were statistically significant between the JSN and no-JSN knees, paired t tests were performed for ThCtAB and tAB values. Due to the relatively large number of tests for the subregions, false discovery rate (FDR) P-values were calculated in addition to raw P-values. These P-values address multiplicity issues by adjusting for the number of tests in a linear step-up procedure33. FDR P-values were calculated using SAS software version 9.1.3 PROC MULTTEST (SAS Institute, Cary, NC).
The subsample of 73 participants included 27 men and 46 women aged 60.8 ± 9.1 years [mean ± SD] with a BMI of 31.2 ± 4.1 kg/m2. All participants had mJSN grade 0 in the no-JSN knee; 45 had mJSN grade 1 in the contra-lateral knee (20 men, 25 women, age 60.7 ± 8.7 y.), 21 mJSN grade 2 (five men, 16 women, 60.1 ± 9.7 y.), and seven mJSN grade 3 (two men, five women, age 63.4 ± 10.4 y).
Cartilage thickness in the MFTC was 190 μm (−5.2%, Z-score = −0.35) thinner in the grade 1 knees, 630 μm (−18%, Z-score = −1.2) thinner in the grade 2 knees, and 1560 μm (−44%, Z-score = −3.0) thinner in the grade 3 knees compared with the contra-lateral grade 0 knees. The differences between mJSN vs contra-lateral no-mJSN knees for the central subregion of MFTC (cMFTC) were 460 μm (−11%, Z-score = −0.6) for grade 1, 1190 μm (−27%, Z-score = −1.6) for grade 2, and 2430 μm (−56%, Z-score = −3.2) for grade 3 knees, respectively.
The largest differences in the MT were observed in the external (Z-score up to −4.6) and central (Z-scores up to −2.1) subregions, and the smallest differences in the internal and anterior subregions (Table I). In the MF, the greatest difference between mJSN vs contralateral no-mJSN knees were observed in the central subregions of the weight-bearing femoral condyle (cMF: Z-score up to −3.7). Overall, observed cartilage thickness differences for mJSN 1 vs mJSN 0 knees were small. Differences in the cMF (Z-scores = −1.5 and −3.2 for mJSN 2 and 3 knees, respectively) were greater than differences in the MT (Z-scores = −0.6 and −2.0 for mJSN 2 and 3 knees) and greater than in the posterior femur (pMF: Z-scores = −0.3 and −1.1 for mJSN 2 and 3 knees) (Table II). When evaluating distinct A–P subregions in the MF, the greatest differences between mJSN and contra-lateral no-mJSN knees were observed in regions located between 30° and 75° at the femoral condyle (Table III, Fig. 3). When analyzing men and women separately in the mJSN grade 1 group, the observations made above did not differ principally between both sexes (data not shown).
The MT subchondral bone area (tAB) did not differ relevantly between grade 1 and grade 0 (−1.0%, Z-score −0.05) and between grade 2 and grade 0 knees (+1.4%, Z-score 0.07), but was considerably larger in grade 3 knees compared to grade 0 knees (+5.7%, Z-score +0.29). In the MF, grade 2 (+4.3%, Z-score +0.24) and grade 3 mJSN knees (+9.1%, Z-score +0.51) displayed considerably larger tABs than contra-lateral grade 0 knees, but not grade 1 mJSN knees (+0.6%, Z-score 0.03).
The objective of this study was to estimate the magnitude and regional distribution of differences in cartilage thickness and subchondral bone area, associated with each of the mJSN grades 1–32,3 observed clinically in radiographic OA. In order to overcome the challenges related to large inter-subject variability in these parameters and potential confounders between groups, this study is the first to employ a between-contra-lateral knee, within-person study design, comparing knees with mJSN and no-mJSN within OA participants with unilateral mJSN. While it is well known that higher JSN grades are associated with greater cartilage loss, the between-knee, intra-person approach is particularly well suited for quantitatively estimating the magnitude of thickness differences for particular JSN grades, as it eliminates between-person confounders, such as age, sex, weight, height, BMI and others. Applying this study design, we find that the intra-person, between-knee differences become greater with higher JSN grades and that the weight-bearing femoral cartilage (cMF) displays greater JSN-related differences than the posterior femoral condyle (pMF) and MT. This indicates that the pMF may represent a less relevant location for monitoring structural progression of knee OA than the cMF. Specific subregions (central cMF, external and central MT) showed the greatest cartilage thickness differences by JSN grade and were more strongly affected than other subregions in the MFTC, suggesting that these subregions may be preferentially affected by JSN-related cartilage loss in knee OA. Knees with higher JSN grades also displayed substantially larger femorotibial subchondral bone areas, suggesting that enlargement of the subchondral bone area may occur at the later stages of radiographic OA.
Between-knee differences in cartilage thickness and subchondral bone areas among health subjects were shown to be substantially smaller than the inter-subject variability of these parameters15. Therefore, quantitative estimates of within-person, between-contra-lateral knee comparisons of cartilage thickness and subchondral bone areas should be more reliable than those originating from between-person comparisons. Also, confounding from differences in age, sex, weight, height, BMI or others, inherent to between-person analyses, could be controlled more effectively using the within-person design than otherwise possible in comparisons between subcohorts. This does not exclude potential confounding from differences at the knee level (e.g., ligament and meniscus status, trauma history, etc.); given the interest to determine structural differences associated with radiographic JSN grades, however, the current analysis provides a more robust estimate of JSN-related differences in cartilage thickness and subchondral bone area than estimates derived from previous cross-sectional between-subject studies9,11.
Previous cross-sectional studies reported significantly lower MRI-based cartilage volume or thickness values in knees with JSN than those without9, or in those with KL grade 3 compared to KL grade 011. These estimates, however, were not specific to JSN grades and may have suffered from confounding of between-person comparisons. The current study shows that JSN grade 1 is associated with approximately 5% difference in weight-bearing femorotibial cartilage thickness, and that the magnitude of thickness difference increases with JSN grade 2 (about 20%) and grade 3 (about 40%). With regard to preferentially affected subregions in the cartilage plates, results from this study are consistent with a previous study in women with and without radiographic OA11.
The femorotibial subregions that were identified as being most strongly affected by JSN-related cartilage thickness differences in the current paper (ccMF, eMT, cMT) also agree with those that have been described to display the greatest longitudinal changes in femorotibial OA in a meta-analysis of three longitudinal studies34. With regard to the A–P (femoral) subregions, the results show that the posterior aspect of the weight-bearing femoral condyle (cMF: 30–75°), but not the posterior aspect of the condyle (pMF) is most strongly affected. This is in agreement with longitudinal changes in the MF23 and supports previous findings that JSN is more severe when knees are flexed to 20–30°35-39. However, it has to be kept in mind that the current results may be specific to the selection of knees based on JSN in fixed flexion radiographs, whereas use of another radiographic technique that applies greater flexion during imaging (i.e., the tunnel view) would have selected knees with somewhat more posterior JSN-related cartilage loss, and a radiographic technique applying less flexion (i.e., extended knee radiographs) selected knees with more anterior JSN-related cartilage loss.
The current paper focused on the relationship between clinically observed JSN grades and both the magnitude and spatial distribution of actual “quantitative” femorotibial cartilage loss. The relatively large standard deviation of the side differences observed highlights previous observations that JSN is not exclusively related to quantitative cartilage loss, but is also influenced by other factors, including technical limitations of radiography (i.e., variable alignment of the tibia with the X-ray beam)4,5, meniscus extrusion6-8, weight-bearing conditions with X-rays but non-weight-bearing conditions with MRI, and others. A recent paper addressed the specific relationship of minimum joint space width (mJSW in weight-bearing Lyon Schuss view) with (sub) regional femorotibial cartilage thickness and meniscal position (medial and posterior subluxation) in non-weight-bearing MRI40. The authors reported that, across OA and non-OA participants with different stages of radiographic disease, two thirds of the variation in mJSW was explained by regional femorotibial cartilage thickness measures, KL grade, and meniscal coverage40. The MT cartilage thickness measures and the central subregion of the weight-bearing medial femur (ccMF) played a consistent role in the variations in mJSW observed across all KL grades; ccMF cartilage and percent meniscal coverage best explained the differences in mJSW found between those subjects with definite JSN and those without40. These findings are consistent with our current observation that cMF (and particularly ccMF) display larger JSN-related differences (Z-scores) than MT and its subregions.
In the current paper, quantitative differences in cartilage thickness were computed between-contra-lateral knees to quantitatively estimate the cartilage thickness loss related to specific JSN grades. In this context, it must be kept in mind that radiographic JSN was measured from weight-bearing radiographs, while MRI was acquired in the supine, non-weight-bearing position. The quantitative MRI-based cartilage parameters might therefore possibly change when being also acquired under weight-bearing conditions. Although the spatial (in-plane) resolution of the sagittal DESSwe sequence was 310 μm × 460 μm, smaller side differences of mean cartilage thickness were reported, because these were averaged over several participants and many thousand thickness measurements in each knee. Given an average size of the tAB of 11.1 cm2 for MT and 17.2 cm2 for MF, the mean thickness value for each subregion (five in MT and nine in MF) relied on approximately 2000 measurements per person. Nevertheless, thickness differences between knees were only reported to the nearest 10 μm.
A limitation of this study is its relatively small sample size, especially for knees with OARSI mJSN grade 3. However, there are few OA subjects who fulfilled the inclusion criteria for this study (mJSN in one knee, no-mJSN in the contra-lateral knee, and no [or less than medial] lateral JSN in either knee), and the cases were selected from a large cohort (first half of the OAI; n = 2678). Cases exhibiting mJSN grade 0 in one and OARSI mJSN grade 3 in the contra-lateral knee are relatively rare (only 6.3% of those in the OAI cohort that exhibit definite radiographic OA in at least one knee), and subjects with such large side difference in mJSN may have a different OA patho-physiology, including post-traumatic OA, previous knee surgery or previous infection. This difference in OA patho-physiology may have potential implications on the spatial distribution of cartilage thickness differences observed, but the “spatial pattern” of relative side differences in cartilage thickness was similar for this group and that with mJSN grade 2.
Previous studies reported contradictory information regarding whether subchondral bone areas are larger in participants with radiographic JSN than in those without9,11 and were not specificto JSN grade. Longitudinal studies reported an increase in subchondral bone area with time, but the observations were made both in OA participants41,42 and in healthy subjects42,43. Therefore, previous studies were not conclusive on whether the increase in subchondral bone area observed longitudinally was specificto OA, or simply a function of the aging process. The results of the current study suggest that lower JSN grades (specifically grade 1) are not associated with increases in subchondral bone size, but higher JSN grades (grade 3 in the tibia, and grades 2 and 3 in the femur) are. These findings suggest that an increase in subchondral bone area occurs in advanced radiographic OA.
In conclusion, this study provides quantitative estimates of differences in cartilage thickness associated with specific medial radiographic JSN grades 1–3. JSN grade 1 was associated with small (Z-scores up to −0.6 only) between-knee differences in mJSN vs no-JSN knees in cartilage thickness and no differences in subchondral bone area. Higher JSN grades were found to be associated with larger cartilage thickness differences, and with substantially larger subchondral bone areas than contra-lateral knees without JSN, suggesting that enlargement of subchondral bone occurs in advanced OA. Within the MFTC, the weight-bearing femoral condyle (cMF) displayed relatively greater JSN-related differences than the MT and posterior femoral condyle (pMF). Specific subregions (central cMF, external and central MT) showed the greatest differences between JSN and no-JSN knees and were more strongly affected than other subregions in the MFTC.
We would like to thank Ana Vaz, MPH, Scientific Communications Associate at Eli Lilly & Company, IN, for her careful language editing of the manuscript. We would also like to thank the following readers: Gudrun Goldmann, Linda Jakobi, Manuela Kunz, Dr Susanne Maschek, Sabine Mühlsimer, Annette Thebis, and Dr Barbara Wehr for dedicated data segmentation.
Funding sources: The image analysis of this study was funded by Eli Lilly & Co, IN. The images were acquired by the OAI, a public–private partnership comprised five contracts (N01-AR-2-2258; N01-AR-2-2259; N01-AR-2-2260; N01-AR-2-2261; N01-AR-2-2262) funded by the National Institute of Health, a branch of the Department of Health and Human Services, and conducted by the OAI Study Investigators. Private funding partners of the OAI include Merck Research Laboratories; Novartis Pharmaceuticals Corporation, GlaxoSmithKline; and Pfizer, Inc. Private sector funding for the OAI is managed by the Foundation for the National Institute of Health. This manuscript has received the approval of the OAI Publications Committee based on a review of its scientific content and data interpretation.
Conflict of interest
Declaration of potentially competing interests: