|Home | About | Journals | Submit | Contact Us | Français|
The Osteoarthritis Initiative (OAI) is aimed at validating (imaging) biomarkers for monitoring progression of knee OA. Here we analyze regional femorotibial cartilage thickness changes over one year using 3 Tesla MRI. Specifically, we tested whether changes in central subregions exceed those in the total cartilage plates.
The right knees of a subsample of the OAI progression subcohort (n=156 age 60.9±9.9 y) were studied. 54 participants had definite radiographic OA (KLG 2 or 3) and a BMI>30. Mean and minimal cartilage thickness was determined in subregions of the medial/lateral tibia (MT/LT), and of the medial/lateral weight-bearing femoral condyle (cMF/cLF), after paired (baseline, follow-up) segmentation of coronal FLASHwe images with blinding to the order of acquisition.
The central aspect of cMF displayed a 5.8%/2.8% change in mean thickness in the group of 54/156 participants, respectively, with a standardized response mean (SRM) of -0.47/-0.31, whereas cartilage loss in the total cMF was 4.1%/1.9% (SRM -0.49/-0.30). In the central MT, the rate of change was -1.6% /-0.9% and the SRM -0.29/-0.20, whereas for the entire MT the rate was -1.0%/-0.5% and the SRM -0.21/-0.12. Minimal thickness displayed greater rates of change, but lower SRMs than mean thickness.
This study shows that the rate of cartilage loss is greater in central subregions than in entire femorotibial cartilage plates. The sensitivity to change in central subregions was higher than for the total cartilage plate in the medial tibia and was similar to the total plate in the medial weight-bearing femur.
Magnetic resonance imaging (MRI) provides the opportunity to measure articular cartilage loss and other structural changes in knee osteoarthritis (OA) directly and three-dimensionally, whereas radiography is limited to the analysis of projections of the bone contours (i.e. joint space narrowing). MRI not only circumvents the challenges in appropriate positioning of the knee versus a two-dimensional film, but also permits one to acquire more comprehensive information on cartilage loss by providing specific information on each of the contacting cartilage plates in the femorotibial (=FT) joint. Moreover, cartilage loss can be measured in subregions (i.e. central, internal, external) of FT cartilage plates1–3, so one can gain additional insight into the spatial distribution of tissue loss throughout the cartilage plates in OA. Assuming that cartilage is not lost homogeneously throughout the plates, this approach may be used to identify subregions with higher rate and sensitivity to change, which may, in turn, permit reductions in sample size in clinical studies for demonstrating, for instance, structure modifying effects of pharmaceutical compounds on disease progression.
We have recently presented a technique by which tibial cartilage can be reliably divided into central, internal, external, anterior, and posterior subregions, and the weight-bearing part of the femoral condyles into central, internal and external subregions, based on segmentation of the total subchondral bone area3. In a recent study by another group2, the highest rates of change (% cartilage loss) were found in the central regions of the FT cartilages at 1.5 Tesla (T). However, these subregions appeared to have no advantage in terms of sensitivity for detecting change due to a proportional increase in the variability of change.
Division of cartilage plates into subregions has the additional advantage that the minimal cartilage thickness can be measured in central areas, whereas for total cartilage plates the minimal cartilage thickness always drops off towards zero towards the joint margins. Analogous to radiography, where minimal joint space width is considered a gold standard for measuring disease progression4, regionalization has the advantage that the minimal cartilage thickness can be monitored three-dimensionally in the central joint areas. In a diseased joint, the minimal cartilage thickness may potentially be more sensitive to change than the mean thickness, because in central areas it is likely located at the site of a lesion, where cartilage loss may occur faster than in other parts of the joint surface.
Here we investigated whether subregional analysis of femorotibial cartilage with 3T MRI, testing central subregions of different sizes3, as well as mean and minimal cartilage thickness3, specific areas/parameters can be identified that provide a higher sensitivity to cartilage loss over time than the analysis of entire femorotibial cartilage plates 5. Specifically, we tested the hypothesis that changes in central subregions of the femorotibial cartilages exceed those in total cartilage plates.
An age and gender stratified subsample of the OAI progression subcohort was studied (OAI public-use datasets 0.1.1, 0.B.1 and 1.B.1) 5 for which baseline and one year follow up MRI data was available. The subsample included 79 women with a mean (±standard deviation) age of 60.3±9.5 years (BMI= 30.3±5.5) and 77 men, aged 62.0±10.2 years (BMI= 30.1±3.7). Participants were aged 45 to 79 years; 79 participants displayed a BMI of >30. The participants all had frequent knee symptoms (pain, aching or stiffness on most days of a month in past year) and radiographic OA (definite osteophytes in the postero-anterior [pa] fixed flexion radiographs 6,7) in at least one knee in the initial readings of the baseline radiographs at the imaging sites. The baseline radiographs were acquired at the same time as the baseline MRIs. In a separate assessment, the baseline radiographs were then read independently by two readers (one musculoskeletal radiologist and one rheumatologist) at Boston University for Kellgren-Lawrence (K–L) grade. When there was a discrepancy in OA status by K–L grade (0–1 versus ≥ 2), readings were adjudicated by consensus with a third reader present. The results of these adjudicated reading were used in the present analysis. No data on knee alignment was available for this cohort.
The MRI sequence used to quantify cartilage morphology (see below) was only available in the right knees, whereas some participants displayed symptomatic and radiographic OA in their left knee. Further the adjudicated central radiographic readings differed slightly from the initial screening readings at the sites. Therefore, not all knees analyzed displayed radiographic knee OA. Of the 156 knees analyzed, 17 showed K–L grade=0, 29 grade 1, 56 grade 2, 47 grade 3 and 7 grade 4). BMI has been identified as a risk factor for OA progression 2,8,9 and in a previous analysis of total cartilage plate changes in this cohort we found a trend for subjects with K–L grade 2 and 3 and high BMI to display greater cartilage thinning over 1 year than in those with low BMI and other K–L grades 5. Also, clinical trials often include subjects with high BMI and definite radiographic OA, but not with end stage radiographic OA (K–L grade=4), because there is little cartilage left to loose in the latter. For this reason, we additionally analyzed regional cartilage change in a ”high risk” subcohort with K–L grade 2 or 3 radiographic OA and a BMI >30 (n = 54).
Double oblique coronal 3D fast low angle shot (FLASH) MR images with water excitation (we), a slice thickness of 1.5 mm and an in plane resolution of 0.31mm × 0.31mm of the right knees were obtained at baseline and at year one follow up using 3T MR scanners (Siemens Magnetom Trio, Erlangen, Germany) and a quadrature transmit-receive knee coils (USA Instruments, Aurora, OH). For further technical details on this imaging please see previous publications5,10–12. The reason for analyzing the FLASH sequence in all right knees rather than any of the other sequences acquired as part of the OAI protocol was that previously published reports on the rate and sensitivity to change of cartilage morphology have been based on the FLASH or similar sequences from other vendors.
All MR acquisitions were reviewed for artifacts, coverage and completeness by the MR technologists and were immediately reacquired if needed. Images were then sent for central archiving and preparation for release at Synarc Inc and the UCSF Coordinating Center. A small sample of the MR images underwent visual quality assurance checks at Synarc and at UCSF. The MR images were provided on an external hard drive by the OAI Coordinating Center and were each quality controlled in detail and converted to a proprietary format at the image analysis center (Chondrometrics GmbH, Ainring, Germany). Segmentation of the FT cartilage plates was performed by seven technicians with formal training and at least 3 years of experience in cartilage segmentation5. Images were read in pairs blinded to the acquisition order. Segmentation included manual tracing of the total subchondral bone area (tAB) and the cartilage joint surface area (AC) of the medial tibia (MT), the lateral tibia (LT), the central (weight bearing) medial femoral condyle (cMF) and the central lateral femoral condyle (cLF) 5,13. The weight-bearing region of the femoral condyles was analyzed between the intercondylar notch and 60% of the distance to the posterior end of the femoral condyles 5,10,11. Quality control of all segmentations was performed by a single expert (S.M.), reviewing all segmented slices of every data set 5. The segmentations were used to compute the cartilage thickness over the entire subchondral bone area (ThCtAB) and included denuded areas as 0 mm cartilage thickness 5,13.
In MT and LT five subregions (central, internal, external, anterior, posterior) were computed based on the tAB (Fig. 1), with the central subregion initially (default) occupying 20% of the total tAB 3. The central tibial region was defined as a cylinder around the center of gravity of the tibial tAB, in which the diameters were adapted to the individual shape of the bone interface area 3 (Fig. 1). Further computations were then performed with the central tibial region occupying 10%, 30%, 40% or 50% of the tibial tAB 3(Fig. 2).
Since the weight-bearing femoral condyles are already limited in their anterior-posterior extension (by the femoral trochlea anteriorly, and the posterior femoral condyle posteriorly), they were subdivided into central, internal, and external strip-like regions of interest, each occupying 33.3% of the tAB in the initial (default) approach (Fig. 1). Further computations were then performed with the central femoral region occupying 25%, 50%, 66% and 75% of the femoral tAB 3 (Fig. 2).
The mean cartilage thickness was computed for all subregions, and the minimal cartilage thickness for the central subregions. Note that the minimal thickness was not determined at a single point, but was the mean of 1% of the smallest values in the central area. The mean change from baseline to follow up, the standard deviation (SD) of change, the standardized response mean (SRM= mean change/SD of change), and the significance of change (paired t-test, without correction for multiple testing) was then calculated for each cartilage plate, subregion and parameter. Note that the mean percent change (MC%) was not obtained by calculating the mean of individual % changes, but by relating the mean change (in mm) to the mean baseline values in the subcohort. For total cartilage plates and for the central subregions, changes were also reported for the medial (MFTC) and lateral (LFTC) FT compartments, by using summed values from MT and cMF, and LT and cLF, respectively11,12. Analysis of variance (ANOVA) of repeated measures was used to identify whether the change in mean cartilage thickness (ThCtAB.Me) between baseline and one year follow up differed between the subregions of each cartilage plate, both in the total sample and in the high risk subcohort. Post hoc tests were then used to identify between which subregions significant differences (in change) occurred; the significance level was set to 5% and p-values were adjusted based on the Bonferroni method. Additionally, a paired (2-sided) t-test was then used to test whether the percent changes in the central areas of MT/LT (20% tAB), and cMF/cLF (33%tAB) were significantly greater than those in the total cartilage plates, with a p-value <0.05 considered statistically significant. 95% confidence intervals were computed for each parameter, also using the observed distribution of percent changes.
In the total weight-bearing femoral condyle (cMF), the mean cartilage thickness decreased by 4.1% (95% confidence interval [CI]: 1.9%–6.4%) over the 12 month observation period in the high risk subcohort (K–L grade 2/3 and obesity; n=54) and by 1.9% (95%CI: 0.9%–2.9%) the total subcohort (n=156). ANOVA of repeated measures identified significant differences in the change of mean cartilage thickness of cMF between baseline and follow up p<0.001 for the total cohort and p<0.01 for the high risk subcohort). The thickness loss in the central subregion of cMF was significantly greater (Bonferroni corrected post hoc test) than that in the external and internal aspect of cMF (Table 1); this applied to both the total cohort (p<0.01) and to the high risk cohort (p<0.05). When analyzing a central subregion of 33% of the tAB in cMF (ccMF33), the reduction in mean cartilage thickness was significantly greater (p<0.05; paired t-test) than for the total plate (cMF) and amounted to 5.8% (95%CI: 2.5%–9.1%) in the high risk cohort and 2.8% (95%CI: 1.4%–4.2%) in the total cohort (Table 1, Fig. 3). The minimal thickness decreased by 9.2% (95%CI: 3.4%–15.1%) and 4.2% (95%CI: 1.5%–7.0%), respectively.. When varying the size of the central area ccMF) the rate of cartilage loss tended to be higher the smaller the region was chosen (5.9% for 25% tAB versus 4.9% for a 75% tAB in the high risk subcohort – Table 1, Fig. 4).
The standardized response mean for mean cartilage thickness of cMF was -0.49 in the high risk subcohort and -0.30 in the total subcohort. When considering ccMF33, the SRM for the mean cartilage thickness was similar (-0.47 and -0.31), and that for minimal thickness was lower (-0.42 and -0.24 - Table 1). The SRM in the external and internal aspect of cMF was lower than for the entire cMF. The SRM varied little with the size of central area (-0.47 at 25% tAB to -0.50 at 75% tAB in the high risk subcohort- Table 1).
In the medial tibia (MT), the mean cartilage thickness decreased by 1.0% (95%CI: 0.1%–2.0%) in the high risk subcohort and by 0.5% (95%CI: 0.01%–1.0%) in the total subcohort. No significant differences in the change of the mean cartilage thickness between the subregions of MT was identified by ANOVA (p=0.08 for the total cohort and p=0.12 for the high risk subcohort), with the results in the subregions being displayed in Table 1. When considering a central area of 20% tAB in MT (cMT20), the reduction in mean cartilage thickness was, however, greater (p=0.01) than for the total MT and amounted to 1.6% 95%CI: 0.1%–3.0%) and 0.9% (95%CI: 0.2%-1.5%), respectively (Table 1, Fig. 3). The minimal thickness decreased by 1.8%/1.5% (neither statistically significant - Table 1). When varying the size of the central area (cMT) the rate of cartilage loss tended to greater the smaller the region chosen (2.1% for 10% tAB versus 1.3% for a 50% tAB in the high risk subcohort – Table 1, Fig. 4).
The standardized response mean for mean cartilage thickness of MT was -0.24 in the high risk subcohort and -0.16 in the total subcohort. When considering cMT (20% tAB), the SRM for the mean cartilage thickness was somewhat higher (-0.29 and -0.20), and that for minimal thickness lower (-0.11 and -0.11 - Table 1). The SRM in the peripheral areas was similar or lower than for the entire MT, and there was little variation with the size of central area (-0.33 at 10% tAB to -0.29 at 50% tAB).
Analysis of MFTC provided similar results to that of cMF, with the mean and minimal cartilage thickness of the central areas yielding a higher rate of change, but a similar SRM to the total MFTC (Table 1).
There was no statistically significant change in cLF, neither for the entire cartilage plate, nor for any of the subregions (Table 2). Also, no significant differences in the change of the mean cartilage thickness between the subregions of cLF was identified by ANOVA (p=0.75 for the total cohort and p=0.91 for the high risk subcohort).
In the lateral tibia (LT), the mean cartilage thickness decreased by 1.1% (95%CI: 0.4%–1.9%) in the high risk subcohort and by 0.7% (95%CI: 0.2%–1.1%) in the total subcohort (Table 2). ANOVA identified significant differences in the change of mean cartilage thickness of LT between baseline and follow up (p<0.05 for the total cohort and for the high risk subcohort, respectively). The change in in cLT was significantly greater (Bonferroni corrected post hoc test) than that in aLT in the total cohort (p<0.01) and in the high risk subcohort (p<0.05; Table 2). When considering a 20% central area (cLT20), the decrease in mean cartilage thickness was significantly (p<0.01) greater than for the total LT and amounted to 1.6% (95%CI: 0.5%–2.7%) and 1.0% (95%CI: 0.2%–1.6%), in the two groups. The minimal thickness decreased by 1.3% and 0.8%, respectively (neither statistically significant, Table 2). When varying the size of the central area (cLT), the rate of cartilage loss tended to be greater for smaller regions of interest (1.7% for 10% tAB versus 1.3% for a 50% tAB in the high risk subcohort, Table 2).
The standardized response mean for mean cartilage thickness of LT was -0.39 in the high risk subcohort and -0.23 in the total subcohort. When considering cLT20, the SRM for the mean cartilage thickness was similar (-0.38 and -0.21), and that for minimal thickness lower (-0.13 and 0.06 - Table 2). The SRM in the peripheral areas tended to be lower than for the entire MT. The SRM varied little with the size of central area (-0.20 at 10% tAB to -0.22 at 50% tAB) for the entire cohort, but the SRM tended to increase with larger central areas of interest in the high risk cohort (-0.34 at 10% tAB to -0.40 at 50% tAB- Table 2).
As cLF, analysis of the LFTC as a compartment did not yield statistically significant changes over the observation period (Table 2).
This is the first study to present longitudinal observations of cartilage change in mediolateral subregions of the weight-bearing femoral condyle, and for minimal cartilage thickness in subregions of tibial and femoral cartilages. Also, this is the first study to examine subregional (specifically central) cartilage thickness changes with 3 Tesla MRI, and the effect of the size of the central subregions on measurements of the rate and sensitivity to change of FT cartilage loss in OA. These observation are important for two reasons: 1) the ability to monitor cartilage change in different subregions of the FT compartments and cartilage plates allows one to more comprehensively describe cartilage loss in OA and therefore better understand the disease and its progression; 2) the ability to determine defined subregions with higher rates of change and in particular sensitivity to change may permit a reduction in the required sample size in clinical studies trying to demonstrate drug effects on disease progression.
Our data show that the central subregions generally display higher rates of change than the total cartilage plates (statistically significant for MT, LT and cMF). The greatest rates of change were observed for “minimal” rather than “mean” cartilage thickness; the sensitivity to change (SRM) of minimal cartilage thickness was, however, inferior to that of the mean thickness, due to a larger inter-subject variability. The sensitivity to change for mean cartilage thickness in the central subregion of the medial tibia was higher than that for the entire cartilage plate, relatively independent of the choice of size of the region (10 to 50% tAB), whereas in cMF and LT the SRM was similar for central subregions and the entire cartilage plate. A limitation of the study is that the biomechanical axis of the leg (neutral, valgus or varus malalignment) was not measured, and future work should therefore address whether the observations made here also apply to subcohorts with different types of knee alignment.
In a previous study, we showed that the algorithm applied here is able to reliably identify regions of interest of a defined proportion of the tAB in FT cartilage plates, and that a test retest precision (RMS CV%) of 1.5% (eMT) to 3.5% (ecMF) can be obtained in the medial FT, and of 1.9% (cLT) to 4.7% (pLT) in the lateral FT compartment 3. The test retest precision error for minimal thickness (central regions) ranged from 5.0% in cLT to 10.1% in ccLF)3. Also, we have reported that total plate cartilage loss tended to be higher in those with a BMI > 30 than in those with lower BMI, and higher in those with K–L grade 2 and 3 than in those with other K–L grades in this cohort 5; however, only trends were observed and non of these or other potential risk factors (age, sex, symptoms) were found to be significantly associated with femorotibial cartilage loss.
In a recent study, Pelletier et al. 2 investigated a subset of 110 patients from a large clinical trial who suffered from symptomatic and radiographic (definite osteophyte) knee OA, a narrower medial than lateral joint space, and a medial joint space width between 2 and 4mm on semiflexed radiographs. The tibial cartilage was divided either into a concentric or a transverse (medial to lateral) central region. The authors 2 also reported that the highest rate of change (over 2 years) occurred in the central areas (-13% in concentric and -15% in transverse cMT, as opposed to - 9.3% in the total MT). The SRM was, however, similar for regional (-1.19 for concentric and - 1.25 for transverse cMT) and for total plate analysis of MT (-1.24). The medial femoral condyle was divided into an anterior, central (cMF) and posterior (pMF) subregion (rate of change = -12.4, -12.0 and -4.4% and SRM = -1.03, -1.04, and -0.56, respectively), but was not divided into medio-lateral subregions, as in the current study. The data were presented as cartilage volume change in these subregions, but not in terms of mean and minimal cartilage thickness.
Despite the use of 3T, the rate of change and SRMs observed in the total and in the high risk subcohort of the present study were lower than those reported by Pelletier et al., 2 even when accounting for the two year observation period in their study. Similar to these authors, however, we find that the central subregions display a higher rate of change than the total cartilage plates, whereas the SRMs are similar for central subregions and the total plate. The only exception to this in our study was the medial tibia, where the SRM for cMT was higher than for the total plate. No significant change was found in the high risk subcohort for the entire MT, whereas a significant loss was observed in cMT.
The minimal central cartilage thickness generally displayed greater rates of change than the mean thickness of the central subregions or total plates, but lower SRMs. In a diseased joint, the minimal cartilage thickness in central areas is likely located at the site of a lesion, where cartilage loss may occur faster than in other parts of the joint. Therefore, it is plausible that the rate of change for minimal cartilage thickness change was higher than that for mean cartilage thickness. However, the intersubject variability in the change in minimal thickness was also higher, likely because of larger changes in subjects without full thickness cartilage loss, or the absence of any loss in minimal thickness in participants with a central denuded area at baseline. In the high risk subcohort, 6 (of 54) participants had already 0mm minimal cartilage thickness in cMT at baseline, and 13 (of 54) in ccMF. Because no loss of minimal cartilage thickness can be measured in these subjects, the standard deviation of the change is larger compared to the standard deviation of change in mean cartilage thickness, with the mean thickness of cMT and ccMF having been greater than zero in all participants. The other potential explanation is the larger precision errors involved in measurement of the minimal cartilage thickness3, given that measurements are averaged over a much smaller area and therefore much more prone to inconsistencies in local segmentation and are more sensitive to partial volume averaging. Note that the average minimal cartilage thickness was 1.36 mm in cMT and 1.09 mm in cMF, so that the size of one pixel (0.31 mm) amounts to 23% and 28% of that value, respectively.
Similar considerations apply to the choice in size of the subregions. Despite the fact that the higher rate of change was generally observed with a smaller central region of interest, this did not translate into a higher SRM, because of the proportional increase in the standard deviation of the change in the smaller area. Again, a potential explanation may be the larger inter-subject variability of cartilage thickness changes measured in smaller areas compared to larger ones, or greater test-retest precision error in smaller areas. This was previously observed in the medial but not in the lateral FT compartment3.
In conclusion, this study shows that the rate of cartilage loss is greater in the central subregions compared to the total cartilage plates. Minimal (central) cartilage thickness displayed a greater rate of change than the mean thickness, but the SRM was less, due to greater variability of the changes. Smaller central subregions generally displayed higher rates of change than larger central subregions, but this did not translate into a higher SRM for the same reason. The sensitivity to change for the central subregions was greater than in the total cartilage plate in the medial tibia and was similar to the entire cartilage plate in the medial femur and lateral tibia.
We would like to thank John Lynch for help in working with the OAI images, the radiograph readers at Boston University, Drs. Piran Aliabadi, Bert Sack and David Felson, and the readers at Chondrometrics GmbH: Gudrun Goldmann, Linda Jakobi, Manuela Kunz, Dr. Susanne Maschek, Sabine Mühlsimer, Annette Thebis, and Dr. Barbara Wehr for dedicated data segmentation.
The OAI is a public-private partnership comprised of 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 Institutes of Health, a branch of the Department of Health and Human Services, and conducted by the OAI Study Investigators. Private funding partners 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 Institutes of Health. This manuscript has received the approval of the OAI Publications Committee based on a review of its scientific content and data interpretation.
Funding Source: Pfizer Inc.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.