Abstract
Objective
To determine whether central (abdominal) or peripheral (thigh) adiposity measures are associated with incident radiographic knee osteoarthritis (RKOA) independent of body mass index (BMI) and whether their relation to RKOA was stronger than that of BMI.
Design
161 Osteoarthritis Initiative participants (62% female) with incident RKOA (Kellgren/Lawrence grade 0/1 at baseline, developing an osteophyte and joint space narrowing grade ≥1 by year-4) were matched to 186 controls (58% female) without incident RKOA. Baseline waist-height-ratio (WHtR), and anatomical cross-sectional areas of thigh subcutaneous (SCF) and intermuscular fat (IMF) were measured, the latter using axial magnetic resonance images. Logistic regression assessed the relationship between each adiposity measure and incident RKOA before and after adjustment for BMI, and area under receiver operating characteristic curves for each adiposity measure was compared to that of BMI using chi-squared tests.
Results
BMI, WHtR, SCF and IMF were all significantly associated with incident RKOA when analysed separately, with similar effect sizes (odds ratio range 1.30–1.53). After adjusting for BMI, odds ratios for WHtR, SCF and IMF were attenuated and no longer statistically significant. No measure of central or peripheral adiposity was significantly more strongly associated with incident RKOA than BMI. Results were similar for men and women.
Conclusions
Although both central (WHtR) and peripheral (SCF and IMF) adiposity were significantly associated with incident RKOA, neither was more strongly associated with incident RKOA than BMI. The simple measure of BMI appears sufficient to capture the elevated risk of RKOA associated with greater amounts of localised adiposity.
Keywords: central adiposity, intermuscular fat, subcutaneous fat, knee, osteoarthritis
INTRODUCTION
Obesity is the strongest risk factor for the development of radiographic knee osteoarthritis (RKOA) in women, and is second only to knee injury in men1. Although mechanical overload of the knee that may lead to altered biomechanics and cartilage breakdown was thought to explain most of the increased risk for RKOA among persons who are obese, recent evidence has suggested that the relationship between obesity and RKOA may also be driven by metabolic factors (i.e., release of pro-inflammatory adipokines such as leptin)1.
The body mass index (BMI) is a simple anthropometric measure most commonly used to define obesity but cannot discriminate adipose from non-adipose body mass, and does not account for patterns of fat distribution throughout the body. This is important as, contrary to adipose tissue, greater muscle mass (that also contributes to greater BMI) is beneficial for knee osteoarthritis2, and adiposity at specific locations (i.e., abdominal or thigh) have distinct properties that may have different associations with disease3. While excessive central adiposity (i.e., abdominal fat including both visceral and subcutaneous depots)4 has been found to be an independent risk factor for end-stage RKOA (i.e., knee replacement surgery)5 and thigh intermuscular fat (IMF) was observed to be associated with RKOA presence and knee functional deficits6, little is known about the impact of central adiposity and peripheral adiposity (i.e., specific thigh fat depots including IMF and subcutaneous fat (SCF)4), on incident RKOA.
In previous work7, we found BMI to be a strong predictor of incident RKOA, in both sexes, although potentially by different mechanisms in men and women (i.e., with increased BMI, muscle quality (specific strength) appears to decrease due to elevated local adiposity in women, but not in men). Determining which measures of body composition are most closely associated with subsequent RKOA development, and whether these differ between men and women, is thus potentially important in the clinical management of the disease at an early (pre-radiographic) stage, and in its prognosis. Further, these relationships may provide clues to better understand the pathogenesis of the disease in view of obesity-related risk factors. Therefore, the aim of this study was to test the hypothesis that central or peripheral adiposity, and specifically intermuscular and subcutaneous adipose tissue, confer greater odds of incident RKOA than BMI, and to explore whether there was any significant difference in the relationship of these adiposity measures with incident RKOA.
METHODS
Participants
This study was ancillary to the Osteoarthritis Initiative (OAI), a multicentre longitudinal study of men and women aged 45–79 years designed to identify risk factors for RKOA incidence and progression (http://oai.epi-ucsf.org/). In the current analysis, we studied all OAI participants who developed incident RKOA, as defined and outlined in a recent report on thigh muscle cross-sectional area and RKOA risk7. Briefly, 186 knees with incident RKOA (i.e., cases), defined as a knee without RKOA at baseline (i.e., Kellgren-Lawrence grade (KLG) 0/1 from fixed-flexion posteroanterior radiographs) that developed both an osteophyte and joint space narrowing (JSN) (Osteoarthritis Research Society International atlas JSN grade ≥1) by 48-month follow-up7, were frequency-matched by baseline KLG 0/1 to 186 control knees without incident RKOA. The left knees of 25 participants with bilateral incident RKOA were excluded from analyses, resulting in 161 incident case knees (161 participants) and 186 frequency-matched control knees (186 participants). The OAI was approved by each study site’s institutional review board, and participants gave informed consent.
Local and central adiposity
From baseline T1-weighted axial-spin-echo thigh MRI acquisitions (imaging dataset 0.E.1), SCF and IMF anatomical cross-sectional areas (ACSAs) were segmented at an anatomically consistent location (33% femoral length; distal-proximal), using custom software8. Briefly, a semi-automated algorithm applied a convex ‘sling’ around the outer circumference and the thigh muscle tissue, separating SCF. Using a signal intensity threshold, IMF was separated from other intermuscular tissue (i.e., vessels, nerves etc.), with the method having shown to display reasonable test-retest precision8. To account for body size differences, SCF and IMF ACSAs were normalized to femoral bone ACSA (Supplementary File).
Central adiposity was estimated using the waist circumference-to-height ratio (WHtR), a well-accepted proxy measure of central obesity9. Waist circumference was measured at the level of the umbilicus between the lower rib and iliac crest manually with a tape measure to the nearest millimeter. To calculate BMI (kg/m2), height was measured barefoot using a stadiometer to the nearest millimeter and weight with lightweight clothing only using a balance beam scale to the nearest 0.1kg.
Statistical analysis
Generalised linear models with logistic regression were used to evaluate the contribution of BMI and each of the three adiposity measures (WHtR, SCF and IMF) to incident RKOA, both unadjusted and adjusted for age (continuous), sex, race (dichotomized as self-reported white vs. other), and knee extensor muscle specific-strength (force ÷ anatomical cross-sectional area [Newtons/cm2]) as previously described in this cohort7. To allow for a direct comparison of the effect of BMI and each adiposity measure on incident RKOA, we report odds ratios (ORs) and 95% confidence intervals (CIs) per standard deviation of each of these. The area under receiver operating characteristic curves (AUC) for each of the adiposity measures association with incident RKOA was compared with that for BMI, and amongst each other, using a chi-squared test (DeLong method). To determine which of central or peripheral adiposity measures conferred greater odds of incident KOA, independent of BMI, logistic regression models for each of the three adiposity measures were adjusted for BMI, similar to our previous analysis of highly correlated metabolic variables in knee OA10. Finally, we constructed a single model with all adiposity measures included and performed a backward stepwise logistic regression analysis (with BMI and demographic covariates forced in and adiposity measures added when p<0.2) to determine whether any of the central or peripheral adiposity measures contributed significantly to the odds of incident RKOA, independent of BMI, and when adjusting for one another. We carefully assessed multicollinearity and goodness-of-fit by computing a variance inflation factor (VIF), and by performing a Pearson goodness-of-fit test for each model. When VIFs were >3.3 (a conservative indication of multicollinearity)11, additional regression models were created using the residuals of the respective adiposity measures to control for weight-related confounding, as done previously10. The sex-specific residuals of the adiposity measures were the difference between the observed and expected values from separate linear regression models, with each adiposity measure as the dependent variable and BMI as the independent variable. We present whole cohort results and evaluate the interaction term between sex and each adiposity measure. Although the interaction term showed that no significant differences in the relationship between each adiposity measure and incident RKOA existed between men and women (p>0.5), as we previously observed potentially different mechanisms by which BMI increases risk of incident RKOA in men and women7, we also present sex-specific results for completeness. Statistical analyses were completed using Stata v14.2.
RESULTS
Of the 161 incident cases and 186 controls, 100 (62%) and 108 (58%) were female, respectively (Table 1). In the whole cohort, BMI, WHtR, SCF and IMF were all significantly associated with incident RKOA when analysed separately, with similar effect sizes, before and after adjustment for age, race and knee extensor specific-strength (OR range 1.30–1.53) (Table 2). No central or peripheral adiposity measure continued to be associated with incident RKOA after adjustment for BMI (Table 2). Direct comparison of AUCs showed that no measure of central or peripheral adiposity was more strongly associated with incident RKOA than BMI, nor was there any significant difference between various central and peripheral adiposity measures (AUC ranged from 0.62 to 0.63, chi-squared p>0.3). Similarly, when considering all adiposity measures and covariates in one model, no central or peripheral adiposity measure influenced the odds of RKOA sufficiently, independent of BMI, to be included in the final model determined by stepwise regression (whole cohort and sex-specific results therefore reflect adjusted model 1 OR for BMI (Table 2)). Sensitivity analysis based on KLG showed similar results for KLG 0 knees as KLG 1 knees (data not shown), with somewhat wider confidence intervals for the KLG 0 knees as a result of fewer participants having KLG 0.
Table 1.
Baseline demographic characteristics*
| All | Women | Men | ||||
|---|---|---|---|---|---|---|
|
|
|
|||||
| Incident RKOA cases (n=161) | Controls (n=186) | Incident RKOA cases (n=100) | Controls (n=108) | Incident RKOA cases (n=61) | Controls (n=78) | |
| Age, years | 61±9 | 60±9 | 61±9 | 61±9 | 61±9 | 59±9 |
| Height, m | 1.68±0.10 | 1.69±0.10 | 1.62±0.06 | 1.62±0.07 | 1.77±0.07 | 1.79±0.06 |
| Weight, kg | 82.0±15.3 | 78.5±15.4 | 76.2±12.8 | 71.0±11.7 | 91.7±14.2 | 89.0±13.8 |
| BMI, kg/m2 | 29.1±4.2 | 27.3±4.0 | 28.9±4.4 | 27.0±4.2 | 29.4±3.9 | 27.8±3.7 |
| Race white/other, n(%) | 134 (83)/27 (17) | 171 (92)/15 (8) | 78 (77)/22 (22) | 98 (90)/10 (9) | 56 (92)/5 (8) | 73 (94)/5 (6) |
| KL grade 0/1, n(%) | 47 (29)/115 (71) | 53 (28)/134 (72) | 29 (29)/72 (71) | 35 (32)/74 (68) | 18 (30)/43 (70) | 18 (23)/60 (77) |
| Waist circumference, cm | 104.2±11.9 | 99.8±11.6# | 103.2±12.3 | 98.5±11.8 | 105.8±10.9 | 101.7±11.2# |
| WHtR | 0.62±0.07 | 0.59±0.07# | 0.64±0.08 | 0.61±0.08 | 0.60±0.06 | 0.57±0.06# |
| SCF absolute, cm2 | 75.10±36.53 | 64.83±31.50 | 93.51±32.77 | 81.79±29.01 | 44.91±17.14 | 41.34±15.89 |
| SCF per femur bone area | 13.10±7.38 | 11.21±3.55 | 17.26±6.13 | 15.18±5.45 | 6.28±2.65 | 5.70±2.24 |
| IMF absolute, cm2 | 13.77±4.05 | 12.39±3.55 | 13.52±4.01 | 11.92±3.20 | 14.17±4.12 | 13.03±3.91 |
| IMF per femur bone area | 2.28±0.72 | 2.03±0.61 | 2.48±0.73 | 2.20±0.61 | 1.96±0.57 | 1.78±0.51 |
| Knee extensor specific-strength, N/cm2 | 6.99±1.97 | 7.41±2.12 | 6.72±2.02 | 7.55±2.35 | 7.44±1.80 | 7.21±1.74 |
values are mean±standard deviation unless indicated otherwise.
1 male control had missing measurement of waist circumference
RKOA, radiographic knee osteoarthritis; BMI, body mass index; KL, Kellgren-Lawrence; WHtR, waist-to-height ratio; SCF, subcutaneous fat; IMF, intermuscular fat; N, Newtons.
Table 2.
Relationship (odds ratio and 95% confidence interval) between each adiposity measure and incident radiographic knee osteoarthritis*
| All (n=347) | Women (n=208) | Men (n=139) | |||||||
|---|---|---|---|---|---|---|---|---|---|
|
|
|
||||||||
| Unadjusted | Adjusted model 1 | Adjusted model 2 | Unadjusted | Adjusted model 1 | Adjusted model 2 | Unadjusted | Adjusted model 1 | Adjusted model 2 | |
| BMI, kg/m2 | 1.39 (1.14, 1.70) | 1.51 (1.20, 1.90) | NA | 1.58 (1.18, 2.11) | 1.43 (1.05, 1.93) | NA | 1.54 (1.08, 2.19) | 1.55 (1.08, 2.22) | NA |
| WHtR | 1.30 (1.08, 1.57) | 1.53 (1.22, 1.93) | 1.30 (0.88, 1.94)# | 1.44 (1.09, 1.92) | 1.43 (1.06, 1.93) | 1.25 (0.77, 2.04) | 1.66 (1.16, 2.39) | 1.65 (1.15, 2.39) | 1.48 (0.72, 3.04)# |
| SCF, cm2 | 1.38 (1.11, 1.71) | 1.35 (1.08, 1.68) | 1.12 (0.86, 1.46) | 1.44 (1.08, 1.92) | 1.39 (1.03, 1.86) | 1.21 (0.83, 1.75) | 1.27 (0.91, 1.77) | 1.27 (0.90, 1.78) | 1.04 (0.70, 1.54) |
| IMF, cm2 | 1.48 (1.18, 1.85) | 1.41 (1.12, 1.78) | 1.20 (0.91, 1.57) | 1.54 (1.14, 2.08) | 1.38 (1.01, 1.88) | 1.22 (0.86, 1.73) | 1.40 (0.99, 1.98) | 1.40 (0.99, 1.98) | 1.12 (0.73, 1.73) |
values are odds ratios and 95% confidence intervals. Odds ratios are based on standardised measures (per standard deviation of each respective adiposity measure).
Adjusted model 1: adjusted for age, race and knee extensor specific-strength (and sex for the whole cohort analysis). Adjusted model 2: adjusted for age, race, knee extensor specific-strength and BMI (and sex for the whole cohort analysis).
Multicollinearity assumptions (i.e., VIF <3.3) met for all regression models when adjusting for BMI, except waist-to-height ratio in total cohort (VIF 4.6) and in men (VIF 4.3). Therefore, waist-to-height ratio residuals were used in the model adjusting for BMI. Pearson goodness-of-fit tests indicated good model fit for all adiposity measures (i.e., all non-significant, p>0.3).
BMI, body mass index; WHtR, waist-to-height ratio; SCF, subcutaneous fat; IMF, intermuscular fat.
In unadjusted sex-specific analyses, BMI displayed a larger effect size than other adiposity measures in women, whereas WHtR was most strongly associated with incident RKOA in men. In contrast to women, greater amounts of peripheral IMF and SCF in men were not significantly linked to incident RKOA (Table 2). However, in both men and women, no single central or peripheral adiposity measure was significantly associated with incident RKOA independent of BMI (Table 2), nor was any significantly more strongly associated with incident RKOA than BMI or significantly different between the respective adiposity measures (AUC for women range 0.61–0.62, and for men 0.58–0.59, chi-squared p>0.5).
DISCUSSION
In this first evaluation of the relationship between specific, local adiposity measures and the risk of incident RKOA, neither central (WHtR) nor peripheral (SCF or IMF) adiposity was observed to be significantly more strongly associated with incident RKOA than BMI. Although both central and peripheral adiposity were significantly associated with the development of RKOA, independent of age, sex, race and muscle specific-strength, they did not provide independent information to BMI, nor was any one adiposity measure more strongly associated with incident RKOA than another. The commonly used measure of BMI thus appears to be sufficient to capture the elevated risk of RKOA associated with greater amounts of localized adiposity.
A limitation of the current study is that different methods were used to estimate central and peripheral adiposity. Thigh MRIs obtained in the OAI enabled distinction between SCF and IMF, whereas in the absence of MRIs or computer tomography images of the trunk in the OAI, central adiposity was estimated from an anthropometric measure (WHtR). While accurately reflecting central adiposity quantity9, this measure is unable to distinguish between visceral and subcutaneous adiposity. Visceral adiposity is known to be more strongly associated with metabolic-related complications12 and may affect RKOA differently compared to the less metabolically active SCF. Further, intramuscular fat, may provide additional insights into the sex-specific risk of RKOA independent of BMI, particularly as intramuscular fat reduces muscle quality (specific strength), which was recently shown to increase incident RKOA risk in women, but not in men7. In other cohorts, thigh intramuscular fat (specifically vastus medialis) was found to be a strong predictor of knee cartilage volume loss13. However, the OAI did not acquire MRI sequences that were technically suitable for intramuscular fat assessment. Finally, small trends in the relationship of the various adiposity measures were observed with incident RKOA that slightly differed between men and women, and we cannot exclude that these may have become statistically significant in a larger sample. However, it is unlikely that a greater sample than the OAI with its 5,000 participants and 4-year radiographic follow-up can be studied in the foreseeable future.
Our findings suggest that specific measures of peripheral and central adiposity, and the commonly used surrogate measure of adiposity (BMI), are similarly correlated with incident RKOA. While these localised measures of adiposity may be associated with incident RKOA through other mechanisms than BMI, such as distinct metabolic (i.e., local adipokines such as leptin) and biomechanical pathways (i.e., local impaired muscle strength), they do not appear to provide additional independent information on RKOA risk than BMI alone. These results extend previous cross-sectional and longitudinal analyses suggesting that precise measurements of fat distribution (i.e., waist circumference, waist-to-hip ratio) offer no advantage over the more simple measure of BMI in assessment of RKOA incidence/presence10, 14. Large population-based cohorts have also shown that BMI and waist circumference share similar risk for progression of RKOA to total joint replacement5, 15. The impact of obesity on RKOA appears to be comparable between men and women in terms of spatial distribution of adipose tissue. Although we observed some apparent sex-specific effects of peripheral adiposity (i.e., SCF and IMF statistically significant in women but not in men), the interaction terms for sex and each adiposity measure were not statistically significant, suggesting that relationships with incident RKOA do not significantly differ between men and women.
In conclusion, peripheral and central adiposity measures were associated with incident RKOA to a similar extent as BMI, and did not provide significant independent information on incident RKOA risk in addition to BMI. These results suggest that a simple measure of BMI appears to be sufficient to capture the elevated risk of RKOA associated with greater peripheral and central adiposity.
Supplementary Material
Acknowledgments
We thank OAI participants, investigators, and clinical center staff for generating this publicly available data set, and Martina Sattler for assistance with MRI segmentations.
ROLE OF THE FUNDING SOURCE
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. 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. The image analysis was supported by the Paracelsus Medical University (PMU) Forschungsforderungsfond (PMU FFF E-11/14/073-WIR) as well as the NIH AR47785 funding. Dr. Culvenor is a recipient of a National Health and Medical Research Council (NHMRC) of Australia Early Career Fellowship (Neil Hamilton Fairley Clinical Fellowship, APP1121173). Dr. Felson is support by NIH AR47785. The sponsors were not involved in the design and conduct of this particular study, in the analysis and interpretation of the data, and in the preparation, review, or approval of the manuscript.
Footnotes
AUTHOR CONTRIBUTIONS
DF, WW, FE conceived the project; WW and TD extracted data from the OAI and assisted with MR imaging segmentation; AGC, DF, FE contributed to data analysis and interpretation. All authors drafted or revised the manuscript for important intellectual content and approved of the final version of the paper. FE managed the project, and obtained project funding. He takes full responsibility for the integrity of the work as a whole, from inception to finished article. felix.eckstein@pmu.ac.at
CONFLICT OF INTEREST STATEMENT
Wolfgang Wirth has a part time employment with Chondrometrics GmbH and is a co-owner of Chondrometrics GmbH, a company providing MRI analysis services to academic researchers and to industry. Torben Dannhauer has a part time employment with Chondrometrics GmbH. Felix Eckstein is CEO of Chondrometrics GmbH; he has provided consulting services to EMD Serono, Bioclinica/Synarc, Samumed, Abbvie and Servier, has prepared educational sessions for Medtronic, and has received research support from Pfizer, Eli Lilly, Merck Serono, Novartis, Stryker, Abbvie, Kolon, Synarc, Ampio, BICL, Orthotrophix and Tissuegene. Adam Culvenor and David Felson have no competing interests.
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.
References
- 1.Felson DT, Lawrence RC, Dieppe PA, et al. Osteoarthritis: new insights. Part 1: the disease and its risk factors. Ann Int Med. 2000;133:635–46. doi: 10.7326/0003-4819-133-8-200010170-00016. [DOI] [PubMed] [Google Scholar]
- 2.Wang Y, Wluka AE, English DR, et al. Body composition and knee cartilage properties in healthy, community-based adults. Ann Rheum Dis. 2007;66:1244–8. doi: 10.1136/ard.2006.064352. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Ferreira I, Snijder MB, Twisk JWR, et al. Central fat mass versus peripheral fat and lean mass: opposite (adverse versus favorable) associations with arterial stiffness? The Amsterdam Growth and Health Longitudinal Study. J Clin Endocrinol Metab. 2004;89:2632–9. doi: 10.1210/jc.2003-031619. [DOI] [PubMed] [Google Scholar]
- 4.Messier SP, Beavers DP, Loeser RF, et al. Knee-joint loading in knee osteoarthritis: influence of abdominal and thigh fat. Med Sci Sports Exerc. 2014;46:1677–83. doi: 10.1249/MSS.0000000000000293. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Wang Y, Simpson AH, Wluka AE, et al. Relationship between body adiposity measures and risk of primary knee and hip replacement for osteoarthritis: a prospective cohort study. Arthritis Res Therap. 2009;11:R31. doi: 10.1186/ar2636. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Maly MR, Calder KM, Macintyre NJ, Beattie KA. Relationship of intermuscular fat volume in the thigh with knee extensor strength and physical performance in women at risk of or with knee osteoarthritis. Arthritis Care Res. 2013;65:44–52. doi: 10.1002/acr.21868. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Culvenor AG, Felson DT, Niu J, et al. Thigh muscle specific strength and the risk of incident knee osteoarthritis: the influence of sex and greater body mass index. Arthritis Care Res. 2017;69:1266–70. doi: 10.1002/acr.23182. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Dannhauer T, Ruhdorfer A, Wirth W, Eckstein F. Quantitative relationship of thigh adipose tissue with pain, radiographic status, and progression of knee osteoarthritis. Invest Radiol. 2015;50:268–74. doi: 10.1097/RLI.0000000000000113. [DOI] [PubMed] [Google Scholar]
- 9.Swainson MG, Batterham AM, Tsakirides C, Rutherford ZH, Hind K. Prediction of whole-body fat percentage and visceral adipose tissue mass from five anthropometric variables. PLoS ONE. 2017;12:e0177175. doi: 10.1371/journal.pone.0177175. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Niu J, Clancy M, Aliabadi P, Vasan R, Felson DT. Metabolic syndrome, its components, and knee osteoarthritis: The Framingham Osteoarthritis Study. Arthritis Rheum. 2017;69:1194–203. doi: 10.1002/art.40087. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Diamantopoulos A, Siguaw JA. Formative versus reflective indicators in organizational measure development: A comparison and empirical illustration. Br J Management. 2006;17:263–82. [Google Scholar]
- 12.Hamdy O, Porramatikul S, Al-Ozairi E. Metabolic obesity: the paradox between visceral and subcutaneous fat. Curr Diabetes Rev. 2006;2:367–73. doi: 10.2174/1573399810602040367. [DOI] [PubMed] [Google Scholar]
- 13.Raynauld JP, Pelletier JP, Roubille C, et al. Magnetic resonance imaging-assessed vastus medialis muscle fat content and risk for knee osteoarthritis progression: relevance from a clinical trial. Arthritis Care Res. 2015;67:1406–15. doi: 10.1002/acr.22590. [DOI] [PubMed] [Google Scholar]
- 14.Abbate LM, Stevens J, Schwartz TA, Renner JB, Helmick CG, Jordan JM. Anthropometric measures, body composition, body fat distribution, and knee osteoarthritis in women. Obesity. 2006;14:1274–81. doi: 10.1038/oby.2006.145. [DOI] [PubMed] [Google Scholar]
- 15.Lohmander LS, Gerhardsson de Verdier M, Rollof J, Nilsson PM, Engstrom G. Incidence of severe knee and hip osteoarthritis in relation to different measures of body mass: a population-based prospective cohort study. Ann Rheum Dis. 2009;68:490–6. doi: 10.1136/ard.2008.089748. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.