Anthropometric Measures, Body Composition, Body Fat Distribution, and Knee Osteoarthritis in Women

Authors

  • Lauren M. Abbate,

    1. Thurston Arthritis Research Center, University of North Carolina at Chapel Hill, Chapel Hill, North Carolina
    2. Department of Epidemiology, University of North Carolina at Chapel Hill, Chapel Hill, North Carolina
    Search for more papers by this author
  • June Stevens,

    1. Department of Epidemiology, University of North Carolina at Chapel Hill, Chapel Hill, North Carolina
    2. Department of Nutrition, University of North Carolina at Chapel Hill, Chapel Hill, North Carolina
    Search for more papers by this author
  • Todd A. Schwartz,

    1. Thurston Arthritis Research Center, University of North Carolina at Chapel Hill, Chapel Hill, North Carolina
    2. Department of Biostatistics, University of North Carolina at Chapel Hill, Chapel Hill, North Carolina
    Search for more papers by this author
  • Jordan B. Renner,

    1. Thurston Arthritis Research Center, University of North Carolina at Chapel Hill, Chapel Hill, North Carolina
    2. Department of Radiology, University of North Carolina at Chapel Hill, Chapel Hill, North Carolina
    Search for more papers by this author
  • Charles G. Helmick,

    1. Centers for Disease Control and Prevention, Atlanta, Georgia.
    Search for more papers by this author
  • Joanne M. Jordan

    Corresponding author
    1. Thurston Arthritis Research Center, University of North Carolina at Chapel Hill, Chapel Hill, North Carolina
    2. Department of Epidemiology, University of North Carolina at Chapel Hill, Chapel Hill, North Carolina
    3. Department of Medicine, University of North Carolina at Chapel Hill, Chapel Hill, North Carolina
    4. Department of Orthopaedics, University of North Carolina at Chapel Hill, Chapel Hill, North Carolina
    Search for more papers by this author

Thurston Arthritis Research Center, University of North Carolina at Chapel Hill, CB# 7280, 3330 Thurston Building, Chapel Hill, NC 27599-7280. E-mail: joanne_jordan@med.unc.edu

Abstract

Objective: Increased BMI is a well-recognized risk factor for radiographic knee osteoarthritis (rKOA); however, the contributions of the components of body composition, body fat distribution, and height to this association are not clear.

Research Methods and Procedures: We examined 779 women ≥45 years of age from the Johnston County Osteoarthritis Project. Body composition was assessed using DXA, and rKOA was defined as Kellgren-Lawrence grade ≥2. Logistic regression models examined the association between rKOA and the fourth compared with the first quartiles of anthropometric, body composition, and fat distribution measures adjusting for age, ethnicity, and prior knee injury.

Results: The adjusted odds ratios and 95% confidence interval of BMI and weight were 5.27 (3.05, 9.13) and 5.28 (3.05, 9.16), respectively. In separate models, higher odds of rKOA were also found for fat mass [4.54 (2.68, 7.69)], percent fat mass [3.84 (2.26, 6.54)], lean mass [3.94 (2.22, 6.97)], and waist circumference [4.15 (2.45, 7.02)]. Waist-to-hip ratio was not associated with rKOA [1.45 (0.86, 2.43)], and percent lean mass was associated with lower odds [0.20 (0.11, 0.35)]. Taller women had higher odds of rKOA after adjustment for BMI [1.77 (1.05, 3.00)].

Discussion: This study confirms that BMI and weight are strongly associated with rKOA in women and suggests that precise measurements of body composition and measures of fat distribution may offer no advantage over the more simple measures of BMI or weight in assessment of risk of rKOA.

Introduction

Arthritis and other rheumatic conditions are highly prevalent in the United States, affecting nearly 70 million people in 2001 (1), and are the leading causes of disability (2). Osteoarthritis (OA)1, the most common of these conditions, was estimated to affect 12.1% of U.S. adults 25 to 74 years of age in 1990 (3). Knee osteoarthritis, in particular, radiographic knee osteoarthritis (rKOA), has been estimated to affect between 4.3% and 33% of the adult population (4, 5, 6), depending on the method of radiographic assessment. Risk factors for rKOA include age, female sex, and obesity (7, 8). The increasing prevalence of overweight and obesity (9) and the aging of the U.S. population indicate that there is potential for increased incidence of rKOA in the U.S. population in the future, especially among women.

Obesity most likely affects rKOA through biomechanical pathways by causing excess forces on the joint (8). Data from the first National Health and Nutrition Examination Survey (NHANES I) indicated that the association between obesity (BMI ≥ 30 kg/m2) and rKOA was not attenuated after adjustment for blood pressure, cholesterol, uric acid, and diabetes, suggesting that obesity does not act through metabolic pathways common to those disorders (10). More recently, data from the Arthritis, Diet, and Activity Promotion Trial indicated that reduced body mass was strongly associated with a reduction in peak knee forces (11). However, other studies have suggested that an obesity-associated mechanism not related to weight bearing may also be operative. Data from the Chingford study of middle-aged women indicated an association between the highest tertile of BMI and hand osteoarthritis, joints that are not directly weight bearing (12), and also suggested that increased blood glucose, a metabolic factor, was associated with rKOA independently of age and BMI (13).

Although BMI has been previously examined in relationship to rKOA (5, 6, 7, 10, 12, 14, 15, 16, 17, 18), the effect of height (19, 20, 21) and body composition components (5, 14, 16) on rKOA have been less widely studied. Three studies have examined the relationship between either skinfold measures (5, 16) or percentage body fat (14), calculated from skinfold measures (22), and rKOA. Measurement of skinfolds is subject to error (23) and does not include all types of body fat or necessarily give an indication of central fat distribution.

Because rKOA has been associated with inflammatory mediators (24, 25), central fat may be important as production of interleukin-6 and other mediators are increased in central fat (26). Additionally, increased central fat may affect gait and balance (27, 28), both of which have been associated with rKOA (29, 30). We know of only two studies that have examined associations of body fat distribution measurements to rKOA in women (12, 14). Neither found an association between rKOA and waist-to-hip ratio (WHR) or waist circumference that was independent of BMI.

It is plausible that there are effects of excess adiposity and fat distribution on rKOA that are not associated with weight bearing. Precise measures of body composition, such as those provided by DXA, have not been examined in relationship to rKOA in a large sample. The purpose of this analysis was to examine the associations of anthropometric, body composition, and fat distribution measures with rKOA in women and to evaluate whether any of these measures was more strongly associated with rKOA than BMI or weight.

Research Methods and Procedures

Participants

The Johnston County Osteoarthritis Project is an ongoing, population-based cohort study designed to estimate the incidence and progression of knee and hip osteoarthritis among African Americans and whites in a rural county in North Carolina. The sampling methods and study protocol have been previously reported (31). At the time of this interim analysis, data were available on fewer men than women, and it was determined that additional recruitment of men would be desirable before initiating analyses in that group. Furthermore, because body composition, fat distribution, and the risk of rKOA all differ by sex, this analysis was restricted to women.

Data for this analysis were from a subset of women in the cohort who responded to an invitation to be examined in the research clinic and had available both scored knee radiographs in the anteroposterior (AP) view and full body DXA scans (n = 785). Of these women, six were excluded for missing anthropometric data, leaving 779 women available for this analysis. This study was approved by the Institutional Review Boards at the University of North Carolina at Chapel Hill and the Centers for Disease Control and Prevention, and written informed consent was obtained from all participants by trained interviewers.

Measurements

The outcome of rKOA was defined as a Kellgren-Lawrence grade ≥2 in either knee (32). All radiographs were taken in the AP view with weight bearing and were read by a single bone and joint radiologist. The inter- and intra-reader reliability has been shown to be excellent, with κ equal to 0.859 and 0.886, respectively (31).

Height, in inches, was measured using a stadiometer, and weight, in pounds, was measured using a balance beam scale. BMI was calculated as weight (kilograms) divided by height squared (meters squared). Waist and hip circumferences were measured in centimeters at the height of the umbilicus and largest area of the hips, respectively. WHR was calculated from these measurements as waist circumference divided by hip circumference. DXA scans were performed using either a Lunar DPX-IQ (n = 486) or Hologic QDR Delphi A (n = 293) to produce total body fat mass, lean mass, and bone mass. Percentages of fat mass, lean mass, and bone mass were calculated by dividing the compartment mass by the total body mass (sum of fat, lean, and bone masses).

Participants completed interviewer-administered questionnaires that included information on demographics, education, history of knee injury, smoking status, and self-reported current recreational activities (33).

Statistical Analysis

Analysis was performed using the Statistical Analysis System, Version 8.02 (SAS Institute, Cary, NC). Age was analyzed as a continuous variable. Ethnicity was categorized as self-reported African American or white, and education was categorized as less than completion of high school or ≥12 years. History of prior knee injury (yes/no) and type of DXA scanner were coded as dichotomous variables. Smoking (never, past, current) and current recreational activity were coded as indicator variables. Because of the distribution of activity in our sample, current recreational activity was categorized into low (0 to 1, no/minimal activity or household work), moderate (2, light household work or yard work), and high (3 to 6, heavy yard work, walking, recreational/competitive sports). Descriptive statistics, age-adjusted means stratified by rKOA status, and Pearson correlation coefficients were calculated.

Indicator variables were used for testing all exposures in the models. Equally sized quartiles were calculated for each of the three anthropometric measures (height, weight, and BMI), the six body composition measures (fat mass, percent fat mass, lean mass, percent lean mass, bone mass, and percent bone mass), and the two body fat distribution measures (WHR and waist circumference), with cut-points based on the entire analysis sample. Quartiles were used because they provided a common unit that made it possible to compare the multiple exposure variables of interest. In addition, categorical analyses were more appropriate than continuous because associations with rKOA did not meet the linearity assumption of the logistic model for some exposures.

Eleven separate multiple logistic regression models were used to evaluate the effects of each anthropometric, body composition, and fat distribution exposure variable modeled in quartiles on rKOA, with the lowest quartile as the referent. Age, ethnicity, history of knee injury, and type of DXA scanner were included in all models. Potential confounding by smoking status, education, and current recreational activity was assessed by examining the change in the odds ratio (OR) estimate with the subsequent removal of each covariate, and all three variables were removed from all models because the estimates for each of the anthropometric, body composition, and fat distribution variables remained essentially unchanged. To determine whether the height, body composition, and fat distribution measurements were associated with rKOA independently of BMI or weight, the models were further adjusted for BMI or weight as continuous variables.

Additional models examined the effect of BMI, modeled in quartiles, with separate adjustment for height and each body composition and fat distribution measure modeled continuously. The effect of weight was assessed using the same method. These models were constructed to evaluate the magnitudes of the BMI and weight estimates independent of height and each body composition and fat distribution measure.

Because BMI, weight, body composition, and fat distribution variables were highly correlated, collinearity between these variables was assessed by computing condition indices based on a principal components eigenanalysis of the correlation matrix and computing a variance inflation factor (VIF) for each model (34). Belsley et al. (35) have suggested that condition indices >30 suggest the presence of moderate or severe collinearity, and Muller and Fetterman (34) have suggested that VIFs >10 warrant caution. For the models designed to examine whether body composition and fat distribution variables had an effect on rKOA that was independent of BMI or weight, an effort was made to reduce collinearity by modeling one variable in quartiles (main exposure) and the other variable of interest (BMI or weight) as continuous.

Results

The mean age (standard deviation) for the 779 women (range, 45 to 90 years of age) was 64.8 (9.4) years, with 212 (27.2%) identifying themselves as African-American. The prevalence of rKOA in the sample was 27.1%. The age-adjusted means and 95% confidence intervals for each of the three anthropometric, six body composition, and two fat distribution measures among those with and without rKOA are shown in Table 1. Compared with those without rKOA, women with rKOA had significantly higher mean BMI, weight, fat mass, percent fat mass, lean mass, bone mass, waist circumference (p < 0.001), and WHR (p < 0.05). Mean height (p = 0.50) was similar in the two groups of women. Compared with women without rKOA, women with rKOA had significantly lower mean percent lean mass (p < 0.001) and mean percent bone mass (p < 0.01; Table 1).

Table 1.  Age-adjusted means for measures of anthropometry, body composition, and fat distribution among women with and without rKOA (n = 779)
 No rKOA (n = 568)rKOA (n = 211)
 Mean95% CIMean95% CI
  1. rKOA, radiographic knee osteoarthritis; CI, confidence interval; WHR, waist-to-hip ratio.

Anthropometry    
 BMI (kg/m2)29.128.6, 29.633.232.4, 34.1
 Weight (kg)74.172.8, 75.485.182.9, 87.2
 Height (cm)159.6159.1, 160.2160.0159.1, 160.8
Body composition    
 Fat mass (kg)30.529.7, 31.336.735.3, 38.1
 Fat mass (%)40.940.4, 41.543.442.5, 44.3
 Lean mass (kg)40.339.7, 40.943.943.0, 44.9
 Lean mass (%)55.955.3, 56.453.552.6, 54.4
 Bone mass (kg)2.32.3, 2.42.52.4, 2.5
 Bone mass (%)3.23.2, 3.33.13.0, 3.1
Fat distribution    
 Waist circumference (cm)91.089.9, 92.199.397.4, 101.1
 WHR0.830.83, 0.840.850.84, 0.86

BMI, weight, and fat mass were most highly correlated with each other, ranging from r = 0.90 to r = 0.93. All correlations with BMI and weight were positive, except for percent lean mass (r = −0.65 and r = −0.61, respectively) and percent bone mass (r = −0.44 and r = −0.41, respectively). WHR was less strongly correlated with BMI (r = 0.28) and weight (r = 0.26) compared with the other measures. Height was only modestly correlated with weight (r = 0.28) and had a weak, negative correlation with BMI (r = −0.09). Although the body composition and fat distribution variables were highly correlated with weight and BMI (ranging from r = 0.39 to r = 0.93), the diagnostic tests for collinearity yielded ranges of condition indices (1.52 to 5.95) and VIFs (1.04 to 5.74) that were within acceptable limits.

Compared with the lowest quartile, the covariate-adjusted ORs for rKOA tended to be higher for the fourth quartile of BMI, weight, fat mass, percent fat mass, lean mass, bone mass, and waist circumference (Figure 1). Trends across quartiles were generally monotonic, although this was not true for percent lean mass or for bone mass. There were no statistically significant associations across quartiles of WHR. For percent lean mass and percent bone mass, women in the highest quartile had lower odds of rKOA compared with those in the lowest quartile. The mean BMI of women with percent lean mass in the highest quartile was 25.0 kg/m2, whereas in the lowest quartile, the mean BMI was 35.7 kg/m2. For percent bone mass, the mean BMI in the highest quartile was 33.9 kg/m2 compared with 27.0 kg/m2 in the lowest quartile.

Figure 1.

Adjusted ORs of rKOA by quartiles of anthropometric, body composition, and fat distribution measures (n = 779). Adjusted for age, ethnicity, history of knee injury, and type of DXA scanner; referent is first quartile.

Adjusted ORs for the fourth compared with the first quartile of each exposure before and after adding BMI or weight to the models as continuous variables are shown in Table 2. When BMI or weight was included, ORs were attenuated and no longer statistically significant for all variables with the exception of height, in which the estimate of effect was slightly increased and statistically significant after adjustment for BMI but not weight.

Table 2.  Adjusted ORs, from 11 separate models, of rKOA for each anthropometric, body composition, and fat distribution measure and with additional adjustment for BMI or weight (n = 779)
 Separate models*Separate models* + BMI (continuous)Separate models* + weight (continuous)
 OR95% CIOR95% CIOR95% CI
  • rKOA, radiographic knee osteoarthritis; CI, confidence interval; WHR, waist-to-hip ratio.

  • *

    Estimates compare fourth to first quartiles and are adjusted for age, ethnicity, prior knee injury, and DXA scanner.

Anthropometry      
 BMI5.273.05, 9.13    
 Weight5.283.05, 9.16    
 Height1.350.82, 2.231.771.05, 3.000.910.53, 1.54
Body composition      
 Fat mass4.542.68, 7.691.420.61, 3.291.180.47, 2.92
 Percent fat mass3.842.26, 6.541.110.54, 2.271.250.64, 2.46
 Lean mass3.942.22, 6.971.510.76, 2.991.070.50, 2.31
 Percent lean mass0.200.11, 0.350.590.28, 1.230.550.27, 1.11
 Bone mass3.721.98, 6.991.580.79, 3.181.110.52, 2.35
 Percent bone mass0.460.24, 0.892.080.92, 4.711.810.82, 4.01
Fat distribution      
 Waist circumference4.152.45, 7.021.310.60, 2.831.220.56, 2.67
 WHR1.450.86, 2.430.950.55, 1.651.000.58, 1.72

The ORs for BMI and weight, with separate adjustment for each of the anthropometric, body composition, and fat distribution variables, are shown in Table 3. Adjustment for fat mass, percent fat mass, lean mass, percent lean mass, bone mass, or waist circumference reduced the OR for rKOA associated with BMI and weight, with the largest reduction occurring with the addition of fat mass. Addition of percent bone mass, height, or WHR slightly increased the OR associated with BMI and weight.

Table 3.  Adjusted ORs of rKOA for BMI and weight with separate, additional adjustment for height and each body composition and fat distribution measure (n = 779)
 BMI*Weight*
Variable added to modelOR95% CIOR95% CI
  • OR, odds ratio; rKOA, radiographic knee osteoarthritis; CI, confidence interval; WHR, waist-to-hip ratio.

  • *

    Estimates compare fourth to first quartiles and are adjusted for age, ethnicity, history of knee injury, and type of DXA scanner.

 5.273.05, 9.135.283.05, 9.16
Height5.523.17, 9.615.563.17, 9.76
Body composition    
 Fat mass2.250.95, 5.352.340.92, 5.92
 Percent fat mass4.782.27, 10.084.482.19, 9.16
 Lean mass3.251.70, 6.243.271.56, 6.85
 Percent lean mass4.642.24, 9.624.402.18, 8.89
 Bone mass3.792.10, 6.863.741.97, 7.11
 Percent bone mass6.923.57, 13.426.193.28, 11.68
Fat distribution    
 Waist circumference2.581.14, 5.832.671.15, 6.20
 WHR5.433.09, 9.535.433.08, 9.55

Discussion

This study indicates that body composition measurements from DXA were not as strongly associated with rKOA as the more simple measures of BMI and weight in a sample of women from Johnston County, NC. Similarly, measures of fat distribution were not more strongly associated with rKOA than BMI or weight. These results suggest that the impact of obesity on rKOA may be primarily through increased mass rather than through systemic or metabolic pathways associated with excess adiposity.

As expected, our results confirmed previous findings of strong associations between BMI or weight and rKOA in women (5, 6, 7, 10, 12, 14, 15, 16, 17, 18, 19). While previous studies using data from NHANES I examined body fat using measures of skinfolds from various sites (5, 16), neither skinfold thickness measured at the subscapular site (16) nor the sum of skinfold thicknesses measured at both the subscapular and triceps sites (5) were associated with rKOA independently of BMI. Similarly, data from the Baltimore Longitudinal Study of Aging showed no association between rKOA and percent body fat calculated from skinfold measurements after adjustment for BMI (14). Our study further supports these findings and suggests that precise measures of fat mass and percent body fat (measured by DXA) are not associated with rKOA independently of BMI.

The use of DXA also allowed us to study lean mass and bone mass, components of body composition not captured in skinfold measurements that have been less commonly examined in relation to rKOA. No directly comparable study has examined lean mass and bone mass. One small study (n = 42) found no significant differences in total body lean mass, as measured by DXA, between postmenopausal women waiting for a knee replacement and similar women waiting for a hip replacement (36). We found that women with rKOA had significantly greater lean mass than those without rKOA. Bone mineral density has been addressed in relation to rKOA in previous studies (10, 36), but absolute bone mass has not been studied. Neither percent lean mass nor percent bone mass, as measured by DXA, has been examined. We suspect that greater percent lean mass and percent bone mass were associated with lower odds of rKOA, primarily because the individuals in the fourth quartiles of these measures were leaner than the individuals in the referent quartiles, and after adjustment for either BMI or weight, the inverse associations between rKOA and these measures were attenuated.

We also examined measures of body fat distribution and found that WHR was not associated with rKOA, a finding that was also reported from analyses of data from the Baltimore Longitudinal Study of Aging (14) and from the Chingford Study (12). In comparison with WHR, our findings and those from the Chingford Study (12) indicate that waist circumference is more strongly associated with rKOA. It is possible that the association between waist circumference and rKOA is related to the increased production of interleukin-6 or other inflammatory mediators or increases in C-reactive protein, which has been shown to be associated with rKOA (25). Another hypothesis is that increased waist circumference may have biomechanistic effects that may alter gait and stance, affecting the quadriceps angle and varus or valgus deformities (27).

While these body fat distribution measures and more complex measures of body composition were associated with rKOA, we found that they offered no distinct advantage over the more simple measures of either BMI or weight for two reasons. First, among the individual models, BMI and weight had the strongest association with the outcome. Second, the estimates for BMI and weight were less attenuated with the addition of any of the other body composition measures than were the estimates for each of the body composition measures with further adjustment for either BMI or weight. Although the addition of fat mass to the models resulted in the largest reduction of the point estimates for BMI and weight, this was likely because of the very high correlations between fat mass and BMI and between fat mass and weight.

Last, we were interested in the relationship between height and rKOA. A few studies have examined the association between radiographic osteoarthritis of various joints and height (19, 20, 21, 37, 38, 39), and only one of these is comparable with our study (19). In our study, height was not associated with rKOA when weight was included in the model; however, when height was modeled in addition to weight adjusted for height (BMI), height seemed to be independently associated with rKOA. This phenomenon has been previously shown by Flegal et al. (40) in a study of the effects of maternal height and weight on infant birth weight and suggests that, to determine the true effect of height on this outcome, BMI should be included in the model. The study by Lau et al. (19), using a similar methodology to our study, specifically addressed the association between standing height and rKOA and found that height was not associated with rKOA; however, the investigators did not adjust for BMI in their analysis. It is possible that their findings would have been similar to ours if the models had included BMI. Height may influence rKOA through its effects on biomechanics as suggested by Hunter et al. (20), and our finding that taller women have increased odds of rKOA suggests that prevention efforts may need to direct attention to taller women.

It is difficult to obtain a clear picture of the strength and independence of the effects of different anthropometric, body composition, and fat distribution measures on rKOA because the measures of interest are highly correlated with each other. To avoid the effects of collinearity, we examined each of the exposure variables in separate models. Furthermore, based on the diagnostic tests, collinearity did not seem to threaten the use of these models. Nevertheless, the results from these models must be interpreted carefully.

This analysis provides the most thorough study of multiple anthropometric and body composition measures and their associations with rKOA to date in a population that is at high risk for this outcome. Additional study of the effects of anthropometry, body composition, and fat distribution on OA in other weight-bearing joints, such as the hip, and in symptomatic OA can lead to a better understanding of the role of these factors in OA incidence and progression. With increases in obesity and the aging of our population, it is critical to further understand how obesity and changes in obesity affect OA outcomes. This study confirms that BMI and weight are strongly associated with rKOA in women and suggests that precise measurements of body composition and measures of fat distribution may offer no advantage over the simpler measures of BMI or weight in assessment of risk of rKOA.

Acknowledgments

This work was supported by the Centers for Disease Control and Prevention/Association of Schools of Public Health S1734, S1733, and S3486, the NIAMS Multipurpose Arthritis and Musculoskeletal Disease Center 5-P60-AR30701, and the NIAMS Multidisciplinary Clinical Research Center 5-P60-AR049465.

Footnotes

  • 1

    Nonstandard abbreviations: OA, osteoarthritis; rKOA, radiographic knee osteoarthritis, NHANES-I, National Health and Nutrition Examination Survey I; WHR, waist-to-hip ratio; AP, anteroposterior; OR, odds ratio; VIF, variance inflation factor.

  • The costs of publication of this article were defrayed, in part, by the payment of page charges. This article must, therefore, be hereby marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

Ancillary