To determine the prevalence of obesity and how accurately standard anthropometric measures identify obesity among men and women with rheumatoid arthritis (RA).
To determine the prevalence of obesity and how accurately standard anthropometric measures identify obesity among men and women with rheumatoid arthritis (RA).
Dual x-ray absorptiometry (DXA) was performed for 141 persons with RA (56 men and 85 women). Two anthropometric proxies of obesity (body mass index [BMI] and waist circumference [WC]) were compared to a DXA-based obesity criterion. Receiver operating characteristic curves determined optimal cut points for each anthropometric measure, relative to DXA. The association of body fat and anthropometric obesity measures with disease status and cardiovascular risk was assessed in multiple regression analyses, controlling for age and glucocorticoid use. All analyses were performed separately for men and women.
A total of 20%, 32%, and 44% of women and 41%, 36%, and 80% of men were classified as obese by BMI, WC, and DXA, respectively. Cut points were identified for anthropometric measures to better approximate DXA estimates of percent body fat (BMI ≥26.1 kg/m2 for women and ≥24.7 kg/m2 for men; WC ≥83 cm for women and ≥96 cm for men). For women and men, higher percent fat was associated with poorer RA status. Anthropometric measures were more closely linked to RA status for women, but identified cardiovascular risk for both women and men.
A large percentage of this RA sample was overfat; DXA-defined obesity was twice as common in men as in women. Utility of revised BMI and WC cut points compared to traditional cut points remains to be examined in prospective studies, but results suggest that lower, sex-specific cut points may be warranted to better identify individuals at risk for poor RA and/or cardiovascular outcomes.
Until recently, most studies of body composition in rheumatoid arthritis (RA) have focused on low lean mass or cachexia that may be caused by chronic inflammation (1, 2). However, newer studies suggest that a substantial portion of individuals with RA may be obese or overfat (3–5), which may in part explain the increased cardiovascular (CV) disease risk seen in RA. Visceral fat appears to carry the greatest CV risk (6–8), and a high prevalence of central or abdominal obesity has been found in RA (9, 10). The strong association between central obesity and CV outcomes suggests that waist circumference (WC), often used as a proxy measure for central obesity or visceral fat mass, may be useful in RA, but this has not been confirmed.
In most large-scale studies, obesity is estimated from body mass index (BMI; weight [kg]/height [m2]). BMI may not accurately reflect the amount of body fat in persons with RA (4, 9), however, because rheumatoid cachexia may occur with little or no weight loss; therefore, an individual may have a BMI within a normal range, but may have greater fat mass than suggested by the BMI. As an example, one recent study reported that ∼33% of an RA cohort was obese by BMI, a proportion that is approximately equivalent to the proportion of obesity in the general population (5). Yet, using a more sensitive measure of body composition, more than one-half of this same sample was resolutely overfat. Use of RA-specific BMI criteria for defining obesity has been suggested (4), but little validation of revised BMI criteria has occurred.
Additionally, there is preliminary evidence of sex differences in the impact of RA on body composition. Giles et al reported that women with RA had significantly greater total body fat than women in a BMI-matched control group, but no differences in total fat were noted between men with RA and male controls (5). Perhaps most important, however, is that men with RA appeared to accumulate greater visceral fat (10).
The goals of these analyses were to 1) determine the proportion of a cohort of men and women with RA who were obese using dual x-ray absorptiometry (DXA) and 2 common anthropometric proxy measures, BMI and WC; 2) evaluate the accuracy of the standard obesity criteria for anthropometric measures for men and women with RA compared to DXA determinations of obesity, and identify criteria that best reflect DXA results; and 3) examine the association of body fat and obesity with measures of RA symptoms and disease activity and CV risk.
Obesity is associated with numerous negative health effects, including heightened cardiovascular risk.
The prevalence of obesity is high in rheumatoid arthritis (RA); there appear to be substantial sex differences in prevalence, with obesity more common in men than in women with RA.
Obesity defined by anthropometric proxy measures identifies poor RA-specific outcomes, as well as cardiovascular risk.
Lower, sex-specific cut points to define obesity by both body mass index and waist circumference may be warranted.
The majority of the research participants were drawn from the University of California, San Francisco (UCSF) RA Panel. The RA Panel was constructed in 1982 from a random sample of rheumatologists practicing in Northern California and originally consisted of 822 patients. There were subsequently 4 additional enrollment periods in 1989–1990, 1995, 1999, and 2003, during which 203, 131, 122, and 169 individuals were enrolled, respectively. The principal data source for the RA Panel is an annual structured telephone interview that includes questions on demographics, RA symptoms, comorbidities, and functioning. Annual retention from year to year has averaged 93%; the 7% attrition includes deaths.
At the end of the telephone interviews in the study years 2007–2009, RA Panel participants who lived in the greater San Francisco Bay area and were willing to travel to UCSF were recruited for in-person assessments (including measurement of body composition) at the UCSF Clinical and Translational Science Institute's Clinical Research Center. Exclusion criteria were non–English speaking, age <18 years, current daily oral prednisone dose ≥50 mg, current pregnancy, uncorrected vision problems that interfered with reading, and joint replacement within 1 year.
A total of 101 participants were recruited from the RA Panel. There were no significant differences between eligible RA Panel members who participated compared to those who did not participate in terms of sex, race/ethnicity, education, RA duration, Health Assessment Questionnaire (HAQ) scores (11), pain severity ratings, depressive symptom scores, or BMI from self-reported height and weight. Participants were, however, significantly younger than nonparticipants (mean age 61 years versus 67 years; P < 0.0001). In 2009, an additional 44 subjects were recruited from the UCSF rheumatology clinic and from another study of RA. Data were not available to compare characteristics of participants recruited from these sources to individuals who declined to participate.
Of 242 eligible individuals, 97 (40.1%) declined participation, primarily because of transportation (n = 36) and scheduling difficulties (n = 38), and 145 individuals completed the study visits. Four participants were excluded from the analysis because they did not complete the body composition assessment. Of the remaining 141 participants, 85 (60.3%) were women and 56 (39.7%) were men. The study was approved by the UCSF Committee on Human Research.
Height was measured with a wall-mounted stadiometer. Weight was measured with subjects wearing light indoor clothing and no shoes. BMI was calculated as weight (kg) divided by height (m2). Obesity by BMI was initially defined as a BMI ≥30 kg/m2 (12). WC was measured with a nonstretch measuring tape that applies a consistent amount of tension to the tape (Gullick II Tape Measure) at the midpoint between the lower border of the ribs and the iliac crest. Two measurements were taken, and the average measure was used. Women with a WC ≥88 cm and men with a WC ≥102 cm were initially classified as obese (12).
Body composition and regional body fat distribution were assessed in the Clinical Research Center using a Lunar Prodigy DXA system. DXA has been validated as a method of assessing body composition in both younger and older persons, has good reported reproducibility, and is sensitive to small changes in body composition (13–16). Precision error (1 SD) for percent fat is 1.4%, for fat mass is 1.0 kg, and for lean tissue mass is 0.8 kg (14). DXA has previously been successfully used to assess body composition in RA (5, 17–22).
There is no agreed-upon standard definition of obesity based on percent body fat (23). We used a standard that linked percent fat to the National Institutes of Health BMI guidelines defining obesity (24). To develop obesity criteria, the average percent fat for individuals with a BMI between 30 and 35 kg/m2 (obese, but not morbidly obese) from 3 samples from the US, UK, and Japan was ascertained by DXA, and aggregated by sex, age, and racial group. Obesity definitions derived thus ranged from 38% fat for African American women ages 20–39 years to 43% fat for white women ages 60–79 years, and from 26% fat for African American and white men ages 20–39 years to 31% fat for white men ages 69–79 years (24). We used those percentages as the criteria for defining obesity from DXA total percent fat, based on individuals' sex, age, and race. Other definitions of obesity based on percent fat have been suggested, but those specify lower fat percentages as obese (25, 26).
Sociodemographic characteristics (e.g., age, race/ethnicity, education, and smoking status) were obtained from self-report. Glucocorticoid use was ascertained at the time of the visit by self-report.
Self-reported RA disease activity was assessed at the visit with the Rheumatoid Arthritis Disease Activity Index (RADAI) (27, 28). RADAI scores range from 0–10, with higher scores reflecting greater disease activity. The RADAI has been shown to be reliable and valid (27, 28). Pain was rated on an 11-point numerical rating scale ranging from 0–10 (where 0 = no pain and 10 = extreme pain) (29). The fatigue severity subscale of the Multidimensional Assessment of Fatigue was used; scores range from 0–10 (where 0 = no fatigue and 10 = severe fatigue) (30, 31). Functional limitations were assessed with the HAQ (scores range from 0–3, with higher scores reflecting greater limitations) (11). Blood was drawn and processed through a commercial laboratory to ascertain the erythrocyte sedimentation rate (ESR).
Data were collected at the clinic visit that permitted calculation of the Framingham CV risk score (32). Blood pressure was measured by registered nurses while the subjects were seated. Serum lipids and high-sensitivity C-reactive protein (hsCRP) level were obtained through nonfasting blood draws. Although fasting measurements of blood lipids are ideal, nonfasting measures of total cholesterol and high-density lipoprotein cholesterol have been found to very closely approximate fasting levels (33). Serum samples were processed by a commercial laboratory. Because hsCRP values were not normally distributed, log values were used in analyses.
Differences in sociodemographic, disease, and body composition characteristics between men and women were assessed with t-tests and chi-square analyses. Receiver operating characteristic (ROC) curves were calculated to determine optimal cutoff points for each anthropometric measure, relative to DXA-determined obesity. Two threshold selection methods were used: the Youden index and a second technique that determines the proximity to perfect correspondence (referred to here as the “Distance to Perfect” index) (34). Briefly, the Youden index determines the maximum vertical distance from the ROC curve to the diagonal reference, or “chance” line; the optimal cutoff point corresponds to the point on the ROC curve farthest from the reference line, which has also been used as a measure of the accuracy of a diagnostic test in clinical epidemiology (35). The Distance to Perfect index selects the point on the ROC curve that is closest to the upper left-hand corner of the graph (0, 1), which represents perfect classification (36), thereby minimizing misclassification. We calculated the sensitivity, specificity, and positive and negative predictive values of both the established and new criteria for each anthropometric measure compared to the DXA-based obesity classification.
To examine the potential usefulness of the original and revised obesity criteria, we first examined the relationship of DXA-based body composition measures (total percent body fat, obesity classification, and percent total body mass from truncal fat) to disease status measures and CV risk using multivariate linear regression, controlling for age, glucocorticoid use, RA duration, and smoking. (The exception was for Framingham risk scores. Because smoking status is used to calculate the risk score, smoking was not included as a covariate in these regression analyses.) Next, we examined the relationship of anthropometric measures to the same disease and CV measures, also controlling for age, glucocorticoid use, RA duration, and smoking. These analyses were intended to identify relationships to DXA-defined body fat estimates, and then determine if anthropometric proxies of obesity were sensitive to these same associations.
Because body composition differences exist between racial and ethnic groups at the population level, we repeated all of the analyses including only white, non-Hispanic subjects (excluding 7 men and 7 women). Results from these analyses did not yield results substantively different from those obtained with the total sample, and are therefore not shown.
Characteristics of the sample are shown in Table 1. There were no significant sex differences in the sociodemographic or clinical variables, with the following exceptions: disease duration was greater for women (21.6 years versus 16.1 years; P = 0.002), a greater proportion of men were positive for anti–citrullinated protein antibodies (100% versus 81.2%; P = 0.0002), and the mean Framingham risk score was higher for women (11.5 versus 9.1; P = 0.006).
|Total (n = 141)||Men (n = 56)||Women (n = 85)||P|
|Age, mean ± SD years||58.0 ± 10.8||56.6 ± 11.1||58.9 ± 10.5||0.21|
|White, % (no.)||90.1 (127)||87.5 (49)||91.8 (78)||0.41|
|Education <12 years, % (no.)||3.6 (5)||3.6 (2)||3.5 (3)||0.99|
|RA duration, mean ± SD years||19.4 ± 11.2||16.1 ± 8.3||21.6 ± 12.3||0.002|
|Current glucocorticoid use, % (no.)||35.5 (50)||33.9 (19)||36.5 (31)||0.86|
|Current dose, mean ± SD mg†||6.3 ± 5.2||7.1 ± 7.3||5.7 ± 3.3||0.42|
|Highest dose in the past year, mean ± SD mg†||17.1 ± 19.4||13.6 ± 13.8||19.9 ± 22.7||0.18|
|Pain rating (range 0–10), mean ± SD||2.6 ± 2.2||2.8 ± 2.3||2.5 ± 2.1||0.38|
|RADAI score (range 0–10), mean ± SD||2.6 ± 1.7||2.7 ± 1.8||2.5 ± 1.7||0.48|
|Fatigue severity (range 0–10), mean ± SD||5.0 ± 2.5||5.4 ± 2.7||4.8 ± 2.4||0.14|
|HAQ score (range 0–3), mean ± SD||0.94 ± 0.67||0.95 ± 0.65||0.94 ± 0.69||0.96|
|ACPA positive, % (no.)||88.7 (125)||100 (56)||81.2 (69)||0.0002|
|Current smoker, % (no.)||5.7 (8)||7.1 (4)||4.7 (4)||0.71|
|Framingham risk score, mean ± SD||10.5 ± 5.1||9.1 ± 5.0||11.5 ± 5.0||0.006|
|hsCRP level, mean ± SD mg/liter||4.8 ± 7.9||4.5 ± 6.5||5.1 ± 8.8||0.66|
The mean ± SD BMI for the total sample was 27.1 ± 6.0 kg/m2 (Table 2), and 28.4% were obese by BMI. The mean BMI for men was significantly higher than that for women (28.6 versus 26.2 kg/m2; P = 0.02), and more men than women were obese by BMI (41.1% versus 20.0%; P = 0.008). Overall, 33.3% (n = 47) met the WC obesity criterion, but there was no difference in the proportion of men and women classified as obese by WC.
|BMI, mean ± SD kg/m2||27.1 ± 6.0||28.6 ± 6.6||26.2 ± 5.4||0.02|
|Obese by BMI ≥30 kg/m2, % (no.)||28.4 (40)||41.1 (23)||20.0 (17)||0.008|
|Obese by waist circumference, % (no.)||33.3 (47)||35.7 (20)||31.8 (27)||0.71|
|Total fat from DXA, mean ± SD %||40.3 ± 8.7||39.9 ± 9.5||40.5 ± 8.2||0.68|
|Trunk fat, mean ± SD %†||20.6 ± 5.4||21.0 ± 5.3||20.4 ± 5.5||0.53|
|Appendicular fat, mean ± SD %‡||18.4 ± 4.7||17.7 ± 5.7||18.9 ± 4.0||0.17|
|Appendicular lean mass, mean ± SD %§||25.2 ± 4.1||26.0 ± 4.3||24.7 ± 3.8||0.08|
|Obese by DXA, % (no.)||58.2 (82)||80.4 (45)||43.5 (37)||< 0.0001|
There were no significant sex differences in total percent fat (39.9% for men versus 40.5% for women; P = 0.68) or percent truncal fat. More than one-half of the total sample (n = 82 [58.2%]) met the DXA criterion for obesity, with significantly more men than women classified as obese (80.4% versus 43.5%; P < 0.0001).
ROC analyses identified new sex-specific obesity definitions for each anthropometric measure, using DXA-defined obesity as the criterion. In each case, the new criteria were lower and more subjects were classified as obese. In addition, correspondence between DXA-defined obesity and anthropometric measures improved for each revised definition, although some improvements were slight.
Revised BMI obesity definitions were ≥24.7 kg/m2 for men and ≥26.1 kg/m2 for women (Table 3). For men, the sensitivity and specificity of the original BMI definition of obesity were 0.47 and 0.82, respectively; the revised BMI definition produced a sensitivity of 0.73 and a specificity of 0.73. For women, the original BMI definition yielded a sensitivity and specificity of 0.46 and 1.00, respectively; the revised BMI definition produced a sensitivity of 0.76 and a specificity of 0.85. Using the revised obesity criteria, 73.2% of men and 81.2% of women were correctly classified.
|Cut point||Classified as obese, %||Correctly classified, %||Sensitivity||Specificity||PPV||NPV|
For WC, the revised obesity definition for men was ≥96 cm, which yielded a sensitivity of 0.50 and a specificity of 0.73 (in contrast to 0.41 and 0.82 for the original definition, respectively). The revised WC obesity definition for women was ≥83 cm, which yielded a sensitivity of 0.78 and a specificity of 0.77 (compared to 0.57 and 0.87 for the original definition, respectively). Using the revised WC criteria, 53.6% of men and 76.5% of women were correctly classified compared to DXA classifications.
For men, total percent fat was significantly associated with higher pain ratings, greater disease activity by the RADAI, and greater fatigue after adjustment for age, glucocorticoid use, disease duration, and smoking (Table 4). Percent truncal fat was also associated with pain rating and fatigue. However, when the DXA obesity cut point was applied to percent fat, no significant associations were noted between obesity and disease status.
|Total fat % (total fat mass/ total mass)†||Obese by DXA||Trunk fat % (trunk fat mass/ total mass)†||Original criteria||New criteria||Original criteria||New criteria|
|Pain rating (range 0–10)||0.8 (0.2, 1.4)‡||0.9 (−0.7, 2.4)||1.2 (0.1, 2.3)§||1.4 (0.02, 2.6)§||1.3 (0.1, 2.7)§||0.8 (−0.4, 2.1)||0.7 (−0.6, 1.9)|
|RADAI score||0.6 (0.1, 1.1)§||0.5 (−0.7, 1.7)||0.8 (−0.1, 1.7)||1.2 (0.2, 2.2)§||1.0 (0.0, 2.0)§||0.6 (−0.4, 1.6)||0.4 (−0.6, 1.4)|
|Fatigue||1.1 (0.4, 1.7)‡||0.7 (−1.0, 2.5)||1.6 (0.3, 2.9)§||1.5 (0.02, 2.9)§||1.3 (−0.1, 2.8)||0.7 (−0.7, 2.2)||0.8 (−0.6, 2.2)|
|HAQ||0.1 (−0.04, 0.3)||−0.2 (−0.7, 0.2)||0.2 (−0.1, 0.5)||0.4 (0.05, 0.8)§||0.2 (−0.2, 0.5)||0.2 (−0.2, 0.6)||0.2 (−0.1, 0.6)|
|ESR||1.2 (−4.5, 6.9)||1.8 (−11.4, 15.1)||4.4 (−5.6, 14.4)||11.6 (0.9, 22.4)§||1.9 (−9.1, 12.9)||9.6 (−1.0, 20.3)||10.1 (−0.2, 20.4)|
|CV risk factors|
|Framingham risk score||0.1 (−0.5, 0.7)||0.8 (−0.6, 2.2)||0.7 (−0.3, 1.8)||1.8 (0.7, 2.9)‡||1.1 (−0.1, 2.3)||1.4 (0.3, 2.5)§||1.8 (0.7, 2.8)‡|
|hsCRP (log)||0.0 (−0.3, 0.3)||−0.5 (−1.0, 0.1)||0.2 (−0.3, 0.6)||0.4 (−0.1, 0.8)||−0.3 (−0.8, 0.2)||0.4 (−0.1, 0.9)||0.4 (−0.1, 0.9)|
|Pain rating (range 0–10)||0.9 (0.3, 1.4)‡||1.3 (0.4, 2.2)‡||1.5 (0.7, 2.2)‡||2.1 (1.0, 3.2)‡||0.9 (−0.03, 1.9)||1.5 (0.6, 2.4)‡||1.0 (0.1, 1.9)§|
|RADAI score||0.7 (0.2, 1.1)‡||0.9 (0.1, 1.6)§||1.2 (0.5, 1.8)‡||1.7 (0.8, 2.6)‡||0.6 (−0.2, 1.4)||1.2 (0.4, 1.9)‡||0.9 (0.1, 1.6)§|
|Fatigue||0.9 (0.3, 1.5)‡||1.2 (0.3, 2.2)§||1.7 (0.9, 2.5)‡||2.4 (1.3, 3.6)‡||1.4 (0.4, 2.4)‡||2.0 (1.2, 2.8)‡||1.5 (0.6, 2.5)‡|
|HAQ||0.1 (−0.1, 0.3)||0.1 (−0.1, 0.4)||0.2 (−0.1, 0.5)||0.2 (−0.2, 0.5)||−0.1 (−0.3, 0.2)||0.3 (0.03, 0.6)§||0.1 (−0.2, 0.4)|
|ESR||3.3 (−1.8, 8.4)||4.0 (−4.5, 12.5)||7.5 (0.1, 14.9)§||11.6 (0.6, 21.5)§||0.4 (−8.5, 9.2)||11.0 (2.3, 19.6)‡||6.3 (−2.1, 14.7)|
|CV risk factors|
|Framingham risk score||0.7 (−0.02, 1.4)||1.2 (0.1, 2.4)§||1.3 (0.3, 2.3)§||1.8 (0.4, 3.2)‡||0.5 (−0.7, 1.7)||1.8 (0.7, 3.0)‡||1.5 (0.3, 2.6)§|
|hsCRP (log)||0.3 (0.1, 0.5)§||0.3 (−0.1, 0.6)||0.4 (0.1, 0.8)§||0.5 (0.1, 1.0)§||0.2 (−0.2, 0.6)||0.3 (−0.1, 0.7)||0.3 (−0.1, 0.7)|
Among women, greater total percent fat and greater percent truncal fat were each significantly associated with higher pain ratings, higher RADAI scores, and greater fatigue. Women who were obese by DXA had significantly greater pain, RADAI scores, and fatigue.
DXA-derived body fat measures were not associated with Framingham risk scores or hsCRP levels for men. Among women, both higher total percent fat and higher percent truncal fat were associated with higher levels of hsCRP. Obesity by DXA and higher percent truncal fat were also associated with higher Framingham risk scores among women.
We examined associations of the same disease status measures with the original and revised BMI and WC obesity definitions. Obesity defined by both the original and revised BMI definitions was associated with significantly greater pain ratings and higher RADAI scores for men, and significantly greater fatigue ratings for women (Table 4). Obesity by the original, but not revised, BMI definition was associated with higher fatigue ratings, HAQ scores, ESRs, and Framingham risk scores for men, and with higher pain ratings, RADAI scores, ESRs, Framingham risk scores, and hsCRP levels for women.
No disease status measure was associated with WC-defined obesity for men, but central obesity by both the original and revised WC definitions was significantly associated with higher Framingham risk scores for men. For women, however, both the original and revised WC obesity definitions were associated with greater pain, higher RADAI scores, and greater fatigue; higher HAQ scores and elevated ESRs were associated with WC obesity by the original definition only. Central obesity defined by both the original and revised WC criteria was associated with elevated Framingham risk scores.
A large portion of this RA sample was obese, including more than one-quarter using the standard BMI definition, one-third using the standard WC definitions, and more than one-half using percent fat from DXA. These rates are similar to those reported in other RA samples, e.g., Giles et al reported that 33% of women and 36% of men with RA were obese by BMI and 57% were obese by DXA (5), and a UK study reported a BMI-defined prevalence of obesity of 31% (37).
Differences between rates of obesity in men and women were striking. Using both BMI and DXA criteria, twice as many men as women were classified as obese. DXA results showed that 80% of men in this RA sample were overfat. Of particular note, there was no significant difference in the total percent fat between men and women in our sample, a finding that is at odds with population studies (38, 39). Other published studies have noted differences in adiposity between men and women with RA. For example, Giles et al found that abdominal visceral adiposity was 51% higher in men with RA compared to men without RA, whereas there was no difference between women with and without RA (10). On the other hand, women with RA had more subcutaneous abdominal fat. Stavropoulos-Kalinoglou et al also noted relatively higher fat distributions in men with RA compared to men without RA (4). Men with RA had total fat percentages 49% higher than controls, while total percent fat for women with RA was 19% higher than controls. For truncal fat, men with RA were 43% higher than controls, while women with RA were 23% higher than controls. However, none of these studies reported disparities in the prevalence of obesity between men and women with RA of the size that we found. Additional research with larger samples will be needed to clarify this unexpected finding.
We identified much lower BMI criteria to define obesity, as well as different criteria for men and women (≥24.7 kg/m2 for men and ≥26.1 kg/m2 for women). Stavropoulos-Kalinoglou et al previously suggested that the BMI cut point for obesity be reduced by 2 kg/m2 (i.e., to 28 kg/m2) for individuals with RA; our analyses support cut points that are even lower. Other investigators have also noted that the current obesity cut point of a BMI ≥30 kg/m2 in non-RA populations is too high, has low sensitivity to detect adiposity in the general population, and is not appropriate for specific ethnic groups, and have identified alternate BMI obesity cut points very similar to those we identified, ranging from 25–25.8 kg/m2 (40–45).
Relationships between DXA body fat measures and RA disease status measures existed for both men and women. Greater total percent fat was associated with higher pain ratings, higher disease activity ratings, and greater fatigue for men and women. Greater truncal fat was also associated with higher pain ratings and greater fatigue for both men and women. Framingham risk scores were associated with DXA-defined obesity and truncal fat for women, but not for men. Among women, inflammatory biomarkers were also associated with percent truncal fat; a similar association was not noted for men. There were few nonobese men, however, which may have limited our ability to find these associations, although Giles et al also noted different patterns of the association of fat with CRP level for men and women (22); adiposity was significantly associated with CRP level for women, but not for men, which is similar to our findings.
We expected the associations between anthropometric obesity proxies and RA disease status and CV risk measures to parallel those seen with measures of body fatness obtained through DXA, and there were important consistencies as well as inconsistencies. BMI appeared to function satisfactorily as a proxy for DXA-derived obesity in these analyses. Men and women who met the standard BMI obesity criterion had significantly greater pain, disease activity, fatigue, and ESRs. Only a few associations were noted with the revised BMI criterion, calling its usefulness into question.
The revised WC cut points improved correspondence with DXA over the original WC cut points, but only marginally. Substantial differences existed between men and women in the associations between central obesity and RA disease measures. There were no significant differences in RA disease measures for men using either WC criterion. In contrast, women who met the original WC obesity criterion exhibited worse RA status on all measures, and differences remained for pain, the RADAI, and fatigue when the revised WC criterion was used.
The relationship between WC and CV risk was more robust. Both men and women who met either WC obesity criterion had significantly higher Framingham risk scores. The revised WC definitions we derived are similar to those proposed by the International Diabetes Federation (46) for whites at low risk for metabolic syndrome (94 cm for men and 80 cm for women). Further research is needed to establish the usefulness of WC as a proxy of obesity in RA and the appropriate criteria to define obesity.
Our revised obesity criteria for both BMI and WC improved sensitivity to detect obesity over the traditional definitions, but decreased specificity. Since BMI and WC might be used as screening tests to identify individuals with potentially harmful levels of body fat, a condition treatable by fairly benign means but associated with a heightened risk of a number of poor health outcomes, the tradeoff of high sensitivity for lower specificity, particularly when specificity is still at an acceptable level, seems appropriate.
This study has several limitations. The sample was relatively small, and we may have lacked statistical power in some cases. These analyses present cross-sectional associations; no causal attributions can be made. Clearly, these analyses need to be repeated with larger, longitudinal samples. The range of disease severity may have limited our findings. Few participants had very active RA, but the restricted range of disease activity should have biased our findings toward the null. Population-based studies have demonstrated racial and ethnic differences in the correspondence of BMI with body fatness. Our sample was primarily white, which limits the extent to which our results, particularly the revised obesity definitions, can be extended to other groups; at the same time, however, our more homogenous sample limited the variability due to racial or ethnic differences. We have only measures of total truncal fat, and were not able to differentiate between subcutaneous and visceral fat. Sex differences in the preponderance of visceral fat have previously been noted in an RA sample, with men having greater visceral fat, and visceral fat is, in turn, most strongly linked to CV events and outcomes (6–8). Glucocorticoid use has often been linked to body composition; cumulative prednisone dose has been linked to visceral fat (10, 47). While we were able to control for glucocorticoid use during the year prior to data collection, we do not have a good estimate of cumulative glucocorticoid use.
Our results do not provide a clear picture of the mechanism whereby obesity might be associated with worse RA disease or CV risk, although we can make a proposal based on analyses of the inflammatory biomarker results. Elevated ESR is often linked to active or severe RA; likewise, elevated CRP level is associated with a longitudinal risk of CV disease. It seems reasonable to project that greater amounts of adipose tissue are a source of inflammation, which then may lead to heightened disease activity and CV risk (3, 22, 48–50). This mechanism has been proposed by several investigators and our cross-sectional associations provide supportive evidence, but future longitudinal studies are needed to further elucidate these pathways. While a paradoxical relationship has been found in which obesity is associated with less severe joint damage in RA, other health effects of obesity in RA appear to parallel those seen in the general population, i.e., increased CV risk, greater functional limitations, and worse disease status.
These findings provide further evidence of a high prevalence of obesity or overfatness among individuals with RA. Sex differences existed in this sample, with men having higher rates of obesity, which may place them at a particularly high risk of CV disease and events. Findings from this study indicate that consideration of separate BMI obesity criteria for men and women may be warranted, although this proposal should be confirmed in other samples. The revised BMI and WC cut points identified for women with RA are very similar to those identified for women with systemic lupus erythematosus using the same methodology (BMI: 26.8 kg/m2, WC: 84.75 cm) (51). Nonetheless, these revised obesity criteria may be overly stringent; those suggested by Stavropoulos-Kalinoglou et al (BMI ≥28 kg/m2) (4) may be more realistic. However, these results do underscore the issue of importance of considering overfatness in RA.
Future research is needed to determine if the relationships noted in these 2 rheumatic conditions are unique, if lower obesity criteria should be considered for connective tissue diseases in general, or if these lower cut points are reflective of changes in body composition in the general population. Regardless, these results suggest that use of more stringent criteria for proxy measures such as BMI among individuals with RA, as has been previously suggested, may be warranted, since more stringent methods appear to identify risk for poor RA outcomes as well as heightened CV risk. Furthermore, anthropometric measures may provide proxy estimates of body composition that are as effective in identifying risk for poor RA and CV outcomes as DXA, but are much less costly and can easily be implemented in clinical settings.
All authors were involved in drafting the article or revising it critically for important intellectual content, and all authors approved the final version to be published. Dr. Katz had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study conception and design. Katz, Criswell.
Acquisition of data. Katz, Trupin, Criswell, Yelin.
Analysis and interpretation of data. Katz, Yazdany, Trupin, Schmajuk, Margaretten, Barton, Yelin.
We gratefully acknowledge the important contributions of Sandi Kaplan, Holly Wing, and Rachel Diskin, who conducted all of the study assessments.