To explore the associations between measures of body composition derived from computed tomography (CT) of the thigh and functional outcomes in patients with rheumatoid arthritis (RA).
To explore the associations between measures of body composition derived from computed tomography (CT) of the thigh and functional outcomes in patients with rheumatoid arthritis (RA).
Patients with RA underwent bilateral midfemoral quantitative CT for measurement of thigh fat area (TFA), thigh muscle area (TMA), and thigh muscle density (TMD). The associations of thigh-composition measures with disability and physical performance, as measured with the Health Assessment Questionnaire (HAQ), the Valued Life Activities (VLAs), and the Short Physical Performance Battery (SPPB) instruments, were explored in the total cohort and in the cohort subgrouped by sex, controlling for pertinent demographic, lifestyle, and RA disease and treatment covariates.
A total of 152 RA patients were studied. Among the potential determinants of TMD, older age, longer duration of sedentary activity, longer duration of RA, higher tender joint count, higher serum interleukin-6 levels, use of glucocorticoids, and nonuse of hydroxychloroquine were all significantly associated with lower TMD in multivariable models. RA characteristics accounted for 63% of the explainable variability in TMD. When comodeled, higher TFA and lower TMD, but not lower TMA, were significantly and independently associated with higher HAQ scores, lower Short Form 36 health survey physical functioning scores, lower composite SPPB scores, and a greater proportion of affected obligatory VLAs.
Thigh CT–derived measures of body composition, particularly fat area and muscle density, were strongly associated with disability and physical performance in RA patients, with RA disease features as potential determinants. Efforts to reduce fat and improve muscle quality may reduce disability in this population with impaired physical functioning.
Rheumatoid arthritis (RA) is a systemic inflammatory disorder that affects 1–2% of adults and frequently leads to progressive joint deformity and disability. This disability is costly, with the direct medical cost of RA disability amounting to between $9,000 and $19,000 per year (1), and the indirect costs, such as loss of work, compounding the expense (2). In addition, greater disability in RA patients portends poorer health outcomes, including higher all-cause and cardiovascular mortality rates (3).
Contributors to disability in RA go beyond articular swelling, tenderness, and deformity. Several of the major extraarticular determinants include mood (4), sleep (5), and the more recently studied body composition. RA patients, on average, have lower body cell mass (6), lower lean mass (7), and higher fat mass (7, 8) as compared to otherwise similar non-RA controls. Lower lean mass and higher fat mass were independently associated with higher levels of disability in RA patients (9). However, distinct from the amount of muscle, the density of muscle may also affect physical functioning. Low muscle density, reflecting reduced muscle contractile units accompanied by fatty replacement (10, 11), was associated with aging, deconditioning, and disuse in the general population (12, 13). As these factors are features of the RA disease state, it is conceivable that low muscle density, in addition to—or even independent of—low muscle mass, may contribute to physical dysfunction in RA patients.
Computed tomography (CT) scanning at the level of the mid-thigh, with quantification of the areas and densities of fat, muscle, and bone, is a validated, reproducible technique for directly assessing whole and regional body composition (14–16). Studies of CT-assessed muscle density in the general population have shown it to be a suitable determinant of muscle quality, in terms of prediction of strength (17, 18), physical performance (18–20), and incidence of mobility limitations (21, 22). However, to date, there have been no studies using CT to assess muscle area or density in the RA population. Most studies have used indirect measures, such as bioelectrical impedance, total body dual x-ray absorptiometry (DXA), or potassium counting.
Therefore, in the present study, we explored the associations of RA disease characteristics with thigh muscle density (TMD), thigh muscle area (TMA), and thigh fat area (TFA), as assessed with quantitative CT. We hypothesized that low thigh muscle density would be associated with disability and low physical performance scores, independently of muscle and fat area.
Study subjects were participants in the Evaluation of Subclinical Cardiovascular Disease and Predictors of Events in Rheumatoid Arthritis (ESCAPE RA) study, a cohort study investigating the prevalence, progression, and risk factors for subclinical cardiovascular disease in RA (23). A total of 197 RA patients completed the baseline study visit, and all of them met the American College of Rheumatology 1987 classification criteria for RA (24), were 45–84 years of age, and did not report having any of the prespecified prior cardiovascular events or procedures. The study was approved by the Institutional Review Board of the Johns Hopkins Hospital. Thigh CT was performed at the third visit for the ESCAPE RA study, for which a total of 158 patients (80%) returned. This third study visit occurred an average of 39 ± 4 (±SD) months postbaseline.
Study participants underwent bilateral mid-thigh CT on the same Aquilion 64-slice CT scanner (Toshiba America Medical Systems). A single transverse section was obtained midway between the greater trochanter and the femoral condyles, localized using a coronal scout film. The scans were analyzed with BonAlyse software, which quantifies areas and densities of muscle, fat, and bone in operator-selected regions. To exclude artifacts and objects outside the patient, the outline of the thigh was traced manually by a technician (HRK) who was blinded with regard to the characteristics of the patients. TFA was defined as the quantity of tissue between −190 and −30 Hounsfield units and included both subcutaneous fat tissue and fat tissue intercalated between muscle tissues. Marrow-associated fat tissue was excluded from the TFA measurement.
Participants also underwent total body DXA on a Lunar Prodigy DXA scanner (GE/Lunar) to measure total and regional fat and lean mass. Anthropometric parameters (height, weight, body mass index [BMI], and waist and hip circumferences) were determined as previously described (25). The BMI was calculated as the body weight in kilograms divided by the square of the height in meters (kg/m2).
We assessed functional outcomes concurrently with the body composition assessments. The Stanford Health Assessment Questionnaire (HAQ) (26) and the physical functioning domain of the Short Form 36 (SF-36) health survey (27) were used to assess disability related to common activities. The HAQ has a range of 0–3, with higher HAQ scores indicating greater disability. The SF-36 physical functioning scores range from 0 to 100, with higher scores indicating less disability.
Physical performance was assessed using the extended Short Physical Performance Battery (SPPB) (28). The SPPB includes assessments of balance (side-by-side, semi-tandem, and one-leg stands), strength (single and repeated chair stands), and gait and endurance (usual and fast paced 6-meter and 400-meter walk tests). Scoring range is 0–4 based on completion/time for each of the 5 testing domains, for a total possible score of 20.
Disability in a broader range of activities, including activities outside of the fundamental physical actions assessed with HAQ, was assessed with the Valued Life Activities (VLA) questionnaire (29). This 29-item instrument assesses limitations in the patient's participation in obligatory activities (i.e., self-care, ambulation, transit; 4 items), committed activities (i.e., housework, preparing meals, shopping, care of family members; 10 items), and discretionary activities (i.e., traveling, leisure activities, recreation, socializing; 15 items) as a consequence of RA. Activities not engaged in by, or those unimportant to, the participant do not count toward the score.
Age, sex, race/ethnicity, and history of current and past smoking were assessed by self-report. Forty-four joints were examined for swelling and tenderness by a single trained assessor, and RA disease activity was calculated using the Disease Activity Score in 28 joints using the C-reactive protein level (DAS28-CRP) (30). Radiographs of the hands and feet were obtained and scored using the van der Heijde modification of the Sharp method (SHS) (31) by a single experienced reader who was blinded with regard to the patient's characteristics.
Physical activity was assessed with the 7-Day Physical Activity Recall Questionnaire (32), and the weekly total of physical activity for intentional exercise activities (moderate or brisk walking for exercise, and moderate or vigorous individual or team sports and conditioning activities) was calculated for each participant. The duration of television watching, a measure of sedentariness, was also assessed by self-report. Depressive symptoms were assessed using the Center for Epidemiologic Studies Depression Scale (CES-D) (33). Fatigue was assessed using the Functional Assessment of Chronic Illness Therapy (FACIT) questionnaire (34). Thyroid disease was classified based on the use of thyroid replacement therapy. Current and past use of glucocorticoids, biologic disease-modifying antirheumatic drugs (DMARDs), and non-biologic DMARDs was determined with the use of detailed examiner-administered questionnaires.
Blood samples were collected while the patient was in a fasting state, and the samples were stored at −80°C. CRP levels were measured by nephelometry and interleukin-6 (IL-6) by chemiluminescent enzyme immunoassay at the Laboratory for Clinical Biochemistry Research (University of Vermont, Burlington, VT). Rheumatoid factor (RF) was assessed by enzyme-linked immunosorbent assay (ELISA), with seropositivity defined as ≥40 units. Anti–cyclic citrullinated peptide (anti-CCP) antibody was assessed by ELISA, with seropositivity defined as ≥60 units. Exon 2 of HLA–DR1 was sequenced for shared epitope alleles as previously described (35).
The distributions of all variables were examined. Univariate linear regression models were constructed to explore the associations of thigh composition outcomes with participant characteristics included as covariates, with calculations of beta coefficients, 95% confidence intervals (95% CIs), and their associated P values. Where required, variables were transformed to normality. To identify indicators of thigh fat and muscle measures, multivariable models were constructed with covariates carried over from univariate models that were significant at P ≤ 0.20 (to allow for residual confounding). More-parsimonious multivariable models were constructed by excluding covariates with the weakest associations with the outcome, with the impact of excluding each covariate tested using Akaike's Information Criterion for nested models. The adjusted coefficient of variability (R2) was used to estimate the total proportion of the variability in the outcome predicted by the modeled covariates. Variance inflation factors (VIFs) were calculated to ensure that variables with excessive collinearity were not modeled simultaneously.
To investigate the association of thigh fat and muscle measures with functional outcomes, multivariable models were constructed with functional measures modeled as outcomes and thigh composition measures modeled as exposures of interest. We considered covariates associated with both the thigh composition measures of interest and the functional outcomes to be confounding covariates. More-parsimonious multivariable models were constructed as described above. Spearman's correlation coefficients were calculated for correlations of anthropometric and DXA-derived body composition measures with functional outcomes as compared to correlations of thigh composition measures. All statistical calculations were performed using Intercooled Stata 10 (StataCorp). A two-tailed alpha value of 0.05 was used throughout.
A total of 152 participants had thigh CTs suitable for body composition analysis. Characteristics of the 152 RA patients are summarized in Table 1. Participants were middle-aged or older (range 47–84 years; mean age 63 years), predominantly female (65%) and Caucasian (88%). On average, the cohort comprised patients with disease of longer duration (median 12 years) with current disease activity in the low-to-moderate range, although high disease activity (DAS28-CRP >5.1) was observed in 17 patients (11%). As a whole, the cohort reported mild-to-moderate disability (median HAQ score 0.69), achieved submaximal scores on physical performance testing (mean SPPB score 12 of a possible 20), and reported a median of 31% of their VLAs as being affected by their RA.
|Characteristic||Total (n = 152)||Women (n = 98)||Men (n = 54)||P|
|Demographic and clinical features|
|Age, mean ± SD years (range 47–84)||63 ± 8||63 ± 8||62 ± 9||0.74|
|Caucasian, no. (%)||133 (88)||84 (86)||49 (91)||0.37|
|Any college, no. (%)||118 (78)||76 (78)||42 (78)||0.97|
|Exercise, minutes/day||29 (9–75)||26 (6–62)||51 (9–112)||0.028|
|TV watching, hours/day||2 (1–3)||2 (1–3)||2 (1–3)||0.66|
|CES-D score||5 (3–10)||6 (3–11)||5 (2–9)||0.063|
|Hormone replacement (women only), no. (%)||9 (9)||9 (9)||NA||–|
|Ever smoker, no. (%)||81 (53)||43 (44)||38 (70)||0.002|
|Current smoker, no. (%)||12 (8)||5 (5)||7 (13)||0.085|
|Thyroid disease, no. (%)||26 (17)||20 (20)||6 (11)||0.14|
|Diabetes mellitus, no. (%)||11 (7)||7 (7)||4 (7)||0.97|
|RA-specific clinical features|
|RA duration, years||12 (8–20)||13 (7–20)||10 (9–18)||0.60|
|RF or anti-CCP seropositivity, no. (%)||113 (75)||74 (76)||39 (72)||0.58|
|Any shared epitope alleles, no. (%)||105 (70)||66 (68)||39 (74)||0.48|
|DAS28-CRP||3.1 (2.3–4.1)||3.3 (2.7–4.3)||2.5 (1.8–3.4)||<0.001|
|CRP, mg/liter||2.4 (0.9–6.2)||2.6 (1.1–7.4)||2.3 (0.8–4.9)||0.11|
|IL-6, pg/ml||4.3 (2.5–9.6)||4.9 (2.5–10.4)||3.8 (2.6–7.3)||0.37|
|Health status assessments|
|Total SHS||13 (3–53)||20 (4–59)||10 (2–29)||0.16|
|Pain, by 100-mm VAS||20 (10–50)||30 (10–50)||10 (10–40)||0.005|
|HAQ score (range 0–3)||0.69 (0.13–1.38)||1.0 (0.38–1.38)||0.19 (0–1.12)||<0.001|
|SF-36 physical functioning score||64 ± 26||61 ± 25||71 ± 26||0.015|
|Physical Performance Battery Score (range 0–20)||12 ± 5||11 ± 5||14 ± 4||0.002|
|Proportion of VLAs Affected||0.31 (0.05–0.71)||0.52 (0.17–0.75)||0.09 (0–0.50)||<0.001|
|Obligatory VLAs||0.20 (0–0.60)||0.20 (0–0.75)||0 (0–0.40)||0.006|
|Committed VLAs||0.50 (0–0.86)||0.60 (0.29–1.0)||0.07 (0–0.67)||<0.001|
|Discretionary VLAs||0.29 (0–0.69)||0.50 (0.13–0.75)||0.09 (0–0.43)||<0.001|
|Medication use, no. (%)|
|Current prednisone||38 (25)||25 (26)||13 (24)||0.82|
|Current nonbiologic DMARDs||130 (86)||88 (91)||42 (78)||0.028|
|Methotrexate||106 (70)||74 (76)||32 (59)||0.028|
|Hydroxychloroquine||29 (19)||20 (21)||9 (17)||0.56|
|Current biologic DMARDs||70 (46)||54 (56)||16 (30)||0.002|
|TNF inhibitors||53 (35)||40 (41)||13 (24)||0.034|
|Thigh CT–derived measures|
|Total thigh area, cm2||201 ± 63||204 ± 70||196 ± 48||0.38|
|Thigh muscle area, cm2||78 ± 34||65 ± 24||100 ± 36||<0.001|
|Thigh muscle density, mg/cm3||39 ± 5||39 ± 6||40 ± 4||0.23|
|Thigh fat area, cm2||124 ± 51||139 ± 54||96 ± 26||<0.001|
|Weight, kg||79 ± 18||73 ± 16||90 ± 16||<0.001|
|Height, meters||1.66 ± 0.10||1.61 ± 0.06||1.74 ± 0.09||<0.001|
|Body mass index, kg/m2||28.5 ± 5.3||28.9 ± 5.7||29.4 ± 4.5||0.084|
|Waist circumference, cm||96 ± 16||91 ± 16||104 ± 13||<0.001|
|Waist-to-hip ratio||0.91 ± 0.10||0.86 ± 0.07||1.01 ± 0.05||<0.001|
|Total body fat, kg||31 ± 11||32 ± 12||29 ± 9||0.12|
|Fat mass index, kg/m2||11.2 ± 4.0||12.1 ± 4.2||9.5 ± 3.0||<0.001|
|Body fat percentage||39 ± 9||43 ± 8||32 ± 7||<0.001|
|Total body lean, kg||44 ± 12||38 ± 6||55 ± 10||<0.001|
|Fat-free mass index, kg/m2||15.9 ± 3.1||14.6 ± 2.0||18.3 ± 3.1||<0.001|
|Trunk fat, kg||16 ± 6||16 ± 6||17 ± 6||0.18|
|Appendicular fat, kg||13 ± 6||15 ± 6||11 ± 4||<0.001|
|Appendicular lean, kg||19 ± 6||16 ± 3||25 ± 4||<0.001|
Univariate analyses of the clinical indicators of thigh fat and muscle measures were determined (data available upon request from the author). Characteristics that yielded at least marginally significant associations in those analyses were carried into multivariable models (Table 2). When comodeled, older age, female sex, longer RA duration, higher tender joint count, and current prednisone use were all significantly and inversely associated with the TMA, while height was positively associated. Together, these 6 characteristics accounted for 47% of the variability in the TMA (adjusted R2 = 0.474), with the RA disease-related characteristics accounting for nearly 20% of the explainable variability (Table 2, TMA model 2). The other characteristics with significant or marginal associations with TMA in univariate models were no longer significantly associated in the multivariable model, and their exclusion did not meaningfully reduce the predictive ability of the model.
|Model 1||Model 2|
|Thigh muscle area (cm2)|
|Age, per year||−0.98||0.001||−1.13||<0.001|
|Height, per meter||56.3||0.086||66.6||0.022|
|Exercise, per 30 minutes||0.62||0.55||–||–|
|CES-D score, per unit||−0.11||0.77||–||–|
|RA duration, per year||−0.49||0.067||−0.59||0.006|
|Any SE alleles||−4.48||0.38||–||–|
|log IL-6, per log unit||−2.36||0.30||–||–|
|Square root swollen joint count||−0.21||0.93||–||–|
|Square root tender joint count||−3.92||0.031||−3.19||0.018|
|log total SHS||−0.89||0.62||–||–|
|Pain (by VAS), per unit||0.12||0.26||–||–|
|Current prednisone use||−15.4||0.007||−15.8||0.001|
|Total adjusted R2||0.473||–||0.474||–|
|R2 for RA characteristics||0.076 (16% of total)||–||0.090 (19% of total)||–|
|Thigh muscle density (mg/cm3)|
|Age, per year||−0.12||0.003||−0.13||<0.001|
|Height, per meter||7.62||0.098||6.90||0.024|
|Exercise, per 30 minutes||−0.003||0.99||–||–|
|TV watching, per 30 minutes||−0.12||0.21||–||–|
|CES-D score, per unit||0.007||0.90||–||–|
|RA duration, per year||−0.04||0.25||−0.06||0.060|
|RF or anti-CCP||0.72||0.34||–||–|
|Any SE alleles||−0.50||0.50||–||–|
|log IL-6, per log unit||−0.96||0.003||−0.92||0.001|
|Square root swollen joint count||0.15||0.66||–||–|
|Square root tender joint count||−0.39||0.11||−0.33||0.040|
|log total SHS||0.008||0.97||–||–|
|Pain (by VAS), per unit||−0.001||0.94||–||–|
|Current prednisone use||−3.72||<0.001||−3.48||<0.001|
|Current HCQ use||1.69||0.035||1.67||0.017|
|Total adjusted R2||0.381||–||0.414||–|
|R2 for RA characteristics||0.210 (55% of total)||–||0.260 (63% of total)||–|
|Thigh fat area (cm2)|
|Height, per meter||74.7||0.17||95.9||0.059|
|Exercise, per 30 minutes||0.57||0.75||–||–|
|RA duration, per year||0.72||0.12||0.76||0.046|
|log IL-6, per log unit||−1.58||0.70||–||–|
|Square root swollen joint count||2.38||0.58||–||–|
|Square root tender joint count||0.24||0.94||–||–|
|log total SHS||1.84||0.55||–||–|
|Pain (by VAS), per unit||0.09||0.62||–||–|
|Current prednisone use||−28.3||0.003||−23.4||0.006|
|Current MTX use||5.71||0.51||–||–|
|Current biologic agent||9.07||0.29||–||–|
|Total adjusted R2||0.201||–||0.222||–|
|R2 for RA characteristics||0.040 (20% of total)||–||0.051 (23% of total)||–|
For the TMD, all of the clinical indicators of the TMA except sex were also associated in the final adjusted model (Table 2, TMD model 2). In addition, higher IL-6 levels were significantly associated with a lower mean adjusted TMD; while participants prescribed hydroxychloroquine (HCQ) had a significantly higher mean adjusted TMD compared to nonusers. Together, the 8 characteristics modeled accounted for 41% of the variability in TMD, with 63% of the explainable variability deriving from RA-related characteristics. For the TFA, female sex, greater height, and longer RA duration were associated with a higher mean adjusted TFA, while the mean adjusted TFA was 23 cm2 lower in current prednisone users compared to nonusers (P = 0.006) (Table 2, TFA model 2). Together, these 4 characteristics accounted for only 20% of the variability in TFA, with RA-related characteristics accounting for 23% of the explainable variability.
TMD was positively correlated with TMA (Spearman's rho = 0.75, P < 0.001) but was not strongly correlated with the TFA (Spearman's rho = 0.13, P = 0.13). TMA and TFA were also uncorrelated (Spearman's rho = 0.06, P = 0.50). Despite a moderate-to-high correlation, the TMD and TMA were not sufficiently collinear to refute comodeling (mean VIF 2.40).
Crude and adjusted associations of the TFA, TMA, and TMD with scores on the HAQ, SF-36 physical functioning, SPPB, and FACIT instruments are summarized in Table 3. For each of the outcomes, higher TFA and lower TMA were significantly associated with higher reported disability and limitation in performance when only these two covariates were comodeled (model 1). Adding TMD to the model (model 2) supplanted nearly the entirety of the association of TMA with the outcomes and eliminated the statistical significance of the remaining TMA associations. In models including potential confounders (model 3), the magnitudes of the associations of TFA and TMD with the outcomes were reduced by 25–75%; however, these associations remained robust in the final models that included only the confounders that showed significant associations with the outcomes (model 4).
|Model 1||Model 2||Model 3||Model 4|
|TFA, per 10 cm2||0.044||<0.001||0.050||<0.001||0.050||<0.001||0.042||<0.001|
|TMA, per 10 cm2||−0.072||<0.001||−0.011||0.62||0.018||0.50||–||–|
|TMD, per mg/cm3||–||–||−0.062||<0.001||−0.047||0.014||−0.041||<0.001|
|Age, per year||–||–||–||–||−0.002||0.74||–||–|
|Height, per meter||–||–||–||–||−1.915||0.003||−1.13||0.018|
|RA duration, per year||–||–||–||–||0.001||0.95||–||–|
|log IL-6, per log unit||–||–||–||–||0.052||0.24||–||–|
|Square root TJC||–||–||–||–||0.122||<0.001||0.112||<0.001|
|Current prednisone use||–||–||–||–||0.217||0.056||0.223||0.044|
|Current HCQ use||–||–||–||–||−0.087||0.44||–||–|
|SF-36 physical functioning|
|TFA, per 10 cm2||−1.37||<0.001||−1.68||<0.001||−1.52||0.001||−1.35||<0.001|
|TMA, per 10 cm2||2.16||<0.001||−0.72||0.40||−1.87||0.093||–||–|
|TMD, per mg/cm3||–||–||2.93||<0.001||2.32||0.004||2.77||<0.001|
|Age, per year||–||–||–||–||−0.06||0.83||–||–|
|Height, per meter||–||–||–||–||68.0||0.010||54.3||0.014|
|RA duration, per year||–||–||–||–||−0.11||0.59||–||–|
|log IL-6, per log unit||–||–||–||–||−1.68||0.36||–||–|
|Square root TJC||–||–||–||–||−4.39||<0.001||−4.21||<0.001|
|Current prednisone use||–||–||–||–||−9.10||0.051||−8.99||0.046|
|Current HCQ use||–||–||–||–||0.97||0.83||–||–|
|Short Physical Performance Battery|
|TFA, per 10 cm2||−0.33||<0.001||−0.38||<0.001||−0.32||<0.001||−0.35||<0.001|
|TMA, per 10 cm2||0.58||<0.001||0.05||0.72||−0.14||0.59||–||–|
|TMD, per mg/cm3||–||–||0.54||<0.001||0.36||0.007||0.43||<0.001|
|Age, per year||–||–||–||–||−0.15||<0.001||−0.15||<0.001|
|Height, per meter||–||–||–||–||8.10||0.067||–||–|
|RA duration, per year||–||–||–||–||−0.02||0.48||–||–|
|log IL-6, per log unit||–||–||–||–||−0.62||0.047||−0.62||0.045|
|Square root TJC||–||–||–||–||−0.37||0.080||–||–|
|Current prednisone use||–||–||–||–||−1.11||0.16||–||–|
|Current HCQ use||–||–||–||–||0.15||0.85||–||–|
|Square root FACIT score|
|TFA, per 10 cm2||0.068||0.002||0.077||0.001||0.048||0.065||0.056||0.009|
|TMA, per 10 cm2||−0.096||0.004||−0.017||0.74||0.075||0.25||–||–|
|TMD, per mg/cm3||–||–||−0.081||0.044||−0.084||0.070||−0.064||0.011|
|Age, per year||–||–||–||–||−0.012||0.40||–||–|
|Height, per meter||–||–||–||–||−1.658||0.28||–||–|
|RA duration, per year||–||–||–||–||0.004||0.76||–||–|
|log IL-6, per log unit||–||–||–||–||0.070||0.52||–||–|
|Square root TJC||–||–||–||–||0.296||<0.001||0.324||<0.001|
|Current prednisone use||–||–||–||–||0.369||0.18||–||–|
|Current HCQ use||–||–||–||–||−0.011||0.97||–||–|
Accordingly, in the final adjusted models, each 10 cm2 higher TFA was associated with a 0.042 unit higher HAQ score, a 1.35 unit lower SF-36 physical functioning score, a 0.35 unit lower SPPB score, and a 0.056 unit higher FACIT score (P < 0.01 for each comparison), while each mg/cm3 increase in the TMD was associated with a 0.041 unit lower HAQ score, a 2.77 unit higher SF-36 physical functioning score, a 0.43 unit higher SPPB score, and a 0.064 unit lower FACIT score (P < 0.05 for each comparison). The adjusted associations of the TFA and TMD with functional outcomes did not significantly differ by sex, and there was no statistical evidence for interaction between the TFA and TMD on any of the functional outcomes (data not shown).
Similar results were observed for the associations of the TFA and TMD with the proportion of obligatory VLAs affected (Table 4). Specifically, on average, each 10 cm2 higher TFA was associated with a 1.28 percentage point higher adjusted proportion of obligatory VLAs affected (P = 0.011), while each mg/cm3 higher TMD was associated with a 1.07 percentage point lower adjusted proportion (P = 0.024) (Table 4, obligatory VLA model 4). The TFA, but not the TMA or TMD, was significantly associated with the proportion of committed and discretionary VLAs affected, with each 10 cm2 higher TFA being associated with a 1.73 and 1.32 percentage point higher adjusted proportion of VLAs affected, respectively (P < 0.05 for each comparison) (Table 4, committed and discretionary VLA model 4).
|Model 1||Model 2||Model 3||Model 4|
|Proportion of obligatory VLAs affected|
|TFA, per 10 cm2||1.45||0.007||1.40||0.009||1.66||0.008||1.28||0.011|
|TMA, per 10 cm2||−2.67||0.001||−1.69||0.062||0.01||0.78||–||–|
|TMD, per mg/cm3||–||–||−1.28||0.010||−1.19||0.021||−1.07||0.024|
|Height, per meter||–||–||–||–||−5.68||0.17||–||–|
|Age, per year||–||–||–||–||0.59||0.082||0.55||0.079|
|RA duration, per year||–||–||–||–||−0.17||0.55||–||–|
|log IL-6, per log unit||–||–||–||–||1.30||0.60||–||–|
|Square root TJC||–||–||–||–||7.18||<0.001||7.14||<0.001|
|Current prednisone use||–||–||–||–||22.39||<0.001||21.90||<0.001|
|Current HCQ use||–||–||–||–||−6.49||0.31||–||–|
|Proportion of committed VLAs affected|
|TFA, per 10 cm2||2.34||<0.001||2.32||<0.001||1.81||0.011||1.73||0.002|
|TMA, per 10 cm2||−3.17||<0.001||−2.70||0.008||−0.31||0.81||−1.40||0.15|
|TMD, per mg/cm3||–||–||−0.61||0.32||−0.25||0.68||−0.38||0.52|
|Height, per meter||–||–||–||–||−6.33||0.12||–||–|
|Age, per year||–||–||–||–||0.37||0.34||–||–|
|RA duration, per year||–||–||–||–||0.10||0.75||–||–|
|log IL-6, per log unit||–||–||–||–||1.13||0.23||–||–|
|Square root TJC||–||–||–||–||8.51||<0.001||8.96||<0.001|
|Current prednisone use||–||–||–||–||4.46||0.46||–||–|
|Current HCQ use||–||–||–||–||−8.99||0.22||–||–|
|Proportion of discretionary VLAs affected|
|TFA, per 10 cm2||1.80||0.001||1.77||0.001||1.46||0.030||1.32||0.015|
|TMA, per 10 cm2||−2.69||0.001||−2.31||0.014||−0.37||0.77||−1.28||0.17|
|TMD, per mg/cm3||–||–||−0.49||0.39||−0.25||0.68||−0.35||0.52|
|Height, per meter||–||–||–||–||−3.21||0.39||–||–|
|Age, per year||–||–||–||–||0.29||0.42||–||–|
|RA duration, per year||–||–||–||–||0.14||0.62||–||–|
|log IL-6, per log unit||–||–||–||–||0.08||0.38||–||–|
|Square root TJC||–||–||–||–||7.08||<0.001||7.21||<0.001|
|Current prednisone use||–||–||–||–||4.02||0.48||–||–|
|Current HCQ use||–||–||–||–||−6.75||0.33||–||–|
We compared the strength of correlation of CT-derived thigh composition measures with functional outcomes against those of anthropometric and DXA-derived measures of body composition (Table 5). In general, anthropometric parameters were not significantly correlated with functional outcomes. In the few cases of significant correlation of anthropometric parameters with functional outcomes (e.g., the BMI with SF-36 physical functioning and the proportion of VLAs affected), the strength of the correlation was weaker than that for the CT-derived measures. Among the DXA-derived measures of fat, the strength of the correlation of the fat mass index, body fat percentage, and appendicular fat with the functional outcomes was comparable or slightly higher than that for the TFA. Appendicular lean mass was the strongest DXA-derived measure, with comparable levels of correlation with functional outcomes as the TMA or TMD.
|HAQ||SF-36 physical functioning||SPPB||Proportion of VLAs affected|
|Thigh CT–derived measures|
|Total thigh area, cm2||0.037||0.65||−0.022||0.79||−0.001||0.99||0.084||0.30|
|Thigh muscle area, cm2||−0.356||<0.001||0.285||<0.001||0.397||<0.001||−0.265||0.001|
|Thigh muscle density, mg/cm3||−0.353||0.373||<0.001||0.460||<0.001||−0.270||<0.001||<0.001|
|Thigh fat area, cm2||0.327||<0.001||−0.225||0.005||−0.291||<0.001||0.307||<0.001|
|Body mass index, kg/m2||0.117||0.14||−0.176||0.027||−0.132||0.10||0.199||0.012|
|Waist circumference, cm||0.031||0.70||−0.120||0.13||−0.107||0.18||0.117||0.14|
|Total body fat, kg||0.251||0.001||−0.247||0.002||−0.189||0.018||0.297||<0.001|
|Fat mass index, kg/m2||0.355||<0.001||−0.327||<0.001||−0.281||<0.001||0.381||<0.001|
|Body fat percentage||0.460||<0.001||−0.369||<0.001||−0.362||<0.001||0.443||<0.001|
|Total body lean, kg||−0.343||<0.001||0.228||0.004||0.271||<0.001||−0.259||0.001|
|Fat free mass index, kg/m2||−0.223||<0.001||0.097||0.228||0.136||0.089||−0.133||0.097|
|Trunk fat, kg||0.155||0.052||−0.188||0.018||−0.117||0.145||0.213||0.007|
|Appendicular fat, kg||0.318||<0.001||−0.271||<0.001||−0.259||0.001||0.345||<0.001|
|Appendicular lean, kg||−0.355||<0.001||0.254||0.001||0.311||<0.001||−0.272||<0.001|
We next explored the linearity of the associations of TMD and TFA with functional outcomes and the combined associations of TMD and TFA (Figure 1). For the HAQ, the change in HAQ score per quartile increase in the TMD or TFA was roughly linear in adjusted analyses (Figure 1A). The mean adjusted HAQ scores were 2-fold higher in the group with the lowest TMD (below the median of 39 mg/cm3) and the highest TFA (above the median of 113 cm2) as compared to the group with the highest TMD and the lowest TFA (1.18 versus 0.55 HAQ units; P < 0.001).
For the SPPB (Figure 1B), the mean adjusted scores were 79% higher for participants in the fourth quartile of the TMD as compared to those in the first quartile (14.3 versus 8.0 SPPB units; P < 0.001); however, the difference in the mean adjusted SPPB scores was the largest between the first and second quartiles of the TMD (8.0 versus 12.9 SPPB units; P < 0.01), and did not differ between the third and fourth quartiles. For the TFA, the mean adjusted SPPB scores were 25% higher for participants in the lowest TFA quartile as compared to those in the highest quartile (14.1 versus 9.8 SPPB units; P < 0.001). As with the TMD, the trend in SPPB scores across quartiles of TFA was not linear, as scores were not significantly lower until the third quartile of the TFA. The mean adjusted SPPB scores were 63% higher in the group with the highest TMD and lowest TFA as compared to the group with the lowest TMD and highest TFA (14.5 versus 8.9 SPPB units; P < 0.001).
For the proportion of obligatory VLAs affected (Figure 1C), per-quartile differences in the associations of the TMD and TFA with the outcome were roughly linear. The proportion of obligatory VLAs affected was 2.4-fold higher, on average, for the group with the lowest TMD and highest TFA as compared to the group with the highest TMD and lowest TFA (45% versus 19%; P = 0.002).
In this investigation, which, to our knowledge, is the first to use mid-thigh quantitative CT to measure body composition in RA patients, we observed that the TMD, more so than the TMA and independently of the TFA, was a strong indicator of functional outcomes and physical performance, with each mg/cm3 increase in the TMD being associated with a lower HAQ score, a lower proportion of valued life activities affected, and higher SF-36 physical functioning and SPPB scores. Moreover, in multivariable analyses, RA disease features accounted for approximately one-fourth of the explainable variability in the TMA and TFA and three-fourths of the explainable variability in the TMD.
In recent years, CT-derived muscle density has been established as a marker of muscle quality (11). Density is estimated from CT-derived attenuation coefficients, and low muscle density reflects increased myocellular lipid content and fatty infiltration of the muscle compartment (10). Low muscle density at various sites has been associated with adverse outcomes in several studies of non-RA patients, including frailty (36), reduced physical performance (18–20), increased incidence of mobility limitations (21, 22), increased risk of hip fracture (37), and increased risk of hospitalization (18). However, until now, there have been no studies assessing muscle density in the RA population, let alone associations with functional outcomes. The findings of the present study confirm the importance of this measure in the RA population. Notably, muscle density was a stronger indicator of functional outcomes and physical performance than muscle area, anthropometric parameters, or DXA-derived measures of lean body mass or fat mass, a finding concordant with that of other studies in the general population (18, 37). It is interesting to note that in these studies, in which the enrolled patients were generally older than our cohort by an average of about a decade, the average thigh muscle area and muscle density were similar to those of our RA cohort (21).
Inflammatory cytokines, such as IL-6 and tumor necrosis factor, have been associated with reduced muscle density in non-RA populations (38, 39). Early studies by Roubenoff and colleagues showing reduced body cell mass in RA patients implicated inflammatory cytokines, suggesting a cytokine-driven state of hypermetabolism as the cause (6). Inflammation likely exerts multiple, complex effects on muscle, such as reduced insulin/growth factor sensitivity and accelerated protein degradation. Inflammatory cytokines are clearly associated with muscle wasting in several pathologic states, such as cancer, heart failure, and sepsis (40). Indeed, in our study, the IL-6 level was independently and inversely associated with muscle density and muscle area. In addition, use of HCQ and use of prednisone were also indicators of thigh muscle area and density. While HCQ has not been studied in this context, it is plausible that its effect on muscle density could be mediated by potentiating the activity of lipolytic enzymes by reducing their lysosomal degradation (41, 42), which could, hypothetically, lead to reduced intramyocellular lipid. However, since the association of HCQ with muscle density could be confounded by factors related to the indication for prescribing the drug (i.e., milder disease, etc.), additional study is needed to explore this association.
It was also notable that both fat and muscle areas were lower in prednisone users in our study. There may be several mechanisms underlying this finding. For one, prednisone-treated patients undergo an apparent redistribution of fat from the periphery to the trunk (43), and for another, glucocorticoid therapy may induce muscle wasting and has been associated with low muscle area (44). These indicators are potential targets for intervention that may improve muscle quality and functional outcomes.
The strong, independent association of lower muscle density and higher fat area with limitation in physical performance and disability suggest that these outcomes might be improved by interventions designed to increase muscle density and/or decrease fat. For example, it has been shown that in a study of non-RA patients, muscle area was the same or greater in obese individuals compared to non-obese controls, but muscle density was lower (45). When obese individuals participated in a 16-week weight loss program, there was an increase in muscle density (46), and other studies have shown that changes in muscle density parallel changes in strength during resistance detraining and retraining (13). From a functional standpoint, resistance exercises have already been shown to be effective in reducing pain and fatigue scores, as well as increasing strength and certain measures of physical performance in the RA population (47–50). However, as no studies in RA patients have used CT to determine body composition, the effects of exercise on muscle density in RA are still unknown. Although the duration of intentional exercise and sedentary activities was associated with some thigh composition measures in univariate analyses, we did not observe strong independent associations in multivariable models. This could indicate a true lack of association, or it may be a reflection of the imprecision of self-reported physical activity among RA patients. A clinical trial using a physical activity intervention would be preferred in order to clarify the role of physical activity in altering body composition in RA patients as a means of improving function.
In addition to CT, we also investigated two other methods of determining body composition: DXA and anthropometry. In general, the anthropometry results were poorly correlated with the functional outcome measures and essentially did not add to the information obtained via CT. Some DXA-derived measures of body composition, however, showed associations with functional outcomes similar to those of the CT-derived measures, and it would appear that appendicular lean mass and appendicular fat mass are potentially reasonable surrogates (9). In clinical practice, neither CT nor DXA is yet practical for use in measuring body composition. It is possible that other anthropometric parameters, such as thigh circumference or caliper skinfold assessment of thigh fat, could be a feasible surrogate for clinical practice; however, these were not evaluated in our study.
In our study, both a lower TMD and a higher TFA were significantly associated with a risk of affected obligatory VLAs, but the TMD was not associated with committed or discretionary VLAs. Obligatory VLAs are those considered necessary for survival and self-sufficiency, including walking to get around, getting around one's community by car or public transportation, and taking care of one's basic needs, such as bathing, washing, getting dressed, or taking care of personal hygiene (51). That these types of activities appear to be more strongly influenced by thigh composition than either committed or discretionary VLAs is intriguing. Since our study population was older and generally had more longstanding RA, it is plausible that our subjects have learned ways to participate in these activities despite their physical performance limitations, or they have already adapted and eliminated these from consideration as being important life activities.
The data from our study are cross-sectional, making these findings hypothesis-generating. Both a longitudinal, prospective cohort study of body composition and, eventually, a randomized controlled trial for interventions would be needed to confirm these postulations. Aside from the cross-sectional nature of our investigation, there are additional limitations. First, our study population was older, with relatively longstanding RA, and while a broad distribution of RA characteristics was represented, most had low-to-moderate disease activity with only mild-to-moderate disability. Second, our CT scans were regional, at only the mid thigh, which may provide a more limited estimate of whole body composition. Analysis of cross-sections from multiple body regions would be a useful future study. Third, we did not include measures of muscle strength, which would have permitted a more complete assessment of muscle quality.
In summary, 3 main conclusions can be drawn from this investigation. First, CT of the mid-thigh is a promising technique for directly assessing regional body composition in RA patients, allowing the determination of muscle area, fat area, and muscle density. Second, RA disease features play an important role in thigh composition, accounting for a large proportion of the explainable variability. Third and perhaps most important, muscle density appears to be a stronger indicator of physical functioning than muscle area in RA patients, with associations that are independent of fat mass. Further studies are needed to determine whether interventions aimed at increasing muscle density will improve the disability and poor physical performance that are so characteristic of this population.
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. Giles 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. Fontaine, Bathon, Giles.
Acquisition of data. Kramer, Bathon, Giles.
Analysis and interpretation of data. Fontaine, Bathon, Giles.
We would like to thank the staff of the Johns Hopkins Bayview Medical Center General Clinical Research Center, the field center of the Baltimore Multi-Ethnic Study of Atherosclerosis (MESA) cohort, and the MESA Coordinating Center at the University of Washington, Seattle. We are indebted to the dedication and hard work of ESCAPE RA staff members Marilyn Towns, Michelle Jones, Patricia Jones, Marissa Hildebrandt, Shawn Franckowiak, and Brandy Miles, as well as to the participants in the ESCAPE RA study who graciously agreed to take part in this research. Drs. Uzma Haque, Clifton Bingham III, Carol Ziminski, Jill Ratain, Ira Fine, Joyce Kopicky-Burd, David McGinnis, Andrea Marx, Howard Hauptman, Achini Perera, Peter Holt, Alan Matsumoto, Megan Clowse, Gordon Lam, and others generously recommended their patients for this study. We would especially like to thank Dr. Luigi Ferrucci for providing the equipment and expertise required to analyze the thigh CT scans in the study.