Effect of body mass index on mortality and clinical status in rheumatoid arthritis

Authors


Abstract

Objective

To study the relative risk (RR) of all-cause and cause-specific mortality in rheumatoid arthritis (RA) associated with body mass index (BMI), and to quantify the clinical and outcome consequences of abnormal BMI.

Methods

We studied mortality in 24,535 patients over 12.3 years, dividing patients into 3 age groups, <50, 50–70, and >70 years and fit Cox regression models separately within each age stratum. We used BMI categories of <18.5 kg/m2 (underweight), 18.5 to <25 kg/m2 (normal weight, reference category), 25 to <30 kg/m2 (overweight), and ≥30 kg/m2 (obesity).

Results

BMI ≥30 kg/m2 was seen in 63–68% and underweight in ∼2%. Reduction in the RR (95% confidence interval [95% CI]) for all-cause (AC) and cardiovascular mortality was seen for overweight (AC 0.8 [95% CI 0.8, 0.9]) and obese groups (AC 0.8 [95% CI 0.7, 0.8]), with and without comorbidity adjustment. Underweight was associated with increased mortality risk (AC 1.9 [95% CI 1.7, 2.3]). By contrast, obesity produced profound changes in clinical variables. Compared with normal weight, the odds ratio in the obese group was 4.8 for diabetes mellitus, 3.4 for hypertension, 1.3 for myocardial infarction, 1.4 for joint replacement, and 1.9 for work disability. Total semiannual direct medical costs were $1,683 greater, annual household income $6,481 less, pain scores 1.1 units higher, Health Assessment Questionnaire 0.28 higher, and EuroQol utility 0.7 units lower in the obese.

Conclusion

Overweight and obesity reduce the RR of all-cause and cardiovascular mortality across different age groups and durations of RA. By contrast, overweight and obesity are associated with substantial increased risks of comorbidity, total joint replacement, greater pain, medical costs, and decreased quality of life.

INTRODUCTION

Body mass index (BMI) is an important factor in all rheumatic diseases, but particularly in rheumatoid arthritis (RA) because of the increased prevalence of cardiovascular disease (CVD) and cardiovascular mortality in RA (1–4), as well as the known association between BMI and cardiovascular risk factors (5, 6). In addition, obesity and overweight place an extra burden on arthritis patients with respect to functional status, pain, work disability, and medical costs (7–9).

BMI is calculated as weight in kilograms divided by height in meters squared and used to express weight adjusted for height (10). BMI is usually classified into 4 groups: underweight <18.5 kg/m2, normal weight 18.5–24.9 kg/m2, overweight 25–29.9 kg/m2, and obese ≥30 kg/m2 (11, 12). The relation between BMI and mortality is complex. BMI has a different effect depending on age, sex, and ethnicity, and the net effect of overweight and obesity on morbidity and mortality is difficult to quantify (6). BMI may influence cause-specific mortality in opposite directions depending on the cause and the degree of BMI deviation from normal (13).

In 2005, Escalante et al reported on 779 mostly Hispanic RA (56%) patients (14). Patients with a BMI >30 (obese) had the lowest mortality, and mortality was greater in each lower BMI category. The authors concluded that BMI had a paradoxical effect on mortality, an effect they thought was mediated in part by comorbidity. In a specific evaluation of cardiovascular mortality in RA, Kremers et al, in an RA incidence cohort in Rochester, Minnesota studied 603 residents ages >18 years who were first diagnosed between 1955 and 1994 (15). Subjects with a low BMI (<20) had an increased risk of cardiovascular mortality, but those with BMI >30 kg/m2 did not have increased risk. However, both the Escalante et al and Kremers et al studies had limited sample sizes.

By contrast, a 2007 population-based study evaluated 36,859 subjects (12,772 deaths) in the 3 combined National Health and Nutrition Examination Surveys (NHANES) (16), beginning with data from 1971 (13). They found that underweight was associated with increased all-cause mortality, overweight was associated with significantly decreased all-cause mortality, and that “obesity also was associated with increased mortality overall, primarily due to its association with CVD mortality.”

To illuminate the relation of BMI and mortality in RA, we used data from 24,535 RA patients participating in the National Data Bank for Rheumatic Diseases (NDB) (17) and examined cause-specific (cardiovascular, cancer, noncardiovascular/noncancer) and all-cause mortality. In addition, we quantified the effect of BMI status on comorbidity, symptoms, treatment, and direct medical costs.

Significance & Innovations

  • Overweight occurs in two-thirds of rheumatoid arthritis (RA) patients.

  • Overweight and obesity reduce the relative risk of all-cause and cardiovascular mortality across different age groups and durations of RA.

  • By contrast, overweight and obesity are associated with substantial increased risks of comorbidity, total joint replacement, greater pain, medical costs, and decreased quality of life.

PATIENTS AND METHODS

We studied 24,535 adult patients with RA who participated in the NDB longitudinal study of RA outcomes (17). Participants were volunteers, recruited from the practices of US rheumatologists, who completed mailed or internet questionnaires about their health at 6-month intervals between 1999 and 2011. The participants were not compensated for their participation. The diagnosis of RA was made by the patient's rheumatologist. The NDB utilizes an open cohort design in which patients are enrolled continuously.

At each assessment, we collected data on height and weight, treatment, utilization of services, and comorbid conditions (18) using a comorbidity score based on the presence of pulmonary disorders, myocardial infarction, other cardiovascular disorders, stroke, hypertension, diabetes mellitus, spine/hip/leg fracture, depression, gastrointestinal (GI) ulcer, other GI disorders, and cancer, as previously described (18). We also collected a measure of lung function, the Medical Research Council dyspnea scale (19), and visual analog scales for pain, global severity, fatigue, and sleep disturbance. Functional status and quality of life measures included the Health Assessment Questionnaire (HAQ) disability index (20), physical component summary (PCS) and mental component summary (MCS) scores from the Short Form 36 version (21, 22), and the EuroQol 5-domain quality of life utility measure (EQ-5D) (23).

Death data.

The NDB obtains death information, including cause of death, from families and physicians. Followup for mortality begins at entry into the cohort. In addition, all patients, whether they are known to be alive or not, are submitted to the US National Death Index (NDI) annually, as the NDI provides cause of death information (24, 25), and we used underlying cause of death data in our cause-specific analyses. The NDI receives all reports of deaths from the states, and represents the repository of all individual death information in the US. The NDI reports lag behind actual deaths. To account for this, we censored patients whose death status was not known as of December 31, 2008. For cause-specific analyses, we classified deaths into 3 categories: cardiovascular, cancer, and all other causes.

Height and weight were obtained by self-report standard. BMI was calculated as weight in kilograms divided by height in meters squared. As noted by Stommel and Schoenborn, “Self-reported BMI values tend to overestimate measured BMI values at the low end of the BMI scale (<22) and underestimate BMI values at the high end, particularly at values >28. The discrepancies also vary systematically with age (younger and older respondents underestimate their BMI more than respondents aged 42–55), gender, and the ethnic/racial background of the respondents” (26). To estimate the proportion of patients in each BMI category (Table 1), we provided original self-report and adjusted BMI data according to Stommel and Schoenborn's analyses to possibly narrow misclassification error (26). Except for this adjustment in Table 1, all other BMI data are shown as the original unadjusted self-reported data.

Table 1. Distribution of BMI and mortality risk according to age and BMI groups*
Age group, yearsUnderweight, BMI <18.5 kg/m2Normal, BMI 18.5–24.9 kg/m2Overweight, BMI 25.0–29.9 kg/m2Obese, BMI ≥30.0 kg/m
  • *

    Values are the relative risk, adjusted for sex (95% confidence interval) unless otherwise indicated. Adjusted body mass index (BMI) category was adjusted for self-reporting. Adjusted all-cause and cardiovascular (CV) mortality categories were adjusted for diabetes mellitus, hypertension, history of myocardial infarction, and current smoking.

  • Significant at P < 0.05.

Age <50, no. (%)152 (2.5)2,312 (37.2)1,657 (26.7)2,095 (33.7)
Age 50–70, no. (%)194 (1.5)3,905 (30.4)4,424 (34.4)4,328 (33.7)
Age >70, no. (%)211 (3.9)2,415 (44.2)1,858 (34.0)984 (18.0)
Total BMI, no. (%)557 (2.3)8,632 (35.2)7,939 (32.4)7,407 (30.2)
Adjusted BMI, no. (%)485 (2.0)7,332 (29.9)8,196 (33.4)8,520 (34.7)
All-cause mortality (n = 3,829 deaths)    
 Age <501.6 (0.7, 3.6)10.8 (0.5, 1.2)1.6 (1.1, 2.2)
 Age 50–702.2 (1.7, 2.9)10.8 (0.8, 1.0)1.0 (0.9, 1.8)
 Age >701.6 (1.3, 2.0)10.8 (0.7, 0.9)0.9 (0.7, 1.0)
 All ages1.9 (1.7, 2.3)10.8 (0.8, 0.9)0.8 (0.7, 0.8)
Adjusted all-cause mortality (n = 3,829 deaths)    
 Age <501.3 (0.6, 3.1)10.7 (0.5, 1.1)1.0 (0.7, 1.5)
 Age 50–702.1 (1.6, 2.8)10.9 (0.8, 1.0)0.9 (0.8, 1.0)
 Age >701.5 (1.2, 1.8)10.8 (0.7, 0.9)0.8 (0.7, 0.9)
 All ages1.9 (1.6, 2.2)10.8 (0.7, 0.8)0.6 (0.5, 0.6)
CV mortality (n = 1,166 deaths)    
 Age <501.0 (0.1, 7.9)10.5 (0.2, 1.4)1.6 (0.8, 3.1)
 Age 50–702.8 (1.7, 4.8)11.0 (0.8, 1.2)1.1 (0.9, 1.4)
 Age >701.5 (1.0, 2.1)10.9 (0.7, 1.0)1.0 (0.8, 1.2)
 All ages2.0 (1.5, 2.7)10.8 (0.7, 1.0)0.8 (0.7, 0.9)
Adjusted CV mortality (n = 1,166 deaths)    
 Age <500.8 (0.1, 6.5)10.5 (0.2, 1.2)1.1 (0.5, 2.4)
 Age 50–702.8 (1.7, 4.7)10.9 (0.7, 1.2)0.9 (0.7, 1.1)
 Age >701.4 (1.0, 2.0)10.8 (0.7, 1.0)0.9 (0.7, 1.1)
 All ages1.9 (1.4, 2.5)10.8 (0.7, 0.9)0.6 (0.5, 0.7)
Cancer mortality (n = 763 deaths)    
 Age <500.9 (0.1, 7.2)10.9 (0.4, 1.9)0.9 (0.4, 2.0)
 Age 50–701.3 (0.7, 2.7)10.8 (0.6, 1.0)0.7 (0.5, 0.9)
 Age >701.7 (1.0, 2.8)11.0 (0.8, 1.3)1.1 (0.8, 1.5)
 All ages1.6 (1.1, 2.4)10.9 (0.8, 1.0)0.7 (0.6, 0.8)
Not CV, not cancer (n = 1,900 deaths)    
 Age <502.1 (0.8, 6.1)10.9 (0.5, 1.6)1.8 (1.2, 2.9)
 Age 50–702.4 (1.6, 3.5)10.9 (0.7, 1.1)1.2 (1.0, 1.4)
 Age >701.7 (1.3, 2.2)10.8 (0.7, 0.9)0.7 (0.6, 0.9)
 All ages2.1 (1.7, 2.6)10.8 (0.7, 0.9)0.8 (0.7, 0.9)

Statistical analysis.

Mortality analyses.

The dates and values of the study variables, except for mortality data, were obtained at the time of the completion of the first NDB questionnaire. We calculated relative risks (RRs) of death using Cox proportional hazard models (Table 1). Because the proportional hazards assumption was not met across age, we divided the data into 3 age groups, <50 years, 50–70 years, and >70 years, and fit cause-specific models separately within each age stratum. For all analyses, we used BMI categories of <18.5 (underweight), 18.5 to <25 (normal weight, reference category), 25 to <30 (overweight), and ≥30 (obese). All models included BMI categories and sex. Covariate adjusted models also included baseline presence or absence of diabetes mellitus, hypertension, history of myocardial infarction, and smoking status.

BMI-clinical variable analyses.

We used a randomly selected observation from the 24,535 RA patients in logistic and linear regression analyses adjusted for age and sex (Tables 2–4). For all analyses, we also calculated the average marginal effect predicted. Margins or marginal effects are adjusted predictions calculated from the previously fit model that include covariate effects. Marginal effects present the results of the analyses in an easy to understand original metric rather than as odds ratios (ORs) or coefficients. To avoid jargon, we characterize them as predicted proportion or value (95% confidence interval [95% CI]) in the Tables.

Table 2. Variables associated with BMI levels*
Variable and BMI categoryOR or coefficient (95% CI)Predicted proportion or value (95% CI)
  • *

    BMI = body mass index; OR = odds ratio; 95% CI = 95% confidence interval; MI = myocardial infarction.

  • Values are the OR and predicted probabilities adjusted for age and sex.

  • Not significant at P < 0.05.

  • §

    Values are the coefficients and predicted values adjusted for age and sex.

Diabetes mellitus  
 Underweight0.61 (0.38, 0.98)0.03 (0.02, 0.04)
 Normal weight1 (ref.)0.05 (0.04, 0.05)
 Overweight2.04 (1.80, 2.32)0.09 (0.09, 0.10)
 Obese4.79 (4.26, 5.39)0.19 (0.18, 0.20)
Hypertension  
 Underweight0.75 (0.61, 0.92)0.19 (0.16, 0.22)
 Normal weight10.23 (0.23, 0.24)
 Overweight1.61 (1.50, 1.73)0.33 (0.32, 0.34)
 Obese3.35 (3.12, 3.60)0.49 (0.48, 0.50)
MI (ever)  
 Underweight0.91 (0.65, 1.26)0.07 (0.05, 0.09)
 Normal weight10.07 (0.07, 0.08)
 Overweight1.16 (1.04, 1.30)0.08 (0.08, 0.09)
 Obese1.42 (1.26, 1.60)0.10 (0.09, 0.11)
Cancer (ever)  
 Underweight1.03 (0.83, 1.30)0.16 (0.13, 0.18)
 Normal weight10.15 (0.15, 0.16)
 Overweight1.04 (0.96, 1.14)0.16 (0.15, 0.17)
 Obese1.02 (0.93, 1.12)0.16 (0.15, 0.16)
Fracture  
 Underweight2.00 (1.20, 3.34)0.08 (0.05, 0.12)
 Normal weight10.04 (0.04, 0.05)
 Overweight1.02 (0.81, 1.28)0.04 (0.04, 0.05)
 Obese1.25 (1.00, 1.55)0.05 (0.05, 0.06)
Substance abuse  
 Underweight0.83 (0.11, 6.25)0.00 (0.00, 0.01)
 Normal weight10.00 (0.00, 0.00)
 Overweight1.04 (0.54, 1.97)0.00 (0.00, 0.00)
 Obese1.12 (0.60, 2.09)0.00 (0.00, 0.00)
Substance abuse ever  
 Underweight0.53 (0.29, 0.97)0.02 (0.01, 0.04)
 Normal weight10.04 (0.04, 0.05)
 Overweight1.02 (0.87, 1.20)0.04 (0.04, 0.05)
 Obese1.12 (0.96, 1.31)0.05 (0.04, 0.05)
Comorbidity index (range 0–9)§  
 Underweight0.25 (0.12, 0.37)1.70 (1.58, 1.83)
 Normal weight11.46 (1.43, 1.49)
 Overweight0.18 (0.14, 0.23)1.64 (1.61, 1.67)
 Obese0.66 (0.61, 0.70)2.12 (2.08, 2.15)
Total joint replacement  
 Underweight1.44 (1.18, 1.76)0.23 (0.20, 0.26)
 Normal weight10.17 (0.16, 0.18)
 Overweight1.05 (0.97, 1.14)0.18 (0.17, 0.19)
 Obese1.39 (1.28, 1.51)0.22 (0.21, 0.23)
Hospitalization  
 Underweight1.77 (1.42, 2.21)0.18 (0.15, 0.21)
 Normal weight10.11 (0.10, 0.11)
 Overweight1.06 (0.96, 1.17)0.11 (0.11, 0.12)
 Obese1.26 (1.14, 1.39)0.13 (0.12, 0.14)
Work disability  
 Underweight1.47 (1.16, 1.86)0.17 (0.13, 0.20)
 Normal weight10.12 (0.11, 0.13)
 Overweight1.18 (1.08, 1.30)0.14 (0.13, 0.15)
 Obese1.93 (1.77, 2.10)0.21 (0.20, 0.21)
Total 6-month medical costs, $§  
 Underweight613.07 (−232, 1,458)7,287 (6,464, 8,110)
 Normal weight06,674 (6,473, 6,875)
 Overweight373 (87, 659)7,048 (6,842, 7,253)
 Obese1,683 (1,336, 2,030)8,358 (8,079, 8,636)
Household income, $§  
 Underweight−6,308 (−8,739, −3,877)42,467 (40,108, 44,825)
 Normal weight048,775 (48,155, 49,395)
 Overweight−1,807 (−2,689, −926)46,967 (46,335, 47,600)
 Obese−6,481 (−7,374, −5,587)42,294 (41,660, 42,928)
Current smoker  
 Underweight1.54 (1.24, 1.91)0.22 (0.18, 0.25)
 Normal weight10.16 (0.15, 0.16)
 Overweight0.85 (0.77, 0.93)0.13 (0.13, 0.14)
 Obese0.76 (0.70, 0.83)0.12 (0.12, 0.13)
Table 3. Severity scales and BMI levels*
Variable and BMI categoryOR or coefficient (95% CI)Predicted proportion or value (95% CI)
  • *

    BMI = body mass index; OR = odds ratio; 95% CI = 95% confidence interval; HAQ = Health Assessment Questionnaire; MRC = Medical Research Council.

  • Values are the coefficients and predicted values adjusted for age and sex.

  • Values are the OR and predicted probabilities adjusted for age and sex.

HAQ disability (range 0–3)  
 Underweight0.30 (0.23, 0.36)1.26 (1.20, 1.33)
 Normal weight0 (ref.)0.96 (0.95, 0.98)
 Overweight0.06 (0.04, 0.08)1.02 (1.01, 1.04)
 Obese0.28 (0.26, 0.31)1.25 (1.23, 1.26)
Pain scale (range 0–10)  
 Underweight0.82 (0.58, 1.07)4.33 (4.09, 4.57)
 Normal weight03.50 (3.44, 3.56)
 Overweight0.33 (0.24, 0.41)3.83 (3.77, 3.89)
 Obese1.07 (0.98, 1.16)4.57 (4.51, 4.64)
Global severity scale (range 0–10)  
 Underweight0.70 (0.47, 0.93)4.03 (3.80, 4.25)
 Normal weight03.33 (3.27, 3.38)
 Overweight0.22 (0.15, 0.30)3.55 (3.50, 3.61)
 Obese0.97 (0.90, 1.05)4.30 (4.24, 4.36)
Fatigue scale (range 0–3)  
 Underweight0.90 (0.64, 1.16)5.03 (4.78, 5.28)
 Normal weight04.13 (4.07, 4.20)
 Overweight0.22 (0.13, 0.31)4.35 (4.29, 4.42)
 Obese0.99 (0.90, 1.09)5.13 (5.06, 5.20)
Sleep scale (range 0–10)  
 Underweight0.47 (0.19, 0.76)3.85 (3.58, 4.13)
 Normal weight03.38 (3.31, 3.45)
 Overweight0.37 (0.27, 0.46)3.75 (3.68, 3.82)
 Obese1.07 (0.97, 1.17)4.45 (4.38, 4.53)
Physical component scale  
 Underweight−4.05 (−5.01, −3.09)34.13 (33.20, 35.06)
 Normal weight038.18 (37.94, 38.43)
 Overweight−1.32 (−1.67, −0.98)36.86 (36.62, 37.10)
 Obese−5.31 (−5.65, −4.97)32.87 (32.64, 33.10)
Mental component scale  
 Underweight−1.90 (−2.91, −0.88)47.57 (46.58, 48.56)
 Normal weight049.47 (49.23, 49.71)
 Overweight−0.48 (−0.83, −0.14)48.98 (48.73, 49.23)
 Obese−2.16 (−2.52, −1.80)47.30 (47.03, 47.57)
EuroQol (range 0–1)  
 Underweight−0.07 (−0.10, −0.05)0.69 (0.66, 0.72)
 Normal weight00.76 (0.76, 0.77)
 Overweight−0.02 (−0.03, −0.01)0.74 (0.74, 0.75)
 Obese−0.07 (−0.08, −0.06)0.69 (0.69, 0.70)
Dyspnea MRC Class 3  
 Underweight1.61 (1.15, 2.26)0.23 (0.17, 0.29)
 Normal weight10.16 (0.14, 0.17)
 Overweight1.55 (1.37, 1.76)0.22 (0.21, 0.24)
 Obese3.67 (3.27, 4.12)0.40 (0.39, 0.42)
Table 4. Rheumatoid arthritis treatment variables and BMI levels*
Variable and BMI categoryOR or coefficient (95% CI)Predicted proportion or value (95% CI)
  • *

    BMI = body mass index; OR = odds ratio; 95% CI = 95% confidence interval.

  • Values are the OR and predicted probabilities adjusted for age and sex.

  • Not significant at P < 0.05.

Biologic therapy  
 Underweight0.90 (0.74, 1.11)0.45 (0.40, 0.50)
 Normal weight10.47 (0.46, 0.48)
 Overweight0.99 (0.92, 1.06)0.47 (0.46, 0.48)
 Obese1.01 (0.94, 1.08)0.47 (0.46, 0.49)
Prednisone  
 Underweight1.20 (1.01, 1.42)0.43 (0.39, 0.47)
 Normal weight10.38 (0.37, 0.39)
 Overweight0.97 (0.91, 1.03)0.38 (0.37, 0.39)
 Obese0.89 (0.84, 0.95)0.36 (0.35, 0.37)
Methotrexate  
 Underweight0.78 (0.66, 0.92)0.46 (0.42, 0.50)
 Normal weight10.52 (0.51, 0.53)
 Overweight1.07 (1.00, 1.13)0.54 (0.53, 0.55)
 Obese1.04 (0.97, 1.11)0.53 (0.52, 0.54)
Strong opioids  
 Underweight2.02 (1.21, 3.38)0.03 (0.02, 0.04)
 Normal weight10.02 (0.01, 0.02)
 Overweight0.97 (0.75, 1.26)0.01 (0.01, 0.02)
 Obese1.55 (1.23, 1.96)0.02 (0.02, 0.03)
Any strength of opioid  
 Underweight1.39 (1.15, 1.69)0.26 (0.22, 0.29)
 Normal weight10.20 (0.19, 0.21)
 Overweight1.17 (1.08, 1.26)0.22 (0.22, 0.23)
 Obese1.69 (1.57, 1.83)0.30 (0.28, 0.31)

We created running line smooths of the y on x variable at the baseline observation (Figures 1 and 2) (18). The smoother is a locally linear function of the predictors for each observation.

Figure 1.

The relation of body mass index (BMI) to age in women (left) and men (right) with rheumatoid arthritis at a random observation. The grey outline surrounding the point estimate line represents 95% confidence intervals. The turning point for each graph is at ∼50 years of age.

Figure 2.

Representation of the U-shaped and monotonic relationships between symptoms and body mass index (BMI) level. HAQ = Health Assessment Questionnaire.

Data were analyzed using Stata, version 12.0. P values less than 0.05 were considered significant, statistical significance was set at the P value less than 0.05 level by 2-tailed test, and confidence intervals were established at 95%.

RESULTS

At entry into the NDB, the mean ± SD age and duration of the 24,535 RA patients were 58.9 ± 13.2 years and 12.4 ± 10.9 years, respectively. Twenty-two percent of patients were men, and the mean ± SD HAQ score was 1.1 ± 0.72. During followup that extended up to 12.3 years, we recorded 3,829 deaths.

BMI and its effect on mortality.

The mean ± SD BMI at entry was 27.8 ± 6.5 units. The majority of patients were overweight (32.4%) or obese (30.2%) according to the US and World Health Organization classifications (11, 12). Underweight and normal-weight patients constituted 2.3% and 35.2% of the sample, respectively (Table 1). Values of BMI adjusted for possible self-reported assessment errors changed the proportions slightly, primarily by increasing the overweight and obese percentages. The adjusted percentages were underweight 2.0%, normal weight 29.9%, overweight 33.4%, and obese 34.7%.

For both men and women, BMI rose to its maximum at ∼50 years of age, as shown in Figure 1. However, peak BMI age in worldwide studies is said to be ∼60 years (5). To examine the simultaneous effect of age and BMI on mortality risk, we divided age at entry into 3 groups, age <50 years, age 50–70 years, and age >70 years.

Compared with those with normal BMI, Table 1 shows that underweight was a substantial risk factor for all-cause mortality when all age groups were considered simultaneously, with an overall RR of 1.9 (95% CI 1.7, 2.3). The RR of all-cause mortality was decreased overall in those who were overweight (RR 0.8 [95% CI 0.8, 0.9]) or obese (RR 0.8 [95% CI 0.7, 0.8]). However, among obese patients, the risk of all-cause mortality was increased (RR 1.6 [95% CI 1.1, 2.2]) in those <50 years of age at study entry, not different from an RR of 1 in those 50–70 years of age, and decreased in patients >70 years of age. Among overweight patients, all-cause mortality risk was decreased in the 50–70 and >70 years age groups, but was not significantly associated with all-cause mortality on those <50 years of age. When all-cause mortality was additionally adjusted for comorbidity, there were no substantial changes in mortality risks.

In general, cause-specific mortality followed the pattern of all-cause mortality for cardiovascular mortality, and the pattern of risk was maintained after additional adjustment for diabetes mellitus, hypertension, history of myocardial infarction, and smoking. However, cancer mortality was less common in obese patients <50 and 50–70 years of age than in the other cause-specific and all-cause groups. The risk of “not cardiovascular, not cancer” mortality was similar to the risk groups for all-cause mortality.

In sensitivity analyses, we examined whether duration of RA at study entry affected the risk of mortality by separately examining patients with 0–2, 2–10, and >10 years duration of RA. Compared with the RRs in the all-ages category of all-cause mortality of underweight 1.9, overweight 0.8, and obese 0.8, RRs for these groups for patients with <2 years RA duration were 2.1, 0.8, and 0.8, for 2–10 years of RA they were 1.8, 0.8, and 0.8, and for >10 years of RA were 2.0, 0.9, and 0.8, respectively. The same levels of statistical significance noted for the all ages analyses in Table 1 were seen in each of the duration group analyses. Similarly, subgroup analyses by nonsmokers, ever smokers, and current smokers did not change the results.

Association of BMI and clinical variables in RA.

Depending on the nature of the variable, clinical variables were associated with linear or sigmoid increases in BMI, as for example, with the risk of diabetes mellitus and hypertension (Table 2 and Figure 2, bottom panel), or with a U-shaped pattern of worse values for underweight as well as overweight and obese patients compared with normal weight individuals (Table 3 and Figure 2, top panel). For each of the 2 patterns, the greater the deviation from normal BMI levels, the greater the change in the study (y) variable.

The strength of associations in Table 2 may be roughly judged by examining the ORs and 95% CIs in the obese category, and the effect of the risk by examining the predicted outcomes. Among comorbidity and outcome variables (Table 2), the strongest associations were seen for hypertension (OR 3.4 [95% CI 3.1, 3.6]), diabetes mellitus (OR 4.8 [95% CI 4.3, 5.4]), and myocardial infarction (OR 1.4 [95% CI 1.3, 1.6]), associations that might be categorized as linear physiologic consequences of BMI. BMI was also an important factor in the risk of total joint replacement, with the marginal effect increasing from 17% in those with normal weight to 22% in the obese. Similarly, being disabled increased from 12–21%, total semiannual direct medical cost increased by $1,683, and annual household income was decreased by $6,481. Of interest, given the general increase in mortality in underweight individuals, we found no association between BMI and cancer. We also found an inverse association between current smoking and BMI, but only 14% of patients were smokers.

Variables that reflected or included abnormalities in functional status or pain, such as fracture, substance abuse, total joint replacement, medical cost, and disability, demonstrated the U-shaped response described above (Tables 2 and 3). Among the severity scales of Table 3, BMI was associated with substantial functional loss (HAQ; OR 0.28 [95% CI 0.26, 0.31]), increased pain (OR 1.1 [95% CI 1.0, 1.2]), fatigue (OR 1.0 [95% CI 0.9, 1.1]), sleep disturbance (OR 1.1 [95% CI 1.0, 1.2]), dyspnea (OR 3.7 [95% CI 3.3, 4.1]), and general quality of life PCS (OR −5.3 [95% CI −5.7, −5.0]) and EQ-5D (OR 0.07 [95% CI −0.08, −0.06]). In addition, underweight was also associated with severity scales, costs, loss of income, and work disability.

Obesity was also associated with use of strong opioids (OR 1.6 [95% CI 1.2, 2.0]) and any opioids (OR 1.7 [95% CI 1.6, 1.8]) (Table 4), but not with methotrexate or biologic therapy. At the study observation, RA treatment percentage was prednisone 37.5%, methotrexate 52.9%, biologic agents 37.7%, nonbiologic disease modifying antirheumatic drugs (DMARDs) 75.9%, and DMARDs or biologic agents 83.8%.

DISCUSSION

In our study of 24,535 individuals and 3,829 deaths in mostly non-Hispanic white (90%) RA patients over a maximum followup of 12.3 years, we found that obesity was common in RA, 30.2% in unadjusted and 34.7% in adjusted BMI estimates. Normal BMI was found in 35.2% and 29.9%, respectively. Underweight, a concern of a number of RA studies (14, 15), was noted in 2.3% and 2.0%. These rates are similar to those noted in the general US population for overweight (33.4%) and obesity (32.9%) (6, 27) and for underweight (1.5%) (28).

BMI overall was associated with an increased risk of all-cause mortality in underweight patients, but in the overweight (RR 0.8 [95% CI 0.8, 0.9]) and obese (RR 0.8 [95% CI 0.7, 0.8]), there was a decreased risk of all-cause mortality (Table 1). These findings, which were generally robust to age, duration, and covariate adjustment, were in accord with the 2005 study of Escalante et al concerning 779 mostly Hispanic (56%) patients (14). However, we did find that when age was taken into account, obese patients <50 years of age also had an increased mortality risk (RR 1.6 [95% CI 1.1, 2.2]) (Table 1). In addition, there was no mortality association in obese patients 50–70 years of age. To summarize, all-cause mortality risk in our study was increased in younger patients with elevated BMI levels, but was generally decreased in older (≥50 years of age) patients with elevated BMI levels. We are unaware of other large studies that have assessed the risk of all-cause mortality associated with RA.

With respect to cardiovascular mortality in RA, Kremers et al, in an RA incidence cohort in Rochester, Minnesota, studied 603 residents ages >18 years who were first diagnosed between 1955 and 1994 (15). Subjects with low BMI (<20) had an increased risk of cardiovascular mortality, but those with BMI >30 did not have increased risk. Both the Escalante et al (14) all-cause mortality and the Kremers et al cause-specific mortality studies had limited sample sizes. Even so, in our larger study, we found no difference in the BMI patterns of mortality risk between all-cause and cardiovascular mortality: increased risk in the underweight and decreased risk in the overweight and obese.

There is substantial controversy in the literature as to whether overweight and obesity are associated with increases or decreases in mortality. In an important population-based study, Flegal et al in 2007 evaluated 36,859 subjects (12,772 deaths) in the 3 combined NHANES studies that began in 1971 (13). They found that underweight was associated with increased all-cause mortality, overweight was associated with significantly decreased all-cause mortality, and that “obesity also was associated with increased mortality overall, primarily due to its association with CVD mortality.” Thus, our result in chronically ill RA patients disagrees with this population-based study in that we did not find an obesity-associated increase in mortality.

Another prospective study of BMI in relation to the risk of death from any cause utilized 527,265 US men and women in the National Institutes of Health-American Association of Retired Persons cohort who were 50–71 years of age at enrollment in 1995–1996 (29). The study participants constituted 18% of possible enrollees from 6 US states and 2 metropolitan areas. The study results showed an increased risk of all-cause mortality for the highest and lowest categories of BMI among both men and women, in all racial or ethnic groups, and at all ages.

There have been a very large number of studies of BMI and mortality. In a study germane to RA, where many patients are elderly, 14,833 subjects ≥75 years of age from 53 family practices in the UK were followed for up to 5.9 years, during which time 6,649 died (30). BMI was not associated with CVD, and BMI was negatively associated with mortality risk in both men and women. Results were essentially unchanged after adjustment for potential confounders. Similar results were obtained from a variety of other studies (30). Ogden et al (6) summarized the available data as follows, “Obesity is associated with a modestly increased risk of all-cause mortality. … The main cause of excess mortality in obesity, relative to normal weight, usually has been found to be cardio-vascular disease. A meta-analysis of 26 studies found that obesity (BMI ≥30) was associated with a RR for all-cause mortality of 1.22, for coronary heart disease mortality of 1.57, and for CVD mortality of 1.48, relative to the normal weight category” (31, 32).

However, the obesity-mortality relationship in some studies (33–35) reflect “a shift to the right in the mortality-BMI curve, implying that minimum mortality occurs at a higher BMI in older subjects than in younger subjects” (5). Ogden et al (6) indicated that the “association of BMI with mortality often has been found to be weak to nonexistent at older ages. …” and “that up to a certain point, higher weights may be associated with improved survival in patients with coronary artery disease so that the net effect of obesity on cardiovascular mortality may not be as great as the net effect of obesity on incident CVD.”

It is not clear why the risk of mortality in the obese should be reduced or at least not increased in RA as was seen in the current study and the 2 other RA studies (14, 15). It seems possible that a reduced association with obesity might be related to the general increase in CVDs that are seen in RA (2, 4, 36). One possible mechanism is that atherogenesis is more dependent on inflammation in RA than in the general population. However, other factors may influence study result differences in ways that may not be entirely clear. Such factors include definitions of BMI groups, comorbidity, age, sex, selection bias, and special populations (5).

While we did not observe that overweight and obesity increased all-cause and cardiovascular mortality in RA, we noted strong effects of BMI on comorbidity, costs, RA outcome, and symptom severity (Tables 2–4) that occurred in 2 patterns. The first pattern of a linear or sigmoid monotonic increase was found in variables in which BMI was associated with a physiologic effect, such as in the development of diabetes mellitus and hypertension and in the consequent increase in myocardial infarction. The second pattern, that of abnormal response in those whose BMI was either low or high compared with normal levels, was found in variables that reflected pain, opioid use, and other measures of distress. Clinically, our results seem to indicate that weight reduction in obesity will result in better overall clinical status and outcome and should be pursued. It is not known whether interventions to increase weight in the underweight will result in benefits, but it seems to be a reasonable approach in the absence of data to the contrary.

Our study has certain limitations. We only used baseline predictor data. It is possible that some results might have been more informative if longitudinal data were available in all patients. Second, death comes sooner to elderly patients, and that might be one reason for age-related differences. Left censoring could have played a role, although when we restricted analyses to patients within 2 years of RA onset, our results were unchanged (data not shown). Finally, our sample was one of generally white volunteers and was not a population-based sample. A very important issue, whether change in BMI for reasons other than illness would have affected mortality, could not be addressed in this study, but is important enough to be recommended for a followup study.

In summary, overweight and obesity reduce the RR of all-cause and cardiovascular mortality across different age groups and durations of RA. By contrast, overweight and obesity are associated with substantial increased risks of comorbidity, total joint replacement, greater pain, medical costs, and decreased quality of life.

AUTHOR CONTRIBUTIONS

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. Wolfe 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. Wolfe, Michaud.

Acquisition of data. Wolfe, Michaud.

Analysis and interpretation of data. Wolfe, Michaud.

Ancillary