Metabolic Syndrome, Adipokines, and Response to Advanced Therapies in Rheumatoid Arthritis
The opinions or assertions presented herein are the private views and opinions of the authors and do not represent the views of the Department of the Veterans Affairs.
Supported by the Corrona Research Foundation and by a Department of Veterans Affairs Clinical Science Research and Development Merit award (grant I01-CX-001703). Dr Baker's work is supported by a Rehabilitation Research and Development Merit award (I01-RX-003644) and a Veterans Affairs Clinical Science Research and Development Merit award (IK2-CX-001703). Dr. Mikuls’ work is supported by the US Department of Defense (PR200793), the National Institute of General Medical Sciences (U54-GM-115458), and the Department of Veterans Affairs Biomedical Laboratory Research and Development (BX006049 and BX005413). Dr Curtis's work is supported by the National Institute of Arthritis and Musculoskeletal and Skin Diseases, NIH (grant P30-AR-072583).
Additional supplementary information cited in this article can be found online in the Supporting Information section (https://acrjournals.onlinelibrary.wiley.com/doi/10.1002/art.43034).
Author disclosures are available at https://onlinelibrary.wiley.com/doi/10.1002/art.43034.
Abstract
Objective
We determined if metabolic syndrome, its components, and adipokines (adiponectin, leptin, and fibroblast growth factor-21) were associated with response to advanced therapies among patients with rheumatoid arthritis (RA).
Methods
This study included participants with RA initiating either tumor necrosis factor inhibitor (TNFi) or non-TNFi biologic therapies from the Comparative Effectiveness Registry to study Therapies for Arthritis and Inflammatory Conditions (CERTAIN) cohort within the CorEvitas registry. Metabolic syndrome was defined according to the National Cholesterol Education Program Adult Treatment Panel III definition. Adipokines were assessed on stored samples from a subsample of responders and nonresponders (n = 200). The primary outcome was the achievement of a change as large as the minimal clinically important difference (MCID) for the Clinical Disease Activity Index (CDAI) at 6 months.
Results
Among 2,368 participants, 687 (29%) had metabolic syndrome. Metabolic syndrome was associated with lower odds of achieving CDAI MCID (odds ratio [OR] 0.69, 95% confidence interval [CI] 0.56–0.86, P = 0.001) with a dose-dependent decrease in response rate according to the number of components present. Model fit was superior for metabolic syndrome compared with body mass index. Associations between metabolic syndrome and MCID achievement were similar between patients receiving TNFi (OR 0.65, 95% CI 0.49–0.87, P = 0.003) versus non-TNF therapies (OR 0.76, 95% CI 0.55–1.04, P = 0.08 [P for interaction = 0.49]). Adipokines were not associated with MCID achievement.
Conclusion
Metabolic syndrome is associated with lower response rates with the initiation of an advanced therapy in RA, with similar effects for both TNFi and non-TNFi agents. Adipokines were strongly associated with metabolic syndrome but were not associated with clinical response.
INTRODUCTION
In current practice in rheumatoid arthritis (RA), it is difficult for physicians to predict who will have refractory disease and demonstrate a poor response to the initiation of therapy. We recently demonstrated that patients with abnormally low weight (body mass index [BMI] < 18.5) as well as patients with obesity and RA were less likely to achieve a meaningful clinical response to biologic disease-modifying therapies from the CorEvitas clinical registry.1 In fact, a number of prior studies have identified reduced response rates among patients with BMI in the obese range.2-4
Although obesity and underweight have been associated with poor response to advanced therapies, our prior study found that BMI did not distinguish those that were more likely to benefit from a tumor necrosis factor inhibitor (TNFi) versus a non-TNFi advanced therapy. In other words, there was no evidence that one treatment would work better than the other in the context of obesity or underweight BMI ranges. In contrast, we recently showed that a low or normal BMI was associated with greater response to TNFi compared with conventional therapies, specifically “triple therapy,”5 suggesting that BMI may be informative as part of a precision medicine approach because it may differentially predict responses among different therapies in RA. Although BMI has been associated with clinical response to therapy in several studies, it is not a comprehensive measure of the metabolic consequences of obesity, and few RA studies to our knowledge have evaluated metabolic syndrome itself (central adiposity, hypertension, lipid abnormalities, and insulin resistance), its individual components, and their independent effects on disease activity.6-9
Adipokines are produced in adipose tissue and serve as metabolic regulators that are strongly associated with obesity and its metabolic complications. The systemic and immune effects of adipokines are not completely understood, but they have been shown to significantly influence the metabolic activities of tissues, including regulating insulin sensitivity.10 In addition, adipokines have been shown to play a role in regulating adaptive and innate immune cells and have been postulated to mediate the inflammatory features of obesity along with circulating cytokines, such as TNF and interleukin-6.11
In a previous study we showed that high levels of adiponectin and leptin predicted a lower likelihood of achieving low disease activity (LDA) over time in a clinical registry independent of BMI.12 Adipokines also differentiated those that were most likely to benefit from a TNFi over triple therapy in one study,5 although we are not aware of studies evaluating adipokines as predictors of response to one advanced therapy over another. These prior observations suggest that adipokines are informative of disease phenotype, although whether these associations are independent of metabolic syndrome is not known.
In this study, we aimed to determine if metabolic syndrome, and high levels of adiponectin, leptin, and fibroblast growth factor (FGF) 21 are associated with a lower likelihood of achieving a clinical response to therapy among patients with RA from the Comparative Effectiveness Registry to study Therapies for Arthritis and Inflammatory Conditions (CERTAIN) cohort. We also aimed to evaluate whether metabolic syndrome and adipokines can identify patients who are more likely to respond to a TNFi over a non-TNFi biologic.
PATIENTS AND METHODS
We used data from the CERTAIN comparative effectiveness study, which was a prospective, protocolized 12-month observational cohort study of adult patients and was conducted as an ancillary study and nested within the CorEvitas (formerly known as Corrona) RA registry.13, 14 Data were collected using structured case report forms that include medication use and dose, height, weight, RA clinical disease activity, function, comorbid illnesses, and adverse events.15 CERTAIN initiations occurred between November 2010 and April 2014. We included participants with a diagnosis of RA initiating either TNFi or non-TNFi biologic therapies within CERTAIN (JAK inhibitors were not included in CERTAIN).
CERTAIN used the CorEvitas network of participating private and academic sites to recruit patients fulfilling the 1987 American College of Rheumatology criteria that had at least moderate disease activity.16 Patients starting or switching biologic agents, either a TNFi or a non-TNFi biologic, were eligible for enrollment. Patient visits and all laboratory blood work, performed centrally in all patients, were mandated every 3 months for 1 year or until discontinuation of the initiated drug. Safety data were collected through 1 year and beyond.
This study was conducted in accordance with the Declaration of Helsinki, and all patients were required to provide written informed consent and authorization before participating. All participating investigators were required to obtain full institutional review board (IRB) approval for conducting noninterventional research involving human subjects. Sponsor approval and continuing review was obtained through a central IRB (the New England Independent Review Board, NEIRB No. 120160610). For academic investigative sites that did not receive a waiver to use the central IRB, full approval was obtained from the respective governing IRBs and documentation was submitted to CorEvitas, LLC, before initiating any study procedures. Patients were not involved in the design, conduct, reporting, or dissemination plans of this research. Data from the CorEvitas RA Registry are available from CorEvitas, LLC, through a commercial subscription agreement and are not publicly available.
Definition of metabolic syndrome
Waist circumference and blood pressure were measured as part of a standard clinical protocol at participating sites. Standard laboratory assessments (glucose and lipid profiles) were performed centrally on stored samples from enrollment (stored for <1 week). Metabolic syndrome was defined according to the National Cholesterol Education Program Adult Treatment Panel III definition, which requires that three of the following criteria be met: (1) waist circumference >40 inches (men) or >35 inches (women), (2) blood pressure >130/85 mm Hg or a diagnosis of hypertension, (3) fasting triglyceride level >150 mg/dl, (4) nonfasting high-density lipoprotein cholesterol level <40 mg/dL (men) or <50 mg/dL (women), and (5) fasting blood glucose level >100 mg/dL or a history of diabetes. Because the fasting glucose level was not available, we used a threshold of 140 mg/dL.17 For analyses that dichotomized metabolic syndrome, we included participants with missing data for one or more of the above if the missing data would not be expected to change the categorization (eg, they met the criteria already with only four of the components available). For BMI, we used the World Health Organization guidelines to categorize BMI as underweight (<18.5 kg/m2), normal weight (≥18.5–25 kg/m2), overweight (≥25–30 kg/m2), obese (≥30–35 kg/m2), and severely obese (≥35 kg/m2).
Adipokine assessments
Adipokines (adiponectin, leptin, and FGF-21) were measured on stored samples in 100 biologic-naive TNFi and 100 biologic-experienced non-TNFi users (powered to detect a moderate effect size [d] of 0.40 between responders and nonresponders). For each therapy, the best 50 responders and worst 50 nonresponders (based on the change in Clinical Disease Activity Index [CDAI]) of those who had remained on therapy at 6 months and had sufficient blood samples were selected. The CDAI minimal clinically important difference (MCID) was used to categorize initiators as responders or nonresponders. The relative decrease (or increase) was used to rank patients within responder/nonresponder groups. Samples were collected at enrollment, using a multianalyte panel from Meso Scale Discovery (Rockville, MD). Adipokine values were log-transformed and standardized so that a 1-unit difference in the value represented a 1 SD difference for all individual analytes. An adipokine score was also generated to reflect the number of adipokines above/below the median in a pattern consistent with metabolic obesity as previously described.5 Thus, the range is from 1 to 4, wherein 1 is the lowest score possible with one additional point given for each of the following: an adiponectin level below the median, a leptin level above the median, and/or an FGF-21 level above the median.
Response outcomes
The primary clinical outcome was the achievement of a change at least as large as the MCID for the CDAI at 6 months.18 The MCID cut points for improvement have been previously described as 12 for patients starting with a high CDAI, 6 for those in moderate CDAI, and 1 for those with low CDAI. The main secondary outcome was the achievement of LDA (CDAI ≤ 10). We also explored associations with the achievement of low individual components of disease activity including pain, patient global scores, evaluator global scores, tender joint counts, and swollen joint counts as defined by Boolean 2.0 criteria.19
For the primary analysis, we imputed or replaced the outcome as nonresponse for binary outcomes if the initiator switched to an alternative therapy before 6 months. This approach assumes that those who did not continue the therapy did not have a meaningful clinical response. If the initiator discontinued therapy but did not start another biologic therapy by 6 months, we used the actual response at 6 months.
Covariable assessments
Several covariables were collected at each visit. We evaluated factors prehypothesized to represent important predictors of response to therapy and included age; sex; race; comorbidities including an adapted Charlson Comorbidity Index, smoking, and CDAI; and the use of other RA therapies, including prednisone, prior biologic experience, and disease duration. Levels of C-reactive protein, anticitrullinated protein antibody (ACPA), and rheumatoid factor were available at baseline.
Statistical analysis
Characteristics of participants stratified by metabolic syndrome status were described and standardized differences presented to assess group differences. The primary analysis used mixed effect logistic regression models to evaluate the association between metabolic syndrome at baseline and the odds of achieving the primary outcome of a change at least as large as the MCID for CDAI. Random effects were used for patients who could contribute more than one initiation and for patients nested within provider. For initial models, we adjusted for prehypothesized factors (age, sex, race, baseline disease activity, and prior biologic experience) as well as characteristics that were significantly different at enrollment between those with and without metabolic syndrome and were not considered likely to be in the causal pathway (comorbidity score).
Secondary analyses evaluated individual components of the metabolic syndrome, evaluated the number of components of metabolic syndrome (to assess dose dependency), and explored model fit compared with BMI alone. Model fit was explored using the Akaike Information Criterion (AIC) and the Bayesian Information Criterion and by testing for additive effects through likelihood ratio testing.
To determine if the effect of metabolic syndrome on clinical response varied by therapeutic mechanism of action, the significance of multiplicative interaction terms between treatment (TNFi vs non-TNFi) and metabolic syndrome was evaluated in regression models. We also tested for interaction by sex and ACPA/rheumatoid factor status.
Associations between adipokines and clinical response were assessed in multivariable logistic regression models adjusting for age, sex, BMI, metabolic syndrome, TNFi (biologic naive) versus non-TNFi (biologic experienced), combination versus monotherapy use, ACPA serostatus, and baseline CDAI. We also explored testing the significance of multiplicative interactions terms between each adipokine and type of therapy (TNFi versus non-TNFi) within these models to assess whether adipokine levels were associated with a superior response to one therapy over another. Because all TNFi initiators were biologic naive and all non-TNFi initiators were bio-experienced within the adipokine subcohort, it was not possible to include both line of therapy and type of therapy in regression models because of collinearity. Analyses were performed using STATA version 18 (College Station, TX).
RESULTS
A total of 2,368 participants in CERTAIN were eligible for the analysis (Supplementary Figure 1; Table 1). Among these, 687 (29%) had metabolic syndrome at enrollment. The components that were most frequently present were a high waist circumference (53%) and high blood pressure (56%) (Supplementary Table 1). Among those with metabolic syndrome, most had three components (61%) (Supplementary Table 2).
| Characteristic | Metabolic syndrome | Standardized difference | |
|---|---|---|---|
| No | Yes | ||
| N | 1,681 | 687 | – |
| Age, mean (SD), years | 56.0 (0.33) | 57.4 (0.44) | −0.11 |
| Female, n (%) | 1,348 (80) | 325 (77) | 0.07 |
| White, n (%) | 1,386 (82) | 561 (82) | 0.02 |
| BMI, mean (SD), kg/m2 | 28.6 (0.16) | 34.5 (0.28) | −0.85 |
| BMI categories, n (%) | 0.91 | ||
| Underweight | 16 (1.0) | 2 (0.3) | – |
| Normal | 509 (30.3) | 41 (6.0) | – |
| Overweight | 607 (36.1) | 163 (23.7) | – |
| Obese I | 287 (17.1) | 191 (27.8) | – |
| Obese II | 262 (15.6) | 290 (42.2) | – |
| Smoking, n (%) | 16 (1.0) | – | – |
| Never | 792 (48) | 313 (46) | 0.06 |
| Past | 530 (32) | 239 (35) | |
| Current | 325 (20) | 127 (19) | |
| Comorbidity score, mean (SD)a | 1.18 (0.01) | 1.48 (0.03) | −0.51 |
| Cancer, n (%) | 97 (6) | 51 (7.4) | −0.07 |
| Diabetes, n (%) | 90 (5) | 292 (43) | −0.97 |
| Cardiovascular disease, n (%) | 129 (8) | 98 (14) | −0.21 |
| Disease duration, mean (SD), years | 8.33 (0.21) | 9.15 (0.37) | −0.09 |
| ACPA or RF positive, n (%) | 1,207 (72) | 487 (71) | 0.02 |
| CDAI, mean (SD) | 29.2 (0.32) | 30.3 (0.48) | −0.09 |
| Log(CRP), mean (SD) | 0.55 (0.63) | 0.70 (0.55) | −0.25 |
| CRP, mean (SD) | 10.1 (19.0) | 10.4 (15.2) | −0.02 |
| Biologic experienced, n (%) | 1,034 (62) | 453 (66) | −0.09 |
| TNFi initiation, n (%) | 986 (59) | 383 (56) | 0.06 |
| Non-TNFi initiation, n (%) | 695 (41) | 304 (44) | |
- * ACPA, anticitrullinated peptide antibody; BMI, body mass index; CDAI, Clinical Disease Activity Index; CRP, C-reactive protein; RF, rheumatoid factor; TNFi, tumor necrosis factor inhibitor.
- a Modified Charlson Comorbidity score.
Several covariates were significantly different between groups (P < 0.05). Participants with metabolic syndrome were slightly older, were more likely to have obesity, had higher overall comorbidity, had higher rates of diabetes and cardiovascular disease, had higher CDAI, had higher C-reactive protein levels, and were somewhat more likely to have previously received an advanced therapy (biologic experienced) (Table 1). Patients receiving a TNFi were also more likely to be using the therapy as a first-line advanced therapy (55%) compared with those receiving a non-TNFi (13%). Non-TNFi biologics included tocilizumab (38%), rituximab (15%), and abatacept (47%).
Metabolic syndrome was associated with a lower rate of MCID achievement as well as a lower rate of achieving LDA in all models (Table 2). For example, in the fully adjusted model, the odds ratio (OR) for MCID achievement was 0.70 (95% confidence interval [CI] 0.56–0.87) for those with metabolic syndrome. The unadjusted rates of achieving MCID at 6 months were 56% for those without and 48% for those with metabolic syndrome. In addition, the OR for LDA achievement was 0.56 (95% CI 0.43–0.74) for those with metabolic syndrome.
| CDAI MCID | CDAI LDA | |||
|---|---|---|---|---|
| OR (95% CI) | P | OR (95% CI) | P | |
| Metabolic syndrome | 0.70 (0.56–0.87) | 0.001 | 0.56 (0.43–0.74) | <0.001 |
| Components | ||||
| Hypertension | 0.92 (0.76–1.13) | 0.43 | 0.74 (0.57–0.94) | 0.02 |
| Diabetes | 0.76 (0.59–0.99) | 0.04 | 0.66 (0.48–0.92) | 0.01 |
| High triglycerides | 0.80 (0.66–0.97) | 0.03 | 0.74 (0.58–0.94) | 0.01 |
| Low HDL | 0.71 (0.56–0.90) | 0.004 | 0.66 (0.49–0.88) | 0.004 |
| High waist circumference | 0.75 (0.61–0.92) | 0.006 | 0.59 (0.46–0.77) | <0.001 |
- * Adjusted for age, sex, CDAI, biologic naive versus experienced, tumor necrosis factor inhibitor versus non–tumor necrosis factor inhibitor, anticitrullinated peptide antibody/rheumatoid factor serostatus, disease duration, and baseline C-reactive protein level. CDAI, Clinical Disease Activity Index; CI, confidence interval; HDL, high-density lipoprotein; LDA, low disease activity; MCID, minimal clinically important difference; OR, odds ratio.
When the individual components of metabolic syndrome were explored separately in regression models, each was associated with lower response (all P < 0.05) with the exception of a diagnosis of hypertension (OR 0.92, 95% CI 0.76–1.13) (Table 2). When all components were included in a single model, high waist circumference (OR 0.80, 95% CI 0.64–0.98) and low high-density lipoprotein were independently associated (OR 0.75, 95% CI 0.58–0.97) (Supplementary Table 3). There was a dose-dependent relationship such that the odds of response were lower when a greater number of components of metabolic syndrome were present (Figure 1).
A regression model including the metabolic syndrome variable had better model fit compared with a model that included BMI category (AIC 3,127 vs AIC 3,135; Bayesian Information Criterion 3,202 vs 3,222). Further, the addition of metabolic syndrome to a regression model that included BMI category significantly improved model fit (likelihood ratio test P = 0.008). In this model, underweight BMI was independently associated with a lower odds of response compared with normal BMI (OR 0.24, 95% CI 0.075–0.77; P = 0.02) whereas obesity categories showed lower efficacy although these differences did not achieve significance with the exception of severe obesity (BMI ≥ 35: OR 0.72, 95% CI 0.53–0.98) (Table 3).
| Odds of CDAI MCID | ||
|---|---|---|
| OR (95% CI) | P | |
| Metabolic syndrome | 0.74 (0.59–0.93) | 0.009 |
| BMI category | ||
| Underweight | 0.25 (0.08–0.79) | 0.02 |
| Normal weight | 1 (reference) | – |
| Overweight | 0.82 (0.63–1.07) | 0.14 |
| Obese | 0.80 (0.53–1.08) | 0.14 |
| Severely obese | 0.72 (0.53–0.98) | 0.04 |
- * BMI, body mass index; CDAI, Clinical Disease Activity Index; CI, confidence interval; MCID, minimal clinically important difference; OR, odds ratio.
The association between metabolic syndrome and MCID was similar between those patients receiving TNFi and non-TNFi advanced therapies (P for interaction = 0.49) (Figure 2). Similarly, there was no difference in the effect of metabolic syndrome on achieving LDA among those receiving TNFi or non-TNFi advanced therapies (P for interaction = 0.26). In stratified analyses, metabolic syndrome was only significantly associated with response among the larger seropositive subgroup, although the association between metabolic syndrome and response was not statistically different in seropositive and seronegative patients (Supplementary Table 4).
Metabolic syndrome was associated with a lower likelihood of achieving low individual components of disease activity, including a lower odds of achieving a low patient global score, low evaluator global score, low swollen joint count, and low pain score (Supplementary Table 5). Although the odds of achieving a low tender joint count was lower among those with metabolic syndrome, this was not statistically significant (OR 0.81, 95% CI 0.61–1.06).
Adipokines and response
Compared with individuals without metabolic syndrome, those with metabolic syndrome also had lower levels of adiponectin (median 14.2, interquartile range [IQR] 10.0–20.0 vs median 24.1, IQR 14.2–37.5 ug/mL; P < 0.001), higher levels of leptin (median 10,014, IQR 4,585–25,435 vs median 5,725, IQR 1,656–11,808 pg/mL; P = 0.01), and higher levels of FGF-21 (median 262, IQR 83–1,162.1 vs median 138 IQR 64–350 pg/mL; P = 0.009) (Supplementary Table 6). Similarly, higher BMI was correlated with lower adiponectin levels (ρ = −0.28, P < 0.001), higher leptin levels (ρ = 0.43, P < 0.001), and higher FGF-21 levels (ρ = 0.15, P = 0.04).
In multivariable models, there was not an association between any of the adipokines and achievement of the MCID for CDAI independent of other clinical measures, such as BMI and metabolic syndrome (Table 4). In these models, those with metabolic syndrome had lower response rates, but this was not statistically significant (OR range 0.53–0.61; P range 0.09–0.19). There was no significant interaction by therapeutic mechanism of action (TNFi-naive vs biologic-experienced non-TNFi; all P > 0.05) (Supplementary Table S2). Those with higher adiponectin levels had a numerically greater likelihood of response among patients who were seronegative (hazard ratio [HR] 2.07, 95% CI 0.82–5.19; P = 0.12), a finding that was significantly different from that observed in seropositive patients (HR 0.95, 95% CI 0.67–1.36; P = 0.78) (P for interaction = 0.04). There was not a significant association between leptin and FGF-21 levels among seronegative patients (Supplementary Table 4).
| Unadjusted (n = 190) | Adjusted (n = 190) | |||
|---|---|---|---|---|
| OR (95% CI) | P | OR (95% CI) | P | |
| Adiponectin (per 1 SD) | 1.04 (0.78–1.47) | 0.81 | 1.08 (0.76–1.52) | 0.67 |
| Leptin (per 1 SD) | 0.92 (0.69–1.23) | 0.58 | 1.09 (0.77–1.55) | 0.62 |
| FGF-21 (per 1 SD) | 0.86 (0.78–1.39) | 0.30 | 0.89 (0.65–1.23) | 0.49 |
| Adipokine score (v. 1) | ||||
| 2 | 0.57 (0.25–1.31) | 0.19 | 0.60 (0.24–1.46) | 0.26 |
| 3 | 0.63 (0.27–1.49) | 0.29 | 0.82 (0.31–2.15) | 0.68 |
| 4 | 0.49 (0.20–1.24) | 0.13 | 0.55 (0.19–1.63) | 0.28 |
- * Each exposure (on the left) is evaluated in its own model adjusted for the below covariates. Two patients were missing data on anticitrullinated peptide antibody/rheumatoid factor serostatus and one was missing disease duration. Adjusted for treatment (tumor necrosis factor inhibitor vs non–tumor necrosis factor inhibitor), age, sex, baseline Clinical Disease Activity Index, metabolic syndrome, body mass index, anticitrullinated peptide antibody/rheumatoid factor serostatus, C-reactive protein level, and disease duration. CI, confidence interval; FGF, fibroblast growth factor; OR, odds ratio.
DISCUSSION
This study demonstrated a strong association between the presence of metabolic syndrome and lower responses to advanced therapies among patients with RA. On the whole, the rate of response was 56% among those without metabolic syndrome, but this was reduced to 48% among those with metabolic syndrome. Model fit was better for metabolic syndrome than that of BMI alone, suggesting that metabolic syndrome is more informative to clinical response than a simple assessment of weight or BMI. Importantly, there was no difference in the association observed according to whether the participants received a TNFi or non-TNFi biologic therapy, suggesting that the observation is common to multiple treatment pathways and not specific to those receiving TNFi therapies. The primary takeaway from this study is to further our understanding of prior work that has suggested an association between obesity and treatment response and to suggest that metabolic factors play an important role.
We observed no significant associations between adipokines and clinical response, although adipokines were associated with metabolic syndrome. In a prior observational study, high adipokine levels were associated with lower rates of remission and LDA over time in a longitudinal cohort.12 In contrast to the current report, the prior study was performed in a larger sample of >1,200 older adults, was conducted over a longer period of follow-up, and was not limited to initiators of new therapies. The current study does not support the hypothesis that adipokines play a causal role in affecting response to therapy that is independent of metabolic syndrome itself. In the current study, we also found that adipokines did not differentiate the likely responders to TNFi versus non-TNFi in this subsample. Although we are not aware of prior studies evaluating adipokines as predictors of response to two different classes of advanced therapies, prior work has shown that BMI and adipokines can help identify those who are most likely to respond to the initiation of a TNFi over conventional therapies (ie, triple therapy).5 A comparison with conventional therapies, as previously studied, was not possible within the CERTAIN data.
A number of prior studies have evaluated obesity and clinical response, summarized in several reviews and meta-analyses.2-4 Our study confirms prior studies and advances the literature by illustrating that this effect is stronger when considering the metabolic implications of obesity, including changes in lipids, blood pressure, insulin resistance, and fat distribution. Few studies have evaluated the impact of metabolic syndrome on disease outcomes in RA, although some observational studies have observed an association between higher disease activity scores and features of metabolic obesity.6, 7
The mechanisms by which metabolic syndrome lowers the likelihood of clinical response are unknown. Some have suggested that obesity (and metabolic syndrome) may drive inflammatory processes that may lead to more resistant disease. However, studies have also shown reduced rates of inflammation detected with advanced imaging and lower rates of radiographic damage among those with obese BMI and, in some studies, among those with greater visceral fat.20-23 Metabolic syndrome is also strongly associated with osteoarthritis, sleep disturbance, and chronic pain syndromes.24-26 Thus, poor response among patients with metabolic syndrome may represent refractory symptoms because of competing comorbid conditions like osteoarthritis that would be unlikely to improve with RA therapies. Indeed, one study showed that low clinical response rates in patients with obesity were not mirrored by similarly poor response rates on imaging,27 suggesting that lower response rates in the context of obesity are driven by factors other than refractory articular inflammation. Although this prior work puts the current observations in context, it is difficult to establish from the current analysis alone whether persistence of subjective symptoms or refractory joint inflammation drive the observed associations, particularly because all components of disease activity appeared to be affected by the presence of metabolic syndrome. Differences in response among those with metabolic syndrome may also reflect differences in disease phenotype. For example, the link between metabolic obesity and psoriatic disease is well established.28 Thus, further study is needed to understand the biologic mechanisms at play.
We are aware of one prior study evaluating the effect of obesity on clinical response across advanced therapies with different mechanisms of action. In that study, also from the CorEvitas registry, the lower response rates for patients with obesity were similarly observed for both TNFi and non-TNFi biologics, suggesting that low response rates for obesity were not specific to drugs with a particular mechanism of action.1 In the current study, there was again no clear benefit of one class of biologic over another in the context of metabolic syndrome or among those with higher/lower levels of adipokines. This observation perhaps supports the concept that comorbidity and chronic pain, rather than more aggressive RA-related inflammation, drive the reduction in response, because such an effect would not be likely to vary across treatments.
This study also confirmed a prior study in the CorEvitas registry that having an underweight BMI is associated with poor clinical response at 6 months.1 It is perhaps not surprising that the current study continues to observe this relationship independent of the presence of metabolic syndrome because metabolic syndrome would be highly unlikely to be observed in this group. Although the effects of obesity on clinical response seem to be at least partially explained by metabolic syndrome, the lower response seen among those with low BMI is not. Patients with RA who have a low BMI are observed to have greater synovitis and osteitis and adverse long-term outcomes, such as radiographic progression.22, 29, 30 Low BMI likely develops over time in the context of more severe and refractory inflammatory disease (ie, cachexia). Thus, low BMI and poor response to therapy may be associated with each other through their shared etiology (ie, an association with a severe disease phenotype).
The primary limitation of the current study is the potential for residual confounding from other exposures that are correlated with metabolic syndrome. Although we assessed and adjusted for a number of potential confounders, unmeasured confounding may be present. Although we used a validated outcome for clinical response, our study did not have imaging outcomes to directly assess the extent of inflammatory arthritis at the articular level. The presence of comorbid diseases in patients with metabolic syndrome might also affect response rates. Patients with metabolic syndrome are more likely to have other comorbidities, including noninflammatory joint conditions. Finally, the subsample evaluating adipokines was smaller, and we may have been underpowered to identify small effects of adipokines on clinical response. Although the cohort was smaller, it was selected carefully from the best and worst responders, suggesting that we might expect a larger effect size in this subcohort. Thus, the lack of association may not be adequately explained by somewhat diminished study power alone.
In conclusion, metabolic syndrome and its components are associated with lower clinical response rates in patients with RA after the initiation of a new advanced therapy; an association that is independent of BMI and increased with the number of components present for the syndrome. Although adipokines were strongly associated with metabolic syndrome, the lack of an independent association with clinical response suggests that they do not drive the relationship between metabolic syndrome and poor response. These observations have implications for clinical practice as well as clinical trial design. Future study may help to determine whether low response rates are caused by refractory inflammatory arthritis versus poor improvement in symptoms because of chronic pain and comorbidity.
ACKNOWLEDGMENTS
The authors would like to thank all the participating providers and patients in the CorEvitas Rheumatoid Arthritis Registry who contributed data for this study.
AUTHOR CONTRIBUTIONS
All authors contributed to at least one of the following manuscript preparation roles: conceptualization AND/OR methodology, software, investigation, formal analysis, data curation, visualization, and validation AND drafting or reviewing/editing the final draft. As corresponding author, Dr Baker confirms that all authors have provided the final approval of the version to be published, and takes responsibility for the affirmations regarding article submission (eg, not under consideration by another journal), the integrity of the data presented, and the statements regarding compliance with institutional review board/Declaration of Helsinki requirements.
