Obesity and response to anti-tumor necrosis factor-α agents in patients with select immune-mediated inflammatory diseases: A systematic review and meta-analysis

Objectives We sought to evaluate the association between obesity and response to anti-tumor necrosis factor-α (TNF) agents, through a systematic review and meta-analysis. Methods Through a systematic search through January 24, 2017, we identified randomized controlled trials (RCTs) or observational studies in adults with select immune-mediated inflammatory diseases–inflammatory bowel diseases (IBD), rheumatoid arthritis (RA), spondyloarthropathies (SpA), psoriasis and psoriatic arthritis (PsA)–treated with anti-TNF agents, and reporting outcomes, stratified by body mass index (BMI) categories or weight. Primary outcome was failure to achieve clinical remission or response or treatment modification. We performed random effects meta-analysis and estimated odds ratios (OR) and 95% confidence interval (CI). Results Based on 54 cohorts including 19,372 patients (23% obese), patients with obesity had 60% higher odds of failing therapy (OR,1.60; 95% CI,1.39–1.83;I2 = 71%). Dose-response relationship was observed (obese vs. normal BMI: OR,1.87 [1.39–2.52]; overweight vs. normal BMI: OR,1.38 [1.11–1.74],p = 0.11); a 1kg/m2 increase in BMI was associated with 6.5% higher odds of failure (OR,1.065 [1.043–1.087]). These effects were observed across patients with rheumatic diseases, but not observed in patients with IBD. Effect was consistent based on dosing regimen/route, study design, exposure definition, and outcome measures. Less than 10% eligible RCTs reported outcomes stratified by BMI. Conclusions Obesity is an under-reported predictor of inferior response to anti-TNF agents in patients with select immune-mediated inflammatory diseases. A thorough evaluation of obesity as an effect modifier in clinical trials is warranted, and intentional weight loss may serve as adjunctive treatment in patients with obesity failing anti-TNF therapy.


Introduction
The global prevalence of obesity is rising, with one in 10 people across the world being classified as obese. [1,2] In the United States, over 35% adults are obese, and increased healthcare spending on this population is estimated to account for nearly a third of the growth in healthcare expenditure. [3] Obesity may contribute to increased risk of developing select immunemediated inflammatory diseases (IMIDs) such as rheumatoid arthritis (RA), psoriasis and Crohn's disease (CD), and approximately 10-50% of patients with IMIDs are obese. [4][5][6][7][8][9] Obesity has been associated with more severe disease activity, inferior quality of life and higher burden of hospitalization in patients with these immune-mediated diseases. [5,6,[10][11][12][13][14] Targeted immunomodulators such as anti-tumor necrosis factor-α (TNF), are the mainstay of therapy for patients with select IMIDs with moderate-severe disease. Clinical response to these agents is seen in 40-80% patients with select IMIDs. [15][16][17][18][19] Population pharmacokinetic studies of different anti-TNF agents have consistently shown that high body weight is associated with accelerated clearance, resulting in lower trough concentrations. [20][21][22] Additionally, obesity, particularly visceral fat, independently contributes to higher systemic inflammatory burden. [23,24] Several observational studies have shown that obesity may be a negative prognostic marker in patients with rheumatic diseases, [11,25] and variably and inconsistently shown an inferior response to biologic agents in obese patients. [26][27][28][29][30][31][32][33] Hence, we sought to systematically review the association between obesity and response to anti-TNF agents across selected IMIDs, and examine whether the effect varies across different diseases and between different anti-TNF agents based on route of administration (subcutaneous vs. intravenous) and dosing scheme (weight-based vs. fixed dosing).

Methods
This systematic review is reported according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) Statement and was conducted following a priori established protocol (S1 Text). [34]

Selection criteria
We included phase 2 or phase 3 randomized controlled trials (RCT) or observational cohort studies, with at least 2 months follow-up, that met the following inclusion criteria: (1) patients with select IMIDs (IBD including CD or ulcerative colitis [UC], RA, psoriasis or psoriatic arthritis [PsA], spondyloarthropathy [SpA]), treated with US Food and Drug Administration (FDA)-approved anti-TNF agents (infliximab, adalimumab, certolizumab pegol, golimumab, etanercept), (b) outcome reported as failure to achieve clinical remission or response or need for treatment modification, and (c) outcomes stratified by baseline body mass index (BMI) or body weight. For our analysis, RCTs were considered as prospective cohort studies, and only patients receiving active intervention with anti-TNF agents were included in analysis; if two or more arms in the RCTs were receiving different active biologic therapy, then they were considered separate cohorts. We excluded cross-sectional studies or case-control studies, studies of non-TNF biologic agents, immunomodulators or small molecules, or studies in which outcomes were not stratified by baseline BMI or weight. We also excluded studies in which the exposure categories did not include any patients with overweight or obesity patients (for example, BMI <20 vs. !20kg/m 2 ). [35] When there were multiple publications from the same cohort, we only included data from the most recent comprehensive report.

Data sources, search strategy and study selection
The search strategy was designed and conducted by an experienced medical librarian with input from study investigators, utilizing various databases from inception to January 24, 2017. The databases included Ovid Medline (January 1, 1946 to January 24, 2017), EMBASE (January 1,1988 to January 24, 2017), Scopus (January 1, 2004 to January 24, 2017), Web of Science (January 1,1990 to January 24, 2017), and Cochrane Central Register of Controlled Trials (January 1,2005 to January 24, 2017). In addition, we searched clinical trial registries (www. clinicaltrials.gov and www.clinicaltrialsregister.eu), conference proceedings and published systematic reviews for additional studies. Controlled vocabulary supplemented with keywords was used to search for observational studies (S2 Text) and RCTs (S3 Text) reporting impact of obesity on response to anti-TNF therapy. Two authors independently reviewed the title and abstract of studies identified in the search to exclude studies that did not answer the research question of interest, based on pre-specified inclusion and exclusion criteria. The full text of the remaining articles was independently reviewed, to determine whether it contained relevant information. Next, we manually searched the bibliographies of the selected articles, as well as review articles on the topic for additional articles. We also performed a manual search of conference proceedings from major gastroenterology, rheumatology and dermatology meetings (Digestive Diseases Week, Annual Meeting of the American College of Rheumatology, Annual European Congress of Rheumatology, and Annual meeting of the American College of Dermatology from 2012 to 2016) for additional abstracts on the topic. Since RCTs may report results stratified by body weight or BMI in subgroup analyses in online supplements alone, we reviewed full-texts of all published phase 2 and 3 RCTs of candidate anti-TNF agents in detail, and on clinical trial registry websites (www.clinicaltrials.gov, www.clinicaltrialsregister.eu).

Data abstraction and risk of bias assessment
After study selection, two authors independently abstracted data on study and patient characteristics, exposure variables, outcomes, confounding variables and statistical analyses, using a standardized data abstraction form. Details of data abstraction are shown in the appendix (S4 Text).
Risk of bias was assessed by 2 investigators independently, using the Quality In Prognosis Studies tool, which evaluates validity and bias in studies of prognostic factors across six domains: participation, attrition, prognostic factor measurement, confounding measurement and account, outcome measurement, and analysis and reporting. [36] reported outcome of interested was chosen: (1) failure to achieve clinical remission, (2) failure to achieve clinical response (based on validated disease activity indices or author-defined criteria); or (3) need for treatment modification (discontinuation or escalation of index therapy due to ineffectiveness and/or switching to alternative therapy, and/or need for surgery).
In order to evaluate stability of the association between obesity and response to anti-TNF therapy, and examine potential sources of heterogeneity, we performed several a priori subgroup analyses based on: type of diseases (IBD vs. RA vs. SpA vs. psoriasis-PsA), route of administration (subcutaneous vs. intravenous), dosing scheme (weight-based dosing vs. fixed dosing), exposure categories (based on BMI: obese vs. normal BMI, obese vs. non-obese [BMI!30kg/m 2 vs. BMI<30kg/m 2 ], overweight vs. non-overweight [BMI!25kg/m 2 vs. BMI<25kg/m 2 ]), outcome definition (based on disease activity indices vs. need for modification of therapy) and study design (RCTs vs. cohort). Sensitivity analysis restricting to studies that adjusted for key confounding variables was performed. We also performed meta-regression to assess whether effect estimates varied depending on prevalence of exposure (proportion of obese patients) or the prevalence of the outcome (proportion who failed index biologic therapy).
We examined dose-response relationship using two approaches. First, we limited analysis to studies reporting the association between 2 or more categories of obesity against a reference category, and compared as nominal groups. For example, for studies using WHO-defined categories, then comparing obese vs. normal BMI and overweight vs. normal BMI; for studies reporting tertiles of BMI or weight, then comparing 3 rd tertile (T 3 ) vs. 1 st tertile (T 1 ) and 2 nd tertile (T 2 ) vs. 1 st tertile (T 1 ). When studies reported multiple different exposure categories, then all categories beyond the bottom two were collapsed into a single category (labeled high BMI or weight) for consistent comparison. Second, studies that reported analyses per 1kg/m 2 change in BMI were pooled separately.

Statistical analysis
We used the random-effects model described by DerSimonian and Laird to calculate summary OR and 95% confidence intervals (CI). [37] Maximally adjusted OR, where reported in studies, was used for analysis to account for confounding variables; where not reported, unadjusted OR were used or calculated based on event rates in different exposure categories. To estimate what proportion of total variation across studies was due to heterogeneity rather than chance, I 2 statistic was calculated. [38] In this, a value of <30%, 30%-60%, 60%-75% and >75% were suggestive of low, moderate, substantial and considerable heterogeneity, respectively. Between-study sources of heterogeneity were investigated using subgroup analyses by stratifying original estimates according to study characteristics (as described above). In this analysis, a p-value for differences between subgroups of <0.10 was considered statistically significant. Publication bias was assessed qualitatively using funnel plots and quantitatively using Egger's regression test. [39] All analyses were performed using Comprehensive Meta-Analysis version 2.0 (Englewood, New Jersey). Since this was a meta-analysis of published studies, institutional review board approval was not required.

Results
From 4242 unique studies identified based on systematic review, 3823 articles were excluded based on title and abstract review (since these were unrelated to biologic therapies or unrelated to diseases of interest). Overall, 419 articles were reviewed in full for eligibility. Of 241 eligible RCTs reviewed in full, 20 (8.3%) reported results stratified by BMI or weight and were included; the remainder did not provide data stratified by BMI or weight. Twenty five RCTs of non-TNF biologics were excluded, and 5 RCTs were excluded since the outcomes of interest were not reported. Of 149 cohort studies, 99 were excluded since they included either non-TNF biologics (n = 37), did not adequately stratify results by BMI or weight (n = 46) or were unrelated to diseases of interest (n = 16). We included 54 cohorts derived from 50 studies for categorical analysis, [14,[26][27][28][29][30][31][32][33] and 9 cohorts from 8 studies for per-unit dose-response analysis. [32,33,42,54,[81][82][83][84] These included 34 observational cohorts (n = 13,336) and 20 cohorts derived from RCTs (n = 6,036). Fig 1 shows the study selection flowchart.

Characteristics of included studies
Tables 1 and 2 provide details of included studies and participants. Overall, these 54 cohorts reported on 19,372 patients; 10, 16, 6 and 22 studies included patients with RA, IBD, SpA and psoriasis-PsA, respectively. Median 23% (IQR, 16-31.5%) patients were classified as obese in these studies. Follow-up ranged from 2-36 months. Overall, the studies were deemed to be at moderate risk of bias, due to failure to adjust for confounding variables and non-standard exposure definition (S1 Table).

Obesity and response to anti-TNF therapy
On meta-analysis, across included IMIDs and across all anti-TNF agents, obesity was associated with 60% higher odds of failure of index anti-TNF therapy (OR, 1.60; 95% CI, 1.39-1.83), with substantial heterogeneity (I 2 = 71%). [14,[26][27][28][29][30][31][32][33] Dose-response relationship. The odds of failing therapy were highest in the third tertile of BMI or weight as compared to patients in the second tertile (21 studies; T 3 vs. Route of administration and dosing regimen. Infliximab is the only anti-TNF agent routinely administered intravenously in a weight-based dose; additionally, a form of golimumab (Simponi Aria 1 ) is occasionally administered intravenously in a weight-based dose; all other agents are administered in a fixed dose independent of body weight, subcutaneously. Hence, route and dosing scheme could not be evaluated independently. Patients with obesity treated with both weightbased dosing regimens ( 80] had higher odds of failing therapy, without significant differences in the summary estimate (p-value comparing effect size for fixed-dose and weight-based dosing = 0.28).   Obesity and response to anti-TNF agents Obesity and response to anti-TNF agents    [14, 26-33, 40, 42, 43, 45-47, 49-51, 53, 55, 61-64, 68, 70, 74, 76-78, 80] On meta-regression, prevalence of obesity in included studies modestly but significantly affected summary estimate (p = 0.06), with smaller effect size observed in studies with lower prevalence of obesity (Panel A in S4 Fig)

Principal findings
In this systematic review with meta-analysis of 54 cohorts with 19,372 patients with select IMIDs treated with anti-TNF agents, we made several key observations related to the influence of obesity on response to anti-TNF therapy. First, we observed that obesity was associated with a 60% higher odds of failing anti-TNF therapy as compared to non-obese and normal BMI counterparts, for most IMIDs, including RA, spondyloarthropathies and psoriasis and psoriatic arthritis, but not IBD. This association was stable across diverse subgroups including study design, exposure definition, outcome measurement, and was independent of other potential confounding variables. Second, a dose-dependent effect was seen, with each unit increase in BMI associated with 6.5% higher odds of failing therapy. Third, this effect was independent route of administration and dosing scheme. It is also important to note that <10% of eligible RCTs reported results stratified by baseline BMI. Our findings of negative impact of obesity on treatment response to anti-TNF agents suggests that obesity is a negative prognostic factor and treatment effect modifier that must be considered in clinical trial design and clinical practice. Obesity and response to anti-TNF agents Obesity is recognized as a perpetual state of chronic low-grade inflammation, through systemic and paracrine increase in levels of cytokines, chemokines and adipokines. [6] Besides its direct impact on inflammation, obesity can also modify pharmacokinetics of anti-TNF and other biologic agents. Population pharmacokinetic studies of all anti-TNF agents have identified high body weight as a risk factor associated with increased clearance of drug, resulting in shorter half-life and lower serum trough drug concentrations. [20][21][22] This effect might be related to rapid proteolysis and to a 'TNF-sink' phenomenon, wherein the clearance of a monoclonal antibody that binds to membrane antigen is faster at low doses as the unbound targets "sop up" antibody, serving as a sink. [85] This may explain why patients with obesity treated even weight-based regimens such as infliximab, had inferior response to therapy.

Comparison with other studies
Our findings expand on prior observations on the potentially negative impact of obesity on overall treatment response in patients with RA and psoriasis. However, these prior studies included all treatment categories, instead of focusing on anti-TNF therapy. Moreover, they were unable to simultaneously study differential impact across different diseases, and with different anti-TNF agent. Though we had hypothesized that obesity would negatively impact Obesity and response to anti-TNF agents response to anti-TNF therapy across all selected diseases, we did not observe such an association in patients with IBD. This may be a true observation. Patients with IBD show a unique locally restrictive form of visceral adipose tissue-creeping fat-whereby mesenteric fat hyperplasia is limited to areas of inflamed bowel. [86,87] It is likely that in IBD patients, this local mesenteric fat plays a more important role than systemic obesity. Additionally, dose of infliximab and adalimumab approved in patients with IBD is higher than that for other rheumatic diseases. Differential effect of confounding by disease severity in IBD and rheumatic diseases may also explain this finding-severe IBD is likely to result in weight loss and misclassification of obesity, whereas, severe rheumatic diseases would likely impact physical activity, promote sedentary lifestyle and contribute to obesity. Alternatively, it is possible that our observation in IBD patients is biased. Most studies in IBD patients presented dichotomous categories, usually as above or below median weight, and did not include patients at extreme categories of BMI, where the negative impact of obesity may be more pronounced. On meta-regression, effect estimates were smaller in studies with lower prevalence of obesity. Five studies in IBD patients were reported only as abstracts, putting them at high risk of bias (though excluding these studies did not modify the summary estimate). Further studies, of both fixed dose therapies and Obesity and response to anti-TNF agents weight-based therapies, preferably using individual participant data from clinical trials are warranted to better understand this association.

Strengths and limitations
The strengths of this systematic review include: (a) comprehensive and systematic literature search with well-defined inclusion criteria, including both RCTs and observational studies as cohort studies, (b) a priori comprehensive sub-group and sensitivity analyses to evaluate the stability of findings and identify potential factors responsible for inconsistencies, (c) assessment of dose-response relationship using two approaches, and (d) simultaneous evaluation of unadjusted (based on raw numbers) and adjusted risk estimates, and hence, being able to evaluate the potential influence of measured confounders on the summary estimate.
There are several limitations in our study. First, the meta-analysis included only cohorts, derived from both RCTs and observational studies, but participants in RCTs were not stratified based on presence or absence of obesity. Most of the included studies did not account for several potential confounding factors including steroid use and baseline disease activity, both of which influence exposure and outcome. Additionally, factors that may be intrinsically related to obesity, such as depression, fibromyalgia, etc., may confound this association Obesity and response to anti-TNF agents between obesity and inferior clinical response to therapy. While sensitivity analysis using the adjusted data had little effect on the summary OR, potentially suggesting that any difference attributable to using different confounders for adjustment is likely small, data regarding several confounders was inadequately reported. Second, substantial heterogeneity was observed in the overall analysis, which was partly explained by difference in outcome definition and prevalence of obesity in included cohorts, as well as different types of diseases. It is important to note that a stronger effect estimate was observed prospective studies that used validated disease activity indices, with fairly consistent outcome definitions, as compared to retrospective studies which used non-validated definitions such as need for escalation of index therapy, switching to alternative therapy or surgery, etc. Fourth, despite absence of small study effects, there was concern for reporting bias. Less than 10% of eligible RCTs results stratified by baseline BMI; when they did, exposure was frequently classified as dichotomous variable, above and below a median weight. Finally, we limited our analysis to only anti-TNF agents, instead of including other non-TNF-biologic agents and small molecules. This was done to minimize conceptual heterogeneity. Negative impact of obesity on response to other therapies has been observed, and merits detailed evaluation.

Implications for clinical practice
While the negative impact of obesity in patients with psoriasis is well-known, our observations that obesity uniformly results in inferior response to anti-TNF therapy across all rheumatic diseases that we studied has important implications for both clinical practice and clinical trial design. In clinical practice, physicians may consider aggressive treatment and close proactive monitoring in patients with obesity treated with anti-TNF agents such as empirically using higher dose of anti-TNF agent in obese patients, frequent therapeutic drug monitoring and/or use of combination therapy with immunomodulators to increase drug concentration and decrease risk of immunogenicity. Clinical trialists should consider obesity as a potential effect modifier and consider obesity as a stratification variable; at the very least, stratified analysis by baseline BMI should be performed and consistently reported. Finally, the presence of obesity may offer a potential therapeutic intervention, either through adjusting biologic dosing for body weight or through directly targeting the obesity itself for treatment in patients with select IMIDs with a multi-disciplinary approach. Small RCTs and cohort studies in patients with psoriasis, psoriatic arthritis and RA have suggested a beneficial effect of intentional weight loss on treatment response to anti-TNF agents. [88][89][90]

Conclusion
In conclusion, obesity is associated with inferior response to anti-TNF therapy in patients with rheumatic diseases, but not in patients with IBD. This effect is independent of route of administration, and is observed in patients treated with weight-based as well as fixed dose agents. However, current findings need to be interpreted with caution due to high heterogeneity and due to lack of adjustment for relevant confounding factors in included studies. Prospective cohort studies and post-hoc analyses of RCTs with individual participant level data are warranted to confirm this association. If this effect is consistent, interventional studies targeting obesity should be explored for difficult-to-treat obese patients with select IMIDs.