Tammy G. Curtice, PharmD, MBA, MS

 To characterize treatment patterns, healthcare resource utilization (HRU), and disease activity among patients with early rapidly progressive rheumatoid arthritis (eRPRA) in the United States when treated with a first-line biologic disease-modifying antirheumatic drug (bDMARD) tumor necrosis factor-α (TNF) inhibitor or first-line abatacept.

 Observational, multicenter, retrospective, longitudinal, medical records—based, cohort study.

 Patients with eRPRA were identified by anti—citrullinated protein antibody positivity, 28-joint Disease Activity Score–C-reactive protein of 3.2 or greater, symptomatic synovitis in 2 or more joints for at least 8 weeks prior to the index date, and onset of symptoms within 2 years or less of the index date. Patients received abatacept or a TNF inhibitor as first-line treatment. Patient characteristics, treatment patterns, HRU, and disease activity following bDMARD initiation were compared across the 2 groups. Odds ratios (ORs) of HRU in the first 6 months of bDMARD treatment were estimated using multivariable logistic regression to adjust for patient mix.

 There were 60 patients treated with abatacept and 192 treated with a TNF inhibitor in the first line. Those treated with first-line abatacept had lower adjusted odds of hospitalization (OR, 0.42; 95% CI, 0.18-0.95), emergency department (ED) visits (OR, 0.39; 95% CI, 0.16-0.93), and magnetic resonance imaging (MRI) (OR, 0.45; 95% CI, 0.21-0.97) than those treated with a first-line TNF inhibitor (all 

 <.05). Adjusted odds of achieving low disease activity as measured by clinical disease activity index within 100 days of bDMARD initiation favored first-line abatacept versus a first-line TNF inhibitor (OR, 4.37; 95% CI, 1.34-13.94; 

 Adjusting for disease severity, patients with eRPRA who were treated with first-line abatacept were less likely to have hospitalizations, ED visits, and MRI use during the first 6 months of bDMARD treatment and more likely to achieve low disease activity within 100 days of bDMARD start compared with those who received a first-line TNF inhibitor.

Current management of rheumatoid arthritis (RA) in the United States is often directed by payer medical policy, typically requiring a tumor necrosis factor-α (TNF) inhibitor as a first- and second-line biologic disease-modifying antirheumatic drug (bDMARD).

The paradigm of this one-size-fits-all approach to the treatment of RA, which then results in prescribers’ formulary constraints, is now challenged by defined clinical subsets of patients who may be better served by alternative treatment sequences.

This study challenges this paradigm by demonstrating benefit observed among patients with early rapidly progressive RA (eRPRA) treated with a non—TNF inhibitor bDMARD.

Patients with eRPRA treated with first-line abatacept were less likely to have hospitalizations, emergency department visits, and magnetic resonance imaging during the first 6 months of bDMARD treatment compared with those who received a first-line TNF inhibitor.

The systemic treatment of rheumatoid arthritis (RA) has evolved from nonsteroidal anti-inflammatory drugs (NSAIDs) to corticosteroids (CSs) to conventional disease-modifying antirheumatic drugs (cDMARDs) to biologic DMARDs (bDMARDs) and Janus kinase (JAK) inhibitors.

 bDMARDs plus JAK inhibitors represent 4 different mechanism of action (MOA) drug classes; one such drug class, the tumor necrosis factor-α (TNF) inhibitor class, dominates current treatment as both the initial and subsequent bDMARD in patients with progressive RA. Whether the result of clinical trial design or the lack of pathobiologic differentiation, patients with RA have, for the past decade, been treated in a restrictive one-size-fits-all paradigm. This treatment paradigm has been codified into payer medical policy that is usually implemented through a pharmacy benefit manager in which formulary restrictions limit symptomatic disease management to the drug sequence of cDMARDs until failure, followed by a TNF inhibitor until failure, and then an alternative TNF inhibitor upon failure. Although traditional medicine challenges the strategy of salvaging the failure of a drug of a specific MOA with another drug of the same MOA (eg, penicillin failure is not followed by alternative penicillin), there has been limited research comparing alternative treatment sequences.

This unusual paradigm in which failure of a drug class is addressed with an alternative drug from the same class persists despite a complex arsenal of alternative therapeutics. cDMARDs (including methotrexate and sulfasalazine) and TNF inhibitors (including adalimumab, certolizumab, etanercept, golimumab, and infliximab) have been widely used and studied over the past 18 years both as single agents and, more recently, in combination. Newer biologics showing favorable safety and efficacy have also steadily emerged.

 Some of the more recent additions to the RA armamentarium are non—TNF inhibitor drugs, including the T-cell costimulation blocker abatacept, the interleukin-6 receptor antibody tocilizumab, and the JAK inhibitor tofacitinib. Challenges to the TNF inhibitor–based paradigm, in the second-line and even the first-line setting, using drug options from this arsenal are beginning to emerge.

Recent clinical and molecular profiling have helped to further define subsets of patients who may not benefit from this uniform treatment strategy, such as patients with early rapidly progressive RA (eRPRA).

Anti—citrullinated protein antibody (ACPA) positivity has been shown to indicate more severe disease progression and more erosive disease, signifying a rapidly progressive form of RA.

 ACPA positivity, as measured by anti—cyclic citrullinated peptide-2 (anti-CCP2) tests, has been reported among 3% to 52% of RA cohorts,

 is likely helpful in identifying patients with eRPRA, and may assist in identifying the most effective treatment sequence for a given patient. Although no definition of eRPRA has been formally established, initial work by the American College of Rheumatology (ACR) has resulted in treatment recommendations stratified by early (<6 months duration) and established (≥6 months duration) RA in the presence or absence of poor prognostic features, including an anti-CCP2 test resulting in ACPA positivity.

The current study sought to identify the clinical subset of patients presenting with eRPRA and to subsequently examine treatment patterns, healthcare resource utilization (HRU), and disease activity following treatment with a first-line bDMARD. The primary aim was to describe HRU and clinical outcomes of patients with eRPRA treated with abatacept and compare them with those patients treated with a TNF inhibitor.

This was an observational, multicenter, retrospective, longitudinal, medical records—based, cohort study of patients with eRPRA treated with a first-line bDMARD. Patients meeting study inclusion and exclusion criteria were identified by providers in Cardinal Health’s proprietary Rheumatology Provider Extended Network, which represents 838 practices comprising approximately 1900 physicians across the 48 contiguous US states. Data were collected by electronic case report forms (eCRFs) abstracted by healthcare providers using patient medical records. Retrospective data were collected at 1 time point, covering at least 6 months of follow-up for each patient.

Patients with eRPRA treated with abatacept as a first bDMARD or a TNF inhibitor as a first bDMARD between February 1, 2012, and January 31, 2017, were eligible for study inclusion; date of initiation of first bDMARD was set as the index date. Patients were included in the study cohort if they met the following criteria: (1) had a diagnosis of RA; (2) were 18 years or older at RA diagnosis; (3) had symptomatic synovitis in 2 or more joints for at least 8 weeks prior to index date; (4) had a 28-joint Disease Activity Score (DAS28)—C-reactive protein (CRP) of 3.2 or greater prior to index date; (5) had ACPA positivity; (6) had onset of symptoms 2 years or less prior to index date; (7) initiated first-line abatacept or a first-line TNF inhibitor between February 1, 2012, and January 31, 2017; and (8) had 6 or more months of data following initiation of abatacept or a TNF inhibitor. Patients were classified into 1 of 2 treatment groups based on their first-line bDMARD treatment: abatacept or TNF inhibitor. Patients remained in this group until discontinuation of first-line abatacept or a first-line TNF inhibitor, death, they were lost to follow-up, or the end of the study period, whichever occurred first.

Treatment patterns included dates of bDMARD initiation and discontinuation and concomitant medications received during bDMARD treatment. Numbers of rheumatology visits, emergency department (ED) visits, inpatient hospitalizations, laboratory tests (CRP, erythrocyte sedimentation rate [ESR]), radiology procedures (magnetic resonance imaging [MRI], ultrasound, x-ray), and surgeries were collected from the patient chart for the 6-month period following initiation of first-line abatacept or a first-line TNF inhibitor.

 Disease activity was measured by DAS28 score, European League Against Rheumatism (EULAR) criteria, and clinical disease activity index (CDAI) at baseline and during the first 100 days of treatment with first bDMARD. The DAS28 was calculated based on swollen joint count (SJC), tender joint count (TJC), ESR, CRP, and physician global assessment at baseline and during treatment with first bDMARD. Low disease activity (LDA) and remission were defined as a DAS28 of 3.2 or less and less than 2.6, respectively. A EULAR response of “good” was defined as a decrease in DAS28 of greater than 1.2 points and resulting in a DAS28 of 3.2 or less, and a EULAR response of “moderate” was defined as a decrease in DAS28 of greater than 0.6 points and resulting in a DAS28 of 5.1 or less. CDAI scores were collected at baseline and during treatment with first bDMARD; LDA and remission were classified as CDAI scores of 2.8 or less and 10.0 or less, respectively.

Descriptive statistics were calculated to summarize patient demographics, baseline clinical characteristics, and clinical outcomes for all patients with eRPRA, as well as for each treatment group. Statistical analyses were performed using the statistical software package SAS 9.4 (SAS Institute; Cary, North Carolina). All tests were 2-sided and α = 0.05.

Patient demographics, baseline clinical characteristics, and clinical outcomes were compared between first-line treatment groups on univariate analyses using 

 tests and Wilcoxon tests for continuous variables and using Pearson χ

 or Fisher exact tests for categorical variables. Times to event for each first-line treatment group were summarized by median time and 95% CIs using Kaplan-Meier estimates; comparisons between times to event were made using log-rank tests.

Odds ratios (ORs) of HRU in the first 6 months of bDMARD treatment were estimated using multivariable logistic regression to adjust for baseline indicators of disease severity and patient mix (age, sex, race, ethnicity, region, payer type, time from RA diagnosis to bDMARD start, baseline concomitant medication use, baseline SJC, baseline TJC, baseline CDAI, and extra-articular manifestations). Differences in adjusted times to event were assessed by Cox proportional hazard models; the same patient covariates listed above were used in those models.

There were 60 and 192 patients who met the criteria for eRPRA who were treated with abatacept and a TNF inhibitor as the first bDMARD, respectively. Mean ages at first bDMARD start were 49.1 and 49.9 years, respectively, and mean times from RA diagnosis to bDMARD initiation were 1.5 and 6.4 months (all 

 >.05). At bDMARD start, patients with eRPRA treated with abatacept and a TNF inhibitor, respectively, had mean SJCs of 6.8 and 7.3, TJCs of 10.4 and 10.2, ESRs of 39.8 mm/hour and 44.0 mm/hour, CRPs of 8.1 mg/L and 8.2 mg/L, and numbers of bony erosions of 1.2 and 1.7 (all 

CS use at initiation of first bDMARD was similar among abatacept users (45%) and TNF inhibitor users (47%) (

 = .89) (Table 1). The most common maintenance doses of oral CS were 5 mg/day and 10 mg/day among 80% and 16%, respectively, of abatacept patients with concomitant oral CS use and 66% and 25% of TNF inhibitor patients with concomitant oral CS use. Approximately one-third (35%) and 43% of patients treated with abatacept and a TNF inhibitor, respectively, were maintained on oral CS use. The majority of patients treated with abatacept and a TNF inhibitor had NSAID use (52% vs 55%; 

 = .66) at the time of first bDMARD initiation. Narcotic analgesics were used among 32% and 23% (

 = .23) of patients treated with abatacept and a TNF inhibitor, respectively, whereas methotrexate was used among 35% and 27%, respectively (Table 1).

Mean duration of first-line treatment was similar across patients treated with abatacept and a TNF inhibitor (15 [range, 5-64] vs 14 [range, 0.23-121] months; 

 = .61). Median time to discontinuation was not reached among patients treated with abatacept (95% CI, 17.5-not reached) and 14.9 months (95% CI, 10.1-18.4; 

 <.01) among those treated with a TNF inhibitor (

). After controlling for patient characteristics, patients treated with first-line abatacept were at a reduced adjusted risk of discontinuation compared with those treated with a first-line TNF inhibitor (hazard ratio, 0.36; 95% CI, 0.15-0.86; 

 = .02). Primary reasons for discontinuation were similar across groups: Of patients treated with first-line abatacept and a first-line TNF inhibitor, 47% and 56% discontinued as primary nonresponders, 33% and 25% discontinued as secondary nonresponders, and 13% and 16% discontinued due to adverse events, respectively.

A lower proportion of patients treated with abatacept had hospitalizations (72% vs 88%), ED visits (75% vs 90%), and MRI use (65% vs 82%) (all 

 <.05), whereas use of ultrasound (82% vs 88%) and x-ray (88% vs 93%) during the first 6 months of bDMARD treatment were similar between the abatacept and TNF inhibitor groups (

). After adjusting for patient characteristics, those treated with first-line abatacept had significantly lower odds of hospitalization (OR, 0.42; 95% CI, 0.18-0.95), ED visits (OR, 0.39; 95% CI, 0.16-0.93), and MRI use (OR, 0.45; 95% CI, 0.21-0.97) compared with those treated with a first-line TNF inhibitor (all 

The majority (70%) of patients treated with first-line abatacept, and 49% of those treated with a first-line TNF inhibitor, achieved CDAI response (LDA or remission) in the first 100 days of the first bDMARD. After adjusting for patient characteristics, patients treated with first-line abatacept were 4.37 times (95% CI, 1.34-13.94; 

 = .01) as likely to achieve CDAI LDA within 100 days of bDMARD compared with those treated with a first-line TNF inhibitor (

By DAS28 response criteria, 33% and 25% of patients treated with first-line abatacept and a first-line TNF inhibitor, respectively, achieved DAS28 response (LDA or remission) in the first 100 days of first bDMARD. Adjusted odds of achieving DAS28 LDA in the first 100 days of first bDMARD were not statistically significantly different between patients treated with first-line abatacept and patients treated with a first-line TNF inhibitor (OR, 2.13; 95% CI, 0.22-20.82; 

The majority (62%) of patients treated with first-line abatacept and 44% of patients treated with a first-line TNF inhibitor achieved EULAR response (moderate or good) in the first 100 days of first bDMARD. After adjusting for patient characteristics, patients treated with first-line abatacept had greater odds of achieving a moderate EULAR response in the first 100 days compared with those treated with a first-line TNF inhibitor, although the difference was not statistically significant (OR, 2.76; 95% CI, 0.92-8.30; 

Results support clinical and economic benefit to patients with eRPRA treated with first-line abatacept compared with a first-line TNF inhibitor in terms of lower odds of hospitalizations, ED visits, and MRI use, as well as improvement in disease activity, early in the course of first-line treatment. Taken together, the findings of this study challenge the prevailing paradigm of a one-size-fits-all approach to the treatment of patients with RA, modeled by treatment with cDMARD then TNF inhibitor then alternative TNF inhibitor, regardless of clinical or biochemical characteristics.

Recent research into clinical and molecular patterns of disease has suggested that patients with RA may be more heterogeneous than previously considered, with distinct clinicopathologic and likely genetic subsets of patients. Martin et al

 identified a subset of patients with older age at diagnosis who are less likely to achieve disease remission with a TNF inhibitor. Emery et al

 described a subset of patients with RA who have early, rapidly progressing disease who similarly fare less well than the general RA population using the classic treatment paradigm. These findings, along with those of the current study, highlight the need to reconsider the treatment paradigm of starting with a TNF inhibitor and using subsequent drugs with the same MOA in light of this more refined understanding of the disease.

Although there are numerous studies reporting the use of a non—TNF inhibitor biologic in the second line following a switch or failure from a TNF inhibitor,

 this study provides data on real-world HRU and clinical outcomes among patients treated with first-line abatacept compared with a first-line TNF inhibitor. Furthermore, the current study identified an important clinical subset of patients with RA—those with eRPRA&mdash;who may benefit from a non—TNF inhibitor in the first line compared with a TNF inhibitor, as supported by the results of this study.

Although patients treated with first-line abatacept or a first-line TNF inhibitor had similar baseline characteristics, adjusted estimates suggest that patients treated with first-line abatacept had lower odds of hospitalization, ED visits, and MRI use in the first 6 months following initiation compared with those treated with a first-line TNF inhibitor. Patients treated with first-line abatacept were also at lower risk of discontinuation. Lower use and a lower maximum dose of CSs were observed among patients treated with first-line abatacept. Finally, the odds of achieving CDAI LDA in the first 100 days of biologic use were significantly greater among those treated with first-line abatacept. Although the results of this research demonstrate both more favorable response to treatment and lower HRU among those treated with first-line abatacept compared with a first-line TNF inhibitor, further research is warranted to establish the association between treatment response to first-line treatment and lower HRU during treatment.

There are several clinical and biochemical characteristics of patients with RA that may characterize them as having eRPRA. One important characteristic is the biomarker ACPA; its positivity is associated with more severe disease progression and greater joint damage.

 When assessing ACPA positivity, recent work by Sokolove et al demonstrated improved disease activity in ACPA-positive patients (compared with ACPA-negative patients) among those treated with abatacept, as well as improved response to abatacept among those with greater concentrations of anti-CCP2 antibodies (compared with patients with lower concentrations)—an association not observed with the TNF inhibitor adalimumab.

 Furthermore, patients with ACPA positivity were found to have greater RA-related healthcare costs compared with those who are ACPA negative.

 Additional work is needed to elucidate the impact of treatment tailored to ACPA status on healthcare costs. Results of the current study begin to answer this by reporting an association of first-line abatacept use with lower rates of hospitalizations and ED visits—2 healthcare resources associated with high costs&mdash;among a cohort of patients with ACPA positivity.

 More information may be needed to fully understand the complex pathways that lead to the generation and persistence of ACPAs in RA and to determine the extent to which ACPAs may drive the disease process.

 However, exploratory analyses demonstrate greater benefit of a non—TNF inhibitor treatment, abatacept, among patients with the highest baseline CCP2 antibody concentrations compared with those with lower CCP2 concentrations.

 The growing recognition of the heterogeneity of RA (eg, biology, level of disease activity, prognosis, comorbidities) among affected patients increasingly suggests that the management of this disease might benefit from a less rigid approach. In the 2012 update of its RA guidelines, the ACR evolved its recommendations to include these advances for the treatment and understanding of RA.

 Although cDMARDs are still considered a first-line treatment, there is a role for biologics in early RA as well. Further, early, aggressive treatment with a cDMARD or bDMARD has been shown to reduce future joint damage and disability.

 Nevertheless, more research is needed to offer the best possible outcomes for patients with RA based on their disease-specific clinical and biologic characteristics, including ACPA positivity.

The transformation of US healthcare delivery to a value-based care model creates additional impetus to reevaluate the long-standing TNF inhibitor—based paradigm. The fact that the current study found that patients treated with first-line abatacept were more likely to have a CDAI response in the first 100 days of treatment and less likely to discontinue first-line treatment with abatacept compared with a first-line TNF inhibitor is value differentiation for patients, their providers, and payers. Although the current study focused on patients with eRPRA, these results align with those reported by Gottenberg et al,

 who showed better effectiveness of abatacept among patients with rheumatoid factor and ACPA positivity—2 biomarkers consistent with eRPRA&mdash;as well as a lower risk of all-cause discontinuation and discontinuation due to ineffectiveness.

There are several important limitations to the current study, which are largely inherent to the retrospective observational design. All data are physician reported. To avoid recall bias, physicians referenced patient medical and office records to complete the eCRF. To address potential errors in the eCRF, data validation mechanisms were employed to disallow invalid values. Complete capture of hospitalizations and ED visits may be limited, given the inherent limitations of collecting HRU data as recorded by 1 provider (eg, resource use was likely limited to those associated with the rheumatology clinic). Differences in patient characteristics at index may confound treatment effects. Multivariable models were fit to adjust for differences in patient characteristics, but residual and/or unobserved confounding effects may exist. Although there is no formal definition of eRPRA, patients were considered to have eRPRA by virtue of meeting all of the study inclusion criteria.

Missing data is often a concern with medical record data source research. In this regard, it is worth highlighting important distinctions in how data were collected in this study from a traditional secondary data analysis of an electronic health record (EHR) data set. Unlike the use of automated extraction of EHR data, data were collected by healthcare providers by performing a review of patient charts/records and entering requested data elements into an eCRF. This review by the treating physician allows for the capture of data elements that may be stored in unstructured fields that often are not captured in an automated extraction of EHR data. This process substantially increases the amount of nonmissing data compared with a natural language abstraction of an EHR data set. Missing data were also minimized by application of the study inclusion criteria; patients had to have known nonmissing data for key baseline variables: diagnosis of RA, swollen/tender joints (ie, synovitis), DAS28 score, anti-CCP2 status, duration of symptoms, bDMARD treatment details, and 6 or more months of follow-up in the participating provider’s clinic.

This study provides data on real-world HRU and clinical outcomes among patients treated with first-line abatacept compared with a first-line TNF inhibitor. Specifically, benefit was demonstrated among an important subset of patients with RA: those with eRPRA. Patients with eRPRA who were treated with first-line abatacept were less likely to have hospitalizations, ED visits, and MRI use during the first 6 months of bDMARD treatment compared with those who received a first-line TNF inhibitor. ACPAs may be a potential biomarker to guide identification of eRPRA. These results challenge the prevailing treatment paradigm in which all patients with RA follow a one-size-fits-all approach initiated by a TNF inhibitor. An alternative approach to managing patients with RA, based on clinical and biochemical characteristics, has important implications in treatment decisions made by healthcare providers and health policy makers for patients with RA and, more specifically, patients characterized as having eRPRA. Further research is warranted to establish the association between treatment response to first-line treatment and lower HRU during treatment.

The authors would like to thank Jill Kaufman for extensive help with recruitment and data collection.

 Cardinal Health Specialty Solutions (AJK, DN, BAF), Dublin, OH; Bristol-Myers Squibb (TGC, KG, KWT, ARS), Princeton, NJ.

 This study was sponsored and funded by Bristol-Myers Squibb (BMS).

 Results from this study have been presented at the Academy of Managed Care Pharmacy Nexus 2017 and European League Against Rheumatism 2018 meetings.

 Dr Curtice was previously an employee of BMS and owns BMS stock; BMS has an approved product for rheumatoid arthritis. Drs Tuell and Szymialis are employees of BMS and own BMS stock. Dr Nero is an employee of Cardinal Health, which performed this study per a contract with BMS. The remaining authors report no relationship or financial interest with any entity that would pose a conflict of interest with the subject matter of this article.

 Concept and design (AJK, KG, KWT, ARS, DN, BAF); acquisition of data (AJK, KG, BAF); analysis and interpretation of data (AJK, TGC, KG, KWT, ARS, DN, BAF); drafting of the manuscript (AJK, TGC, KG, KWT, BAF); critical revision of the manuscript for important intellectual content (AJK, TGC, KG, ARS, DN, BAF); statistical analysis (KG, DN); provision of patients or study materials (KG); obtaining funding (KG); administrative, technical, or logistic support (AJK, TGC, KG, ARS, BAF); and supervision (TGC, KG).

Articles

This study suggests that lower healthcare resource use and achieving low disease activity are associated with first-line abatacept compared with a first-line tumor necrosis factor-α inhibitor for patients with early rapidly progressive rheumatoid arthritis.

Tammy G. Curtice, PharmD, MBA, MS

Articles

Real-World Outcomes Among Patients With Early Rapidly Progressive Rheumatoid Arthritis