Impact of Adherence to Golimumab on Disease Flares in Rheumatoid Arthritis: Results from a Canadian Observational Study

Introduction

Rheumatoid Arthritis (RA) is a chronic inflammatory autoimmune disorder that primarily affects the joints symmetrically, causing pain, swelling, and stiffness.^1,2 It is characterized by progressive inflammation and hyperplasia of the synovial tissue, leading to cartilage and bone damage, as well as the production of rheumatoid factor and anti-citrullinated protein antibodies.^3,4 These symptoms limit physical function and significantly reduce patients’ quality of life.³

Initial RA treatment typically includes conventional synthetic disease-modifying antirheumatic drugs (csDMARDs) such as methotrexate and leflunomide, used alone or in combination. For patients who do not respond to first-line treatments, biologic DMARDs (bDMARDs) including tumor necrosis factor inhibitors (TNFi), interleukin-6 inhibitors (IL-6i), T-cell costimulation modulators, or targeted synthetic DMARDs (tsDMARDs) like Janus kinase inhibitors (JAKi), may be prescribed.^3,5,6

Therapeutic adherence, defined as the extent to which a patient’s medication-taking matches their prescribed regimen, is essential for achieving maximum drug effectiveness and optimal treatment outcomes.⁷ Previous studies have shown that consistent adherence allows for sustained control of disease activity, lower rates of disease flares, reduced progression of joint damage, and improved quality of life in patients with RA.^8–12 Conversely, non-adherence or low adherence can compromise treatment efficacy, leading to inadequate therapeutic response, increased disease activity, and greater risk of flare-ups.^9,13,14 RA flares are defined as periodic, significant worsening of disease activity, characterized by severe and prolonged symptoms that may necessitate treatment adjustments or new therapies.^15–17 These flare-ups can result in intensified pain, increased joint damage and radiographic progression, stiffness, reduced mobility, and general malaise.^18–20

Previous research has shown that treatment adherence often decreases over time,^21,22 with non-adherence rates ranging from 9.6% to 55% in patients with RA.^23–26 Although adherence varies by treatment class and route of administration,^27,28 non-adherence to bDMARDs remains a common issue in RA. Understanding real-world patterns and impact of non-adherence can help inform strategies to improve patient outcomes.

Golimumab is a tumor necrosis factor-alpha (TNF-α) inhibitor often used in a monthly dosing regimen in combination with methotrexate for RA treatment.²⁹ Studies have shown that golimumab can slow disease progression and alleviate associated symptoms, and it has been reported to have a higher adherence rate compared with other bDMARDs.^30–33 This study aimed to assess adherence rates to golimumab treatment among Canadian patients with RA followed in routine clinical care, and to evaluate the association between non-adherence and the incidence of disease flares.

Methods

Study Design and Population

GO FAR (NCT03729349) was a prospective, multi-center, non-interventional real-world study that was carried out in 27 centres across Canada between January 2019 and August 2023. Eligible patients were adults with confirmed RA diagnosed by the treating physician, for whom the treating rheumatologist decided to initiate golimumab treatment prior to and independently of participating in the study. Patients were excluded if they (1) had a diagnosis of axial spondyloarthritis, ankylosing spondylitis or psoriatic arthritis; (2) had received any investigational product in the last 3 months prior to study initiation; or (3) were currently participating in another clinical trial, either investigational or observational.

Baseline demographics, disease characteristics and concomitant rheumatologic medication use were gathered at the initial visit. Clinical disease assessments, patient assessments of adherence and disease flares, and adverse events (AEs) were recorded twice in 6-month intervals during the 12-month observational period.

Ethics approval for the study was obtained from each participating site’s independent ethics committee or institutional review board and all study participants provided written informed consent prior to enrollment.

Outcome Measures

Adherence with treatment was assessed with the Compliance Questionnaire in Rheumatology (CQR).³⁴ CQR is a patient reported compliance/adherence measure designed to assess patient’s compliance/adherence with anti-rheumatic treatments. It contains 19 items scored on a 4-point Likert scale, where a total score of 0 indicates no adherence and 100 indicates complete adherence. In a previous study, multiple linear regression analysis generated a weighted CQR score, which showed good validity in detecting satisfactory (>80%) compliance/adherence and correct dosing.³⁵ Using the previously identified optimal threshold score (Z_k of –0.58489) for the discrimination of satisfactory adherence, patients were classified as having high or low predicted adherence at baseline, 6 and 12 months.

Disease flaring was assessed by using Rheumatoid Arthritis – Flare Questionnaire (RA-FQ) mean score and a flare was defined as a positive response to question 7 (Are you having a Flare?) of the RA-FQ.³⁶ RA-FQ is a patient reported flare questionnaire with 7 items that measures 5 components including (a) pain, physical activity, fatigue, stiffness, participation in daily activities over the past week using a numeric rating scale of 0–10, where a score of 0 indicates none and a score of 10 indicates severe symptoms; and (b) self-reporting of symptoms duration >7 days and self-reporting of a flare. A higher score in the first 5 questions and a positive response to questions 6 and 7 indicated worsening flare.

Both questionnaires were completed by the patient in his or her local language (English or French).

Statistical Analysis

The analysis population (effectiveness analysis set; EAS) included patients who received at least 1 dose of golimumab and had both a baseline and at least 1 post-baseline assessment (6- or 12-month visit) within the specified 1-year visit window. Summary statistics were produced for all study variables, including the mean (standard deviation [SD]) or the median (interquartile range [IQR]) for continuous variables, depending on data distribution, and frequency distributions for categorical variables. The efficacy data were summarized by level of adherence, categorized as high or low based on baseline or visit adherence.

The impact of treatment adherence on flares was analyzed by comparing (a) the mean RA-FQ scores and (b) the proportions of patients with a positive answer to question 7 of the RA-FQ between the baseline predicted low and high adherence groups using the differences in means/proportions along with the corresponding 95% confidence interval (CI). Additionally, a secondary analysis was conducted to assess the impact of treatment adherence on flares using visit-predicted adherence instead of baseline-predicted adherence.

Similar methods were used to assess the association between treatment adherence and (a) disease activity, measured with the Clinical Disease Activity Index (CDAI), Physician Global Assessment (MDGA), and Patient Global Assessment (PtGA); (b) glucocorticoids use, and; (c) incidence of AEs and serious AEs (SAEs).

All analyses were performed using SAS software, version 9.4 (SAS Institute, North Carolina, US).

Results

Baseline Patient Characteristics

A total of 215 patients were enrolled, of whom 78.6% completed the study. Primary reasons for study discontinuation included withdrawal by the subject (5.1%), physician decision (4.2%), and loss to follow up (3.3%). The EAS included 180 patients, the majority of whom were female (66.7%) and white (84.7%), with a mean (SD) age of 58 (11.8) years, a BMI of 30.0 (7.2) kg/m², and median (IQR) time since RA diagnosis of 4.0 (0, 45.7) years. Mean (SD) CDAI levels of 29.1 (12.3) were indicative of high disease activity at baseline. Most (77.2%) patients at baseline were biologic-naive; 40.0%, 36.7% and 16.1% had received prior methotrexate, csDMARDs and glucocorticoid medication, respectively (Table 1).

Table 1 Baseline demographics and disease characteristics of patients in the effectiveness analysis set and the high and low baseline adherence groups

Treatment Adherence

Based on the CQR score at baseline, 88 (52.4%) patients were classified as having high adherence. Baseline demographics and disease characteristics were generally comparable between patients in the high and low adherence groups, although the patients with high adherence were older and exhibited slightly higher disease activity (Table 1). Over time, the proportion of patients in the high adherence group increased to 60.9% at 6 months and 67.4% at 12 months (Figure 1).

Figure 1 Proportion of patients with high and low predicted adherence in the effectiveness analysis set based on CQR response at baseline, month 6 and 12. Abbreviations: CQR, Compliance Questionnaire in Rheumatology; High, high predicted adherence; Low, low predicted adherence.

Impact on RA Flares

At 6 months, the mean (SD) RA-FQ score was 23.8 (13.2) in the high baseline adherence group and 22.5 (13.1) in the low baseline adherence group (p=0.56); at 12 months, the corresponding mean (SD) scores were 19.9 (13.5) and 20.8 (12.9) (p=0.70), respectively (Figure 2A). RA flares were reported in 31.8% of patients at month 6 and 26.2% at month 12, with a numerically lower incidence observed in the high adherence group compared to the low adherence group. Specifically, at 6 months, 28.2% of patients in the high adherence group reported flares, versus 35.7% in the low adherence group; at 12 months, these rates were 24.7% and 30%, respectively (Figure 2B).

Similar to the primary analysis using baseline-predicted adherence, no meaningful differences in the mean (SD) RA-FQ scores were observed in the high vs low visit-predicted adherence groups at 6 months (24.3 [12.8] vs 20.9 [12.9]; p=0.12], and at 12 months [20.6 (13.5) vs 19.7 (12.7); p=0.71; Figure S1A]. In terms of disease flaring, 31.8% of patients reported a flare at month 6 and 26.2% at month 12, with slight numerical but not statistical differences observed between the high and low visit-predicted adherence groups at both time points: 32.6% vs 28.6% at month 6, and 28.7% vs 23.9% at month 12 (Figure S1B).

Figure 2 Disease flaring by baseline predicted adherence in low and high adherence groups by (A) RA-FQ mean score and (B) proportion of patients reporting flare. Values above bars are mean differences between low and high (95% CI). Abbreviations: CI, confidence interval; High, high predicted adherence; Low, low predicted adherence; RA-FQ, Rheumatoid Arthritis Flare Questionnaire.

Impact on Disease Activity

Disease activity levels at 6 months were comparable in the high and low baseline-predicted adherence groups based on CDAI [13.8 (10.8) vs 12.9 (10.5)], MDGA [4.5 (2.7) vs 4.3 (2.7)], and PtGA [2.9 (2.6) vs 2.6 (2.4)] (Figure 3). At 12 months, patients in the high baseline-predicted adherence group had significantly lower CDAI (8.4 [7.3] vs 11.9 [10.2]; Δmeans [95% CI]: −3.5 [−6.9, −0.1]) and numerically lower levels of MDGA (3.3 [2.5] vs 3.7 [2.74]) and PtGA (1.8 [1.9] vs 2.2 [2.5]).

Figure 3 Clinical disease activity by baseline predicted adherence in low and high adherence groups measured with (A) CDAI, (B) MDGA, and (C) PtGA. Values above bars are mean differences between low and high (95% CI). Abbreviations: CDAI, clinical disease activity index; High, high predicted adherence; Low, low predicted adherence; MDGA, physician global assessment; PtGA, patient global assessment; SD, standard deviation.

No differences at either 6 or 12 months were observed in the high vs low adherence group when using the visit-predicted adherence (Figure S2).

Impact on Glucocorticoid Use

Comparable rates of glucocorticoid use were observed in the high vs low baseline-predicted adherence groups at month 6 (18.2% vs 20%; p=0.76) and month 12 (15.9% vs 18.8%; p=0.63; Figure 4A). However, a lower mean (SD) dose of oral glucocorticoids was used in high vs low baseline-predicted adherence patients in the first 6 months (6.4 [4.2] vs 12.6 [16.0] mg prednisone equivalent), as well as between months 6 and 12 (6.0 [3.8] vs 9.5 [11.9] mg prednisone equivalent).

Figure 4 Proportion of patients using glucocorticoids in low and high adherence groups by (A) baseline predicted and (B) visit predicted adherence. Values above bars are differences in proportions between low and high (95% CI). Abbreviations: CI, confidence interval; High, high predicted adherence; Low, low predicted adherence.

In the visit-predicted adherence groups, a lower rate of glucocorticoid use was observed in the high vs low adherence groups at 6 months (16.3% vs 30.5%; p=0.04) but not at 12 months (23.2% vs 21.7%; p=0.85; Figure 4B).

Impact on AEs

A small numerical difference in reported AEs was observed between baseline-predicted adherence groups, with AEs reported for 44.3% of patients in the high adherence group and 53.8% in the low adherence group (Figure 5A). No differences in AE incidence were observed in high vs low predicted adherence at 6 and 12 months (Figure 5B and C).

Figure 5 Proportion of patients reporting treatment emergent adverse events in low and high adherence group (A) by baseline predicted adherence (B) from baseline to the first 6 months by month 6 predicted adherence (C) from month 6 to month 12 by month 12 predicted adherence. Values above bars are differences in proportions between low and high (95% CI). Abbreviations: AE, adverse event; CI, confidence interval; Discont, discontinuation; Gol, golimumab; High, high predicted adherence; Low, low predicted adherence; SAE, serious adverse event.

The most commonly reported AEs by preferred term were drug ineffective (12.6% of patients), rheumatoid arthritis (8.8% of patients), and arthralgia (4.7% patients), while the most common serious AEs were appendicitis and pneumonia, each occurring in 0.9% of patients (Table S1).

Discussion

In this prospective, multicentric, non-interventional study of patients with RA treated with golimumab in routine care across Canada, 52% to 67% of patients were classified in the high predicted adherence group during the 12 month period of the study. These results align with previous studies showing similar adherence rates to golimumab in real-world settings.^37–39 Higher RA-FQ scores were observed at baseline compared with the 6 month and 12 month visits, irrespective of adherence group, consistent with prior findings that disease flares are more common at the beginning of biologic treatment. The decline in RA-FQ scores over 12 months was numerically greater among patients in the high predicted adherence group, suggesting that timely adherence to biologic therapy may reduce flare incidence and, consequently, the need for glucocorticoids. Patient support programs (PSPs) have been developed to educate patients on the importance of adherence, address concerns about treatment that could act as a barrier to uptake, and provide reminders to improve medication-taking behavior. Our findings underscore the importance of PSPs in improving patient outcomes, particularly during the early stages of treatment.

Disease flare-ups, as measured by the RA-FQ questionnaire, were comparable between patients in the low and high baseline adherence groups. These results suggest that variations in the adherence to golimumab treatment may not have significant impact on disease control. Notably, a greater proportion of patients in the low adherence group used glucocorticoids, which may have contributed to the comparable rate of flares between adherence groups. Furthermore, some differences in CDAI levels but not MDGA or PtGA were observed at 12 months in patients with high vs low baseline-predicted adherence. Physicians may have used glucocorticoids as a compensatory mechanism to manage disease activity in patients with lower adherence to golimumab and/or early symptoms of disease flares. Previous studies have shown an association between adherence to biologic therapy and steroid use.⁴⁰ However, long-term glucocorticoid use is not a sustainable treatment strategy, due to the risk of severe and potentially irreversible adverse effects. Thus, high adherence to bDMARDs can effectively control inflammation, reducing the need for glucocorticoids and mitigating their associated risks. Alternatively, the observed association between low adherence and glucocorticoid use may be influenced by adverse reactions associated with glucocorticoids or by the effects of polypharmacy. Further studies are needed to confirm this relationship and explore strategies to enhance adherence to biologic therapies, thereby reducing reliance on glucocorticoids and improving long-term disease outcomes.

No differences were observed in the incidence of treatment-emergent AEs/SAEs between low and high adherence groups, which aligns with the comparable rates of flare-ups. Approximately 68% of patients receiving golimumab have previously reported at least one AE, and 4.4% have reported at least one SAE, which is consistent with the safety profile of other TNF-α inhibitors.⁴¹

Real-world studies such as GO FAR provide valuable insights that complement data from clinical trials, offering a more comprehensive understanding of RA treatment effectiveness. Adherence rates and patient behavior in the real-world may differ significantly from clinical trials due to the inclusion of more diverse patient populations and the absence of strict visit schedules, and understanding these differences is essential for optimizing treatment strategies. The GO FAR study, thus critically contributes to the understanding of golimumab effectiveness in real-world practice.

The GO FAR study did not assess reasons for low adherence such as comorbid conditions, concomitant medication use, access to healthcare services, mental health, patient education, disease activity, complex treatment regimens, incidence of AEs, and other lifestyle factors, and did not account for potential confounding variables that could influence the incidence of disease flares such as rheumatoid factor and anti-citrullinated potein positivity. Furthermore, actual adherence to treatment was not measured; only predicted adherence was reported, which may not fully reflect real-world adherence behavior or its variability over time. Similarly, given that disease flaring was assessed by the patient, misclassification bias of the outcome due to co-occurring conditions such as osteoarthritis and fibromyalgia, cannot be excluded. The study was not powered for inferential hypothesis testing and conclusions were constrained by the limited sample size and the short timeline of the study; it is possible that a larger and/or longer study would have allowed the detection of a significant difference between the high and low adherence groups. Finally, despite reasonable efforts to contact patients, approximately 20% did not have follow-up assessments, which may have introduced selection bias.

Conclusion

In this real-world study, no significant differences were observed in RA-FQ score, the proportion of patients reporting disease flares, or AE rates between patients with RA with low vs high predicted adherence to golimumab treatment. The more frequent use of glucocorticoids within the first 6 months in patients with low predicted adherence warrants further investigation to better understand this association. These findings underscore the importance of identifying barriers to adherence and implementing strategies, such as PSPs, to improve adherence to bDMARDs, particularly in the early stages of treatment. Enhancing adherence could ultimately help to reduce glucocorticoid dependence and improve long-term outcomes in RA management.