Background

Ankylosing spondylitis (AS) is a chronic inflammatory rheumatic disease characterized by inflammatory back pain that can lead spinal ankylosis [1]. The introduction of anti-tumor necrosis factor (TNF) agents has significantly transformed the management of AS patients, especially for patients who do not respond well to or cannot tolerate non-steroidal anti-inflammatory drugs (NSAIDs) [2]. Currently, six anti-TNF agents have been approved for treatment of AS: adalimumab, etanercept, infliximab, golimumab, and certolizumab. Despite the benefits of these anti-TNF agents in controlling inflammation and symptoms, their effectiveness and persistence can vary, with some patients experiencing therapeutic failure, adverse events, or discontinuation of treatment for various reasons.

Measuring drug persistence (also referred to as drug survival), the time from treatment initiation to discontinuation, offers essential insights into their safety and effectiveness [3]. In general, drug retention rates for anti-TNF treatments tend to be higher in AS patients compared to individuals with rheumatoid arthritis (RA) [4,5,6]. Data from 12 European registries, encompassing 24,195 axial spondyloarthritis (axSpA) patients who initiated their first anti-TNF treatment, revealed an overall 12-month retention rate of 80%, with individual registries ranging from 71 to 94%. Over 24 months, the overall retention rates were 73%, with individual registries showing retention rates varying from 71 to 94% [7]. Research on the long-term retention rates of first-line anti-TNF agents in AS is relatively scarce, and the available studies have shown varying results, including an eight-year anti-TNF retention rate of 57.2% in a study with 316 axSpA patients and a 10-year anti-TNF survival rate of 28.9% in another study with 231 AS patients [4, 8].

Awareness of individual variability in drug retention has naturally led to an increased interest in understanding the factors associated with the response to anti-TNF agents. Several factors have emerged as predictors of a favorable response to anti-TNF treatment and improved survival rates. These factors include male sex [9,10,11,12], young age [9, 10, 13], high baseline levels of inflammatory markers [10, 12,13,14], a shorter disease duration, human leukocyte antigen (HLA)-B27 positivity [9, 11, 13], and non-smoking status [15].

However, applicability of these previous studies has been limited, due to relatively short observation periods. It is crucial to acknowledge that reasons for discontinuation of anti-TNF agent treatment can vary between short-term and long-term treatment periods [8]. Recognizing the clinical differences between discontinuations due to clinical remission and those due to lack of efficacy, it becomes evident that a deeper investigation into the causes of discontinuation is necessary, rather than focusing solely on the retention rate.

Therefore, the objective of this study was to explore and identify factors associated with cause-specific discontinuation after long-term treatment with first-line anti-TNF agents in AS patients.

Methods

Patient enrollment

We conducted a retrospective cohort study including AS patients who initiated first-line anti-TNF treatment with either adalimumab, etanercept, infliximab, or golimumab between 2004 and 2018. All patients met the 1984 modified New York criteria for AS [16]. To focus on the cause-specific discontinuation of anti-TNF treatment in the long term, we excluded patients who stopped using the anti-TNF agent within two years of initiation. Patients with inflammatory bowel disease were also excluded.

The index date was defined as that when the anti-TNF agent was initiated. Patients were observed until their last visit, discontinuation or switching of the anti-TNF agent, or end of the study in September 2022.

Data collection

We collected information on various demographics, including age and smoking status, as well as AS-related clinical factors including symptom and disease duration, HLA-B27 positivity, peripheral arthritis, uveitis, use of NSAIDs, and the specific type of anti-TNF agent used. To assess disease activity, we recorded the serum concentration of C-reactive protein (CRP) and the Bath Ankylosing Spondylitis Disease Activity Index (BASDAI) score at the index date and at three months after initiating anti-TNF treatment. To evaluate structural damage, the modified Stoke Ankylosing Spondylitis Spinal Score (mSASSS) was used [17].

Discontinuation of treatment was defined as either stopping or switching anti-TNF therapy. The reasons for discontinuation were categorized as follows: (1) clinical remission, (2) loss of efficacy, (3) adverse events, and (4) other reasons, including loss to follow-up and issues related to cost or reimbursement. Given the retrospective nature of our study, a comprehensive chart review was conducted, and each case was evaluated by a rheumatologist for classification into one of these categories.

Statistical analysis

Enrolled patients were divided into four groups according to type of anti-TNF agent; adalimumab, etanercept, infliximab, and golimumab groups. Demographic and clinical characteristics of each group are described in a descriptive analysis and all data are shown as median interquartile range [IQR] or frequency (percentage). The four groups were compared using the Kruskal-Wallis test for non-normally distributed numerical variables and chi-square test and Fisher’s exact test for categorical variables.

Competing risks were taken into account for survival analysis. Gray’s test was performed to analyze differences in the cumulative incidence between each anti-TNF agent group. We used Cox cause-specific hazards models to identify factors associated with discontinuation of anti-TNF treatment for specific causes, including clinical remission, loss of efficacy, overall adverse events, infection-related adverse events, and other reasons. An exploratory analysis was performed via univariable Cox cause-specific hazard models to determine which factors should be included in multivariable modes. Significant clinical factors including age, sex, disease duration, and type of anti-TNF agent or factors at a P-value < 0.2 in the univariable models were considered as candidates for the multivariable model.

Statistical analyses were performed using SAS version 9.4 (SAS Institute Inc., Cary, NC, USA) and R version 4.2.2 (R Foundation for Statistical Computing, Vienna, Austria). A P-values < 0.05 was regarded as statistically significant.

Results

Demographic and clinical characteristics of patients

Among 556 screened patients, 122 who discontinued anti-TNF treatment within two years and five with inflammatory bowel disease were excluded, resulting in the inclusion of 429 AS patients for the analysis (Fig. 1). Table 1 provides a summary of the demographic and clinical characteristics of the enrolled patients. The median age of the patients was 32.1 (26.9–38.5) years, with 88.8% being male. The median symptom duration was 10.4 (4.9–15.4) years. The most commonly prescribed anti-TNF agent was etanercept (176, 41.0%), followed by adalimumab (121, 28.2%), infliximab (89, 20.8%), and golimumab (43, 10.0%). At baseline, the median BASDAI score was 7.0 (6.0–8.0), and the level of CRP was 2.1 (0.9–4.6) mg/dL. After three months of anti-TNF treatment, the median BASDAI score decreased to 2.8 (2.0–4.0), and the level of CRP decreased to 0.8 (0.8–0.8) mg/dL. The median overall survival duration for anti-TNF agent treatment was 10.6 (7.9–14.5) years. During the observational period, 103 (24.0%) patients discontinued first-line anti-TNF treatment, with reasons of loss of efficacy (36, 34.9%), clinical remission (31, 30.1%), other reasons (21, 20.4%), and adverse events (15, 14.6%).

Fig. 1
figure 1

Flow chart of patient selection

Ankylosing spondylitis, AS; TNF, tumor necrosis factor

Full size image
Table 1 Clinical characteristics of patients
Full size table

Cause-specific discontinuation of anti-TNF treatment

The overall survival duration differed among the anti-TNF agents, with adalimumab having a duration of 10.2 (7.8–12.0) years, etanercept 13.5 (10.2–16.7) years, infliximab 10.0 (7.4–12.5) years, and golimumab 7.7 (7.0-9.1) years (P < 0.001). The proportion of patients with uveitis was higher in the etanercept group (37.2% for adalimumab, 51.7% for etanercept, 40.5% for infliximab, and 23.3% for golimumab, P = 0.003). Other clinical characteristics, such as age, sex, baseline mSASSS, BASDAI, CRP, and smoking status, were comparable between the groups (Table 1).

The reasons for discontinuation varied depending on the anti-TNF agent used (P = 0.005). The most common reason for discontinuation of adalimumab was clinical remission (14, 45.2%), followed by other reasons (7, 22.6%), adverse events (6, 19.4%), and loss of efficacy (4, 12.9%). For etanercept, loss of efficacy (19, 38.0%) was the primary cause of discontinuation, followed by other reasons (12, 24.0%), clinical remission (11, 22.0%), and adverse events (8, 16.0%). In the case of infliximab, loss of efficacy (13, 68.4%) was the leading cause of discontinuation, followed by clinical remission (4, 21.0%), adverse events (1, 5.3%), and other reasons (1, 5.3%). Finally, for golimumab, the reasons for discontinuation were clinical remission (2, 66.7%) and other reasons (1, 33.3%).

The Kaplan-Meier survival curve for anti-TNF agents did not show significant differences among the agents (Fig. 2). Therefore, cause-specific discontinuation was compared based on the type of anti-TNF agent using the cumulative incidence function (Fig. 3). There were no significant differences in discontinuation due to clinical remission, overall adverse events, and other reasons (P = 0.083, P = 0.253, and P = 0.387, respectively). However, discontinuation associated with loss of efficacy differed significantly according to the type of anti-TNF agent used (P = 0.009).

Fig. 2
figure 2

Kaplan-Meier survival curve for anti-TNF agents

Full size image
Fig. 3
figure 3

Cumulative incidence function for individual anti-tumor necrosis factor agents

Full size image

Predictors for cause-specific discontinuation

We conducted univariable and multivariable cause-specific hazard models to identify predictors associated with cause-specific discontinuation. The multivariable cause-specific hazard models are presented in Table 2, while the univariable hazard models for each cause-specific discontinuation can be found in Supplementary Tables 1–5.

Table 2 Multivariable cause-specific hazard model
Full size table

In comparison to adalimumab, the use of etanercept was associated with a decreased likelihood of discontinuation due to clinical remission (Hazard ratio [HR] 0.45, 95% confidence interval [CI] 0.21–0.99, P = 0.048). Higher baseline BASDAI (HR 1.31, 95% CI 1.04–1.65, P = 0.023) and the use of infliximab (HR 4.53, 95% CI 1.45–14.16, P = 0.009) were predictors for discontinuation due to loss of efficacy. Increased age was a predictor for discontinuation due to infection-related adverse events (HR 1.07, 95% CI 1.02–1.12, P = 0.005). Current smokers were more likely to discontinue anti-TNF treatment due to other reasons (HR 6.22, 95% CI 1.82–21.28, P = 0.004).

Discussion

In this study using real-world data, we investigated cause-specific discontinuation in 429 AS patients receiving first-line anti-TNF treatment. We specifically focused on the long-term use of anti-TNF agents; therefore, only patients who continued treatment for at least two years were included; 21.9% (122 of 556) discontinued the therapy within the first two years. During an overall survival period of 10.6 (7.9–14.5) years, we observed a discontinuation rate of anti-TNF inhibitors in 24.0% of patients. The cumulative incidence function for cause-specific discontinuation was similar across types of anti-TNF agents, with the exception of discontinuation due to loss of efficacy. Specifically, the use of infliximab was associated with a higher risk of discontinuation due to loss of efficacy, as well as higher baseline BASDAI scores. The predictors for discontinuation varied depending on the specific cause, highlighting the complexity of treatment response and patient factors influencing therapy discontinuation. Etanercept use was associated with a decreased likelihood of discontinuation due to clinical remission compared to the use of adalimumab. Older age was associated with discontinuation caused by infections, while current smokers were more likely to discontinue anti-TNF treatment for other reasons.

Taking a closer look at the reasons for discontinuation, the discontinuation in the present study was primarily attributed to loss of efficacy (34.9%), followed by clinical remission (30.1%), other reasons (20.4%), and adverse events (14.6%). In contrast, a study conducted on Korean AS patients by Kim et al., reported a different ranking of reasons, with other reasons (32.6%) and loss of efficacy (32.6%) being the most common, followed by adverse events (23.6%), and clinical remission (11.2%) [11]. Remarkably, our study observed a relatively lower rate of discontinuation due to adverse events, which may be attributed to our research design; We excluded patients who discontinued anti-TNF agents within the first two years. It is important to note that adverse events, such as skin reactions, allergies, and severe infection, tend to be more frequent during the initial stages of anti-TNF therapy [18,19,20]. Additionally, our extended follow-up duration may have contributed to the relatively high rate of clinical remission. The longer follow-up duration likely led to more discontinuation attempts among individuals who had experienced sustained remission over an extended period [2].

Interestingly, our study demonstrated variations in discontinuation rates among anti-TNF agents. Infliximab showed a greater discontinuation rate due to loss of effectiveness compared to other anti-TNF agents, with similar results reported in previous studies [4, 5, 21]. This phenomenon may be attributed to drug antibodies that can potentially diminish efficacy by neutralizing the therapeutic effects or hastening its elimination from the body. Infliximab, in particular, is known to have the highest incidence of drug antibody formation among anti-TNF agents [22]. In contrast, etanercept, which is associated with the lowest generation of drug antibodies [22], was linked to a reduced likelihood of discontinuation due to clinical remission compared to the use of adalimumab. This observation may be explained, in part, by considerations related to uveitis. Uveitis incidence tends to increase with disease duration [23], and given the focus of our study on the long-term use of anti-TNF agents, the unfavorable results may be attributed to uveitis. Etanercept has demonstrated clear inferiority in controlling uveitis compared to other anti-TNF agents such as infliximab or adalimumab [24, 25]. While adalimumab does induce drug antibody formation, albeit to a lesser extent than infliximab [22], it is not associated with an increased risk of uveitis. Moreover, the less frequent dosing regimen of adalimumab compared to etanercept may facilitate easier tapering when patients achieve clinical remission. A multicenter retrospective study demonstrating the eight-year retention rate of first-line anti-TNF in SpA patients showed similar findings. Adalimumab demonstrated the highest persistence, while infliximab exhibited the lowest retention rate in AS, although this trend did not reach statistical significance [8].

Higher baseline BASDAI scores, consistently demonstrated in prior research, have been linked to an increased risk of discontinuation due to loss of efficacy [10, 26,27,28]. Conversely, elevated objective inflammatory markers like CRP and acute inflammation detected on MRI have consistently shown a strong association with a clinical response to anti-TNF agents or treatment continuation [10, 12, 14, 29,30,31,32]. Consequently, recent guidelines emphasize the importance of evaluating these objective signs of inflammation before initiating biologics, as patients displaying these indicators tend to experience more favorable treatment outcomes [2]. The BASDAI solely reflects the patients’ perspective, which may account for the disparities observed between BASDAI and objective inflammation markers in predicting treatment outcomes [33].

Smoking is a well-established poor prognostic factor in AS, linked to increased disease activity and radiologic progression [34]. An observational cohort study, based on the Danish nationwide DANBIO registry, revealed that both current and former smokers exhibited significantly poorer patient-reported outcomes at baseline, shorter treatment adherence, and inferior treatment responses compared to individuals who had never smoked [15]. Notably, our study found that current smokers were more likely to discontinue their anti-TNF treatment for other reasons including loss to follow-up, cost, or reimbursement issues. Given the strong recommendation for smoking cessation as a crucial component of AS patient treatment [35], we can interpret the link between current smoking and discontinuation for other reasons as a potential sign of suboptimal treatment adherence.

There was no significant difference in overall adverse events among the anti-TNF agent groups. However, older age was a significant risk factor for discontinuation due to infection-related adverse events, in line with the well-established understanding that older age is associated with an increased risk of infections [36]. Remarkably, our study revealed that only three cases of infections resulted in discontinuation of anti-TNF therapy. This finding aligns with prior researches, suggesting that TNF inhibitors are generally safe for AS treatment [37,38,39]. In a recent meta-analysis of 3,564 AS patients from 18 randomized controlled trials, anti-TNF treatment was not associated with a significant increase in the risk of serious infection or upper respiratory tract infection compared to the placebo group (Odds ratio [OR] = 1.44, p = 0.36 and OR = 1.22, p = 0.16, respectively). The sole statistically significant difference was the higher occurrence rate of nasopharyngitis among AS patients treated with anti-TNF agents, at 8.69%, in contrast to the placebo rate of 5.67% (OR = 1.53, p = 0.03) [40].

The present study had several limitations. Firstly, it is a single-center observational study, which may introduce selection bias and sampling bias. Second, our results may be influenced by reimbursement policies in South Korea. Third, the number of patients using Golimumab and the duration of observation were limited, primarily due to the delayed approval of Golimumab compared to other biologics. Golimumab has been reported to exhibit better drug persistence than other anti-TNF agents [11, 21, 41], but this association was not observed in our study. Fourth, in cases where patients sustained clinical remission, dose tapering or extension of the interval of anti-TNF use is common practice. These alterations might contribute to the persistence of treatment. However, we did not account for doses and intervals of anti-TNF treatment. Additionally, extra-musculoskeletal manifestations, including uveitis, can influence treatment decisions and outcomes. The retrospective nature of our study meant that pre-existing uveitis could not be clearly determined. However, considering that we frequently encounter situations where uveitis develops over the follow-up period, even if not initially present. We believe that our strategy better reflects the real-world scenarios faced by AS patients and their healthcare providers. Despite these limitations, the value of this study lies in its examination of extended real-world retention of anti-TNF agents and the analysis of cause-specific discontinuation. Through the exclusion of early discontinuations, it provides valuable insights into the extended phase of anti-TNF use. With a median overall survival of 10.6 years, it effectively bridges a preceding research gap and enables a more in-depth exploration of long-term anti-TNF users.

To conclude, our study of 429 AS patients on first-line anti-TNF treatment for at least two years reveals a favorable long-term treatment retention rate, with a 24.0% discontinuation rate over an overall survival period of 10.6 years. Predictors for discontinuation differed by cause, highlighting the complexity of treatment response and emphasizing the need for personalized approaches to treatment management. Further research is warranted to uncover the underlying mechanisms behind these discontinuation patterns and to refine strategies for optimizing patient outcomes in the realm of AS treatment.