Switching TNFα inhibitors: Patterns and determinants

Abstract The aim of this study was to assess switching patterns and determinants for switching in patients initiating TNFα inhibitor (TNFα‐i) treatment. Patients were included who started TNFα‐i treatment between July 1, 2012 and December 31, 2017, from three Dutch hospitals, and were diagnosed with rheumatic diseases (RD), inflammatory bowel disease (IBD), or psoriasis. Outcomes were switching, defined as initiating another biological; switching patterns including multiple switches until the end of follow‐up; determinants for first switch, assessed using multivariate logistic regression. A total of 2228 patients were included (median age 43.3 years, 57% female), of which 52% (n = 1155) received TNFα‐i for RD, 43% (n = 967) for IBD, and 5% (n = 106) for psoriasis. About 16.6% of RD patients, 14.5% of IBD patients, and 16.0% of psoriasis patients switched at least once, mainly to another TNFα‐i. TNFα‐i dose escalation (OR 13.78, 95% CI 1.40–135.0) and high‐dose corticosteroids initiation (OR 3.62, 95% CI 1.10–12.15) were determinants for switching in RD patients. TNFα‐i dose escalation (OR 8.22, 95% CI 3.76–17.93), immunomodulator initiation/dose escalation (OR 2.13, 95% CI 1.04–4.34), high‐dose corticosteroids initiation (OR 6.91, 95% CI 2.81–17.01) and serum concentration measurement (OR 5.44, 95% CI 2.74–10.79) were determinants for switching in IBD patients. Switching biological treatment occurred in about one in six patients. RD patients with TNFα‐i dose escalation and/or high‐dose corticosteroids initiation were more likely to switch. IBD patients with TNFα‐i or immunomodulator initiation/dose escalation, high‐dose corticosteroids initiation or serum concentration measurement were more likely to switch. These findings might help clinicians anticipating switching in TNFα‐i treatment.


| INTRODUC TI ON
Tumor necrosis factor (TNF) α inhibitors have revolutionized the treatment of several immune-mediated inflammatory diseases (IMID), such as rheumatic diseases (RD), inflammatory bowel disease (IBD), and psoriasis. Five TNFα inhibitors are currently available for patient care in Europe: adalimumab and infliximab are, among others, approved for RD, IBD, and psoriasis, etanercept and certolizumab pegol are approved for RD and psoriasis and golimumab is approved for RD and IBD. [1][2][3][4][5] TNFα inhibitors are advised as first-line biological treatment in IMID when conventional immunomodulator treatment, such as methotrexate or azathioprine, does not achieve sufficient clinical benefit. TNFα inhibitors may improve clinical signs and symptoms and make low disease activity and remission realistic objectives for patients suffering from IMIDs. [6][7][8][9][10] However, although many patients benefit from TNFα inhibitor treatment, several patients experience a lack of efficacy or bothersome side effects. 11,12 For those patients, switching to another biological drug, or to a Janus Kinase (JAK) inhibitor, is recommended. The choice for switching to a second TNFα inhibitor or to a biological drug belonging to another mechanistic class depends on the indication of use and on the reason for switching. For example, the IBD guideline advices on switching based on response to the TNFα inhibitor, drug concentrations and presence of antibodies, 6 whereas the rheumatoid arthritis (RA) guideline does not provide a strategy in choosing between another TNFα inhibitor or to a biological drug belonging to another mechanistic class. 7 In clinical practice, switching to another biological treatment frequently occurs. A previous study in RD patients showed that 67% of the patients remained persistent users (percentage of patients on the same biological drug after 12 months of initiation) 13 of their first TNFα inhibitor, 13% had switched to another biological drug (other TNFα inhibitor or biological belonging to another mechanistic class) and 20% had discontinued biological treatment. 14 A study in IBD patients reported a 1-year persistence of TNFα inhibitors of 48.5% for CD and 44.8% for UC. Switching to another biological drug occurred in 19.4% of CD and 20.3% of UC patients. 15 One-year persistence was higher in psoriasis patients; 77.4% of patients were persistent users, 17.5% had switched to another biological drug and 5.1% had discontinued biological treatment. 16 Several determinants for TNFα inhibitor treatment discontinuation in IMID have been identified. For example, women are at a 1.3 to 1.8 times higher risk for discontinuation than men. [17][18][19] Concomitant use of methotrexate decreases the risk of discontinuation in RD patients by 22%, 20 and in psoriasis patients by 66.2%. 19 The risk of TNFα inhibitor treatment discontinuation additionally increases by 1.4-6.0% per year with increasing age. 19,20 The aforementioned studies mainly focused on biological treatment discontinuation and determinants thereof, or only on the first biological treatment switch, in a single indication. However, little has been studied on the patterns of multiple switches of biological treatment across multiple indications and on determinants specifically for switching biological treatment. Data on switching patterns, including information on the type of biological drug, and more knowledge on determinants for switching may support more efficient treatment with biological drugs.
The aim of this study was to assess switching patterns and determinants associated with switching in patients who initiated TNFα inhibitor treatment for IMID between 2012 and 2017.

| Design and setting
This cohort study included patients from three large hospitals in the Netherlands: the Spaarne Gasthuis, the Medisch Spectrum Twente (MST), and the University Medical Center Utrecht (UMC Utrecht).
The Spaarne Gasthuis and the MST are both large teaching hospitals; the UMC Utrecht is an academic teaching hospital.
Dispensing data from the outpatient pharmacy from the Spaarne Gasthuis, the MST and the UMC Utrecht were obtained from CompuGroup Medical (CGM). Hospital and laboratory data from the Spaarne Gasthuis and the MST were obtained directly from the hospital and pharmacy information systems, that is, through Epic (Spaarne Gasthuis) and Vipharma, and GLIMS (MST).
Hospital and laboratory data from the UMC Utrecht were obtained from the Utrecht Patient Oriented Database (UPOD). UPOD is an infrastructure of relational databases comprising data on patient characteristics, hospital discharge diagnoses, medical procedures, medication orders, and laboratory tests for all patients treated at the UMC Utrecht since 2004. UPOD data acquisition and management was in accordance with current regulations concerning privacy and ethics. The structure and content of UPOD are described in more detail elsewhere. 21 Since January 1, 2012, all outpatient-administered biological drugs have been exclusively dispensed by the outpatient pharmacy of the hospital where a patient is treated due to reimbursement regulations in the Netherlands. Consequently, the outpatient pharmacy contains a complete overview of all biological drugs used in the home setting. 22

| Study population
All new users of TNFα inhibitors (etanercept, infliximab, adalimumab, certolizumab, and golimumab), treated for RD, IBD, or psoriasis, be- For all patients included, date of birth, gender, treatment indication defined as RD, IBD, or psoriasis (derived from the specialism of the prescriber of the TNFα inhibitor), type of biological drug, dose and dosing regimen, dispensing date (outpatient biological drugs), or administration date (biological drugs administered at the hospital ward), having TNFα inhibitor serum concentration or anti-drug antibodies measured, use of immunomodulators and high-dose corticosteroids were collected.

| Switching patterns
For each patient treatment episodes were constructed, defined as the duration of use of a single type of biological drug over time. For outpatient biological drugs, this was the time between the first dispensing of that biological drug until the end of the duration of the last dispensing. For biological drugs administered at the hospital ward, this was the time between the first administration of that biological drug until the last administration plus the standard dosing interval. A maximum permissible gap of 90 days (outpatient biological drugs) or twice the length of the standard dosing interval (biological drugs administered at the hospital ward) was allowed to correct for potential temporary treatment interruptions (e.g., due to surgery or infections).
From these treatment episodes, switching patients were identified, defined as starting a treatment episode of another biological drug (or a JAK inhibitor) within the maximum permissible gap of the previous one. In addition, patients who did not switch were identi- The following biological drugs were included in the analysis: abatacept, anakinra, belimumab, brodalumab, canakinumab, guselkumab, ixekizumab, rituximab, sarilumab, secukinumab, ustekinumab, vedolizumab (biological drugs), and baricitinib and tofacitinib (JAK inhibitors).

| Determinants for switching
Determinants for switching from the first TNFα inhibitor to another biological drug (or JAK inhibitor) were explored in a nested casecontrol analysis. Cases were defined as patients who switched at least once during follow-up. Patients who did not switch were included as controls. Up to four controls were randomly selected for each case by using incidence density sampling. Cases and controls were matched by the type of TNFα inhibitor at the index date, treatment in the same hospital and the date of initiation of treatment (± 3 months). Controls could be selected more than once, and patients who became cases could be selected as controls at earlier time points.
The following determinants for switching were explored: age at index date (continuous, years); gender (categorical); dose escalation of TNFα inhibitor within 60 days before the switch (yes or no); initiation or dose escalation of treatment with immunomodulator within 60 days before the switch (yes or no); initiation of treatment with high-dose corticosteroids within 60 days before the switch (yes or no); and TNFα inhibitor serum concentration measurement (or antidrug antibodies for the specific TNFα inhibitor) within 60 days before the switch (yes or no).
Dose escalation of outpatient-administered TNFα inhibitors was defined as having any increase in dose or shortening of dosing interval of the index TNFα inhibitor in the 60-day period before the switch. For infliximab, dose escalations were defined as either a minimum 25% increase 23 in dose in the 60-day period before the switch or an increase in dosing interval of a minimum of 8 days to overcome rounding up and dose increases due to an increased weight of the patient or logistic issues.

| Data analysis
Descriptive statistics were used to present the baseline characteristics of the patients. Treatment patterns were presented in a Kaplan-Meier curve for persistent use of index TNFα inhibitor. Switch to another biological drug or JAK inhibitor and discontinuation of index TNFα inhibitor without switching were presented in cumulative incidence curves.
Determinants for switching were analyzed with conditional logistic regression, stratified per indication. All possible determinants were first analyzed univariately, and determinants with a p-value of <.1 in the univariate analysis were analyzed using multivariate conditional logistic regression.
In a sensitivity analysis, the impact of changing the definition of new users was assessed by only including patients who did not use any biological drug for RD, IBD, or psoriasis 12 months prior to the date of inclusion. This was done to discriminate prevalent users of TNFα inhibitor from new users.
Data were analyzed using R version 3.6.1 (R Foundation for Statistical Computing).

| RE SULTS
A total of 2228 patients were included, with a median age of 43.3 years, 56.6% of the patients being female (Table 1). Of the in-

| Switching patterns
Approximately 16% of patients switched from the initial TNFα inhibitor to another biological drug, which was comparable across indications, as shown in Figure 1

| Determinants for switching
The assessment of determinants showed that patients suffering from RD who had a dose escalation of their TNFα inhibitor (OR 13.78, 95% CI 1.40-135.0) or initiated high-dose corticosteroid treatment (OR 3.62, 95% CI 1.10-12.15) were more likely to switch biological treatment ( Table 2).  (Table 3).

TA B L E 1 Patient characteristics of the cohort at baseline
The study did not include a sufficient number of cases with psoriasis to allow for a case-control analysis in this group of patients.
The sensitivity analysis produced similar results both for the treatment patterns and determinants analysis as the main analysis (Table S1-S3).

| DISCUSS ION
In this study, we investigated switching patterns and determinants for switching in patients with RD, IBD, or psoriasis initiating treat- were associated with switching as well in IBD patients. These factors are possible markers for disease worsening and, consequently, switching. In RD and IBD patients, disease flares are often treated by initiating high-dose corticosteroids or immunomodulators. 28,30 However, in contrast to RD, if an IBD patient experiences a flare, measuring serum drug concentrations (and anti-drug antibodies), and intensifying the dose are also commonly used strategies. 29  We additionally did not have information on the reason for switching to another biological drug or the discontinuation of biological treatment. As the reason for switching treatment influences the choice of second-line biological drug, this information might add to the understanding of the switching patterns seen.
Finally, as the indication for TNFα inhibitor treatment was derived from the specialism of the prescriber, we were unable to make distinctions between the individual RD, such as RA, AS, psoriatic arthritis, and juvenile idiopathic arthritis. RA and AS are the most prevalent rheumatic diseases, 31 we believe that these are also the most prevalent types of RD in our cohort. As the biological treatment strategies in RA and AS are comparable and there are no differences in reimbursement regulations between these indications, we believe that aggregating all types of RD has little impact on our results.
In conclusion, this large study of real-life data on biological use demonstrated-specific switching patterns of patients who ini-

D I SCLOS U R E
The authors declare they have no conflict of interest.

E TH I C A L A PPROVA L
The Institutional Review Board of the Spaarne Gasthuis approved the study protocol. This study was additionally reviewed by the Medical Research and Ethics Committee UMC Utrecht (protocol reference number 19-049/C), which concluded that ethical approval was not required, as the study did not fall under the scope of the Dutch Medical Research Involving Human Subjects Act.

DATA AVA I L A B I L I T Y S TAT E M E N T
Research data will not be shared.