Background
Crohn’s disease (CD) and ulcerative colitis (UC) are chronic progressive inflammatory diseases of the gastrointestinal (GI) tract, collectively referred to as inflammatory bowel disease (IBD). While CD can affect the entire GI tract, UC mostly remains limited to the colonic mucosa [
1]. IBD is estimated to affect more than 6.8 million individuals globally [
1]. In the United States (US), 3.1 million adults are known to be affected (2015 estimates); age-standardized prevalence in the US was estimated to be 464.5 patients per 100,000 population [
1,
2]. IBD imposes significant health and economic burden on communities worldwide, and substantially impacts patients’ quality of life [
1].
Conventional treatments for both CD and UC include corticosteroids, aminoacylates, antibiotics, and immunomodulatory drugs [
3]. Advanced therapies, such as tumor necrosis factor inhibitors (TNFi), interleukin (IL) 23 p40 inhibitors, integrin inhibitors or Janus kinase inhibitors (only for UC) are usually reserved for patients with moderate-to-severe IBD, who do not adequately respond to conventional therapies [
3,
4]. Despite the availability of newer therapies, TNFi drugs remain the first-line treatment of moderate-to-severe IBD. These drugs have been shown to be well-tolerated and effective for inducing and maintaining the remission of CD and UC [
5,
6]. However, about 10–40% of patients with IBD are primary non-responders [
7] and up to 50% of patients experience secondary loss of response after 12 months on therapy [
8]. Inadequate response to advanced therapies indicates a need for newer therapies to improve the management in patients with CD or UC. Therefore, investigating treatment effectiveness in patients with CD and UC would help clinicians make more informed treatment decisions and contribute to value-based reimbursement models of care.
In the recent past, various algorithms have been used to assist researchers and clinicians in identifying patients with CD or UC, treatment patterns, treatment safety and healthcare resource utilization from insurance claims database studies [
9‐
13]. While effectiveness and safety outcomes associated with effective treatment in CD and UC have been studied using these types of data [
14,
15], methods of evaluating the inadequate response to medications in patients with CD and UC from claims databases are limited. The present study investigated the frequency of, and factors associated with, inadequate response over 1 year after advanced therapy initiation in adult patients from the US. Inadequate responses were identified by using a claims-based algorithm originally developed by Curtis et al. [
16], validated for rheumatoid arthritis and adapted for use in IBD.
Methods
Data source and population
This was a retrospective claim-based cohort study that utilized longitudinal claims data from the HealthCore Integrated Research Database® (HIRD®) from January 1, 2016 to August 31, 2019. The HIRD® contains data from January 2006 on patient enrollment, inpatient and outpatient medical care, prescription, and health care utilization. It is a large longitudinal medical and pharmacy claims database of health plan members comprising all regions of the US.
The data were accessed and used in full compliance with the relevant provisions of the Health Insurance Portability and Accountability Act. The study was conducted under the research provisions of Privacy Rule 45 CFR 164.514(e). Researchers’ access to claims data was limited to data stripped of identifiers to ensure confidentiality. An Institutional Review Board did not review the study since only this limited data set was accessed. This study was conducted in accordance with the ethical principles that have their origin in the Declaration of Helsinki and that are consistent with Good Pharmacoepidemiology Practices as well as legal and regulatory requirements.
Adult patients aged ≥ 18 years with CD (International Classification of Diseases, 10th Revision, Clinical Modification [ICD-10-CM] diagnosis codes: K50.x) or UC (ICD-10-CM diagnosis codes: K51.x) who initiated an advanced therapy during the index period of July 1, 2016 through August 31, 2018 were included in the study. Index date was defined as the first observed occurrence of a claim (medical or pharmacy) for any eligible advanced therapy during the index period. For patients who started more than one therapy, only the earliest one observed was used. Included patients were enrolled in commercial, Medicare Advantage, or Medicare Supplemental plus Part D insurance plans for ≥ 6 months before the index date (pre-index period) and ≥ 12 months after index date (follow-up period). Eligible patients were required to have ≥ 2 medical claims for CD or UC from a provider of any specialty at least seven days apart during the study period, of which ≥ 1 claim occurred during the pre-index period.
In this study, advanced therapies for CD included TNFi (adalimumab, certolizumab, infliximab) and non-TNFi (natalizumab, ustekinumab, vedolizumab). For UC, advanced therapies included TNFi (adalimumab, golimumab, infliximab), non-TNFi (vedolizumab; ustekinumab as a potential switcher but not index drug), and other therapies (tofacitinib). Conventional therapies included 5-aminosalicylic acid derivatives (mesalazine and sulfasalazine) and immunosuppressants (azathioprine, methotrexate, mycophenolate, cyclosporine, tacrolimus, 6-mercaptopurine).
Patients were excluded if they had claims for ≥ 1 advanced therapy during the 6-month pre-index period to identify new initiators of advanced therapy. Patients who had evidence for other autoimmune diseases including psoriasis, lupus, ankylosing spondylitis, psoriatic arthritis, or rheumatoid arthritis (defined as ≥ 2 claims on different dates for the same disease) were also excluded in order to avoid misclassification of the estimated response rate (e.g., related to non-adherence) due to multiple indications.
Criteria of inadequate response
The algorithm to identify inadequate response to index advanced therapy was derived from a claims-based algorithm originally developed by Curtis et al. [
16] and validated for rheumatoid arthritis. The first claim for advanced therapy is set as index date. Some modifications were made to the algorithm for UC and CD. The absence of all criteria listed in Table
1 denoted adequate response (stable disease); presence of one or more of them denoted inadequate response. For example, low index therapy adherence reflects inadequate response. All criteria were calculated based on the 1-year follow-up period for each patient. Details of the algorithm used are presented in Additional file
1: Table S1. In brief, the parameters of the algorithm included low adherence (defined as proportion of days covered [PDC] < 80%), switched/added new advanced therapy/new biologic, added a new conventional therapy, increased dose/frequency of advanced therapy/biologics, addition or dose increase of oral glucocorticoids, used a new pain medication, or had surgery for UC or CD.
Table 1
Inadequate response criteria evaluated over 1-year follow-up for both Crohn’s disease and ulcerative colitis
Criteria based on the reference algorithm [ 16] |
Low adherence to index advanced therapy (defined as proportion of days covered [PDC] < 80%) |
Switch/add non-index advanced therapy |
Add new conventional therapy (methotrexate, sulfasalazine, and others) |
Dose or frequency increase of index advanced therapy (> 20% higher than the index dose) |
Addition or dose increase of oral glucocorticoid |
Additional criteria for this study |
Use of pain medication classa not observed at pre-index period |
Use of surgery (Current Procedural Terminology codes for surgery are presented in Additional file 1: Table S2) |
Variables measured
Patient characteristics (age, sex, geographic region, Quan-Charlson Comorbidity Index [QCI] score, and specific comorbidities), provider specialty, and prior/historical treatments were assessed during pre-index period or on the index date. Prior “historic exposure” was defined as any claim from start of the patient’s continuous plan enrollment time up until the beginning of the 6-month pre-index period. Inadequate response to advanced therapies was recorded during 1-year follow-up and compared between patients receiving TNFi and non-TNFi.
Statistical analysis
Descriptive statistics including mean, standard deviation (SD), median, and absolute/relative frequencies for continuous and categorical data, respectively, were reported. Patient characteristics were statistically compared between responders and inadequate responders using Chi-square tests or Fisher’s exact tests for dichotomous variables, and t-tests or Wilcoxon tests for continuous variables. Characteristics during pre-index or on index date associated with inadequate response to the index advanced therapy were identified by a multivariable logistic regression model using a stepwise selection with entrance and exit p value cut-offs of 0.15. Index drug class, age, and gender were included in the model a-priori. Variables with prevalence rate < 1% in any group (responder or inadequate responder) were excluded from model selection. Goodness-of-fit statistics including C-statistic (higher score preferred) and Hosmer–Lemeshow test (p value > 0.05 preferred) were reported for each logistic regression. An alpha level of 0.05 was considered as statistically significant without any adjustments made for multiple comparisons. Sample selection and creation of analytic variables were performed using the Instant Health Data platform (Panalgo, Boston, MA). Statistical analyses were performed using SAS version 9.3 (SAS Institute Inc., Cary, NC, USA).
Discussion
In this retrospective claim-based study, an algorithm was developed to assess the extent of inadequate response to prescription treatments for the management of CD and UC. Adherence, increase in dose/frequency, and addition of new medications were the key indicators of inadequate response in this study. Results showed that most of the patients with CD or UC had initiated TNFi as first advanced therapy (~ 80%), of which ~ 50% started with adalimumab. However, a considerable proportion (> 60%) of patients did not adequately respond during the 1-year follow-up. The majority of both CD and UC inadequate responders were classified as such based upon low adherence to their index treatment (> 40%). The next most common reasons for inadequate response were switching to or adding new advanced therapy and increasing dose of existing therapies or glucocorticoids.
Results of this study are consistent with those observed in claims-based algorithm studies in rheumatoid arthritis [
16] and in ankylosing spondylitis and psoriatic arthritis [
17]. Approximately 70% of patients with rheumatoid arthritis, ankylosing spondylitis and psoriatic arthritis responded inadequately to advanced biologic treatments, where the most common reason for inadequate response was low adherence to the index medication [
16,
17]. Adherence to medication is a major problem in the management of IBD as well which can lead to adverse clinical outcomes including an increase in disease activity, relapse, and loss of response to TNFi [
18]. We used low adherence as a proxy for inadequate response to treatment as patients are unlikely to persist taking a medication that is ineffective or causes intolerance or adverse events, any of which may result in lack of effectiveness. Switching to advanced therapy, adding a new conventional therapy, increasing dose or frequency of medication including oral glucocorticoid, use of pain medication, are other prominent indicators to show that current therapies are not sufficient. In a proportion of patients, need for surgery may also indicate the insufficiency of current therapies. Inadequate response to TNFi typically leads to discontinuation of treatment. Other studies using real-world data showed that approximately half of patients with IBD receiving infliximab or adalimumab discontinued treatment during 1-year follow-up, and a substantial percentage were switched to a nonbiologic [
10,
19]. Moreover, in patients initiating a new biologic therapy for IBD, the likelihood of providers tapering therapy within the first year due to adequate response, such that it would result in low measured adherence, is less.
Inadequate response to advanced therapy cannot be explained by biology alone; algorithm-based studies are required to identify factors influencing inadequate response. In our study, factors influencing inadequate response in patients with CD included being female, historical use of TNFi, and having a consumer-driven health plan. While in patients with UC, these factors included baseline use of TNFi and higher QCI score. Consistent with the observations from some clinical trials, such as the one comparing infliximab plus azathioprine combination therapy [
20], patients taking any conventional therapy (CD) or azathioprine (UC) in the pre-index period in our study were more likely to be responders. A similar finding was also reported in a retrospective cohort study based on German claims data: Patients with UC who received conventional therapies at index showed less likelihood of experiencing inadequate response versus patients who were not on conventional therapy (Hazard ratio, 0.73; 95% CI 0.59–0.90); initial concomitant use of conventional therapies was associated with fewer dose augmentations [
21].
To our knowledge, this is the first study to use a claims-based algorithm to serve as a proxy for the clinical effectiveness of treatments for CD and UC in the United States. Results of the study suggest that there is a large unmet need in the management of CD and UC for more effective therapies and disease management strategies which can sustain remission. In addition, this study showed that algorithms are a promising proxy to investigate inadequate response to advanced therapies utilizing claims databases, which could be useful for value-based contracts and other innovative reimbursement schemes relevant for health plans and other stakeholders with access to claims data.
Our study also has some limitations associated with claims-database analysis. Patients identified in this study may not be representative of the US population who receive health care through government organizations or who lack health insurance. The relationships described between baseline patient characteristics and responder status represent associations rather than causal chains; several important confounders, such as disease activity metrics (e.g., CDAI) or provider behaviors, are unavailable or incomplete in claims data. For patients who used multiple advanced therapies over time, we focused on the first of them to assess adequate response and did not examine subsequent treatments’ response rate. Prescription claims include medications dispensed by pharmacies, and do not necessarily reflect the actual consumption by patients. The possibility of coding errors associated with incorrect diagnoses cannot be completely avoided. Furthermore, the modified algorithm used in this analysis has yet to be validated against clinical UC or CD disease activity measures or biomarkers. The therapeutic landscape of CD and UC is constantly changing with the introduction of novel treatments whose indications also change over time; therefore, the use of therapies as assumed in this study may not fully reflect actual indications at each time point of the study period. Nevertheless, this study provides important insights into the ways in which biologic agents are currently used in clinical practice. Future studies are warranted for the validation, augmentation, and application of these algorithms in CD and UC.
Open AccessThis article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit
http://creativecommons.org/licenses/by/4.0/. The Creative Commons Public Domain Dedication waiver (
http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.