Tuberculosis risk in patients with Crohn’s disease on biologics: a retrospective analysis of the Japanese Medical Claims Database
Article information
Abstract
Background/Aims
Treatment using tumor necrosis factor-α (TNF-α) inhibitors is one of the risk factors for active tuberculosis (TB) in patients with Crohn’s disease (CD). Biologics, such as ustekinumab (UST) and vedolizumab (VDZ), are less likely to cause opportunistic infections. However, large-scale studies for active TB and biologics other than TNF-α inhibitors are limited. We aimed to investigate the association between biologics and active TB utilizing a Japanese medical claims database.
Methods
We analyzed retrospectively the association of the risk of active TB development with treatment using TNF-α inhibitors and other biologics (UST and VDZ) in patients with CD using the Japanese Medical Data Vision (MDV) database between April 2008 and June 2022. The durations of each biologic and biologic-free treatment were calculated for each patient. Univariate and multivariate analyses were performed using the Cox proportional hazards model, with the utilization of biologics considered as time-dependent covariates.
Results
We included 28,811 patients with CD in MDV database. Finally, 17,169 patients were analyzed. In total, 7,064 patients were categorized as biologic-naïve, while 10,105 were classified as biologic-experienced. Seventeen patients developed active TB, including 7 on infliximab, 5 on adalimumab, and 5 on no biologics. None of the patients treated with UST and VDZ developed active TB. Multivariate analysis suggested that TNF-α inhibitors were the risk factors for active TB (hazard ratio, 3.66; P=0.020).
Conclusions
TNF-α inhibitors, but not UST or VDZ, are risk factors for active TB in Japanese patients with CD.
INTRODUCTION
Tuberculosis (TB), caused by Mycobacterium tuberculosis, remains a major cause of death worldwide. According to the World Health Organization, TB affects an estimated 10.6 million individuals, resulting in 1.4 million deaths in 2021 [1]. Furthermore, approximately one-quarter of the global population is estimated to have latent tuberculosis infection (LTBI), a condition characterized by no symptoms, despite a continuous immune response to M. tuberculosis [2]. LTBI reactivation occurs when the immune response decreases for various reasons, triggered by factors such as human immunodeficiency virus (HIV) infection or the use of immunosuppressive agents [3].
Crohn’s disease (CD) is characterized by idiopathic inflammation of entire digestive tract, marked by relapses and remissions, often necessitating long-term immunosuppressive therapies [4]. The pathogenesis of underlying CD is unclear; however, various cytokines, such as tumor necrosis factor-α (TNF-α), are associated with chronic intestinal inflammation. Infliximab (IFX) and adalimumab (ADA) are TNF-α inhibitors, which were approved for moderate-to-severe CD in Japan in 2002 and 2010, respectively. They are highly effective in induction and maintenance therapy, leading to a paradigm shift in treatment [5,6]. They play a central role in CD, despite the availability of new biologic agents.
TNF-α plays an important role in tumor and infection immunity. However, patients on TNF-α inhibitors are susceptible to opportunistic infections caused not only by viral and bacterial infections but also by antimicrobial infections. For instance, Keane et al. [7] demonstrated study an association between IFX and TB development in patients with rheumatoid arthritis and CD. Thereafter, numerous epidemiologic studies have reported on the risk of TNF-α inhibitors in the development of antimicrobial infections [8].
Ustekinumab (UST), a monoclonal antibody for interleukin (IL)-12 and IL-23, was approved for inflammatory bowel disease (IBD) in several countries in 2016. Notably, long-term studies have reported on the high safety profile and efficacy of UST [9]. Additionally, vedolizumab (VDZ), an antibody to integrin alpha 4 beta 7, was approved for moderate-to-severe CD in Japan in 2019. Clinical trials have demonstrated the long-term safety of VDZ for IBD [10]. Overall, these drugs are expected to have a lower risk of infection [11,12].
In patients with IBD treated with IFX or ADA, the TB risk is correlated with the local burden of TB [13]. Data from each region is essential for understanding the rate of TB reactivation by each biologic.
However, no large-scale studies have been conducted on the association between biologics other than TNF-α inhibitors and active TB development in patients with CD, particularly in Japan. Therefore, in this study, we aimed to investigate the risk of developing active TB in patients with CD treated with biologics using a Japanese clinical medical claims database.
METHODS
1. Study Design and Data Sources
In this retrospective cohort study, we used the real-world Japanese Medical Data Vision (MDV) database. The MDV database consists of health care claims data from acute care hospitals that use the Diagnosis Procedure Combination (DPC) system. As of March 2022, the database includes approximately 39.4 million inpatients or outpatients from 463 Japanese DPC hospitals, accounting for 26% of the Japanese hospitals that use the DPC system. Approximately 35% of the patients are aged ≥ 65 or over years. The database consists of information on patient characteristics, diagnosis, drug prescriptions, and medical procedures. The diagnoses were recorded according to the International Classification of Diseases, 10th Revision (ICD-10), and Japanese-specific codes. We could not trace the data continuously if patients were transferred to another hospital.
2. Study Population
We obtained the data of patients with a confirmed CD ICD-10 code (K-50) from the MDV database between April 2008 and June 2022. We extracted data on the age, sex, date of diagnosis, and date of prescription. However, since the MDV database includes data before CD diagnosis, we analyzed data after the CD ICD-10 code (K-50) was listed in the data of each patient.
The exclusion criteria were as follows: (1) patients with ICD-10 code (K-51) for ulcerative colitis (confirmed cases) due to diagnosis with indeterminate colitis; (2) patients without prescription data for CD, such as elementary diet, 5-aminosalicylic acid, steroids, thiopurines (azathioprine or 6-mercaptopurine), IFX, ADA, UST, or VDZ; (3) ≤ 12 weeks between the date when the CD ICD-10 code was first listed and the date of the last prescription; and (4) the date of active TB development, as the risk for TB could not be assessed for patients with short-term data. We defined 12 weeks as the threshold, as the longest administration interval for these medications is 12 weeks of UST. Risankizumab and upadacitinib had not been approved for CD in Japan as of June 2022.
3. Defining Active TB Development
We defined active TB development as patients who were newly prescribed 3 or 4 types of anti-TB drugs among isoniazid (INH), rifampicin, pyrazinamide, streptomycin and ethambutol, in addition to the TB ICD-10 codes (A15-19). Patients with an intestinal TB ICD-10 code (A18.3) (confirmed cases) were excluded because they may have been prescribed anti-TB drugs. Moreover, we excluded patients who were prescribed INH for purposes other than active TB treatment. In such cases, INH is prescribed to prevent LTBI reactivation. Therefore, we excluded these patients because we did not intend to investigate of the preventive effects of INH on LTBI reactivation.
4. Follow-up Duration
The duration of biologics use was calculated from the first to the last prescription day for each biologic. When another biologic was prescribed before the last prescription of 1 biologic, the duration of the biologic was calculated as the number of days other than the duration of the second biologic. In biologic-naïve patients, the entire data duration was biologic-free. In biologic-experienced patients, a period beyond 12 weeks before the first prescription or after the last prescription of biologics was defined as a biologic-free duration. Patients who developed active TB were followed up until it. For patients without active TB, data were obtained until the last day of prescription or June 2022.
5. Corticosteroid or Thiopurine Use
We investigated the use of corticosteroids or thiopurines to evaluate TB risk, particularly in relation to concomitant use with biologics. The patients were classified based on the use of systemic corticosteroids or thiopurines whether prescribed even once during both biologic and biologic-free. Notably, budesonide exerts fewer corticosteroid-related adverse effects because it is metabolized in the liver rapidly [14]; thus, it was not included as a systemic corticosteroid use in this study.
6. Comorbidities
We investigate whether diabetes mellitus ICD-10 codes (E10-E14) and HIV infection ICD-10 codes (B20-B24) to examine the influence of comorbid diabetes or HIV infection, which have been reported as risk factors for TB reactivation [15].
7. Statistical Analysis
Clinical characteristics are summarized as medians and interquartile ranges (IQRs) for continuous variables and as counts and percentages for categorical variables. Continuous variables were evaluated using the Wilcoxon rank-sum test, whereras categorical variables were evaluated using the chi-square test. We compared biologic-naïve and biologic-experienced patients based on their age at the first listing of a diagnosis of CD. Moreover, we compared the age of biologic-experienced patients at the first initiation of each biologic. When the number of events was 0, we calculated the 95% confidence interval (CI) based on the rule of 3 (i.e., 0–3/total person-years × 100,000) [16].
We conducted univariate and multivariate analysis using the Cox proportional hazards model, considering TNF-α inhibitors (IFX and ADA), UST, and VDZ as the time-dependent covariates. The duration of the biologic-free period was used as a reference. In the multivariate analyses, we included the sex, age, use of corticosteroids, and use of thiopurine as covariates, as these variables are the known risk factors for active TB development [17-19].
8. Ethical Statements
The data was anonymized, and informed consent was waived according to Ethical Guideline for Medical and Health Research Involving Human Subject in Japan.
RESULTS
1. Patient Demographics and Characteristics
We identified 28,811 patients from the database between April 2008 and June 2022. First, we excluded 11,123 patients based on the exclusion criteria. Second, 519 patients were excluded based on the development of active TB. Finally, we analyzed 17,169 patients (Fig. 1). Table 1 summarizes the characteristics of the eligible patients. The median age of all patients was 37 years, and 69.2% were men. In total, 7,064 patients were categorized as biologic-naïve, and 10,105 patients were categorized as biologic-experienced. The mean age of the biologic-naïve patients was significantly higher than that of the biologic-experienced patients (P<0.001). Moreover, the proportion of men was significantly higher among the biologic-experienced patients (P<0.001). Table 2 summarizes the characteristics of biologic-experienced patients. Approximately, 1,579, 335, and 45 patients had a prescription history of 2, 3, and 4 types of biologics, respectively. No patients with HIV infection ICD-10 codes were listed in the database.
Flowchart of study population. Only confirmed cases diagnosed by ICD-10 codes were included in the analysis. Due to the overlap in patients of each exclusion criterion, the total numbers of excluded patients were not 11,123 and 519, respectively. ICD-10, International Classification of Diseases, 10th Revision; CD, Crohn’s disease; MDV, Medical Data Vision; TB, tuberculosis; INH, isoniazid.
Eligible Patients Characteristics
Biologic-Experienced Patients Characteristics
2. Incidence
Seventeen patients developed active TB. The characteristics of patients are shown in Supplementary Table 1. All patients developed TB of respiratory organs. Of the 17 patients, 7 were on IFX. Among these patients, 1 patient each reported the concomitant use of corticosteroids, thiopurine, and both corticosteroids and thiopurines. Five patients were on ADA, of whom 1 patient had a history of concomitant use of corticosteroids, and the other had a history of concomitant use of thiopurines. Of the 5 biologic-free patients, 2 had a history of both IFX and ADA prescription over 10 months before active TB onset, whereas 3 patients were biologic-naïve. Meanwhile, among the 3 biologic-naïve patients, 1 was on thiopurines. Notably, none of the patients treated with UST or VDZ developed active TB. The median duration from the initiation of TNF-α inhibitors to the onset of active TB was 142.5 days (IQR, 107.0–370.8 days), whereas the median biologic-free duration was 304.0 days (IQR, 122.0–874.0 days) for 5 biologic-free patients. Fig. 2 shows the Kaplan-Meier curves of active TB-free survival for patients on TNF-α inhibitors, UST, VDZ, and biologic-free.
Kaplan-Meier curves of active tuberculosis-free survival for patients on tumor necrosis factor-α (TNF-α) inhibitors, ustekinumab (UST), vedolizumab (VDZ), and biologic-free.
Table 3 summarizes the incidence of active TB. The overall incidence of active TB was 21.3 per 100,000 person-years (95% CI, 11.2–31.5). The incidence per 100,000 person-years was 12.9 (95% CI, 1.6–24.3) and 33.1 (95% CI, 14.4–51.9) for biologic-free patients and patients on TNF-α inhibitors, respectively.
Incidence of Active Tuberculosis
3. Cox Proportional Hazards Analysis
We analyzed the risk factors for active TB after excluding the duration of UST or VDZ use as none of the patients treated with UST or VDZ developed active TB. The univariate analysis indicated that there were no significant risk factors for active TB development; however, in the multivariate analysis, TNF-α inhibitors were associated with active TB development (hazard ratio, 3.66; 95% CI, 1.23–10.93; P=0.020) (Table 4). Concomitant corticosteroid, thiopurines, or comorbid diabetes were not associated with active TB development.
Univariate and Multivariate Analysis for Active Tuberculosis Development with the Cox Proportional Hazards Model Considering TNF-α Inhibitor Use as a Time-Dependent Covariate
DISCUSSION
In this retrospective cohort study using MDV database, TNF-α inhibitors, but not UST or VDZ, were identified as risk factors for active TB development in patients with CD. Although this result is compatible to previous studies [7-12], to our knowledge, this is the first large-scale study using medical claims database on Japanese patients. This novel study demonstrated the association between biologic use and active TB in Japanese patients with CD. With the increasing number of patients treated with biologics, it is imperative for clinicians to recognize the risk of TB associated with biologics even in Japan where the prevalence of TB have been decreasing.
In this study, the overall incidence of active TB was 21.3 (95% CI, 11.2–31.5), compared to 9.2 according to the Japanese annual report of 2021 [20]. Thus, considering that many of the patients in the MDV database used in the present study were patients with CD receiving immunomodulatory therapy, we are confident that the data obtained in this study are reliable.
TNF-α inhibitors play a central role in CD treatment; they bind to soluble and transmembrane TNF-α, preventing TNF-α from binding to its receptors. Simultaneously, these inhibitors bind to transmembrane TNF-α and induce antibody-dependent cellular cytotoxicity-mediated cytotoxicity to prevent the overproduction of TNF-α [21]. However, TNF-α is also an important cytokine associated with mycobacterial infection; TNF-α activates alveolar macrophages and induces granuloma formation [22,23]. Consequently, TNF-α inhibitors impede the immune response to TB and increase its risk.
The incidence of TB infections depends on the TB prevalence in the respective geographical region. Therefore, in regions with low TB transmission, active TB primarily develops owing to LTBI reactivation. To prevent LTBI reactivation, numerous guidelines recommend LTBI screening by combining patient clinical data, chest radiography, and interferon-gamma release assay (IGRA) and/or tuberculin skin test (TST) [24-26]. However, LTBI treatment regimens vary among countries. For instance, in several countries, including Japan, patients diagnosed with LTBI receive INH for 6 to 9 months [24,25,27,28]. Furthermore, Carmona et al. [29] reported that active TB infections in patients on TNF-α inhibitors decreased after LTBI screening before introducing biologics. However, the implementation of LTBI screening and treatment with INH remains incomplete in many healthcare settings [29]. IGRA and TST can provide false-negative or false-positive results, particularly in patients receiving the Bacillus Calmette-Guérin vaccination or on immunosuppressive agents. Additionally, numerous idiopathic inflammatory diseases, such as IBD, require immunosuppressive and immunoregulatory therapy, including corticosteroids, which may also result in false-negative results. Therefore, it is difficult to exclude LTBI by screening with IGRA or TST alone.
UST is a monoclonal antibody against the p40 subunits of IL-12 and IL-23. Although some studies have reported active TB associated with UST in patients with psoriasis [30,31], there are few reports in patients with CD [32]. Both human and mouse models have demonstrated the role of IL-12 in generating a protective interferon-gamma-mediated immune response against TB infections [33,34]. Furthermore, IL-23 is central to anti-TB immunity, despite marginal effects [35]. Nonetheless, no patients treated with UST developed active TB, aligning with findings from a report in China, which identified no instance of patients with active TB development among 721 patients with CD treated with UST [36]. Possible explanations include the relative unimportance of IL-12/23 in anti-TB immunity, the antibody dosage and affinity, and the absence of cytotoxicity via antibody-dependent cellular cytotoxicity compared to TNF-α inhibitors. Nonetheless, further prospective studies with more cases are warranted.
VDZ is a gut-selective monoclonal antibody to integrin α4β7. As its mechanism of action, VDZ inhibits lymphocyte binding to mucosal addressin cell adhesion molecule-1 and migrates to the intestinal mucosa. Notably, VDZ poses a low risk for active TB in patients with IBD [10,11,37,38]. Based on our findings, no patients with CD who were receiving VDZ developed active TB, a result that is consistent with previous reports. Furthermore, a study from South Korea demonstrated that no patients developed active TB among 181 patients with CD treated with VDZ [39].
Compared with TNF-α monotherapy, the concomitant use of corticosteroids, methotrexate, or azathioprine with TNF-α inhibitors increases the risk of serious infection [18,19]. However, in our study, the concomitant use of corticosteroid or thiopurine with TNF-α inhibitors was not associated with active TB. We hypothesized that this lack of association could be attributed to considering the concomitant administration as a moderator variable. However, this approach may have overlooked precise quantification of the duration and dosage of corticosteroids and thiopurines, potentially affecting the accuracy of our findings.
This study has some limitations. Firstly, active TB was diagnosed using ICD-10 codes and prescription data, which may have introduced the minor possibility of an inaccurate diagnosis. Secondly, we could not consider LTBI screening because IGRA and TST results were not included in the MDV database. However, we excluded patients who were receiving INH to prevent LTBI reactivation. Therefore, we investigated the risk of TB caused by biologics without prophylaxis. Thirdly, owing to the characteristics of MDV database, we could not obtain follow-up data upon patient transfer another hospital. Moreover, in some cases, active TB treatment was initiated rapidly after data initiation. Thus, although these patients may have been transferred owing to active TB development, they were not included in the analysis. This is because we excluded patients without prescription data ≥ 12 weeks before the onset of active TB. Finally, potential confounding variables such as disease severity of CD and lifestyle habits like smoking or alcohol drinking were not considered because of the characteristics of database. Additionally, we were not able to investigate dose-dependent effect of TNF-α inhibitors because the database did not include body weight data for some patients. Despite these limitations, our study’s strength lies in the enrolment of a relatively large number of patients, including those treated not only with TNF-α inhibitors but also with UST and VDZ, to evaluate the risk of active TB infection.
In conclusion, TNF-α inhibitors, but not UST or VDZ, are risk factors for active TB in Japanese patients with CD. This result suggests the importance of LTBI screening before the initiation of TNF-α inhibitors and careful follow-up during the administration. Treatment with UST and VDZ may be a better option for immunocompromised hosts, such as older adults or individuals requiring multiple immunosuppressive therapies.
Notes
Funding Source
The authors received no financial support for the research, authorship, and/or publication of this article.
Conflict of Interest
No potential conflict of interest relevant to this article was reported.
Data Availability Statement
Not applicable.
Author Contributions
Conceptualization: Fujimoto K, Hosomi S, Ohfuji S, Fujiwara Y. Data curation: Fujimoto K. Formal analysis: Fujimoto K, Hosomi S, Ohfuji S. Investigation: Fujimoto K. Methodology: Fujimoto K, Hosomi S, Ohfuji S. Project administration: Hosomi S. Supervision: Fujiwara Y. Writing-original draft: Fujimoto K. Writing-review & editing: Hosomi S, Kobayashi Y, Nakata R, Nishida Y, Ominami M, Nadatani Y, Fukunaga S, Otani K, Tanaka F, Ohfuji S, Fujiwara Y. Approval of final manuscript: all authors.
Supplementary Material
Supplementary materials are available at the Intestinal Research website (https://www.irjournal.org).
Supplementary Table 1.
Characteristics of Patients Who Developed Active Tuberculosis