Drug Repurposing Prediction and Validation From Clinical Big Data for the Effective Treatment of Interstitial Lung Disease

Abstract

Interstitial lung diseases (ILDs) are a group of respiratory disorders characterized by chronic inflammation and fibrosis of the pulmonary interstitial tissues. Although the etiology of ILD remains unclear, some drug treatments are among the primary causes of ILD. In the present study, we analyzed the FDA Adverse Event Reporting System and JMDC Inc. insurance claims to identify a coexisting drug that reduced the incidence of ILD associated with the use of an anti-arrhythmic agent, amiodarone, and found that the thrombin inhibitor dabigatran prevented the amiodarone-induced ILD in both clinical datasets. In an experimental validation of the hypothesis, long-term oral treatment of mice with amiodarone caused a gradual decrease in body weight caused by respiratory insufficiency. In the lungs of amiodarone-treated mice, infiltration of macrophages was observed in parallel with a delayed upregulation of the platelet-derived growth factor receptor α gene. In contrast, co-treatment with dabigatran significantly attenuated these amiodarone-induced changes indicative of ILD. These results suggest that dabigatran is effective in preventing drug-induced ILD. This combinatorial approach of drug repurposing based on clinical big data will pave the way for finding a new treatment with high clinical predictability and a well-defined molecular mechanism.

Keywords: adverse event; amiodarone; chronic inflammation; dabigatran; pulmonary fibrosis; real-world data; retrospective analysis.

FIGURE 1

Increased incidence of interstitial lung disease (ILD) with the prescription of drugs and confounding effects of concomitant drugs on amiodarone-induced ILD in the FDA Adverse Event Reporting System (FAERS) data. Volcano plots for visualizing the reporting odds ratio (ROR, on a log scale) and its statistical significance (absolute Z score) are shown. Each circle indicates an individual drug, and the size of the circle reflects the number of patients taking the drug. (A), Strong and significant increases in the ROR for ILD were seen in patients using amiodarone and bleomycin. Overall values are presented in Supplementary Table S1. (B), Within the population taking amiodarone, confounding effects of concomitantly used drugs on the incidence of amiodarone-induced ILD were calculated thoroughly and are plotted. Overall values are presented in Supplementary Table S2. (C), Effects of amiodarone, dabigatran, and their combination on the report proportion of ILD in FARES data.

FIGURE 2

Time distribution of the event in JMDC claims data. (A,B), The intervals from the insurance enrolment of a patient to the first prescription of amiodarone (A) and to the initial diagnosis of interstitial lung disease (ILD) (B) were analyzed, and the number of patients is shown on a monthly basis. From these data, patients who had been enrolled in health insurance for 0–2 months were omitted in the following analysis to allow a run-in period. (C), Sequence symmetry analysis showing the causal relationship between the start of amiodarone treatment and the onset of ILD in an observation period of ±36 months. At month 0, the precise chronological order was unknown, and the data were omitted from the analysis. (D), Kaplan-Meier curves for the cumulative incidence ratio of ILD in patients taking amiodarone are shown individually in two populations, one without (red) and one with (blue) co-prescribed dabigatran. Dotted lines show the number of patients at risk as a ratio to the initial number (n ₀) of patients.

FIGURE 3

Effects of dabigatran on amiodarone-induced lung toxicity. Kaplan-Meier survival data (A) and the body weight changes (B) of the male C57BL/6J mice orally treated with amiodarone (300 mg kg⁻¹ day⁻¹) and dabigatran (60 mg kg⁻¹ day⁻¹) or both (A + D) 5 times per week on weekdays for 4 weeks (n = 18–28 per group). Body weight data are shown as the mean ± SEM of those survived 26 days (n = 15–18 per group). Statistical significance was tested by two-way ANOVA with multiple comparisons. (C), Representative images of Iba1-immunostaining (left) and bright field (right) of mouse lung section on day 26 after chronic oral treatment with amiodarone and dabigatran or both. White arrowheads indicate alveolar macrophage infiltration. Scale bar = 20 μm. (D), Numbers of Iba1-positive macrophages in the immunostained images (n = 6–7 per group). Statistical significance was tested by two-way ANOVA with Tukey’s multiple comparisons.

FIGURE 4

Heatmap of gene expression in the lung tissue (on day 26) of mice treated with vehicle (1 VEH), amiodarone (2 AMI), dabigatran (3 DAB), or amiodarone + dabigatran (4A + D). Log2-expression values calculated from RNA-Seq results are standardized across treatment groups. Red and blue colors represent larger and smaller, respectively, standardized values in Z scores. In addition, the differences between AMI and VEH groups (2–1) and A + D and AMI groups (4–2) are shown. Genes are ordered by the difference between the AMI and VEH groups (2–1) from high to low separately in inflammation- (A) and fibrosis- or both- (B) associated gene sets.

FIGURE 5

Quantitative RT-PCR results showing the expression of inflammation-associated genes in the lung of mice after a repetitive, oral treatment with amiodarone (Amio, 300 mg kg⁻¹ day⁻¹) and dabigatran (Dab, 60 mg kg⁻¹ day⁻¹) or both (A + D) 5 times per week for 4 weeks. Expression levels of Pdgfra (A), Pdgfc (B), Mmp12 (C), and Timp1 (D) were normalized to the expression of Pum1 in each sample from individual mouse (n = 8–11), and are represented as the relative ratio to those of vehicle group (Veh). Individual data are shown with the mean ± SEM. Statistical significance was tested by one-way ANOVA with Tukey’s multiple comparisons. ***, p < 0.001; ns, not significant.