Nintedanib and ischemic colitis: Signal assessment with the integrated use of two types of real‐world evidence, spontaneous reports of suspected adverse drug reactions, and observational data from large health‐care databases

Abstract Purpose Statistical screening of Vigibase, the global database of individual case safety reports, highlighted an association between the MedDRA Preferred Term (PT) “colitis” and nintedanib. Nintedanib is a protein kinase inhibitor authorized in accelerated regulatory procedures for the treatment of idiopathic pulmonary fibrosis (IPF). The aim of this report is to describe the integration of two types of real‐world evidence, spontaneous reports of adverse drug reactions (ADR), and observational health data (OHD) in the assessment of a post‐authorization safety signal of ischemic colitis. Methods Assessment of the statistical signal of “nintedanib – colitis” was undertaken using data from VigiBase, OHD from the Observational Heath Data Sciences and Informatics (OHDSI) collaborative, published literature, and openly available regulatory documents. Evidence synthesis was performed to support Bradford Hill criteria in causality assessment. Results Evidence for strength of association, specificity, consistency, and analogy was found upon review of the case series. OHD was used to calculate incidence rates of colitis in new users of nintedanib across multiple populations, supportive of consistency, and further evidence for strength of association. Literature review identified support for biological plausibility and analogy. Signal assessment was supplemented with characterization of real‐world users and exploration of potential risk factors using OHD. Conclusions An integrated approach using two forms of real‐world data, spontaneous reports of ADRs and data from observational databases allowed a comprehensive and efficient signal assessment of nintedanib and colitis. Further exploration of the complementary use of real‐time OHD in signal assessment could inform more efficient approaches to current signal management practices.


| INTRODUCTION
Nintedanib is a small molecule receptor tyrosine kinase inhibitor (TKI) blocking vascular endothelial growth factor receptors (VEGFR 1-3), fibroblast growth factor receptors (FGFR 1-3), and platelet-derived growth factor receptors (PDGFR) α and β kinase activity. It received marketing approval from the European Medicines Agency (EMA) and the United States Food and Drug Administration (USFDA) in 2014 for the treatment of IPF, a rare disease characterized by an excess of fibroblasts, leading to progressive fibrosis of the lung. Marketing approval was granted by the EMA after an accelerated assessment and the USFDA after a priority review; less than 1000 patients received nintedanib in pivotal trials that provided the basis for licensure. 1,2 Ischemic colitis is a condition caused by insufficient blood flow to the large intestine, which results in mucosal ulceration, inflammation, and hemorrhage. Causes are physiological (hypotension, secondary to embolus/thrombosis) or iatrogenic (secondary to medicines, surgery). Ischemic colitis manifests with diarrhea, colicky abdominal pain, and rectal bleeding. Patients with ischemic colitis should be recognized quickly, as colonoscopic evaluation is recommended within 48 hours of symptom onset. Colonoscopy can distinguish those cases that may be treated with conservative management from those that require emergency resection. The complications arising from ischemic colitis include obstruction, necrosis, and perforation. 3 The aim of this report is to describe the assessment of a signal of colitis and nintedanib with the complementary use of spontaneous reports of ADRs and OHD.

| METHODS
Statistical screening in VigiBase using the VigiRank algorithm 4 was performed in December 2019 and highlighted the drug-adverse event (AE) combination of "nintedanibcolitis" for review. The drug-AE pair was assessed for the potential for a causal relationship using the Bradford Hill criteria. 5 Additional analyses within VigiBase and several observational databases (insurance claims databases from the US) available via the Observational Heath Data Sciences and Informatics (OHDSI) collaborative, 6 as well as reviews of the licensure application and published medical literature, were performed to explore the potential for a causal relationship between nintedanib and colitis.

| RESULTS
Clinical review of the 25 individual case safety reports (ICSRs) for the drug-AE pair revealed different types of colitis, such as ischemic or inflammatory, or was further unspecified. Given that nintedanib acts upon the VEGF receptor, the association between nintedanib and the more specific clinical concept of "ischemic colitis" was explored.
One case from the original "colitis" case series had information sufficient to be considered as "ischemic colitis" (colonoscopy results were provided). Subsequent exploration within VigiBase used the MedDRA SMQ, ischemic colitis (narrow); nine additional cases were identified. The final case series of ischemic colitis included a total of 10 cases coded to different MedDRA PT: five cases of "colitis ischemic," two of "intestinal ischemia," one reporting both "colitis ischemic" and "intestinal ischemia," one of "intestinal infarction," and one of "colitis." Within these 10 reports, seven reports described males and three described females (Table 1). Ages ranged between 53 and 78; age was not included in one report. Reports originated from five countries (one country in the Americas, three countries in Europe, and one country in Asia). Time to onset was from 3 days to 5 months. In seven cases, the indication for nintedanib was IPF; in the three other cases, non-small cell lung cancer, bronchial carcinoma, and glioblastoma.
Diagnostic results were reported in three cases (6, 8, and 10). Fatal outcomes were reported in four cases. Limited additional information was available in most cases given that narratives had not been shared into Vigibase. Potential confounding/risk factors for ischemic colitis were advanced age, concomitant use of corticosteroids in two cases (1 and 5) and medical history of hyperlipidemia in two cases (2 and 5).
Observational data were used to explore "real-world use" of nintedanib after marketing approval. Use of nintedanib started in 2014 with most users in the 70-to 80-year age range and in more males than females. New users of nintedanib had no prior history of colitis and did not significantly differ from users of the alternative therapy for IPF, pirfenidone.  VEGF inhibition could contribute to perforation by inducing regression of normal blood vessels in the gastrointestinal tract 9 (presumptively via an intermediate step of ischemia). Specificity was explored in the spontaneous database using the MedDRA SMQ "ischemic colitis"; evidence of disproportionality was found for multiple PTs within this SMQ, including "colitis ischemic" (IC025 0.32), "large intestine perforation" (IC025 1.67), and "large intestinal hemorrhage" (IC025 0.39).
Evidence for strength of association is found in the disproportionality in Vigibase and measures of incidence from observational data (eg, 56.74 cases per 1000 person-years from one insurance system). Consistency is found in both ICSRs (reports from numerous countries) and OHD (increased incidence observed in multiple insurance systems). A consideration of analogy is feasible given the availability of other antiangiogenic agents which antagonize VEGF (aflibercept and bevacizumab) or inhibit the VEGF receptor (regorafenib and sorafenib Additionally, OHD could allow for a preliminary estimation of the incidence risk ratio from a self-controlled "pre-exposure" comparative cohort. An important limitation in this pilot study was the differences in the amounts of data and medical terminology between the different data sources. For example, further specification of "colitis" was not possible in the OHD due to the relatively small number of patients taking nintedanib. Furthermore, clinical reasoning is required to translate signals identified in MedDRA terminology into explorations of databases using ICD coding.

| DISCUSSION
Several publications have explored different approaches to the use of observational data in signal detection. A large European initiative, the EU-ADR project, has explored the use of real-world healthcare data in signal detection; the findings revealed that potential for signal detection within real-world data was found to be dependent on frequency of the event and utilization of drugs in the general population. 12 Other proposals have included a signal detection strategy that requires signaling from both an adverse event reporting system and electronic health records 13  ORCID Rebecca E. Chandler https://orcid.org/0000-0002-9334-9046