Nonrandomized Comparison of Adverse Events Following Facility- and Home-Infused Biologics Using Real-World Data

Observational studies of adverse events for medical therapies have an important role in pharmacovigilance. The discipline referred to as pharmacoepidemiology describes the real-world use of medications in clinical practice and both their expected and unexpected outcomes. This discipline provides important information about new medication years after the foundational clinical trials and regulatory approval of drugs.

In the comparative safety study by Baker et al,¹ the authors compare the rate of emergency department (ED) visits and hospitalizations between patients receiving biologic therapies at home or in a health care facility for a variety of immune-mediated diseases, including rheumatological, gastroenterological, and neurological conditions. Administrative health claims data on 57 220 patients who received 752 150 biologic infusions from 2007 to 2017 were leveraged to better understand the safety of these medications. In multivariable analyses, the authors found a statistically significant 25% increase in odds of an ED visit or hospitalization following home infusion.

We describe the pharmacoepidemiological context of this study and important questions it raises about using real-world data to assess drug safety with respect to: (1) depletion of susceptibles; (2) causal attribution of adverse events, and (3) understanding drug discontinuation. Patients who may have first been safely administered a biologic medication in a facility and then go on to have a home-based infusion could have a lower likelihood of infusion-related acute events. In pharmacoepidemiological studies, particularly when investigating medication adverse events, this phenomenon is referred to as a “depletion of susceptibles.”² This occurs in follow-up studies where exposure to a drug in a cohort of patients is associated with an early increase in the incidence rate of an acute event, followed by a decrease in incidence with longer duration since first exposure and possibly a lower risk of acute events associated with subsequent administrations of a medication. The remaining patient population would be depleted of susceptible patients who experienced the event early and discontinued treatment.

Renoux et al³ illustrated time-dependent bias from depletion of susceptibles among hormone therapy users and the risk of venous thromboembolism in the Million Women’s Study cohort. New medication users at risk of this adverse outcome had a rapid increase and subsequent stabilization of the cumulative incidence function. However, it is important to note that if depletion of susceptible patients was prominent in the study by Baker and colleagues, and patients at lower or stable risk were enriched among the home infusion group, the significantly increased risk observed could be an underestimation of the true risk associated with home-based biologic infusions.

One drawback to the approach used by Baker et al, common to many administrative claims–based studies of patients with chronic diseases, is the lack of information on full disease history, including the length of time since diagnosis, prior treatments, and related adverse events. This is a consequence of only being able to observe patients’ data while they are continuously enrolled on the health plans in the database. The natural history of immune conditions is long and often progressive or degenerative, meaning that a longer duration and possibly higher severity of illness confers a higher risk of adverse outcomes like ED visits or hospitalization, a primary outcome of this study. Another limitation was the authors’ inability to differentiate whether the biologic medication infused was a first-ever treatment or a continuation of previous treatment. It was not known if these characteristics differed between those administered infusions at home and those at health care facilities. Thus, comparisons between patients who are starting a new treatment vs those receiving these chronic treatments over years are infeasible. The investigators used other means of accounting for possible differences in illness severity by adjusting for recent use of glucocorticoids at a prednisone-standardized dosage, which could be related to both underlying disease severity and the incidence of infusion reactions.

This investigation of all-cause ED visits and hospitalizations also raises the interesting question of optimal surrogates for identifying drug-related adverse events in the absence of more specific outcomes data. Patients receiving these biologic agents may present to the hospital for a variety of reasons including drug-related adverse events, disease sequelae, or other unrelated health conditions.

True causal attribution of adverse events to a drug of interest is difficult to ascertain when using real-world data. Limiting the time period of observed ED visits and hospitalizations to the same and next day after infusion suggests a temporal relationship attributing that event to drug administration. However, the reliability of this attribution is uncertain, and details of diagnoses associated with these events may provide additional information. Alternatively, specific diagnosis codes can be used to determine more targeted adverse events. A similar study⁴ of patients with inflammatory bowel disease specifically investigated disease-related ED visits and hospitalizations following home- vs hospital-based infusions of biologic agents. Foley et al⁵ determined the rates of infusion reactions associated with cetuximab using frequency of outpatient treatments, ED visits, and hospitalizations for hypersensitivity and allergic reactions. In their study, they used International Classification of Diseases codes consistent with signs and symptoms of infusion reactions. Access to more detailed data from the electronic health records (EHR) or patient registries can provide further insight on the timing, severity, and management of drug-related adverse events, increasing confidence in the association of these events with the drug administration.

Claims data are limited by primary purpose, which is not for research but for reimbursement for health care services. The investigators designed their study in such a way as to minimize the limitations of using claims data, but this data source is still limited. EHR and registry data are 2 alternative real-world data sources that may offer complementary information about treatment discontinuation that is otherwise unavailable.

The secondary outcome in the study by Baker et al, drug discontinuation, is particularly suited for investigation with real-world data because treatment with biologics may last many years and exceed randomized controlled trial follow-up.⁶ Specifically, EHR and registry data may shed light on reasons for drug discontinuation, whether for toxic effects, loss of efficacy, patient preference, or otherwise, given access to more detailed clinical or patient-reported data. Additionally, EHR or registry data may be generally better suited to identifying when a drug is discontinued because clinical documentation often states that the treatment was halted and why.

Only a randomized clinical trial could establish a causal association between home infusion of biologic agents and subsequent adverse events and discontinuation of the agent, but conducting such a trial at this scale would be costly and may not reflect the uncontrolled conditions that occur in clinical practice. This study provides insight into outcomes related to where these biologic infusions occur, in the home or clinical setting, using the data from patients in real-world settings. With the expansion of home health care coverage and greater utilization of home-based infusion services for patients amid the COVID-19 pandemic,⁷ the evidence provided here is highly relevant to understanding the risks and benefits of these disease-altering biologic therapies in different treatment settings.

Published: June 3, 2021. doi:10.1001/jamanetworkopen.2021.11156

Open Access: This is an open access article distributed under the terms of the CC-BY License. © 2021 Calip GS et al. JAMA Network Open.

Corresponding Author: Gregory S. Calip, PharmD, MPH, PhD, Flatiron Health Inc, 233 Spring St, New York, NY 10013 (gregory.calip@flatiron.com).

Conflict of Interest Disclosures: Drs Calip, Yerram, and Ascha reported current employment with Flatiron Health Inc, which is an independent subsidiary of the Roche group, and stock ownership in Roche. No other disclosures were reported.