This year at ICPE All Access, Aetion’s scientific research was represented across symposia, presentations, and posters, including some initial findings from our COVID-19 research collaboration with the U.S. Food and Drug Administration (FDA) and our continued efforts to advocate for best practices in pharmacoepidemiologic methods. In this two-part series, we break down the key takeaways from each presentation. This is part two; see our full list of posters and presentations here.
Pharmacoepidemiologists are continually working to improve upon existing research methods to ensure their practices are as efficient and accurate as possible, and that they align with the latest scientific and regulatory standards. This work is essential to advance the use of principled methods for real-world evidence (RWE) across stakeholders, especially as regulators continue to set standards around key questions like how to define fit-for-purpose real-world data (RWD), and how to construct scientifically valid RWE studies.
The research highlighted below—prepared for ICPE All Access 2021 and primarily powered by Aetion Evidence Platform®—shares efforts to promote methods best practices and learnings from methods-focused analyses, featuring co-authors from Aetion, biopharma, and academia.
COVID-19 confidential: Enabling trust in COVID-19 research with fit-for-purpose RWD
Nicolle Gatto, PhD, MPH, FISPE; Jessica Franklin, PhD; Elizabeth M. Garry, PhD, MPH; David Martin, MD, MPH; Aloka Chakravarty, PhD, MStat
As researchers worked to generate RWE to answer critical and timely questions about COVID-19 and its treatments, they encountered unique challenges when selecting fit-for-purpose RWD to conduct the analyses.
In this symposium, representatives from FDA, Moderna, Optum, and Aetion shared their respective experiences determining data fitness for purpose amidst the continually evolving landscape of the pandemic. Panelists advocated for a structured process for defining fit-for-purpose RWD in order to yield valid and transparent RWE to address COVID-19-related research questions.
RWE blueprints: Decoding SPACE, SPIFD, STaRT-RWE, and other tools promoting principled pharmacoepidemiology
Elizabeth M. Garry, PhD, MPH; Nicolle Gatto, PhD, MPH, FISPE; Shirley Wang, PhD, MSc; Ulka Campbell, PhD, MPH
The number of guidance documents and frameworks that aim to set standards around the use of RWE has grown rapidly in recent years, with varying levels of specificity and agreement with each other. As such, a number of structured processes have emerged to help researchers ensure their work follows the tenets of principled database epidemiology.
In this symposium, panelists discuss a number of these structured processes, templates, and frameworks, and how researchers can refer to them to support the planning, design, conduct, and reporting of RWE studies. Frameworks discussed include the Structured Preapproval and Postapproval Comparative study design framework to generate valid and transparent real-world Evidence (SPACE), the Structured Process to Identify a Fit-for-purpose Data (SPIFD) framework, and the Structured Template and Reporting Tool for Real-World Evidence (STaRT-RWE).
Propensity score modeling to reduce channeling bias when exposure is rare
Julia C. Pitino, BS; Jennifer Polinski, ScD, MPH, MSc; Brenda Hinman-McIlroy, DO, MPH, FACOG; Michael C. Snabes, MD, PhD; Dionna Attinson, MPH; Stephanie E. Chiuve, ScD
Channeling bias occurs when treatments indicated for similar populations are prescribed to particular groups over others depending on certain baseline characteristics. It’s especially common in rare treatments indicated for a small subset of patients with a common disease, or in first-in-class treatments. In these cases, channeling bias in post-marketing surveillance of treatments can introduce challenges in contextualizing rates of adverse events, and researchers need further methodology to quantify the underlying, baseline risks of adverse events in the treatment-indicated population.
In this study, authors from Aetion and AbbVie sought to generate an appropriate comparator group consisting of patients who were demographically and clinically similar to the rare treatment initiators, but did not receive treatment. They focused on elalogix usage among patients with endometriosis (EM), and defined a population that was clinically similar to the treatment-indicated population—known as a counterfactual comparator group—using propensity score modeling to match patients unexposed to elalogix to exposed patients.
The authors demonstrated that the counterfactual group can inform the benefit-risk profile of the treated population prior to initiating a comparative safety analysis, and can account for differences in comorbidities and disease severity in patients with the same disease who initiate treatment, compared to those who do not. This method represents a useful tool for researchers looking to contextualize rates of adverse events among a treated population, either prior to or alongside a true comparative safety analysis for a drug in the post-market setting.
A counterfactual approach to minimize channeling bias in post-market safety surveillance
Julia C. Pitino, BS; Jennifer Polinski, ScD, MPH, MSc; Brenda Hinman-McIlroy, DO, MPH, FACOG; Michael C. Snabes, MD, PhD; Dionna Attinson, MPH; Stephanie E. Chiuve, ScD
Elagolix is an approved treatment for moderate to severe EM pain, but it may increase risk of mood disorders. Prior evidence has shown that EM pain is associated with a greater risk of mood disorders overall, and elagolix initiators may have more severe pain than the general population of patients with EM.
In this study, authors from Aetion and AbbVie sought to clinically describe the elagolix-indicated population and determine the factors that predict whether a patient will be treated with elagolix. They then generated a counterfactual comparator group of patients who did not take elagolix, but were clinically similar to those who did. The authors used this group to quantify the baseline rate of mood disorders relative to the general EM population, thereby examining channelling bias and contextualizing the increased risk of mood disorders among elagolix users. Ultimately, the authors found that women with EM and similar clinical profiles to patients treated with elagolix may have a higher baseline risk of mood disorders relative to the general EM population.
These methods may be useful to researchers studying a rarely used treatment indicated for a common disease, or when studying a poorly described population. Here, the study design allowed the authors to dive deep into the data to identify clinical characteristics that differed between the exposed and unexposed populations, which is critical for post-market drug safety analyses.
Impact of increasing enrollment requirements in administrative claims data
Liza R. Gibbs, BS; Irisdaly Estevez, MPH; Katherine K. Gilpin, BA; Elizabeth M. Garry, PhD, MPH
Pharmacoepidemiologic studies using claims data often specify a minimum amount of time that patients must be enrolled in a health plan prior to their cohort entry date. Differing study goals may affect researchers’ rationale behind longer or shorter enrollment requirements, but it is not always clear the impact that varying enrollment requirements might have on who is included in the study cohort.
In this analysis, authors used two claims-based cohorts to evaluate the impact of increasing baseline enrollment requirements on the distribution of baseline characteristics. The findings indicate that manipulating the length of required minimum enrollment not only impacts the sample size, but can change the demographic makeup of a cohort.
Use of high-dimensional Propensity Scores (hdPS) in a Japanese National Claims Database
Jocelyn R. Wang, MS; Sahar Syed; Elizabeth M. Garry, PhD, MPH
In comparative safety and effectiveness studies using RWD, researchers have used the high-dimensional propensity score (hdPS) algorithm extensively to control for confounding in U.S. claims data. However, the performance of hdPS in Japanese claims data, which differs in clinical practice and health care data management from the U.S., has not been evaluated.
This study assessed the confounding control imposed by hdPS in a national Japanese claims database. The authors tested this using the known protective effect of COX-2 inhibitors on severe gastrointestinal complications as an example, evaluating the treatment effect estimates of COX-2 inhibitors versus non-selective traditional NSAIDs using different confounding control strategies. Their findings indicated that use of the hdPS algorithm in addition to covariates determined a priori achieved findings that were most similar to expectations, thereby demonstrating the utility of hdPS in Japanese claims data, despite the differences in clinical practice and health care data management.