Skip to Main Content


Yale University-Mayo Clinic CERSI

Center of Excellence in Regulatory Science and Innovation

The Yale-Mayo CERSI conducts high-quality, high-impact collaborative research to support several areas of focus in the FDA strategic plan for regulatory science. Research topic areas include: advancing clinical and post-market surveillance of drugs and biologics, advancing clinical and post-market surveillance of medical devices and diagnostic tests, development and application of novel analytics, and fostering patient-centered decision making.

For comments or questions pertaining to our CERSI research projects, contact the Yale-Mayo Clinic CERSI at

Current Projects

Advancing Clinical and Post-Market Surveillance of Drugs and Biologics

Characterizing use, safety and efficacy of brand-name and generic drugs used to treat hypothyroidism

FDA Priority Area/Regulatory Science Challenge: Generic drugs are approved based on bioequivalence to the brand-name agents. However, there are sometimes concerns among patients and clinicians that generic and brand-name drugs are not equivalent and may have differing effects. One example of this is the clinical preference to prescribe brand-name L-thyroxine as opposed to the generic formulation of the drug. This disconnect between FDA and expert recommendations is likely to cause confusion among patients and may be resolved through comparative effectiveness and safety research among currently available L-thyroxine products.

This project uses Real World Evidence via structured electronic health record and administrative claims data to better understand brand and generic use of FDA-regulated drugs to treat hypothyroidism and examine comparative effectiveness and safety. The proposed project will advance our knowledge of the effectiveness and safety of generic drugs.

Project Description: Using a large administrative claims data source that includes information on privately insured and Medicare Advantage enrollees of all ages, we will characterize patterns of use of generic and brand-name L-thyroxine products and then compare the effectiveness and safety of generic and brand-name L-thyroxine among new users.


Use of instrumental variable approaches to assess the safety and efficacy of brand-name and generic drugs used to treat hypothyroidism

FDA Priority Area/Regulatory Science Challenge: Generic drugs are approved based on bioequivalence to the brand-name agents. However, there are sometimes concerns among patients and clinicians that generic and brand-name drugs are not equivalent and may have differing effects. One example of this is the clinical preference to prescribe brand-name L-thyroxine as opposed to the generic formulation of the drug. This disconnect between FDA and expert recommendations is likely to cause confusion among patients and may be resolved through comparative effectiveness and safety research among currently available L-thyroxine products.

This project uses Real World Evidence via structured electronic health record and administrative claims data to better understand brand and generic use of FDA-regulated drugs to treat hypothyroidism and create ways to aid in the implementation of instrumental variable methods for future research. The proposed project will advance our knowledge of methods that can be used to compare the safety and effectiveness of generic drugs. It will also provide a toolkit for the use of instrumental variable approaches for future research.

Project Description: Building on our ongoing research, we will test different methods to compare the effectiveness and safety of generic and brand-name drugs using a large administrative claims data source that includes information on privately insured and Medicare Advantage enrollees of all ages. Further, we will create packages that will ease the implementation of instrumental variable methods for future research. Collaborators from the University of Washington are involved in this project.

Linking data sources to elucidate non-fatal and fatal opioid-related overdose epidemiology and the role of FDA-regulated products

FDA Priority Area/Regulatory Science Challenge: Overdose, or the syndrome where a drug causes loss of consciousness/coma, is the leading cause of accidental death in the U.S. It is clear that medications- not just illegal drugs- are increasingly involved in overdoses. As such, the FDA has pledged to re-examine the role of medications, particularly opioids and benzodiazepines, in the dramatic increase in overdoses.

One of the main challenges in studying opioid use is that, while data may show who has experienced an overdose, it has been very difficult to assess what medications they may have had access to or what they were taking. The primary reason for this difficulty is that there are many different sources of data on medication receipt and it is challenging to link these datasets. In this project, those challenges will be overcome by linking a variety of datasets that will generate a more accurate assessment of which medications at which doses put patients at risk for- or protect them from- overdose. This foundational work will ultimately allow for identification of specific FDA-regulated products (e.g. long and short acting opioid analgesics, abuse-deterrent formulations, benzodiazepines) and individuals who use them at highest risk for overdose and the development of multi-system prevention efforts.

Project Description: The goal of this project is to characterize the role of FDA-regulated controlled substances in overdose epidemiology. This will first be addressed by gaining access to data from several independently maintained state databases. Next the project will link data from these databases to characterize morbidity and mortality risks associated with opioid products. Collaborators from the University of Connecticut are involved in this project.


Characterization and analysis of high incidence of potentially unsafe prescribing of some Extended-Release (ER) opioid analgesics using Natural Language Processing (NLP) of Electronic Health Record (EHR) clinical notes

FDA Priority Area/Regulatory Science Challenge: In 2012, the FDA approved a Risk Evaluation and Mitigation Strategy (REMS) to provide prescriber education to help reduce adverse outcomes resulting from misuse and abuse of extended-release (ER) opioid analgesics. Two recent studies conducted by the FDA, one using the Medicare database and the other using the Sentinel database, focused on inappropriate prescribing of ER opioid analgesics that require prior opioid tolerance to patients who do not appear (based on prescription dispensing records) to be opioid-tolerant. Those studies suggest a high incidence of potentially unsafe prescribing behavior for ER opioid analgesics to opioid non-tolerant patients, but recommend further validation in other datasets. Because prior analyses were performed using administrative claims data only, there is interest in determining whether information from EHRs could provide additional evidence of opioid tolerance in patients whose claims data lacks that evidence.

This project uses Real World Evidence to determine whether additional data from EHRs would enable enhanced analysis of the causes, potential clinical drivers, and verification and adjudication of the reasons for apparent widespread unsafe practice of prescribing ER opioid analgesics that require prior opioid tolerance to patients who are opioid-non-tolerant.

Project Description: The objective of this new study is twofold: (1) examine ER opioid analgesic prescribing patterns in the OptumLabs claims database of commercially insured and Medicare Advantage patients using the Willy and LaRochelle approach, and (2) provide further insight to understand the context of prescribing behavior using clinical data from Electronic Health Records (cdEHR) including data extracted from raw provider notes using advanced NLP techniques. Collaborators from OptumLabs are involved in this project.


Understanding the contribution of laboratory data linked to administrative claims: a case study looking at renal function and oral anticoagulant performance in patients with atrial fibrillation

FDA Priority Area/Regulatory Science Challenge: The FDA has long relied on electronic healthcare data (e.g., data from claims and EHRs) to examine the safety of drugs in the post-market setting as part of its pharmacoepidemiology contracts and Sentinel Initiative. The passing of the 21st Century Cures Act and the proposed commitments included in the reauthorization of the Prescription Drug User Fee Act require FDA to give further attention and issue guidance on this topic. Recently, linking multiple types of electronic healthcare data is getting more attention, due to its ability to improve the measurement of exposures, outcomes, and confounders.

This project seeks to understand how linking laboratory data to insurance claims can help examine a drug's performance after approval. Lifelong oral anticoagulation is recommended in most patients with atrial fibrillation (AF) to prevent stroke. However, treatment decisions can be complicated by the presence of chronic kidney disease, as poor renal function increases the risks of both stroke and bleeding (a major complication of oral anticoagulation treatment), and may change the risk-benefit ratio of different treatment options. By using Real World Evidence to evaluate the performance of oral anticoagulants after FDA approval in patients with AF, the project will improve understanding of oral anticoagulant safety and effectiveness in relation to renal function as well as help providers and patients make informed decisions and achieve better outcomes.

Project Description: We will conduct a case study looking at renal function and the performance of oral anticoagulant drugs in patients with AF. This project proposes to answer two important questions pertaining to the impact of renal function on oral anticoagulation treatment in patients with AF. First, in patients with severe-to-no renal impairment, we will assess the comparative effectiveness and safety of different oral anticoagulant drugs across the range of renal function. Second, we will use novel analytic methods to identify patient characteristics that contribute to the heterogeneity in treatment effects.

We will answer these questions by leveraging the power of a large observational database, OptumLabs Data Warehouse, which contains over 160 million privately insured and Medicare Advantage enrollees of all ages, races, and from 50 states and the USRDS data.

Non-Federal Entity Collaborators: Brahmajee Nallamothu, MD, MPH, and Rajiv Saran, MD, MS, MBBS- Co-Investigators (University of Michigan)


Characterizing safety and efficacy of brand-name and generic drugs used to treat hypothyroidism among patients who switch therapy formulation

FDA Priority Area/Regulatory Science Challenge: Generic drugs are approved based on bioequivalence to the brand-name agents. However, there are sometimes concerns among patients and clinicians that generic and brand-name drugs are not equivalent and may have differing effects. One example of this is the clinical preference to prescribe brand-name L-thyroxine as opposed to the generic formulation of the drug. This disconnect between FDA and expert recommendations is likely to cause confusion among patients and may be resolved through comparative effectiveness and safety research among currently available L-thyroxine products.

This project uses Real World Evidence via structured electronic health record and administrative claims data to better understand the safety and effectiveness among patients who switch from brand to generic formulations to treat hypothyroidism. Additionally, this project will use novel measures to examine relevant outcomes among all patients.

Project Description: Using a large administrative claims data source that includes information on privately insured and Medicare Advantage enrollees of all ages, we will examine the effectiveness and safety of generic and brand-name L-thyroxine among adult patients. We will first examine the effectiveness and safety among patients who switch from brand to generic L-thyroxine within 1 year of treatment initiation and then among all patients who switched from brand to generic formulations regardless of treatment duration. Lastly, we will use a novel measure to estimate the overall time spent on generic and brand L-thyroxine for those patients on treatment for at least one year to examine relevant clinical outcomes.


Real-world data to assess variation in opioid prescribing and use for acute pain in diverse populations

FDA Priority Area/Regulatory Science Challenge: Many studies have described the differences reported by patients between the amount of opioid analgesic prescribed and the amount that they actually used to manage acute pain. However, these studies have generally assessed use after a limited number of surgical procedures, used small groups of patients at single institutions, and have not considered diverse populations that may have different demographics and social or cultural norms regarding opioid analgesic use. To better inform prescribing guidelines and public health measures, data are needed from diverse populations of patients on their use of opioid analgesics to manage acute pain, trajectories of pain experienced and response to opioids, and how patients dispose of these medications when no longer needed. Additionally, these data can help identify patient factors that predict variation in opioid analgesic use to incorporate into prescribing guidelines.

The congressional SUPPORT Act (Section 3002) has tasked FDA with developing evidence-based opioid analgesic prescribing guidelines for treating specific acute pain diagnoses where such guidelines do not exist. As part of this task, FDA will support development of real-world data on patient-reported opioid analgesic use to manage acute pain to help develop evidence-based recommendations for opioid analgesic-prescribing for specific conditions or procedures commonly associated with acute pain.

Project Description: Using a novel patient-centered health data-sharing platform (Hugo), the Yale-Mayo Clinic CERSI will enroll 1,550 patients who have been prescribed short-acting opioid analgesics after receiving care for new onset pain in the emergency department, primary care, or dental care offices at one of five diverse hospital systems. Patients will be followed for 180 days to collect information on pain control and opioid use through survey questionnaires sent via the Hugo platform. Additionally, patients will connect their electronic medical records and pharmacy data, when available, to the Hugo platform, using their patient portal accounts. Patients will also use wearable devices to gather additional insights into their activity and sleep patterns during the study period. At the end of follow-up, the CERSI will assess what patients did with their unused opioids. Non-federal collaborators from Cedars Sinai Medical Center, Monument Health, University of Alabama Birmingham (UAB), and UAB Dentistry are involved in this project.

Quantifying the relationship between inappropriate prescribing of opioid-tolerant-only medications to patients without prior opioid tolerance and opioid-related harms

FDA Priority Area/Regulatory Science Challenge: In 2012, the FDA approved a Risk Evaluation and Mitigation Strategy (REMS) to provide prescriber education to help reduce adverse outcomes resulting from misuse and abuse of extended-release opioid analgesics. A recently completed project using the OptumLabs claims data looked at prescribing of opioids that are only intended for use in people who are opioid tolerant. The project found that more than half of patients starting these drugs had no evidence of opioid tolerance. Such use is inconsistent with the labelling of these drugs and may create safety risks; however, the magnitude of this risk and prevalence of harms are not well described.

Project Description: This project will use claims data to measure the risk of opioid-related harms associated with the use of opioid-tolerant-only formulations by opioid naïve patients, and identify risk factors associated with these harms. Ultimately the findings from this study will help to meet the goals of the ongoing REMS for Opioid Analgesics efforts to reduce the risk of abuse, misuse, addiction, overdose, and deaths due to prescription opioid analgesics by understanding the outcomes of such medications.

Trends in opioid use among patients with cancer

FDA Priority Area/Regulatory Science Challenge: The CDC reports that in the US, an average of 130 people die every day from opioid overdose. In Connecticut, the drug overdose mortality rate in 2016 was 25.1 per 100,000, compared to a national rate of 17.1 per 100,000. While much of the opioid-related morbidity and mortality stems from the illicit use of synthetic opioids, opioid over-prescribing within the healthcare system has led to a substantial public health crisis, including misuse, overdose, and death. In response, urgent efforts are being deployed at institutional, state, and federal levels to improve opioid stewardship and reduce over-prescription. However, measures to restrict access to opioids for people who may potentially misuse them may inadvertently limit access to patients who need these medications for otherwise intractable pain.

Little is known about how efforts to improve overall opioid stewardship have affected cancer-related prescribing practices across the continuum of the disease. For instance, patients with advanced or metastatic cancer are at high risk of experiencing acute and chronic pain that can impact functioning and quality of life. In contrast, opioid over-prescription is associated with potential harm for patients with localized cancers, who seldom experience enduring cancer-related pain. While national efforts have decreased overall opioid prescribing, the impact of these efforts on the care of patients with cancer is unclear.

Project Description: This project will conduct a population-level analysis using SEER-Medicare data to assess trends in receipt of new opioid use and new persistent opioid use among Medicare beneficiaries. The project will also conduct a health system analysis using linked Flatiron Health-EMR data to assess trends in new opioid prescription and new persistent opioid use among patients with cancer in a large cancer care delivery system. This project aims to understand the impact of national efforts to decrease overall opioid prescribing on cancer patients for whom these medications may be intended or required.

Sex differences in immune profiles of PASC before and after vaccination

FDA Priority Area/Regulatory Science Challenge: Women are more likely than men to be affected by Post-Acute Sequelae of SARS-CoV-2 Infection (PASC), yet the reasons are unknown. Such knowledge has implications for diagnostic and therapeutic strategies. During acute COVID, men have a greater risk of adverse events and sex differences in immune responses have been reported, with men having greater innate cytokine and women greater T cell responses. Moreover, emerging, but preliminary surveys find that ~40% of long haulers report symptom improvement after receiving mRNA vaccines. Based on these observations, we hypothesize that a persistent viral reservoir and/or autoimmune responses may cause PASC. Vaccines may alleviate PASC due to the induction of robust antibodies and T cells to the spike protein, leading to clearance of the viral reservoir and/or modulating autoreactive B and T cell function. There are likely sex differences in the causes of PASC and, if vaccination changes PASC in some people, there likely are sex differences in this response.

Project Description: This project will investigate sex differences in the immune profile and viral load surrogates of people with PASC and compare with those without PASC but evidence of prior infection. We will also study changes in symptoms/immune profiles after vaccination for those with PASC. We will specifically determine whether changes occur through antigen-specific adaptive immune stimulation or innate immune signals leading to transient immunomodulation and whether there are sex differences. This project aims to understand the sex differences in immune response in PASC before and after vaccination to determine how baseline immune responses differ from those of acute COVID, as well as to determine how vaccination alters the baseline immune status of long COVID patients and how such changes correlate with symptom changes.

A mixed methods research design to identify factors influencing prescriber decision-making about pain management and opioid prescribing

FDA Priority Area/Regulatory Science Challenge: Misuse of prescription opioids remains a public health crisis. Appropriate short-term use of these medications in opioid-naïve patients is indicated in select health care settings, but intentional short-term use is emerging as an under-recognized segue to unintended prolonged opioid use (UPOU). Clinical strategies aimed at preventing UPOU in health care settings are lacking due, in part, to the absence of information about how this poorly understood clinical phenomenon actually develops.

Investigators at Mayo Clinic organized a group of thought leaders to develop a conceptual framework to guide the study of UPOU and to identify potential targets for interventions to reduce UPOU. The framework is comprised of three domains, including (1) patient characteristics; (2) practice environment characteristics; and (3) opioid prescriber characteristics that interact to either facilitate or impede UPOU. This framework was leveraged to conduct a systematic review to investigate the characteristics of physicians who prescribe long-term opioid therapy for chronic pain. Several physician-level characteristics were identified but, in addition, important patient and environmental factors were identified which could influence provider decision-making and prescribing behavior. The findings of this systematic review provide novel preliminary data that are critical for developing the semi-structured interviews that are the foundational components of the individual informant interviews that will be used to identify the barriers and facilitators influencing decision-making regarding pain management and opioid prescribing.

Project Description: The goal of this project is to identify and quantify the barriers and facilitators influencing prescriber decision-making about pain management and opioid prescribing. We will leverage our innovative work surrounding UPOU, our experience using mixed methods research techniques, and novel pain medicine expertise to achieve this goal. First, we will identify factors that influence prescriber decision-making regarding pain management and opioid prescribing behavior, using participatory research methods utilizing informant interviews and focus groups to identify the facilitators and barriers influencing prescriber decision-making regarding pain management and opioid prescribing. Second, we will develop and field test targeted surveys aimed at quantifying the newly identified barriers and facilitators to optimal decision-making regarding pain management.

Advancing Clinical and Post-Market Surveillance of Medical Devices and Diagnostic Tests

Post-market surveillance with a novel mHealth platform

FDA Priority Area/Regulatory Science Challenge: Medical devices play an important role in advancing patient care and reducing morbidity and mortality. All devices must receive FDA approval before they can be marketed. Once medical devices are marketed, it is necessary and important to monitor their safety and effectiveness in real-world clinical practice. Safety concerns may emerge when these devices are used in significantly more patients than were studied before marketing and when longer duration of follow-up is available. FDA therefore requires post-market surveillance of medical devices to ascertain if devices perform as intended and detect any unexpected or serious adverse effects. While the FDA currently employs multiple post-market surveillance strategies, none of these mechanisms can capture longitudinal patient-reported outcomes nor can they integrate data from multiple sources.

This project aims to explore whether a new mobile health technology could aid in the FDA's post-market surveillance of medical devices by obtaining health data from medical records allowing for better insights into patients' outcomes, as well as information on patient-reported symptoms and experiences. The proposed project will inform how the FDA may use novel and emerging technologies as it increasingly adopts a life-cycle evaluation approach to medical device regulation.

Project Description: A novel health data sharing platform (Hugo) has been developed that unobtrusively enables patients to provide their own outcomes (through short questionnaires and through synchronizing data from mobile health trackers) to the FDA after they have received a procedure that utilizes medical devices. In addition, with user permission, this application draws data from the electronic health record (EHR) to complement patient-reported data. In this project, we will conduct a pilot study testing this mobile health application to enable the FDA to conduct post-market surveillance of two procedures that use medical devices: the multiple devices (including sutures and stapler) used to perform bariatric surgeries (either sleeve gastrectomy or gastric bypass) in patients seeking weight loss and an ablation catheter when used in patients with atrial fibrillation seeking a return to sinus rhythm. Patients will be enrolled before receiving each of the devices and then will be asked to report specific symptoms related to their need for the procedure and those that may be expected at baseline (enrollment, which is pre-procedure), and 1, 4, and 8 weeks post-procedure.

Additionally, patients will be asked 2-3 short questions every 3-4 days for the first 30 days post-procedure related to post-procedure symptoms. We will also test if these patients' EHR data from multiple health systems where they receive care can be synchronized into a research-ready database. Patients will also be provided with syncable devices to provide additional insights into their health and health outcomes. Finally, we will test the feasibility of obtaining medication data from pharmacies or the current needs to create a functional system that can integrate pharmacy data into the mobile application. Integration of these multiple data sources (patient-reported outcomes, wearable/mobile device data, EHR data, and pharmacy data) have the potential to ultimately enable a more robust and thorough post-marketing surveillance strategy by leveraging the potential of digital health technologies. Non-federal entities involved in this project include Me2Health (the developers of the Hugo health data sharing platform), Johnson & Johnson (collaborator- provides input on project and funds to Me2Health for development of Hugo health data sharing platform) and AliveCor (donated Kardia Mobile devices for this project). Identifier: NCT03436082

Non-Federal Entity Collaborators: Karla Childers, MSJ, Paul Coplan, ScD, MBA, and Stephen Johnston, MSc (Johnson and Johnson) and Sanket Dhruva, MD, MHS- Co-Investigator (University of California, San Francisco)


Real world short- and intermediate-term safety outcomes following atrial fibrillation ablation

FDA Priority Area/Regulatory Science Challenge: Atrial fibrillation (AF) is the most common abnormal heart rhythm in humans and is a leading cause of stroke. Over the last two decades, catheter ablation has emerged as an effective modality for treating some patients with AF and its use has been rapidly increasing, especially with the introduction of new technology that allows more effective treatment. However, the "real-world" incidence of short- and intermediate- term complications following AF ablation is not well-characterized in the modern era. Specifically, the development of atrioesophageal fistula (AEF) - an abnormal connection between the heart and the esophagus (food pipe) - is a rare but deadly complication of catheter ablation of AF. It is widely believed that the diagnosis of AEF is often missed and/or underreported, and that the true incidence of AEF in the "real world" is higher than the limited study data suggest. This research proposal aims to use Real World Evidence to understand the incidence of short- and intermediate-term complications of catheter AF ablation, with an emphasis on AEF.

Project Description: Using claims data from the OptumLabs database, we will first characterize the 30- and 90-day rates of acute care use (including ED visits and hospitalizations) for patients undergoing AF ablation from 2011-2017. Next, using an algorithm, we will characterize the 30- and 90-complications that are likely to be related to AEF.

Non-Federal Entity Collaborators: James Hummel, MD- Co-Investigator (University of Wisconsin School of Medicine) and Sanket Dhruva, MD, MHS- Co-Investigator (University of California, San Francisco)


Implementation of a stakeholder survey to facilitate enhanced pediatric medical device innovation

FDA Priority Area/Regulatory Science Challenge: Since its inception in 1976, the Center for Devices and Radiological Health (CDRH) at the FDA has been dedicated to the protection and promotion of public health. While the number of devices approved for adults continues to rise, little change has occurred in the number of devices developed for children. Over the years, stakeholders have raised concerns about a lack in the availability of devices designed and tested for children. To encourage the development of devices for underserved populations, Congress passed the Safe Medical Devices Act of 1990, and in 2007 Congress passed the Pediatric Medical Device Safety and Innovation Act (PMDSIA). Despite these actions, minimal improvement has been seen over the last 30 years in the number of devices developed for children. Pediatric device development is hindered by a number of factors including challenges in consent and enrollment in clinical trials, variations in anatomy and pathophysiology, and small, heterogeneous and geographically disparate populations.

Creating a marketplace that supports development of technologies which serve the complexities of children may accelerate medical technology innovation for all Americans. This project aims to both identify the current challenges and barriers for the development and approval of pediatric medical devices by interviewing key stakeholders as well as to prioritize the barriers to market that will inform regulatory decision-making and promote innovation.

Project Description: The goal of this project is to survey members of industry to better inform the growth of a marketplace that supports the development of devices for pediatric populations that are approved or cleared and labeled for use in pediatrics. The project will use both qualitative and quantitative methods to gain insight from key stakeholders from the device industry, research community, investors, and payers.

Non-Federal Entity Collaborators : Andrew Lo, PhD- Co-Investigator (Massachusetts Institute of Technology)
Evaluating mobile health tool use for capturing patient-centered outcome measures in heart failure patients

FDA Priority Area/Regulatory Science Challenge: Heart failure (HF) is a highly prevalent disease that also carries high morbidity and mortality. Improvements in mortality and healthcare utilization, including hospitalizations, remains the gold standard outcome for HF drug approvals. However, it is difficult to improve mortality as the only endpoint due to the variation in the age groups and comorbidities of the population and ineffectiveness to alter all-cause mortality. Considering these issues, there is a need for exploration of complimentary endpoints. The FDA recognizes the importance of developing patient-centric endpoints that are relevant to patients beyond mortality and hospitalizations. Patients with HF have substantially reduced functional capacity and quality of life (QoL) and it is imperative to explore interventions that impact endpoints that directly measure how a patient feels or functions on a daily basis.

Recently, new mobile health technologies have emerged as clinical tools and offer an opportunity to overcome the challenges in measuring functional capacity and recording symptoms. These technologies capture and integrate data from disparate sources reflecting patients' functional status and symptomatology and have the potential of serving as surrogate endpoints for new HF therapy approvals.

Project Description: The goal of this project is to test the feasibility and reliability of capturing quantifiable measures of functional capacity and QoL using a wearable sensor in HF patients for a period of 60 days. Acute Decompensated Heart Failure (ADHF) patients will be recruited post-discharge from National Heart Centre and National University Hospital in Singapore. Patients will be monitored at home using the Biofourmis' BiovitalsHF™ platform which will capture biosensor data from two wearable devices: Everion® and Apple Watch Series 4. Patients will also use the BiovitalsHF™ smartphone application to capture electronic patient reported outcomes (ePROs) such as medication adherence, symptoms, the Kansas City Cardiomyopathy Questionnaire (KCCQ) responses, and perform the guided mobile-based 2-minute-step-test.

Non-Federal Entity Collaborators: Kuldeep Singh Rajput, Trace Brookins, Rachel Chan (Biofourmis)

Development and Application of Novel Analytics

Utilization and adverse events associated with mechanical circulatory support devices among patients with acute myocardial infarction and cardiogenic shock undergoing PCI

FDA Priority Area/Regulatory Science Challenge: When the heart muscle is severely and/or suddenly weakened, it is unable to deliver blood, oxygen, and nutrients to the body, a condition referred to as cardiogenic shock. There are many causes of cardiogenic shock, with the most common being a heart attack. However, a clinically meaningful classification system for cardiogenic shock has yet to be developed and all patients who have an abnormally low blood pressure or oxygen requirement are treated as having cardiogenic shock. It is highly likely, though, that there are distinct subgroups of patients experiencing cardiogenic shock that represent distinct combinations of risk factors and cardiac status. Because all patients with cardiogenic shock are lumped together, many of the clinical trials that have been conducted to investigate the impact of various interventions have failed to show benefit. This may derive from the fact that there is a mixture of different types of patients or different types of cardiogenic shock, some of which may respond quite favorably to an intervention while others may not have a favorable response.

This project uses Real World Evidence from the National Cardiovascular Data Registry (NCDR) to identify subgroups of cardiogenic shock patients undergoing percutaneous coronary intervention and characterize the usage of different mechanical circulatory support devices. The proposed project will enhance regulatory science by improving understanding of cardiogenic shock, characterizing contemporary utilization patterns of mechanical circulatory support devices, and identifying differences in the utilization and adverse events associated with devices when used in patients with cardiogenic shock, which may inform benefit-risk decisions. Ultimately, this work is expected to inform efforts to improve health outcomes for patients with cardiogenic shock.

Project Description: This project proposes to advance our understanding of cardiogenic shock with the ultimate aim of enabling patients and providers to estimate risk and develop optimal, individualized treatment plans. Specifically, we will use the NCDR CathPCI and Chest Pain-MI registries, two national registries of patients with acute myocardial infarction (Chest Pain-MI) and patients undergoing stent procedures (CathPCI) to determine the utilization patterns of devices in cardiogenic shock. We will then use advanced analytic methods to identify distinct subgroups of patients and test for differences between subgroups. Collaborators from Texas A&M University are involved in this project.

Non-Federal Entity Collaborators: Jeptha Curtis, MD, Frederick Masoudi, MD, MSPH, and John Messenger, MD- Co-Investigators (American College of Cardiology), Sanket Dhruva, MD, MHS- Co-Investigator (University of California, San Francisco), and Saket Girotra, MBBS, MS- Co-Investigator (University of Iowa)


Understanding the use of existing real-world data for medical product evaluation

FDA Priority Area/Regulatory Science Challenge: Clinical trials are considered the gold standard for understanding the safety and efficacy for any clinical or health system intervention, and results from clinical trials nearly always form the basis of FDA regulatory evaluations. Likewise, clinical trial data has traditionally been prioritized when making clinical practice guidelines and treatment decisions. However, in recent years the potential to use observational research for medical product evaluation has improved due to availability of clinical data, as well as the increasing detail provided in clinical data. In addition, advances in computing and the development of statistical methods are increasingly making it possible to use large-scale, real-world data to inform our understanding of the safety and effectiveness of medical products.

Project Description: The goal of this proposed research is to better understand the potential advantages and limitations of applying observational research methods to the use of existing real-world data for medical product evaluation. Focusing on drugs that have been approved for use by the FDA, this proposed research will use OptumLabs claims data to predict the trial populations and results of ongoing clinical trials used for regulatory evaluations, particularly those focused on drug safety. This proposal addresses a key methodological gap: existing work applying observational research methods to real-world data has focused on replicating the results of completed clinical trials whose results are already known. The two trials being replicated are PRONOUNCE (protocol can be found here) and GRADE (email for the protocol).

Non-Federal Entity Collaborators: William Crown, PhD, Co-investigator (Brandeis University), Sanket Dhruva, MD, Co-investigator (University of California, San Francisco), Eric C. Polley, PhD, Co-Investigator (University of Chicago)

Bayesian adaptive basket trial designs for neoantigen based immunotherapy with borrowing strength across subpopulations within the trial and from external controls

FDA Priority Area/Regulatory Science Challenge: Cancer immunotherapy has changed the landscape of modern oncology. For example, immune checkpoint inhibitors have emerged as an effective form of immunotherapy. More recently, much attention has been emphasized on tumor-specific antigens (TSAs); these newly formed TSAs are called neo-antigens. The recent advancement of molecular diagnosis technologies such as the next generation sequencing dramatically increase the abilities to identify neo-antigens specific to a sub-group of tumors, generated by genetic variations. As a result, a range of cancer treatment strategies are being developed to target distinct antigens and tumors. The focus of this research project is thus novel early phase trial design for master protocols intended to allow the rapid identification of those drugs that will most likely continue to develop and possibly join the immunotherapeutic arsenal in the near future.

However, initial studies of immune checkpoint inhibitors were developed at a rapid pace in large basket protocols, but lack appropriate statistical property such as early stopping for futility in disease-specific cohorts. In addition, they have significantly improved outcomes in some patients, but not all, likely due to the tumor heterogeneity. Similarly, neo-antigen based trials often suffer from smaller treatment-eligible subpopulations; in addition, basket trials enroll patients with multiple distinct cancers, making randomization not practical. Therefore, basket trials typically use single-arm designs and hence are prone to selection bias and confounding.

Project Description: This study aims to develop a Bayesian basket trial design borrowing information across subpopulations within the trial, and to evaluate its efficiency, along with developing a Bayesian basket trial design borrowing information from external controls, and to evaluate its validity. We propose a novel basket design framework that will make sequential adaptive subgroup-specific decisions while possibly clustering subtypes that have similar response to treatment. We also propose to extend and compare existing Bayesian methods of historical borrowing in settings where randomization is not practical as in a basket trial. The goal of this project is to improve trial efficiency by allowing information borrowing across multiple cancers expressing the same antigen (public neoantigen), but also account for cancer heterogeneity, and to glean information from external data sources while minimizing the bias/confounding due to lack of randomization.

Informatics driven real world analysis of SARS-CoV-2 serologic response and in vitro diagnostic accuracy

FDA Priority Area/Regulatory Science Challenge: The SARS-CoV-2 pandemic has rapidly and dramatically changed healthcare and daily life. Many questions about the diagnostics for SARS-CoV-2 and the immune response to infection remain, including how long patients remain infected, the accuracy of various testing methodologies over time, if protective antibodies are formed, and how long an individual may be immune after infection. These unknowns make it difficult to develop definitive plans for a safe reopening for healthcare systems that are resuming routine services and, further, the phased reopening of society. The rapid implementation of serologic assays that have been approved through Emergency Use Authorization pathways has further complicated the assessment of SARS-CoV-2 prevalence, as largely unknown accuracy and likely variation between assays make it difficult to interpret population-level results.

Correlation of diagnostic test results with clinical disease is critical to guide our interpretation of serologic assays within the clinical laboratory. As a number of assays have been developed that assess different antibodies and epitopes, with significant uncertainty about the overall accuracy of and correlation between these assays, the data provided from this project will provide valuable information from a real-world data set to better understand serologic results. This will address several key areas of impact, including the advancement of regulatory science and inform regulatory decision-making for a critical public health need.

Project Description: The goal of this project is to gain a better understanding of the performance characteristics of commercial assays. Specifically, we will implement a real-time, digital phenotyping approach to identify patients with prior SARS-CoV-2 infection and assess the performance of various clinical assays. We will identify 1) patients with a previous SARS-CoV-2 nucleic acid test and 2) a diagnosis-based phenotype based on ICD-10 codes that are provider entered or extracted by natural language processing. We will compare RT-PCR results with subsequent SARS-CoV-2 serological testing using various commercial assays. We will also identify specimens appropriate for longitudinal serologic testing from patients who meet the phenotype definition beginning 8 weeks after the time of initial test positivity or diagnosis and banked for ongoing assay validation within the clinical laboratory. Protocol available here.

Evaluate application of artificial intelligence to adaptive enrichment clinical trials

FDA Priority Area/Regulatory Science Challenge: There are substantial uncertainties when designing randomized controlled trials (RCTs), leading to an increased risk of negative trials, even if a treatment is effective. Over the past decade, adaptive trials have become increasingly used to reduce resource use, shorten trial duration, and minimize patient burden. The FDA issued guidance on how to plan and implement adaptive designs, but it did not mention the use of artificial intelligence (AI), which offers great potential to increase the efficiency of RCTs. One type of adaptive trials is adaptive enrichment - if the interim analysis shows one subgroup has a more favorable response, the trial can be “enriched” by modifying eligibility criteria to either solely or predominantly enroll patients from this subgroup.

Pre-specifying the subgroups in adaptive enrichment can often be challenging due to the lack of prior knowledge on biological mechanisms that lead to heterogeneous treatment effect (HTE). Unlike conventional trials where the learning of HTE occurs after the trial is completed, AI could potentially identify HTE as the data accumulate during the trial and facilitate the subsequent enrollment to enrich certain subgroups that are more likely to respond. However, little is known about whether such AI methods are helpful and how to best implement the AI methods to meet regulatory requirements.

Project Description: The goal of this proposed research is to evaluate the use of AI to facilitate adaptive enrichment designs for clinical trials, which will inform methods and development of standards for future regulatory submissions that utilize AI. We will use a causal AI method to discover HTE subgroups for adaptive enrichment. This project addresses a key gap between the development of AI methods and the applications of these in adaptive enrichment clinical trials. The findings will inform development of standards for a data-driven approach to identify subgroups for adaptive enrichment, which will increase in trial efficiency. With data generated from this study, FDA will also be able to better evaluate RCTs that use AI, determine whether the methods are trustworthy, and provide guidance to the medical community and the industry for using AI in adaptive trials.

Identifying, selecting, and utilizing quantitative bias analysis methods

FDA Priority Area/Regulatory Science Challenge: When new therapeutics are approved by the FDA, the usual requirement is that at least two well-controlled clinical trials (i.e. Phase III pivotal trials) are conducted that independently provide evidence of efficacy. However, over the past decade, FDA has developed a framework and guidance for utilizing real-world data (RWD) and observational studies to inform regulatory decision-making, including the approval of new indications for approved drugs and monitoring postmarket safety and adverse events. However, unlike randomized controlled trials with double-blind allocation, observational studies have inherent methodological limitations that can generate bias and confounding.

In order to rely on observational evidence, it is important to evaluate the potential impact of bias arising from systematic errors. Therefore, quantitative bias analysis (QBA) methods have been developed to evaluate the potential impact of bias arising from systematic errors in observational studies. However, there are a number of challenges that have undermined the widespread adoption of QBA in observational studies, including analytical complexities, difficulties establishing bias parameters, and concerns about the potential misuse/misinterpretation of methods and results. In order for observational studies to inform regulatory decision-making, it is necessary to have a comprehensive understanding of QBA methods that can be used to evaluate the robustness of observed associations.

Project Description: The goals of this project are to (1) systematically identify, summarize, and compare QBA approaches, with a focus on describing the applicable study designs, types of biases addressed, mathematical formulas and bias parameters, data formats, and model assumptions; (2) develop a user-friendly decision tree which will allow investigators to identify QBA methods based on different study characteristics; and (3) test the feasibility of using the QBA decision tree to select appropriate QBA methods for published results from various observational studies. Overall, these evaluations will provide a robust understanding of the types of analytical methods of QBA that can be utilized when conducting observational studies. Protocol available here.

Community-level emerging substance misuse simulation

FDA Priority Area/Regulatory Science Challenge: The FDA Office of Translational Sciences and Center for Drug Evaluation and Research are developing and validating a community- level substance use model to help inform the FDA’s approach to addressing the national polysubstance epidemic at the community level. This work was launched with FDA partnering with INOVA and the City of Alexandria focused on leveraging real-world data from a community, FDA Opioid Data Warehouse (ODW), and research papers to enhance, calibrate, and measure the veracity of the FDA’s agent-based model (ABM) and approach developed by previous research.

Project Description: This project aims to analyze polydrug use prior to concurrent use of specified classes of non-opioid drugs as influences on the timing and likelihood of transitions of use and outcomes such as treatment outcomes, overdose, and fatality. This project also aims to assess naloxone use patterns/harm reduction strategies by developing a series of simulations that evaluate the effectiveness of naloxone use by examining (a) strength/dose/packaging of naloxone; (b) distributions strategies – such as hospital emergency departments, ambulance calls, police overdose care and transportation reports, health clinics, schools, workplaces, post-incarceration and/or home health settings; (c) effectiveness on different types of opioids and populations; (d) presence of bystanders at an overdose event and bystander possession of naloxone or basic lifesaving skills.

Evaluate the Application of Machine Learning Algorithms to the Management of Postpartum Hemorrhage

FDA Priority Area/Regulatory Science Challenge: Postpartum hemorrhage (PPH) remains a leading cause of maternal morbidity and mortality in the United States, accounting for 10.7% of pregnancy-related deaths from 2014-2017 according to the CDC. The FDA Office of Women’s Health (OWH) has identified PPH as a priority in addressing maternal morbidity and mortality and improving care and outcomes for diverse populations of women. PPH is typically managed in a stepwise manner, in which specific actions may be taken as the condition progresses. However, despite this widely implemented pattern of care, there is substantial variation in its execution regarding sequential order, timing, and pace. This variation likely contributes to differences in maternal morbidity and mortality outcomes. More sophisticated analytic tools are needed to inform clinical decisions in PPH management and developing these tools requires rigorous examination of trajectories of obstetrical care. This project will use machine learning techniques including traditional supervised approaches in combination with a Bayesian Network to examine care patterns in the management of PPH across several diverse populations of women. Use of machine learning as a novel analytic tool will help determine the sources of variation in PPH management with the goal of refining current clinical practices to improve maternal care and outcomes.

Project Description: This project will use existing Epic electronic medical record (EMR) datasets from two academic medical centers (Yale University and The University of Pennsylvania) to create a reliable and validated data source. Advanced machine learning techniques will be applied to the Epic data to evaluate patient attributes associated with different patterns of PPH management. Finally, Epic data will be evaluated to determine if certain trajectories of PPH management are associated with increased risk of hysterectomy, other severe maternal morbidities, and maternal death. This study aims to answer the following: 1) are specific clinical and sociodemographic attributes associated with variations in PPH management, 2) are variations/deviations from accepted patterns of PPH management associated with higher rates of severe maternal morbidity, including the need for hysterectomy, and 3) can specific policies/ guidelines be developed to reduce deviations in care with the aim to improve outcomes for diverse populations of women? The ultimate goal is to use advanced machine learning analytic techniques to better understand variations in PPH management in order to inform the development or modification of regulatory standards that involve FDA-approved products related to the management of pregnancy complications such as PPH.

Non-Federal Entity Collaborators: Heather Burris, MD, MPH (University of Pennsylvania), Kevin Dysart, MD (Nemours Children’s Hospital), Sara Handley, MD, MSCE (University of Philadelphia), Kathryn McKenney, MD, MPH (University of Colorado Medicine)

Real-world outcomes of novel PET imaging tracers for prostate cancer

FDA Priority Area/Regulatory Science Challenge: New imaging tools have the potential to change cancer management in ways that are not well understood. Recently, imaging tools- such as positron emission tomography (PET) scans, have been developed help diagnose and treat several types of cancer. By better identifying cancer, it is believed that these tests will also improve how patients are treated as well as the outcome of their disease. Molecular imaging is particularly promising for prostate cancer, the second leading cause of cancer death in males, and a disease in which precisely locating areas of spread (metastasis) has previously been difficult. One of the most promising new PET scans is known as Prostate Specific Membrane Antigen (PSMA). PSMA-PET appears to be more accurate at finding small amounts of prostate cancer in the body as compared with standard tests such as Computerized Tomography (CT scans).

The effects of widely using PSMA-PET in patients with prostate cancer are not known but require attention. This is because no studies have yet examined the long-term outcomes of patients after PSMA-PET imaging to determine whether these scans help patients lead longer lives or reduce the burden from their disease. In addition, PSMA-PET, like any diagnostic test, is imperfect and there are risks of missing or incorrectly identifying areas as cancerous. As a result, patients may start to begin treatments with potential side effects earlier, which would impact their quality of life. In this study we will examine several unknown questions about PSMA-PET relating to (1) how testing affects cancer staging (2) how testing affects how prostate cancer is treated, and (3) how testing may affect patient outcomes after treatment.

Project Description: The primary goal of this project is to create new knowledge about the scientific results associated with PSMA-PET imaging for patients with prostate cancer. In the first phase of this study, we want to reduce uncertainty about the short-term clinical decisions that occur after PSMA-PET imaging in prostate cancer. By studying groups of patients who have previously undergone PSMA-PET imaging we will learn about the results of their testing and what treatment decisions were made afterward. In the second phase of this study, we will create decision models to estimate the long-term effects of PSMA-PET imaging for patients, focusing on how testing may impact length of treatment, side-effects, quality of life and overall survival.

Development of optical and digital microscopy hardware, software, and statistical methods to enable computational pathology in low resources

FDA Priority Area/Regulatory Science Challenge: Many patients are seen in low resource community care centers and practices close to their homes. There have been significant advances in technology, such as the creation of digital imaging microscopes and artificial intelligence software, to improve the health of patients. But these expensive platforms are not possible options for most low resource environments. Not having access to these tools leads to further health inequities.

This project aims to improve general knowledge around these new tools by expanding data with real world evidence from community healthcare providers. The data collected will be used to develop affordable tools and best practices for low resource environments.

This project will focus on the common practice of using a panel of experts to score the density of immune cells in digital pathology images of H&E slides that will be used as a reference standard for software developers who will be creating tools for digital images for use in clinical care. Including community healthcare providers in this work will reveal the challenges and limitations arising in low resource environments. In this project we will summarize best practices and create user-friendly tools to expand understanding of how artificial intelligence models are developed and can be used in clinical practice.

Project Description: This project will focus on developing tools and a guide for studies that are used to assess the performance of artificial intelligence software tools. This includes designing hardware that can be controlled virtually through the web, software, and statistical methods that can be used in all environments, especially low resource. This study will be a powerful method to determine the validity of artificial intelligence software tools.

We will create a virtual bridge between existing software on the precision FDA website for immune cell density scoring and a physical microscope for remote control of the microscope. We will create a software tool for controlling a local microscope by a local computer. Another software tool will be created to connect the local computer to special web-based digital pathology software. The online software will be modified to accept the virtual bridge and to control the microscope from the precision FDA website. These tools are currently being used independently of each other for viewing and collecting data from pathologists. After we integrate these tools together as one unit, we will develop worklists that specify the tasks that need to be completed for each slide. Following this, we will create task-specific workflows with the previous worklists.

A panel of experts will also be used to score immune cell density in these images. This data will be used to train pathologists to the task so they may participate in studies to assess performance of related artificial intelligence software. These evaluations will help in the development of tools that can be used to assess performance of a diverse population of pathologists using artificial intelligence software.

Fostering Patient-Centered Decision Making

Qualitative analysis of gender differences in heart failure PROs

FDA Priority Area/Regulatory Science Challenge: Cardiovascular disease is the number one cause of death for women in the world. In recent years, approximately 1 in every 4 deaths in the United States results from heart disease, with similar rates among men and women. Heart failure (HF) patient-reported outcomes (PROs), like the Kansas City Cardiomyopathy Questionnaire (KCCQ), have been shown to correlate with the New York Heart Association classification system and to predict hospitalization and death. Instruments like the KCCQ are regularly used as endpoints in clinical studies and patient care for their utility. The differences in presentation and symptoms of the disease in female and male patients are well documented. However, because the instrument was developed and validated on a majority male population (69% male in the original study), there are large differences in scores between male and female patients with the same level of disease severity. Regulatory decisions being made upon these results may not adequately represent the symptoms and experiences of all patients.

An improved understanding of the KCCQ to address both male and female patients will advance clinical study design and analyses of HF trials. Furthermore, this project will also address improved health communication by initiating the work to make KCCQ clear and applicable to female patients, improving both patient care and regulatory decisions.

Project Description: This project will focus on the initial qualitative analysis methods to understand concept and item interpretation by patients. The qualitative work in this study will further our understanding in KCCQ response differences (in particular, potential for response bias) between genders. This project aims to further validate the use of the KCCQ among female HF patients and provide an example of how to account for gender differences in patient-reported outcomes.


Methods to capture post-market patient preference information

FDA Priority Area/Regulatory Science Challenge: Patient preferences strongly influence use (or disuse) of medical products after they are approved for market. While the FDA incorporates patient preferences in the regulatory process, their efforts have mostly focused on pre-approval evaluations. However, these preferences may not be appropriate or helpful for understanding and acting on patient preferences once a drug or device becomes available. Post-market preferences are impacted by lived experiences and expectations of illness and treatment, and arise in the context of making real rather than hypothetical decisions. The opioid epidemic is an example in which post-market patient preferences are a significant factor in appropriate medication use or abuse, as a large proportion of patients sought out opioid products for pain management instead of other pain management strategies.

The methods for understanding post-market decisional preferences are underdeveloped. Certain methods, such as focus groups, often only reveal speculative preferences with respect to the decisions. Approaches more central to decision-making may include direct from encounter (DFE) methods- a third party observing and analyzing the encounter; or, immediately post-encounter methods (IPE)- ‘debriefing’ through key informant interview or by self-reported survey immediately after the encounter.

Project Description: This study aims to characterize differences between DFE and IPE approaches when used to evaluate real-world device decision preferences in terms of: a) time demands on patients and evaluators (including instrument administration and analysis); b) preferences identified: number, nature; c) identification of the contextual factors that bear on expressions of preference e.g. living circumstances; d) the extent of post-hoc justifications for preferences; and e) identification of interactions of preferences (e.g. preference for higher out-of-pocket cost to avoid undesirable device effects). The goal of this project is to provide initial real-world evidence of the strengths and weaknesses of DFE and IPE evaluation approaches that can be used to inform the funding and development of these methods or their enhancements toward establishing their validity and responsiveness.

Quantifying physical function in cancer patients undergoing chemotherapy using clinician-reported, patient-reported, and wearable device data sources

FDA Priority Area/Regulatory Science Challenge: Patients’ performance status has long been an important prognostic measure in cancer clinical trials and is often used as a stratification factor. In addition to its prognostic importance, physical function and the ability to take care of oneself has been noted to be a meaningful outcome that is important to cancer patients. Recognizing the importance of symptoms and function and how these outcomes may be influenced by both safety and efficacy of cancer therapies, disease symptoms, symptomatic adverse events, physical function (PF) and ability to work and perform leisure activities have been identified as core clinical outcomes. In order for PF to be considered for regulatory and treatment decisions, accurate measurement of this core outcome is needed.

There are several approaches to quantitatively assess PF in cancer patients including using wearable devices as well as other measures of clinician-reported outcomes (ClinRO), patient-reported outcomes (PRO), and performance outcomes (PerfO, such as a 6 minute walk test). There are different strengths and limitations to data from each of these four distinct PF assessment modalities, and an improved understanding of their logistical feasibility, sensitivity, complementarity and meaningful levels of change are needed in order to design appropriate endpoints and analyses for regulatory review.

Project Description: This study aims to characterize assessment challenges with various PF modalities, compare change over time between various PF modalities, identify clinically meaningful changes in the various assessment methods, explore ability of PF changes to predict serious adverse events, hospitalization or emergency department visits, and use structured exit interviews to understand burden and usability of electronic PROs and wearable devices. The goal of this project is to evaluate the measurement characteristics of these different sources of physical function data in a cancer population undergoing cytotoxic chemotherapy. Click here for more information.

Non-Federal Entity Collaborators: William Wood, MD (University of North Carolina at Chapel Hill)


Patient and provider informed labeling of AI/ML-based software to enable transparency and trust

FDA Priority Area/Regulatory Science Challenge: Artificial Intelligence (AI) is increasingly used in health care products and services to improve prevention, diagnosis, treatment, and maintenance of health conditions. While there are many promising applications of AI and Machine Learning (AI/ML) in health care, the device community has yet to develop a trustworthy and transparent approach to informing patients and providers about AI/ML. Current device labeling does not always address the unique challenges of communicating AI/ML-based software, such as data capture, training data sources, model accuracy, potential biases, opting out of use, and AI autonomy. This gap can hinder shared patient decision-making and public trust in AI/ML enabled medical devices. Additionally, although existing research has begun to explore patient and provider perspectives on AI transparency and trust, data on how to convey needed information about AI/ML to cultivate merited trust remain sparce and largely unactionable. More systematic research, engaging both patients and providers, is needed to uncover the specific content and communication requirements for trusted consumer understanding and use of AI/ML-based medical devices and to support patient-informed labeling.

Project Description: The goal of this project is to build a dynamic patient-centered digital labeling approach for AI/ML-based medical devices, which can be publicly available and adjustable to suit various types of AI/ML-based software. We will engage patients, providers, and subject matter experts in two objectives using online focus groups and crowdsourcing methods. First, we will develop hypothetical AI/ML-based software and device scenarios to gauge patient and provider attitudes toward AI/ML and identify labeling needs. Second, we will create prototype digital labels through iterative information design and pilot-test the labels to assess patient acceptance, usability, and comprehension. The ultimate goal is to create a documented approach to developing provider and patient-centered labeling of AI/ML-based software. This dynamic approach will provide the core ingredients of patient-facing AI/ML-based software labels and allow for tailoring recommendations, thereby addressing the gap in patient-centered communication and transparency about AI/ML software.

Improving the Diagnosis and Treatment of Women with Myocardial Ischemia and No Obstructive Coronary Artery Disease

FDA Priority Area/Regulatory Science Challenge: Ischemic heart disease (IHD) is a leading cause of morbidity and mortality in women. However, there are significant disparities in the outcomes of women with IHD. This is partly because women present with unique symptoms that are not fully consistent with traditional models of IHD, and women are less likely than men to have obstructive atherosclerotic coronary artery disease (CAD) which is the most widely recognized form of IHD.

Instead, women often suffer from myocardial ischemia and no obstructive coronary artery disease (INOCA). Although recognition of and research on INOCA has increased in recent years, there is no specific functional status measure, limiting healthcare providers’ ability to evaluate the course of illness or effectiveness of therapies for INOCA. This project will develop measures to better assess INOCA, thereby improving evaluation, diagnosis, and treatment of women with the disease.

Project Description: The goal of this project is to develop and validate a patient reported outcome measure (PROM) that can assess the overall health status of women with INOCA. The PROM will be in the form of a self-report questionnaire and will measure patients’ frequency/stability of symptoms, physical limitation, perception of illness, psychological toll, and functional limitation. To achieve this goal, we will conduct structured interviews with women with INOCA. Using the constant comparison method, essential concepts from the interview data will be coded and compared over successive patient interviews to extract recurrent themes. These themes will contribute to initial development of the PROM of which we will evaluate content and construct validity with INOCA patients and further refine the PROM based on results. Finally, we will evaluate criterion validity of the PROM for use with specific types of INOCA (i.e., coronary microvascular dysfunction, coronary vasospasm, or other disorders of coronary physiology). The resulting PROM may be used by healthcare providers and their patients to improve the assessment of symptoms and impact of therapies for women with INOCA.

Qualitative techniques to define meaningful within-patient change in symptoms of advanced cancer patients

FDA Priority Area/Regulatory Science Challenge: FDA Priority Area/Regulatory Science Challenge: Patients receiving cancer treatment often experience changes in cancer symptoms or treatment side effects, such as fatigue, nausea, or pain. Questionnaires called “patient-reported outcome measures” are commonly used to collect information from patients on their symptoms and side effects throughout the course of treatment and even after treatment is completed. It is important for cancer care teams to understand how to interpret the information from these questionnaires. Specifically, they need to know whether the amount of change in cancer symptoms or treatment side effects is meaningful to patients with cancer. A better understanding of meaningful change could help cancer care teams evaluate the impact of different cancer treatments on their patients’ well-being and it could help cancer care teams to support patients more effectively during their treatment. Defining the concept of meaningful change and determining how much change is considered meaningful are challenging. The research in this area so far is limited, in that it rarely considers the perspective of patients. Qualitative research can be used to obtain a deeper understanding of the patient experience by generating in-depth information about the experiences, perspectives, priorities, preferences, and feelings of patients, in their own words. This study will apply qualitative research methods in selected groups of patients receiving cancer treatment to gain insight into the definition and measurement of meaningful change in cancer symptoms and treatment side effects.

Project Description: This study investigates how best to communicate the complex concept of meaningful change to patients and how best to incorporate patients’ perspectives when establishing meaningful change thresholds. We will use multiple qualitative methods (such as semi-structured interviews and focus groups) to develop and test our methods for eliciting patient perspectives for defining meaningful change in selected patient-reported outcome measures. Analyses of interview and focus group data will focus on patients’ interpretation of meaningful change and how it translates to the change in actual questionnaire score in selected patient-reported outcome measures. The outcome of this project will be to (1) define meaningful change from patient perspectives, (2) understand the amount of change in selected patient-reported outcome measures that patients consider meaningful, and (3) evaluate new methods for engaging patients in studying meaningful change.

A positive deviance approach for representing women, older adults and patients identifying as racial and ethnic minorities in oncology research

FDA Priority Area/Regulatory Science Challenge: Policy efforts to improve representation and diversity in clinical research span decades, yet disparities in enrollment persist. New cancer therapies are often tested in clinical trials that enroll patients who are younger, healthier, and more likely to report White race than patients with conditions who are not in a research study. Women are under-represented in trials for some types of cancer too. When the demographics of the trial participants do not represent the demographics of patients for whom treatment is indicated, clinicians, payers, and patients can be left wondering how to interpret the results from the trial and how to apply them to different groups of patients. Research examining diversity in clinical research has shown some pharmaceutical companies are better than others at enrolling representative participants. This research group proposes to study the factors that distinguish these exceptional performing sponsors from others, using a positive deviance approach, to determine shared behaviors, strategies and contexts enabling them to perform better that can be generalized and implemented by the broader research community. We anticipate findings will help advance health equity through better inclusion of under-represented populations in cancer drug research and development.

Project Description: We will use a positive deviance approach to identify shared strategies enabling trial sponsors to perform exceptionally, that is better than their peers, on enrolling diverse and representative participants in cancer research, that can be generalized and implemented by the broader research community to produce similar results. This will involve, first, identifying positive deviant sponsors, those demonstrating exceptionally high performance on adequately representing older adults, women, Asian, Black and LatinX identifying patients, via a cross sectional study. The study will compare trial participant demographics, for pivotal trials supporting novel cancer drugs approved by the FDA in 2012 through 2021, to those of the patients with the disease or condition targeted in a study, by constructing a participant-to-prevalence ratio (PPR) for each trial. Second, we will conduct a qualitative study to identify practices, strategies, and organizational contexts positive deviant trials and sponsors share, allowing them to achieve top performance.

Longitudinal analysis & visualization of patient-reported physical function & symptom data

FDA Priority Area/Regulatory Science Challenge: Physical function and side effects of treatment as reported by patients through surveys are important in understanding how well patients tolerate cancer treatment. As such, the FDA Oncology Center of Excellence has recommended physical function and side effects be measured by patient surveys during most or all cancer treatment clinical trials. However, there is a lack of standardized approaches to analyzing and displaying survey data on physical function and side effects in a way that is valuable to patients, clinicians, and decision makers at the FDA. Additionally, little is known about how physical function and side effects change over time in patients with rare cancers, leaving these patients with little to no information about what to expect on their treatments.

This study aims to better understand physical function and side effects in patients with rare cancers and determine how to communicate patient-reported physical function and side effects for patients, clinicians, and decision makers at the FDA. The content of this project aligns with work by colleagues in the FDA Oncology Center of Excellence, including the FDA’s Project Patient Voice (PPV), a pilot online platform for patients and clinicians to visualize survey data from cancer trials.

Project Description: In the proposed study, we will develop tools to statistically analyze and visualize patient survey data on physical function and side effects from multiple trials in rare cancers including amyloidosis, myeloproliferative neoplasms, carcinoid, sarcoma, and esophageal cancer.

This study aims to (1) develop novel graphics of physical function and side effect data collected by patient surveys from rare cancer clinical trials and (2) to obtain initial feedback from patient advocates, clinicians/clinical investigators, and decision makers at the FDA on these visualizations. The ultimate goals of this project are to refine multiple depictions over time of physical function and side effect data from patient surveys from rare cancer trials, develop graphical representations that may complement those in FDA’s Project Patient Voice, and obtain preliminary feedback from patients, clinicians, and decision makers at the FDA on their interpretability and ability to reflect information on the tolerability of a cancer treatment.

Universal Common Data Model Mapping – Implementation and Validation
The ability to access population-scale data is essential for many emerging areas of biomedical research. Precision medicine, which focuses on identifying the optimal treatment pathway for very refined patient phenotypes, or those with specific genetic variation, require large sample sizes to identify differences in patient outcomes. Similarly, other biomedical research areas that have a limited number of events, such as drug development and pharmacovigilance surveillance, or projects that rely on high dimensional data, have an intrinsic need for very large sample sizes. Often, individual researchers and institutions are unable to identify a sufficient number of participants within local patient populations. However, data sharing and integration between clinical research data management systems (CDMs) and electronic health record (EHR) systems, which would allow systems to coordinate and collaborate to conduct research of this type, remains a challenging issue. To address this problem, the medical informatics community has made significant advances through the development of data and semantic interoperability standards to encourage multi-institutional collaborations and enable researchers to rapidly identify relevant patient cohorts. In this project, we will validate the accuracy and efficiency of these systems for drug safety surveillance, by implementing a universal CDM mapping approach and comparing it to gold standard data from the clinical data warehouse.