FDA Perspective on the Regulation of Artificial Intelligence in Health Care and Biomedicine.

Importance: Advances in artificial intelligence (AI) must be matched by efforts to better understand and evaluate how AI performs across health care and biomedicine as well as develop appropriate regulatory frameworks. This Special Communication reviews the history of the US Food and Drug Administration's (FDA) regulation of AI; presents potential uses of AI in medical product development, clinical research, and clinical care; and presents concepts that merit consideration as the regulatory system adapts to AI's unique challenges. Observations: The FDA has authorized almost 1000 AI-enabled medical devices and has received hundreds of regulatory submissions for drugs that used AI in their discovery and development. Health AI regulation needs to be coordinated across all regulated industries, the US government, and with international organizations. Regulators will need to advance flexible mechanisms to keep up with the pace of change in AI across biomedicine and health care. Sponsors need to be transparent about and regulators need proficiency in evaluating the use of AI in premarket development. A life cycle management approach incorporating recurrent local postmarket performance monitoring should be central to health AI development. Special mechanisms to evaluate large language models and their uses are needed. Approaches are necessary to balance the needs of the entire spectrum of health ecosystem interests, from large firms to start-ups. The evaluation and regulatory system will need to focus on patient health outcomes to balance the use of AI for financial optimization for developers, payers, and health systems. Conclusions and Relevance: Strong oversight by the FDA protects the long-term success of industries by focusing on evaluation to advance regulated technologies that improve health. The FDA will continue to play a central role in ensuring safe, effective, and trustworthy AI tools to improve the lives of patients and clinicians alike. However, all involved entities will need to attend to AI with the rigor this transformative technology merits.

Why Evidence Generation Should Matter to Payers and How They Can Help.

Importance: The US leads the world in bringing new medical products to market, but the ability to generate evidence to inform clinical practice in postmarket settings needs improvement. Although a diverse group of stakeholders is working to improve postmarket evidence generation, the role of private payers has been underappreciated. Observations: Payers are crucial allies in improving evidence generation because better data would better inform coverage decisions, their policies and practices influence the conduct of care and research, and their claims data are a source of real-world evidence used in medical product evaluation. In addition, payers have a stake in improving evidence generation because the kinds of evidence needed to inform health care and coverage decisions are often not available when a product enters the market and may not be generated without their involvement. Here, we describe several key steps payers could take to improve evidence generation, including participating in efforts to reduce administrative and financial barriers to the conduct of clinical trials, directly incentivizing evidence generation on high-priority questions by funding potential cost-saving trials, increasing engagement with the medical products industry on evidentiary needs for coverage decisions, and improving usability of claims data by reducing data lags and routinely recording unique device identifiers. Broad payer engagement with US Food and Drug Administration recommendations regarding evidence generation will ensure that the opportunities to participate in clinical research are extended to all communities and that evidence needed to inform care is generated in trials and surveillance systems that reflect the clinical reality across the US. Conclusions and Relevance: Increasing payer involvement in evidence generation can benefit all participants in the medical innovation ecosystem. The importance of payers in these efforts will continue to grow in response to imperatives to increase integration of care and research, engage a diverse set of communities in clinical research, and move toward alternative payment models.

The Integration of Clinical Trials With the Practice of Medicine: Repairing a House Divided.

Importance: Optimal health care delivery, both now and in the future, requires a continuous loop of knowledge generation, dissemination, and uptake on how best to provide care, not just determining what interventions work but also how best to ensure they are provided to those who need them. The randomized clinical trial (RCT) is the most rigorous instrument to determine what works in health care. However, major issues with both the clinical trials enterprise and the lack of integration of clinical trials with health care delivery compromise medicine's ability to best serve society. Observations: In most resource-rich countries, the clinical trials and health care delivery enterprises function as separate entities, with siloed goals, infrastructure, and incentives. Consequently, RCTs are often poorly relevant and responsive to the needs of patients and those responsible for care delivery. At the same time, health care delivery systems are often disengaged from clinical trials and fail to rapidly incorporate knowledge generated from RCTs into practice. Though longstanding, these issues are more pressing given the lessons learned from the COVID-19 pandemic, heightened awareness of the disproportionate impact of poor access to optimal care on vulnerable populations, and the unprecedented opportunity for improvement offered by the digital revolution in health care. Four major areas must be improved. First, especially in the US, greater clarity is required to ensure appropriate regulation and oversight of implementation science, quality improvement, embedded clinical trials, and learning health systems. Second, greater adoption is required of study designs that improve statistical and logistical efficiency and lower the burden on participants and clinicians, allowing trials to be smarter, safer, and faster. Third, RCTs could be considerably more responsive and efficient if they were better integrated with electronic health records. However, this advance first requires greater adoption of standards and processes designed to ensure health data are adequately reliable and accurate and capable of being transferred responsibly and efficiently across platforms and organizations. Fourth, tackling the problems described above requires alignment of stakeholders in the clinical trials and health care delivery enterprises through financial and nonfinancial incentives, which could be enabled by new legislation. Solutions exist for each of these problems, and there are examples of success for each, but there is a failure to implement at adequate scale. Conclusions and Relevance: The gulf between current care and that which could be delivered has arguably never been wider. A key contributor is that the 2 limbs of knowledge generation and implementation-the clinical trials and health care delivery enterprises-operate as a house divided. Better integration of these 2 worlds is key to accelerated improvement in health care delivery.

Now is the time to fix the evidence generation system.

Despite enormous advances in biomedical science, corresponding improvements in health outcomes lag significantly. This is particularly true in the United States, where life expectancy trails far behind that of other high-income countries. In addition, substantial disparities in life expectancy and other health outcomes exist as a function of race, ethnicity, wealth, education, and geographic location. A major reformation of our national system for generating medical evidence-the clinical research enterprise-is needed to facilitate the translation of biomedical research into useful products and interventions. Currently, premarket systems for generating and evaluating evidence work reasonably well, but the postmarket phase is disaggregated and often fails to answer essential questions that must be addressed to provide optimal clinical care and public health interventions for all Americans. Solving these problems will require a focus on three key domains: (1) improving the integration of and access to high-quality data from traditional clinical trials, electronic health records, and personal devices and wearable sensors; (2) restructuring clinical research operations to support and incentivize the involvement of patients and frontline clinicians; and (3) articulating ethical constructs that enable responsible data sharing to support improved implementation. Finally, we must also address the systemic tendency to optimize individual components of the clinical research enterprise without considering the effects on the system as a whole. Overcoming suboptimization by creating incentives for integration and sharing will be essential to achieve more timely and equitable improvement in health outcomes.

Avoiding the Coming Tsunami of Common, Chronic Disease: What the Lessons of the COVID-19 Pandemic Can Teach Us.

During the past year, clinicians and the public have been focused on the coronavirus disease 2019 (COVID-19) pandemic and its associated societal and economic effects. However, once the acute phase of this crisis has passed, we will face an enormous wave of death and disability as a result of common chronic diseases (CCDs), with cardiometabolic diseases at the crest (Figure). A tsunami results when an earthquake on the ocean floor creates huge waves that can wreak devastation far distant from the original upheaval, especially when warnings are ignored. Similarly, underlying global and national demographic and risk factor profiles have for some time presaged an overwhelming burden of CCDs. However, although the pandemic has created additional impetus that unless heeded will amplify the consequences of this burden, the rapid adaptations and innovations in care and research prompted by the urgent response to it may also offer us the means to stem this flood.

Opioid Use and Its Relationship to Cardiovascular Disease and Brain Health: A Presidential Advisory From the American Heart Association.

The misuse of opioids continues to be epidemic, resulting in dependency and a recent upsurge in drug overdoses that have contributed to a significant decrease in life expectancy in the United States. Moreover, recent data suggest that commonly used opioids for the management of pain may produce undesirable pharmacological actions and interfere with critical medications commonly used in cardiovascular disease and stroke; however, the impact on outcomes remains controversial. The American Heart Association developed an advisory statement for health care professionals and researchers in the setting of cardiovascular and brain health to synthesize the current literature, to provide approaches for identifying patients with opioid use disorder, and to address pain management and overdose. A literature and internet search spanning from January 1, 2012, to February 15, 2021, and limited to epidemiology studies, reviews, consensus statements, and guidelines in human subjects was conducted. Suggestions and considerations listed in this document are based primarily on published evidence from this review whenever possible, as well as expert opinion. Several federal and institutional consensus documents and clinical resources are currently available to both patients and clinicians; however, none have specifically addressed cardiovascular disease and brain health. Although strategic tools and therapeutic approaches for recognition of opioid use disorder and safe opioid use are available for health care professionals who manage patients with cardiovascular disease and stroke, high-quality evidence does not currently exist. Therefore, there is an urgent need for more research to identify the most effective approaches to improve care for these patients.

Importance of Social Determinants in Screening for Depression.

IMPORTANCE: The most common screening tool for depression is the Patient Health Questionnaire-9 (PHQ-9). Despite extensive research on the clinical and behavioral implications of the PHQ-9, data are limited on the relationship between PHQ-9 scores and social determinants of health and disease. OBJECTIVE: To assess the relationship between the PHQ-9 at intake and other measurements intended to assess social determinants of health. DESIGN, SETTING, AND PARTICIPANTS: Cross-sectional analyses of 2502 participants from the Baseline Health Study (BHS), a prospective cohort of adults selected to represent major demographic groups in the US; participants underwent deep phenotyping on demographic, socioeconomic, clinical, laboratory, functional, and imaging findings. INTERVENTIONS: None. MAIN OUTCOMES AND MEASURES: Cross-sectional measures of clinical and socioeconomic status (SES). RESULTS: In addition to a host of clinical and biological factors, higher PHQ-9 scores were associated with female sex, younger participants, people of color, and Hispanic ethnicity. Multiple measures of low SES, including less education, being unmarried, not currently working, and lack of insurance, were also associated with higher PHQ-9 scores across the entire spectrum of PHQ-9 scores. A summative score of SES, which was the 6th most predictive factor, was associated with higher PHQ-9 score after adjusting for 150 clinical, lab testing, and symptomatic characteristics. CONCLUSIONS AND RELEVANCE: Our findings underscore that depression should be considered a comorbidity when social determinants of health are addressed, and both elements should be considered when designing appropriate interventions.

Reimagining What We Measure in Atherosclerosis-a "Phenotype Stack".

The term phenotype is so commonly used that we often assume that we each mean the same thing. The general definition, the set of observable characteristics of an individual resulting from the interaction of their genotype with the environment, is often left to the eye of the beholder. Whether applied to the multiple levels of biological phenomena or the intact human being, our ability to characterize, classify, and analyze phenotype has been limited by measurement deficits, computing limitations, and a culture that avoids the generalizable. With the advent of modern technology, there is the potential for a revolution in phenotyping, which incorporates old and new in structured ways to dramatically advance basic understanding of biology and behavior and to lead to major improvements in clinical care and public health. This revolution in how we think about phenotypes will require a radical change in the scale at which biomedicine operates with significant changes in the unit of action, which will have far-reaching implications for how care, translation, and discovery are implemented.

Call to Action: Rural Health: A Presidential Advisory From the American Heart Association and American Stroke Association.

Understanding and addressing the unique health needs of people residing in rural America is critical to the American Heart Association's pursuit of a world with longer, healthier lives. Improving the health of rural populations is consistent with the American Heart Association's commitment to health equity and its focus on social determinants of health to reduce and ideally to eliminate health disparities. This presidential advisory serves as a call to action for the American Heart Association and other stakeholders to make rural populations a priority in programming, research, and policy. This advisory first summarizes existing data on rural populations, communities, and health outcomes; explores 3 major groups of factors underlying urban-rural disparities in health outcomes, including individual factors, social determinants of health, and health delivery system factors; and then proposes a set of solutions spanning health system innovation, policy, and research aimed at improving rural health.

Impact of Regulatory Guidance on Evaluating Cardiovascular Risk of New Glucose-Lowering Therapies to Treat Type 2 Diabetes Mellitus: Lessons Learned and Future Directions.

Responding to concerns about the potential for increased risk of adverse cardiovascular outcomes, specifically myocardial infarction, associated with certain glucose-lowering therapies, the US Food and Drug Administration and the Committee for Medicinal Products for Human Use of the European Medicines Agency issued guidance to the pharmaceutical industry in 2008. Glucose-lowering therapies were granted regulatory approval primarily from smaller studies that have demonstrated reductions in glycated hemoglobin concentration. Such studies were overall underpowered and of insufficient duration to show any effect on cardiovascular outcomes. The 2008 guidance aimed to ensure the cardiovascular safety of new glucose-lowering therapies to treat patients with type 2 diabetes mellitus. This resulted in a plethora of new cardiovascular outcome trials, most designed primarily as placebo-controlled noninferiority trials, but with many also powered for superiority. Several of these outcome trials demonstrated cardiovascular benefits of the newer agents, resulting in the first-ever cardiovascular protection indications for glucose-lowering therapies. Determining whether the guidance continues to have value in its current form is critically important as we move forward after the first decade of implementation. In February 2018, a think tank comprising representatives from academia, industry, and regulatory agencies convened to consider the guidance in light of the findings of the completed cardiovascular outcome trials. The group made several recommendations for future regulatory guidance and for cardiovascular outcome trials of glucose-lowering therapies. These recommendations include requiring only the 1.3 noninferiority margin for regulatory approval, conducting trials for longer durations, considering studying glucose-lowering therapies as first-line management of type 2 diabetes mellitus, considering heart failure or kidney outcomes within the primary outcome, considering head-to-head active comparator trials, increasing the diversity of patients enrolled, evaluating strategies to streamline registries and the study of unselected populations, and identifying ways to improve translation of trial results to general practice.

Plasma Volume Status and Its Association With In-Hospital and Postdischarge Outcomes in Decompensated Heart Failure.

BACKGROUND: Prior analyses suggest an association between formula-based plasma volume (PV) estimates and outcomes in heart failure (HF). We assessed the association between estimated PV status by the Duarte-ePV and Kaplan Hakim (KH-ePVS) formulas, and in-hospital and postdischarge clinical outcomes, in the ASCEND-HF trial. METHODS AND RESULTS: The KH-ePVS and Duarte-ePV were calculated on admission. We assessed associations with in-hospital worsening HF, 30-day composite cardiovascular mortality or HF rehospitalization and 180-day all-cause mortality. There were 6373 (89.2%), and 6354 (89.0%) patients who had necessary characteristics to calculate KH-ePVS and Duarte-ePV, respectively. There was no association between PV by either formula with in-hospital worsening HF. KH-ePVS showed a weak correlation with N-terminal prohormone BNP, and with measures of decongestion such as body weight change and urine output (r < 0.3 for all). Duarte-ePV was trending toward an association with worse 30-day (adjusted odds ratio 1.07, 95% confidence interval [CI] 1.00-1.15, P = .058), but not 180-day outcomes (adjusted hazard ratio 1.03, 95% CI 0.97-1.09, P = .289). A continuous KH-ePVS of >0 (per 10-unit increase) was associated with improved 30-day outcomes (adjusted odds ratio 0.75, 95% CI 0.62-0.91, P = .004). The continuous KH-ePVS was not associated with 180-day outcomes (adjusted hazard ratio 1.05, 95% CI 0.98-1.12, P = .139). CONCLUSIONS: Baseline PV estimates had a weak association with in-hospital measures of decongestion. The Duarte-ePV trended toward an association with early clinical outcomes in decompensated HF, and may improve risk stratification in HF.

Understanding the use of observational and randomized data in cardiovascular medicine.

The availability of large datasets from multiple sources [e.g. registries, biobanks, electronic health records (EHRs), claims or billing databases, implantable devices, wearable sensors, and mobile apps], coupled with advances in computing and analytic technologies, have provided new opportunities for conducting innovative health research. Equally, improved digital access to health information has facilitated the conduct of efficient randomized controlled trials (RCTs) upon which clinical management decisions can be based, for instance, by permitting the identification of eligible patients for recruitment and/or linkage for follow-up via their EHRs. Given these advances in cardiovascular data science and the complexities they behold, it is important that health professionals have clarity on the appropriate use and interpretation of observational, so-called 'real-world', and randomized data in cardiovascular medicine. The Cardiovascular Roundtable of the European Society of Cardiology (ESC) held a workshop to explore the future of RCTs and the current and emerging opportunities for gathering and exploiting complex observational datasets in cardiovascular research. The aim of this article is to provide a perspective on the appropriate use of randomized and observational data and to outline the ESC plans for supporting the collection and availability of clinical data to monitor and improve the quality of care of patients with cardiovascular disease in Europe and provide an infrastructure for undertaking pragmatic RCTs. Moreover, the ESC continues to campaign for greater engagement amongst regulators, industry, patients, and health professionals in the development and application of a more efficient regulatory framework that is able to take maximal advantage of new opportunities for improving the design and efficiency of observational studies and RCT in patients with cardiovascular disease.

Call to Action: Urgent Challenges in Cardiovascular Disease: A Presidential Advisory From the American Heart Association.

Although advances in care have spurred improvements in cardiovascular outcomes, cardiovascular disease remains the leading cause of death in the United States and around the world. Previous declines in cardiovascular disease mortality have slowed and even reversed for certain demographics. Further concerns exist with regard to cardiovascular drug innovation, quality of care, and healthcare costs. The Value in Healthcare Initiative-Transforming Cardiovascular Care, a collaboration of the American Heart Association and Duke University, Robert J. Margolis, MD, Center for Health Policy, aims to increase access to and affordability of cardiovascular treatment and to decrease barriers to care. The following Call to Action describes trends in cardiovascular care, identifies gaps in areas of cardiovascular disease prevention and treatment, highlights challenges with medical product innovation, and finally, outlines a series of learning collaboratives that will aid in the development of road maps for transforming cardiovascular care.

Frequency, Regional Variation, and Predictors of Undetermined Cause of Death in Cardiometabolic Clinical Trials: A Pooled Analysis of 9259 Deaths in 9 Trials.

BACKGROUND: Modern cardiometabolic clinical trials often include cardiovascular death as a component of a composite primary outcome, requiring central adjudication by a clinical events committee to classify cause of death. However, sometimes the cause of death cannot be determined from available data. The US Food and Drug Administration has indicated that this circumstance should occur only rarely, but its prevalence has not been formally assessed. METHODS: Data from 9 global clinical trials (2009-2017) with long-term follow-up and blinded, centrally adjudicated cause of death were used to calculate the proportion of deaths attributed to cardiovascular, noncardiovascular, or undetermined causes by therapeutic area (diabetes mellitus/pre-diabetes mellitus, stable atherosclerosis, atrial fibrillation, and acute coronary syndrome), region of patient enrollment, and year of trial manuscript publication. Patient- and trial-level variables associated with undetermined cause of death were identified using a logistic model. RESULTS: Across 127 049 enrolled participants from 9 trials, there were 9259 centrally adjudicated deaths: 5012 (54.1%) attributable to cardiovascular causes, 2800 (30.2%) attributable to noncardiovascular causes, and 1447 (15.6%) attributable to undetermined causes. There was variability in the proportion of deaths ascribed to undetermined causes by trial therapeutic area, region of enrollment, and year of trial manuscript publication. On multivariable analysis, acute coronary syndrome or atrial fibrillation trial (versus atherosclerotic vascular disease or diabetes mellitus/pre-diabetes mellitus), longer time from enrollment to death, more recent trial manuscript publication year, enrollment in North America (versus Western Europe), female sex, and older age were associated with greater likelihood of death of undetermined cause. CONCLUSIONS: In 9 cardiometabolic clinical trials with long-term follow-up, approximately 16% of deaths had undetermined causes. This provides a baseline for quality assessment of clinical trials and informs operational efforts to potentially reduce the frequency of undetermined deaths in future clinical research.

Technology-Enabled Clinical Trials: Transforming Medical Evidence Generation.

The complexity and costs associated with traditional randomized, controlled trials have increased exponentially over time, and now threaten to stifle the development of new drugs and devices. Nevertheless, the growing use of electronic health records, mobile applications, and wearable devices offers significant promise for transforming clinical trials, making them more pragmatic and efficient. However, many challenges must be overcome before these innovations can be implemented routinely in randomized, controlled trial operations. In October of 2018, a diverse stakeholder group convened in Washington, DC, to examine how electronic health record, mobile, and wearable technologies could be applied to clinical trials. The group specifically examined how these technologies might streamline the execution of clinical trial components, delineated innovative trial designs facilitated by technological developments, identified barriers to implementation, and determined the optimal frameworks needed for regulatory oversight. The group concluded that the application of novel technologies to clinical trials provided enormous potential, yet these changes needed to be iterative and facilitated by continuous learning and pilot studies.

Levels of evidence supporting drug, device, and other recommendations in the American Heart Association/American College of Cardiology guidelines.

Clinical guideline documents reflect the evidence supporting clinical practice, but few recommendations in cardiovascular guidelines are supported by evidence from randomized controlled trials (RCTs), the highest level of evidence. Incentives for generating evidence from RCTs differ by topic of guideline recommendation, and it is uncertain whether evidence supporting guideline recommendations differs based on the topic of the recommendation. METHODS: We abstracted recommendation statements from current ACC/AHA guideline documents (2008-2019). Two reviewers independently characterized each statement into categories based on its primary topic: pharmaceutical, device, non-invasive or minimally invasive therapeutic procedure, surgery, diagnostic invasive procedure or non-invasive imaging, laboratory, care strategies, health services or policy, history/physical examination, lifestyle or counseling. We determined the number and proportion of recommendations in each category characterized as level of evidence (LOE) A (supported by multiple RCTs), B (supported by a single RCT or observational data), and C (supported by expert opinion or limited data). RESULTS: Of 2934 recommendations from 29 clinical guideline documents, 784 (26.7%) were primarily about pharmaceuticals, 503 (17.1%) diagnostic invasive procedure or non-invasive imaging, 366 (12.5%) devices, 349 (11.9%) care strategies, 274 (9.3%) surgery, 216 (7.4%) therapeutic procedures, 162 (5.5%) lifestyle interventions or counseling, 160 (5.5%) health services, care delivery, or policy, 83 (2.8%) laboratory, and 37 (1.3%) elements of the history and physical. Across all recommendations, 257 (8.8%) were characterized as LOE A, with considerable variability by topic. 25.9% of lifestyle/counseling recommendations, 16.9% of lab recommendations, and 14.7% of drug recommendations were classified as LOE A, but <8% of recommendations in all other categories, including 5.5% of device recommendations, 6.0% of therapeutic procedure recommendations, 2.6% of surgery recommendations, and 5.0% of health services or policy recommendations. CONCLUSION: Less than 10% of current ACC/AHA guideline recommendations are supported by high quality evidence from RCTs, with substantial variability by topic and multiple areas with very few recommendations supported by high-quality evidence. Development and implementation of inexpensive methods for generating a higher volume of RCT evidence to support clinical practice are needed, especially in areas where there are not strong incentives to conduct RCTs.