Advancing pediatric medication safety using real-world data: Current problems and potential solutions

INTRODUCTION

There are substantial limitations in evaluating medications for pediatric use premarket; safety concerns are often identified after Food and Drug Administration (FDA) approval.^1,2 Children are often excluded from prelicensure trials, and fewer than half of medications contain information specific to use in children on the label.³ As a result, 45% of outpatient pediatric prescriptions are off-label, including 83% of neonatal prescriptions, and off-label outpatient use is increasing.⁴ Additionally, nearly 80% of hospitalized children received at least one drug off-label.⁵ Information on drug–drug interactions (DDIs) is often based on adult data, and the rate of adverse drug events (ADEs) related to DDIs in children is unknown.⁶ However, these off-label uses allow for harnessing of real-world data to study postmarket efficacy and safety in children. There is a critical need to better assess the real-world safety and effectiveness of these and other commonly used medications in children. We discuss challenges and potential solutions to using real-world data to improve pediatric drug safety.

CHALLENGES IN THE CURRENT APPROACH

Limitations of randomized controlled trials in detecting pediatric adverse drug events

Prelicensure clinical trials are usually powered to establish efficacy and safety under ideal conditions. These studies, however, provide little to no information on uncommon ADEs, potential drug–drug interactions, and long-term safety. Participants in clinical trials are skewed toward homogeneous, healthier, nonpediatric populations. As a result, important vulnerable populations such as infants and children, and those with comorbidities who require multiple medica- tions are often underrepresented.⁶ Additionally, most clinical trials are not adequately powered, nor are their endpoints of sufficient duration, to identify uncommon but serious medication safety events. New medications are often compared to placebo, and there is a general lack of comparative effectiveness data when medications are approved, which may not reflect real-world clinical practice.⁷ Finally, prelicensure clinical trials of new oral medications are typically performed in the outpatient population and applied to the hospital setting after approval. There are additional barriers to performing hospital-based clinical trials, including obtaining informed consent in the settings of acute and critical illnesses.⁸

Problems with postmarket methods to detect safety events in children

A “data gap” exists between when a drug is approved in adults and when clinical trials or postmarket studies become available in children. Thus, clinicians routinely prescribe medications off-label to children when clinically indicated. No systemic mechanism exists to capture these data and identify possible ADEs. The current postmarket detection of safety events relies heavily on spontaneous, voluntary reporting from providers and patients, such as those collected in the FDA Adverse Event Reporting System (FAERS). This passive system detects important safety signals that may be evaluated in future studies. Limitations of these voluntary reports include reporting and attention biases, incomplete reporting information, lack of population-based event rates, and inability to determine causality, especially in the hospital setting where patients have an acute illness and multiple medication exposures.

Active “pharmacovigilance” detection systems to identify safety signals leading exist. For example, active vaccine surveil- lance programs focusing on children include the Centers for Disease Control (CDC) Vaccine Safety Datalink and the Vaccine and Immunization Surveillance in Ontario.⁹ Drug-related systems, including the recently expanded FDA Sentinel program, include whole population surveillance with the ability to use EHR and claims data to evaluate specific pediatric concerns. The FDA Sentinel Initiative was established in 2008 following the 2007 FDA Amendments Act. The FDA Sentinel program utilizes information from various data partners, including academic institutions, administrative database companies, as well as health insurance and pharmaceutical companies, to perform postmarket drug and medical product safety surveillance and evaluations. However, these surveillance and evaluation systems in pediatric populations are limited by a lack of patient-level data or maternal–child dyad identification, which leads to high-cost safety evaluations using multiple data partners. These systems typically do not capture medication exposures in the hospital setting, making monitoring hospital-based adverse drug events difficult.

The hospital information void

Most clinical trials and FDA safety reports occur in the outpatient setting. Randomized clinical trials are less common in hospitalized, and critically ill patients, especially children, and interpreting safety events in these settings are challenging. Hospitalized children are often simultaneously exposed to multiple medications, have underlying conditions, and often undergo surgical procedures, making the evaluation of safety events using observational data difficult. Furthermore, many claims databases lack detailed inpatient pharmacy records, thus limiting their use for large-scale evaluation of hospital interventions.

PROPOSED SOLUTIONS

A critical need exists to improve and complement current systems to better evaluate the safety and efficacy of new and existing medications in children. The availability of patient reporting tools integrated electronic health records, and advanced pharmacoepidemiologic methods now provide potential solutions to overcome the limitations of the current safety monitoring system (Table 1).

TABLE 1.

Potential solutions to improve drug safety and effectiveness evaluations in the pediatric population.

Proposed solutions	Details	Example
Active collection of postmarket adverse events	Active patient-reported safety events	V-safe after vaccination health checker
	Active pharmacovigilance using real-world data	FDA sentinel program
Operationalize and better leverage	Link electronic health records, registries, and claims	Tennessee Medicaid Mother-Child Linked Cohort
existing data	databases	Clinical Practice Research Link (UK) Pediatric Health Information System-The Cystic Fibrosis Foundation Patient Registry Linkage
	Develop national database and registries using universal data-sharing agreements	Korean National Health Insurance Service (KNIS) database
		Finnish, Swedish, and Danish National Registries
	Enhance public-private partnerships across	FDA Sentinel Initiative
	insurance and health systems	Israel-Pfizer Real-World Epidemiological Evidence Collaboration Agreement
	Promote common data models and interinstitutional partnerships	FDA Sentinel Initiative Common Data Model
		The Observational Health Data Science and Informatics (OHDSI) Program
Incorporate children into clinical trials using novel methods	Evaluation of pediatric medications in the	Pragmatic Clinical Trials
	real-world setting
	Leveraging heterogeneity of pediatric care	High-Efficiency RandOmIzed Controlled (HEROIC) trials
	Safety-centered outcome trials	AHRQ Trigger Tools

1. Pro-active, patient-engaged collection and detection of postmarket ADEs

Current systems rely heavily on passive, voluntary reporting of postmarket safety events, which are subject to reporting bias, missing information, duplicate reports, under-reporting, and variable reporting over time. Transitioning from passive to active systems using patient and provider-targeted collection would improve reporting quality. Systematic collection of information on medication utilization and reporting of “no safety events” would provide a useful denominator for calculating event rates. A successful example of this is the V-safe program, wherein the CDC provided personal web and phone-based “check-ins” following COVID-19 vaccination.¹⁰ In the hospital setting, providers could play a greater role than patients in reporting certain events depending on the clinical context. A similar ADE feature could be added to medication-related phone applications, such as the World Health Organization Medsafe App.¹¹ Use of the programs could be either voluntary (opt-in vs. opt-out) or required for specific populations of interest, including in the hospital setting where patient-reported outcomes are less frequently captured.

Active, pragmatic, and targeted collection of ADEs in the real-world setting from patients and providers in the early postmarket years, especially for medications approved based on surrogate endpoints, those with short follow-up durations, and those used off-label in children, would improve detection of safety events and better evaluate medication effectiveness relative to market alternatives. A central body, such as the FDA or CDC, could help prioritize medications of concern or those with limited pediatric data. The choice of medication, adverse event, and population to be evaluated will likely be informed by safety concerns identified in adult clinical trials, drug mechanisms of action, as well as safety signals from drug adverse event surveillance systems. Timely event reporting to either a regulatory body or neutral third party (e.g., Harvard Pilgrim Institute within the FDA Sentinel Initiative) would be needed for data evaluation and signal detection.

2. Improve and better leverage real-word data sources

Existing real-world data sources include electronic health records, administrative and claims databases, and patient registries. However, these data are limited by a lack of patient-level data, and the inability to follow patients across institutions, settings of care, payors, and time. Linking administrative and claims databases to multi-institutional electronic health records and patient registry data would provide the necessary granularity to perform robust drug safety research. Additionally, linking maternal records to pediatric records would enable the study of perinatal ADEs and potentially long-term child outcomes related to pregnancy.¹² The 21st Century Cures Act and recent improvements in EHR interoperability allow for multi-institutional data collection. Given that hospital medication information is available in the EHR but not administrative databases, these linkages will be particularly useful in facilitating hospital-based safety studies.

Furthermore, leveraging common data models to standardize methods for abstracting data elements, can significantly increase the quality, diversity, and generalizability of results. The Observational Health Data Science and Informatics (OHDSI) Program and FDA Sentinel Initiative, recently enhanced from Mini-Sentinel, are shining examples of how the use of common data models, using claims data, can enhance drug utilization and safety research.^13,14 For example, an early success of the Sentinel Initiative was identifying the rare but serious association between the RV5 pentavalent rotavirus vaccine (RotaTeq) and intussusception in young children.¹⁵ Established national registries of health information, allowing for long-term longitudinal follow-up over time and across healthcare systems, has greatly benefited medical research in European and Asian countries, such as Pacurariu et al.¹⁶ These national registries have helped solve the “denominator” problem found in passive systems such as FAERS and smaller disease-specific registries. Ideally, public and private partnerships across industries, including payors, health systems, and governments, could synergize to greatly enhance the data collection infrastructure, speed of data acquisition, and quality of available data allowing for robust, population health research.¹⁷

3. Inclusion of children in postmarket clinical trials using novel methods

Children must routinely be included in clinical trials of both new and previously authorized medications. The application of pragmatic clinical trials and other study designs, using real-time data for research in the context of routine clinical care, is one path to include children in clinical studies. These “pragmatic” study designs are enticing as they can evaluate medication exposure in routine clinical settings to determine comparative effectiveness, safety, and costs.¹⁸ Similarly, clinical decision support and electronic health record tools can be used for screening, enrollment, and automated data collection in pragmatic trials. For example, the EHR may be leveraged to screen for inclusion/exclusion criteria or passively capture outcome events, such as the use of AHRQ triggers tools to capture safety-centered outcomes. High-Efficiency RandOmIzed Controlled (HEROIC) trials, in which a small number of participants are enrolled from many separate sites, is another novel approach that takes advantage of rapid enrollment and care heterogeneity to answer pragmatic questions.¹⁹ The HEROIC design may be particularly used in performing hospital-based pediatric clinical trials as serious adverse drug events in hospitalized children occur less often than in adults. These new methods are typically less costly to perform than traditional clinical trials, allowing for postmarket medication safety evaluation. Changing current approaches may require regulatory bodies, such as the FDA, to require pediatric data upon approval or mandate the completion of expedited postmarketing studies in children to ensure that robust safety and efficacy data become available in a timely manner. Prioritizing training for investigators in these new methods, as well as supporting funding for pediatric clinical trials, are also key to increasing inclusion of children in these studies.

CONCLUSION

There is a critical need to enhance the collection, evaluation, and dissemination of real-world postmarket information to improve the safety of medications prescribed to children. Only through large-scale, systematic investment in the management and analysis of clinical data can we develop the necessary infrastructure to apply new pharmacoepidemiologic findings to prevent harm and improve outcomes at the bedside.