Pediatric Pharmacoepidemiology and Drug Development From a Regulatory Perspective

Safe and effective pediatric medical products developed in an efficient manner ensure the availability of medicines for children. Objectives for the field of pediatric pharmacoepidemiology include fostering collaborations among academia, industry, and government in pediatric medical product development, establishing generally accepted priorities within this field, and utilizing organized widely accessible pediatric registries—especially within subspecialties. The tools in pharmacoepidemiology include commonly used data resources in healthcare, such as Electronic Medical Records (EMRs) and claims data generated from sources such as Medicare and Medicaid. Data collected in the context of routine care are more abundant, accessible, inexpensive, and may be more generalizable than data collected through randomized controlled trials (RCTs). In addition, the advantages of conducting pediatric pharmacoepidemiologic studies, include the following: (1) the eligibility criteria for treatment in RCTs are almost always narrower than the criteria used for treatment in the general population, thus potentially limiting the generalizability of RCT results to broader populations; (2) adequately powered pharmacoepidemiologic studies can detect and evaluate rare and delayed events that are often not identified in shorter-term RCTs. Like adults, children should have access to products that have undergone rigorous evaluation.1 Adequate and well-controlled investigations of products intended for pediatric use must demonstrate substantial evidence of effectiveness and clinical benefit,2 and have sufficient safety data to support an adequate risk–benefit assessment for the intended population.1, 2 There are disadvantages to pediatric pharmacoepidemiology, as compared to RTCs. Pharmacoepidemiologic studies are not randomized. The observed effects could be related to residual confounding, selection bias, information bias, and other forms of bias. For example, it may be difficult to interpret differences between age groups in observational settings due to the lack of randomization, making any differences subject to confounding, and leading to difficulties in interpretability of the findings. Methods for limiting confounding and other sources of bias in observational settings, such as active-comparator/new-user designs, propensity scoring, target trial emulation, and outcome validation, have been described elsewhere.3, 4 In recent years, the terminology regarding commonly used data resources in healthcare used in pharmacoepidemiology has been formalized. As defined in the FDA guidance, real-world data (RWD) are data relating to patient health status and/or the delivery of health care routinely collected from a variety of sources.5 Real-world evidence (RWE) is the clinical evidence about the usage and potential benefits or risks of a medical product derived from analysis of RWD.5 Common sources of RWD in pediatric pharmacoepidemiologic studies include claims databases, electronic health records data, and registries. Multidatabase studies, such as those conducted through the FDA Sentinel Initiative (https://www.sentinelinitiative.org/), can enable the study of especially large pediatric populations, which can enhance both statistical power and generalizability of pharmacoepidemiologic studies. Pharmacoepidemiologic studies may also be used to evaluate specific medical product use in pediatric populations such as Drug Utilization Review studies (DURs).6 These studies have the usual challenges in pharmacoepidemiology of lack of randomization and also have pediatric interpretability challenges (see next paragraph). Generally, the principles of data collection and analysis of RWD/E are similar in adults and children. The content of the data may differ depending on the age group of the children, the subspecialty, and the setting (school vs. retirement community). Some outcomes are almost exclusively pediatric (e.g., rheumatic fever, Kawasaki disease). The prevalence, clinical presentation, and treatment outcomes in children, usually necessitate independent validation in pediatric populations. Growth or development is key in children, which may be important as ways to express exposure (e.g., weight-based dosing), outcomes of interest, or effect modifiers. Educational outcomes (e.g., achievement, attainment) and education-related issues (e.g., drug holidays during breaks from school) may be important variables in children. Other key variables (depending on the specific pediatric population and research question) include prenatal events, gestational age or weight at birth, mode of birth delivery, family history, and passive tobacco exposure. And additional designs, such as sibling-comparison designs are frequently used in pediatric pharmacoepidemiology. Structured, generalizable RWD approaches, such as registries, may be able to be harnessed for regulatory purposes.7 Continued efforts are needed to ensure data quality, integrity, and accessibility by specialists, healthcare practitioners, and the public. Quality registries provide equal outcome measures for some eligible children in the United States though for wider proportions of the population in some other countries. Fit-for-purpose, prospective registries should incorporate data that are specific to the questions that are intended to be answered just as they would be for an RCT. On the contrary, in non-prospectively designed pharmacoepidemiologic data sources, for example, claims databases, vital signs, disease measures, and patient-reported outcomes are generally not available unless there is a possibility to link claims data with patient medical records. In EMRs in the United States, not all claims databases are able to be linked to patient medical records. One registry that collects data in one subspecialty, pediatric rheumatology, is well organized and provides comparable data to RCTs for pediatric rheumatology.7 Specific data that would be clinically useful in pediatric rheumatology data sources include age and other demographics, vital signs (e.g., temperature), anthropomorphic measurements (e.g., height, weight), disease assessments (e.g., joint counts, validated measurements), laboratory values, radiographic information, and patient/caregiver-reported outcomes. Other registries that have the potential to span much pediatric morbidity and mortality in the United States and could potentially expand care substantially are widely accessible pediatric registries that might cover illnesses such as orphan diseases, hematology, nephrology, endocrine, cardiovascular, and other individual subspecialty registries have the potential to be “think tanks” for that subspecialty and could be a resource for primary care physicians. Pharmacoepidemiologic studies completement data collected in RCTs, particularly with respect to safety in pediatrics. A larger sample size can enable enhanced safety signal detection and study of rare events. Additionally, well-designed comparative effectiveness studies that move toward addressing confounding by indication or disease severity and other potential sources of bias (e.g., selection bias, immortal time bias) through new-user designs and target trial emulation have been described. Note that close emulation of observational studies has resulted in comparable findings to those in RCTs.8 Pharmacoepidemiologic studies have successfully been used in children to support clinical trial designs. For example, the use of well-designed, carefully collected, fit-for-purpose RWD has been used to create an external control arm that can be used as a comparator arm in single-arm registration trials for rare, life-threatening diseases.2 Natural history data were used as an external comparator for a single-arm study to approve Brineura (cerliponase alfa) for the treatment of Batten disease, a rare, life-threatening condition in children. However, the use of RWD in this context must be carefully considered and agreed upon with regulatory authorities because the control arm data are not concurrently collected which can lead to bias and incorrect conclusions about the effect of the treatment. FDA has issued guidance, “Considerations for the Design and Conduct of Externally Controlled Trials for Drug and Biological Products,” that provides more detailed information on the appropriate use of this type of study design.9 RWD can also be used to provide evidence to support a pediatric extrapolation approach. Pediatric extrapolation is defined as “an approach to providing evidence in support of effective and safe use of drugs in the pediatric population when it can be assumed that the course of the disease and the expected response to a medicinal product would be sufficiently similar in the pediatric [target] and reference (adult or other pediatric) population,”10 and can be used to support meeting substantial evidence standards when used appropriately. “In the development of the pediatric extrapolation concept, a review of data from RWD sources including but not limited to electronic health records, claims databases, and registries, can be considered”.10 Additionally, given the challenges in using RWD, it is important to evaluate whether the scientific question and the methodological considerations will be appropriate for the regulatory requirements (e.g., good clinical practice). The use of sound pharmacoepidemiologic principles and rigorous approaches in study design and conduct is of paramount importance for regulatory considerations of pediatric RWD/RWE. The use of pediatric pharmacoepidemiology studies, including the use of RWD to evaluate drug outcomes, is currently dynamic and developing novel approaches to addressing common data gaps. The selection of the RWD sources should be driven by the specific study questions at hand. Additionally, innovative strategies are essential for integrating clinical care data with clinical trial data, and for linking broader data sources, such as claims data with electronic health records or registry data. This integration will provide access to larger, more complete datasets with complementary information essential for conducting rigorous, well-powered pharmacoepidemiologic research. Given the challenges of conducting pediatric studies in small, more heterogeneous populations, these advances hold particular promise in pediatric drug development. Continued efforts are needed to ensure data quality, accessibility, and integrity. Pediatric pharmacoepidemiology studies, including the use of RWD/RWE, can be used to support but not replace controlled clinical trials for regulatory decision making in all phases of pediatric therapeutics development. The potential applications of pediatric pharmacoepidemiology have been expanding over recent decades with the availability of improved analytic approaches and computing technology. Rigorous and judicious use of fit-for-purpose pediatric pharmacoepidemiology has the potential to increase the speed and efficiency of pediatric therapeutics development with the goal of efficient availability of safe and effective medicines to all children who need them. No funding was received for this work. The authors declared no conflict of interest. Rutgers University receives, on Dr. Horton's behalf, research funding from the Childhood Arthritis and Rheumatology Research Alliance and the Arthritis Foundation.