For all novel drugs (including new molecular entities and therapeutic biologics) approved by the FDA’s Center for Drug Evaluation and Research (CDER) from 2011 to 2023, we identified the pediatric postmarket studies required by the FDA at the time of original drug approval [18]. Approval letters include a specific section detailing pediatric studies required under PREA, and these were reviewed to identify and collect information on required pediatric studies, including study type, pediatric age groups to be studied, and due dates for submission of study reports to the FDA. While PREA study requirements can be issued also for other types of approvals (e.g., new indications, dosage forms, or routes of administration), we limited our analysis to novel drug approvals to focus on pediatric studies associated with products entering the market during a defined recent time period. Preclinical animal studies were excluded as were studies associated with drugs that were discontinued during the study period, since pediatric studies would no longer be expected for these drugs. All analyses were prespecified, although a written analysis plan is not available. This study is reported as per the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guideline (S1 Checklist).

Using the FDA’s Postmarketing Requirements and Commitments Database, information was collected on the status of studies through December 31, 2024 [19]. Status categories include pending, ongoing, delayed, submitted, fulfilled, and released. Submitted studies are those for which a study report has been provided to the FDA, while fulfilled studies are those for which the study report has been reviewed and approved by the FDA as meeting the terms of the requirement. Studies can be released if the FDA determines that the postmarket study would no longer be feasible or provide useful information. We considered studies to be ongoing if their status was listed as pending, ongoing, or delayed, and completed once they were submitted. For studies that had been submitted, we noted the date of study report submission and considered this as the study completion date. Planned study duration was defined as the time from drug approval to the original study report due date and observed study duration as time from approval to study report submission.

Since information on studies that are fulfilled or released are removed from the database after one year, we consulted archived versions of the database available on the FDA website to obtain information no longer available in the database. For studies with missing information on due dates (n = 24) across these sources, we obtained this information from the FDA through a Freedom of Information Act request.

Sponsors can request extensions to study due dates, which if approved by the FDA, result in a new study report due date. Studies can receive multiple extensions and revisions to the study report due date. Information on study extensions granted by the FDA was obtained from a publicly available FDA database that tracks pediatric study extensions for drugs approved by CDER [20]. Studies can also be replaced if the sponsor requests modifications, such as changes to participant age groups or study endpoints. Information on these study replacements, including revised study due dates, was obtained from the FDA Postmarketing Requirements and Commitments Database. In these cases, we noted the study replacement and considered the new study a continuation of the original study requirement. In a sensitivity analysis, study completion was also assessed excluding studies that were replaced to account for potential bias toward higher completion rates.

To supplement the pediatric study descriptions provided in FDA approval letters, we used study records in ClinicalTrials.gov to collect additional information on study primary endpoints. Using the study endpoints, studies were classified into mutually exclusive categories of primary efficacy, safety, or pharmacokinetic/pharmacodynamic (PK/PD) studies. Pediatric age groups were classified as neonates (0–<1 month), infants (1 month–< 2 years), early childhood (2–< 6 years), late childhood (6–< 12 years), and adolescent (12–< 18 years), based on conventional clinical definitions [21]. Therapeutic categories were assigned according to the Anatomic Therapeutic Classification System [22].

For all drugs with an associated pediatric study requirement, we determined whether pediatric labeling changes resulting from the PREA-required studies had been made as of December 31, 2024. Under PREA, sponsors are required to add information generated from the required studies to the drug label, irrespective of the study findings on drug benefit [20]. The FDA maintains a Pediatric Labeling Changes Spreadsheet, which contains information on all pediatric labeling changes made as a result of studies conducted under PREA [23]. We used this data source to identify relevant labeling changes and reviewed the drug labels to determine the type and time to first PREA-specific labeling additions, as well as the pediatric age groups receiving labeling information. A pediatric study was considered to have resulted in labeling changes if any result from the study was added. Since only two drugs in the sample had pediatric studies associated with more than one indication, all labeling additions were considered at the drug level.

We performed descriptive statistics to describe study characteristics, completion of studies, and addition of pediatric prescribing information to drug labels. In assessing study completion and labeling additions, we excluded studies that had not yet reached their expected completion date and limited to the subset of studies with due dates by December 31, 2024. We performed a multivariable linear regression analysis to assess planned study duration, modeling study duration in years as the dependent variable, and including predictors for study type, therapeutic area, and binary indicators for whether a given study included participants from each pediatric age group. Since the pediatric age group indicators are not mutually exclusive and a single study can include participants from multiple age groups, the coefficients for these variables represent the additional association of including each age group while adjusting for the others. We controlled for drug type (small molecule drug versus biologic) and study start year. To account for potential non-independence among studies of the same drug, standard errors were adjusted for clustering by drug. We also assessed the linearity assumption for study start year by comparing a linear specification to a spline-based model; the spline model did not improve fit, supporting the adequacy of treating approval year as a linear term in all the models.

Completion and on-time completion of pediatric studies were evaluated using multivariate logistic regression models, using the same set of predictor and control variables and adjusting for potential non-independence of studies of the same drug. Results are reported as odds ratios with 95% confidence intervals for each level relative to the reference group. In some cases, categories with very sparse data or complete separation (e.g., no events or all events within a group) led to non-convergence of the logistic regression model, and odds ratios could not be reliably estimated.

Among drugs with pediatric labeling additions, differences in mean time to labeling addition across labeling types were assessed using one-way ANOVA. This analysis was restricted to drugs with observed labeling additions, for which all event times were fully observed and uncensored. To examine predictors of time to labeling addition across all drugs (including those without an addition by the study cutoff), we used Cox proportional hazards models. Logistic regression was used to evaluate factors associated with the presence of any pediatric labeling addition. All models were adjusted for drug type and the year of approval. For the Cox proportional Hazards model, the proportional hazards assumption was verified using Schoenfeld residuals, which indicated non-proportionality for approval year. Accordingly, we modeled approval year as a time-varying covariate by including an interaction with log(time).

We used Kaplan–Meier analyses to estimate the cumulative incidences of pediatric study completion and of labeling additions over time. Since the FDA does not set expected dates for labeling additions, the latest expected study completion date among all studies associated with a drug was used as the comparison for the date of labeling addition. Differences between expected and observed study completion dates were evaluated using the Wilcoxon signed-rank test, indicating whether the median observed completion date differed from the expected date. This test was also used to assess differences between the latest expected study completion date associated with a drug and the date of labeling change. Analyses were performed using R Studio Version 2024.12.1 + 563.

The development of the research question was informed by concerns regarding evidence-based use of therapeutics in pediatric patients. Patients were not involved in the design or conduct of the study, nor were they advisers in this study.

There were 552 drugs approved by CDER over the study period, of which 179 (32.4%) were associated with pediatric study requirements under PREA at the time of approval. Thirteen of these drugs were eventually discontinued, yielding a study cohort of 166 drugs associated with a total of 338 pediatric study requirements (mean of 2 study requirements per drug).

About half the 338 PREA-required studies were primary efficacy studies (N = 175, 51.8%), with another 32.5% (n = 110) comprised of primary safety studies, and 15.7% (n = 53) primary PK/PD studies (Table 1). Inclusion in studies increased with increasing pediatric age, with 18.3% (n = 62) of studies enrolling neonates and 79.3% (n = 268) adolescents. Studies were concentrated across a few therapeutic areas, with almost half studying either nervous system agents (n = 92, 27.2%) or anti-infectives (n = 70, 20.7%), followed by 15.1% (n = 51) focused on alimentary tract drugs.

The mean planned duration for the pediatric studies was 5.6 years (95% CI, 5.3–5.9), with the longest duration seen for efficacy studies at 6.5 years (95% CI, 6.0–6.9), followed by safety studies at 5.3 years (95% CI, 4.8–5.8), and PK/PD studies at 3.2 years (95% CI, 2.7–3.8). There were no significant differences in mean planned duration across the different age groups, though we did observe differences based on the drug’s therapeutic area. Studies for anti-infectives, hematologic agents, and the miscellaneous “other” category showed shorter planned durations compared with nervous system drug studies (differences of −1.23 years [95% CI, −2.08 to −0.37], –1.42 years [95% CI, −2.68 to −0.16], and −1.10 years [95% CI, −1.99 to −0.22], respectively), while other therapeutic areas did not differ significantly.

Of the 338 pediatric studies, 222 had a due date by December 31, 2024, and were included in the analysis examining study completion. Median follow-up time for these studies was 9.1 years (IQR 6.3–11.1). Among the studies, 111 (50.0%) had obtained at least one extension to the study due date, 29 (13.1%) had been replaced by a new study with a later due date, and 126 (56.8%) had either an extension or a replacement resulting in an extended timeline. The mean change to due dates was 2.9 years (SD 2.0).

A total of 122 (55.0%) studies were completed, with 54 (24.3%) completed by the original due date (Table 2). Compared with primary efficacy studies, primary safety studies had higher odds of completion (adjusted OR, 2.45; 95% CI, 1.07–5.62) and primary PK/PD studies showed even greater odds of completion (adjusted OR, 6.49; 95% CI, 2.37–17.80). There was no evidence of differences in study completion across pediatric age groups. For therapeutic area, compared with anti-infectives, studies of nervous system drugs (adjusted OR, 0.34; 95% CI, 0.13–0.85) and immunomodulating agents (OR, 0.18; 95% CI, 0.04–0.79) had lower odds of completion, while other therapeutic areas showed no significant differences. On-time completion did not differ across study types, though with respect to pediatric age groups, those involving early childhood participants had higher odds of on-time completion (adjusted OR, 3.61; 95% CI, 1.49–8.73) compared with studies in neonates. Therapeutic area was also associated with timely completion, with studies for nervous system drugs showing lower odds of on-time completion (adjusted OR, 0.25; 95% CI, 0.08–0.79) compared with studies of anti-infectives. Results were similar in a sensitivity analysis excluding 49 studies that were replaced (S1 Table).

Fig 1 shows Kaplan–Meier curves for the cumulative incidence of study completion based on expected and observed completion dates. By 5 years after drug approval, 44.4% of pediatric studies were expected to be completed based on the original due date, compared with 24.8% observed to be completed. This difference increased over time and by 10 years, 92.0% of studies were expected to be completed, but only 59.5% were completed. Overall, studies were completed significantly later than expected (Wilcoxon signed-rank test, p = 0.01), indicating systematic delays relative to planned timelines.

The 222 pediatric studies due by December 31, 2024, corresponded to 117 drugs (Table 3). Just over half of these drugs (n = 64, 54.7%), had pediatric information added by December 31, 2024, with a mean of 5.6 years (SD 2.7) to the addition of labeling information. The most common type of labeling change was the addition of approval for pediatric use (n = 45/64, 70.3%), followed by expansion of the approved pediatric age range (n = 12/64, 18.7%), and addition of information describing lack of safety and effectiveness in pediatric patients (n = 7/64, 10.9%). There were significant differences in time to labeling additions based on type of labeling change (ANOVA, p < 0.001), with the shortest duration observed for expansion of approved pediatric age ranges at a mean of 3.8 years (SD 1.3), followed by new pediatric approval at 5.7 years (SD 2.6), and addition of information on lack of pediatric safety and efficacy, which took a mean of 8.3 years (SD 3.3).

The likelihood of a pediatric labeling addition did not differ by age group. In contrast, labeling additions varied substantially by therapeutic area. Compared with anti-infectives, drugs in the alimentary tract and metabolism (adjusted OR, 0.08; 95% CI, 0.01–0.42), nervous system (OR, 0.13; 95% CI, 0.02–0.70), and immunomodulating agent (OR, 0.03; 95% CI, 0.00–0.25) classes had lower odds of receiving a pediatric labeling addition. Similarly, in time-to-event analyses, the rate of pediatric labeling additions did not differ across age groups, though differences were seen based on therapeutic area. Drugs in the alimentary tract and metabolism (adjusted HR, 0.16; 95% CI, 0.07–0.40), nervous system (HR, 0.22; 95% CI, 0.09–0.55), and immunomodulating agent (HR, 0.07; 95% CI, 0.02–0.25) classes had slower rates of pediatric labeling additions compared to anti-infective drugs.

Kaplan–Meier curves are shown in Fig 2 for the cumulative incidence of expected pediatric study completion and observed labeling additions over time. At 5 years after drug approval, 30.7% of drugs were estimated to have had all pediatric studies completed, but only 14.9% of drugs had labeling changes. By 10 years, 90.4% of drugs were expected to have had all pediatric studies completed, but only 52.8% of drugs had pediatric information added to labels. Comparison of the latest planned study completion date with the date of labeling changes indicated that labeling updates occurred significantly later than expected (Wilcoxon signed-rank test, p = 0.04).