CDER scientists are developing a quantitative framework to describe the exposure-response relationship and the difference in exposure-response between populations to support extrapolation of efficacy in drug development programs from adults to pediatric populations.
The Scientific Challenge
Pediatric drug development often encounters challenges that don’t typify adult drug development. The pediatric patient populations available for conducting clinical studies, for example, tend to be smaller, and the ethical considerations surrounding pediatric trials may impose particular restrictions and concerns. As a result, investigators have in many instances opted to extrapolate evidence for drug approvals in adults and older pediatric patients to inform aspects of drug review relevant to younger pediatric patients. The use of extrapolation of efficacy into pediatric drug development programs has been found useful when disease progression and treatment response are similar in adult and pediatric populations.
Several methods have proven useful in extrapolating adult data into matters of pediatric drug assessments, one of which draws from analyses of exposure-response similarity between adults and children. In this method, a measure of exposure, such as the maximum concentration (Cmax), is plotted against a measure of treatment response for all the patients in both populations. The relationship between the exposure and response in each population can be presented as an exposure-response curve, and comparison of the two exposure-response curves may then provide a basis for determining the suitability of extrapolating data from adults into pediatric patients.
Historically, exposure-response curves have been inspected visually to assess comparability. Visual inspections, however, are somewhat subjective, potentially undermining the interpretation of clinical trial data and leading to irreproducible or unwarranted conclusions. The development of a quantitative framework to assess exposure-response similarity in pediatric and adult patients, based on accumulated experience in pediatric drug development programs, is thus an important goal. A quantitative framework can provide more objective criteria for assessments of exposure-response similarity, strengthening the scientific evidence available for determining whether drugs approved in adult populations can be assessed as safe and effective for pediatric use.
Non-inferiority Paradigm to Assess Exposure-Response Similarity Between Pediatric and Adult Patients
The main basis for the quantitative approach is based on a concept often used in clinical trials called non-inferiority. In drug development, there may be circumstances where it is appropriate to test a new drug against a control drug that is expected to provide a similar level of benefit. Statistically, however, showing a test drug has the same level of benefit as a comparator drug requires an extremely large number of patients. In order to make such comparisons clinically practical, we can instead arrange to assess whether the test drug is non-inferior to (i.e., no worse than) the control drug, and in such an approach we prespecify a “non-inferiority margin.” The non-inferiority paradigm is directly relevant to exposure-response comparisons between adults and children, as we are frequently concerned in determining whether we can expect drug exposure to elicit similar pharmacological effects regardless of age. In this context, the objective, ideally, is to evaluate the entire exposure-response curves for both populations, which differs from the standard non-inferiority approach where the overall estimated benefit is the objective of the comparison. The elaboration of the entire exposure-response curve, however, is arduous, so that a more limited number of exposure quantiles (10th, 50th and 90th percentiles) may instead be selected along with non-inferiority margins selected for those exposure estimates.
Based on the published literature and FDA reviews between 1998 and 2016, CDER investigators identified 10 drugs that were approved in both pediatric and adult patients and for which visual inspections similar exposure-response profiles for pediatric and adult patients. The investigators used these 10 drugs to develop a quantitative framework. See Figure 1 for details.
Figure 1. Workflow for quantitative assessment of comparing exposure-response in pediatric and adult patient. In usual circumstances, the objective of a non-inferiority comparison is to determine if a drug is effective by showing that the efficacy difference between the test drug and the known effective comparator is no more than a prespecified, clinically acceptable margin that is called the non-inferiority margin. The CDER investigators aimed to evaluate whether the entire E-R curve for pediatrics was noninferior to adults. To do so, the estimated efficacy difference between pediatric and adult patients was calculated at the 10th, 50th, and 90th percentiles of the adult exposure range. For each percentile, the point estimate and the confidence interval of the efficacy difference between the pediatric and adult patients were computed using PK/PD modeling and simulations. To determine non-inferiority, the point estimates and confidence intervals were compared with a pre-specified non-inferiority margin, i.e., a numeric value that had been calculated based on adult data or adopted from historical reports for each drug and indication. To assess the uncertainty of the point estimates, the probabilities of E-R non-inferiority at the 3 exposure points were calculated using bootstrap and Bayesian methods.
The comparison of the E-R relationships in pediatric and adult patients using the non-inferiority approach is shown in Figure 2. The investigators analyzed results for five antiepileptic drugs with linear exposure-response relationships (Figure 2A) and for three drugs (darunavir, esomeprazole, and infliximab) with nonlinear exposure-response relationships (Figure 2B). Note that because the drugs analyzed in Figure 2A were antiepileptics and the clinical endpoint was frequency of seizures, a lower value of efficacy in the pediatric population minus efficacy in adults (Eped-Eadult) indicates a more favorable pediatric response compared with adults). In most cases, the confidence intervals for the estimates of the difference in the drug’s efficacy of children and adults were much wider than the selected non-inferiority margins, indicating that a larger sample size would be required to make these assessments feasible using a typical non-inferiority approach. Figure 3 shows the results of the probability estimates of E-R non-inferiority between adult and pediatric populations at 10%, 50%, and 90% exposure percentiles using bootstrap (Figure 3A) and Bayesian (Figure 3B) methods. These two methods were used independently to estimate the probability that the observed data were consistent with a finding of non-inferiority between the adult and pediatric response.
Figure 2. E-R comparison between pediatric and adult patients for the linear and nonlinear E-R relationship. E-R comparison between pediatric and adult patients for drugs with linear and nonlinear E-R relationships. (A) the estimated differences for the antiepileptic drugs and (B) the estimated differences for darunavir, esomeprazole, and infliximab. The red, green, and blue colors indicate the 10th, 50th, and 90th percentiles of the entire adult exposure range. Each colored dot and its horizontal line represent the point estimate of efficacy difference between pediatric and adult patients and its 95% confidence interval, respectively. The black vertical bar on the plot was the predefined non-inferiority margin. Eped, efficacy of pediatric patients; Eadult, efficacy of adult patients.
Figure 3. Probability estimates using bootstrap and Bayesian methods. The probability of E-R non-inferiority between the adult and pediatric populations at 10%, 50%, and 90% exposure percentiles were calculated by bootstrap (A) and Bayesian (B) methods separately. The black, dark gray, and light gray colors indicate the 10th, 50th, and 90th percentiles of the entire adult exposure range.
For the bootstrap approach, the probability estimates, that is, the proportion of bootstrap samples for which the non-inferiority criteria were met ranged from 52.9% to 100% among the eight drugs. For the Bayesian approach, the probability estimates were similar. Using these methods, the E-R relationships can be considered sufficiently similar to support efficacy extrapolation if the estimated probability of achieving E-R non-inferiority between adult and pediatric population at the 10%, 50% and 90% exposure percentiles exceed prespecified thresholds. In typical non-inferiority trials, a threshold of 97.5% would be used. However, in pediatric development programs practical considerations may make this threshold infeasible and alternatives may need to be considered.
The Significance of the Non-inferiority Paradigm to Assess Exposure-Response Similarity and Dose Between Pediatric and Adult Patients
An international discussion is ongoing regarding the amount of additional data needed to support pediatric extrapolation. With an abundance of evidence to support similarity of the disease and response to treatment between the adult and pediatric patients for certain conditions, less additional data may be needed to provide evidence to support pediatric drug approval. Furthermore, ICH E11(R1) guidelines recommend the use of “quantitative methods” to generate evidence necessary to assess the potential similarities and differences between the source and target populations. This study introduced such a quantitative approach to assess the similarity of E-R between pediatric and adult populations.
How does this research advance pediatric drug development?
For many drugs that have been evaluated in adult patients, evidence as to their efficacy in children is lacking. Based on accumulated experience in pediatric drug development programs, CDER researchers have developed a framework to assess exposure-response similarity in pediatric and adult patients. This framework for applying a non-inferiority paradigm can provide quantitative evidence to help regulators make important decisions about the appropriate use of these drugs in children.
Qunshu Zhang, James Travis, Jian Wang, et al. Applying the Noninferiority Paradigm to Assess Exposure-Response Similarity and Dose Between Pediatric and Adult Patients. The Journal of Clinical pharmacology 2021, 61(S1) S165–S174. Published 2021. This article is a U.S. Government work and is in the public domain in the USA. DOI: 10.1002/jcph.1885