Sie können Operatoren mit Ihrer Suchanfrage kombinieren, um diese noch präziser einzugrenzen. Klicken Sie auf den Suchoperator, um eine Erklärung seiner Funktionsweise anzuzeigen.
Findet Dokumente, in denen beide Begriffe in beliebiger Reihenfolge innerhalb von maximal n Worten zueinander stehen. Empfehlung: Wählen Sie zwischen 15 und 30 als maximale Wortanzahl (z.B. NEAR(hybrid, antrieb, 20)).
Findet Dokumente, in denen der Begriff in Wortvarianten vorkommt, wobei diese VOR, HINTER oder VOR und HINTER dem Suchbegriff anschließen können (z.B., leichtbau*, *leichtbau, *leichtbau*).
In pediatric trial design, it is particularly essential to maximize data utilization and ensure robust designs while minimizing sample collection. A common criterion to justify pediatric pharmacokinetic study designs is based on parameter precision (PP) evaluation, recommended by the US Food and Drug Administration. Here, we propose an alternative approach to design evaluation based on accuracy for dose selection (ADS).
Methods
This work was conducted using a simulation-and-reestimation framework, based on a real-case scenario of designing the single-dose pharmacokinetic study of the anti-tuberculosis (TB) drug pretomanid, with the aim of selecting doses for the next multidose long-term study. The study powers were computed using the ADS approach under scenarios with (1) real-case conditions, (2) high variabilities, (3) available options of tablet doses for selections. The study power using a PP approach was computed to compare with the ADS approach.
Results
The ADS approach suggested that the design selected accurate doses with study power >80% in almost all dosing weight groups, whereas the PP approach found the design underpowered for clearance. The ADS-based power was decreased to ~65% in the smallest weight groups given high variability. Varying the options of dose levels affected the ADS-based power non-monotonically, although fewer levels generally yielded higher power.
Conclusion
The ADS approach practically evaluates the precision in dose selection, providing a directly relevant decision criterion for designing pediatric pharmacokinetic studies and could be an alternative for power evaluation when the study is focused on determining doses using discrete tablet sizes.
A common criterion to justify pediatric pharmacokinetic study designs is based on the evaluation of pharmacokinetic parameter precisions.
We developed a novel alternative approach focusing on accuracy of dose selection (ADS), evaluating the design ability to accurately select doses that achieve target exposures per dosing group.
The developed ADS approach practically evaluated the precision in dose selection, providing a directly relevant decision criterion for designing pediatric pharmacokinetic studies, and could be an alternative for power evaluation in studies focused on determining doses using discrete tablets.
1 Background
Off-label use of medicines remains a significant issue in the pediatric population because of a lack of pediatric data, despite continuous advocacy for pediatric drug development by regulatory authorities [1, 2]. Adult data generally cannot fully account for the unique physiological and developmental differences in the pediatric population, underscoring the need for pediatric studies [3]. According to the rules of extrapolation from adults to children published by US Food and Drug Administration (FDA) in 1994, pediatric studies typically focus on characterizing pharmacokinetics, optimizing dosing, and investigating the safety profiles in children when disease progressions and exposure–response relationships are assumed to be similar between adults and children [4‐6]. According to the Pediatric Research Equity Act of 2003, information on drug safety and effectiveness in relevant pediatric weight and age ranges is a prerequisite for regulatory submission and approval of new drugs or formulations that could provide meaningful therapeutic benefit in children [7, 8].
Anzeige
The design of pediatric clinical trials is complicated by many factors. Before conducting a pediatric clinical trial, the risk–benefit balance must be assessed to ensure that the information to be gained will outweigh the potential risks. For instance, although single-dose studies rarely provide direct therapeutic benefits, their value lies in offering critical information for determining safe dosing regimens for future long-term use [9]. Resource limitation is another aspect to consider, including the limited capacity of specialized clinical sites, the significant challenge of recruiting pediatric patients, and lack of funding to support research due to limited market profitability [2]. Moreover, ethical practices, such as ensuring minimally invasive sampling methods and reducing the number of blood samples, are crucial, especially in neonates and infants, where intensive sampling is particularly challenging. Leveraging optimal designs and advanced modeling techniques allows for robust study design with minimized sample collections and maximizing data utilization [1, 10].
It can be a challenge to justify the sample size in a pharmacokinetic study in general, as these studies typically lack a clear hypothesis to evaluate [10]. Parameter precision represented by confidence interval (CI) width has been suggested as a target to justify the sample size of pharmacokinetic studies [11]. To justify the design of pediatric pharmacokinetic studies, a common criterion recommended by FDA guidance is that the power to achieve 95% CIs contained within 60–140% of the geometric mean estimates of main pharmacokinetic parameters of interest in each (weight/age) sub-group should be at least 80% [12, 13]. This approach focuses on parameter precision (PP). However, the PP criterion does not directly investigate a clinically relevant target such as dose selection. Here, we propose an alternative novel approach of study or sample size evaluation that is based on accurate dose selection (ADS). This approach evaluates the power of the given study design and sample size to identify accurate doses for each subgroup. By accurate doses, we mean doses that would generate desirable exposure levels.
The first application of the ADS approach was in the design of a pediatric pharmacokinetic and safety study of pretomanid. Pretomanid is an anti-tuberculosis (TB) drug that was approved for the treatment of drug-resistant TB in adults in 2019, in combination with bedaquiline and linezolid [14]. To date, no pediatric data for pretomanid are available. A pediatric investigation plan for pretomanid approved by the European Medicines Agency is actively under way to investigate the use of pretomanid in children for the same indication as in adults [15]. Pretomanid is now available as a 200 mg solid-tablet formulation, a scored dispersible tablet formulation of pretomanid with different dose strengths (10 and 50 mg) was developed in light of the pediatric investigation plan to facilitate administration in younger children [16]. Benefiting from the new formulation, the first-in-pediatric single-dose pharmacokinetic study of pretomanid (IMPAACT 2034, ClinicalTrials.gov ID: NCT05586230) is currently ongoing with the objective to select doses for a subsequent long-term study.
In this work we aimed to (1) illustrate the ADS-based evaluation approach through the real-case scenario of designing the single-dose pharmacokinetic study of pretomanid by simulations and re-estimations and (2) compare the ADS-based approach and the standard PP-based evaluation approach under various circumstances.
Anzeige
2 Methods
The schema of the procedure of this simulation study is illustrated in Fig. 1 and described in detail here, including the setup for patient population and study design, followed by the evaluation process using the two approaches. The given study design was evaluated using both ADS- and PP-based approaches for study power. All simulations and estimations were performed using the software NONMEM 7.4.4 and R version 4.3.3 [17, 18].
Fig. 1
The flowchart of the simulation study for the study design evaluation using both accurate dose selection (ADS)-based and parameter precision (PP)-based approaches. CI confidence interval, CL clearance, PK pharmacokinetic, SIR sampling importance resample, SSE stochastic simulation and estimation, V distribution
To populate the simulations, a pool of virtual pediatric patients with TB was generated, with age uniformly distributed between 0 and 18 years. Sex was assigned with a 50/50 probability. Given age and sex, individual body weights were calculated through a TB-adjusted least mean squares method based on growth curves from the World Health Organization and the National Health and Nutrition Examination Survey [19, 20]. The distribution of the simulated population is visualized in Fig. S1 in the electronic supplementary material (ESM), with further details of the simulation. From this population, 30,000 children with weights above 4 kg were selected. Patients “enrolled” in the virtual single-dose studies that comprise the simulation investigations here were sampled from this simulated population pool.
2.2 Study Design
The trial design to be evaluated was a phase I study of a single dose of pretomanid added to an optimized background regimen in children with rifampicin-resistant TB. The primary purpose of the trial was to evaluate the pharmacokinetics of pretomanid in children with rifampicin-resistant TB to identify weight-banded doses to be evaluated in a multidose study covering the intended treatment duration. These doses were expected to generate exposures matching those reported in adults.
A sample size of 36 was proposed, with nine patients per study weight cohort of 4–12, 12–20, 20–31, and ≥ 31 kg. Children would be given weight-banded doses according to six pre-defined weight groups: 4–6, 6–8, 8–12, 12–20, 20–40, and ≥ 40 kg. A pharmacokinetic sampling schedule at 1, 3, 6, 9, 24, and 48 h post-dose was pre-determined based on practical considerations and preliminary model-based simulations. The dose for each dosing weight group was selected to achieve a target exposure, as described separately below in Sect. 2.4.
The choices of sample sizes for the study design were ultimately justified by the ADS approach described more fully below.
2.3 Scaled Adult Pharmacokinetic Model
2.3.1 Adult Pharmacokinetic Model
The simulation study was performed using a population pharmacokinetic model for pretomanid that was developed based on data from 133 adult participants combined from a 26-week phase III study (Nix-TB) and a phase I relative-bioavailability study (CL-011) [16, 21]. The model described variabilities across the two formulations (dispersible tablet formulation and solid-tablet formulation) at doses of 10, 50, and 200 mg and two populations (healthy volunteers and patients with highly resistant TB). The model included one compartment for disposition, first-order elimination, and first-order absorption with a transit delay compartment. The estimated typical half-life was 15 h in adults, and time to maximum concentration (tmax) was around 6 h after dose. Allometric scaling with weight was applied to clearance (CL) and volume of distribution (V), with coefficients of 0.75 and 1, respectively. Here, CL and V should be interpreted as apparent oral values at a dose of 200 mg. Relative bioavailability varied with dose as F = (dose/200)0.081, translating to, for example, 11% and 22% lower bioavailability at 50 mg and 10 mg compared with the 200 mg dose of pretomanid. Additional parameters included interindividual or interoccasion variability (IIV or IOV) of the typical pharmacokinetic parameters and residual unexplained variability in measured concentrations. See Table S1 in the ESM for the model’s parameters.
2.3.2 Scaling to Children
The scaling was based on the model’s allometric scaling factors, and a function added to the model, describing the maturation of metabolic enzymes that was assumed to affect CL in the youngest children. Pretomanid is eliminated by metabolism through a complex set of pathways, for which only a 20% contribution of cytochrome P450 (CYP)−3A4 has been explicitly quantified [22]. Therefore, 20% of the maturation was attributed to a reported function for CYP3A4, and the remaining 80% was assumed to follow a generic moderate rate of maturation, described by the function for paracetamol [23, 24]. These maturation components together generated a fraction between 0 and 1, with 1 representing 100% of adult metabolic activity level. As a function of age (Table S1 in the ESM), this fraction reaches 50% at ~3 months and 90% at ~16 months after birth.
This scaled adult pharmacokinetic model was used as the basis for all the following work. For the simulation exercise described here, the point estimates of the fixed effect parameters in this model are assumed to be the “true” parameter values (θ*).
2.4 Optimal Group Doses
Based on the above model, optimal group doses (\({GD}_{m}^{*}\)) were selected for children in the six different dosing weight groups of the virtual population, m = 1, …, 6. The objective of dose selection was to yield area under the concentration–time curves from time 0 to infinity (AUC0–∞) after a single dose close to 50.9 μg/mL·h, the geometric mean of steady-state AUC from time 0 to 24 h (AUC0–24,ss) observed in an intensively sampled subset of participants in the Nix-TB study [21]. Limited by the tablet formulations, pretomanid was dosed only in multiples of 5 mg (half of the 10 mg scored dispersible tablet) up to 200 mg (approved dose for adults). The \({GD}_{m}^{*}\), m = 1, …, 6, was selected as the amount (DOSEm) that minimized the root mean squared error (RMSEm) defined by:
where \({N}_{gr{p}_{m}}\) denotes number of virtual patients out of the 30,000 who fall in the mth dosing group; AUCi denotes the model-predicted AUC0–∞ of the ith subject (AUC0–∞ after a single dose is equivalent to AUC0–24,ss), based on weight (WTi) and age (AGEi), given θ*; and AUCtarget is 50.9 μg·h/mL.
2.5 Re-estimation
2.5.1 Simulated Dataset
In a simulated trial, 36 virtual patients were randomly “recruited” into the defined study weight cohort from the virtual pediatric population of 30,000 patients. Each such virtual participant in the mth weight group was assigned the dose \({GD}_{m}^{*}\), and pretomanid concentrations were generated according to the study design’s sampling schedule using the complete “true” pharmacokinetic model, including covariate relationships, IIV/IOV, and residual unexplained variability.
2.5.2 Re-estimated Pharmacokinetic Model
Using the data from each simulated clinical trial, the parameters of the pharmacokinetic model (including the parameters of the non-CYP3A4 maturation component and the allometric scaling factors) were re-estimated. The re-estimated parameters were then submitted for evaluation by both the ADS approach for accuracy of dose selection and the PP approach for precision of parameter estimation.
2.6 ADS Approach
2.6.1 Estimated Group Doses
With the re-estimated parameters from each simulated clinical trial, a set of optimal doses for the mth dosing group, \({\widehat{GD}}_{m}\) (m = 1, …, 6), was estimated using the same methodology as described above for the determination of \({GD}_{m}^{*}\), replacing \({{\varvec{\theta}}}^{\boldsymbol{*}}\) for each of the 30,000 virtual patients with \({\widehat{{\varvec{\theta}}}}^{\boldsymbol{*}}\) determined from the re-estimated parameters. As for \({GD}_{m}^{*}\), doses were tested only among multiples of 5 mg from 5 to 200 mg.
2.6.2 Ratios of Estimated to Optimal Group Doses
\(T\) He ratio of \({\widehat{GD}}_{m}\) to \({GD}_{m}^{*}\) for each dosing group was calculated and denoted as DRm. The dose selection for each dosing group was judged as having sufficient accuracy if DRm was within the limit of 60–140%.
2.7 PP Approach
The PP-based approach was performed following the methodology presented by Wang et al. [13]. The same simulated datasets were used as in the ADS-based approach, that is, those generated as described in Sect. 2.5.
2.7.1 Uncertainty Methods
To obtain 95% CIs of parameters of interest, parameter uncertainties around the re-estimated parameters were computed by using the “sandwich” R−1·S·R−1 matrix in $COV of NONMEM, assuming a multivariate normal distribution. Other methods to obtain parameter uncertainty, including different options in $COV of NONMEM, sampling importance resample, and stochastic simulation and estimation, were also investigated [25].
2.7.2 Parameter Precision of CL and V
The 95% CI (for the parametric uncertainty distribution) or 2.5th/97.5th quantiles (for the non-parametric empirical uncertainty distribution) of CL and V, given the simulated trial, were summarized. The study design was judged as able to provide sufficiently precise parameters if the 95% CI for both CL and V was within the limit of 60–140%.
2.8 Study Power Calculation
In total, 500 clinical trials were simulated using different random seeds, following the same process described in Sect. 2.5. For the ADS approach, the study power was summarized for each dosing group as the proportion of simulated trials concluding sufficient accuracy, that is, that the ratio DRm of the estimated group dose \({\widehat{GD}}_{m}\) to the optimal group dose \({GD}_{m}^{*}\) was between 60 and 140%. For the PP-based approach, the study power was summarized as the proportion of simulated trials providing sufficiently precise parameters, that is, that the 95% CIs for CL and V were contained within 60–140% of their point estimates. The study design was judged as adequate with sufficient sample size if the calculated study power was at least 80%.
Anzeige
2.9 Sensitivity Analyses
To study a scenario with higher variance, the IIV of CL was doubled from 33 to 66% coefficient of variation, and IOV of relative bioavailability was doubled from 8 to 16% coefficient of variation. For the ADS approach, the sensitivity to dose granularity was investigated by varying the minimal available tablet sizes from 0.1 to 25 mg, including the current 5 mg increment. For example, a 25 mg minimal tablet size would contribute to dose selection among multiples of 25 mg, up to 200 mg.
3 Results
3.1 Base Scenario
The power given the planned study design and evaluation approaches in different scenarios is shown in Fig. 2. The distribution of the enrolled patients across 500 simulated trials is summarized per dosing weight band and per age subgroup as Fig. S2 in the ESM.
Fig. 2
Study powers given two approaches under scenarios with high variabilities. Error bars represent 95% prediction intervals of the calculated power based on a 500-simulation size. ADS accurate dose selection, CL clearance, CV coefficient of variation, F bioavailability, PP parameter precision, V distribution
In the base scenario, ADS assessed that the study design was sufficiently powered (above 80%) to select doses accurately (close to optimal group doses) based on study information in weight bands > 6 kg. The ADS approach indicated that the power for selecting accurate doses was less than 80% in the dosing group < 6 kg. This is due to the greater impact of a 5 mg difference (one “step”) at lower doses. For example, the ratio of \(\widehat{GD}\) of 150 mg to \(G{D}^{*}\) of 155 mg would be 0.97, whereas the ratio of \(\widehat{GD}\) of 10 mg to \(G{D}^{*}\) of 15 mg would be 0.67, even though the difference was only 5 mg. Additionally, according to the study design, the smallest study weight cohort 4–12 kg was separated into three dosing weight bands (4–6, 6–8, 8–12 kg). Consequently, as presented in Fig. S2 in the ESM, 70% of the simulated trials included only one or no patient weighing < 6 kg, making it challenging to gather sufficient data to achieve study power per dosing weight group. The power for the highest weight group (≥ 40 kg) was not compromised by being at the edge of the study weight range with limited data availability, as the dosing options for this group were also restricted to a maximum of 200 mg (the approved dose for adults).
3.1.2 PP
With uncertainty computed using the NONMEM $COV sandwich matrix, the parameter precision was powered below 80% for CL with the planned study design across weight bands, as multiple parameters, including those determining the maturation function and the allometric scaling factor, were poorly estimated with large uncertainty using such limited data. The PP-based power for the lightest group was as low as below 30%, implying that much larger sample sizes were needed for this single-dose study across weight bands to provide the required precision in clearance. A power of 80% in obtaining precise V was achieved, as fewer parameters were involved for V in the estimation. Using different uncertainty estimation methods influenced the PP-based power heavily, as shown in Fig. S3 in the ESM. The sampling importance resample method yielded the lowest power nearly uniformly. The superiority of the sandwich-matrix or stochastic simulation and estimation method varied with the parameter (CL or V) and body weights.
3.2 Sensitivity to Variabilities
In the scenario with high variability of either CL or bioavailability, PP-based study power was generally reduced by 10–30% across dosing weight bands <40 kg and by 40% in those ≥ 40 kg. In contrast, the ADS-based power was only slightly decreased by about 10% in groups <12 kg. The power for precision in V was slightly affected by the increasing variability of CL but decreased more pronouncedly with the increasing variability of bioavailability. This was potentially because the disposition parameters (CL and V) were estimated as apparent values that were relative to bioavailability, thus both affected by the variability in bioavailability.
3.3 Sensitivity to Tablet Size
As the ADS-based approach evaluated deviations between selected and optimal doses, we investigated its sensitivity to the availability of dose-strength options for selection, that is, minimal tablet size, as shown in Fig. 3. In general, higher power was achieved when the options of dose levels were few (e.g., minimal tablet size of 25 mg would provide eight options within the dosing range of 25–200 mg) and vice versa (e.g., minimal tablet size of 1 mg would provide 200 options up to 200 mg). Decreasing the minimal tablet size to fine steps of 0.1 mg did not significantly worsen the power. However, the relationship between power and minimal tablet size was not monotonic, as observed by comparing scenarios with 10 mg and 5 mg (Fig. 3). For example, when aiming at a \(G{D}^{*}\) of 20 mg, \(\widehat{GD}\) with 5 mg minimal size having one-step difference (i.e., 15 or 25 mg) would result in a ratio of 75 or 125%, whereas with 10 mg minimal size, the ratio for one-step difference as 10 or 30 mg would be 50 or 150%, beyond the limit of 60–140%. These results highlight the advantage of the ADS-based approach in considering tablet size and practically feasible dosing options when evaluating study power.
Fig. 3
Study powers given the accurate dose selection (ADS)-based approach under scenarios with different minimum tablet sizes
In this work, we proposed an approach to evaluate pediatric pharmacokinetic study designs based on the ADS, offering insights from a pragmatic application perspective, treating pharmacokinetic exposure and dosing as direct targets. Such an evaluation approach is useful since it directly relates the study power to the outcome of interest. In addition to focusing on a directly clinically relevant endpoint, this approach allows for the incorporation of various factors influencing dose selection, such as available dosage options limited by formulation. We compared the ADS approach to the standard approach in terms of the study power based on the precision of parameter estimates, PP. Our comparison of the two approaches used an actual case study from pretomanid as a test scenario. The PP approach indicated that the study is underpowered for CL precision. The ADS-based approach offered a different perspective, demonstrating sufficient power to select doses accurately with the given study design of 36 participants in a single-dose study, regardless of pharmacokinetic variabilities. Our workflow for the ADS approach adhered to the common practice in dosing simulation and selection, that is, considering the available pharmacokinetic model as a true basis for selecting doses. We believe this approach reflects the general consideration in pediatric dosing selection and indeed much of model-based drug development and even optimal design theory more broadly. The implicit assumption (which we here render explicit) of using a “true” model in any simulation study is that the study design will work well if the true state of nature is not too different from what is being assumed. In situations where the model structures or covariate relations are uncertain, additional simulation scenarios or sensitivity analyses should be considered to build more confidence, independent of whether the ADS or PP evaluation approach is used. Currently, the actual single-dose pediatric study of pretomanid (IMPAACT 2034; NCT05586230) is actively enrolling, with the study protocol based on the results from the ADS-based evaluation. This study received regulatory oversight from the Division of AIDS (DAIDS) of the National Institute of Allergy and Infectious Diseases. The protocol was reviewed by both DAIDS and the US FDA, as well as institutional review boards and ethics committees, and officially approved by DAIDS.
The minimal available tablet size affected the study’s power to select optimal doses. Tablets are commonly used as the dosage form across various therapeutic areas. The ADS approach successfully incorporates tablet size into power calculations, enabling more informative and relevant evaluation. Our results suggest that, when designing a study to confirm or optimize doses, having fewer dose level options would result in a smaller sample size. However, it should be noted that the ADS approach selects a dose for a given group to minimize the deviation (RMSE) of the predicted exposures within that group from the target exposure given the available doses, but the approach is agnostic as to whether that minimized deviation is small enough for comfort. Having more doses available might yield a smaller minimized variability. Moreover, a non-monotonic pattern of tablet size versus study power was observed in this work, so it is important to perform the analysis to investigate the impact of tablet sizes on the study power, if selection of tablet sizes is an option.
It may be argued that our comparison of ADS and PP is unfair in that ADS is based on point estimates of doses and PP is based on CIs of parameters, and CIs reflect more uncertainty than point estimates. Against that charge, two defenses may be brought. First, the PP approach is the standard approach for determining sample sizes in pediatric trials, so any new approach should be compared to it. Second, the ADS approach is appropriate in circumstances such as its use in the pretomanid single-dose pharmacokinetic trial, where a trade-off of increased efficiency and relevance at the expense of decreased certainty is acceptable because follow-up is planned, namely, a multiple-dose study where more data will be collected. Because with AUC0–∞ as the target exposure there is an equivalence between dose and clearance (AUC0–∞ = dose/(CL/F)), the ADS approach is about picking the right dose with a stipulated risk of error, whereas the PP approach is about estimating the dose with a stipulated level of precision.
What would happen with the ADS approach if, instead of point estimates of dose, CIs of dose were used? This was investigated as described in Resource 1 in the ESM. It can be seen there that for the design of the pretomanid single-dose study, such an ADS-with-uncertainty approach has power similar to, and even somewhat less than, that of the PP approach. However, the ADS-with-uncertainty approach introduces the criterion of precision, like the PP approach. When accuracy is a satisfactory criterion, as in the pretomanid example, ADS could be sufficient and more efficient.
Anzeige
To comply with the published PP approach, both maturation function parameters and allometric scaling factors were estimated in the simulation work, which is an extreme setting given the current relatively small clinical trial sample size, especially the lack of information in the lightest dosing weight groups across simulated trials (Fig. S2 in the ESM) [13]. Consequently, the clearance was underpowered as informed by the PP approach. Although the presented trial enrolled participants per weight groups, a further study design criterion to ensure a large enough sample size in those aged < 2 years may improve the estimation of maturation function.
When using the PP-based approach, one concern was the determination of parameter uncertainty. This work demonstrated that study power conclusions varied greatly depending on the uncertainty method used. In addition to differences in assumptions and characteristics between these methods, the high variation in parameter re-estimation (due to limited data support for complex model parameterization) may also contribute to the heterogeneous results observed across uncertainty methods. Nevertheless, no official guidelines specify which uncertainty method should be used to best reflect the “true” uncertainty. The article by Wang et al. [13] introduced the PP approach and demonstrated the delta method for estimating joint uncertainty distribution; however, this approach relies on linear approximations, which may not be valid for models with nonlinearity or high variability. Broeker and Wicha [26] compared various uncertainty methods and showed various behaviors of methods for different sample sizes. These findings emphasize the importance of considering the selection of uncertainty methods in the PP approach. Although acquiring parameter uncertainty can sometimes be difficult under such a complex model in this work, the ADS approach again offers a pragmatic alternative by prioritizing simplicity and feasibility.
The proposed ADS approach as an alternative criterion for study design can be generalized to broader applications outside the current test scenario of pretomanid. The current work illustrated the ADS approach with AUC0-∞ as the exposure target for dosing selection, but the ADS approach is compatible with any model-derived pharmacokinetic metric. If the target metric depends on multiple parameters, for example, maximum concentration, the ADS approach offers the advantage of evaluating doses by jointly considering the contributions from multiple parameters to the exposures. In contrast, the PP approach, which typically assesses the precision of individual parameters, may not fully reflect the overall precision at the concentration level. When aiming to achieve several exposure targets, the ADS approach would require a clear weighting strategy across targets to define “optimal” dosing, much representing the practice in real-world dosing selection. In such cases, the PP approach would need to incorporate joint parameter precisions when exposures of interest for a drug are driven by several parameters, a situation that can be more complex in models with multiple compartments.
5 Conclusion
The ADS approach practically evaluates the accuracy of dose selection, providing a more directly relevant decision criterion for designing pediatric pharmacokinetic studies. It could be a good alternative for power evaluation of pediatric pharmacokinetic study design when the study is focused on determining doses using discrete tablet sizes, dosing options are limited by available dosage strengths, and where an approach that picks doses with a stipulated risk of error instead of estimating doses with a stipulated level of precision is acceptable.
Declarations
Funding
Open access funding provided by Uppsala University. This work was supported by TB Alliance (Global Alliance for TB Drug Development) with funding from Australia’s Department of Foreign Affairs and Trade, the Gates Foundation [OPP1129600], the Foreign, Commonwealth and Development Office (United Kingdom), Germany’s Federal Ministry of Education and Research through KfW, Irish Aid, and the United States Agency for International Development. This work was supported in part by the Gates Foundation [OPP1129600]. The conclusions and opinions expressed in this work are those of the author(s) alone and shall not be attributed to the Foundation. Under the grant conditions of the Foundation, a Creative Commons Attribution 4.0 License has already been assigned to the Author Accepted Manuscript version that might arise from this submission. The computations were enabled by resources in projects [NAISS 2023/22-1058] and [NAISS 2024/22-1458] provided by the National Academic Infrastructure for Supercomputing in Sweden (NAISS) at Uppsala Multidisciplinary Center for Advanced Computational Science (UPPMAX), funded by the Swedish Research Council through grant agreement no. 2022-06725.
Conflicts of Interest
Elin M Svensson is an editorial board member of Clinical Pharmacokinetics and so was not involved in the selection of peer reviewers for the manuscript or any of the subsequent editorial decisions. Yuanxi Zou, Jerry Nedelman, and Mats O Karlsson have no competing interests to declare that are relevant to the content of this article.
Ethics Approval
Not applicable.
Consent to Participate
Not applicable.
Consent for Publication
Not applicable.
Availability of Data and Material
Not applicable.
Code Availability
Code is provided upon request.
Authors’ Contributions
JN conceived the concept underlying this new approach. All authors contributed to the design of the study. YZ performed the modelling and simulation with input and support from EMS, MOK, and JN.
Open Access This article is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License, which permits any non-commercial use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by-nc/4.0/.
Bi Y, Liu J, Li L, et al. Role of model-informed drug development in pediatric drug development, regulatory evaluation, and labeling. J Clin Pharmacol. 2019;59(S1):S104–11. https://doi.org/10.1002/jcph.1478.CrossRefPubMed
Dunne J, Rodriguez WJ, Murphy MD, et al. Extrapolation of adult data and other data in pediatric drug-development programs. Pediatrics. 2011;128(5):e1242–9. https://doi.org/10.1542/PEDS.2010-3487.CrossRefPubMed
US Food and Drug Administration (FDA). Regulations requiring manufacturers to assess the safety and effectiveness of new drugs and biological products in pediatric patients. Final rule. Fed Regist. 1998;63(231):66631-66672. https://pubmed.ncbi.nlm.nih.gov/10338880/. Accessed 11 Jan 2025.
Ogungbenro K, Aarons L. How many subjects are necessary for population pharmacokinetic experiments? Confidence interval approach. Eur J Clin Pharmacol. 2008;64(7):705–13. https://doi.org/10.1007/s00228-008-0493-7.CrossRefPubMed
Wang Y, Jadhav PR, Lala M, Gobburu JV. Clarification on precision criteria to derive sample size when designing pediatric pharmacokinetic studies. J Clin Pharmacol. 2012;52(10):1601–6. https://doi.org/10.1177/0091270011422812.CrossRefPubMed
R Core Team. A language and environment for statistical computing. R Foundation for Statistical Computing. 2024. https://www.r-project.org/. Accessed 19 Feb 2025.
Johnson TN, Rostami-Hodjegan A, Tucker GT. Prediction of the clearance of eleven drugs and associated variability in neonates, infants and children. Clin Pharmacokinet. 2006;45(9):931–56. https://doi.org/10.2165/00003088-200645090-00005.CrossRefPubMed
