Development of in vitro/in vivo correlation for long-acting injectable suspensions via bootstrapping

Abstract

In vitro/in vivo correlation (IVIVC) is of great importance to expedite the development life-cycles for pharmaceutical industry and possibly to reduce the burden for regulatory assessment, especially for long-acting injectable (LAI) drug products. However, owing to the inherent complexity and relative novelty, no regulatory guidance is available for IVIVC development of LAIs. Moreover, LAI dosage forms are prone to exhibit much higher inter-patient variability compared to conventional formulation types. Currently, relevant publications and an industry survey indicate most of the researchers tend to use mean (both in vitro and in vivo) data to establish IVIVCs, which may discount the inherent variability of LAIs, and hence may result in stringent IVIVCs with poor predictive performance. Therefore, it is critical to use statistical tools such as bootstrapping to represent highly variable in vivo pharmacokinetic profiles and more accurately develop IVIVC for various LAIs. The objective of this work is to use an innovative bootstrapping strategy for IVIVC development and evaluation to capture the inherent high inter-subject variabilities observed with four LAI aqueous suspensions. By using bootstrapping, developed IVIVC models were able to predict the in vivo performance of LAI suspensions. More importantly, it demonstrated the feasibility of using such a technique to reflect the variability of pharmacokinetic characteristics of LAI suspensions.

Introduction

Long-acting injectables (LAI) are types of dosage forms that are able to deliver drug with high potency over a long duration varying from weeks to months.1 This long acting property can significantly unburden the dosing regimen, reduce the peak – trough fluctuation, and greatly improve the patient compliance.2 However, the extended duration of action poses significant challenges for product development and regulatory evaluation.3 For drug product approval, both pre-clinical and phase I, II and III clinical trials are essential to determine the safety and efficacy of a formulation. Specifically for a LAI, this can be extremely lengthy and expensive. It is therefore crucial to develop a reliable prediction tool such as in vitro/in vivo correlation (IVIVC) for product understanding and decision making throughout the drug product life cycle.

IVIVC is a mathematical model to correlate the in vivo performance (percent absorption and plasma drug concentration time (Cp-t) profiles) of a drug product with its in vitro characteristics (e.g. release profiles).4 According to the FDA guidance, a validated IVIVC may serve as a surrogate for bioequivalence studies during initial approval process or for post-approval changes to facilitate formulation screening, reduce the pre-clinical and clinical tests, and expedite development processes.4 However, owing to the complexity in the release profiles of LAI (burst release, multi-phasic release characteristics, and long release duration), it is challenging to establish a reliable IVIVC.5,6 Currently, only a limited number of publications are available for developing IVIVCs for LAI dosage forms.7, 8, 9 Only one IVIVC guidance is provided by FDA for extended-release oral dosage form for model establishment and validation.4 Thus far, researchers have referred to this guidance to develop IVIVC for non-oral products including LAIs. In addition, an industry-wide survey conducted recently shows that 40% of respondents never developed IVIVC or IVIVR for non-oral dosage forms, which may be attributed to the lack of research and regulatory references.10

Moreover, according to reported retrospective pharmacokinetic analyses for LAI suspensions, these dosage forms are prone to exhibit much higher inter-patient variability compared to conventional formulation types.11, 12, 13, 14 Johnson et al. reported substantial variability in paliperidone palmitate exposure, with Cmax spanning 6.5–113 ng/mL (median 40.1 ng/mL) and AUC0-t spanning 64,417–216,117 ng·day/mL (median 131,651 ng·day/mL). Likewise, Kraft et al. reported 44% and 48% coefficient of variations (CVs) for Cmax and AUC0-t in a long-acting lopinavir suspension.15 Nevertheless, IVIVC assessments generally rely on mean in vitro and in vivo data, which can mask the true extent of variability in LAIs’ PK. For example, Jain et al. reported an IVIVC model for LAI suspensions developed using mean in vitro and in vivo profiles and validated per conventional method and criteria based on FDA guidance.4,9 The validation of this IVIVC model with external formulations was unsuccessful (>15 % PE) per the FDA IVIVC evaluation criteria even though the predicted pharmacokinetic parameters (Cmax and AUC0-t) were well within the range of observed standard deviation 9. It is likely, or at least partly possible that current conventional IVIVC method development and evaluation may be overly stringent owing to inputting mean values instead of considering the whole data set with high variability. Moreover, there is no harmonized industrial-wide strategy for mean or individual data input to the PK model based on a recent survey.10 FDA’s guidance states that “methodology for the evaluation of IVIVC predictability is an active area of investigation and a variety of methods are possible and potentially acceptable”.4 It is important to investigate the feasibility and understand the effect of taking inherent variations into consideration for data input, IVIVC model development and assessment.

Bootstrapping is a randomized resampling process that creates a bootstrap population to draw meaningful inferences from highly variable data. This way the parameter estimates, such as Cmax and AUC0-t, better represents the overall variability, hence, improves the robustness of a model including IVIVC.16 In general, the objective of bootstrapping is not to reduce but account for and quantify variability within the IVIVC framework by generating confidence intervals and evaluating the variability in parameter estimates. This method has been widely used for formulation screening, process development, and dissolution data comparison.17, 18, 19 Moreover, this technique has been applied for IVIVC model development for oral dosage form.20,21 For example, to understand if a specific subject or a specific formulation has large impact on determining the slope of the IVIVC linear regression, U.S. FDA reported a strategy, that is, using bootstrapping to sequentially eliminating each study subject or each formulation from the whole data set using bootstrapping, and then recalculating IVIVC.22 The developed IVIVC was evaluated by comparing the prediction range (lower to upper 95%) to the whole clinical data set instead of just comparing mean profiles.

The objective of this work was to understand the feasibility of developing and evaluating IVIVC for LAIs with high PK variability via the bootstrapping technique. Previously, we have developed and evaluated a conventional IVIVC for microsuspension formulations via a classic two-stage (deconvolution and correlation) method using mean in vitro/in vivo profiles and mean PK parameters Cmax and AUC0-tbased on FDA guidance.4,9 In this study, a similar two-stage strategy together with bootstrapping technique will be applied to develop IVIVC on our previously investigated LAI microsuspension formulations. Individual in vitro and in vivo data (release and absorption profiles) were randomly sampled and replaced from the whole data set, which includes all in vitro/in vivo data. The sampled data was used to develop a linear correlation function between select in vitro/in vivo data. This process was repeated to exhaust all random sampling combinations. Parameters of each developed correlation function were used to determine a distribution of the estimated slopes of the bootstrapped IVIVC with a confidence interval. The developed IVIVC was evaluated by t-test as well as by determining whether observed mean PK parameters fall within the ranges [95% confidence interval (95% CI)] of predicted absorption profiles. This work may provide an alternative option for drug development and regulatory assessment.

Continue reading here

Bo Wan, Shalini Raj Unnikandam Veettil, Yi Li, Tien Ho, Chris Foti, Krutika Harish Jain, Development of in vitro/in vivo correlation for long-acting injectable suspensions via bootstrapping, Journal of Pharmaceutical Sciences, 2026, 104277, ISSN 0022-3549, https://doi.org/10.1016/j.xphs.2026.104277.