Discussion
In this study, popPK models for tezacaftor-ivacaftor and its active metabolites were successfully developed in children with real-world data using prior information from adolescent/adult models. Substantial variability in AUC was observed both within and across age and dosing groups. In general, AUC corresponded well with reported values in the product information, except for tezacaftor in 12-17y and ivacaftor in 6-11y≥30kg. Cmax and half-lives also corresponded closely with reported values, though in children 6-11y≥30kg Cmax tended to be higher and in children 6-11y<30kg half-lives tended to be shorter. (3) Also, a strong correlation between Cmin and AUC was found for tezacaftor-ivacaftor.
Although in this study sparse data was available, the prior subroutine was essential to successfully develop full popPK models, that also described the absorption phase and the distribution to peripheral compartments. It was not possible to obtain these extended models without prior information, emphasizing the importance of prior information in popPK modeling when data are limited, as is frequently the case in pediatric studies. When data are sparse, there are two methods to stabilize difficult-to-estimate parameters: 1. Fix them to a previous value described in literature; 2. ‘Inform’ them about the previous values. The latter strategy minimized bias in situations where the parameters differed somewhat between the preceding population and the population from which the sparse data were taken. (6)
The key covariate of relevance for this analysis was weight as predictor of CL/F(*fm), Q/F(*fm) and V/F(*fm), and was pre-defined. This is fairly typical in pediatric studies due to the large weight range, partially explaining the IIV in PK parameters. No other covariates explaining IIV were identified, likely due to the study being underpowered for this type of analysis. Interestingly, despite the real-world context, AUC variability in our studies (CV=16-88%) was in close agreement with the reported values in the product information. (3) A notable contribution of this study is the first real-world AUC data for children in the 30-40 kg weight class, as the registration studies relied on model-based predictions in this weight group due to other weight-dose group categorizations. (3, 5) Additionally, our findings of elevated AUC and Cmax in children 6-11y≥30kg raise questions about whether the tezacaftor-ivacaftor dose is appropriate for all children within this subgroup, as they receive the adult dose. Lowering the dose could reduce the risk of overexposure and possible development of side effects, as well as saving costs. The observation of a consistent relationship between Cmin and AUC in both this study and our previous work in adults, suggests that Cmin could be a valuable predictor of AUC for TDM in this population. (12)
This study benefits from several strengths, including the use of real-world data in children 6-17y that allowed the analysis of AUC in the 30-40 kg weight class, where limited data previously existed. Also, real-world PK data are relevant to account for the difference between controlled clinical trials and the complexity of drug use in diverse, every day settings. Next, the selective use of informative prior information from adolescents/adults was used to support portions of the model that were not well defined from the currently available pediatric data (Ka, D1, Vp, Q) and allowed for the estimation of the remaining parameters (CL, V). Limited sampling strategies were applied to reduce the burden of PK sampling in this age group. As well as the use of DBS sampling (at home), which served as a feasible PK sampling method and patients experienced this as less invasive than a venipuncture. (8) Especially in the era of changing CF care this method is preferable, as pwCF have better outcomes and will probably visit the hospital less frequent. (13) Furthermore, intake with fatty food was registered, which allowed for additional control over factors influencing drug absorption as fatty food increases the absorption of tezacaftor-ivacaftor. (3) Concomitant fatty food intake was not further investigated in the covariate analysis, as the absorption parameters were estimated with prior information and covariates can only be applied on parameters estimated without priors. (6)
Despite this, the study has some limitations. The sample size was small, due to faster access to ETI than expected, which limited the recruitment as some children did not start with tezacaftor-ivacaftor awaiting ETI. This also immediately implies that tezacaftor-ivacaftor is hardly used, since the introduction of ETI. However, the findings in this study are still useful because elexacaftor had little effect on the PK of tezacaftor-ivacaftor, as indicated in the registration report. (4) Furthermore, due to the short study duration, only one AUC curve was acquired for certain patients who used home-based DBS collections. This also resulted in incomplete covariate data in those patients (e.g. liver-enzyme measurements), as they did not visit the hospital during the study visit. This could have resulted in an underpowered covariate analysis, prohibiting a thorough assessment of the IIV in PK of tezacaftor-ivacaftor. This also made it impossible to assess the inter-occasion variability. Last, self-reported adherence could have influenced the results of the study, however this reflects a real-world situation.
To conclude, this study is the first to describe the popPK of tezacaftor-ivacaftor and its active metabolites in cwCF based on real-world data. The selective use of prior information from adolescent/adult models enabled the development of stable and robust models. The popPK models developed in this study could be used as a basis for more personalized medicine. Future applications of such TDM models could enhance dose optimization, particularly for children experiencing suboptimal efficacy, adverse effects, drug-drug interactions, or where adherence is a concern.