Two key modifications were made to the published model based on information presented by Bansal et al.. First, the percent contributions of cytochrome P450 and UGT enzymes to CBD clearance were redefined as 20% and 80%, respectively, based on updated in vitro human liver microsome studies. Second, UGT1A7 was removed as an important contributing enzyme and the relative percent contributions of UGT1A9 and UGT2B7 to overall CBD clearance were redefined . To capture these changes, the percent contributions to CBD clearance as determined from each study were updated, and the fraction metabolized in the liver via the enzymes were adjusted to the revised percent contributions. While the update in CBD clearance pathways and fraction metabolized did not affect the overall pharmacokinetic simulations with CBD administered alone, it did increase accuracy regarding the drug–drug interaction predictions . As a result, we chose to use the updated literature for extrapolating to pediatrics. Following these modifications, anatomy and physiology were scaled for different infant ages, and growth and maturation of relevant processes including metabolic capacity, glomerular filtration rate, protein binding, and body composition, were adjusted for. Variability was applied to the anatomy and physiology to produce a virtual infant population. For the user defined protein, UGT1A9, activity was found not to be age dependent, and thus ontogeny was described with a linear function and geometric standard deviation of 1.5.
For model evaluation, two studies reported on the pharmacokinetics of CBD administered in children; however,indoor cannabis grow system the experimental data were not consistent. Particularly, the area under the curve AUC 0 − τ on day 1 presented by Wheless et al. vastly differed from the AUC0 − τ reported in adults and children aged 4–11 years reported by Devinsky et al.with similar weight-normalized doses. Thus, evaluation of the pediatric PBPK model was performed with Devinsky et al., where children aged 4–11 years were randomized to receive one of three doses of CBD oral solution daily . Using the developed pediatric PBPK model, infant populations of 200 individuals using the National Health and Nutrition Examination Survey population were simulated per age group in days: > 0 to 7, > 7 to 14, > 14 to 30, > 30 to 60, and > 60 to 365. Administration of CBD to these virtual breastfed infants differed from that given to adults. In the adult oral model, CBD solution was described as a dissolution-precipitation process that was dose specific and fit to describe the data. For extrapolation to pediatric populations, CBD was assumed to remain as a solution because of the small doses that breastfed infants receive. As CBD exhibits non-linear kinetics, each infant was assigned a daily dose of CBD solution until steady state was reached and AUC0 − τ, where τ = 24 hours, was taken. This process to simulate doses was performed with all CBD concentrations in milk and for each of the subgroups.The HMB dataset contained 200 breast milk samples of CBD concentrations obtained from 181 unique breastfeeding mothers. Participant demographics of the 181 mothers presented per breast milk sample are provided in Table 1. Of these samples, 124 from 118 participants had only one maternal type of administration. The three methods to account for BLQ values produced similar results. Thus, LLOQ/2 with LLOQ as 0.1 ng/mL was applied. Only concentrations above the LLOQ were used in the subgroup analyses as BLQ values tended to produce unsatisfactory residual distributions when incorporated into the log-linear regression models. The proportion of BLQ values were similar across subgroups .
Descriptive plots of each assessed subgroup using the exposure-concentration subset while accounting for BLQ values are presented in Fig. 1. A backward step-wise elimination procedure was performed involving TAD, administration type, and their interactions. An interaction with administration type and dose frequency was not feasible for model testing because of the low sample size. The final model included administration type, which exhibited satisfactory residual behavior. Post-hoc pairwise comparisons between administration types across the three p value adjustment methods suggested that oil versus joint/blunt, joint/blunt versus pipe, and edible versus pipe had significantly different estimated marginal means. Therefore, administration type was grouped into two contrasting subgroups, oil or pipe and joint/blunt or edible, for subsequent dose simulations. Goodness-of-fit plots, estimated marginal means, their 95% confidence intervals, and model estimates are presented in Figs. S1 and S2, and Table S1 of the Electronic Supplementary Material. Administration type was not found to be significant after controlling for TAD, and thus BLQ values had the same chance of occurring for all administration types. The model performed well with the Hosmer–Lemeshow test resulting in a p value of 0.768. The distributions of CBD in milk concentrations and administered doses to virtual breastfed infants are presented in Table 2. The developed pediatric PBPK model evaluation with Devinsky et al. results are shown in Table 3. Predicted AUC0 − τ were comparable to observed AUC0 − τ, which provided confidence in the ability of the model to accurately predict exposures in pediatrics. Pediatric PBPK model-predicted, daily, steady-state AUC0 − τ of breastfed infants across the age groups for all CBD concentrations, joint/blunt or edible exposure only, and oil or pipe exposure only compared to children administered the CBD therapeutic dose are presented in Fig. 2.
Calculated UARs for each age group are shown in Table 4. Through use of real-world CBD concentrations in breast milk, this study provided additional information on potential concentrations of CBD exposure in breastfed infants. By examining the relationship between the maternal type of administration and concentrations in breast milk, it was determined that oil or pipe tended to result in higher predicted concentrations as compared with joint/blunt or edible forms. Additionally, this work found that a significant proportion of breast milk samples contained BLQ values that likely contributed to the low predicted exposures to breastfed infants. The longer the TAD, the greater the presence of BLQ concentrations were in breast milk. Moreover, BLQ values had the same chance of occurring for all administration types. Knowledge about the impact of TAD on BLQ concentrations across administration types could have clinical advice implications, such as the existence of optimal breastfeeding times when taking CBD and CBD-containing products. A strength of this study was based on the ability of the PBPK model to predict AUC0 − τ reasonably in adults. This increased our confidence especially in the AUC0 − τ predictions in children aged 4–11 years for model evaluation. Although geometric mean AUC0 − τ was predicted to be 1.7-fold less than observed in Devinsky et al.for 10-mg/kg/day dosing, our findings were relatively in line with Wheless et al.. It is worth noting that the pediatric CBD PBPK model developed by Bansal et al. predictions were in line with Devinsky et al. at this dose, which suggests a potential significant role of accumulation and time-dependent auto-inhibition. Yet, since their model tended to over predict in all other doses in children, and our model predictions were acceptable in adults, further research is needed to confirm the potential CBD impact on enzymes responsible for its own metabolism. Beyond the ability to predict exposures,cannabis grow set up the UAR accounts for the anatomy and physiology of breastfeeding infants; age-dependent factors, such as milk intake volumes as a function of age; and variability in the infant and maternal population, such as maternal pharmacogenotypes that could lead to an increased presence of medication in breast milk. The UAR was calculated using the pediatric PBPK model-predicted exposures in virtual breastfed infants. This novel metric offers an improvement over current metrics that focus solely on the potential dose received by the breastfed infant, without accounting for exposure . The UAR calculated for CBD revealed that even the exposures of the most vulnerable breastfed infant are well below the exposures of children aged 4–11 years receiving the lowest approved dose for approved indications. This finding serves as additional exposure information to healthcare providers to consider when discussing CBD use by mothers in relation to their breastfeeding infants. For context, our group has simulated breastfeeding exposures for lamotrigine and escitalopram in previous work. Predicted breastfeeding infant exposures tended to reach levels of exposure from adults taking therapeutic doses for lamotrigine, but not for escitalopram.
The UAR was also calculated for lamotrigine and was determined to be relatively high for some age groups. These observations were in line with adverse reactions reported for lamotrigine and escitalopram, with more observed in the former than the latter. Thus, the UAR serves as a useful tool to anticipate potential responses in breastfeeding infants. In regard to CBD, it would be of interest to follow-up in future studies assessing breastfeeding infant adverse reactions and effects on neurodevelopment to understand the relationship between the UAR results of this study with response information. This work recognizes the great uncertainty of CBD bio-availability in breastfed infants. In adults, bio-availability is low and greatly impacted by food. To address this issue, we used the idea that breastfed infants receive small doses of CBD and thus the precipitation-dissolution precipitation cycle experienced in adults was not expected. Therefore, CBD was given as an oral solution without the dissolution complexities. Moreover, because a solution is already dissolved, the food effect was not relevant in our virtual breastfed infants. As a result, our work was conservative with the pediatric PBPK model-predicted 0- to 1-year-old infant bio-availability being 0.54–0.68, as compared with 0.24 in adults administered the 200-mg oral solution. Even with this larger infant bio-availability, the UAR was still very low. The low sample size per subgroup serves as a limitation to this study. Although administration type was found to be a significant subgroup, further data to support this finding are warranted. Likewise, larger sample sizes are needed to assess other potential subgroups , such as those given by dose frequency. It is possible that oil or a pipe maternal type of administration tended to have higher dose frequencies. Similarly, the relationship between TAD and BLQ concentrations in milk could be influenced by dose frequency. However, analyses with the limited dose frequency information we had suggest this not to be the case. A limitation of the parent study is that maternal exposure information on dose, timing, and type of administration relied on a maternal report and may therefore be inaccurate. Furthermore, maternal administration information was typically measured in the previous 2 weeks prior to milk sample collection. As a result, less data were acquired on the long-term frequency of use, which may contribute to the infant dose. A further limitation to this study relates to the inability to validate our workflow with CBD concentrations measured in breastfed infant plasma. As these data have not been reported in the literature, we were not able to check whether the pediatric PBPK model-predicted infant plasma concentrations were in line with observed concentrations. Future studies should focus on collecting and analyzing plasma concentrations from infants breastfed by mothers taking CBD or CBD-containing products to confirm our results. Because the study of CBD in milk concentrations was based on highly dispersed observational data, it can only shed light on the potential association between concentrations and administration types and any statement on causality should be avoided. Nevertheless, this study was able to draw conclusions on infant exposures from the real-world maternal use of CBD and CBD-containing products, which can be insightful to healthcare providers in advising breastfeeding mothers taking CBD and CBD-containing products. Future studies investigating the relationship of maternal administered CBD doses to potential breastfed infant responses, particularly neurodevelopmental delay, will add to our understanding of the CBD dose–exposure–response relationship in infants. For example, low predicted exposures of CBD in breastfed infants as compared with children receiving therapeutic doses for approved indications may still have the potential for adverse effects, if infants are more susceptible in early brain development. Another future direction includes studying CBD metabolites, especially 7-hydroxycannabidiol, which is known to have the activity and potential to accumulate in breast milk. Further cannabinoids are another area of focus. Notably, tetrahydrocannabinol is observed to have substantially greater concentrations in milk as compared with CBD. Similar studies to our presented work can be applied to 7-hydroxycannabidiol and tetrahydrocannabinol to provide a fuller perspective on cannabis use during breastfeeding.