Page 79 - Read Online

Page 30 Pandey et al. J Transl Genet Genom 2021;5:22-36 I http://dx.doi.org/10.20517/jtgg.2020.45

understood. The variability in the level of protein, lipids, and small molecules involved in ADME of the
drug, such as the transporter protein OATP1B, might be responsible for the variability in patients’ PK.
Therefore, understanding the ADME processes will help in building a mechanism-based model and
strengthen the predictive ability of the population PK model. The studies highlighted here implemented PK
modeling strategies; however, there is a need to link the PK model with the PD model.

Pharmacodynamic modeling
The majority of the models developed for HU describe the PK and the variability in the PK profile of
patients. There is very little research done in the area of PD modeling of HU. There is a difference in the
timescale of PK and PD variables. The changes in PK variables occur on the scale of hours, while changes in
PD variables occur over weeks. The long-term usage of HU causes a significant increase in the percentage
[28]
of HbF and MCV of RBC .
[32]
Ware et al. did a univariate and multivariate analysis between PK and PD variables and covariates and
PD variables. The PD variables were HbF% at MTD and the MTD itself. Multivariate modeling identified
five significant variables related to HbF% and MTD, as listed in Table 2. However, multivariate linear
[32]
regression could not adequately predict HbF% and MTD .
The mechanism through which HU generates a specific response is not fully understood. To predict the
change in response without knowing the exact mechanism of the drug action, four basic types of structural
models (turnover models) have been used for describing drug response as a function of drug plasma
[56]
concentration . A general form of these models is shown in the following equation:

(11)

These four turnover models involve: (1) stimulation or inhibition of the rate of production (K ); and (2)
in
stimulation or inhibition of the rate of elimination (K ) of the response variable (R) by the drug plasma/
out
biophase concentration. The turnover models have been investigated to correlate the change in MCV and
[33]
HbF% with HU . In sickle cell patients, two models, for two response variables, HbF% and MCV, were
tested: (1) HU-mediated stimulation of K ; and (2) HU-mediated inhibition of K . For HbF% dynamics,
out
in
the elimination rate was modeled as K (1 - I ), where (1 - I ) is an inhibitory function. This model
max
out
max
could not correlate HbF% and plasma drug concentration for the given dataset. For MCV dynamics, the
-
-γ
-γ
elimination rate was modeled as K (1 - βy ), where y is the average drug concentration and (1 - βy ) is the
out
inhibitory function. To explain the variability in responses of patients, the NLME model was used . The
[33]
proportional residual error model in the form of Equation (4) was used to explain RV for HbF% and MCV.
The exponential models in the form of Equation (2) were used to describe the IIV for both HbF% and
MCV model parameters (K , K , I , and β). For HbF%, K has an exponential dependence on ΔMCV
max
in
in
out
(change in MCV/day) as a significant covariate. For MCV, β has an exponential dependence on ΔHbF%
(change in HbF%/day) as a significant covariate. The comparison between the population PD model results
and the observed HbF% and MCV data using VPC showed that the model’s performance was acceptable .
[33]
The study also compared two dosing regimens: (1) a daily dose of 1000 mg for seven days; and (2) a daily
[33]
dose of 1000 mg for five consecutive days followed by an interruption of two days . For patients with
the highest HbF% level, continuous dosing resulted in stronger response (quantified by a higher HbF%)
as compared to dosing with a two-day interruption. The effect of missing the dose on two days in a week
was cumulative as the difference in HbF% continued to increase with every week. The two different dosing
[33]
schemes did not seem to alter MCV dynamics significantly .
The percentage changes in HbF and MCV increase in HU treatment. There are very few PD models
developed, and only one of them describes the change in the HU response variables with the change in