DATA ANALYSIS
The primary intent of the data analysis is to
assess whether dosage adjustment is needed for pregnant patients, and, if so, to
develop dosing recommendations for such patients based on gestational age or
trimester. The analysis, specifically modeling and dosing recommendations, will
depend on the study design characteristics. The categorization of pregnancy
status, either as nominal (e.g., trimester) or continuous (e.g., week of
gestation) data will direct the type of analysis performed. The Agency
encourages giving special analytical considerations to longitudinal study
designs and the baseline (e.g., postpartum) comparisons. The data analysis
typically consists of the following steps:
·
Estimation of PK parameters
·
Development of dosing recommendations
A. Parameter Estimation
The Agency recommends that total and unbound
plasma concentration data (and urinary excretion data if collected) be used to
estimate PK parameters of the parent drug and metabolite(s). Standard PK
parameters of a drug include the area under the plasma concentration curve (AUC),
peak concentration (Cmax), plasma clearance (CLT) or
apparent oral clearance (CL/F), renal clearance (CLR), apparent
volume of distribution (VZ/F or Vss/F), and terminal
half-life (t1/2). It is recommended that PK parameters be expressed
in terms of total and unbound concentrations and when applicable (e.g., oral and
renal clearance, expressed in terms of body weight, L/hr/kg). For drugs and
metabolites with a relatively low extent of plasma protein binding (e.g., extent
of binding less than 80 percent), description and analysis of the PK in terms of
total concentrations can be sufficient. Noncompartmental and/or compartmental
modeling approaches to parameter estimation can be employed.
B. Development of Dosing Recommendations
Specific dosing recommendations should be
constructed based on study results. Typically the dose should be adjusted to
produce a comparable range of unbound plasma concentrations of drug or active
metabolites in both controls and pregnant patients. Simulations are encouraged
as a means to identify doses and dosing intervals that achieve that goal for
pregnant patients at different trimesters or gestational ages.
One approach might be for the sponsor to
recommend, prior to the conduct of the studies, specific no effect
boundaries for the ratio of a PK measurement from pregnant patients and
controls, such as (AUCu,pregnant/AUCu,control) or (Dpregnant/Dcontrol). If the
90 percent confidence interval for the ratio of PK measurements falls within
these boundaries, the sponsor might claim no effect of pregnancy on PK,
and it would be reasonable to conclude that no dosage adjustment is needed for
pregnancy. The sponsor might determine no effect boundaries from
population or individual PK/PD relationships, dose-finding studies and/or
dose-response studies which are conducted as part of drug development.
Another approach might be for the sponsor to
assume no effect boundaries of 80-125 percent for Cmax
and AUC without further justification, recognizing that the small sample sizes
in pregnancy studies coupled with high intersubject variability can preclude
meeting the 80-125 percent no effect boundaries.
For some drugs, pregnancy may not alter PK
sufficiently to warrant dosage adjustment. A sponsor might make this claim by
providing an analysis of the study data to show that the PK measurements most
relevant to therapeutic outcome in pregnant patients are similar or equivalent
to those in the comparator group.
|
