Population pharmacokinetics of peginterferon α2a in patients with chronic hepatitis B

There were significant differences in response and pharmacokinetic characteristics to the peginterferon α2a treatment among Chronic Hepatitis B (CHB) patients. The aim of this study is to identify factors which could significantly impact the peginterferon α2a pharmacokinetic characteristics in CHB patients. There were 208 blood samples collected from 178 patients who were considered as CHB and had been treated with peginterferon α2a followed by blood concentration measurement and other laboratory tests. The covariates such as demographic and clinical characteristics of the patients were retrieved from medical records. Nonlinear mixed-effects modeling method was used to develop the population pharmacokinetic model with NONMEM software. A population pharmacokinetic model for peginterferon α2a has been successfully developed which shows that distribution volume (V) was associated with body mass index (BMI), and drug clearance (CL) had a positive correlation with creatinine clearance (CCR). The final population pharmacokinetic model supports the use of BMI and CCR-adjusted dosing in hepatitis B virus patients.


Patients and treatment
All hospitalized cases will come from 302 Military Hospital of China, who are diagnosed with chronic Hepatitis B and treated with peginterferon α2a. Individuals were considered Chronic hepatitis B and in line with interferon treatment indications of "Guide of chronic Hepatitis B Prevention, China, 2012".

Exclusion criteria
1) Combined with HCV, HDV or HIV; 2) Receiving other drug treatment which may affect the pharmacokinetics or pharmacodynamics of interferon; 3) Combined with hepatic carcinoma, severe primary disease of heart, kidney, lung, endocrine, blood, metabolism and gastro intestine; 4) Pregnant or lactating women; 5) Received other antiviral therapy 3 months before this trial; 6) Medication compliance is poor.

Treatment
Peginterferon α2a was subcutaneously injected into patients once a week.

Blood collection point
The T max of peginterferon α2a is about 72 hours. In order to ensure that the blood collection point evenly distributed at the absorption phase, near the peak concentration and distribution phase, every patient will be randomly assigned to three groups after administration with peginterferon α2a: blood collection within 48 hours, between 48 hours and 96 hours and after 96 hours. Because the nonlinear mixed-effects model can fit the pharmacokinetic curve, specific blood collection time will be determined by research doctors according to the negotiation with patients.
With the consent of the patient, multiple blood samples could be collected at different phases or different hospitalizations, but the maximum is 4 times.

Blood sample management
Collecting elbow vein blood 5 ml, centrifugal separation of serum, and stored in the -70 ℃ refrigerator to be tested. Peginterferon α2a concentrations in the serum samples were analyzed using a commercial Human IFN-α Multi-Subtype ELISA Kit (product # 41105) with a detection limit of 15 pg/mL manufactured by Pestka Biomedical Laboratories, Inc.

Clinical information collection
For the Clinical Laboratory has passed the ISO 15189 certification, clinical characteristics of the patients, including age, body weight, serum creatinine, creatinine clearance, body mass index, height, sex, aspartate transaminase, alanine transaminase and disease grade were retrieved from medical records. Laboratory results from the records are tested in one week before blood sample collection.

Basic pharmacokinetics model
Nonlinear mixed-effect modeling method was employed to develop the basic pharmacokinetic model for peginterferon α2a. All the plasma concentration-time data set was fitted using the NONMEM software (Version 72, ICON Development Solutions, Ellicott City, MD, USA) with first-order conditional estimation with Interaction (FOCE-I) approach. IIV was described by an exponential variability model as follow: P i =P×e ηi Equation 1 where P means the typical value of parameter and P i is the ith patient's individual parameter. IIV was assumed to follow a log-normal distribution, and the random variable ηi is normally distributed with mean 0 and variance of ω 2 . Combined error model (proportional error and additive error) was used to calculate the residual error of the pharmacokinetic model: C ij= C ij P × (1+ε 1ij ) + ε 2ij Equation 2 C ij P and C ij represent model prediction and individual observation in ith patient's jth concentration, respectively. ε 1 characterizes the proportional error and is normally distributed with mean 0 and variances of σ 1 2 . ε 2 describes the additive error and is also distributed with mean 0 and variances of σ 2 2 . One-and two-compartmental open models with first-order absorption and elimination were attempted to fit the data set. Model comparisons were made using the objective function value (OFV) for model discrimination, with the significance level was selected a priori at 0.05 (df = 2, ∆OFV = 5.99).

Final pharmacokinetic model
Based on the basic pharmacokinetic model, effects of covariates, including age, body weight, serum creatinine, creatinine clearance, body mass index, height, sex, aspartate transaminase, alanine transaminase and disease grade on the basic model were investigated. For the categorical covariates, such as sex and disease grade, they were incorporated using indicator variables. The other covariates were continuous and they were included into the model in the following ways:

Equation 3
COV and MEAN COV mean the covariate and the mean value of this covariate, respectively. θ is the coefficient representing the relationship between the COV and P i . The effects of covariates were investigated using the forward inclusion-backward elimination approach. A forward inclusion was used and covariates with a change in the OFV of ≥ 3.38 (P = 0.05) were incorporated one at a time. After adding all the "significant" covariates from the forward inclusion in one model, the backward elimination step was performed. Covariates that caused a change ≥ 6.63 (P = 0.01) in the OFV when eliminated were kept in the model.

Model evaluation and validation
Visual method was used to evaluate the basic and final population pharmacokinetic models. Scatter plots of observation (dependent variable, DV) versus prediction (PRED), the conditional weighted residuals (CWRES) against PRED and TIME (time after dose) were drawn by Microsoft Office Excel 2007 software. Furthermore, the stability of the final model was assessed using the bootstrap technique. 1000 data sets were generated using the re-sampling method, and they were analyzed using NONMEM software. After obtaining the mean and standard error of the fixed-effect parameters, the population estimates obtained from the final model were compared with the median, 2.5% and 97.5% percentiles (95% confidence interval, 95%CI) of the bootstrap replicates.