Patients and drug administration

A total of 661 women with twin gestation and 134 women with triplet gestation were recruited into 2 randomized controlled trials. The women received weekly injections of either 250 mg 17-OHPC in 1 mL castor oil or 1 mL castor oil alone from the time of enrollment (16 0/7 to 20 6/7 weeks’ gestation) until 35 weeks, unless delivery occurred earlier. Data that were recorded for each patient included maternal age, parity, race, body mass index (BMI), gestational age at enrollment, and gestational age at each blood sampling and at delivery. These data were evaluated as covariates in the pharmacokinetic analysis. This study was approved by the institutional review boards of each clinical site and of the data-coordinating center. Consent was given before enrollment into the study. The parent trials were registered at Clinical Trials.gov ( NCT00099164 ).

Pharmacokinetic sampling schedule

Among women who were recruited for the primary randomized controlled trials and who received all their scheduled injections of 17-OHPC, 97 women with twins and 53 women with triplets were undelivered and had a single blood sample drawn from 24-28 weeks’ gestation (epoch 1) for measurement of 17-OHPC concentration. Among these, 70 women with twins and 26 women with triplets were undelivered and had a second sample taken at 32-35 weeks’ gestation (epoch 2) for analysis. The infrequent (sparse) sampling described is useful in comparing plasma 17-OHPC concentrations over time and between groups but does not lend itself to classic pharmacokinetic analysis that requires frequent sampling during 1 dosing interval ( Appendix ). Fifteen of the 97 women with twins agreed to have a single blood sample taken daily for 7 consecutive days over a dosing interval of 1 week from 24-28 weeks’ gestation (intensive sampling). The first blood sample was drawn minutes before a scheduled injection. Recruitment of these 15 women was masked to treatment arm; therefore, women who received either 17-OHPC or placebo were included. All of the 15 women had received a minimum of 4 weekly injections of 17-OHPC from the time of enrollment in anticipation that steady state concentration in those women who received 17-OHPC would be achieved by the start of the pharmacokinetics study. Analysis of these plasma samples was not undertaken until completion of the clinical trials so that masking of treatment arm was maintained.

Sample analysis

For all 17-OHPC measurements, blood was collected in 10-mL tubes with ethylenediaminetetraacetic acid as the anticoagulant and centrifuged within 1 hour at 3500 g for 10 minutes. The supernatant plasma was aliquoted to 1-mL tubes and frozen at –70°C until analyzed by high-performance liquid chromatography with tandem mass spectrometry. The assay method was reported previously. The lower limit of detection of the assay for 17-OHPC was 1 ng/mL, inter- and intraassay variability at 10 ng/mL were 7.9% and 5.2%, respectively. The analyst and the clinical centers that were involved in recruitment remained masked to the treatment assignment until the analyses were completed.

Noncompartmental pharmacokinetic analysis

Nine of the 15 subjects who underwent “intensive sampling” (daily for 7 days) had received placebo and therefore were not included in the pharmacokinetic analysis, although their plasma samples had been analyzed. Pharmacokinetic parameters for the 6 women who received 17-OHPC and underwent “intensive sampling” were estimated with the standard noncompartmental approach that is implemented in WinNonlin software (version 4.0; Pharsight Corp, Mountain View, CA). Trough concentration, maximum concentration and time to maximum concentration were determined from the observed data. The elimination rate constant (λz) was determined by log-linear regression of the terminal linear disposition phase. Half-life was estimated by 0.693/λz. Area under curve (AUC) was calculated by the linear trapezoidal method. Apparent clearance (clearance/bioavailability) was estimated by dose/AUC; the apparent volume of distribution (volume of distribution/bioavailability) was calculated by dose/(λz AUCinfinity).

Population pharmacokinetic analysis

Population pharmacokinetics involves the pharmacokinetic evaluation of patients who are representative of the target patient population that is being treated with a medication. This method is used commonly to define patient-specific factors (such as weight, race, or sex) that contribute to the variability in drug concentrations. Large populations of subjects with fewer samples that are taken from each subject can be analyzed with this method. With a sufficiently large population, estimates of the impact of covariates on pharmacokinetic parameters can be made without the collection of numerous samples from each subject.

For the current study, analysis was performed by means of nonlinear mixed effects modeling that was implemented in the Monolix software (Monolix Group, Paris, France). Both sparsely sampled and intensively sampled data that were used in the noncompartmental analysis were included in the dataset.

Preliminary analysis for the structural model was performed by a comparison of a 1-compartment model with 2- and 3-compartment models. A proportional error model was used for describing residual variability. Interindividual variability was assumed to have a log-normal distribution. The patient characteristics BMI, race, fetal number, and parity were evaluated as covariates during the model-building process, which was described by Lavielle and Mentre. The goodness of fit of the final model was evaluated by an inspection of the following charts: scatterplots of predictions (population and individual) vs individual observations, population-weighted residuals vs predictions and independent variable (time), and absolute individual-weighted residuals vs individual predictions. In addition, model validation was performed with prediction distribution errors and visual predictive check, which were obtained by conducting 1000 Monte Carlo simulations of the dataset for the final model.

Further, simulations were carried out based on the final model to evaluate the effect of changes in BMI and dosing regimen on the plasma concentration time profiles of 17-OHPC.

Statistical analysis

GraphPad Prism (version 4.01; GraphPad Software, Inc, La Jolla, CA) was used for the statistical tests for significance. Nonparametric (Mann Whitney U ) tests were used for group comparisons. The Kruskal-Wallis test with Dunn’s multiple comparison was used for testing the equality of population medians in multiple groups. We considered probability values of < .05 to be significant.