Introduction

Sildenafil is the only phosphodiesterase 5 (PDE5) inhibitor currently approved as an oral treatment for pulmonary arterial hypertension in adults. Inhibition of PDE5-mediated breakdown of cyclic guanosyl monophosphate results in increased intracellular cyclic guanosyl monophosphate, leading to smooth muscle relaxation.¹ Given the high levels of PDE5 in the pulmonary endothelium, PDE5 inhibition results in selective vasodilation of the pulmonary vascular bed and reduction in pulmonary arterial pressure. Persistent pulmonary hypertension of the newborn (PPHN), a disorder in which the pulmonary artery pressure does not drop after birth, is diagnosed in 1.9/1,000 live births.² The syndrome includes pulmonary vasoconstriction, right-to-left shunting, and severe hypoxemia without evidence of congenital heart disease. The only treatment for PPHN that has been shown to improve oxygenation and reduce the need for the highly invasive extracorporeal membrane oxygenation procedure is inhaled nitric oxide.^3,4 However, the cost of inhaled nitric oxide treatment is high, and rebound PPHN following treatment withdrawal is common. Sildenafil has not been evaluated in randomized clinical trials for PPHN treatment, but an increasing body of evidence supports its efficacy in hypoxemia, PPHN, and similar indications in neonates.^5,6,7,8

This study is the first reported characterization of sildenafil pharmacokinetics in neonates. Sildenafil undergoes almost complete hepatic metabolism in adults, with <1% of an intravenous dose excreted unchanged in urine and feces. N-demethylation to UK-103320, which appears to be the only pathway of sildenafil biotransformation, is catalyzed primarily by CYP3A4 (79%) and CYP2C9 (20%).^9,10 UK-103320, which has half the in vitro potency of sildenafil, is the primary circulating metabolite and is extensively metabolized.¹⁰ Based on the disposition of sildenafil in adults, renal clearance of sildenafil in neonates may be assumed to be negligible, and metabolic clearance would be expected to be catalyzed by hepatic CYP3A4 and, to a smaller extent, CYP2C9. The fetal CYP3A isoform, CYP3A7, expressed predominantly during the early postnatal period may also potentially contribute to sildenafil metabolism in neonates, but substrate specificity of sildenafil for CYP3A7 has not been reported. The developmental expression of CYP enzymes in fetal and neonatal livers is not well understood, and most available information on the ontogeny of the major CYP enzymes, 3A, 2C, and 2D6, is based on in vitro studies using postmortem tissue.^{11,12,13,14,15,16} These in vitro studies reported a rapid increase in expression of CYP3A4 and CYP2C9 immediately after birth, from substantially lower levels in the fetal liver.^11,12,13 A concurrent decrease in CYP3A7 expression after the first week of life was also reported.¹¹ A recent longitudinal study of in vivo CYP2D6-mediated dextromethorphan O- and N-demethylation in subjects 2 weeks to 12 months of age demonstrated that CYP2D6 activity was detectable and concordant with genotype by 2 weeks of age and showed no further change up to 12 months.¹⁷ Therefore, rapid developmental changes in the expression of CYP enzymes may potentially influence sildenafil's pharmacokinetics during the early postnatal period. The objective of this analysis was to characterize the population pharmacokinetics of sildenafil in neonates, including potential effects of CYP ontogeny on metabolic clearance during the early postnatal period.

Results

In this open-label study consisting of eight escalating-dose groups, 36 full-term neonates diagnosed with PPHN were administered intravenous sildenafil within 72 h of birth. One subject died during the study from her underlying disease, and data from this subject were not included in the pharmacokinetic analysis. The dosing regimen consisted of a loading infusion of fixed duration ranging from 5 min to 3 h among treatment groups, followed by a continuous maintenance infusion of variable duration, ranging from 2.6 to 168 h for individual subjects. Mean duration of maintenance infusion for all treatment groups was 77 h, and, except for one subject, all subjects received the maintenance infusion for at least the minimum duration of 48 h specified in the protocol. One group of four subjects received the maintenance infusion only. Doses administered in each treatment group are summarized in Table 1. Observed sildenafil and UK-103320 plasma concentrations in each treatment group, and corresponding predictions from the final pharmacokinetic model, are plotted vs. postdose time in Figures 1 and 2, respectively.

Figure 1.

Observed (symbols) and individual predicted (lines) sildenafil plasma concentrations from the final model vs. time after initiation of intravenous sildenafil infusion, shown by treatment groups used for dose escalation. Observed and predicted values for individual subjects in each treatment group are identified by different colors.

Full figure and legend (34K)

Figure 2.

Observed (symbols) and individual predicted (lines) UK-103320 plasma concentrations from the final model vs. time after initiation of intravenous sildenafil infusion, shown by treatment groups used for dose escalation. Observed and predicted values for individual subjects in each treatment group are identified by different colors.

Full figure and legend (33K)

Table 1 - Mean rate and duration of intravenous sildenafil loading and maintenance infusions administered to neonates in each treatment (dose) group.

Full table (19K)

The base model included two-compartment disposition of both sildenafil and UK-103320, with estimated correlation between interindividual random effects on clearances of sildenafil and UK-103320 and between central volumes of distribution of sildenafil and UK-103320. Stepwise inclusion of two-compartment disposition for sildenafil and UK-103320 resulted in objective function reductions of 293 and 101, respectively, indicating statistically significant (P < 0.001, df = 2) improvements compared to models with one-compartment disposition. The Akaike Information Criterion was substantially improved with the inclusion of two-compartment disposition of sildenafil (-184.68) in a model with one-compartment disposition of both parent and metabolite (104.34). Further improvement in the Akaike Information Criterion was observed in a model with two-compartment disposition of both parent and metabolite (-281.69). Interindividual variability in peripheral compartment parameters for sildenafil and UK-103320 could not be estimated.

Covariates evaluated on clearance of sildenafil and UK-103320 included postnatal age, weight at screening, and gender (Table 2). Only weight at screening and gender were evaluated as covariates on central volume of distribution of sildenafil and UK-103320. An increase in sildenafil clearance with postnatal age was the only significant covariate effect in the final model. This predicted covariate effect was consistent with the observed gradual decrease in sildenafil concentrations in individual subjects during the course of the continuous infusion (Figure 1), indicating an increase in clearance with time.

Table 2 - Summary of demographic covariates for all study subjects.

Full table (21K)

Goodness-of-fit diagnostic plots for sildenafil and UK-103320 from the final model (Figures 3 and 4, respectively) indicate adequate fits to sildenafil and UK-103320 plasma concentration–time data. Because postnatal age increased with sampling time, weighted residuals of sildenafil data from the base and final models were plotted vs. postnatal age (Figure 5). An improvement in the weighted residual plot and a decrease in the residual variability in sildenafil concentrations (43.4% coefficient of variation (CV) in the base model vs. 39.8% CV in the final model) were observed in the final covariate model relative to the base model. Parameter estimates from the base, full, and final pharmacokinetic models, and 95% bootstrapped confidence intervals on the final model parameters, are shown in Table 3. A simulation-based predictive check using the final model parameters showed good concordance between distributions of observed and simulated values of dose-normalized area under the curve from time 0 to time of last quantifiable concentration (AUC(0–tlqc)/D), in the case of both parent and metabolite (Figure 6).

Figure 3.

Observed vs. (a) population predicted and (b) individual predicted sildenafil concentrations, and (c) weighted residuals vs. time, from the final model. A line of unity is shown in each of the observed vs. predicted plots, and a horizontal line at y = 0 is shown in the weighted residual plot. A trend line (loess smooth) is shown in bold in the weighted residual plot.

Full figure and legend (28K)

Figure 4.

Observed vs. (a) population predicted and (b) individual predicted UK-103320 concentrations, and (c) weighted residuals vs. time, from the final model. A line of unity is shown in each of the observed vs. predicted plots, and a horizontal line at y = 0 is shown in the weighted residual plot. A trend line (loess smooth) is shown in bold in the weighted residual plot.

Full figure and legend (30K)

Figure 5.

Plots of weighted residuals for sildenafil concentrations vs. postnatal age from the (a) base model and (b) final model. A horizontal line at y = 0 and a bold trend line (loess smooth) are shown in each plot.

Full figure and legend (40K)

Figure 6.

Predictive check of the final pharmacokinetic model. Histogram of individual dose-normalized area under the concentration–time curve from time 0 to time of last quantifiable concentration (AUC(0–tlqc)/D) values for (a) sildenafil and (b) UK-103320 from 1,000 simulated data sets. Median observed and simulated AUC(0–tlqc)/D values are shown by vertical lines, and quantile–quantile plots of the observed and simulated values are inset in each panel.

Full figure and legend (15K)

Table 3 - Summary of base, full, and final pharmacokinetic model parameter estimates for sildenafil and UK-103320 in neonates.

Full table (101K)

The evaluated covariates only partly explained the high interindividual variability in sildenafil and UK-103320 pharmacokinetics observed in this study. Interindividual variability (%CV) in sildenafil UK-103320 pharmacokinetic parameters was lower in the full model, but covariates other than postnatal age were not significant in the final model. The inability to detect significant covariates, other than postnatal age, was possibly due to the relatively small sample size and high residual variability in this study, and the potential effect of other covariates on sildenafil and UK-103320 pharmacokinetics in neonates cannot be ruled out on the basis of this study. An example of an influential covariate that was not significant in the final model was body weight, which showed substantial positive correlation with sildenafil clearance and central volume as well as with UK-103320 clearance and central volume in the full model (Table 3). Gestational age was highly correlated with body weight, such that the individual effects of these covariates were not identifiable. Therefore, only body weight was evaluated because it was considered a more useful covariate for purposes of potential dose adjustment. Based on the predicted relationship between postnatal age and sildenafil clearance, the typical value of sildenafil clearance increased from 0.84 l/h for a 1-day-old subject to 2.58 l/h at 7 days of age (Figure 7).

Figure 7.

Predicted relationship between sildenafil clearance and postnatal age for a typical neonate. The bold line represents the mean, and the shaded region represents the 90% confidence interval.

Full figure and legend (15K)

Plasma protein binding of sildenafil and UK-103320 were determined in 19 and 17 subjects, respectively, using pharmacokinetic samples collected at the end of the maintenance infusion in each subject. Mean (s.d.) protein binding of sildenafil was 93.9 (2.46)% and that of UK-103320 was 92.0 (3.07)%. The plasma protein binding of sildenafil in neonates appears to be lower than that previously reported for sildenafil in adult plasma (96%), consistent with the expectation of lower plasma protein content in neonates compared to adults.^18,19

The majority of adverse events reported in this study were of mild to moderate severity. The adverse events related to sildenafil treatment were hypotension (three subjects), labile blood pressure (one subject), and patent ductus arteriosus (one subject).

Discussion

Continuous sildenafil infusions of up to a maximum duration of 7 days were permitted in this pharmacokinetic study, primarily for the purpose of providing a potential clinical benefit to study participants. A continuous infusion of relatively long duration was also advantageous for characterization of potential time-dependent changes in the pharmacokinetics of sildenafil and UK-103320 within individuals. The increase in postnatal age of individual subjects was significantly correlated (P < 0.001) with sildenafil clearance, and we hypothesize that the rapid maturation of sildenafil metabolic clearance was due primarily to developmental changes in the expression of hepatic CYP enzymes. The effect of other physiological and anatomical changes that occur in the early neonatal period (hemodynamic changes, for example) on liver function and drug clearance are not well understood and therefore cannot be ruled out as potential contributory factors. Because the hepatic extraction ratio of sildenafil in adults (~30% of liver blood flow) is in the low to intermediate range, hemodynamic changes would not be expected to have a substantial effect on clearance but could partly contribute to the overall change.

Another potential explanation is the improvement in the clinical condition of the subjects, which could have resulted in an increase in metabolic activity. Although these other potential covariates are difficult to measure precisely to quantitatively evaluate their effects in the pharmacokinetic model, they would be expected to cause a time-dependent increase in the clearance of UK-103320 as well, given that this metabolite, like the parent, undergoes almost complete hepatic elimination in adults. The absence of a significant time-dependent increase in UK-103320 clearance in this analysis suggests that the increase in sildenafil clearance may be due mostly to maturation of specific enzymes involved in its metabolism. This hypothesis is supported by previously reported in vivo and in vitro evidence of the ontogeny of CYP3A4. CYP3A4 mRNA levels in human liver samples were shown to increase rapidly from low levels at birth to adult levels within 1 week, with a slower increase in in vitro functional CYP3A4 activity to 30–40% of adult values by 1 month of age.^11,20 In vitro functional activity of CYP3A7 was found to be the highest in the first week after birth, gradually declining to very low levels in adults. Total CYP3A protein content remained constant in liver samples from fetuses and 1-day-old to adult subjects, suggesting a switch from CYP3A7 to CYP3A4 expression in humans immediately at birth. In a population pharmacokinetic study of midazolam (a CYP3A4 substrate) in 0- to 10-day-old neonates, clearance was found to be ~1/7th to 1/5th of that in older children and adults, consistent with other reports that showed low expression of CYP3A4 at birth and very low in vitro affinity of CYP3A7 for midazolam.^21,22 In a report evaluating allometric scaling methods, allometrically scaled clearance of alfentanil (a CYP3A4 substrate) at birth was shown to be ~1/6th of that in adults.²³ The allometrically scaled clearance value was shown to mature to adult levels at ~1 month of age, with a predicted maturation half-time of 4.7 days.

The model-predicted sildenafil clearance in a 1-day-old neonate in this study was 0.84 l/h, which is equivalent to 8.05 l/h/70 kg, based on allometric scaling using the 3/4th power of body weight and assuming a neonatal body weight of 3.44 kg (mean weight in this study). Sildenafil clearance 24 h after birth was therefore ~1/3rd of reported adult sildenafil clearance (24 l/h/70 kg), suggesting that CYP3A7, the predominant hepatic enzyme at this stage of life, may have greater affinity for sildenafil than for midazolam and alfentanil. We have used allometrically scaled pharmacokinetic parameter values for comparison to adult values in this report because it is generally accepted that drug clearance tends to correlate nonlinearly with body weight (approximately to the 3/4th power), across the range of pediatric to adult body sizes, once the drug elimination pathways have matured.²⁴ Although sildenafil pharmacokinetic data are not available in older pediatric subjects to confirm that the allometrically scaled clearance of sildenafil remains constant through adulthood, this has been shown to be true for alfentanil.²³ Therefore, it would be appropriate to use allometric clearance values to compare maturation of drug clearance to the level expected in older children and adults. An increase in midazolam clearance with postnatal age was not detected in the midazolam population pharmacokinetic study, in which continuous infusions were administered to 0- to 10-day-old neonates.²¹ This appears to contradict the hypothesis of rapid maturation of CYP3A4-mediated in vivo clearance after birth. However, it should be noted that given the high interindividual variability in neonatal pharmacokinetics reported in the midazolam study, and observed in this study as well, relatively dense pharmacokinetic sampling in the presence of continuous infusions of long duration or recurring bolus doses within individual subjects may be necessary to detect a significant effect.

In the midazolam study, the average number of samples per subject was fewer than three, and only 8% of the samples were obtained after >4 days of infusion, which may explain the inability to detect a significant time-dependent effect on drug clearance. The maintenance infusion duration in our study was >4 days in seven subjects, with a mean duration of 3.2 days in all patients, and sildenafil and UK-103320 concentrations were each quantified in an average of 8.9 samples per patient. Given that the mean postnatal age at start of infusion was 1.4 days (range 0.4–3 days), the model-predicted increase in sildenafil clearance in this study is consistent with previous reports of the increase in CYP3A4 mRNA, as well as in vitro and in vivo CYP3A4 functional activity, starting soon after birth.¹¹ The model-predicted typical sildenafil clearance in a 1-day-old subject increased threefold to a value of 2.58 l/h (24.7 l/h/70 kg) at 7 days of age. The allometrically scaled clearance estimate in a 7-day-old neonate was similar to adult sildenafil clearance (24 l/h) calculated from a previous population pharmacokinetic report of oral sildenafil, assuming absolute bioavailability of 40% in adults.^25,26

Because sildenafil pharmacokinetic data are not available beyond the early neonatal period, it is not possible to conclude whether allometrically scaled sildenafil clearance had reached a plateau at adult levels by the first week of birth or whether it may increase further in older neonates before returning to adult levels. However, if CYP3A7 contributes partly to sildenafil metabolism in early neonates, which appears possible, this component of sildenafil clearance may be expected to decline after the first week of birth. This decline may be compensated by further increase in CYP3A4-mediated clearance of sildenafil, at least up to 1 month of age, based on the maturation profile of in vivo alfentanil clearance shown previously.²³ The contribution of CYP2C9 ontogeny to sildenafil metabolism may also be important, as CYP2C9 is the higher-affinity pathway in adults and contributes to about 20% of total sildenafil clearance.⁹ The ontogeny of CYP2C9, based on in vitro functional activity, is very similar to that of CYP3A4, being very low at the time of birth and increasing to about 30% of adult values at 1 month of age.¹³ Therefore the relative contribution of CYP3A4 and CYP2C9 to sildenafil metabolism in neonates may be expected to be similar to that in adults.

Total volume of distribution of sildenafil in neonates (22.4 l or 456 l/70 kg) was higher than that estimated in adults from previous reports (105–150 l), potentially due to lower plasma protein binding in neonates compared to adults. Although extracellular and total body water spaces in neonates are larger than in adults, the low hydrophilicity of sildenafil (cLog P = 2.7) suggests that distribution into these spaces may be limited.^19,25,26,27 Mean distribution and terminal half-life values of sildenafil and UK-103320 were calculated from the corresponding eigenvalues of the rate constant matrix derived from mean pharmacokinetic model parameters. The mean distribution half-life of sildenafil was 6.8 h, and typical terminal elimination half-life values were 55.9 and 47.7, respectively, in 1- and 7-day-old neonates. The terminal elimination half-life of sildenafil estimated in neonates was longer than that previously reported in adults (3–4 h). The clearance of UK-103320 in neonates (3.80 l/h or 36.4 l/h/70 kg) was lower than that in adults (159 l/h, data not published), and total volumes of distribution of UK-103320 in neonates (682 l/70 kg) and adults (606 l, data not published) were similar. As a result, the mean terminal elimination half-life of UK-103320 (10.9 h) was longer than that previously reported in adults (3–5 h).²⁶ The mean distribution half-life of UK-103320 in neonates was 0.7 h. The metabolite-to-parent ratio and the contribution of UK-103320 to PDE5 inhibition in neonates are likely to be higher than those in adults. Assuming that a dosing regimen consisting of a 3-h loading infusion followed by a 165-h maintenance infusion is initiated in a 1.44-day-old neonate, the concentration of UK-103320 would be expected to be 11% of parent at the end of the loading infusion and 76% of parent at the end of the maintenance infusion, thereby contributing up to 38% of the in vivo activity of sildenafil. In adults, UK-103320 exposure is about 15% of sildenafil exposure after intravenous dosing.

The high unexplained interindividual and residual variability in neonatal pharmacokinetics of sildenafil observed in this study could have resulted from the variable clinical condition and unstable hemodynamics of very ill patients, genetic or acquired differences in physiological and anatomical maturation among neonates, including maturation of hepatic enzymes, and potential interactions between the patients' clinical condition and maturation processes.²⁸ Although the specific influence of some of these covariates on drug pharmacokinetics is not well defined, they could potentially affect drug disposition in a number of ways. These covariates are difficult to measure precisely and may be changing rapidly within individuals, making it impossible to evaluate their effect on sildenafil pharmacokinetics. Similarly, high interindividual variability has been previously reported for other drugs administered to neonates in intensive care settings, resulting in problems with achieving precisely targeted concentrations with continuous infusions.²¹ It was difficult to achieve targeted concentrations during dose escalation in this study as well, as is evident from the overlap in plasma concentrations among dose groups (Figure 1), even though there was no evidence of dose dependence of sildenafil pharmacokinetics during analysis of the whole data set. The large between-patient variability in sildenafil exposure observed in this study also reduced the clinical importance of the observed within-patient decrease in sildenafil exposure, especially given the good toleration profile of sildenafil. Although body weight was not a significant covariate in the final model, the relatively high correlation of pharmacokinetic parameters with body weight in the full model suggests that adjustment of neonatal doses by body weight at start of infusion (on a per-kilogram basis) may have reduced variability in sildenafil exposure between subjects.

Sildenafil infusions were well tolerated in this study. The most common adverse event related to sildenafil treatment was hypotension, which is an expected finding based on the mechanism of action of sildenafil and the underlying compromised hemodynamics in patients with PPHN. Sildenafil treatment–related hypotensive effects appeared to occur most commonly during relatively rapid loading infusions administered over 5 min and 30 min, especially at the higher dose levels, whereas a slower loading infusion administered over 3 h at the highest dose level did not appear to cause hypotension in the four subjects who received the regimen. It may be theorized that the vasodilatory effect of sildenafil on the pulmonary vasculature resulted in a rapid reversal of the right-to-left shunt, which in turn resulted in an initial, temporary lower systemic blood pressure.

Methods

Ethics. This study was conducted in compliance with the ethical principles originating from the Declaration of Helsinki and with local laws and regulations relevant to the use of new therapeutic agents in the countries of conduct. Before starting the study, the sponsor had written and dated approval or a favorable opinion from the Institutional Review Board/Independent Ethics Committee for each study center. Written informed consent was gained from each subject's legally acceptable representative (parent/guardian) before a subject entered the study.

Study design. This was an open-label, multicenter study in 36 neonatal subjects with PPHN or hypoxic respiratory failure. Following a screening period up to 72 h of age, subjects received intravenous sildenafil infusions for up to 7 days, with or without the addition of standard treatment (inhaled nitric oxide and/or extracorporeal membrane oxygenation). Plasma samples (0.5 ml) for determination of sildenafil and UK-103320 plasma concentrations were collected every 24 h during the maintenance infusion, just prior to the end of the maintenance infusion, and 1, 4, 8, 12, 24, 48, and 72 h after the end of infusion. Plasma samples were also collected at 5 and 30 min after the end of the loading infusion or, when a loading infusion was not used, 6 and 12 h after the start of the maintenance infusion. In one group of four subjects receiving a 3-h loading infusion, samples were collected 30 min after the start and at the end of the infusion.

Determination of sildenafil and UK-103320 plasma concentrations and plasma protein binding. Samples for analysis of sildenafil and UK-103320 in plasma were prepared using solid-phase extraction followed by reconstitution. Sildenafil and UK-103320, along with their respective internal standards, sildenafil-d₈ and desmethyl sildenafil-d₈, were quantified using liquid chromatography and tandem mass spectrometry. Method imprecision (%CVs) for the analysis of plasma quality control samples at concentrations of 3, 30, and 350 ng/ml were 7.7, 4.3, and 6.2%, respectively, for sildenafil and 4.3, 3.9, and 6.2%, respectively, for UK-103320. The mean inaccuracy (bias) of the assay ranged from –1.9 to +2.6% for sildenafil and –1.3 to +2.0% for UK-103320. The lower limit of quantification of both sildenafil and UK-103320 was 1.0 ng/ml. Samples for determination of protein binding were prepared by ultrafiltration, followed by solid-phase extraction of the plasma ultrafiltrate and analysis by liquid chromatography and tandem mass spectrometry. Method imprecision values (%CVs) for the analysis of plasma ultrafiltrate quality control samples at concentrations of 0.6, 4, 40, and 75 ng/ml were 7.2, 5.2, 2.8, and 2.7%, respectively, for sildenafil, and 8.3, 1.8, 3.5, and 1.5%, respectively, for UK-103320. The mean inaccuracy (bias) of the assay ranged from -13.3 to +1.7% for sildenafil and -7.8 to +2.7% for UK-103320.

Pharmacokinetic model development. Model development was performed using nonlinear mixed-effects modeling software, version V, level 1.1 (NONMEM; Icon Development solutions, Ellicott City, MD).²⁹ Natural log-transformed plasma concentration–time data for sildenafil and UK-103320 were described simultaneously using the ADVAN5 subroutine within the PREDPP module. Because biotransformation to UK-103320 is the primary route of elimination of sildenafil in adults, it was assumed, for purposes of model development, that sildenafil was completely metabolized to UK-103320 in neonates. Both one- and two-compartment disposition of sildenafil and UK-103320 were evaluated. The structural model was parameterized in terms of sildenafil clearance, central volume of distribution of sildenafil, intercompartmental clearance of sildenafil, peripheral volume of distribution of sildenafil, UK-103320 clearance, central volume of distribution of UK-103320, intercompartmental clearance of UK-103320, and peripheral volume of distribution of UK-103320. Interindividual variability in clearance of parent and metabolite and central volume of distribution of parent and metabolite was modeled using multiplicative exponential random effects. A full-block (unstructured) allowing estimation of all variance–covariance elements was initially evaluated and reduced, if necessary, to avoid model ill-conditioning. Residual variability was modeled using an additive error model on the log scale. Residual variability was estimated separately for observed sildenafil and UK-103320 concentrations.

Covariates were added to the base model simultaneously to form the full model. Once the full model was developed, Wald's approximation method (WAM) was used to test for predictive covariates.³⁰ When using the WAM algorithm, the top 15 "optimal" models determined on the basis of maximizing Wald's approximation to the Schwarz's Bayesian criterion were evaluated using NONMEM. A comparison of the conditional rankings of the actual Schwarz's Bayesian criteria (based on the 15 NONMEM runs) to the WAM-based rankings was performed as a diagnostic to assess the performance of the WAM algorithm, and the model ranked first on the basis of the NONMEM runs was considered the final model. A Spearman's rank correlation exceeding 0.5 was used as an indicator of good performance of the WAM.³¹ The models' goodness of fit to the data was evaluated using the following criteria: change in objective function, visual inspection of different diagnostic plots, precision of parameter estimates, and decreases in both interindividual variability and residual variability. To avoid model ill-conditioning, the covariance matrix of the estimates at every stage of model development was inspected to verify that extreme pairwise correlations ( > 0.90) of parameters were not encountered. Additionally, the condition number of the correlation matrix of the parameter estimates was to be <1,000 whenever possible.

A predictive check of the final pharmacokinetic model was performed using stochastic simulations. A total of 1,000 data sets were simulated from the model using the mean model parameters and the matrix. AUC(0–tlqc)/D of sildenafil and UK-103320 was calculated for each subject in each simulated data set using the total dose of sildenafil (D) received. Simulated AUC(0–tlqc)/D values were compared graphically to those calculated from the original data set. AUC(0–tlqc)/D was used as the metric for the predictive check because other metrics, such as observed concentrations, were not suitable due to substantial variation of doses and regimens between subjects.

Conflict Of Interest

This study was sponsored by Pfizer Inc, manufacturer of Revatio and Viagra, brand names of sildenafil. All authors were employees and shareholders of Pfizer at the time the study was conducted.

References

1. Arnal, J.F., Dinh-Xuan, A.T., Pueyo, M., Darblade, B. & Rami, J. Endothelium-derived nitric oxide and vascular physiology and pathology. Cell. Mol. Life Sci. 55, 1078|[ndash]|1087 (1999). | Article | PubMed | ISI | ChemPort |
2. Walsh, M.C. & Stork, E.K. Persistent pulmonary hypertension of the newborn. Clin. Perinatol. 28, 609|[ndash]|627 (2001). | Article | PubMed | ChemPort |
3. The Neonatal Inhaled Nitric Oxide Study Group (NINOS). Inhaled nitric oxide in full-term and nearly full-term infants with hypoxic respiratory failure. N. Engl. J. Med. 336, 597|[ndash]|604 (1997). | Article | PubMed | ISI |
4. Clark, R.H. et al. Low-dose nitric oxide therapy for persistent pulmonary hypertension of the newborn. N. Engl. J. Med. 342, 469|[ndash]|474 (2000). | Article | PubMed | ISI | ChemPort |
5. Atz, A.M. & Wessel, D.L. Sildenafil ameliorates effects of inhaled nitric oxide withdrawal. Anesthesiology 91, 307|[ndash]|310 (1999). | Article | PubMed | ISI | ChemPort |
6. Abrams, D., Schulze-Neick, I. & Magee, A.G. Sildenafil as a selective pulmonary vasodilator in childhood primary pulmonary hypertension. Heart 84, E4 (2000). | Article | PubMed | ChemPort |
7. Atz, A.M., Lefler, A.K., Fairbrother, D.L., Uber, W.E. & Bradley, S.M. Sildenafil augments the effect of inhaled nitric oxide for postoperative pulmonary hypertensive crises. J. Thorac. Cardiovasc. Surg. 124, 628|[ndash]|629 (2002). | Article | PubMed |
8. Karatza, A.A., Bush, A. & Magee, A.G. Safety and efficacy of sildenafil therapy in children with pulmonary hypertension. Int. J. Cardiol. 100, 267|[ndash]|273 (2005). | Article | PubMed |
9. Hyland, R., Roe, E.G., Jones, B.C. & Smith, D.A. Identification of the cytochrome P450 enzymes involved in the N-demethylation of sildenafil. Br. J. Clin. Pharmacol. 51, 239|[ndash]|248 (2001). | Article | PubMed | ChemPort |
10. Warrington, J.S., Shader, R.I., Von Moltke, L. & Greenblatt, D.J. In vitro biotransformation of sildenafil (viagra): Identification of human cytochromes and potential drug interactions. Drug Metab. Dispos. 28, 392|[ndash]|397 (2000). | PubMed | ChemPort |
11. Lacroix, D., Sonnier, M., Moncion, A., Cheron, G. & Cresteil, T. Expression of CYP3A in the human liver|[mdash]|evidence that the shift between CYP3A7 and CYP3A4 occurs immediately after birth. Eur. J. Biochem. 247, 625|[ndash]|634 (1997). | Article | PubMed | ISI | ChemPort |
12. Koukouritaki, S.B. et al. Developmental expression of human hepatic CYP2C9 and CYP2C19. J. Pharmacol. Exp. Ther. 308, 965|[ndash]|974 (2004). | Article | PubMed | ChemPort |
13. Treluyer, J.M., Gueret, G., Cheron, G., Sonnier, M. & Cresteil, T. Developmental expression of CYP2C and CYP2C-dependent activities in the human liver: in-vivo/in-vitro correlation and inducibility. Pharmacogenetics 7, 441|[ndash]|452 (1997). | Article | PubMed | ChemPort |
14. Treluyer, J.-M., Jacqz-Aigrain, E., Alvarez, F. & Cresteil, T. Expression of CYP2D6 in developing human liver. Eur. J. Biochem. 202, 583|[ndash]|588 (1991). | Article | PubMed | ISI | ChemPort |
15. Leeder, J.S. et al. Variability of CYP3A7 in human fetal liver. J. Pharmacol. Exp. Ther. 314, 626|[ndash]|635 (2005). | Article | PubMed | ISI | ChemPort |
16. Jacqz-Aigrain, E., Funck-Brentano, C. & Cresteil, T. CYP2D6- and CYP3A-dependent metabolism of dextromethorphan in humans. Pharmacogenetics 3, 197|[ndash]|204 (1993). | Article | PubMed | ChemPort |
17. Blake, M.J. et al. Ontogeny of dextromethorphan O- and N-demethylation in the first year of life. Clin. Pharmacol. Ther. 81, 510|[ndash]|516 (2007). | Article | PubMed | ISI | ChemPort |
18. Walker, D.K. et al. Pharmacokinetics and metabolism of sildenafil in mouse, rat, rabbit, dog and man. Xenobiotica 29, 297|[ndash]|310 (1999). | Article | PubMed | ISI | ChemPort |
19. Kearns, G.L., Abdel-Rahman, S.M., Alander, S.W., Blowey, D.L., Leeder, J.S. & Kauffman, R.E. Developmental Pharmacology|[mdash]|Drug disposition, action, and therapy in infants and children. N. Engl. J. Med. 349, 1157|[ndash]|1167 (2003). | Article | PubMed | ISI | ChemPort |
20. de Wildt, S.N., Kearns, G.L., Leeder, J.S. & van den Anker, J.N. Cytochrome P450 3A: ontogeny and disposition. Clin. Pharmacokinet. 37, 485|[ndash]|505 (1999). | Article | PubMed | ISI | ChemPort |
21. Burtin, P. et al. Population pharmacokinetics of midazolam in neonates. Clin. Pharmacol. Ther. 56, 615|[ndash]|625 (1994). | PubMed | ChemPort |
22. Williams, J.A. et al. Comparative metabolic capabilities of CYP3A4, CYP3A5, and CYP3A7. Drug Metab. Dispos. 30, 883|[ndash]|891 (2002). | Article | PubMed | ISI | ChemPort |
23. Anderson, B.J. & Meakin, G.H. Scaling for size: some implications for pediatric anesthesia dosing. Pediatr. Anesth. 12, 205|[ndash]|219 (2002). | Article |
24. Anderson, B.J. & Holford, N.H. Mechanism-based concepts of size and maturity in pharmacokinetics. Annu. Rev. Pharmacol. Toxicol. 48, 303|[ndash]|332 (2008). | Article | PubMed | ChemPort |
25. Milligan, P.A., Marshall, S.F. & Karlsson, M.O. A population pharmacokinetic analysis of sildenafil citrate in patients with erectile dysfunction. Br. J. Clin. Pharmacol. 53 (Suppl 1),45S|[ndash]|52S (2002). | Article | PubMed | ChemPort |
26. Nichols, D.J., Muirhead, G.J. & Harness, J.A. Pharmacokinetics of sildenafil citrate after single oral doses in healthy male subjects: absolute bioavailability, food effects, and dose proportionality. Br. J. Clin. Pharmacol. 53(Suppl 1), 5S|[ndash]|12S (2002). | Article | PubMed | ChemPort |
27. Muirhead, G.J., Wulff, M.B., Fielding, A., Kleinermans, D. & Buss, N. Pharmacokinetic interactions between sildenafil and saquinavir/ritonavir. Br. J. Clin. Pharmacol. 50, 99|[ndash]|107 (2000). | Article | PubMed | ChemPort |
28. Abdel-Rahman, S.M., Reed, M.D., Wells, T.G. & Kearns, G.L. Considerations in the rational design and conduct of phase I/II pediatric clinical trials: avoiding the problems and pitfalls. Clin. Pharmacol. Ther. 81, 483|[ndash]|494 (2007). | Article | PubMed | ChemPort |
29. Beal, S.L., Sheiner, L.B., Boeckman, A.J. (eds). NONMEM Users Guides, (1989|[ndash]|2006). (Icon Development Solutions, Ellicott City, Maryland).
30. Kowalski, K.G. & Hutmacher, M.M. Efficient screening of covariates in population models using Wald's approximation to the likelihood ratio test. J. Pharmacokinet. Pharmacodyn. 28, 253|[ndash]|276 (2001). | Article | PubMed | ChemPort |
31. Conover, W.J. Practical Nonparametric Statistics 2nd edn. 252 (Wiley, New York, 1980).

Acknowledgments

We thank Dr Gregory Kearns for valuable advice on the design of this study. We express our gratitude to the study investigators, their staff, and Pfizer employees who helped with the conduct of the study and the study participants and their parents.