Population pharmacokinetics refers to the study of drug kinetics in target populations with unique pathophysiology that might affect the drug dose–concentration relationship.Reference Jackson 1 Children with palliated single ventricle heart defects have very unique physiology including a propensity for elevated venous/hepatic pressures with associated hepatic congestion. These factors may alter drug pharmacokinetics, particularly of drugs undergoing hepatic metabolism, and therefore these patients represent an ideal population for population pharmacokinetic assessment.Reference Gewillig 2 – Reference Gentles, Gauvreau and Mayer 7
Sildenafil is a phosphodiesterase type-5 inhibitor that is often used to lower pulmonary vascular resistance in children and adults with single ventricle heart defects.Reference Beghetti 8 – Reference Uzun, Wong and Bhole 11 Sildenafil undergoes predominantly hepatic metabolism (cytochrome P450 3A4 [major route] and cytochrome P450 2C9 [minor route]) and is converted to an active metabolite – des-methyl-sildenafil – which has ~50% of the in vitro potency of phosphodiesterase type-5 as the parent drug.Reference Walker, Ackland and James 12 In adult patients with hepatic congestion secondary to pulmonary arterial hypertension or hepatic dysfunction (e.g. cirrhosis), sildenafil clearance is reduced by 50–80% with effects on the clearance of both sildenafil and des-methyl-sildenafil.Reference Muirhead, Wilner and Colburn 13 Although single ventricle patients frequently demonstrate hepatic dysfunction and congestion, no previous studies have evaluated sildenafil pharmacokinetics in single ventricle patients. Sildenafil dosing in children has been the source of recent controversy after the “sildenafil in Treatment-Naive Children, Aged 1–17 Years, with Pulmonary Arterial Hypertension” (STARTS) trials demonstrated increased mortality in children with pulmonary hypertension randomised to medium- or high-dose sildenafil when compared with low-dose therapy. 14 , Reference Barst, Ivy and Gaitan 15
In the present study, we sought to determine the pharmacokinetics of intravenous sildenafil in children with surgically palliated single ventricle heart defects. We tested the hypothesis that clearance of sildenafil and des-methyl-sildenafil would be directly related to surgical stage and hepatic pressures.
Materials and methods
Study population
The study design, samples analysis (limits of quantification=0.05 µg/L), and detailed cohort demographics have been previously described.Reference Hill, Barker and Fleming 16 , Reference Hill, Tunks and Barker 17 In brief, blood samples were collected as part of a prospective dose escalation pharmacokinetic and haemodynamic efficacy study of intravenous sildenafil. Children aged between 6 months and 10 years, children post stage II or stage III single ventricle surgical palliation, and children undergoing electively scheduled cardiac catheterisation were eligible for inclusion. Children with significant hepatic dysfunction, defined as having either aspartate aminotransferase or alanine aminotransferase⩾twice the upper limits of normal were excluded from participation. Dosing groups included 0.125 mg/kg (n=2), 0.25 mg/kg (n=5), 0.35 mg/kg (n=8), and 0.45 mg/kg (n=5). The study was approved by the Duke University Medical Center Institutional Review Board, and written informed consent for trial participation was obtained from the parent or guardian of each patient.
Population pharmacokinetics model development
Sildenafil and des-methyl-sildenafil concentration–time data were analysed using non-linear mixed effects modelling with Phoenix NLME 1.2 software (Certara, St. Louis, Missouri, United States of America) using the first order conditional estimation with interaction algorithm. We explored one-, two-, and three-compartment structural pharmacokinetic models for sildenafil and des-methyl-sildenafil, 100% conversion of sildenafil to des-methyl-sildenafil – which is represented by the clearance of sildenafil to des-methyl-sildenafil – or <100% conversion – which is represented by clearance sildenafil to des-methyl-sildenafil in addition to a sildenafil elimination clearance parameter, and proportional versus proportional plus additive residual error models.
Random effects on structural model parameters were considered supported by the data if shrinkage was <30% and condition number was <1000. Weight was included a priori as covariates for structural model parameters using a fixed (3/4 or 1) or estimated exponent. Diagnostic plots used for model evaluation included the following: observed versus population-predicted concentration and versus individual-predicted concentration; conditional weighted residuals versus population-predicted concentration and versus time after last dose; random effects and conditional weighted residuals histograms; and observed versus population-predicted and individual-predicted concentrations by patient. In addition, precision of parameter estimates and objective function values were used to assess the goodness-of-fit model.
Once the base model was selected, covariates were investigated for their influence on pharmacokinetic parameters. The continuous covariates evaluated were age, weight, cardiac index, which is calculated from catheterisation data at the time of sildenafil administration, serum creatinine, and hepatic pressure, which is directly measured at the time of sildenafil administration, and were centred around the median. Categorical covariates included surgical stage, race, and sex. In the final model, comparisons were made between individuals with hepatic pressures ⩾10 mmHg versus those with hepatic pressures <10 mmHg based on an a priori estimation of the approximate cut-off point for abnormal hepatic (central venous) pressures in children of similar age. Inter-individual variability estimates in pharmacokinetic parameters were plotted against covariates, and those with a discernible physiological and graphical relationship were evaluated for inclusion in the final model. The threshold for significance of a single covariate was reduction of the objective function by >3.84 (p<0.05). A forward-addition (p=0.05), backward-elimination (p=0.01) approach to covariate selection was planned for use if more than one covariate were found to be significant.
Model evaluation
Base and final model performance was evaluated based on successful minimisation, goodness-of-fit plots, and precision of parameter estimates. The final model was further evaluated with bootstrap procedures and visual predictive check. The precision of the final population pharmacokinetic model parameter estimates was evaluated using non-parametric bootstrapping (1000 replicates) to generate the 95% confidence intervals for parameter estimates. For the visual predictive check, the final model was used to generate 1000 Monte Carlo simulation replicates of sildenafil exposure, and the simulated results were compared with those observed in the study. The number of observed concentrations outside the 90% prediction interval for each time point was quantified.
Dose-exposure assessment
Individual pharmacokinetic parameters from the final model were used to simulate sildenafil and des-methyl-sildenafil concentration–time profiles after a single dose. Using Phoenix WinNonlin 6.3 software and simulated concentration–time profiles, elimination rate constants were calculated from linear regression of log concentration versus time in the elimination phase, area under the curve was calculated using the trapezoidal rule and linear up log down method, and elimination half-life (t1/2) was calculated as ln(2)/elimination rate constant assuming linear kinetics. As des-methyl-sildenafil has 50% activity compared with sildenafil, combined area under the curve was calculated as follows: (sildenafil area under the curve)+([des-methyl-sildenafil area under the curve]/2).
Results
A total of 20 children were enrolled in the study. Indications for cardiac catheterisation included the following: pre-Fontan evaluation (n=9), haemodynamic assessment secondary to relative cyanosis (n=3), pulmonary artery evaluation (n=4), poor function by echocardiogram (n=2), aortic arch evaluation (n=1), or suspected high pulmonary vascular resistance (n=1). Demographic features and physiological parameters are summarised by dosing group in Table 1.
Study population
Median (range) for continuous variables or n (%) for categorical variables.
Overall, 140 samples (73 sildenafil and 67 des-methyl-sildenafil) were above the limits of quantification (0.05 µg/L) and 20 (7 sildenafil, 13 des-methyl-sildenafil) were below. A single outlier peak sildenafil concentration and all samples below the limits of quantification were excluded. Therefore, the analysis included 72 sildenafil and 67 des-methyl-sildenafil samples from 20 children. The median (range) number of sildenafil and des-methyl-sildenafil samples per child was 4 (2–5) and 3.5 (2–5), respectively. The median (IQR) sildenafil and des-methyl-sildenafil sampling times for the first, second, third, fourth, and fifth samples were 20 (20, 20) minutes, 60 (43, 64) minutes, 1.9 (1.4, 3.0) hours, 4.1 (1.8, 5.1) hours, and 18.8 (17.3, 20.8) hours after dose, respectively. Concentration–time profiles stratified by surgical stage are shown in Figure 1. The median (range) sildenafil and des-methyl-sildenafil concentrations were 106 (1.59–775) and 16.6 (1.08–96.2) µg/L, respectively.
Concentration–time profiles. A, C=linear y-axis scale; B, D=log y-scale; circles=surgical stage 2; DMS=desmethylsildenafil; SIL=sildenafil; triangles=surgical stage 3.
Population pharmacokinetic model development
Figure 2 summarises the final structural model, which was a two-compartment model for sildenafil and a one-compartment model for des-methyl-sildenafil, with assumed 100% sildenafil to des-methyl-sildenafil conversion and metabolite clearance from the body represented by des-methyl-sildenafil clearance. The data only supported the addition of inter-individual variability parameters for sildenafil to des-methyl-sildenafil conversion and des-methyl-sildenafil clearance (i.e. shrinkage <30%). Body weight was included as a covariate for all base model parameters; addition of an allometric scaling (exponent=3/4) was not included as it did not improve the model fit (decrease in objective function value of −1.7).
Final structural pharmacokinetics (PK) model. C1=sildenafil central compartment; C2=sildenafil peripheral compartment; CL12=sildenafil intercompartmental clearance; C3=DMS central compartment; CLSIL-to-DMS=sildenafil clearance (conversion to DMS); CLDMS=metabolite (DMS) clearance; DMS=desmethylsildenafil; SIL=sildenafil.
Data describing model-building steps and model evaluation are included in the online supplement. In the base model, surgical stage demonstrated a suggestive graphical relationship with the inter-individual variability for conversion of sildenafil to des-methyl-sildenafil. Gender, surgical stage, and mean hepatic pressure demonstrated a suggestive graphical relationship for inter-individual variability for des-methyl-sildenafil clearance (Supplementary Figure 1). In the univariable screen, mean hepatic pressure and surgical stage were significant covariates for des-methyl-sildenafil clearance; after inclusion of hepatic pressure, no additional covariates were significant (Supplementary Figure 2 and Supplementary Table 1). Body weight was used as a covariate for all final model parameters. The final model demonstrated adequate goodness-of-fit with no significant deviation from the unity line for sildenafil model-predictions versus observed concentrations and no significant deviation from zero for residuals (Supplementary Figure 3A–D). There was under-prediction of the highest des-methyl-sildenafil concentrations found in surgical stage 3 children (Supplementary Figure 4A, B) and absence of bias in residuals (Supplementary Figure 4C, D).
Model evaluation
The number of observed concentrations outside the visual predictive check 90% prediction interval for sildenafil and des-methyl-sildenafil were 6/72 (8%) and 8/67 (12%), respectively, indicating good model predictive performance (Supplementary Figure 5). Relative standard errors of bootstrapped parameter estimates were <20%, and the percentage difference between model and bootstrapped median parameter estimates was ⩽5%, with the exception of the correlation coefficient, indicating precise estimation of population model parameters (Supplementary Table 2).
Dose-exposure assessment
Individual Bayesian parameter estimates are included in Table 2. As suggested by the final model, clearance of des-methyl-sildenafil decreased and des-methyl-sildenafil half-life (t1/2) increased in children with HP ⩾10 mmHg. Following a single dose of 0.35 mg/kg and using individual Bayesian pharmacokinetics parameter estimates, predicted area under the curve for sildenafil, des-methyl-sildenafil, and combined for the study population increased with increasing HP (Figs 3a–c). Median (range) predicted combined area under the curve was 533 µg*h/L [284–1046] and 792 µg*h/L [417–1431] (~1.5-fold difference) for children with hepatic pressure <10 mmHg and ⩾10 mmHg, respectively. Following the same single dose of 0.35 mg/kg in children with hepatic pressure <10 mmHg and 0.25 mg/kg in children with HP >10 mmHg, median (range)-predicted AUCTOTAL was similar between dose groups (533 µg*h/L [284–1046] and 565 [298–1022] µg*h/L, respectively [Figs 3d–f]).
Predicted total exposures in the study population with and without dose reduction for hepatic pressures >10 mmHg. A-C=single dose of 0.35 mg/kg; D-F=single dose of 0.35 mg/kg for HP <10 mmHg and 0.25 mg/kg for HP ⩾10 mmHg; Circles=DMS; Diamonds=AUCTOTAL=AUCSIL+AUCDMS/2; HP=hepatic pressure; Triangles=SIL.
Individual PK parameters
CL=clearance; DMS=desmethylsildenafil; HP=hepatic pressure; PK=pharmacokinetics; SIL=sildenafil; t1/2=elimination half-life
Discussion
This is the first population pharmacokinetic analysis of intravenous sildenafil in children and the first population pharmacokinetic analysis of any kind in children or adults with palliated single ventricle heart defects. These patients have very unique physiology, often demonstrating chronically elevated central venous pressures and secondary hepatic congestion.Reference Rychik, Veldtman and Rand 4 – Reference Baek, Bae and Ko 6 We demonstrated delayed clearance of the active sildenafil metabolite, des-methyl-sildenafil, with a direct relationship with increased hepatic pressures. The consequences are potentially clinically important with an estimated 1.5-fold increase in drug exposure (area under the curve) in individuals with hepatic pressures ⩾10 mmHg when compared with those with hepatic pressures <10 mmHg.
The only previous study regarding intravenous sildenafil in the paediatric population focused on term neonates with persistent pulmonary hypertension of the newborn, demonstrating typical clearance and volume of distribution at steady state (central+peripheral) for a 3-day-old neonate of 0.54 L/hour/kg and 7.8 L/kg, respectively.Reference Mukherjee, Dombi and Wittke 18 Studies in adults with pulmonary hypertension have reported approximate weight-normalised clearance of 0.38–0.59 L/hour/kg, with a half-life of 2.2–3.9 hours.Reference Muirhead, Rance and Walker 19 – Reference Nichols, Muirhead and Harness 21 The sildenafil clearance value of 0.62 L/hour/kg reported in this study is similar to clearance values in the healthy adult (0.59 L/hour/kg) and neonatal studies (0.54 L/hour/kg). Based on allometric scaling, we anticipated higher weight-normalised clearance values in children relative to adults; however, the altered physiology and morbidity related to single ventricle physiology may result in the lower-than-expected sildenafil clearance that we observed.
Interestingly, we detected a covariate effect of increased hepatic pressure on clearance of des-methyl-sildenafil, but not for the clearance of sildenafil itself. According to the model, for a child of given weight, des-methyl-sildenafil clearance is predicted to decrease ~7-fold as hepatic pressure increases from 4 to 18 mmHg. We hypothesise that increased hepatic pressure selectively impaired des-methyl-sildenafil clearance as a result of decreased biliary clearance. In mouse, rat, and dog, des-methyl-sildenafil is excreted to the bile and found in faeces, whereas sildenafil is cleared primarily by metabolism.Reference Walker, Ackland and James 12 , 22 Potentially selectively impaired des-methyl-sildenafil clearance due to increased hepatic pressure could be explained by unimpaired access of sildenafil to sites of metabolism in hepatocytes, but impaired des-methyl-sildenafil clearance through bile, secondary to increased hepatic pressure. This is consistent with a study in adults comparing oral sildenafil kinetics in individuals with and without liver cirrhosis. Cirrhotic individuals demonstrated reduced metabolism and clearance of both sildenafil and des-methyl-sildenafil, but the effect was more substantial for des-methyl-sildenafil (48% reduction in clearance for des-methyl-sildenafil versus 24% for sildenafil).Reference Muirhead, Wilner and Colburn 13 Potentially sildenafil clearance was unaffected in the present study because hepatic blood flow was not sufficiently affected in the range of the hepatic pressure in our study cohort. Sildenafil is an intermediate-to-high extraction ratio drug with clearance of 41 L/hour following intravenous administration in healthy adults,Reference Muirhead, Rance and Walker 19 compared with typical liver blood flow of 90 L/hour. As such, we would expect sildenafil clearance to be potentially impacted only if increased hepatic pressure resulted in significantly decreased liver blood flow. Nevertheless, the small sample size could also play a role in our inability to detect a relationship between sildenafil clearance and elevated hepatic pressure. In addition, none of our subjects demonstrated overt liver dysfunction as we excluded those with levels of either AST or ALT⩾twice the upper limits of normal, and only two study participants demonstrated a value for either AST or ALT that was outside the normal reference range.
It is notable that inclusion of surgical stage in addition to hepatic pressures did not significantly change our overall model. Stage III surgical palliation significantly alters venous physiology, typically raising central venous and hepatic pressures;Reference Gewillig 2 , Reference Beghetti 8 however stage II patients can demonstrate elevated hepatic pressures as a result of impaired ventricular diastolic function, and our results indicate that in these patients sildenafil dosing should be adjusted to account for reduced clearance. These findings have broader implications for both children and adults with pulmonary hypertension where central venous and hepatic pressures may also be substantially elevated. Potentially, sildenafil dosing in these patients might also require adjustment based on the degree of elevation in hepatic pressures.
Beyond clearance, volume of distribution and half-life are also critical determinants of drug kinetics. Sildenafil is likely distributed to tissues, and our data demonstrate a similar volume of distribution relative to adults. In previous studies in healthy adults, volume of distribution has been reported as ~105 LReference Nichols, Muirhead and Harness 21 compared with typical total body water of ~42 L. Weight-normalised total volume of distribution in the present study (1.81 L/kg) was within 20% of the volume of distribution reported in the above adult studies (~1.5 L/kg), but was ~5-fold lower than that reported in neonates.Reference Mukherjee, Dombi and Wittke 18 This finding is consistent with elevated total body water in neonates relative to children and adults. Similarly, the sildenafil half-life reported in this study (median 2.9 hours) is also within the range observed in healthy adults (2.2–3.9 hours), but is substantially below that reported for neonates (48–56 hours). Consistent with the association between increased hepatic pressure and reduced des-methyl-sildenafil clearance found in this study, des-methyl-sildenafil half-life in this study (median 3.3 hours) was prolonged relative to healthy adults (2.3 hours).Reference Muirhead, Rance and Walker 19
Overall, our pharmacokinetic data demonstrate the critical importance of population-specific pharmacokinetic assessment, particularly in patients with unique physiology that may affect drug metabolism. Population pharmacokinetics using sparse sampling methodologies have been widely applied to other patient populations but have not been commonly used in children with heart disease.Reference Marsot, Boulamery and Bruguerolle 23 – Reference Valitalo, Ranta and Hooker 26 Single ventricle patients are increasingly treated with sildenafil to lower pulmonary vascular resistance.Reference Beghetti 8 – Reference Uzun, Wong and Bhole 11 We have previously demonstrated that intravenous sildenafil acutely improves pulmonary blood flow and cardiac output in these patients, whereas others have demonstrated that sildenafil improves exertional tolerance and myocardial performance.Reference Hill, Barker and Fleming 16 , Reference Hill, Tunks and Barker 17 , Reference Goldberg, French and McBride 27 , Reference Giardini, Balducci and Specchia 28 Nevertheless, there are important safety concerns associated with sildenafil drug accumulation in children. The STARTS trial demonstrated increased mortality associated with medium- or high- dose oral sildenafil when compared with low-dose therapy.Reference Barst, Ivy and Gaitan 15 These findings prompted the US Food and Drug Administration to issue a safety warning recommending against the use of sildenafil in children. 29 This regulatory action has been contentious; however, the European Medicines Agency reviewed the same data and yet approved the use of sildenafil for paediatric use at low doses (10 mg three times daily for patients <20 kg and 20 mg three times daily for patients >20 kg). 30 Despite difficulties in interpreting the STARTS trial results, the findings highlight the critical importance of dosing adjustments in populations with delayed sildenafil clearance.
Although our study focused on intravenous sildenafil, if our hypothesised mechanism of reduced biliary clearance of des-methyl-sildenafil is accurate, then findings may also apply to oral dosing. In healthy adults, oral sildenafil is well-absorbed but undergoes significant first-pass metabolism with a reported bioavailability ranging from 25 to 63%. 22 Peak levels are seen 30–120 minutes (median 60 minutes) after oral dosing (versus 20 minutes after intravenous dosing in the present study) and the recommended intravenous dose is half the oral dose. Our data suggest that a child with significantly elevated Fontan pressures receiving “low dose” oral sildenafil could be exposed to drug levels (combined sildenafil+des-methyl-sildenafil) corresponding to higher doses. Reassuringly, in the STARTS trial, mortality was increased only in the subset of children with idiopathic pulmonary hypertension. There are limited data for comparison of sildenafil clearance and volume between similar populations after oral and intravenous doses, and pharmacokinetics and longer-term efficacy studies are needed to evaluate both oral and intravenous sildenafil in Fontan patients.
There are important limitations to this analysis. The sample size was relatively small, and because this study was conducted in children we employed a sparse sampling strategy with a median of 4 sildenafil samples per child. Sparse sampling is considered an appropriate approach to pharmacokinetics analysis in children where blood draws must be limited, and we used a sampling and population pharmacokinetics approach that has been endorsed by both the US Food and Drug Administration and the European Medicines Agency.Reference Marsot, Boulamery and Bruguerolle 23 – Reference Valitalo, Ranta and Hooker 26 , 31 , 32 Another limitation is that the study population included patients with differing surgical anatomy (approximately half were status post stage II surgery and the remainder status post stage III surgery). Surgical stage uniquely affects the physiology; however, we did not detect a covariate effect of surgical stage on pharmacokinetic parameters during model development after accounting for hepatic pressures. Finally, a two-compartment model for sildenafil and one-compartment model for des-methyl-sildenafil described the data appropriately, with precise model parameters (model estimates nearly identical to bootstrap estimates) and good model performance (good overlap of observed and simulated data on visual predictive checks). On the other hand, there was under-prediction of the highest observed des-methyl-sildenafil concentrations, occurring in stage III children. This likely resulted from the inability to incorporate inter-individual variability in volume of distribution of des-methyl-sildenafil in this model.
In conclusion, the major findings of this analysis include similar weight-normalised sildenafil clearance in children with palliated single ventricle heart defects when compared with healthy adults and neonates. Volume of distribution and half-life were also similar to healthy adults, but when compared with neonates volume of distribution was almost 5-fold lower, resulting in a much shorter half-life of 2.9 hours (versus 48–56 hours for neonates). Notably, we demonstrate an inverse relationship between hepatic pressure and clearance of des-methyl-sildenafil with estimated exposures ~50% greater in those with hepatic pressures ⩾10 mmHg when compared with children with hepatic pressures <10 mmHg. These data highlight the critical importance of pharmacokinetic analyses in patient populations with unique physiology, particularly because higher doses of sildenafil have been associated with increased mortality in previous studies in children. In our opinion, there is a critical need for pharmacokinetics and longer-term safety and efficacy studies in single ventricle patients for sildenafil as well as other drugs that undergo hepatic metabolism.
Acknowledgements
None.
Financial Support
K.D.H., J.S.L., and M.C.W. received support from the NIH-funded Duke Clinical and Translational Science Award (UL1TR001117). KDH and MCW received salary support from the Department of Health and Human Services Food and Drug Administration (1UO1FD004858-01). M.C.W. receives support from the US National Institutes of Health. M.R.S. was supported by training grant T32GM086330 from the National Institute of General Medical Sciences. This study was partially funded by an investigator-initiated research grant from Pfizer Incorporated, with additional funding provided by the Duke University Department of Pediatrics.
Conflicts of Interest
None.
Ethical Standards
The authors assert that all procedures contributing to this work comply with the ethical standards of the United States guidelines on human experimentation and with the Helsinki Declaration of 1975, as revised in 2008, and has been approved by the institutional review board of Duke University Medical Center.
Clinicaltrials.gov identifier: NCT01169519
Supplementary material
To view supplementary material for this article, please visit http://dx.doi.org/10.1017/S1047951115000359
