Physiologically-based pharmacokinetic modeling to investigate the effect of maturation on buprenorphine pharmacokinetics in newborns with neonatal opioid withdrawal syndrome

Abstract

Neonatal opioid withdrawal syndrome (NOWS) is a major public health concern whose incidence has paralleled the opioid epidemic in the US. Sublingual buprenorphine is an emerging treatment for NOWS, but given concerns about long-term adverse effects of perinatal opioid exposure, precision dosing of buprenorphine is needed. Buprenorphine pharmacokinetics (PK) in newborns, however, is highly variable. To evaluate underlying sources of PK variability, a neonatal physiologically-based pharmacokinetic (PBPK) model of sublingual buprenorphine was developed using Simcyp (v19.1). The PBPK model included metabolism by cytochrome P450 (CYP) 3A4, CYP2C8, UDP-glucuronosyltransferase (UGT) 1A1, UGT1A3, UGT2B7, and UGT2B17, with additional biliary excretion. Maturation of metabolizing enzymes was incorporated, and default CYP2C8 and UGT2B7 ontogeny profiles were updated according to recent literature. A biliary clearance developmental profile was outlined using clinical data from neonates receiving sublingual buprenorphine as NOWS treatment. Extensive PBPK model validation in adults demonstrated good predictability, with geometric mean (95% confidence interval [CI]) predicted/observed ratios (P/O ratios) of area under the curve from zero to infinity (AUC_0–∞), peak concentration (C_max), and time to reach peak concentration (T_max) equaling 1.00 (0.74–1.33), 1.04 (0.84–1.29), and 0.95 (0.72–1.26), respectively. In neonates, the geometric mean (95% CI) P/O ratio of whole blood concentrations was 0.75 (0.64–0.87). PBPK modeling and simulation demonstrated that variability in biliary clearance, sublingual absorption, and CYP3A4 abundance are likely important drivers of buprenorphine PK variability in neonates. The PBPK model could be used to guide development of improved buprenorphine starting dose regimens for the treatment of NOWS.

Introduction

The opioid epidemic is a devastating and ongoing US public health emergency. In 2019, close to 50,000 people died from opioid overdose.¹ Moreover, 6.6% of women reported prescription opioid use during pregnancy, 21.2% of which disclosed opioid misuse.² Recent studies indicate that the coronavirus disease 2019 (COVID-19) pandemic, with its detrimental toll on healthcare systems, mental health, and socioeconomic wellbeing, has further fueled the opioid epidemic,³ and secondarily exacerbated the risk of maternal opioid use. Neonatal opioid withdrawal syndrome (NOWS) is a constellation of symptoms caused by prolonged in utero exposure to opioids. The incidence of NOWS has paralleled the expanding opioid epidemic in the US, increasing from 4.0 to 7.3 per 1000 hospital births between 2010 and 2017.⁴ NOWS is marked by gastrointestinal dysfunction and neurologic excitability, and symptoms include poor weight gain, tremors, and fever.⁵ In those infants whose symptoms are insufficiently controlled through non-pharmacological approaches, pharmacological treatment with either morphine, methadone, or buprenorphine is initiated.⁶

Although buprenorphine has been a standard of care in treating adults with opioid use disorder, it is just emerging as treatment for infants with NOWS.⁷ Buprenorphine undergoes substantial first-pass metabolism, attaining a bioavailability of only 6–15% following oral administration.^8,9 Sublingual administration negates this to a certain extent. Buprenorphine is primarily metabolized by cytochrome P450 (CYP) 3A4 and CYP2C8 to norbuprenorphine,¹⁰ which both undergo glucuronidation catalyzed by UDP-glucuronosyltransferase (UGT) 1A1, UGT1A3, UGT2B7, and UGT2B17.¹¹ In addition, buprenorphine and its metabolites are cleared through biliary excretion. Following parenteral administration, 33–47.2% of the dose is recovered in feces as unconjugated buprenorphine,^8,12 although this may have initially been excreted as glucuronide conjugate and subsequently undergone hydrolysis by the gut microbiome back to buprenorphine. However, unchanged buprenorphine is also found in meconium obtained from newborn infants exposed to buprenorphine in utero,¹³ despite fetal intestines being sterile. This strongly suggests prenatal biliary excretion of unchanged buprenorphine. To date, the transport mechanism behind biliary excretion has not been elucidated.¹⁴

Although buprenorphine pharmacokinetics (PK) and pharmacodynamics (PD) have been extensively studied in adult populations, clinical data are still very limited in neonates.¹⁴ Physiologically-based pharmacokinetic (PBPK) modeling has emerged as a suitable approach to characterize and predict the effect of developmental changes on drug disposition in neonates and children.¹⁵ PBPK modeling and simulation is a mathematical modeling approach that conjoins human body characteristics with drug physicochemical properties to conceptualize PK behavior. Its application is specifically beneficial when studying neonatal PK, since, due to ethical and practical considerations, collection of clinical PK data in this vulnerable patient population is exceedingly limited. Using PBPK modeling and simulation, real-life patients can be recreated as virtual populations with matching demographic properties, but minor interindividual differences reflective of the expected variability between individuals of the same demographic profile, creating for each patient what has been coined “virtual twins”.¹⁶ These virtual twins can then be used to study the effect and expected variability of different clinical scenarios on individual drug disposition.

Previous studies have shown that buprenorphine PK following sublingual administration in neonates is highly variable,^17-19 but factors driving this variability have not been fully identified. This hampers the development of evidence-based and personalized treatment of NOWS with buprenorphine, which, considering concerns about long-term adverse effects of pre- and postnatal opioid exposure, is expected to improve clinical outcomes. Various PBPK models of buprenorphine have been developed, but these models could not describe sublingual drug absorption²⁰ or were not validated in neonatal populations.^21,22 Therefore, this study aimed to develop a neonatal PBPK model of sublingual buprenorphine to identify the most likely underlying sources of PK variability. The role of biliary clearance and its possible maturation was specifically examined, as this could be an essential elimination route in neonates with underdeveloped buprenorphine metabolizing enzymes.

Methods

Adult PBPK model development

A buprenorphine PBPK model was developed and evaluated using Simcyp (v19.1, Simcyp Limited, Sheffield, UK). The workflow for adult and neonatal PBPK model development is shown in Figure 1. The present PBPK model was based on a model developed by Johnson et al.,²² which was expanded by incorporating in vitro enzyme kinetics data (K_m and V_max) for CYP2C8, UGT1A3, UGT2B7, and UGT2B17, besides CYP3A4- and UGT1A1-mediated metabolism.^10,11 Drug physiochemical and physiological parameters used to construct the buprenorphine compound file are shown in Table 1.^{8,10,11,22-30} Incorporation of sublingual absorption into the PBPK model is described in the Supplementary Methods.

Figure 1 –

Schematic representation of the physiologically-based pharmacokinetic (PBPK) model development process. CYP, cytochrome P450; IV, intravenous; PK, pharmacokinetics; PNA, postnatal age at start of treatment; SL, sublingual; UGT, UDP-glucuronosyltransferase.

Table 1.

Input data for the buprenorphine PBPK model

Parameter	Value
Physiochemical
Molecular weight (g mol−1)	467.6
LogP	4.9823
Compound type	Ampholyte23
pKa (acid; phenol)	9.6223
pKa (base; amine)	8.3123
Blood binding
Blood-to-plasma ratio	124
fu, plasma	0.0425
Gastrointestinal tract absorption (first order model)
fa	1a
ka (h−1)	0.33b
fu, gut	1a
Qgut (L/h)	16.8c
Peff, man (10−4 cm/s)	6.83c
Caco-2 7.4:7.4 (10−6 cm/s)	66.726
Lungd absorption (first order model)
fa	1a
ka (h−1)	0.5e
Proportion of dose inhaledsublingual tablet	0.3f
Proportion of dose inhaledsublingual solution	0.6f
Distribution (minimal PBPK model)
kin (h−1)	0.6422
kout (h−1)	0.0722
Vsac (L/kg)	1.3322
Vss (L/kg)	2.724
Elimination
CYP2C8
Vmax (pmol/min per mg protein)	176.310
Km (μM)	12.410
CYP3A4
Vmax (pmol/min per mg protein)	52010
Km (μM)	13.610
UGT1A1
Vmax (pmol/min per mg protein)	287011
Km (μM)	66.411
UGT1A3
Vmax (pmol/min per mg protein)	28611
Km (μM)	20211
UGT2B7
Vmax (pmol/min per mg protein)	17311
Km (μM)	13.811
UGT2B17
Vmax (pmol/min per mg protein)	17211
Km (μM)	9.611
fu, mic	0.129
CLrenal (L/h)	0.54g
CLbiliary (μl/min per million hepatocytes)	5122

CL_biliary, biliary clearance; CL_renal, renal clearance; CYP, cytochrome P450; f_a, fraction absorbed; f_{u, gut}, fraction unbound in enterocytes; f_{u, mic}, fraction unbound in in vitro microsomal incubation; f_{u, plasma}, fraction unbound in blood plasma; k_a, first-order absorption rate constant; K_m, Michaelis-Menten constant; k_in and k_out, first-order rate constants describing the transfer of buprenorphine to a single adjusting compartment; PBPK, physiologically-based pharmacokinetic; P_{eff, man}, human jejunum effective permeability; Q_gut, nominal flow in gut model; UGT, UDP-glucuronosyltransferase; V_max, maximum metabolic rate; V_sac, volume of single adjusting compartment; V_ss, volume of distribution at steady state.

^aAssumed value.

^bEstimated by sensitivity analysis by Johnson et al.²² using pharmacokinetic (PK) data from the Suboxone^® new drug application (NDA).⁸

^cSimcyp predicted.

^dSublingual route of administration is not available in Simcyp; absorption across oral mucosa is therefore mimicked by employing the inhalation module.

^eEstimated by sensitivity analysis by Johnson et al.²² using PK data from McAleer et al.³⁰

^fProportion of dose inhaled reflects the percentage that is sublingually absorbed and roughly equals bioavailability (30% for tablets;²⁷ twice as high for solution)²⁸ due to extensive first-pass metabolism following oral administration.

^gCalculated by Johnson et al.²² based on mass balance study where 1% was excreted unchanged in urine,⁸ with total plasma clearance of 54.1 L/h.²⁴

Adult PBPK model validation

Following a comprehensive literature search of buprenorphine PK data in healthy adult volunteers, the predictive performance of the PBPK model following intravenous and sublingual administration was assessed by determining fold-differences between predicted and observed (P/O ratios) area under the curve from zero to infinity (AUC_0–∞), clearance (CL) or apparent clearance (CL/F), peak concentration (C_max), and, in case of sublingual administration, time to reach peak concentration (T_max). The noncompartmental analysis performed to calculate unreported PK parameter values, together with trial and virtual population construction are detailed in the Supplementary Methods.

Scaling to pediatric and neonatal populations

PK simulations in pediatric and neonatal populations were conducted using the Simcyp Pediatric simulator. Virtual neonates were assumed to be full-term infants and were redefined over time (i.e., physiological parameters were adjusted throughout the virtual trial to reflect growth and maturation during the trial). The effect of developmental changes in metabolizing enzyme expression was accounted for by incorporating ontogeny profiles defined by the Simcyp Pediatric simulator. The preset ontogeny profiles for all enzymes involved in buprenorphine metabolism were evaluated using human expression data in the literature and adjusted if needed (see details in the Supplementary Methods). As a result, the developmental profiles for UGT2B7 and CYP2C8 were modified according to recently published data by Bhatt et al.³¹ and Song et al.³² (Figure S1), respectively. Moreover, although the Simcyp Pediatric simulator’s default CYP3A4 developmental profile³³ was used for final analyses, the PBPK model’s predictability was additionally tested using the CYP3A4 ontogeny profile described by Upreti and Wahlstrom.³⁴ Final ontogeny profiles for all buprenorphine metabolizing enzymes are shown in Figure S2.

Maturation of biliary clearance

Development of biliary clearance was investigated following a manual parameter estimation approach³⁵ using buprenorphine blood concentration data from a prospective study in which neonates received sublingual buprenorphine solution as NOWS treatment.¹⁸ Buprenorphine concentrations were predicted in each patient (n = 19) using a realistic virtual neonatal population of the same postnatal age (PNA), sex, and weight, following virtual twin methodology.¹⁶ A total of 100 simulations (10 individuals × 10 trials) were run for each patient. To evaluate the effect of conceivable maturation, concentrations were predicted with 11 degrees of biliary clearance, increasing from 0 to 51 μl/min per million hepatocytes (the previously estimated adult extent).²² The ratio between predicted and observed whole blood concentrations (concentration fold-difference) was determined under each degree of biliary clearance. To capture saturable maturation of biliary clearance, a sigmoid E_max model was fitted through mean concentration fold-differences closest to 1 (i.e., no difference between predicted and observed buprenorphine concentrations) across 0.5 day PNA bins. Following development of a biliary clearance maturation profile, relative individual contributions of metabolic and elimination pathways to total buprenorphine clearance across age was determined.

Pediatric and neonatal PBPK model validation

Published buprenorphine PK data in pediatric populations were collected and used for PBPK model validation, similar to as described for adult PBPK model validation. In neonates, the predictive performance of the PBPK model was assessed with and without incorporating the defined biliary clearance developmental profile by determining concentration fold-differences using data from the aforementioned neonatal study,¹⁸ according to virtual twin methodology.¹⁶ The neonatal PBPK model was additionally evaluated externally by comparing predicted and observed plasma concentrations in preterm neonates following continuous intravenous buprenorphine administration. Due to sparse sampling in neonates, buprenorphine concentrations were used for PBPK model evaluation rather than PK parameters such as AUC_0–∞, CL, C_max, and T_max.

Sensitivity analysis

Sensitivity analyses were performed to investigate the effect of the fraction of the dose absorbed through oral mucosa (across a range of 10–99.9%), extent of biliary clearance (0–200%), cardiac output (50–150%), and hepatic buprenorphine metabolizing enzyme abundances (12.5–200%) on CL/F in different neonatal age cohorts (1 day, 1 week, and 1 month PNA). Likewise, the impact of hepatic enzyme abundances on the buprenorphine fraction escaping hepatic metabolism (F_h), as well as intestinal UGTs and CYP3A4 abundances on the fraction escaping gut metabolism (F_g), were investigated across a range of 12.5–200%. More details on the sensitivity analysis are provided in the Supplementary Methods.

Statistical analysis

The predictive performance of the PBPK model was deemed adequate if the geometric mean of PK parameter P/O ratios and concentration fold-differences fell between 0.5-fold and 2-fold. In addition, the fraction of concentration fold-differences falling within this 2-fold range was determined. The statistical analysis is discussed in further detail in the Supplementary Methods.

Results

Predictive performance in healthy volunteers following intravenous and sublingual administration

The developed adult PBPK model was first validated by determining P/O ratios of AUC_0–∞, CL, and C_max following intravenous administration of buprenorphine in healthy adult volunteers (Table S1). For intravenous adult PBPK model validation, 14 separate PK studies, including a total of 103 subjects and 123 concentration-time profiles, were used.^9,24,36-40 Age and buprenorphine dose ranged from 20 to 67.5 years and from 0.3 to 16 mg, respectively. The P/O ratios of AUC_0–∞, CL, and C_max fell within 0.5-fold and 2-fold in 14 (100%), 14 (100%), and 11 (78.6%) out of 14 PK studies, respectively. Geometric mean (95% CI) AUC_0–∞, CL, and C_max P/O ratios were 0.90 (0.78–1.05), 1.07 (0.91–1.26), and 0.69 (0.41–1.17), respectively, suggesting adequate predictive performance of AUC_0–∞, CL, and C_max following intravenous administration in healthy adult volunteers.

The adult PBPK model was subsequently validated by determining P/O ratios of AUC_0–∞, CL/F, C_max, and T_max following sublingual buprenorphine administration in healthy adult volunteers (Table 2). Fourteen separate PK studies, which included 151 concentration-time profiles of 127 subjects, were used for sublingual adult PBPK model validation.^{9,30,38,39,41-45} Age ranged from 19 to 66 years and doses from 0.4 to 4 mg as sublingual tablet or solution. AUC_0–∞, CL/F, C_max, and T_max P/O ratios fell within 0.5-fold and 2-fold in 11 (78.6%), 10 (71.4%), 14 (100%), and 12 (85.7%) out of 14 PK studies, respectively. Geometric mean (95% CI) AUC_0–∞, CL/F, C_max, and T_max P/O ratios were 1.00 (0.74–1.33), 1.00 (0.75–1.32), 1.04 (0.84–1.29), and 0.95 (0.72–1.26), respectively, indicating good predictability of all PK parameters following sublingual administration in healthy adult volunteers.

Table 2.

Predicted and observed buprenorphine pharmacokinetic parameters following sublingual administration in healthy adult volunteers

Clinical trial	Dose (mg)	Route of administration	n	Female (%)	Mean age [range] (years)		AUC0–∞ (ng⋅h/mL)	CL/F (L/h)	Cmax (ng/mL)	Tmax (h)
Bullingham et al.9	0.4	sublingual, tablet	5	60	64.2	Observed	4.54	88.1a	0.50	3.50
						Predicted	4.40	90.9	0.41	1.17
						P/O ratio	0.97	1.03	0.83	0.33
Bullingham et al.9	0.8	sublingual, tablet	5	60	66.0	Observed	8.06	99.3a	1.04	3.20
						Predicted	5.97	134.0	0.69	1.29
						P/O ratio	0.74	1.35	0.66	0.40
McAleer et al.30	2	sublingual, tablet	36	0	(19–42)	Observed	9.33b	214.4b	1.60	1.50
						Predicted	10.2	210.3	1.24	1.34
						P/O ratio	1.10	0.98	0.78	0.89
Dong et al.41	2	sublingual, tablet	8	51.2	27.7	Observed	10.9	213.0	1.65	1.00
						Predicted	10.7	201.9	1.34	1.32
						P/O ratio	0.98	0.95	0.81	1.32
Harris et al.42	4	sublingual, tablet	8	12.5	(22–42)	Observed	18.6b	215.6b	1.84	1.06
						Predicted	20.6	208.8	2.50	1.33
						P/O ratio	1.11	0.97	1.36	1.25
Ciraulo et al.43	4	sublingual, tablet	23	30.4	34.5	Observed	9.73b	410.9b	2.00	1.09
						Predicted	21.5	200.6	2.60	1.34
						P/O ratio	2.21	0.49	1.30	1.23
Walsh et al.44	2	sublingual, solution	4	0	32.3 (28–41)	Observed	40.3b	49.6b	4.17b	1.00b
						Predicted	18.5	115.7	2.27	1.34
						P/O ratio	0.46	2.33	0.54	1.34
Mendelson et al.38	2	sublingual (3 min hold), solution	6	16.7	29 (21–38)	Observed	14.3	139.9a	1.60	1.25
						Predicted	10.4	208.7	2.28	1.32
						P/O ratio	0.72	1.49	1.43	1.06
Mendelson et al.38	2	sublingual (5 min hold), solution	6	16.7	29 (21–38)	Observed	13.2	151.5a	1.72	1.62
						Predicted	10.4	208.9	2.28	1.32
						P/O ratio	0.78	1.38	1.33	0.81
Schuh & Johanson45	2	sublingual, solution, m.d.	14	21.4	40 (20–50)	Observed	19.6b	126.3b	2.05b	1.93b
						Predicted	27.4	72.9	2.60	1.31
						P/O ratio	1.40	0.58	1.27	0.68
Walsh et al.44	4	sublingual, solution	4	0	32.3 (28–41)	Observed	104.8b	38.2b	6.73b	1.00b
						Predicted	37.1	115.6	4.53	1.33
						P/O ratio	0.35	3.03	0.67	1.33
Kuhlman et al.39	4	sublingual, solution	6	0	34.4 (27-40)	Observed	23.9	210.4	3.31	0.71
						Predicted	36.9	116.1	4.52	1.33
						P/O ratio	1.55	0.55	1.37	1.87
Schuh & Johanson45	4	sublingual, solution, m.d.	14	21.4	40 (20–50)	Observed	32.0b	153.4b	2.74b	1.50b
						Predicted	54.9	72.8	5.21	1.31
						P/O ratio	1.72	0.47	1.90	0.87
Harris et al.42	4	sublingual, solution	12	16.7	(23–34)	Observed	24.6b	162.5b	3.56	1.09
						Predicted	36.3	118.0	4.56	1.32
						P/O ratio	1.47	0.73	1.28	1.21
						Geo. meanc (95% CI)	1.00 (0.74–1.33)	1.00 (0.75–1.32)	1.04 (0.84–1.29)	0.95 (0.72–1.26)

AUC_0–∞, area under the curve from zero to infinity; CI, confidence interval; CL/F, apparent clearance; C_max, peak concentration; m.d., multiple doses; P/O ratio, fold-difference between predicted and observed values; T_max, time to reach peak concentration.

^aCalculated following CL/F = Dose/AUC_0–∞.

^bPharmacokinetic parameter manually calculated through noncompartmental analysis.

^cGeometric mean of P/O ratios.

Predictive performance in pediatric subjects following intravenous administration

Pediatric PBPK model-based predicted PK parameters were subsequently compared to those reported in the literature. One clinical trial investigating buprenorphine PK in pediatric patients was identified.⁴⁶ In this study, 10 children aged 4.6 to 7.5 years received buprenorphine intravenously at a dose of 3 μg/kg. Predicted and observed AUC_0–∞ and CL were 1.37 and 0.82 ng⋅h/mL (P/O ratio = 1.66) and 46.2 and 78.0 L/h (P/O ratio = 0.59), respectively, and therefore met the PBPK model acceptance criteria. C_max was not reported in this study. No other clinical PK studies were identified to test the model’s predictability in pediatric subjects.

Biliary clearance maturation profile

A biliary clearance maturation profile was estimated using clinical PK data obtained from neonates with realistic PBPK model-based predictions following virtual twin methodology. A total of 52 buprenorphine whole blood concentrations obtained from 19 neonates (0.49–5.47 days PNA at time of blood sampling) were available to determine concentration fold-differences.¹⁸ Mean concentration fold-differences across 0–5.5 days PNA and biliary clearance of 0–51 μL/min per million hepatocytes are shown in Figure 2A. The sublingual PBPK model with minimal biliary clearance (0 μl/min per million hepatocytes) yielded concentration fold-differences closest to 1 in neonates younger than 3 days PNA. Blood concentrations of older neonates were best predicted with higher degrees of biliary clearance, approaching the adult extent at roughly 4 days PNA. A sigmoid E_max model with F_birth, Adult_max, Age₅₀, and n set to 0.00, 0.91, 3.09 days, and 20.1, respectively, reasonably captured fold-differences closest to 1 across PNA (Figure 2B), thereby outlining a developmental profile of biliary clearance. Relative individual contributions of metabolic and elimination pathways to total clearance across age are shown in Figure S3.

Figure 2 –

(a) Concentration fold-differences across postnatal age (PNA; columns) and 11 increasing degrees of biliary clearance (rows), where a biliary clearance of 51 μl/min per million hepatocytes equals the adult level. Concentration fold-differences close to 1 (i.e., no difference between predicted and observed buprenorphine concentrations) are denoted by green shaded cells, whereas yellow, orange, red, and violet shaded cells indicate an increasing bidirectional deviation from 1. The number of concentration fold-differences (observations) per PNA age bin is shown underneath each column. Concentration fold-differences closest to 1 are bolded and underlined per column. (b) Fraction of adult biliary clearance across PNA. Open circles represent the extent of biliary clearance as a fraction of the adult level attaining a concentration fold-difference closest to 1 per age bin, where the size of the circle corresponds to the number of observations per bin. Solid line represents the sigmoid E_max model that best captures the developmental trajectory of biliary clearance.

Predictive performance in neonates following sublingual and intravenous administration

After maturation of biliary clearance was outlined, the PBPK model’s predictive performance was evaluated in neonates with and without considering development of biliary clearance. Fold-differences between the observed concentration and the PBPK model-based prediction across PNA are shown in Figure 3. Assuming no developmental trajectory of biliary clearance, lower concentration fold-differences were observed in younger neonates, indicating systemic bias in the model prediction (Figure 3A). By including the developed biliary clearance maturation profile, concentration fold-differences were consistent across 0–5.5 days PNA (Figure 3B). The geometric mean (95% CI) concentration fold-difference was 0.56 (0.47–0.66) without biliary clearance maturation, while inclusion of biliary clearance development resulted in a significantly greater geometric mean (95% CI) concentration fold-difference of 0.75 (0.64–0.87, p < 0.001). In addition, a higher proportion of concentration fold-differences fell between 0.5-fold and 2-fold (69.2% vs. 55.8%, p = 0.078), indicating that incorporating biliary clearance maturation improves the PBPK model’s accuracy. Due to the young age of the neonatal study population, using the CYP3A4 ontogeny profile described by Upreti and Wahlstrom³⁴ instead of the Simcyp Pediatric simulator’s default³³ yielded nearly identical results (Figure S4). PBPK model-predicted and observed concentration-time profiles for the youngest and oldest neonates of the study cohort, and two newborns in between, are shown in Figure 4. Figure S5 shows these profiles overlayed with PBPK model-predicted concentration-time profiles in the scenarios that no maturation of biliary clearance is assumed and the CYP3A4 ontogeny profile described by Upreti and Wahlstrom³⁴ is used.

Figure 3 –

Buprenorphine concentration fold-differences (ratio between predicted and observed buprenorphine concentrations; open circles) across postnatal age (PNA). Dotted lines represent linear regression, with Spearman’s correlation coefficients denoted by r_s in the left upper corners of the lower panels. Blue shaded area indicates a 2-fold difference between predicted and observed concentrations. (a) Assuming no developmental trajectory of biliary clearance, represented by the horizontal solid line intercepting the y-axis at 1 in the upper panel, the concentration fold-differences are weakly correlated with PNA. (b) Incorporating the outlined biliary clearance maturation profile of Figure 2, represented by the sigmoid curve in the upper panel, results in no correlation with PNA.

Figure 4 –

Physiologically-based pharmacokinetic (PBPK) model-predicted and observed concentration-time profiles following sublingual administration of buprenorphine in neonates. Solid line and blue shaded area represent the mean and 5th to 95th percentile range of the virtual twin population (n = 100), respectively. Open circles represent observed buprenorphine concentration-time points obtained from a recent clinical trial.¹⁸ Concentration-time profiles of the (a) youngest and (b) oldest neonate of the study cohort, and (c, d) two newborns in between, are shown. Corresponding postnatal age (PNA) at the start of the treatment is shown in the upper left corners of the panels.

The neonatal PBPK model incorporating maturation of biliary clearance was externally evaluated using data from a clinical trial in which 12 preterm newborns aged 1 to 27 days PNA received intravenous buprenorphine.⁴⁷ Predicted and observed concentration-time profiles for all individual neonates are shown in Figure S6. Observed concentrations were highly variable. PBPK model-predicted concentration-time profiles were similar to observed concentration-time profiles in roughly half of the study cohort; the other concentration-time profiles were not well captured by the PBPK model, indicating the presence of unexplained variability not accounted for in the PBPK model.

Sensitivity analysis

The PBPK model-predicted buprenorphine CL/F was sensitive to changes in the fraction of dose absorbed through oral mucosa (Figure 5A), extent of biliary clearance (Figure 5B), and cardiac output (Figure 5C) in virtual neonates aged 1 day, 1 week, and 1 month PNA. The relative change in CL/F emerging from modifying the fraction of dose absorbed through oral mucosa was similar across the three age groups, whereas changing the extent of biliary clearance had the most profound effect in neonates of 1 day PNA. Conversely, the relative change in CL/F caused by altering the degree of cardiac output was greatest in neonates of 1 month PNA. Moreover, the PBPK model-predicted CL/F was sensitive to CYP3A4 abundance in neonates aged 1 day PNA (Figure 5D), but its impact diminished in neonates of 1 week PNA (Figure 5E) and 1 month PNA (Figure 5F). The predicted CL/F was not sensitive to changes in CYP2C8, UGT1A1, UGT1A3, UGT2B7, and UGT2B17 abundances in neonates aged 1 day, 1 week, and 1 month PNA. PBPK model-based simulations indicated F_h to be 0.859, 0.651, and 0.567 in neonates aged 1 day, 1 week, and 1 month PNA, respectively. F_g was estimated to be consistent and considerably lower, equaling 0.037, 0.038, and 0.041 in neonates aged 1 day, 1 week, and 1 month PNA, respectively. F_h was moderately sensitive to hepatic CYP3A4 abundance in neonates aged 1 day PNA, where its impact diminished in older neonates. In contrast, F_g was sensitive to abundances of intestinal UGTs across all neonatal ages (Figure S7).

Figure 5 –

Sensitivity analysis of the physiologically-based pharmacokinetic (PBPK) model in virtual neonates. Upper row shows relative change in apparent clearance (CL/F) across (a) fraction absorbed through oral mucosa, (b) biliary clearance, and (c) cardiac output in neonates of 1 day postnatal age (PNA; closed circles with dotted lines), 1 week PNA (closed squares with dashed lines), and 1 month PNA (closed triangles with solid lines). Change is relative to the values defined in the PBPK model (i.e., 60%, 51 μL/min per million hepatocytes, and 100% for fraction absorbed through oral mucosa, biliary clearance, and cardiac output, respectively), where intersecting horizontal and vertical solid narrow lines indicate the origin. Lower row shows relative change in CL/F across increasing abundances of cytochrome P450 (CYP) 3A4, CYP2C8, UDP-glucuronosyltransferase (UGT) 1A1, UGT1A3, UGT2B7, and UGT2B17 in neonates aged (d) 1 day PNA, (e) 1 week PNA, and (f) 1 month PNA. Change is relative to an abundance of 100% (i.e., at full maturation 137, 24, 48, 23, 71, and 5.9 pmol/mg protein for CYP3A4, CYP2C8, UGT1A1, UGT1A3, UGT2B7, and UGT2B17, respectively), where intersecting horizontal and vertical solid narrow lines indicate the origin.

Discussion

In this study, we successfully developed a neonatal PBPK model of sublingual buprenorphine to investigate the effect of maturation and patient characteristics on buprenorphine PK in neonates. The model included enzyme-mediated metabolism by CYP3A4, CYP2C8, UGT1A1, UGT1A3, UGT2B7, UGT2B17, with additional biliary excretion. The Simcyp Pediatric simulator’s ontogeny profiles for CYP2C8 and UGT2B7 were updated according to recent literature.^31,32 The PBPK model was first thoroughly vetted using published data obtained from adult and pediatric subjects receiving buprenorphine intravenously. Subsequently, the model’s sublingual absorption component was extensively validated using 14 PK studies in which 127 healthy adult volunteers received buprenorphine as sublingual tablet or solution. Geometric mean P/O ratios of AUC_0–∞, CL/F, C_max, and T_max approached, or indeed reached, unity, indicating good PBPK model performance. Following verification, the predictability was evaluated in 19 neonates treated with sublingual buprenorphine solution. Through the incorporation of biliary clearance maturation, PBPK model-predicted buprenorphine concentrations were unbiased across PNA, with 69.2% falling within 2-fold of the observed value, attaining a geometric mean (95% CI) concentration fold-difference of 0.75 (0.64–0.87), thereby demonstrating reasonable predictability of the central tendency of buprenorphine concentrations in neonates following sublingual administration.

This is the first study to outline a developmental profile of buprenorphine biliary clearance. Our results indicate that biliary clearance is minimal at birth, but steeply rises in the first days of life, approaching the adult extent at approximately 4 days PNA. Johnson et al. have described similarly rapid developmental trajectories of biliary elimination of azithromycin, ceftriaxone, and digoxin, for which maturation occurs at birth or soon thereafter.²² Through population PK modeling of buprenorphine, Ng et al. reported an immediate increase in body weight-adjusted total clearance after birth, reaching 90% of the adult value at 10 days PNA.¹⁷ This increase may be in part explained by the onset of biliary elimination as described in our study.

An age-dependent effect on bioavailability could have influenced the outline of biliary clearance maturation. However, given the consistently minimal F_g (0.037–0.041 across three neonatal age groups), bioavailability of swallowed buprenorphine is expected to be negligible in all neonates. Moreover, the extent of sublingual buprenorphine absorption is thought to be consistent among neonates, and may be similar to that in adults, as no data indicate transmucosal drug absorption is significantly different between infants and adults.⁴⁸ Therefore, we believe overpredicted buprenorphine biliary clearance, rather than underpredicted total bioavailability, explains underprediction of buprenorphine blood concentrations in young neonates (Figure 3A), which is corrected when using the outlined biliary clearance maturation profile (Figure 3B).

Although the developmental mechanism underlying buprenorphine biliary clearance remains unclear, various factors could help explain its observed maturation. In the fetus, hepatobiliary function is immature.⁴⁹ Postnatally, bile flow is significantly lower at birth than at one week PNA, as demonstrated in preterm and term non-human primates.⁵⁰ Moreover, enteral feeding is known to stimulate bile flow.⁵¹ Lastly, hepatic active transport mechanisms may be underdeveloped in the first days of life, but given that no sinusoidal or canalicular transporters of buprenorphine have been identified,¹⁴ it is difficult to test the validity of this assumption. Therefore, we postulate buprenorphine biliary clearance development could be driven by a postnatal rise in bile flow and therefore hepatobiliary function, possibly combined with maturation of an unknown hepatic active transport process.

Sensitivity analysis showed that PBPK model-predicted buprenorphine CL/F was sensitive to changes in the fraction of the dose absorbed through oral mucosa. Sublingual buprenorphine bioavailability is highly variable between subjects, as demonstrated by Bullingham et al., who reported bioavailability ranging between 15.7 and 94.4%.⁹ Therefore, variation in the fraction of the dose absorbed through oral mucosa seems to be an essential driver of PK variability and a better understanding of factors influencing transmucosal absorption could improve the PBPK model’s predictive performance. We hypothesize that sublingual bioavailability is significantly greater in sleeping neonates than in those awake, since newborns may be less likely to swallow the dose.⁵² The awake or sleep state during dosing was not documented in the clinical trial¹⁸ that provided the data for the present study, but future studies logging this could help elucidate the effect it might have on sublingual absorption, thereby refining the current PBPK model.

Sensitivity analysis additionally showed that the extent of biliary clearance, cardiac output, and CYP3A4 abundance influence PBPK model-predicted buprenorphine CL/F. The effect of cardiac output, and thus hepatic blood flow, was greatest in neonates aged 1 month PNA, which is in line with increased hepatic intrinsic clearance as metabolizing enzymes mature. Conversely, the impact of biliary clearance and CYP3A4 abundance was most significant in neonates of 1 day PNA, which reflects age-dependent relative contributions of metabolic and elimination pathways as shown in Figure S3. Although not yet clinically confirmed, CYP3A4 ontogeny and genotype could therefore be essential drivers of buprenorphine PK variability in the first days of life, which is the target population for NOWS. CYP3A4 activity shows substantial interindividual variability that is not well explained by genetic variation only.⁵³ Future studies could utilize relatively noninvasive phenotyping probes to determine CYP3A4 activity in neonates, such as the urinary 6β-hydroxycortisol (6β-OHF) to cortisol ratio,⁵⁴ since cortisol undergoes CYP3A4-mediated metabolism to 6β-OHF. Subsequent incorporation of individual CYP3A4 activity in the PBPK model could enhance the predictive performance of the PBPK model.

The present neonatal PBPK model has a few limitations. The developmental trajectory of biliary clearance proposed in this study is based on a limited number of observations and could not be externally validated given the lack of separate NOWS-diagnosed neonatal buprenorphine time-concentration data. Additional studies are warranted to verify the current results. Measuring buprenorphine and its metabolites in neonatal feces could further substantiate the maturation profile. Moreover, the neonatal buprenorphine blood-to-plasma ratio was not determined in the clinical study¹⁸ that provided the data for the present analysis, and was assumed to be equal to the adult blood-to-plasma ratio. However, different hematocrit, plasma protein binding, and red blood cell participation in neonates may affect the blood-to-plasma ratio, and could therefore influence PBPK model-based simulations. In addition, F_g values reported in this study are influenced by the enzyme abundance per microsomal protein per intestine (MPPI), which is calculated in the Simcyp Pediatric simulator. Simcyp is currently refining the MPPI algorithm. Estimated values for F_g might therefore slightly change in future versions of Simcyp. Furthermore, the PBPK model-based simulations in this study did not take the neonate’s gestational age (GA) into account, as the module used to mimic sublingual administration was not supported for preterm populations in the Simcyp Pediatric simulator. Buprenorphine blood concentrations used for neonatal PBPK model development were obtained exclusively from term and moderate to late preterm (32–37 weeks GA) neonates.¹⁸ Therefore, the implication of not considering GA in model development is likely small. However, it could help explain why the external evaluation of the PBPK model using results reported by Barrett et al. was unsuccessful for some subjects, as all subjects in that clinical trial were very preterm neonates (27–32 weeks GA).⁴⁷ In addition, the indication for buprenorphine use in the study of Barrett et al.,⁴⁷ which was to achieve sedation under mechanical ventilation, was unlike that in the clinical trial of Mizuno et al.,¹⁸ namely NOWS symptom control. Lastly, Barrett et al. measured buprenorphine plasma concentrations through a radioimmunoassay,⁴⁷ whereas Mizuno et al. utilized high-performance liquid chromatography-tandem mass spectrometry,¹⁸ which hampers clear comparison between the determined buprenorphine concentrations. Collectively, this underscores that a detailed understanding of the patient’s clinical background is needed to optimize the PBPK model performance.

In conclusion, a neonatal PBPK model of sublingual buprenorphine was successfully developed and validated. Moving PBPK modeling forward towards application in personalized medicine requires a thorough understanding of factors driving PK in the individual subject. Neonatal PBPK model-based simulations identified the fraction of the dose absorbed through oral mucosa, CYP3A4 ontogeny and genotype, and, as novel source, the extent of biliary clearance as most likely underlying mechanisms of buprenorphine PK variability in newborns. Although the developed PBPK model does not account for all observed PK variability in neonates with NOWS, the central tendency of buprenorphine exposure across PNA was adequately captured. The present PBPK model can therefore guide the development of improved starting dose regimens of buprenorphine for the treatment of NOWS.

Study highlights.

What is the current knowledge on the topic?

Neonatal opioid withdrawal syndrome (NOWS), a major public health issue resulting from the opioid epidemic, is increasingly treated with sublingual buprenorphine. Physiologically-based pharmacokinetic (PBPK) modeling can fill in knowledge gaps concerning its variable pharmacokinetics (PK) in neonates.

What question did this study address?

Through the development of an extensively validated neonatal PBPK model of sublingual buprenorphine, this study investigated the effect of maturation and patient-specific characteristics on buprenorphine PK.

What does this study add to our knowledge?

This study demonstrated that the extent of biliary clearance, proportion of the dose absorbed through oral mucosa, and CYP3A4 abundance are likely drivers of buprenorphine PK variability. Inclusion of a newly outlined biliary clearance maturation profile significantly improved the PBPK model’s predictability in neonates early after birth, signaling rapid development in the first days of life.

How might this change clinical pharmacology or translational science?

The developed neonatal PBPK model enables investigation of buprenorphine PK in newborns, and can be used to support development of improved starting dose regimens for the treatment of neonates with NOWS.