Oral drug absorption in pediatrics: the intestinal wall, its developmental changes and current tools for predictions.

The dissolution, intestinal absorption and presystemic metabolism of a drug depend on its physicochemical characteristics but also on numerous physiological (e.g. gastrointestinal pH, volume, transit time, morphology) and biochemical factors (e.g. luminal enzymes and flora, intestinal wall enzymes and transporters). Over the past decade, evidence has accumulated indicating that these factors may differ in children and adults resulting in age-related changes in drug exposure and drug response. Thus, drug dosage may require adjustment for the pediatric population to ensure the desired therapeutic outcome and to avoid side-effects. Although tremendous progress has been made in understanding the effects of age on intestinal physiology and function, significant knowledge gaps remain. Studying and predicting pharmacokinetics in pediatric patients remains challenging due to ethical concerns associated with clinical trials in this vulnerable population, and because of the paucity of predictive in vitro and in vivo animal assays. This review details the current knowledge related to developmental changes determining intestinal drug absorption and pre-systemic metabolism. Supporting experimental approaches as well as physiologically based pharmacokinetic modeling are also discussed together with their limitations and challenges.

Introduction

The number of reports discussing drug pharmacokinetics in pediatrics has grown substantially. More drugs have been studied in children in the past decade than in the preceding 50 years 1. This growing interest has roots in the emerging evidence of developmental changes in drug pharmacokinetics and pharmacodynamics, in the increasing regulatory pressure, and in the overwhelming medical need for safer and more effective pediatric medicines.

Until recently, drugs were developed and approved for use in adults only. Drug labels were silent on pediatric use and generally ignored potential age‐dependent effects on drug disposition and drug response 2. The treatment of children often employed unlicensed or off‐label drugs 3, with a dosage formulation not tailored for pediatrics 4. Drug dosages were based on data obtained in adults, after a simple adjustment for body weight. Such an approach has limited value as many physiological processes do not develop linearly with body size and may show dramatic age‐related differences. Occasionally, the lack of pediatric data led to sub‐therapeutic outcomes or unanticipated side‐effects. The gray baby syndrome in neonates receiving chloramphenicol is an exemplary illustration of such unwanted toxicity due to unadjusted dosing 5.

For years, the paucity of pediatric clinical trials and the very few drugs licensed for pediatric use contrasted with the considerable number of medicines prescribed, averaging 0.8 to 3.2 per child per year 3. Meanwhile, the World Health Organization has persistently drawn attention to the millions of childhood deaths annually in low income countries, as a result of untreated diseases. Despite their unequivocal value, investigations of pediatric medicines are hampered by numerous barriers. These include the ethical and technical issues limiting the access to pharmacokinetic studies in healthy pediatric subjects, the scarce availability of high quality tissue samples for in vitro assays, the poor translatability of animal data and the questionable financial return on investment due to the smaller size of the pediatric patient population 6, 7.

In response to the above challenges, the US Food and Drug Administration (FDA) and the European Medicines Agency (EMA) pressed the pharmaceutical industry to generate pediatric data for new and established medications (2002 Best Pharmaceuticals for Children Act and 2003 Pediatric Research Equity Act in US; 2007, European pediatric Regulation in EU) 7, 8. In the absence of a specific waiver, the sponsor must submit a pediatric development plan at the same time as conducting studies supporting adult use. Also, to incentivize pharmaceutical companies, patent extension and market exclusivity are offered for drugs with approved pediatric use. Finally, private and public funding was assigned for research on the treatment of pediatric patients using off‐patent drugs 9. Eventually, all these concerted efforts from regulatory authorities, the pharmaceutical companies and health care providers has started to show benefit and has generated pediatric data for hundreds of drugs 1.

Most medicines delivered to children are administered orally. Intestinal absorption is influenced by the drug characteristics (e.g. physicochemical properties) but also by physiological parameters (e.g. gastrointestinal fluid composition and volume, transit time, microbiota, drug metabolizing enzymes and drug transporters) and environmental factors (e.g. food, drug formulation). It is now recognized that these factors may vary considerably with age and result in age‐related changes in drug absorption. The focus of the present review is to summarize the age‐related changes in the physiological processes, drug metabolizing enzymes and drug transporters determining intestinal drug absorption and presystemic metabolism. The review is complementary to recent publications in this area 10, 11, 12, 13, 14 and intends to focus specifically on case reports, knowledge gaps and discrepancies, and to discuss the available experimental approaches (ex vivo, animal studies, clinical studies, physiologically based pharmacokinetics PBPK) together with their limitations.

Age‐mediated changes in physiological processes determining drug intestinal absorption

pH

Gastric pH is important for drug stability, drug dissolution and drug ionization, all of which affect the absorption process. At birth, gastric pH is neutral (pH 6–8), presumably as a result of amniotic fluid in the stomach 15. There is some controversy over the variations of fasting gastric pH during childhood. Review articles often report a decline to pH 2–3 within hours of birth, a rise to neutral pH after 24–72 hours followed by a progressive decrease (over several weeks to several years) to finally reach acidic values similar to those measured in adults 6, 13, 16, 17, 18, 19. Contrasting with these reviews, a comprehensive examination of original reports shows that gastric pH is continuously maintained around acidic values (< 3) from preterm neonates to adults (Figure 1), a finding already reported elsewhere 14, 20.

Figure 1

Age‐related changes in gastric pH under fasting conditions. Error bars indicate pH range (solid symbols) or standard deviation associated with mean pH (open symbols). References: ▼ 203, ● 204, ♦ 205, ■ 206, ▲ 207, ○ 208, □ 209, ◊ 25 (this latter reference reports pH data for subjects aged 2 months to 18 years)

The supposed transient neutral gastric pH in neonates is regularly discussed for its potential impact on drug disposition, especially for its positive effect on the oral bioavailability of acid‐labile drugs (e.g. ampicillin, penicillin or nafcillin) 17. A higher pH would also favor the ionization, and thus reduce the absorption of weak acids as possibly observed with phenytoin 21 or nalidixic acid 22. The higher gastric pH in neonates might also negatively impact on the absorption of weak bases by decreasing their solubility, as observed with ketoconazole 23. Overall, a clear consensus is still missing concerning the ontogeny of gastric pH and its impact on drug absorption.

Gastric volume

In addition to gastric pH, gastric volume is a key determinant of oral drug solubility. According to the biopharmaceutics classification system (BCS), highly soluble drugs in adults are those where the maximum dose unit is soluble in 250 ml aqueous liquid, corresponding to the initial gastric volume 24. The drug solubility normalized to the dose and to the gastric volume is given by the Dose number D 0. When normalized per kg body weight, the fasted gastric content volume is broadly similar in children (0.25–0.56 ml/kg, 25, 26, 27) and adults (0.36–0.50 ml/kg, 28, 29) while the drug dosage may differ. As a result, some drugs have a BCS solubility class that changes with age. Chloramphenicol, erythromycin, cefalexin and prednisolone gradually alter from a low solubility classification in 6‐month old children (D 0 of 1.3, 2.1, 2.3 and 3.8) to high solubility in adults (D 0 ≤ 1) 30. These findings should be taken cautiously as many challenges may confound pediatric BCS consideration, including the lack of a standardized fluid volume ingested by pediatric patients, the limited literature data across pediatric subpopulations and the knowledge gaps for some pediatric physiological parameters 31.

Gastric emptying and intestinal transit

Gastric emptying (GE) and intestinal motility influence the rate and extent of intestinal drug absorption. The GE rate is affected by numerous factors including feeding status, food composition, posture, activity, appetite, psychomotor development, disease states and the method for measuring GE (see 32 for a brief review). Gestational/postnatal age has been also proposed as affecting GE. In adults, the GE of liquid is biphasic with an initial rapid phase (< 10 min) followed by an slower phase 33. Review articles often report that GE is much slower and linear below the age of 6–8 months, as a result of immature motility neuro‐regulation 19, 34. A scintigraphic study in newborns and young infants (aged 1–10 weeks) given milk demonstrated a monophasic GE with a half‐life of 45–141 min 35. An ultrasound study in newborns showed age‐dependent changes in the preprandrial GE half‐time (from 72 min in preterm newborns of 28–32 weeks to 83 min in full‐term newborns) 36. Prolonged GE in the pediatric population is anticipated to decrease the absorption rate and delay the absorption of drugs where GE is rate limiting. This appears to be the case for paracetamol which shows a prolonged absorption half‐life in children under 3 months 37 or phenobarbital, sulfonamides and digoxin which all have an absorption rate that increases from 3 weeks to 1 year of age 38. For similar reasons, cisapride has its plasma peak concentration delayed in neonates when compared with older children 39. The assertion that GE changes during childhood has been challenged in a recent meta‐analysis of 49 published clinical studies (1457 individuals aged from 28 weeks of gestation to adults) 40. Age was found not to be a significant covariate for GE. However, the authors acknowledged some limitations to the analysis, e.g. the large diversity of the study protocols, and the fact that age and meal type might have been partially confounded. Overall, the effect of age on GE and the consequences on drug disposition remain poorly understood. As concluded by the authors of the meta‐analysis, a well‐controlled prospective scintigraphy study measuring GE across a wide range of age is warranted.

The small intestinal and colonic transit time may vary depending of the methodology used. However, within the same method, transit time for the small intestine and colon appears to be minimally affected by age 41, 42.

Biliary function

The bile acid pool size and bile flow in newborns are decreased compared with older children and adults 43, 44. In addition, the ileal reabsorption of bile salts remains immature until several months of age 45. These properties contribute to lower duodenal bile acid concentrations 46. Bile salts are critical to the intestinal absorption of lipid‐soluble xenobiotics by facilitating their solubility and dissolution rate. Although direct clinical evidence is scarce, these compounds are expected to show impaired intestinal absorption in younger age groups. It is tentatively assumed that the above mechanism might account for the 2‐fold decrease in the oral bioavailability of the water insoluble antiviral pleconaril in neonates when compared with older children 47.

Pancreatic enzymes

The pancreatic exocrine function is another key determinant of intestinal absorption that changes with age. Pancreatic enzyme levels (e.g. lipase, trypsin) are typically low at birth and develop during the first year of life 48, 49, 50. This might have an impact on the disposition of drugs sensitive to hydrolysis, as exemplified by prodrugs. Decreased intra‐duodenal hydrolysis was thought to be at least partially responsible for the incomplete, prolonged and erratic absorption of chloramphenicol palmitate after oral dosing to neonates 51.

Intestinal villous morphology and surface area

The villous pattern of the small intestine begins at 8 weeks of gestation 52. Crypt fission peaks at 6–12 months of age 53. Young children tend to show short ridge‐shaped villi and longer crypts 54. The typical adult profile with elongated finger‐ or leaf‐shaped villi appears after 3 years of age 55. Together with a reduced number of villi, these differences translate into a reduced surface‐to‐volume ratio in children when compared with adults 56. A retrospective population pharmacokinetic analysis in children and adolescents (aged 3 months to 18 years) with epilepsy revealed that the absorption rate constant of levetiracetam (LEV) increased over 3‐fold within the investigated age range 57. This finding was found difficult to interpret as LEV is a BCS class 1 drug (high solubility, high permeability) with restricted absorption unexpected. However, considering its physiochemical properties (LogD of −0.45, MW of 170), LEV might not easily dissolve in the lipid bilayer of the intestinal cell membranes and might be partly absorbed by the paracellular pathway, as demonstrated for other hydrophilic small molecular weight compounds 58, 59. To be fully absorbed, these compounds diffuse further down the length of the villi and require a larger absorptive surface area 60, 61. Thus, it could be tentatively hypothesized that the immature intestinal villi in children, with a lower surface/volume ratio and a reduced absorptive surface, may influence LEV absorption.

Intestinal permeability

The potential age‐related changes in the intestinal paracellular and transcellular permeability of drugs have been largely ignored. The polar hydrophilic antidiuretic furosemide was described to be absorbed, at least in part, by the paracellular route 62. In adults, its absorption is incomplete and its oral bioavailability is variable and rather low (mean value of ca. 56%) 63. Data about the pharmacokinetics of furosemide in very young children are scarce. Mirochnick et al. 64 reported an average oral bioavailability of 84% when furosemide is delivered to children of 39 weeks gestational age. Whether this higher bioavailability originates from higher intestinal paracellular permeability remains to be confirmed as other explanations cannot be excluded 65.

Gastrointestinal lumen and microbiota

At birth, the intestinal tract is generally sterile. Microbial colonization occurs within a few hours 66 with a microbiota composition that varies and evolves depending upon the mode of birth delivery (vaginal versus cesarean) 67, the first feeding (breast‐milk vs infant formula) 68, 69, and many other environmental factors (e.g. hospitalization, diet, exposure to antibiotics and environmental toxicants) as listed in recent reviews 70, 71, 72. Both gestational and postnatal age also have significant effects on the intestinal microbiota. Penders et al. demonstrated a higher Clostridium difficile colonization rate in preterm infants when compared with healthy full term infants 73. This finding is likely to be related to the use of antibiotics in neonatal intensive care units, rather than to maternal transmission 74. Koening et al. reported a gradual change in the diversity of the intestinal microbiota with four discrete steps that evolve up to 3 years of age where an adult‐like profile is achieved 75. Each step corresponds to a change in the children's diet (e.g. breast milk, formula feeding, adult diet), which coincides with a modification of the relative abundance of the bacteria phyla (i.e. firmicutes, bacteroidetes, actino‐ and proteobacteria). Other authors reported that adult characteristics were reached after the first year of life 66.

The intestinal microbiota has various important functions, some with impacts on drug disposition, e.g. regulation of intestinal barrier development, bile salt metabolism, gastrointestinal motility and drug metabolism. The impact of gut microbiota on drug metabolism is largely overlooked despite being recognized since the 1960s 76 and potentially affecting the disposition of up to one third of drug candidates 77, 78. The effects of gut flora ontogeny on drug metabolism are even less well‐documented beyond the often cited digoxin case example. After oral dosing to adults, digoxin is partially eliminated as reduced metabolites containing a saturated lactone ring (e.g. dihydrodigoxin). This reaction is primarily supported by anaerobic bacteria of the gut flora. Linday et al. 79 investigated the in vivo pharmacokinetics of digoxin in over 200 subjects, from 3.5 weeks of age to adults. Reduced metabolites were not detected below 8 months of age and the adult pattern was observed after 16 months (Figure 2).

Figure 2

Excretion of digoxin reduced metabolites at different ages. Digitalized children and adults were examined for metabolites recovered in urine samples. Subjects were classified as excretors when digoxin reduced metabolites accounted for at least 5% of the total drug‐related material. Adapted from 79

Intestinal wall drug metabolizing enzymes

Although less abundant than in the liver, cytochrome P‐450 enzymes (CYPs) are also expressed in the intestinal wall and can potentially decrease the oral bioavailability of a wide variety of drugs such as midazolam and cyclosporine 80. CYP3A4/5 accounts for 80% of the total immuno‐quantifiable intestinal CYPs, with its expression decreasing from the duodenum to the ileum 81.

The ontogeny of hepatic drug metabolizing enzymes has been thoroughly documented 82, 83, 84, 85. On the other hand, there are far fewer data on intestinal metabolizing enzyme ontogeny, probably as a result of the scarcity of quality pediatric intestinal tissue samples. Unfortunately, the ontogeny profile of drug metabolizing enzymes varies across organs 86, 87 and the developmental changes at the intestinal level cannot be extrapolated from another tissue. The precise mechanisms of this organ‐specific developmental regulation remain to be determined. In his pioneering and comprehensive work using biopsies and surgical specimens, Johnson et al. 88 demonstrated that duodenal CYP3A4 protein and activity change with age. Both endpoints gradually increased from barely detectable levels in the fetus to maximal levels by the age of approximately 2 years (Figure 3). A subsequent study investigated 59 duodenal biopsies from donors of 1 month to 17 years of age 89. As demonstrated by immunohistochemistry, CYP3A4 protein was detected in all enterocytes from 6 months of age. CYP3A4 and CYP3A5 mRNA were elevated during the first year of life and then gradually decreased. Of note, duodenal samples showed low levels of CYP3A7, the fetal liver CYP3A isoform. More recently, Chen et al. 86 investigated a total of 112 liver and intestinal samples from human donors (1–198 days of age). While the level of liver CYP3A4 mRNA was positively correlated with age, duodenal CYP3A4 mRNA showed a more complex and almost opposite pattern. Duodenal CYP3A4 mRNA expression tended to increase between day 1 to 96, and then gradually decreased with age. Overall, when compared with the 1–70 days age group, adults showed a 67% decrease in CYP3A4 mRNA. The same adult group showed, however, higher CYP3A4 protein level suggesting that a translational efficiency increases with age. However, these observed changes were not statistically significant and were hampered by a large inter‐individual variability. Overall, the above data indicate that intestinal CYP3A4 functional activity increases with postnatal age, a finding that cannot be predicted from mRNA measurement.

Figure 3

Age‐related changes in human intestinal CYP3A4 activity as measured by the rate of testosterone 6‐ß hydroxylation. Sample size given by the numbers in brackets.Reproduced from 88

Midazolam is extensively transformed by CYP3A4/5. In healthy subjects, about 50% of an oral dose is cleared by intestinal CYP3A4 metabolism, as demonstrated by comparing the kinetics after oral versus intravenous with and without co‐administration with CYP3A4 inhibitors 90, 91. De Wildt et al. 92 showed that the midazolam oral bioavailability was substantially higher in preterm infants (49%) when compared with the values reported in adults (24–38%) 93, 94. Similarly, the CL/F after oral dosing was nearly 10‐fold lower in infants than in older age groups 92. The authors concluded that these findings were due to the ontogeny of both liver and intestinal CYP3A4 activity. Beyond the often‐cited midazolam example, many other drugs are extensively transformed by intestinal CYP3A4 80, 95 and could potentially show ontogeny in their disposition. Sirolimus has a low oral bioavailability (14%) mostly linked to its extensive intestinal CYP3A4 metabolism (84% intestinal extraction) 96. Recent pharmacokinetic data from a multicenter phase II clinical trial in pediatric patients showed a clear effect of age on sirolimus metabolism 97. The formation rate of the two most abundant circulating metabolites, 16‐O‐demethylsirolimus and 24‐hydroxysirolimus, gradually increased with age. The median metabolite ratios in children (3–14 years) was about 2‐fold lower than in adults (25–48 years) which might be explained by immature intestinal CYP3A4. The HIV protease inhibitor saquinavir belongs to those drugs with over 80% of the oral dose cleared by CYP3A4‐mediated intestinal clearance 95. Grub et al. reported that the CL/F of an oral dose of saquinavir was higher in children (3–13 years) than in adolescents or adults. The authors tentatively assumed a higher intestinal clearance and/or lower bioavailability in children 98. However, this hypothesis is difficult to reconcile with the lower intestinal CYP3A4 activity reported in pediatrics, although it could be argued that the studies explored different age ranges. In adults, the anxiolytic buspirone has an oral bioavailability as low as 4% 99 which parallels its 79% CYP3A4 first‐pass intestinal extraction 95. A pharmacokinetic study in children 2–6 years old receiving a single oral dose of buspirone did not reveal any age‐related changes in CL/F 100, with values indistinguishable from those measured in adults 95.

As illustrated with all the above examples, the impact of intestinal CYP3A4 ontogeny on drug disposition is not yet fully understood and requires further investigations which should also explore potential confounding factors (e.g. pediatric formulations, feeding regimen, disease states).

Drug metabolism in the intestinal wall is not restricted to CYPs and involves other enzymes that are far less well documented with respect to maturation. Glutathione‐S‐transferase (GST) is expressed in the liver but also along the length of the small intestine. GST is responsible for the metabolic elimination of busulfan, an alkylating agent used in hematologic malignancies. Hassan first reported an age effect on busulfan pharmacokinetics after oral dosing. The exposure was found to be lower in children under 5 years of age when compared with older children and adults (2 to 3‐fold difference), even when the values were normalized for body surface or body weight 101 (Figure 4). Subsequent in vitro studies demonstrated that intestinal GSTA1 participates in the first‐pass extraction of bisulfan with the activity being up‐regulated in young children under 5 years of age when compared with older subjects 102.

Figure 4

Age‐related changes in busulfan plasma exposure of after oral dosing. C max values were normalized for the dose and body weight. Adapted from 101

Carboxylesterase‐2 (CES2) is highly expressed in the intestinal wall of adults, with lower levels in the liver, and the enzyme is known for contributing to the first‐pass metabolic hydrolysis of various drug substrates such as the ester prodrugs of the angiotensin receptor blockers candesartan and olmesartan 77. CES2 mRNA and protein were measured in duodenal samples from human donors of various ages 86. Both endpoints were found to increase with time to reach adult levels at 0.5–1 year of age. Despite the reported ontogeny of intestinal CES2, the prodrugs candesartan cilexetil and olmesartan medoxomil show the same pharmacokinetic profile after oral dosing to children when compared with adults 103, 104. Valacyclovir, an ester prodrug of acyclovir, is also claimed to be cleared by intestinal CES2 105. Surprisingly, its exposure after oral dosing consistently increases with age (from 2 to 18 years) 106, which contradicts CES2 ontogeny data. It could tentatively be assumed that valacyclovir hydrolysis is preferably supported by an enzyme distinct from CES2 107, with a different ontogeny pattern.

Phosphate prodrugs, designed to improve the absorption of water‐insoluble compounds, are based on a prior intestinal dephosphorylation step that is supported by the brush border enzyme alkaline phosphatase. The hydrolysis results in intraluminal supersaturation of the active moiety, which in turn favors the absorptive process. This approach has been used successfully with fosamprenavir, the phosphate prodrug of the water‐insoluble amprenavir 108. Of interest, a population pharmacokinetic analysis of fosamprenavir revealed an effect of age on body weight‐adjusted CL/F 109. The 1.8‐fold higher CL/F in the youngest children (2 month) compared with children >4 years possibly reflects a lower F because of immature intestinal alkaline phosphatase activity 110. The exact mechanisms, however, remain unclear; especially considering that the enzyme has multiple roles in the maintenance of the intestinal function (pH, absorption of lipids, modulation of gut flora) 111.

Glucuronidation enzymes (UGTs) are highly expressed in the intestinal wall and are responsible for the low oral bioavailability of some drugs such as raloxifene 112. While the ontogeny of hepatic UGTs has been documented 113, there is little published data about the intestinal counterparts. Court et al. investigated the mRNA levels of 19 UGT isoforms in various adult and fetal tissues including small intestine and colon 114. Unfortunately, developmental changes were discussed for liver and nasal tissues, not for intestine. The mRNA of UGT2B7, the enzyme primarily involved in the metabolism of zidovudine 115 and chloramphenicol 116, was found to be highly expressed in adult but not in fetal liver. Interestingly, the oral bioavailability of zidovudine is significantly higher in infants ≤14 days (89%) when compared with older children (61%) 117. The ‘gray‐baby’ syndrome (see Introduction) occurred following chloramphenicol administration to neonates. This serious and life‐threatening syndrome is believed to result from reduced metabolic clearance, and hence high exposure, of chloramphenicol associated with immature UGT2B7 118. Although immature hepatic UGT2B7 is likely to be involved in zidovudine and chloramphenicol findings, the intestinal form could also contribute.

Other drug metabolizing enzymes demonstrate little change with age, including epoxide hydrolase, glutathione peroxidase 119 and alcohol dehydrogenase 120. The ontogeny of the other intestinal drug metabolizing enzymes remains to be examined. Developmental patterns based on mRNA expression are not necessarily predictive of the changes occurring at the protein or activity level. This is not unexpected given the numerous regulatory mechanisms occurring after mRNAs are manufactured (see 121 for a review).

Intestinal wall transporters

Transporters are membrane‐bound proteins playing a major role in the cellular uptake and efflux transport of endogenous and exogenous compounds. Numerous transporters belonging to the ATP binding (P‐gp, BCRP, MRPs) or solute carrier super‐families (PEPT1, OATPs) are expressed in the intestine and are important determinants of oral drug pharmacokinetics (see 122 for review).

P‐glycoprotein (P‐gp) is, by far, the most studied intestinal efflux transporter and is known to markedly affect the oral bioavailability of various drugs, especially those with poor solubility and low passive diffusion permeability. Several studies have demonstrated that intestinal P‐gp is undetectable in early fetal life, then increases with age to reach adult expression after 12 weeks of gestation 123, 124. As expected, there are no age‐related differences in intestinal P‐gp expression between neonates, infants, children and adults 125, 126. It should be pointed out that all these investigations focused on mRNA and therefore need careful interpretation. Indeed, it has been demonstrated that mRNA levels for intestinal transporters might not correlate with the corresponding proteins or functional activities 121, 127. Where available, the pharmacokinetic behavior of P‐gp substrates in the pediatric population vs adults is more informative. Cyclosporin has a poor oral bioavailability as a result of high intestinal CYP3A4 first‐pass metabolism combined with effective P‐gp‐mediated efflux. The bioavailability of cyclosporine was found to be constant from 0.4 to 18 years of age, which suggests an absence of changes in intestinal P‐gp and/or CYP3A4 activities in this age range 81. Aliskiren is another P‐gp substrate with an oral bioavailability as low as 3% 128 with the possibility of drug interactions when co‐administered with P‐gp inhibitors 129, 130. A juvenile rat toxicity study revealed a substantial increase in plasma exposure and severe toxicity when aliskiren was delivered to pups compared with adult animals. Immaturity of intestinal P‐gp was identified as the causative mechanism 131. Conversely, the pharmacokinetics and safety profile of aliskiren in pediatric patients (6–17 years old) were found similar to adults 132. Although incomplete, these data may suggest species differences in the developmental pattern of intestinal P‐gp, with less impact in humans.

When compared with P‐gp, there are fewer data on the other intestinal transporters. Immunodetectable breast cancer resistant protein (BCRP), another efflux transporter limiting the intestinal absorption of drugs, was demonstrated as early as 6 weeks of gestation 123. As for BCRP, data related to the multi‐drug associated protein 1 (MRP1) efflux transporter are scarce and only one study reported immunodetectable protein levels comparable to adults from 7 weeks of gestation 123. Mooij et al. 125, 133 reported stable mRNA expression of MRP2 from neonates to adults. The expression of influx oligopeptide transporter PEPT1 was investigated in intestinal tissue samples across the pediatric age range 133. PEPT1 mRNA expression of the neonates was only marginally lower (0.8‐fold) than the older children. Immunostaining demonstrated specific apical localization of PEPT1 in neonates, comparable to that observed in adults 133, 134. The organic anion transporting polypeptide (OATP) 2B1 uptake transporter appears to have a different ontogeny pattern compared with the transporters described earlier. OATP2B1 mRNA expression was found to be 3‐fold higher in neonates compared with adults 125, 133.

Again, mRNA may not be predictive of transporter protein level or activity, and should be interpreted cautiously. Hopefully, absolute transporter protein abundance data are becoming more and more available thanks to the recent developments in mass spectrometry‐based targeted proteomics 135, 136. Compared with mRNA, abundant protein level is usually considered to be a more suitable parameter of functional activity. However, some recent reports have emerged showing a disconnection between protein expression and activity for some transporters 137. One of the proposed hypotheses relates to post‐translational modifications (glycosylation, phosphorylation and ubiquination) as described for P‐gp, PEPT2, OATP2B1 and BCRP 138, 139. Because of the limitations of the expression data, more functional and pharmacokinetic studies are needed to understand age‐related changes in intestinal transporter activity. Of note, pharmacokinetic data across age are typically missing for drugs that are substrates of intestinal transporters other than P‐gp.

Experimental approaches to measure developmental changes in intestinal functions (Table 1)

Many in silico (e.g. calculated logP and PSA) and in vitro cellular assays (e.g. Caco‐2, MDCK cell lines, expressed metabolizing enzymes and transporters) have been successfully applied to characterize and predict the intestinal absorption, metabolism and transport of drugs (see 140, 141, 142 for review). These assays cannot directly assess effects of age on drug disposition. Instead, they allow a mechanistic understanding of the drug identifying the properties that might limit its availability (e.g. low solubility, instability in the gastrointestinal tract, efflux, intestinal first‐pass). Also, these assays are useful to identify the specific absorption and dissolution mechanisms, enzymes and transporters involved. Once identified, the properties of concern can be scrutinized for potential developmental changes using the relevant ex vivo, in situ or in vivo approaches.

Ex vivo and in situ methods using intact tissue specimens have been developed to study intestinal drug disposition. Although these methods have some limitations and are not always amenable to humans, they can be instrumental for exploring age effects, as well as species differences, effects of formulation, food, co‐administrations and disease states on intestinal functions (see 143 for review). Brush‐border membrane vesicles (BBMV) consist of apical membrane preparations isolated from intestinal segments through a number of homogenization and purification steps 144. The BBMV method was developed to study transport activities across the apical membrane in various species including human 145, 146. When compared with the other techniques listed here below, BBMV are easier to set up and less demanding in terms of tissue quality. Unfortunately, since the apical and basolateral membranes are separated, the use of BBMV is restricted to the study of active uptake into or efflux from enterocytes, not to the prediction of intestinal absorption in vivo. BBMV have been used to investigate developmental changes in intestinal brush border functions of mice 147, rats 148, 149, minks 150, piglets 151 and calves 152. To the best of our knowledge, the technology has never been used to study ontogeny of intestinal function in humans.

Everted sac preparations are based on intestinal segments, inverted over a glass rod to expose the serosa to the outside, and tied off at the ends 143, 144. An oxygenated solution is injected into the inner compartment and then the sac is placed in a buffered medium. Transport of drugs can be studied in either direction, mucosa to serosa or serosa to the mucosa. The method allows investigations of intestinal drug absorption, transport and gut wall metabolism. Unfortunately, the method suffers numerous drawbacks, e.g. the limited availability of human tissues, rapid deterioration of intestinal function after tissue removal (within minutes), limited viability of the intestine under in vitro conditions, lack of digestive enzymes, and a small serosal compartment that may confound the kinetics. So far, the method has only been applied to animal species such as mice, rats, guinea‐pigs, rabbits, sheep and pigs 153.

The single‐pass intestinal perfusion technique is based on the cannulation and perfusion of a segment of the intestine 143, 154. The disappearance of the drug from the perfusate is measured to determine the rate and extent of absorption. The method can also be used to investigate intestinal metabolism and active transport. While ex vivo isolated intestinal perfusion is only applicable to laboratory animals, in situ perfusion can be used across all species including human and is recommended by the FDA for classifying drugs based on BCS 154. In situ perfusion was used to demonstrate the higher intestinal permeability in suckling rats compared with older rats 155. In another rat study, the jejunal absorption of metoprolol and antipyrine was found to be constant over 5–30 weeks of age 156. In situ perfusion studies in human adults remain scarce due to the ethical issues and limited availability of healthy volunteers. So far, the method has not been reported in children.

The Ussing chamber is a powerful and well validated in vitro method involving a small section of intestinal mucosa mounted between two chambers (i.e. donor and receiver compartments). Samples are collected from the receiver side at intervals and analysed for drug appearance 143. The Ussing chamber technique has been successfully applied to animal and human tissues to study drug permeability, metabolism and transport for many compounds. Recently, using a large set of compounds, the in vitro permeability as measured in pig intestinal tissues was found to correlate well with the values measured in human samples 157. As for the other in vitro methods, the Ussing chamber technique requires fresh surgical samples (to be used within 1 hour following surgical excision) which hampers studies with human material 158. Endoscopic biopsies have been introduced as an alternative amenable to healthy adult volunteers 159 or diseased children 160, 161, 162. Using the Ussing chamber technique, rat intestinal segments and various marker drugs, Annaert et al. 163 showed a decrease in the paracellular pathway in aging rats. This finding was paralleled with an increase in the transcellular flux and P‐gp mediated‐efflux. There has been no reported attempt to use human excised intestinal segments to study developmental changes in intestinal drug absorption, metabolism or transport.

The access to human surgical samples is restricted and limits the use of the above mentioned in vitro methods. In most cases, specimens originate from patients undergoing gastrointestinal surgery 159 or from organ donors 164. The conditions of the human tissue samples are variable due to differences in diet, medication, environmental conditions, surgical procedure, separation from adjacent diseased tissue, which all may affect the tissue morphology, the functional activity and the data collected in vitro 157. Finally, irrespective of the technique, the tissue has a limited viability in vitro and only incubations of less than 3 hours can be envisaged 158. Biopsy samples are easier to obtain but provide a small amount of tissue (2–21 mm2) and so may not be compatible with all ex vivo methodologies or expression analysis (e.g. mRNA and protein). Healthy intestinal biopsies can be obtained in adult volunteers, but rarely from healthy children, which limits the investigation on age‐dependent changes in the control population. Very recently, precision‐cut intestinal slices have re‐emerged as a tool for in vitro metabolism, transport, drug interaction and toxicity studies both in preclinical animal species and in humans 165. It allows the study of numerous compounds in parallel using tissues from the same individual. Although potentially powerful, this latter technique is likely to have the same shortcomings as other methods using excised tissues (i.e. limited access to healthy specimens).

In vivo studies using preclinical animals are routinely used to explore the pharmacokinetic profile of drug candidates and predict the clinical outcome. Unfortunately, species differences in drug disposition usually make animal studies poorly translatable to humans. The most commonly used species are mice, rats, dogs and monkeys, but also recently minipigs 166. There are reports describing a good correlation between oral drug absorption in rodents and humans 167, however, using different sets of test compounds, other studies concluded poor predictivity 168, 169, 170. Similarly, absorption in monkey appears to be predictive of humans 171, except for highly transformed compounds 172. Drug absorption in dogs was first described as poorly predictive of the human outcome 173. Subsequent studies indicated that the paracellular transport route for polar compounds is overestimated in dogs. Consequently, when polar compounds are excluded from the dataset, dog might be a more relevant to humans 168. Overall, the translation of drug absorption and intestinal permeation from animals to humans remains challenging. This is not unexpected given the numerous processes involved and their propensity for species differences 174. The use of laboratory animals to predict age‐related changes in human drug absorption is likely to be even more hazardous. Although never fully scrutinized, the ontogeny of the various gastrointestinal functions is likely to have a time course that varies across species. This deserves more attention as it may improve the design and interpretation of toxicity studies in juvenile animals 87.

Pediatric clinical pharmacokinetic studies

Pending more predictive preclinical models and a better understanding of the ontogeny of intestinal processes, there is an overwhelming need for more pharmacokinetic trials in the pediatric population, especially those that measure changes in absorption parameters with age. Clinical data are necessary to provide a rational basis for dosage adjustment in this vulnerable population, to generate scientific evidence for the developmental changes in intestinal processes (e.g. population pharmacokinetic analysis), and to gradually fill the knowledge gaps in the field. Often, because of the paucity of clinical data, the pharmacokinetic profile in the pediatric population is predicted from adult data assuming that the absorption rate constant and the fraction absorbed are conserved across the age spectrum. A recent review of 1081 registered trials in children revealed that only 24% of the trials incorporated pharmacokinetic measurements 175. Also, 74% of the eligible trials were conducted in children aged over 2 years, while changes in drug disposition and drug response tend to predominate in younger children. The limited number of pediatric studies, especially in very young children, illustrates the numerous obstacles associated with those studies, from practical and methodological difficulties to regulatory requirements and ethical concerns.

Pediatric trials typically involve fewer study participants compared with trials in adults 176. Innovative study designs (e.g. adaptative design, Bayesian approach, randomized withdrawal design) are required to accommodate the small group sizes, and possibly facilitate ethical acceptance.

Blood sample collection in children is more challenging than in adults. The blood volume that can be safely collected at a time in newborns is much smaller than in adults (1 ml versus up to 10 ml, respectively) 16. Venipuncture is also technically more challenging in children, a needle prompting an emotional response in the child and the parents, with the practitioner being reluctant to have more than one or two attempts at venipuncture before sampling 177. Microsampling from finger and heel prick can be used as an alternative to venipuncture. A novel dried blood sampler (termed volumetric absorptive microsampler) was recently introduced to overcome the technical issues observed when taking a sub‐punch of dried blood spot samples 178. Hopefully, the newer bioanalytical techniques are sensitive enough to accommodate those smaller volume samples.

From a regulatory perspective, drugs investigated in pediatric trials should either provide a direct benefit to the subjects (precluding studies in healthy volunteers) or be used at a dose where risks are minimal. The latter includes the use of microdosing studies in children who have a disorder or condition which is the object of the research 179. A proof of concept pediatric microdose study investigating [14C]‐paracetamol in diseased children and using accelerated mass spectrometry was recently reported 180. Microdosing in control healthy children is not authorized. In addition, as for adults, microdosing has some limitations with respect to saturable processes such as intestinal transport and metabolism 181, 182, 183.

The absence of age‐appropriate formulations has been identified as another factor limiting pharmacokinetic studies in the pediatric population 16. Pediatric trials with oral drugs often require dedicated formulations differing from those used in adults. Pediatric formulations typically include liquids, chewable tablets, rapidly dissolving tablets, and more palatable flavors.

Physiologically based pharmacokinetics (PBPK)

Traditionally, the drug dosage to be administered to children has been estimated from adult data using allometric scaling related to body weight or body surface area 184. Recently, a move towards more mechanistic approaches has been made, integrating developmental changes in key anatomical, physiological and biochemical processes relevant for drug disposition 185. Nowadays, physiologically based pharmacokinetic (PBPK) modeling is routinely used for this mechanistic purpose, and is available as various ready‐to‐use ‘designed’ software (e.g. Simcyp®, PKSim®, Gastroplus®) 186.

Initially, PBPK software primarily focused on adult populations. Subsequently, pediatric modules have been added to integrate the physiological differences between adults and children. These modules, derived from the most recent literature, describe in mathematical terms the relationships between age and various physiological and biochemical variables such as body weight, body height, organ volume, cardiac output, renal function, plasma protein binding and some key drug metabolizing enzymes (e.g. CYPs) 187, 188, 189. The gastrointestinal function is covered as well including e.g. age‐related changes in organ volume, absorptive surface area, permeability, enzymes activity, pH, intestinal fluid content and transit time 190, 191.

There are several recent examples illustrating the value of PBPK for predicting drug disposition in the pediatric population, and for supporting pediatric approval 192, 193, 194, 195. Often, the strategy used to predict disposition in pediatrics involves an initial step where a PBPK model is built to describe the drug pharmacokinetics in adults 195, 196, 197. It is an iterative process where predicted and observed data are compared with further improve the model. Then, once the model has been qualified in adults, its physiological variables are scaled down to the pediatric population, whereas the drug‐related parameters remain unchanged. Such a workflow allows the sponsor and the regulatory authorities to gain confidence in the understanding of the drug pharmacokinetics before performing studies in the vulnerable pediatric population.

Although successful predictions for pediatrics have been reported, one should recognize the knowledge gaps that still limit the predictive value of PBPK. Indeed, the predictability of PBPK models is strongly dependent on the understanding of the drug properties and of the key physiological processes involved in its disposition. Unfortunately, as highlighted earlier, most of the developmental changes affecting drug absorption are still poorly understood. A recent study from a PhRmA initiative 198 showed that the current methods to predict the pharmacokinetics in adults were about three times less predictive for oral when compared with intravenous drugs. The plasma time course profiles after oral dosing were accurately predicted in only 23% of the cases, illustrating the need for a better understanding of the gastrointestinal physiology, its sensitive factors as well as the interplay between all the underlying processes driving drug absorption. Modeling for pediatrics is even more challenging considering the additional hurdle of our limited understanding of the developmental changes in gastrointestinal processes. This is nicely illustrated with voriconazole which shows an oral bioavailability of 45–66% in pediatrics to be compared with 96% in adults, a difference that cannot be explained by age‐dependent changes in hepatic clearance. The data were recapitulated in a retrospective PBPK analysis incorporating an extensive intestinal first‐pass in pediatric, but not in adult, subjects 199. This would imply that the enzymes involved in voriconazole extraction are overexpressed in the pediatric intestine, which either has not been unambiguously established (CYP3A4) or has not been studied (CYP2C19 and FMO3) 200, 201. This case study illustrates how PBPK modeling is undoubtedly useful for mechanistic investigations but still suffers limitations for prospective predictions of complex drugs.

In order to fill these needs, the Innovative Medicines Initiative (IMI) launched a dedicated initiative in 2012, the Oral Biopharmaceutical Tools (OrBiTo) project 202. The project is intended to refine existing tools and to explore newer assays in order to better understand and predict drug absorption. Ultimately, the project plans to include the full complexity of drug absorption integrating drug release and dissolution (as a function of both pH and intestinal content), potential precipitation all along the gastrointestinal tract, potential impact of gut flora, intestinal enzymes and transporters. Because the IMI is Europe's largest public‐private initiative in life sciences combining academia, pharmaceutical companies and software companies, the OrBiTo project will be an important and useful step to better understand and predict the pharmacokinetic profile of oral drugs, including in pediatrics.

Conclusion

A recent FDA survey showed that 42% of recently completed pediatric trials have failed to demonstrate either safety or efficacy, and pediatric approval could not be obtained 175. One quarter of the efficacy failures was attributed to suboptimal exposure, probably from a misapprehension of age‐related changes in drug disposition.

Thanks to coordinated efforts from the pharmaceutical industry, academics and regulatory agencies, enormous progress has been made in our understanding of the age‐related changes in drug disposition, particularly those relating to liver drug oxidation and transport. So far, ontogeny of intestinal drug absorption and pre‐systemic metabolism has been overlooked with gaps in our knowledge remaining.

Specific research efforts are needed on the age‐dependent expression of intestinal metabolizing enzymes, beyond CYP3A4, and drug transporters. Although mRNA data can be easily generated, it may not be predictive. Thus, protein quantitation using LC–MS‐MS and ex vivo functional assays should be given a high priority. Expression data are often reported for fetal and neonatal intestinal samples while investigations should cover the whole age range, with more information in infants and children. The relative lack of physiological data across age groups restricts the use of PBPK modeling to predict drug disposition in the pediatric population. Although encouraged by regulatory agencies, clinical trials in the pediatric population are still limited. More clinical data in the pediatric population will help validating PBPK models and building more evidence based information on the effects of age on intestinal absorption. All these information gaps need to be filled to ensure the development of safer and more efficacious drugs for children.

Conflict of Interest

The authors have no conflict of interest.

Table 1

Experimental methods to study developmental changes in intestinal drug absorption, metabolism and transport

Model	Applications	Limitations	Examples of use for studying ontogeny of intestinal function
Enzyme and transporter expression in tissue collection and biopsies	mRNA, protein and activity related to metabolizing enzymes and transporters	mRNA might be not predictive of functional activity Difficulties to access healthy tissue samples	CYP3A4 protein and activity measured in 74 histologically normal pediatric biopsies 88
Brush‐border membrane vesicles	Active uptake	Limited scope	Intestinal uptake of folate in rats of varying ages 148
Ex vivo everted sac	Permeability, metabolism and transport	Not amenable to human	Intestinal transport and metabolism of prostaglandin PGF2 in rats of varying ages 210
Ex vivo Ussing chamber	Permeability, metabolism and transport	Difficulties to access healthy tissue samples	Intestinal permeability of various compounds measured in rats of varying ages 163
In situ intestinal single pass perfusion	Permeability, metabolism and transport	Complex surgical procedures and instrumentation Hardly amenable to children	Intestinal permeability of various compounds measured in rats of varying ages 156
In vivo animal pharmacokinetic studies	Absorption, transport, metabolism	Animal data not translatable to human. Ontogeny time frame different than in humans	Ontogeny of intestinal P‐gp responsible for the higher bioavailability and higher toxicity of aliskiren in rat pups when compared with adult animals 131
PBPK	Full integration of all the data. Predicted drug concentration over time in plasma and various tissues	Relative lack of physiological and biochemical data across age groups	Age‐dependent change in the intestinal first‐pass metabolism of voriconazole. Knowledge gaps discussed 199, 200, 201
Pharmacokinetic trial in children		Restricted to diseased subjects; small sample size; ethical barriers	Pediatric microdose study of [14C] paracetamol using accelerated mass spectrometry detection 180