Quantifying breast milk intake by term and preterm infants for input into paediatric physiologically based pharmacokinetic models.

Abstract Despite the many benefits of breast milk, mothers taking medication are often uncertain about the risks of drug exposure to their infants and decide not to breastfeed. Physiologically based pharmacokinetic models can contribute to drug‐in‐milk safety assessments by predicting the infant exposure and subsequently, risk for toxic effects that would result from continuous breastfeeding. This review aimed to quantify breast milk intake feeding parameters in term and preterm infants using literature data for input into paediatric physiologically based pharmacokinetic models designed for drug‐in‐milk risk assessment. Ovid MEDLINE and Embase were searched up to July 2, 2019. Key study reference lists and grey literature were reviewed. Title, abstract and full text were screened in nonduplicate. Daily weight‐normalized human milk intake (WHMI) and feeding frequency by age were extracted. The review process retrieved 52 studies. A nonlinear regression equation was constructed to describe the WHMI of exclusively breastfed term infants from birth to 1 year of age. In all cases, preterm infants fed with similar feeding parameters to term infants on a weight‐normalized basis. Maximum WHMI was 152.6 ml/kg/day at 19.7 days, and weighted mean feeding frequency was 7.7 feeds/day. Existing methods for approximating breast milk intake were refined by using a comprehensive set of literature data to describe WHMI and feeding frequency. Milk feeding parameters were quantified for preterm infants, a vulnerable population at risk for high drug exposure and toxic effects. A high‐risk period of exposure at 2–4 weeks of age was identified and can inform future drug‐in‐milk risk assessments.

Despite the many apparent benefits of breast milk, mothers taking medication often have difficulty deciding whether to breastfeed their infant. Uncertainty regarding the safety of breastfeeding while on medication has been frequently cited as a reason for mothers not to initiate or continue breastfeeding (Li, Fein, Chen, & Grummer-Strawn, 2008;Nordeng, Koren, & Einarson, 2010;Teich, Barnett, & Bonuck, 2014). Concerned mothers may choose not to breastfeed due to the risk of exposing the infant to the drug through milk, which has led to serious toxicity, including death in some reported cases (Beauchamp, Hendrickson, Horowitz, & Spyker, 2019;Koren, Cairns, Chitayat, Gaedigk, & Leeder, 2006;Madadi et al., 2009;Schultz, Kostic, & Kharasch, 2019). Alternatively, mothers may discontinue taking their medication even though the resultant infant exposure to medications may actually be low. As examples, breastfeeding mothers have been shown to be noncompliant to oral antibiotics (Ito, Koren, & Einarson, 1993) and to antidepressant therapies that may have been relatively safe for the infant after a risk-benefit ratio assessment (Boath, Bradley, & Henshaw, 2004).
As a strategy to reduce uncertainties surrounding maternal drug use during breastfeeding, risk assessments can be performed. In these assessments, an infant dose is estimated to determine the amount of drug an infant would ingest through milk (Hale & Rowe, 2016). Integrating the infant dose with a physiologically based pharmacokinetic (PBPK) model can lead to a metric of exposure that, when linked to a measure of safety, forms the basis for the risk assessment.
The incorporation of PBPK models in the process of drug development has become increasingly prevalent in the past two decades (Rowland, Peck, & Tucker, 2011). PBPK models have the ability to provide in silico estimates of drug exposure given the proper parameterization with host physiology and drug properties (Maharaj & Edginton, 2014). In order to fully exploit the utility of PBPK models in quantifying drug uptake in breastfed neonates, an accurate measure of infant feeding parameters, volume and frequency of maternal milk intake, is needed. Essentially, dose to the infant through breast milk is calculated by multiplying daily volume of milk intake by the drug concentration in milk (Anderson & Sauberan, 2016). Knowledge of the total daily milk intake combined with information on intake frequency can help identify the peak serum concentration an infant would receive after feeding (C max ) and contribute to an assessment of drug safety. Total daily milk intake divided by the frequency gives volume per feed, which essentially determines the dose of the xenobiotic to the infant. Higher doses lead to higher peak concentrations and may factor in to a decision about risk during breastfeeding. The daily milk intake volume of 150 ml/kg is commonly used to determine infant dose, a value first proposed by Wilson, Brown, Hinson, and Dailey (1985) in 1983 and solidified by the WHO in 1988 (Bennett, 1988). However, as suggested by Anderson and Sauberan (2016), and clearly demonstrated in longitudinal data from the United States (Neville et al., 1988), feeding volumes are not constant across postnatal ages (PNAs) and often have large interindividual variability. It is therefore fundamental to capture representative intake volumes to inform more appropriate risk assessments.
Although feeding volumes have been captured by several reviews and reports (Butte, Lopez-Alarcon, Garza, & Expert Consultation on the Optimal Duration of Exclusive Breastfeeding, 2002;da Costa et al., 2010;Galpin et al., 2007;Hester, Hustead, Mackey, Singhal, & Marriage, 2012;Reilly, Ashworth, & Wells, 2005; U.S. EPA, 2011), they focus solely on term infants, whereas milk intake by preterm neonates remains an unexplored area. Notwithstanding the lack of reviews for preterm infants, the need to study this population in relation to breast milk and maternal medication should be emphasized.
Breast milk is particularly beneficial to preterm infants in providing appropriate nutrition during their time growing ex-utero in a crucial period of accumulating nutrient reserves typical for the developing fetus (Bouyssi-Kobar et al., 2016;Carlson & Ziegler, 1998;Ehrenkranz et al., 2006) and reducing necrotizing enterocolitis and sepsis, which is more prevalent in this population (Underwood, 2013). Further attention to preterm infants is also warranted because they are more vulnerable than term infants to toxicity from drug exposure through breast milk. Preterm infants have reduced capacities for drug metabolism in the liver and drug excretion in the kidneys, and as a result, eliminate drugs more slowly from the body (Kearns et al., 2003;Reiter, 2002). In comparison to term infants, their further lowered ability to eliminate drugs may lead to high sustained drug concentrations in plasma, especially over multiple doses or feeds.

Key messages
• Physiologically based pharmacokinetic (PBPK) models can be used to predict infant exposure as part of drug-in-milk risk assessments.
• Existing reviews approximating breast milk intake for input into these PBPK models are unable to address exclusive breastfeeding across all ages, preterm infants and feeding frequency.
• Results from this literature review were used to construct a nonlinear regression equation on the weight-normalized human milk intake and determine a weighted mean feeding frequency of exclusively breastfed term infants.
• Preterm infants fed with similar feeding parameters to term infants on a weight-normalized basis.
• A high-risk period of exposure at 2-4 weeks of age was identified.
In this review, a comprehensive search of the literature was performed to retrieve estimates of human milk intake and frequency for term and preterm infants as a function of PNA, as inputs for PBPK models with the purpose of subsequent drug-in-milk risk assessments.

| Eligibility criteria
Studies reporting term or preterm infants receiving human milk with data on their volume or frequency of intake were of interest. Term infants were defined as >37 gestational age (GA) at birth, and those of ≤37 GA at birth were identified as preterm infants. For term infants, articles were included if data were provided on infants of any age who were exclusively breastfeeding (EBF) or infants >6 months PNA who were partially breastfeeding (PBF). These criteria were selected to reflect the AAP and WHO recommendations and produce conservative estimates for subsequent risk assessments. For preterm infants, articles were included if data were provided on infants who were exclusively human milk-fed or were PBF with a diet that consisted mainly of breast milk, where information regarding the proportion of human milk and other sources of nutrients in the diet were provided. Only articles with volume data presented in relation to individual infant's body weights, presented as weight-normalized human milk intake (WHMI), were included to reduce interinfant and intrainfant variability of daily milk consumption observed in absolute milk intake measurements. Included studies were those that measured volume and frequency of intake for at least 24 hr, as intake of breast milk tends to differ throughout the day (Kent et al., 2006). Articles were excluded if infants had significant birth complications or were otherwise unhealthy, and studies in which the intake of breast milk was influenced by the study investigators, including non-ad libitum feedings and interventions that significantly increased the milk expression of mothers.

| Search strategy
The search strategies consisted of MeSH headings and text words related to premature and term infants, breastfeeding and volume and frequency of intake. The Ovid MEDLINE and Embase databases were searched up to July 2, 2019. Results were limited to the English language. Complete search strategies are provided in Appendix A. In addition to the searches, reference lists of key studies on volume or frequency of breastfeeding and the grey literature were used to identify studies.

| Study selection and data extraction
Two investigators (C. H. T. Y. and S. F.) screened title, abstract and full text for inclusion. The results were not screened in duplicate. Data extraction was performed by one investigator (C. H. T. Y.).

| Data analysis
Daily WHMI and human milk feeding frequency from each study were presented as mean ± standard deviation (SD) ml/kg/day and number of feeds/day, respectively. When only a median value was reported for a feeding parameter, the mean was assumed equal to the median based on an assumption of normally distributed data as demonstrated previously with individual subject measurements in five of the included studies (Arcus-Arth, Krowech, & Zeise, 2005). Studies reporting volume of milk intake in grams were converted to millilitres using the density of milk (1.03 g/ml). PNA in days were approximated from alternative sources of infant age, such as corrected age (CA) and GA, where appropriate. Data without quantitative summaries in the literature were digitized using Plot Digitizer (v2.6.8 by Joseph Huwaldt). Data were graphically represented and analysed using MATLAB R2018b. A sample-size-weighted nonlinear regression was performed to quantify the WHMI by EBF infants across all studies using the Intiquan Toolbox (IQMTools v1.2.2.2 by Henning Schmidt and colleagues). The following function containing an integrated form of a first-order increase followed by a first order decline was selected to represent the general shape of the data using the least number of parameters: where WHMI is in ml/kg/day and t is age of the infant in days. The unknown parameters (θ 1 , θ 2 and θ 3 ) were fitted using the observed data for breast milk volume, weighted by the sample size in each dataset. The cost function to be minimized was the sum of squared error. A simulated annealing temperature-based approach was used to explore the parameter spaces. The optimization was repeated 50 times with randomized parameter start values to confirm that a global minimum had been achieved and to explore any potential correlations between parameters.

| RESULTS
The literature search identified 2,257 nonduplicate results, and 17 articles were identified through other sources. Title and abstract screening of 2,274 records resulted in 2,054 articles excluded at this stage. In assessing the eligibility of full-text articles, 220 results were screened and 164 were excluded. The review process resulted in 52 studies presented in 56 articles. A Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) study selection flow diagram is provided in Figure 1, and characteristics of the included studies are described in Table 1.

| Volume of human milk intake
Twenty-eight and seven studies reported on the WHMI for term and preterm infants, respectively. The mean WHMI of term infants EBF (of all ages) and PBF (>6 months of age) according to age in days are presented in Appendix B. Table 2 reports the WHMI of preterm infants exclusively and partially breastfed across all PNAs. Figure 2 shows mean WHMI plotted against PNA and log-transformed PNA for term (Figures 2a and 2c) and preterm (Figures 2b and 2d) infants.
Across all studies, mean WHMI increases from birth until reaching a maximum of 152.6 ml/kg/day at 19.7 days of age, and then declines thereafter ( Figure 2a). Figure 3 presents the results of nonlinear regression modelling on the WHMI for EBF term infants over PNA  reporting on the milk intake of preterm infants described in Table 2, two provided WHMI volumes from EBF preterm infants in the neonatal intensive care unit (NICU) setting (Carnielli et al., 1996;Hendrickse, Spencer, Roberton, & Hull, 1984). The remaining studies with milk intake at the NICU presented preterm infants PBF with mother's own milk or donor human milk and preterm formula (Atkinson, Bryan, & Anderson, 1981;Cabrera Lafuente et al., 2018;Itabashi, Miura, Okuyama, Takeuchi, & Kitazawa, 1999). Two studies included infants who were also parenterally fed, which is reflected in their substantially lower WHMI as compared with those of term infants (Cabrera F I G U R E 1 Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) flow diagram of breast milk feeding parameter study selection T A B L E 1 Characteristics of studies quantifying breast milk feeding parameters in term and preterm infants Nine term infants, with five of the infants feeding exclusively on human milk and measured at approximately 1.5, 3 and 4 months of age. The remaining four infants were fed human milk and supplemental foods.
EBF volume  Cross-sectional, observational U.S. To combine the doubly labelled water method with conventional indirect calorimetry to explore possible differences in energy utilization between breastfed and formulafed infants.
Forty term infants, of which 20 were exclusively breastfed human milk since birth.
Ten of the breastfed infants were observed at 1 month, and a second set of 10 infants were observed at 4 months of age.
EBF frequency  Cross-sectional, observational U.S. To investigate the energy utilization of breastfed and formula-fed infants, and determine whether energy utilization was different between the two feeding groups.
Sixty-five healthy term infants whose mothers were recruited from a milk bank program were observed at 1 and 4 months of age. Of the participating infants, 32 were exclusively breastfeeding (44% male).

EBF volume
Cabrera Lafuente et al.

Spain
To describe the intake of mothers' own milk and its composition according to gestational age and postnatal age in preterm infants and to correlate them with neonatal weight, length and morbidities.
One-hundred-andseventy-six preterm infants (born <32 weeks GA; 52% male) were included, where the majority of infants received partial feeding of mothers' own milk. The infants were followed at 72 hr, 15 and 30 days and monthly until discharge (90 days).

Italy
To determine whether feeding very low birth weight infants exclusively with preterm human milk can ensure constant plasma and red blood cell long chain polyunsaturated fatty acids levels during the first month of life.
Twenty-two preterm infants (born 750-1,750 g), of which 14 infants received mother's own milk and eight received preterm human milk, were followed for 4 weeks.  Casey, Neifert, Seacat, and Neville (1986) Longitudinal, observational U.S. To report the intakes of milk, energy and some selected nutrients to estimate nutrient intake in the young neonate.
Eleven healthy term infants from mothers with uncomplicated pregnancies, who were put to the breast within two hours of birth and were followed for 5 days. The infants were exclusively breastfed, except for two who were given water or glucose water 3-6 times after breast feeds.

Honduras
To examine breast milk intake, total energy intake and infant growth among breastfed infants randomly assigned to receive nutritionally adequate, hygienically prepared complementary foods beginning at 16 weeks or to be exclusively breastfed until 26 weeks.
Fifty healthy term infants whose mothers were recruited from two public hospitals in Honduras. Infants were exclusively breastfed and at 16 weeks were randomly assigned to continue exclusively breastfeeding (control group), to introduce solid foods with ad libitum breastfeeding, or to introduce solid foods with maintenance of preintervention breastfeeding frequency. The infants were followed until 26 weeks of age.
EBF frequency de Carvalho, Robertson, Merkatz, and Klaus (1982) Longitudinal, observational U.S. To provide normative data for true demand breastfeeding mothers during the first 14 days postpartum, and to determine whether the frequency and duration of breastfeeding affect milk production at 1 month.
Forty-six term infants (59% male) whose mothers were encouraged to nurse on true demand, were observed during the first 14 days after delivery.
EBF frequency Lonnerdal (1991a, 1991b) Longitudinal, observational U.S. To present data on indexes of functional outcomes to judge whether a particular pattern of intake or growth, particularly for breastfeeding infants, is advantageous in a given environment.

| Frequency of human milk feeding
Twenty-four studies and one study reported on the frequency of human milk intake throughout the day for term and preterm infants, respectively. The mean feeding frequencies of term and preterm infants EBF and preterm infants PBF according to PNA in days are presented in Appendix C. There were no studies reporting on the feeding frequencies of term infants past 6 months of age. Figure 4a presents the daily feeding frequencies of term infants plotted against PNA. Feeding frequency in term infants increased until approximately 10 days of age and was relatively stable thereafter. The average feeding frequency across all ages of EBF term infants, weighted by sample size in each dataset, was 7.7 feeds/day (range: 4.3 to 13.8 feeds/day).
The mean feeding frequencies at 16, 21, and 26 weeks PNA from a single study (Cohen, Brown, Canahuati, Rivera, & Dewey, 1994) were much greater than values of other EBF term infants. PBF term infants tended to either feed at a high (>9 feedings/day) or low frequency (<6 feedings/day). Figure 4b shows the feeding frequency of preterm infants from a single study by Oras et al. (2015). The study reported the frequency of EBF preterm infants postdischarge from the NICU directly feeding at the breast, or at the breast and expressed milk (from the tube, cup and/or bottle feeding), were 12.5 and 14 feeds/day, respectively. At 2 months CA, the infants were feeding at the breast 9 feeds/day and at the breast with expressed milk 10 feeds/day. At 6 months CA, infants who were consuming expressed milk were fed 11.5 feeds/day. Figure 4, EBF preterm infants tend to feed at a greater frequency than preterm infants who were PBF (Figure 4a) and the average feeding frequency of term infants (Figure 4b). Conversely, PBF preterm infants were feeding at a lower frequency than the average for term infants, except at post-discharge (Figure 4b).

| DISCUSSION
In this review, human milk feeding parameters of daily WHMI and frequency of feeds as a function of age were quantified for term and preterm infants with literature data for inputs into PBPK models designed for drug-in-milk risk assessment. This review is the first to perform a literature search to report on daily WHMI and feeding frequencies for preterm infants, and feeding frequencies for term infants.
Through the literature search process, weight-adjusted intake values were obtained and reported for term and preterm infants. For term infants, a regression model described average volume intake levels in those exclusively breastfed from birth to 1 year of age. Frequency of feeding was described for both term and preterm infants across PNAs.
The rate of human milk consumption by term infants from approximately 20 days to 6 months PNA was consistent with published analyses that also examined WHMI (Figure 3). identified. This suggests that EBF term infants are at greatest risk of drug exposure at 2-4 weeks of age. These results are consistent with a review by Anderson and Valdes (2015), reporting a maximum average volume of intake between 170 and 184 ml/kg/day at 4 to 35 days of age from a longitudinal study in 13 lactating women (Neville et al., 1988). In a review by McNamara and Abbassi (2004), the reported peak intake volume was approximately 173.8 ml/kg/day at 1 month of age from four studies (Butte, Garza, Stuff, Smith, & Nichols, 1984;Hopkinson, Schanler, Fraley, & Garza, 1992;Neville, 1999;Rattigan, Ghisalberti, & Hartmann, 1981). Furthermore, the increased risk at 2-4 weeks of age is reflected in a review of case reports and studies At 27-28 days PNA (n = 22) Milk intake (ml/kg/day): 177.0 ± 9.7 Milk intake included mother's own milk and/or donor milk when mother's milk was insufficient.
Setting: NICU GA age at birth (weeks): 30 Birth weight (g): NR, however, infants were LBW and <1500 g at birth.
Infants fed nasogastrically or nasojejunally as much milk as they would tolerate. Maximum volume offered on the first day was 90 ml/ kg and increased by 30 ml/kg/day in a stepwise fashion. Infants in the own mother's expressed breast milk group fed predominantly breast milk. These infants received additional breast milk from donor milk, where necessary. In the event that donor milk was unavailable, formula was provided. Infants ceased to receive IV dextrose after the third week.
Own mother's expressed breast milk group (n = 10): At lower clearance in preterm infants relative to term infants, this group is at risk for higher exposure and toxic effects of the drug.
Human milk feeding frequency of EBF term infants slightly increased in the first 10 days of life and subsequently declined and stabilized ( Figure 3). Daily frequency of feeds as a function of age shown in Figure 3 suggests that feeding frequency was fairly constant over 6 months of age. In contrast, frequencies were either high (>9 feeds/day) or low (<6 feeds/day) across ages past 6 months for PBF  (2019) term infants. This stark contrast between feeding frequencies could be due to differences in PBF patterns of developed and developing countries, and rural and urban communities. Studies reporting higher frequencies were conducted in developing countries or rural communities (Amatayakul et al., 1999;Martinez & Chavez, 1971;van Steenbergen, Kusin, Kardjati, & Renqvist, 1991), whereas the lower frequencies were reported in studies conducted in developed countries and urban communities (Howie, McNeilly, Houston, Cook, & Boyle, 1981;Michaelsen, Larsen, Thomsen, & Samuelson, 1994;Nommsen, Lovelady, Heinig, Lönnerdal, & Dewey, 1991;Pao, Himes, & Roche, 1980;Paul, Black, Evans, Cole, & Whitehead, 1988). Similarly, the relatively high mean feeding frequencies of 13.1 to 13.8 feeds/day in those EBF were from term infants of a rural community (Cohen et al., 1994). More research may uncover the influence of cultural practices on feeding frequency, such as time spent at home, support from family members and willingness to breastfeed.
A single study by Oras et al. (2015) reported on the frequency of human milk feeding in the preterm population. EBF preterm infants tended to feed at a greater frequency than PBF preterm infants and the average feeding frequency of term infants (Figure 4b). Conversely, PBF preterm infants tended to feed less frequently than the term infant average (Figure 4b).
This review is not the first to identify quantified milk feeding parameters as inputs into PBPK models for predicting infant drug exposures through breast milk. In fact, existing literature shows that published PBPK models have used different values of milk intake feeding parameters as inputs. Schreiber (1993) used PBPK modelling to predict infant exposure to perchloroethylene, where an infant weighing 7.2 kg was assumed to ingest 700 ml/day of breast milk for 12 months postpartum. Equivalent to an average WHMI of 97.2 ml/kg/day, infant daily dose would be largely underpredicted at 2-4 weeks to 6 months PNA according to the WHMI findings in this review ( Figure 3). In one case report, obstructive jaundice and hepatomegaly were observed in an infant receiving 1.4 mg/kg/day of tetrachloroethylene (Bagnell & Ellenberger, 1977). Updating the calculations for the infant dose through breast milk with the WHMI from this review yields a daily intake of 1.3 mg/kg/day, which was previously calculated as 0.82 mg/kg/day (Schreiber, 1993). This demonstrates the potential influence of different intake values and the importance of identifying an appropriate milk volume in such risk assessments. In another study by Delaney et al. (2018), the authors used values reported by Kent et al. (2006) to simulate out the variability in feeding parameters, employing a mean ± SD of 76 ± 12.6 ml/feed and 11 ± 3 feeds/day. Interestingly, the calculated mean of these parameters across all age groups evaluates to 153 ml/kg/day.
While generous, this metric overestimates doses for children >2 months of age. For this reason, milk intake feeding parameters that capture changes across PNA, such as the regression equation as derived from the volume of intake literature values in this review, would be an improvement.
Although mean WHMI and frequency of feeds were obtained through the literature, variation around these values was not explored in this review. Future research focusing on the variability of milk feeding parameters as inputs into the PBPK models will be important to subsequent drug-in-milk risk assessments. Particularly, these efforts can help identify infants who are outliers and may be at highest risk for receiving toxic effects of the drug.

| CONCLUSION
In summary, the volume and frequency of human milk intake in term and preterm infants were quantified to provide dose information for paediatric PBPK models that will be used to inform infant exposure F I G U R E 4 Daily frequency of human milk feeding of term and preterm infants by age. (a) Human milk feeding frequency of term infants by the mean number of feeds per day of each study's reported age group. (b) Exclusively and partially breastfed preterm infants with human milk feeding frequency across postnatal ages superimposed on term infant data. Data are presented from a single study by Oras et al. (2015) and as median number of feeds per day and subsequent risk assessment. The derived nonlinear regression equation of WHMI can be used to describe the volume of intake for EBF term infants. Because the WHMI of preterm infants were consistent with the observed WHMI of EBF term infants, the nonlinear regression equation may be applicable to preterm infants. For daily frequency of feeds, a weighted mean of 7.7 feeds/day can be used for EBF term infants across all ages. The data from Oras et al. (2015) provided context in preterm infant milk intake feeding frequencies, however, more data are needed to inform the frequency of feeds in this population.

ACKNOWLEDGMENT
This study was funded by the Canadian Institutes of Health Research, grant PJT-159782.

CONFLICTS OF INTEREST
The authors declare that they have no conflicts of interest.