Molecular descriptors that influence the amount of drugs transfer into human breast milk

https://doi.org/10.1016/S0731-7085(02)00037-7Get rights and content

Abstract

Most drugs are excreted into breast milk to some extent and are bioavailable to the infant. The ability to predict the approximate amount of drug that might be present in milk from the drug structure would be very useful in the clinical setting. The aim of this research was to simplify and upgrade the previously developed model for prediction of the milk to plasma (M/P) concentration ratio, given only the molecular structure of the drug. The set of 123 drug compounds, with experimentally derived M/P values taken from the literature, was used to develop, test and validate a predictive model. Each compound was encoded with 71 calculated molecular structure descriptors, including constitutional descriptors, topological descriptors, molecular connectivity, geometrical descriptors, quantum chemical descriptors, physicochemical descriptors and liquid properties. Genetic algorithm was used to select a subset of the descriptors that best describe the drug transfer into breast milk and artificial neural network (ANN) to correlate selected descriptors with the M/P ratio and develop a QSAR. The averaged literature M/P values were used as the ANN's output and calculated molecular descriptors as the inputs. A nine-descriptor nonlinear computational neural network model has been developed for the estimation of M/P ratio values for a data set of 123 drugs. The model included the percent of oxygen, parachor, density, highest occupied molecular orbital energy (HOMO), topological indices (χV2, χ2 and χ1) and shape indices (κ3, κ2), as the inputs had four hidden neurons and one output neuron. The QSPR that was developed indicates that molecular size (parachor, density) shape (topological shape indices, molecular connectivity indices) and electronic properties (HOMO) are the most important for drug transfer into breast milk. Unlike previously reported models, the QSPR model described here does not require experimentally derived parameters and could potentially provide a useful prediction of M/P ratio of new drugs only from a sketch of their structure and this approach might also be useful for drug information service. Regardless of the model or method used to estimate drug transfer into breast milk, these predictions should only be used to assist in the evaluation of risk, in conjunction with assessment of the infant's response.

Introduction

Breastfeeding is an essential physiologic process that provides nutrition and protects a child against infection and immunologic disorders. The frequency of various diseases and metabolic disorders are less in a breastfed infant. When drugs are administered to a nursing mother, a small part of them may appear in her milk [1]. Most drugs pass into breast milk, but the dose is reduced and usually does not produce a pharmacological effect. Certain drugs, however, do reach greater levels in milk than in the mother. Because of the infant's small size and the difference in metabolism between infants and adults, occasionally this transfer of medication can be harmful to the infant [2], [3], [4]. The amount of drug excreted into milk depends on a number of kinetic factors: (1) lipid solubility of the drug; (2) molecular size of the drug; (3) blood level attained in the maternal circulation; (4) protein binding in the maternal circulation; (5) oral bioavailability in the infant and the mother; and (6) half-life in the maternal and infant's plasma compartments.

Human milk is a suspension of protein and fat globules in a carbohydrate-based solution [5]. The mechanisms by which medications are transferred into breast milk are no different than those governing passage into any other maternal body fluid or organ system. Drugs enter milk primarily by passive diffusion reaching concentration equilibrium with the concentration in the blood, but also by active secretory methods that can concentrate the drug in the breast milk [6], [7], [8].

The most important determinant of drug penetration into milk is the mother's plasma level. Drugs enter milk and, in most cases, exit milk as a function of the mother's plasma level. Of the many factors, perhaps the two most important and useful are the degree of protein binding and lipid solubility. Drugs that are extremely lipid soluble penetrate milk in higher concentrations (i.e. CNS active drugs). Protein binding also plays a very important role. Drugs that have high maternal protein binding almost invariably produce lesser levels in milk. Most drugs circulate in the maternal plasma bound to a large molecular weight protein called albumin. An unbound component remains freely soluble in the plasma and transfers into milk, while the bound fraction stays in the maternal plasma unable to reach the tissues. For weak acids and bases, excretion into breast milk is governed by factors such as their pKa, their concentration in plasma and the pH of the milk and plasma. In some instances, drugs become ion trapped in milk. Due to the lower pH of human milk, the physicochemical structure of the drug changes and prevents its perfusion back into the maternal circulation. Hence, they become ion trapped in the milk. Also, drugs may concentrate in the milk due to specialized transport systems, which ‘pump’ substances into the milk. In addition, a small water-soluble molecule, such as alcohol, may pas into the milk through aqueous pores in the membrane.

Assessing the risk of maternal medication to the breast-feed infant requires knowledge of the concentration of drug that might be present in the milk. This concentration can be calculated for a particular drug if the milk to plasma (M/P) concentration ratio is known for the drug. The M/P value is an attempt to identify the equilibrium concentration between breast milk and blood. It is equivalent to the drug concentration in the breast milk divided by the maternal serum concentration. As almost all drugs pass into milk from maternal plasma by passive diffusion, the M/P ratio is affected by the composition of the milk (aqueous, lipid, protein and pH) and the physicochemical characteristics of the drug (degree of ionization, molecular weight protein binding and lipophilicity).

Lower molecular weight drugs with high lipophilicity that are less protein-bound, nonionized, are more likely to diffuse into breast milk. The more lipid-soluble a medication, the more likely it will diffuse into milk. Breast milk has a lower pH (7.08) than plasma (7.42), causing weak bases to pass more readily into the milk than weak acids. Drugs that tend to concentrate in milk are weak bases, with low plasma protein binding and high lipid solubility. Additionally, medications may be transferred into breast milk incorporated within fat globules or bound to proteins, primarily casein and lactalbumin. Highly protein-bound drugs, though, are unlikely to cross extensively into breast milk, since these drugs bind preferentially to serum albumin [9], [10]. Finally, large molecules are not able to diffuse passively into the milk.

There is a tremendous need for more information on the safety of drugs while breastfeeding. The milk to plasma drug concentration ratios has been determined experimentally for many drugs. This ratio is most reliable when it comes from studies where the area under the concentration–time profiles has been measured over a whole dose interval. Unfortunately, most of the data on breastfeeding and drugs is based on single case reports or small case series. M/P data based on single time point concentration measurements in the two phases can be misleading because the time course of concentrations in milk and plasma may not parallel each other. Using these experimental data, empirical regression equations have been developed relating milk plasma ratio to the drug pKa, the octanol water partition coefficient and plasma protein binding [11]. A log-transformed phase-distribution model appears to have good predictive performance [12]. The disadvantage of these methods is that plasma protein binding for the drug must be known or experimentally determined. Since these physico-chemical drug properties are not always available, a theoretical method that could predict the milk plasma ratio from the drug structure would be of interest.

The aim of this research was to simplify and upgrade a previously developed genetic neural network (GNN) model for prediction of M/P ratio given only the molecular structure of the drug [13]. The model is based only on theoretical molecular descriptors that can be calculated directly from molecular structure. This approach has potential use for drug information services when experimental physico-chemical properties of the drug are not available and experimental milk plasma ratios have not been investigated.

An artificial neural network is a biologically inspired computer program designed to simulate the way in which the human brain processes information. ANNs are composed of hundreds of single processing elements (PE). Each PE has weighted inputs, transfer function and one output. PEs are connected with coefficients (weights) and are organized in a layered feed forward topology, the input layer, the output layer and the hidden layers between them. The number of layers and the number of units in each layer determines the function complexity.

Neural networks gather their knowledge by detecting the patterns and relationships in data and learn (or are trained) through experience with appropriate learning exemplars. The input layer neurons receive data from a data file. The output neurons provide the ANN's response to the input data. Units in the input layer do not process. They simply pass an output value onto units in the second layer. Each hidden or output unit performs a biased weighted sum of inputs and passes this activation signal through an activation function (also known as a transfer function) to produce their output.

Multilayer perceptron (MLP) is the most commonly used type of feed-forward network. MLPs use a linear activation signal or post synaptic potential (PSP) to combine incoming inputs and usually a non-linear activation function (also known as a transfer function). PEs are defined by their weights and threshold. Linear PSP units perform a weighted sum of their inputs, biased by the threshold value. Thus, a parameterized model is formed, with the weights and thresholds (biases) as the free parameters of the model. The standard activation function for MLPs is the logistic function. It is an S-shaped (sigmoid) curve, with output in the range (0.1).

When the network is executed, the input variable values are placed in the input units and then the hidden and output layer units are progressively executed. Each unit in the proceeding layers calculates its PSP of the neuron by taking the weighted sum of the outputs of the units in the preceding layer and subtracting the threshold. The activation signal is than passed through the activation function to produce the output of the neuron. When the entire network has been executed, the outputs of the output layer act as the output of the entire network.

Section snippets

Experimental

Neural Networks TM (StatSoft Inc., Tulsa) was used for building the QSAR model. For calculating drug properties from molecular structure, BioMed CAChe Project leader (Fujitsu America, Inc.) and ACD/ChemSketch (Toronto, Canada) were used.

Results and discussion

The first step in developing QSAR was to calculate molecular descriptors. Seventy-one calculated structural features, including constitutional, topological, geometrical, quantum chemical and physicochemical descriptors, were generated for each drug molecule. The next step was to select descriptors that effect drug passage into breast milk.

Selection of the important molecular descriptors and examination of the variable contribution to the model through output sensitivity is an important aspect

Conclusion

A nine-descriptor nonlinear computational neural network model has been developed for the estimation of M/P ratio values for a data set of 123 drugs. Unlike previously reported models, the QSPR model described here does not require experimental parameters and could potentially provide useful prediction of M/P ratio of new drugs and reduce the need for actual compound synthesis and M/P ratio measurements. Model can be used to estimate the activities of other molecules only from a sketch of their

References (163)

  • C.Y. Oo et al.

    Active transport of cimetidine into human milk

    Clin. Pharmacol. Ther.

    (1995)
  • F.W. Kari et al.

    Active transport of nitrofurantoin across the mammary epithelium in vivo

    J. Pharmacol. Exp. Ther.

    (1997)
  • V.S. Toddywalla et al.

    Active transport of nitrofurantoin across a mouse mammary epithelial monolayer

    J. Pharmacol. Exp. Ther.

    (1997)
  • P.O. Anderson

    Drug use during breast-feeding

    Clin. Pharm.

    (1991)
  • S. Kacew

    Adverse effects of drugs and chemicals in breast milk on the nursing infant

    J. Clin. Pharmacol.

    (1993)
  • J.T. Wilson

    Drugs in Breast Milk

    (1981)
  • H.C. Atkinson et al.

    Clin. Pharmacokin.

    (1990)
  • J.C. Dearden

    Partitioning and lipophilicity in quantitative structure–activity relationships

    Environ. Health Perspect.

    (1985)
  • M.R. Riazi et al.

    Estimating diffusion-coefficients of dense fluids

    Ind. Eng. Chem. Res.

    (1993)
  • A. Leo et al.

    Comparison of parameters currently used in the study of structure–activity relationships

    J. Med. Chem.

    (1969)
  • G. Grassy et al.

    Computer-assisted rational design of immunosuppressive compounds

    Nat. Biotechnol.

    (1998)
  • V.K. Gombar et al.

    Quantification of molecular shape and its correlation with physico-chemical properties

    Indian J. Chem.

    (1987)
  • L.B. Kier et al.

    Molecular Connectivity in Structure–Activity Analysis

    (1986)
  • A. Verloop et al.
  • Z.N. Stowe et al.

    Sertraline and desmethylsertraline in human breast milk and nursing infants

    Am. J. Psychiatry

    (1997)
  • L.L. Altshuler et al.

    Breastfeeding and sertraline: a 24-hour analysis

    J. Clin. Psychiatry

    (1995)
  • P.O. Anderson

    Alcohol and breastfeeding

    J. Hum. Lact.

    (1995)
  • R.A. Lawrence

    Breastfeeding: A Guide for the Medical Profession

    (1994)
  • V.S. Toddywalla et al.

    Can chronic maternal drug therapy alter the nursing infant's hepatic drug metabolizing enzyme pattern?

    J. Clin. Pharmacol.

    (1995)
  • A. Taddio et al.

    Acyclovir excretion in human breast milk

    Br. J. Clin. Pharmacol.

    (1994)
  • K. Bork et al.

    Transfer of aciclovir from plasma to human breast milk

    Arzneim.-Forsch.

    (2000)
  • K.L. Wisner et al.

    Antidepressant treatment during breast-feeding. Review

    Am. J. Psychiatry

    (1996)
  • U. Breyer-Pfaff et al.

    Secretion of amitriptyline and metabolites into breast milk

    Am. J. Psychiatry

    (1995)
  • T.F. Bader et al.

    Amitriptyline in human breast milk and nursing infant's serum

    Am. J. Psychiatry

    (1980)
  • W.B. Pittard et al.

    Amitriptyline excretion in human milk

    J. Clin. Psychopharmacol.

    (1986)
  • H. Nau

    Clinical pharmacokinetics in pregnancy and perinatology. II. Penicillins. Review

    Dev. Pharmacol. Ther.

    (1987)
  • E. Steiner et al.

    Amphetamine secretion in breast milk

    Eur. J. Clin. Pharmacol.

    (1984)
  • P.E. Branebjerg et al.

    Blood and milk concentrations of ampicillin in mothers treated with pivampicillin and in their infants

    J. Perinat. Med.

    (1987)
  • R.M. Welch et al.

    Excretion of drugs in human breast milk

    Drug Metab. Rev.

    (1981)
  • S.H. Erickson et al.

    Aspirin in breast milk

    J. Fam. Pract.

    (1979)
  • K.F. Ilett

    Drug distribution in human milk

    Aust. Prescr.

    (1997)
  • W.B. White et al.

    Atenolol in human plasma and breast milk

    Obstet. Gynecol.

    (1984)
  • H. Liedholm et al.

    Accumulation of atenolol and metoprolol in human breast milk

    Eur. J. Clin. Pharmacol.

    (1981)
  • D. Ortega et al.

    Excretion of lidocaine and bupivacaine in breast milk following epidural anesthesia for cesarean delivery

    Acta Anaesthesiol. Scand.

    (1999)
  • G.G. Briggs et al.

    Excretion of bupropion in breast milk

    Ann. Pharmacother.

    (1993)
  • E.E. Tyrala et al.

    Caffeine secretion into breast milk

    Arch. Dis. Child.

    (1979)
  • J.E. Ryu

    Caffeine in human milk and in serum of breast-fed infants

    Dev. Pharmacol. Ther.

    (1985)
  • M. Perez-Reyes et al.

    Presence of delta9-tetrahydrocannabinol in human milk

    New Engl. J. Med.

    (1982)
  • R. Shimoyama et al.

    Monitoring of carbamazepine and carbamazepine 10,11-epoxide in breast milk and plasma by high-performance liquid chromatography

    Ann. Clin. Biochem.

    (2000)
  • S. Pynnonen et al.

    Carbamazepine: placental transport, tissue concentrations in foetus and newborn, and level in milk

    Acta Pharmacol. Toxicol. (Copenhagen)

    (1977)
  • Cited by (60)

    • The European Human Biomonitoring Initiative (HBM4EU): Human biomonitoring guidance values (HBM-GVs) for the aprotic solvents N-methyl-2-pyrrolidone (NMP) and N-ethyl-2-pyrrolidone (NEP)

      2021, International Journal of Hygiene and Environmental Health
      Citation Excerpt :

      Due to the low log Pow of NMP and NEP and their relatively short half-lives, the accumulation in fat is not very likely and the possibility of a transfer into human milk fat is reduced (Breitzka et al., 1997). But as both NMP and NEP as well as their metabolites have a very small molecular weight and through their solvent properties, it cannot be excluded that they may pass the human membranes into human breast milk via aqueous pores according to their concentration gradient (Kustrin et al., 2002), but at the moment no studies exist to confirm this hypothesis. Due to the reproductive toxicity and classification as SVHC of NMP and the chemical analogy to its substitute NEP, the German cooperation on human biomonitoring between the Federal Ministry for the Environment, Nature Conservation and Nuclear Safety and the German Chemical Industry Association (VCI) saw the necessity to develop appropriate analytical methods to monitor NMP and its substitute NEP in urine of the general population.

    • ExHuMId: A curated resource and analysis of Exposome of Human Milk across India

      2021, Chemosphere
      Citation Excerpt :

      In contrast, there is no clear difference between mean molecular weight for human milk contaminants and chemicals with low risk in Vasios et al. In sum, our observations confirm previous observations (Agatonovic-Kustrin et al., 2002; Zhao et al., 2006; Heinzow, 2009; Anadón et al., 2011; Vasios et al., 2016) on physicochemical properties of chemicals with high risk of transfer to human milk from maternal plasma. Overall, these results give insights into the effect physicochemical properties can have in the transfer of environmental chemicals into human milk, and further, can enable the prediction of such chemicals.

    • Distribution of hydrophilic and lipophilic antibacterial drugs in skim milk, cream, and casein

      2018, Journal of Dairy Science
      Citation Excerpt :

      Drug distribution to the different fractions of milk varies according to both the route of administration and the dose (Ziv and Rasmussen, 1975; Power et al., 2013a). In particular, protein binding degree and lipophilicity play an important role in drug distribution in the milk (Agatonovic-Kustrin et al., 2002). The passage ratio of hydrophilic drugs, such as β-lactam and aminoglycoside, to udder tissue are low, whereas that of lipophilic drugs, such as macrolides and fluoroquinolones are high (Benchaoui, 2010).

    View all citing articles on Scopus
    View full text