Volume 34, Issue 3 pp. 209-223
REVIEW ARTICLE
Free Access

Patient variation in veterinary medicine – Part II – Influence of physiological variables

S. MODRIC

S. MODRIC

Center for Veterinary Medicine (CVM), Office of New Animal Drug Evaluation, Food and Drug Administration (FDA), Rockville, MD, USA

Search for more papers by this author
M. MARTINEZ

M. MARTINEZ

Center for Veterinary Medicine (CVM), Office of New Animal Drug Evaluation, Food and Drug Administration (FDA), Rockville, MD, USA

Search for more papers by this author
First published: 18 November 2010
Citations: 24
M. Martinez, Center for Veterinary Medicine (CVM), Office of New Animal Drug Evaluation, Food and Drug Administration (FDA), 7500 Standish Place, HFV-130, Rockville, MD 20855, USA E-mail: [email protected]

The views expressed in this article are those of the authors and do not reflect the official policy of the FDA. No official support or endorsement by the FDA is intended or should be inferred.

Abstract

Modric, S., Martinez, M. Patient variation in veterinary medicine – Part II – Influence of physiological variables. J. vet. Pharmacol. Therap.34, 209–223.

In veterinary medicine, the characterization of a drug’s pharmacokinetic properties is generally based upon data that are derived from studies that employ small groups of young healthy animals, often of a single breed. In Part I of the series, we focused on the potential influence of disease processes, stress, pregnancy and lactation on drug pharmacokinetics. In this Part II of the series, we consider other covariates, such as gender, heritable traits, age, body composition, and circadian rhythms. The impact of these factors with respect to predicting the relationship between dose and drug exposure characteristics within an animal population is illustrated through the use of Monte Carlo simulations. Ultimately, an appreciation of these potential influences will improve the prediction of situations when dose adjustments may be appropriate.

Introduction

Although most of the data collected in veterinary species are derived from animals of similar physiological characteristics, there are readily identifiable factors that can influence drug clearance, bioavailability, and volume of distribution. These factors include genetic variability (e.g. breed effects), disease/stress, specific physiological conditions (such as pregnancy and lactation), hepatic and renal function, environment, food, gender, age, and circadian rhythms.

Within veterinary medicine, pharmacokinetic (PK) data are typically generated in small groups of normal healthy animals, and it is often assumed that these data will reflect the drug’s kinetic properties across the intended patient population. However, the population inferential value of these datasets is rarely assessed. In Part I of this series (Martinez & Modric, 2010), we explored published data describing the potential impact of altered physiological states, such as those associated with disease, stress, pregnancy, and lactation, on drug PK. In Part II, we continue this assessment with an evaluation of the influence of ‘normal’ physiological variables. These include such factors as gender, heritable traits, age, body composition, and circadian rhythms. Although none of these factors represent a pathological state or altered physiology, each has been shown to potentially influence the PK and/or pharmacodynamic (PD) characteristics of drugs both in human and in veterinary medicine. In the last section of this review, the results of Monte Carlo simulations are provided to illustrate how failure to identify the presence of subpopulations can distort expectations and render flawed population predictions.

As with Part I, we have included tables (Tables 1–4) which contain the results of PubMed searches conducted by including the variable of interest, the word ‘pharmacokinetics’ and the major veterinary species (dog, cat, swine, horse cattle) in the search line. For each factor, Tables include the number of hits collected from each search, and the number of relevant papers contained within these hits. A relevant paper is defined as a manuscript that addresses how these variables influence drug PK in that animal species. The citations for each of those relevant articles are provided. Inclusion of these papers is not intended to reflect our endorsement of their conclusions or of the experimental methodologies employed. Rather, it reflects our effort to provide as inclusive a list of references as possible. Relevant information derived from our search is provided in the body of this manuscript.

Table 1. Impact of gender on drug pharmacokinetics in major veterinary species
Species No. hits No. relevant hits Reference
Dog 45 7 Bruss et al. (2004)
Doursout et al. (1990)
Hay Kraus et al. (2000)
Izzat et al. (1989)
Lin et al. (1996)
Shiraga et al. (1995)
Vanapalli et al. (2002)
Cat 7 1 Lainesse et al. (2007)
Swine 16 2 Huxley et al. (2005)
Madej et al. (2005)
Horse 3 0 N/a
Cattle 5 1 Czerniak (2001)
Table 2. Impact of age on drug pharmacokinetics in major veterinary species
Species No. hits No. relevant hits Reference
Dog 154 20 Adusumalli et al. (1992, 1993)
Alberola et al. (1993)
Arnold et al. (1985)
Bruenger et al. (1983, 1991)
Ecobichon et al. (1988)
Hastings et al. (1985, 1986)
Moura et al. (1993)
Murphy et al. (1987)
Nakanomyo et al. (1992)
Ritschel and Vachharajani (1993)
Ritschel et al. (1988, 1991)
Singh et al. (1978)
Tryfonidou et al. (2002)
Weber et al. (2002)
Yang et al. (1992)
Yoshida et al. (1998)
Cat 76 3 van Hoek et al. (2007)
Seguin et al. (2004)
Swenson et al. (1990)
Swine 188 12 Bird and Hartmann (1996)
Kinoshita et al. (1995)
Tagawa et al. (1994)
Farmer et al. (1993)
Riond et al. (1992)
Nielsen et al. (1991)
Wilson et al. (1989)
Weström et al. (1989)
Gyrd-Hansen et al. (1984)
Friis et al. (1984)
Friis (1981)
Svendsen (1976)
Horse 64 12 Adamson et al. (1991)
Burrows et al. (1992)
Carrillo et al. (2005)
Clarke et al. (1992)
Cummings et al. (1990)
Duffee et al. (1997)
Ensink et al. (1994)
Golenz et al. (1994)
Jensen et al. (1990)
Lees et al. (1985)
Norman et al. (1997)
Perez et al. (1994)
Cattle 192 20 Aschbacher and Feil (1994)
Baroni et al. (1996, 2008)
Depelchin et al. (1988)
Igarza et al. (2002, 2004, 2006)
Janus and Antoszek (1999, 2000)
Janus and Suszycka (1996)
Janus et al. (1992)
Kaartinen et al. (1997)
Nouws et al. (1986, 1989, 1991)
Shoaf et al. (1987, 1989)
Soback et al. (1987)
Sutter et al. (1993)
Volner et al. (1990)
Table 3. Impact of body composition on drug pharmacokinetics in major veterinary species
Species No. hits No. relevant hits Reference
Dog 33 3 Kaiyala et al. (2000)
Lallemand et al. (2007)
Weiss (2008)
Cat 3 1 Center et al. (2000)
Swine 11 3 Craven et al. (2001, 2002a,b)
Horse 3 0 N/a
Cattle 11 1 Lee et al. (2008)
Table 4. Impact of circadian rhythm on drug pharmacokinetics in major veterinary species
Species No. hits No. relevant hits Reference
Dog 24 3 Hardie et al. (1991), Rackley et al. (1988, 1991)
Cat 263 0 N/a
Swine 319 1 Prémaud et al. (2002)
Horse 157 0 N/a
Cattle 430 0 N/a

Whenever possible, veterinary examples are provided. However, in some cases veterinary examples were not available or human examples were more illustrative. Therefore, to provide a comprehensive discussion of each variable, both human and veterinary examples are explored. Finally, as emphasized in Part I, most of the discussions are addressed from the perspective of target animal safety and effectiveness evaluation, but the implications of altered PK on human food safety and withdrawal times should also be considered when reviewing the information in this article.

Physiological Variables That Can Influence Drug Kinetics

Heritable traits

Variations in DNA sequence can have a major impact on how individuals respond to disease, environmental factors (including bacteria, viruses, toxins, and chemicals), and drugs. Single nucleotide polymorphisms (SNPs) are small variations in the DNA sequence that occur when a single nucleotide is altered. Although SNPs alone can neither explain total genetic diversity nor the genetic susceptibility to complex diseases and adverse drug reactions, by some accounts they are responsible for up to 90% of all human genetic variations (Human Genome Project, http://www.ornl.gov/sci/techresources/Human_Genome/home.shtml).

Genetic-related variability in drug response often reflects population differences in the activity of drug transporters and/or metabolizing enzymes. Variation in the distribution and frequency of occurrence of mutated alleles of genes encoding for drug-metabolizing enzymes is known to alter the rate and extent of drug metabolism. The cytochrome P450 (CYP 450) enzymes, which are involved in the metabolism of >80% of all clinically used drugs (both in human and veterinary medicine), are known to be highly polymorphic, with the notable exception of CYP1A1 and CYP2E1 (Cropp et al., 2008).

Genetic diversity within veterinary species and breed-related differences in pharmacological responses to xenobiotics have been reported across a range of species, including cattle (Sallovitz et al., 2002), sheep (Ammoun et al., 2006), chickens (Opdycke & Menzer, 1984), pigs (Sutherland et al., 2005), and dogs (Paulson et al., 1999;. With regard to food-producing animals, Sallovitz et al. (2002) reported a significantly slower absorption and lower systemic availability of moxidectin in Aberdeen Angus as compared with Holstein calves. Depelchin et al. (1988) reported that antipyrine elimination in Friesian calves was twice that reported in the Blue White Belgian breed, suggesting a breed-related difference in hepatic microsomal oxidative activities. Ripoli et al. (2006) compared the polymorphism of the microsomal enzyme diacylglycerol O-acyltransferase (DGAT1) in 14 populations of cattle from Argentina, Bolivia, and Uruguay. DGAT1 catalyzes the triglyceride synthesis and has an important role in determination of the milk fat content. Previous work on the amino acid sequence for DGAT1 resulted in the identification of a SNP that has been associated with elevated milk fat content (Grisart et al., 2002). Based on a gene frequency analysis, Ripoli et al. (2006) reported significant differences among the breeds in the activity of the DGAT1 enzyme.

Because of the extensive selection pressure exerted by the restrictive breeding of dogs, genetic variability has been particularly accelerated in dogs. Analogous to the Human Genome Project, the Dog Genome Project at the National Human Genome Research Institute resulted in identification of over 2.5 million distinct SNPs mapped to the draft genome sequence, corresponding to an average density of approximately one SNP per kb (Lindblad-Toh et al., 2005). Recently published work has identified 155 genomic regions that present strong signatures of recent selection pressure in dogs (Akey et al., 2010) induced by domestication and carefully controlled breeding patterns.

An extensive summary of canine breed-related differences in drug metabolism and its influence on drug response has been reviewed elsewhere, with references to breed idiosyncrasies in the P450 enzyme system, thiopurine methyltransferase, N-acetyltransferase, as well as breed-related difference in physiology (Fleischer et al., 2008). For example, a significant pharmacogenetic variation in the gene encoding for CYP2D15 enzyme (counterpart to human CYP2D6) has been reported in purebred Beagles. Beagles that have the wild-type gene for CYP2D15, are extensive metabolizers, with a terminal elimination half-life (t½) of celecoxib ranging from 1.5 to 2 h vs. 5 h in poor metabolizers (Paulson et al., 1999). In terms of the breed effect on the PD response, the opioid dysphoria observed in certain dog breeds such as the Labrador Retriever, Alaskan Malamute and Siberian Husky may be attributable to a SNP in the canine μ-opioid receptor gene (Hawley & Wetmore, 2010).

With regard to safety concerns, a 4-bp deletion in the ABCB1 (formerly MDR-1) gene, which encodes for P-glycoprotein (P-gp), results in a nonfunctional gene (Mealey, 2001). This mutation affects 30–50% of the Collie population, as well as other divergent breeds (Neff et al., 2004; Mealey, 2006). Dogs that are homozygous recessive for this mutation exhibit neurotoxicity to the avermectin class of drugs, as well as to many other P-gp substrates. This is due to the inability of the efflux drugs to prevent these P-gp substrates from entering into the central nervous system (CNS). This efflux pump defect can also influence tissue drug concentrations, which may influence drug safety and effectiveness in these dogs (Martinez et al., 2008). In addition, there may be a different genetic mutation in some dogs of noncollie breed that, while not related to the mutation previously recognized in collies, leads to similar pump defects (Bissonnette et al., 2009). This includes subchronic signs of neurotoxicity to macrocyclic lactones, which is a typical clinical sign of the ABCB1 mutation.

Recent efforts to analyze genetic differences across equine breeds (Wade et al., 2009) will likely give rise to the identification of mutations that result in breed-related differences in disease and drug metabolism in horses.

Because of the tremendous increase in information on the influence of genetic attributes on breed idiosyncrasies, a detailed discussion of the relationship between heritable traits and drug exposure–response relationships would require volumes, not pages, of text. Nevertheless, although breed alone cannot explain all genetic variability reported for various animal species, it does provide a starting point for assessing the diversity that can influence drug safety and effectiveness.

Gender

In human medicine, women generally have lower total body weight and organ size, a higher percentage of body fat, lower glomerular filtration rate (GFR), and slower gastric motility when compared with men (Meibohm et al., 2002; Schwartz, 2003). In terms of gender differences in the PK response, men tend to have higher CYP1A2 and CYP2E1 activity. P-gp activity also appears to be greater in men as compared with women. Similar disparities have been observed for glucuronosyltransferases and sulfotransferases. On the contrary, women have higher levels of CYP2D6 activity.

Gender disparity in PK has been identified for numerous drugs; however, it is not common for these differences to result in a substantially different pharmacological response (Thürmann, 2007). Interestingly, a Bayesian statistical analysis of sex differences in the adverse events in human patients showed that, although the adverse events were reported at similar frequencies for men and women, those that are reported by women tended to be of a more serious nature (Miller, 2001).

Gender differences have also been occasionally observed in pharmacological response. For example, women tend to differ from men in their responses to pain therapy, glucose management, and arrhythmia susceptibility (Beierle et al., 1999). Gender differences in drug-induced QTc prolongation (Meibohm et al., 2002) can lead to a greater risk of drug-induced cardiotoxicity in females vs. males across both human and veterinary species. Other examples of PK- and PD-related gender effects in human medicine can be found elsewhere (Meibohm et al., 2002). Accordingly, within the human drug approval process, there is now greater attention given to potential gender differences in drug response. A discussion of observed gender differences in human drug approvals can be found at http://www.fda.gov/cder/reports/womens_health/women_clin_trials.htm.

A listing of the veterinary research articles published on the effect of gender on drug PK is summarized in Table 1. Although findings similar to those reported in human medicine have been observed in veterinary species, it should be noted that many domestic animals (both companion and food-producing) are castrated prior to reaching full maturity. The removal of sex organs results in a depletion of the sex hormones, thereby influencing the magnitude of observed gender effects. Nevertheless, castration does not completely eliminate gender effects, as shown by Hutson et al. (2008) in humans undergoing medical castration. These investigators hypothesized that therapeutic reduction in testosterone concentrations would affect the metabolism of other drugs, as testosterone is a substrate of the CYP3A4 drug-metabolizing enzyme. However, their study results showed that the decrease in testosterone concentrations did not lead to a significant change in the activity of the CYP3A4 enzyme.

Another major physiologic difference between humans and veterinary species is that, unlike humans, many domestic animals are seasonal breeders, with one or more estrous cycles occurring during certain periods of the year. Although seasonal breeders have dormant phases of reproductive cycle, which could suggest potential differences in the PK response due to the different levels of sex hormones, no studies have specifically compared the impact of gender on drug PK between seasonal breeders and continuous breeders.

Among veterinary species, gender-linked differences in PK and drug metabolism have been reported in cats (Erichsen et al., 1980; Lainesse et al., 2007), cattle (Dacasto et al., 2005), dogs (Hay Kraus et al., 2000; Bruss et al., 2004), ferrets (Court, 2001), and fish (Vega-Lopez et al., 2007). Janus and Antoszek (1999) reported marked sex-linked differences in plasma antipyrine clearance and urinary excretion of the main metabolites of antipyrine in cattle over 12 months of age, with females being the more active metabolizers. With respect to the PD-related gender differences, Doursout et al. (1990) reported a statistically significant difference in the response of male vs. female dogs to centrally administered angiotensin II. In male dogs, angiotensin II induced parallel pressor and dipsogenic responses, whereas no hypertension and no increase in fluid intake was observed in females, thus providing an evidence of the role of gender in the physiological properties of centrally administered angiotensin II.

Because the extent and direction of gender differences vary among the veterinary species (Witkamp et al., 1991), the interspecies extrapolation of gender effects should be performed with caution (Christian, 2001). This is especially true when attempting to extrapolate gender effect data derived from rodents. Rats in particular tend to express gender-related differences that do not necessarily translate to similar gender differences in other species (Niwa et al., 1995; Reinoso et al., 2001; Martinez, 2005). This species-specific gender effect has been linked to gender-related differences in the daily rhythm of rat hepatic enzymes and a secretion of the growth hormone (Furukawa et al., 1999; Czerniak, 2001).

Age

Geriatrics. Changes in PK that occur with maturity and senescence are well recognized in both animals and humans, with the following three age-related physiologic processes having the most profound effect on drug PK: (i) decreased plasma protein binding, which affects drug distribution; (ii) declining liver function, which affects drug metabolism; and (iii) impaired kidney function, which delays drug elimination. In addition, age-related changes in body composition (less lean body mass and more fat tissue), as well as a decrease in gastrointestinal motility, further affect the fate of drugs in an aging body (Hilmer et al., 2007). The result is often an increased risk of drug toxicity. Moreover, the heterogeneity observed in older populations dramatically increases in whatever physiological variable is being studied. For this reason, the biggest shift that occurs between drug PK in young and old subjects is often a marked increase in the magnitude of PK variability rather than a substantial shift in the mean dose exposure–response relationship (Kinirons & O’Mahony, 2004).

A significant decrease in hepatic metabolizing capacity that occurs with aging, combined with decreased renal function, can markedly affect the ability of aging individuals to clear substances from the systemic circulation. In general, the t½ of drugs processed by the P450 enzyme system or via renal elimination is 50–75% longer in those human patients ages 65 and older as compared with their younger counterparts (Ginsberg et al., 2005). In addition, liver and kidney diseases are much more common in elderly individuals, further compromising organ and systemic clearance (Lam et al., 1997; Regev & Schiff, 2001).

Another organ system profoundly affected by the human aging process is the CNS, leading to greater risk of drug-related toxicities. With advancing age, the CNS undergoes a variety of changes, including neuronal loss, altered neurotransmitter and receptor levels, and a decreased ability to adapt to changes induced by xenobiotics. For example, a single therapeutic dose of flurazepam administered to people over the age of 60 caused measurable side effects in nearly half of the people, whereas only 5–10% of younger individuals experienced side effects (Liu & Christensen, 2002). Bartels et al. (2008) recently reported on a marked age-related decrease in the P-gp activity of the blood–brain barrier (BBB). The authors suggested that an age-related decrease in BBB functions could explain the increased risk of neurodegenerative diseases in the elderly.

In contrast to the age-related decrease in P-gp activity in the brain, both mouse and human data suggest an association between aging and the increase in P-gp expression in lymphocytes (Witkowski & Miller, 1993; Gupta, 1995). Because P-gp is expressed in a wide variety of tissues and cells, altered expression of P-gp with advancing age may underlie many drug interactions and altered drug effects in older people (Kinirons & O’Mahony, 2004).

In addition to its effect on PK, senescence can affect receptor numbers and affinities in both humans (McLean & Le Couteur, 2004) and animals (The Merck Manual, 2006). For example, the effect of verapamil on prolongation of the PR interval of ECG is significantly less in geriatric adults. However, the optimal verapamil dose tends to remain unchanged because of the reduced verapamil clearance observed in older people (Abernethy et al., 1993). Aging has also been linked to down-regulation of beta-adrenergic receptors, elevated plasma noradrenaline levels, and reduced cAMP response to adrenergic stimulation, possibly explaining the reduced bronchodilatory response of older people to beta-agonists (Scarpace et al., 1991).

As with research of other factors influencing drug PK and PD, we found that the systematic evaluation of effects of aging is lacking in veterinary species. Most of the information available to veterinarians is based on clinical experience with geriatric pets rather than on research findings. Therefore, it has been recommended that veterinarians prescribing drugs for geriatric dogs and cats rely on information from the human drug package insert as guidelines for dose adjustment (Dowling, 2005).

A common practice among veterinary practitioners is to use lower doses of drugs in geriatric patients based on clinical experience from either human or veterinary medicine, Yamashita et al. (2009) recently showed that the minimum alveolar concentration of sevoflurane needed in old dogs is approximately 17–20% lower than that needed in young dogs, confirming clinical observations of increased anesthetic potency with advanced age. This observation is consistent with the increase in CNS drug sensitivity in the human geriatric population. On the contrary, there is evidence suggesting that, despite similar histological changes observed in the glomerulus of aging dogs and humans, and a well characterized decrease in GFR in healthy aging humans (Hoang et al., 2003), only very limited data confirm a corresponding decrease in the canine GFR with age. Bexfield et al. (2008) compared GFR in 118 healthy dogs ages 0–14 years and showed that on the average, less than a 1% decrease in GFR/kg body weight or GFR/extracellular fluid volume occurred as a function of age. However, for dogs weighing 1.8–12. 4 kg, a small negative trend in renal function was observed with age. The authors hypothesized that the decrease in canine GFR associated with aging may not occur to the same extent as that seen in humans due to the dogs’ shorter life expectancy.

Immunosenescence, a gradual deterioration of the immune system associated with aging, has also been reported to affect the PD characteristics of a drug response. Immunosenescence has been studied in both animal models and in humans, and is considered by some to be a major contributory factor to the increased frequency of morbidity and mortality among elderly people. Both cellular- and humoral-mediated responses may be abnormal in the elderly, but the effect of aging on the immune response is highly variable. In their reviews of the effect of aging on host defenses, Scordamaglia et al. (1991) and Pawelec et al. (2002) focused on the role of T cells in aging, as they serve as a marker of the immune response and are also one of the main targets of many therapeutic drugs. Factors contributing to T cell immunosenescence may include an altered production of T cell progenitors, decreased levels of newly generated mature T cells, aging of resting immune cells, and disrupted activation pathways in immune cells. Similar processes have been reported in other types of immune cells, contributing to the increased incidence of morbidity and mortality from infectious disease, and possibly autoimmunity and cancer among elderly people (Mocchegiani & Malavolta, 2004).

Neonates and infants. As compared with the human adult, neonates have higher skin surface area per body weight, a greater percentage of body water, less body lipid, lower renal blood flow, and functionally immature hepatic and renal functions (Bartelink et al., 2006). In humans, the hepatic expression of CYP1A2, 2C, 2D6, 2E1, and 3A4 each develop gradually and at different rates in the postnatal period. Conversely, the expression of CYP3A7 progressively diminishes with maturity (Gow et al., 2001). Hepatic glucuronidation in the human neonate is relatively immature at birth, whereas hepatic sulfation activity is considerably more mature. Aside from immature hepatic and renal function, neonates and infants tend to have a prolonged gastric residence time, a higher gastric pH, slower intestinal motility, and in general, a slower oral absorption of compounds as compared with that observed in adults.

Because of developmental changes in absorptive capacity, first-pass metabolism, distribution, and elimination processes, administration of drugs on the mg/kg basis in young individuals can be inadequate to produce the desired clinical response in adults. Furthermore, because intestinal transit time and absorptive surface area are lower in young organisms as compared with mature ones, drug delivery systems that may be suitable for use in adults may not deliver the total dosage in children. Ultimately, in a manner similar to that seen in the elderly population, drug PK characteristics in infants and neonates are typically more variable than that seen in adults due to differences in rates of development of key physiologic and metabolizing systems.

There is only limited information on the differences that occur in adult vs. immature animals. A summary of studies comparing pharmacokinetics in adult vs. young animals or in various stages of maturation is provided in Table 2. Although these processes are analogous to the early development of human newborns, it should be noted that, in general, young animals are born much more precocious than humans and a rapid maturation of the major elimination organs occurs in the first 6 months of life.

The prolonged elimination t½ in early life, caused mainly by immature hepatic pathways or underdeveloped renal excretion processes, is one of the most pronounced PK differences between newborn and adult animals. Nouws (1992) reviewed the effect of age on half-lives of several antimicrobial drugs in young calves and pigs and reported a steady decrease in t½ for 17 out of 20 antibiotics when comparing 1–2 days of age, week 1, week 2–4, and >8 weeks of age. In addition to the slower elimination in infant animals, the absorption processes are also underdeveloped, including the gastrointestinal blood perfusion and motility. Moreover, the development of reticuloruminal activity takes place over several weeks, during which time a microbial population is established, which can affect the fate of drugs (Nouws, 1992).

Changes in body fluid composition are also responsible for age-related differences in drug PK. A greater proportion of aqueous fluids in the muscle interstitial tissues of newborns as compared with mature animals results in the enhanced spread of drugs at intramuscular injection sites, which in turn leads to an increase in the absorptive surface area and therefore enhanced drug absorption (Nouws et al., 1986). The greater total body water at birth leads to a greater drug volume of distribution for polar compounds (Short & Davis, 1970). In contrast, the amount of body fat is low at birth and increases progressively with maturity, which could lead to a lower volume of distribution for lipophilic compounds in neonates (Nouws, 1992).

While some enzymes are fully developed at birth, the microsomal oxidation activity of the P450 isoenzymes and glucuronide conjugation reactions are underdeveloped in the neonate. In rodents and pigs, the activities of oxidative and conjugative liver enzymes attain adult levels within the 2 months after birth (Short & Davis, 1970). Development of the oxidative metabolic pathways in calves is also a gradual process, achieving maximum capacity at 3–12 weeks of age (depending on the enzyme system) (Davis et al., 1973; Nouws, 1992). Galtier and Alvinerie (1996) studied the age-related changes in hepatic drug metabolizing activities of Lacaune ewes and found that hepatic total P450 concentrations reached maximum levels at 6 years of age, whereas the levels of CYP3A peak during the first 4 weeks of life. The activity of the Phase 2 enzymes also changes with age, with the glutathione S-transferase activity peaking around 11 months of age, and activity of glucuronyl transferase starting to decline at the same time. The results of this study clearly illustrate vast differences in the rate of maturation of various enzyme systems associated with drug elimination.

Age and maturation can also affect active transport mechanisms. For example, calcium absorption, which is characterized by both active and passive processes, varies as a function of maturation. The active component comprises a relatively more significant component of total calcium absorption in neonates, but decreases in its relative contribution to total calcium absorption as the dog matures (Tryfonidou et al., 2002). Several researchers have recently discussed the effect of aging on the expression and activity of the efflux transporter P-gp (Mangoni, 2007; Bartels et al., 2008). In both T-lymphocytes and natural killer cells, the highest levels of activity (as measured by the ability to prevent entry of a fluorescent marker) were highest in human cord blood and progressively declined with age (Machado et al., 2003; Giraud et al., 2009).

Body composition and exercise

Modification of drug dosages in obese human or veterinary patients has become a subject of major concern, particularly for drugs with a narrow therapeutic index. In general, the PK behavior of drugs characterized by low lipophilicity is rather predictable. As they are mostly distributed in lean tissues, dosage can be based on the ideal bodyweight (IBW) (Cheymol, 2000). Therefore, it has been suggested that lean body mass may be a better predictor of drug dosage than either total bodyweight or body surface area (BSA), for hydrophilic compounds (Morgan & Bray, 1994). However, as drug lipophilicity increases, dose adjustment for total body weight may be necessary, although it is the physico-chemical attributes of the drug rather than the degree of lipophilicity that determines the extent to which a drug will distribute into body fat (Cheymol, 2000; Hanley et al., 2010).

Fat is a rather stable reservoir because of a relatively low blood flow. Nevertheless, drug distribution into fat tissue may be rapid. For example, as much as 70% of thiopental is present in body fat only 3 h after administration (Buxton, 2006). Body composition has been shown to affect the PK of drugs, irrespective of the route of administration. Chan et al. (2003) demonstrated that the bioavailability of human chorionic gonadotropin, used to induce oocyte maturation during in vitro fertilization, is significantly affected by the body mass index: for both the intramuscular and subcutaneous routes, the bioavailability was markedly reduced in obese women.

Body condition has also been found to influence the PK of various drugs in veterinary species (as summarized in Table 3), especially highly lipophilic drugs. In the development of new veterinary compounds, where a persistent effect is desirable, the use of lipophilic actives or vehicles is a common method of extending drug availability and effectiveness (Hennessy, 1997). The long residence time, which allows for persistent effectiveness, is influenced by the route of administration, dose, formulation, and animal body composition. Because of their high lipophilicity, macrocyclic lactones are selectively distributed into the fat tissue from which they are gradually released, thus contributing to their prolonged persistent effects (McKellar & Benchaoui, 1996; Craven et al., 2002a,b).

Body condition score and body fat composition affect the rate and extend of exposure of topical eprinomectin in goats (Dupuy et al., 2001), subcutaneous ivermectin, doramectin and moxidectin in sheep (Echeverría et al., 2002; Barber et al., 2003), subcutaneous ivermectin and moxidectin in in pigs (Craven et al., 2002a,b), and oral moxidectin in dogs (Lallemand et al., 2007). These observations raise safety concerns about the use of these compounds in very thin animals, where low body fat can lead to high concentrations of drug in the blood. In addition to the safety risks in thin animals, moxidectin levels were found to be less persistent in thin as compared with fat animals, suggesting concerns about the product’s effectiveness, and possibly contributing to the emergence of resistance to these compounds. It is important to note that of the macrocyclic lactones used in veterinary medicine, moxidectin has higher lipophilicity than any of the other avermectins.

Similarly, amikacin, a molecule, that is sparingly soluble in water, has a substantially lower volume of distribution in Greyhounds as compared with Beagles (KuKanich & Coetzee, 2008), reflecting breed differences in lean body mass (Fleischer et al., 2008). Similar differences in the volume of distribution of Greyhounds vs. other canine breeds have been shown for several other lipophilic compounds (Robinson et al., 1986; Sams & Muir, 1988; Zoran et al., 1993).

Unlike highly lipophilic drugs, hydrophilic compound typically have a lower volume of distribution per kg body weight (Wright et al., 1991). For example, when gentamicin is dosed on the basis of total body weight to cats, there is a high risk of drug toxicity due to an over-estimation of the volume of distribution. Therefore, a more appropriate dosing for drugs such as cancer chemotherapeutics may be calculated on the basis of BSA rather than total body weight. However, such dosing strategies may not be equally applicable across all drug types and animal species. Frazier and Price (1998) concluded that dosing on the basis of BSA may be appropriate for drugs that are eliminated unchanged by glomerular filtration or are degraded via nonenzymatic processes. In contrast, it may be sub-optimal when administering drugs that are extensively metabolized (which tend to be more highly lipophilic). Therefore, it is not surprising that PK and pharmacological studies in humans do not support the global application of BSA dosing strategies for chemotherapeutic agents (Kouno et al., 2003; Kaestner & Sewell, 2007).

Because of their unique physiology, there have been reports on the impact of body composition (in terms of degree of hydration) on drug PK in camels. When tylosin tartrate was administered by i.v. injection to normal vs. water-deprived camels, serum tylosin concentrations in the water-deprived camels were significantly higher, the rate of drug elimination was slower, the volume of distribution was significantly smaller, and the total body clearance was significantly slower as compared with that observed in normal camels (Ziv et al., 1995). Similar results were reported in cattle after the administration of caffeine (Janus et al., 2001). It was suggested that even short-term (4-day) water or food deprivation can lead to an inhibition of the P450 system.

Finally, exercise-induced increases in muscular blood flow has a complex effect on drug PK, ranging from an increase in the binding of digoxin (leading to lower serum drug concentrations) to a decrease in hepatic blood flow which results in a decrease in the clearance of high extraction ratio drugs and consequently higher serum drug concentrations (e.g. propranolol) (Khazaeinia et al., 2000; Lenz et al., 2004). Exercise can also decrease the clearance of drugs that undergo renal elimination and can increase biliary-related drug clearance (Khazaeinia et al., 2000; Lenz et al., 2004). On the contrary, chronic exercise can increase drug absorption by increasing collateral blood flow and by changing gastrointestinal transit times. Chronic exercise can change drug distribution characteristics by increasing lean body mass, decreasing body fat, increasing plasma protein, and increasing plasma volume (Persky et al., 2003).

Circadian rhythm

In most mammalian species, numerous physiological activities and organ systems are influenced, at least to some extent, by circadian rhythms, including heart rate, blood pressure, liver and renal plasma flows, bile and urine production, intestinal peristalsis, secretion of digestive enzymes into the GI tract, major endocrine functions, immune cell production, and metabolism (Levi & Schibler, 2007; Xu et al., 2008). It is, therefore, no surprise that the effect of circadian rhythms have been found to influence drug response (Beauchamp & Labrecque, 2007).

Chronopharmacology is the study of rhythmic, predictable-in-time differences in the effects and/or PK of drugs. Chronotoxicology describes the effect of the body’s endogenous circadian rhythms on drug toxicity. Daily variations in drug effectiveness and variations in PK across a wide range of therapeutic classes have been observed in humans (Lemmer, 1996). In fact, most of our knowledge with regard to chronopharmacology and chronotoxicology is built upon information gathered in people. For example, in a study of 179 patients with serious infection, Prins et al. (1997) determined that gentamicin- or tobramycin-induced nephrotoxicity was significantly greater when administered between midnight and 7:30 a.m. as compared with dosing at any other time of day. Other researchers have confirmed that the nephrotoxicity of aminoglycosides is minimized when they are administered in the early afternoon (Rougier et al., 2003; Beauchamp & Labrecque, 2007). Beauchamp and Labrecque (2007) suggested that possible mechanisms of the chronotoxicity of aminoglycosides include the daily variation in aminoglycoside pharmacokinetics, food intake, and the effect of diurnal variations in urine pH.

Chronotoxicity has been studied extensively in rats and mice. For example, it has been shown that approximately 75% of rodents studied survived the administration of 3′azido-3′-deoxythymidine when administered at the beginning of the resting period, but only 13% of treated rats survived when the drug was administered during the early phase of activity (Zhang et al., 1993). Similar results were collected when amphotericin B was administered to rats at resting vs. active periods (Skubitz et al., 1986; LeBrun et al., 1996).

Research on the effects of circadian rhythm on drug PK/PD is complicated by the circadian pattern associated with many diseases. In human medicine, disease areas for which chronotherapy has been identified as a major consideration include cancer, immune diseases, and cardiovascular diseases (Xu et al., 2008). With regard to pain control, Bruguerolle and Labrecque (2007) suggest that a primary reason that treatments fail to provide adequate pain control is a failure to recognize time-of-day patterns in pain intensity and the medication requirements based on the circadian rhythm of pain and PK. While rodent species exhibit their highest pain threshold at the end of the resting period and their lowest threshold at the end of the activity period (Frederickson et al., 1977; Kavaliers & Hirst, 1983; Gschossmann et al., 2001), human studies show a morning peak and/or an evening trough of pain (Bruguerolle & Labrecque, 2007). Bruguerolle & Labrecque further concluded that the impact of circadian rhythm is dependent upon the type of pain. For example, the perception of pain peaked in the morning in patients with angina pectoris, myocardial infarction, migraine, rheumatoid arthritis, and toothache, whereas it peaked in the evening/nighttime in patients with biliary colic, cancer, and intractable pain.

Circadian rhythms not only affect different types of pain differently, but also influence the selection of analgesics. For opioids, the circadian rhythmicity of pain perception seems to be related to the time-dependent variation in the brain levels of opioid peptides, which in humans are higher at the beginning of the day and lower in the evening (Labrecque & Vanier, 2003). Based on the recent review of pain and chronotherapy, animal and human studies are in agreement about the role of circadian variations in the PK and PD of NSAIDs, with bioavailability being greatest in the morning (Bruguerolle & Labrecque, 2007). For instance, Ohdo et al. (1995) demonstrated that the mortality induced by acetylsalicylic acid is highest when administered at the end of the resting period of rats; others have demonstrated that the effectiveness of NSAIDs varies by 40–50% according to the time of their administration (Labrecque et al., 1995).

With regard to veterinary species, as seen in Table 4, diurnal variability has been extensively studied in both food and companion animals. However, studies have focused primarily on such factors as endogenous substances (hormones), brain activity, sleep, exercise, physiological processes (including heart rate, blood pressure, body temperature, milk production), and reproduction. Only a handful of studies report chronopharmacology and chronotoxicology evaluations. In fasted dogs administered pentazocine in the morning (8 a.m.), or during the dark cycle (8–9 p.m.), there was a statistically significant difference in the estimates of volume of distribution and t½ (highest at night), although Cl and AUC remained unchanged (Ritschel et al., 1980). The time of cisplatin intravenous administration also affected drug PK and renal toxicity in dogs, where renal clearance was highest at night but nephrotoxicosis and free cisplatin drug concentrations were greatest in the morning (Hardie et al., 1991). In pigs, methotrexate (but not vinorelbine) exhibited very marked circadian rhythms in the plasma drug concentration–time profiles, with the highest concentrations generally occurring around midnight (Prémaud et al., 2002).

Part of the problem with studying diurnal rhythms in animals is that under real-life conditions, the inherent variability (particularly in companion animals) is confounded by human interactions. Diurnal pH changes in the duodenum of the conscious dogs are linked to digestion (Itoh et al., 1980). Similarly, studies to track canine diurnal changes in plasma concentrations of adrenal corticotrophic hormone (ACTH) and cortisol appear to be confounded by human activity (Castillo et al., 2009). However, dogs do show marked diurnal variation in theophylline plasma drug concentrations: in a study following constant aminophylline infusion for 48 h, peak concentrations occurred between 24 and 30 h after administration and troughs at approximately 36–42 h after the start of drug administration (Rackley et al., 1988). In one dog, whose infusion was initiated in the p.m. rather than the a.m., times of peaks and troughs relative to the time of start of infusion were reversed, indicating that it was more the time of day rather than time relative to the start of infusion that determined the time of peak and trough concentrations. Interestingly, this variation may be, at least in part, related to food intake because despite its intravenous administration, the major route of theophylline elimination (the kidney) is affected by food. This link appears to be a consequence of food-induced changes in urine pH, resulting in a greater proportion of the drug existing in its ionized state and therefore an increase in its urinary elimination (Rackley et al., 1991). Thus, at least in part, food consumption had an important role in promoting the chronopharmacokinetics of this compound. Similar results on the crucial importance of the food effect in the temporal variation of renal toxicity were reported for amphotericin B in rats (LeBrun et al., 1999).

In cats, inherent diurnal patterns in cardiovascular variables and motor activity appeared to be negated by the effects of human interactions (Brown et al., 1997). Hormonal releases (e.g. melatonin and prolactin) are regulated primarily by the duration of the light/dark cycles (Leyva et al., 1984). Therefore, when evaluating chronopharmacology and chronotoxicology in domesticated animals, there is a fine line between evaluating the inherent patterns that might exist in the wild vs. the patterns that are likely to due to human–animal interactions.

Despite these challenges, physiological information in veterinary species and chronopharmacology and chronotoxicology information derived in humans and rodents confirm that for some compounds, circadian rhythms can influence drug exposure, safety, and effectiveness. Accordingly, this phenomenon should not be ignored as a potential source of variability in veterinary medicine.

Population Predictions

Identifying variables can markedly influence drug exposure characteristics is an important step in predicting the relationship between drug physico-chemical characteristics, its elimination pathways, and the variability likely to occur when that drug is used under clinical field conditions. This is particularly important when considering the very small clinical trials generally conducted in veterinary medicine. The small sample size associated with the clinical trial population renders it unlikely that the study will have the power to identify covariates that may necessitate dose adjustments. Data from clinical trials are generally pooled together and analyses generated with an assumption that all of the data are collected from a single normal (log-normal) distribution. With this practice in mind, we considered it important to demonstrate how the typical practice of pooling data can lead to large errors in population predictions.

To illustrate how failure to identify the presence of subpopulations can greatly distort expectations, we simulated a population of 10 000 dogs, where 70% of the dogs were extensive metabolizers (EM) and 30% of the dogs were poor metabolizers (PM). The relationship between the clearances of the two subpopulations (clearance of EM group = 2 ×  clearance of PM group) and the between-animal variability associated with each of the clearance estimates (25% CV) was based upon the relationship between PM and EM reported by Paulson et al. (1999), where they examined polymorphism in the metabolism of celecoxib in dogs. The results of this simulation are provided in Fig. 1. A summary of how to interpret the symbols used in boxplots is provided in the appendix.

Details are in the caption following the image

Illustration of the importance of recognizing the presence of a disparate subpopulation. The distribution of 10 000 AUC values was simulated on the basis of 30% of the population being poor metabolizers (PM) and the other 70% being extensive metabolizers (EM). Based upon the data from Paulson et al. (1999), we assumed that the clearance of the EM group was approximately double that of the PM group. As with the celecoxib example, it was assumed that the %CV for clearance was about 25% in both groups.

The important point to note is that a failure to recognize the presence of two subpopulations (fast and slow metabolizers) can introduce substantial bias into our dose-exposure evaluations. When the EM and PM groups are separated, the AUC values collected in the EM group are markedly lower than are those collected in the PM group. Accordingly, for one or both groups, dose adjustments may be needed to achieve the expected safety and effectiveness. The divergence in exposure that can occur within the patient population would be overlooked if we simply pooled data from all 10 000 simulated animals and ignored the pharmacokinetic differences.

While this is simply a simulated example, the hope is that it will underscore the importance of considering the numerous variables that can affect the dose exposure–response relationship in veterinary medicine, as described in this manuscript and in Part I of this series.

Conclusions

Numerous factors can influence drug PK and the response to pharmacological therapy. Identifying these variables is an important step in predicting the relationship between drug physico-chemical characteristics, the elimination pathways, and the variability likely to occur when a drug is used in the target population. Considering the small number of animals generally included in veterinary clinical field trials, potentially important sources of population variability in the PK and PD response may be undetected. For that reason, it is our ultimate hope that this series on patient variation will challenge veterinary pharmacologists to consider these factors across a range of target animal species. Understanding and appreciating the impact of patient variability will ultimately allow us to develop more reliable predictions of drug safety and effectiveness under the clinical conditions of use.

Appendix

Interpretation of a boxplot graph

A boxplot provides a graphical presentation of descriptive statistics such as the mean, lowest 25% (first quartile, or Q1), upper 75% (third quartile, or Q3), and median values. The box itself contains the middle 50% of the data and is called the interquartile range (IQR). It represents values associated with the middle 50% of the population (i.e. Q3–Q1). Within this box, the median is represented by a line and the mean is represented by a plus sign (+). If the median line within the box is not equidistant from the edges, then the data are skewed. Any data observation which is less than the Q1 or higher than the Q3 (but is greater than Q1 − 1.5 × IQR and less than Q3 + 1.5 ×  IQR) is connected to the box with a vertical line (‘whisker’). Extreme outliers are more than three times the IQR from Q1 and Q3. These values are indicated by an asterisk. Values that are more than 1.5 times the IQR (but less than three times the IQR) from Q1 and Q3 are indicated by an open circle (○).

    The full text of this article hosted at iucr.org is unavailable due to technical difficulties.