Abstract
Identifying any extrahepatic excretion phenomenon in preclinical species is crucial for an accurate prediction of the pharmacokinetics in man. This understanding is particularly key for drugs with a small volume of distribution, because they require an especially low total clearance to be suitable for a once-a-day dosing regimen in man. In this study, three animal scaling techniques were applied for the prediction of the human renal clearance of 36 diverse drugs that show active secretion or net reabsorption: 1) direct correlations between renal clearance in man and each of the two main preclinical species (rat and dog); 2) simple allometry; and 3) Mahmood's renal clearance scaling method. The results show clearly that the predictions to man for the methods are improved significantly when corrections are made for species differences in plasma protein binding. Overall, the most accurate predictions were obtained by using a direct correlation with the dog renal clearance after correcting for differences in plasma protein binding and kidney blood flow (r2 = 0.84), where predictions, on average, were within 2-fold of the observed renal clearance values in human.
Introduction
In the last 10 years, the pharmaceutical industry has become a very competitive area facing new challenges. The costs of discovering and developing a new chemical entity have quadrupled since 1987, mainly because of attrition during the development process (DiMasi et al., 2003). Although reasons for attrition are varied, including portfolio decisions and lack of clinical efficacy of the biological mechanism, many reasons for compound failure are still related to unfavorable absorption, distribution, metabolism, and excretion properties of new drug candidates. This has generated intense efforts to identify potential absorption, distribution, metabolism, and excretion challenges with significant intellectual and financial investment in optimizing pharmacokinetic parameters of molecules at the earliest stages of discovery, before any costly clinical trials are conducted.
The concept of body clearance not only includes metabolic clearance (hepatic or extrahepatic) but also the excretion of unchanged drug in the bile or the urine. Therefore, identifying and understanding any extrahepatic excretion phenomenon in preclinical species is crucial for an accurate prediction of the pharmacokinetics in man. This is particularly the case for drugs with a low volume of distribution, because they require an especially low total clearance to be suitable for a once-a-day dosing regimen in man. As a consequence, the probability of encountering renal clearance in man and the impact it would have on the half-life requires investigation at an early stage in the discovery process.
Renal clearance of xenobiotics involves several major processes, i.e., passive glomerular filtration, active tubular secretion, and passive and active reabsorption. As a result of the complexity of this phenomenon, various animal scaling approaches have been used in this study for the prediction of renal clearance in man for a diverse set of 34 marketed and 2 experimental drugs (Beaumont et al., 2000; Webster et al., 2003). The most basic animal scaling technique consists of using pharmacokinetics parameters obtained in preclinical species as a prediction for human (Boxenbaum, 1982). Furthermore, simple allometry scaling uses the combination of values measured in several preclinical species to predict man (Boxenbaum, 1980). This approach, based on body weight, has been used to predict total clearance in man (Kaye et al., 1997). This scaling method is based on the following function: Y = a(W)b, where Y is clearance, W is body weight, and a and b are the coefficient and exponent of the allometric equation, respectively. However, pharmacokinetic parameters cannot always be accurately predicted using this approach, and several modified methods have been suggested over the years.
Mahmood (1998) published a study in which he adapted simple allometry to predict the renal clearance of 10 actively secreted drugs. To that end, a normalizing factor was applied to the renal clearance of species that exhibit active renal secretion. This factor took into account glomerular filtration rate (GFR), kidney blood flow (KBF), body weight, and kidney weight. Despite the various studies published on the use of allometric scaling for the prediction of clearance, surprisingly, very little is known about the ability of this technique to predict accurately the renal clearance of a diverse set of drugs (acids, bases, neutrals, and zwitterions) that exhibit both active secretion and net reabsorption. The aim of this study was to use measurements of renal clearance in the two most common preclinical species (rat and dog) and to investigate direct correlations with renal clearance in man, simple allometry, and Mahmood's approach using a normalizing factor to assess the best method for predicting human renal clearance in man.
Materials and Methods
The renal clearance data for a diverse set of 36 drugs in rat, dog, and human were obtained from the literature or measured in-house. Plasma-free fraction in the three species was also obtained for the set of 36 drugs and was used in conjunction with the renal clearance measurements for scaling purposes. Where no data were available in the literature, plasma protein binding (PPB) was measured in-house.
Chemicals.
All compounds were obtained from Sigma-Aldrich (Poole, UK) with the exception of ramatroban and ampicillin, which were synthesized at AstraZeneca (Loughborough, UK). Sodium bicarbonate, ethanol, and phosphoric acid were obtained from Fisher Scientific (Loughborough, UK). Methanol was purchased from Romia (Cambridge, UK). Formic acid was obtained from VWR International (Poole, UK).
Pharmacokinetics Studies.
All in vivo work was subject to internal ethical review and conducted in accordance with Home Office requirements under the Animals Scientific Procedures Act (1986).
Bile duct-cannulated rats.
Male rats were obtained from Charles River (Margate, UK). They were housed in a light-controlled room, kept at a temperature of 18 ± 2°C, and kept at a level of 55 ± 10% humidity. They received a RM3 diet (Special Diet Services, Witham, UK) and had access to water ad libitum. After at least 1 week of acclimatization, rats (250–350 g) were surgically prepared under isoflurane anesthesia. The bile duct was cannulated, and cannulae were also implanted into the jugular vein (dosing cannula) and carotid artery (blood sampling cannula). Dose solutions were administered via the intravenous (jugular vein) cannula as a bolus dose of 1 mg/kg. Serial plasma samples (200–300 μl) were taken from the intra-arterial (carotid artery) cannula. Approximately 200 μl of blood was drawn through the cannula before taking a sampling aliquot to ensure circulating blood was sampled through the cannula. Blood samples were taken over the time course of 2, 4, 8, 15, 30, 60, 120, 180, 240, 300, 360, and 420 min. Bile was collected over 7 h. Rats were kept in metabolism cages over 7 h for urine collection onto dry ice.
Bile duct cannulated dogs.
Male beagle dogs were obtained from Harlan (Bicester, UK). They were housed in a light-controlled room, kept at a temperature of 18 ± 2°C, and kept at a level of 55 ± 10% humidity. They received a Teklad 2021 diet (Harlan) and had access to water ad libitum. After at least 4 weeks of acclimatization, a chronic bile duct cannulation was performed on two dogs following the technique described by Kissinger et al. (1998). Dogs were allowed to recover from surgery for a minimum of 1 month.
The dose (1 mg/kg) was administered to the dogs (n = 2) via a 30-min infusion in the cephalic vein. Approximately 2.5 ml of blood was collected on EDTA via the jugular vein 0, 15, 30, 60, 120, 180, 300, 420, 720, and 1440 min after the beginning of the infusion. Bile was collected over 24 h. Dogs were kept in metabolism cages over 24 h for urine collection onto dry ice.
Sample analysis.
Plasma, bile, and urine samples were prepared within 30 min of collection according to the following procedure. Blood was centrifuged at 1110g for 10 min at 4°C. Plasma was transferred into plain polypropylene tubes, each prepared in advance to contain 1.5 μl of concentrated phosphoric acid to stabilize any potential acyl glucuronide metabolites. Samples were then immediately frozen upright on dry ice and stored at −20°C. Bile and urine samples were treated with phosphoric acid, frozen, and stored as described previously for plasma.
Concentrations of drugs were determined by a high-performance liquid chromatography -mass spectroscopy method. Compounds were extracted from their biological matrix (50 μl) after addition of methanol containing an internal standard (150 μl). Samples were frozen for at least 1 h at −20°C to precipitate the proteins. They were then centrifuged at 2050g for 20 min at 4°C. Ten microliters of supernatant was injected into the mass spectrometer. The mobile phases were water and methanol, both containing 0.1% formic acid (v/v). The column used was a Waters Symmetry C8 (3.5 μm, 2.1 × 30 mm; Waters, Milford, MA). For detection, we used a Quattro Ultima (Micromass; Waters) in negative electrospray ionization mode with data analysis on Quanlynx software (version 4.0, Micromass; Waters).
Pharmacokinetic analysis.
Pharmacokinetic parameters were calculated using WinNonlin (version 3.2; Pharsight, Mountain View, CA). Plasma clearance (CLp) was determined from the dose and plasma AUC0-∞ with less than 20% extrapolation in all cases. Renal clearance was calculated using eq. 1.
where CLr and CLp are the renal and total clearances, respectively, expressed in milliliters per minute per kilogram.
Plasma protein binding determination.
Where no data were available in the literature, PPB was measured in human, dog, and rat following the method described by Fessey et al. (2006).
Animal Scaling.
Renal clearance in man was predicted by using the following three allometric approaches. Scaling was performed with both unbound and bound renal clearance to assess the influence of differences in plasma protein binding.
Method I: Direct correlations.
Renal clearance values in man were estimated from the rat or dog renal values. Renal clearances were predicted directly from the preclinical species using the following simple equation (eq. 2).
Corrections for PPB differences were calculated using eq. 3.
Additional corrections for kidney blood flow differences were calculated as follows (eq. 4).
where CLrHuman and CLrSpecies, fuHuman and fuSpecies, and KBFHuman and KBFSpecies are the renal clearance, fraction unbound in plasma, and kidney blood flows in man and rat or dog, respectively.
Method II: Simple allometry.
The following allometric equation (eq. 5) was used to predict renal clearance in man.
where CLr is renal clearance, W is body weight, and a and b are the coefficient and exponent of the allometric equation, respectively. When correcting for PPB, unbound renal clearance in human was predicted from unbound renal clearance in rat and dog using eq. 5. Predicted total renal clearance values were then estimated using eq. 6.
Method III: Mahmood's Renal clearance scaling method.
Mahmood's method to predict renal clearance is based on the simple allometry method, but it takes into account the kidney physiological differences between human and preclinical species. First, a species-specific factor (SSF) is calculated for every species (see eq. 7). This coefficient is then divided by the value obtained for human to produce a correction factor (see eq. 8).
Values used for rat, dog, and human KBF were 52 (Hsu et al., 1975), 22 (Keil et al., 1989), and 16 ml · min−1 · kg−1 (Wolf et al., 1993), respectively. GFR values used for rat, dog, and human were 5.2, 3.2, and 1.8 ml · min−1 · kg−1 (Mahmood, 1998), respectively. Renal clearance values of all the species that exhibit active secretion (CLr > GFR × fu) are normalized using the above correction factor and used in a simple allometry (see eq. 5). When using unbound clearances, predicted values were treated as described previously (see eq. 6).
Statistics.
The accuracy of prediction was compared between the three different methods. Precision of each approach was estimated by measuring the root mean squared error (rmse) and the average fold error (afe), respectively.
The bias of each method was expressed as follows:
This parameter explains how the data are related to the line of unity. A bias greater than 1 means that the method overpredicts the observed results. However, if the bias tends to 0, then the method underpredicts the actual results.
Results
Human, dog, and rat renal clearance and plasma-free fraction values for the 34 marketed and 2 experimental drugs are shown in Table 1. Data from this table with references can be found in Supplemental Table 1. Figure 1 shows that 22, 20, and 26 of the 36 drugs have unbound renal clearance values greater than GFR for human, dog, and rat, respectively. Therefore, the 36 drugs represent chemistries of different charge type and exhibit examples of active secretion and net reabsorption.
Renal clearance and plasma-free fraction values
Human unbound renal clearance versus rat (open symbols) or dog (filled symbols) unbound renal clearance. Circles, acids; squares, bases; triangles, neutrals; diamonds, zwitterions; solid line, human GFR; dotted line, dog GFR; dashed line, rat GFR.
The predictions from the different scaling methods versus the measured human renal clearance values are represented graphically in Figs. 2 and 3, and the statistical analysis of these predictions is summarized in Table 2. Rat and human renal clearances are weakly correlated with one another (r2 = 0.52) (Fig. 2a; Table 2). Human renal clearance covers 3 orders of magnitude versus 5 for rat, leading to a flat relationship. Moreover, the high/low clearance compounds in rat tend to be severalfold higher/lower than their corresponding human values. The overall bias of the human versus rat correlation is 2.5, suggesting that, on average, rat has higher renal clearance than human (Table 2). Correcting for differences in PPB slightly tightens the correlation (r2 = 0.56); however, several organic anions still seem to be underpredicted in man using the adjusted rat renal clearance (Fig. 2c). Whereas some improvement is afforded by adjusting rat renal clearance for PPB differences, an overall bias still exists. Adjusting rat renal clearance for both PPB and KBF differences reduces the bias down to 1.4 (Fig. 2e; Table 2). The majority of the human and rat data now lie along a line of unity; however, the organic anions, ibuprofen, indomethacin, losartan, and ramatroban, remain outliers. Therefore, using this method would predict human renal clearance well for a diverse set of drugs but may also underpredict certain organic anions. Dog and human renal clearances are also weakly correlated (r2 = 0.51) (Fig. 2b; Table 2). The overall bias of the human versus dog correlation is 2.3, suggesting that, on average, dog has higher renal clearance than human (Table 2). Correcting for differences in PPB dramatically tightens up the correlation (r2 = 0.84), and bias is reduced to 1.3 (Fig. 2d). The aforementioned organic anions are no longer outliers, and bias is further reduced to 1.1 by adjusting dog renal clearance for both PPB and KBF differences (Fig. 2f; Table 2). The human and adjusted dog data now lie along a line of unity with an afe of 2.2, suggesting that this method predicts human renal clearance very well for a diverse set of drugs.
Measured human renal clearance versus measured renal clearance from rat (a), dog (b), rat corrected (corr) for fu differences (c), dog corrected for fu differences (d), rat corrected for both KBF and fu differences (e), and dog corrected for both KBF and fu differences (f). Circles, acids; squares, bases; triangles, neutrals; diamonds, zwitterions.
Measured human renal clearance versus predicted renal clearance from simple allometry (a), simple allometry corrected (corr) for fu differences (b), Mahmood renal allometry (c), and Mahmood renal allometry corrected for fu differences (d). Circles, acids; squares, bases; triangles, neutrals; diamonds, zwitterions.
Statistical analysis of scaling methods
Applying simple allometry to the rat and dog renal clearance data to predict man does not improve the accuracy of the results. The average fold errors obtained with the simple allometry technique are 4.9 and 3.4 when using bound and unbound data, respectively (Fig. 3, a and b; Table 2). However, there is no bias in the predicted human renal clearance from unbound allometry compared to the observed values. Finally, Mahmood's method offers a slight improvement over simple allometry with average fold errors of 4.1 and 3.2 when using bound and unbound data, respectively (Fig. 3, c and d; Table 2). The two allometry methods show weaker statistics compared with predicting human renal clearance from dog renal clearance after adjusting for PPB and KBF differences.
Discussion
This study highlights the important influence that plasma protein binding has on active secretion as well as passive renal clearance processes in rat, dog, and human. It is generally accepted that only unbound drug is subject to glomerular filtration; however, there is less agreement on the influence of plasma protein binding on active secretion. Bow et al. (2006) recently demonstrated that the in vitro uptake of the acid molecule Ochratoxin A by organic anion transporters (OATs) 1, 3, and 4 is virtually eliminated by an albumin concentration equivalent to 10% of that present in plasma. These recent findings suggest that the extent of plasma protein binding can affect the active secretion of drugs. The evidence presented herein strongly suggests that corrections should be applied for differences in plasma protein binding when predicting human renal clearance from rat and dog data.
The bias that is observed when predicting human renal clearance from either rat or dog renal clearance is in best accordance with interspecies differences in KBF and less so with GFR. The ratio of rat-to-human KBF is similar to the ratio of rat-to-human GFR, and both are consistent with the bias observed between the two species. However, KBF is very similar between dog and human, 22 versus 16 ml · min−1 · kg−1, which is consistent with the small bias of 1.3 (Table 2) obtained when predicting human renal clearance directly from dog and correcting for free fraction. GFR in dog is approximately 2-fold greater than in human, and if this correction was applied, it would lead to an underprediction of renal clearance in man. One may expect GFR to be appropriate for correcting compounds that undergo exclusively passive filtration; however, the renal clearance of actively secreted compounds will be restricted by the flow of blood through the kidney rather than filtration through the glomerulus.
The excellent correlation that is observed between human and dog renal clearance, after corrections for plasma-free fraction differences, suggests good species crossover between the transporters involved in any active processes for this diverse set of compounds. Renal clearance in the rat generally correlates well with man; however, certain organic anions seem to be outliers. Therefore, the extent of renal clearance in man may be underpredicted when using rat as a species for certain chemistries. There is some evidence in the literature to suggest that rat may not have good crossover to man. Tahara et al. (2005) showed that there is a poor correlation between human and rat OAT3-mediated transport activities for nine substrates. Also, Kato et al. (2002) showed that the renal clearance of a series of organic anions, which are all substrates for organic anion-transporting polypeptide 1(oatp1), was much higher in female than in male rats. It was also shown that expression of oatp1, which is localized at the apical plasma membrane of the kidney, was higher in the kidneys of male rats. A large quantity of the literature rat renal clearance data used in this work was obtained from male rats. Therefore, one plausible hypothesis for the underprediction of human renal clearance from rat, for the organic anions, is that reuptake within the kidney is much lower in human relative to male rats. The disconnect between rat and human for the organic anions will distort predictions using simple allometry and Mahmood's method, decreasing the accuracy of these two techniques, compared to the dog approach. Finally, very little has been published to date on the prediction of human renal clearance for organic anion compounds specifically. Mahmood applied his method to 10 molecules, of which only 3 of these were acids and did not correct for differences in PPB. Moreover, Mahmood (2006) suggests that the free fraction corrected intercept method (FCIM) postulated by Tang and Mayersohn (2005) may not be suitable for renally secreted drugs. Our study shows that Mahmood's approach is reasonably predictive, except for nonsteroidal anti-inflammatory drugs such as ibuprofen, as long as corrections are made for PPB differences. An analysis of our data using FCIM gave very similar statistics to Mahmood's approach correcting for differences in PPB, suggesting that FCIM is just as reasonable as Mahmood's approach for predicting renally secreted drugs.
In the absence of species differences in PPB, laboratories would be best served to apply the direct correlation between human and dog renal clearance without the need for PPB data. When observing significant differences in PPB, the data suggest that correcting for these differences is essential for predicting human renal clearance. This makes it particularly suitable for drug discovery; indeed, it permits an early screening of test compounds for their risk of exhibiting high renal clearance in dog and hence in man. Moreover, it also meets the requirement of reducing animal use. However, a caveat must be applied to this method when used to test compounds that may exhibit or be suspected of having differences in renal transporter activity between dog and human.
Authorship Contributions
Participated in research design: Paine, McGinnity, and Riley.
Conducted experiments: Ménochet and Denton.
Performed data analysis: Paine, McGinnity, and Denton.
Wrote or contributed to the writing of the manuscript: Paine and Riley.
Acknowledgments
We thank Roger E. Fessey for generating the plasma protein binding data.
Footnotes
Article, publication date, and citation information can be found at http://dmd.aspetjournals.org.
doi:10.1124/dmd.110.037267.
↵
The online version of this article (available at http://dmd.aspetjournals.org) contains supplemental material.-
ABBREVIATIONS:
- GFR
- glomerular filtration rate
- CLr
- renal clearance
- FCIM
- free fraction corrected intercept method
- fu
- unbound plasma fraction
- KBF
- kidney blood flow
- PPB
- plasma protein binding
- rmse
- root mean square error
- afe
- average fold error
- OAT
- organic anion transporter
- oatp1
- organic anion-transporting polypeptide 1.
- Received November 16, 2010.
- Accepted February 28, 2011.
- Copyright © 2011 by The American Society for Pharmacology and Experimental Therapeutics
