Article pubs.acs.org/molecularpharmaceutics
Assessing the Risk of pH-Dependent Absorption for New Molecular Entities: A Novel in Vitro Dissolution Test, Physicochemical Analysis, and Risk Assessment Strategy Neil R. Mathias,*,† Yan Xu,† Dhaval Patel,† Michael Grass,‡ Brett Caldwell,‡ Casey Jager,‡ Jim Mullin,‡ Luke Hansen,‡ John Crison,† Amy Saari,† Christoph Gesenberg,† John Morrison,§ Balvinder Vig,† and Krishnaswamy Raghavan† †
Drug Product Science & Technology Department, Bristol-Myers Squibb Co., New Brunswick, New Jersey 08903, United States Preclinical Candidate Optimization Department, Bristol-Myers Squibb Co., Wallingford, Connecticut 06492, United States ‡ Bend Research Inc., 64550 Research Road, Bend, Oregon 97701, United States §
ABSTRACT: Weak base therapeutic agents can show reduced absorption or large pharmacokinetic variability when coadministered with pH-modifying agents, or in achlorhydria disease states, due to reduced dissolution rate and/or solubility at high gastric pH. This is often referred to as pH-effect. The goal of this study was to understand why some drugs exhibit a stronger pH-effect than others. To study this, an API-sparing, two-stage, in vitro microdissolution test was developed to generate drug dissolution, supersaturation, and precipitation kinetic data under conditions that mimic the dynamic pH changes in the gastrointestinal tract. In vitro dissolution was assessed for a chemically diverse set of compounds under high pH and low pH, analogous to elevated and normal gastric pH conditions observed in pH-modifier cotreated and untreated subjects, respectively. Represented as a ratio between the conditions, the in vitro pH-effect correlated linearly with clinical pH-effect based on the Cmax ratio and in a non-linear relationship based on AUC ratio. Additionally, several in silico approaches that use the in vitro dissolution data were found to be reasonably predictive of the clinical pH-effect. To explore the hypothesis that physicochemical properties are predictors of clinical pH-effect, statistical correlation analyses were conducted using linear sequential feature selection and partial least-squares regression. Physicochemical parameters did not show statistically significant linear correlations to clinical pH-effect for this data set, which highlights the complexity and poorly understood nature of the interplay between parameters. Finally, a strategy is proposed for implementation early in clinical development, to systematically assess the risk of clinical pH-effect for new molecular entities that integrates physicochemical analysis and in vitro, in vivo and in silico methods. KEYWORDS: pH-effect, risk assessment, in vitro−in vivo correlation, dissolution, supersaturation, precipitation, physicochemical properties, achlorhydria
■
INTRODUCTION Gastric pH-modifying agents such as proton pump inhibitors, H2-antagonists, and antacids are frequently prescribed for prophylaxis or treatment of acid-related disorders such as gastroesophageal reflux disease, gastric bleeding, and gastric ulcers.1,2 When coadministered with other therapeutic drugs, they can significantly alter the drug’s pharmacokinetics and lead to reduced efficacy. One mechanism for the negative pharmacokinetic impact is physicochemically driven, where the elevated gastric pH induced by the pH-modifying drug lowers the dissolution rate and/or solubility of the drug in the stomach. This results in poor absorption and/or high pharmacokinetic variability if a formulation is not optimized to perform in these variable gastric pH conditions.3,4 Another mechanism is drug−drug interaction between the therapeutic drug and a proton-pump inhibitor or H2-antagonist, due to inhibition of cytochrome P450 enzymes and/or efflux transporters, which can lead to compromised pharmacokinetics.2,3,5 © 2013 American Chemical Society
The latter mechanism is not addressed in this article. Understanding the factors that account for compromised dissolution rate and/or solubility is the central focus of this article. In addition to co-medication, suppressed gastric acid secretion, a condition that affects populations such as the elderly, those inflicted with certain disease states, and patients treated with certain drug classes (e.g., HIV protease inhibitors),4,6 also known as achlorhydria, can result in a similar negative pharmacokinetic impact. Several publications Special Issue: Impact of Physical Chemical Drug-Drug Interactions from Drug Discovery to Clinic Received: Revised: Accepted: Published: 4063
July 23, 2013 September 10, 2013 September 13, 2013 September 13, 2013 dx.doi.org/10.1021/mp400426f | Mol. Pharmaceutics 2013, 10, 4063−4073
Molecular Pharmaceutics
Article
Figure 1. Schematic diagram of in vitro microdissolution pH-shift test.
predicting the magnitude of the risk and understanding the key drug properties that drive it. While the physicochemical properties that impact a drug’s sensitivity to food have been studied with statistical and in vitro−in vivo correlation analyses,23,24 predictions of clinical pH-effects have been particularly challenging due to the complex and poorly understood interplay of a drug’s ionization characteristics, dissolution rates, precipitation and redissolution kinetics, permeability, and the variable pH and transit time induced by the pH-modifying agent.3,25,26 To enable such an analysis, dissolution characteristics of chemically diverse drugs that exhibit a wide range of clinical pH-effects were needed in a common biorelevant in vitro test. This study attempts to provide such an analysis, by generating dissolution data and integrating it with in silico predictions to correlate the results to the overall clinical pH-effects. The primary goal of this study was to understand why some weak base compounds have a strong clinical pH-effect while others show mild or no effect. To accomplish this, we developed a novel in vitro microdissolution test to study the kinetics of dissolution and precipitation. The prerequisites of such an in vitro test were that it had to be efficient, relatively simple, and API-sparing, have a reasonable throughput, and be biorelevant capturing the dynamic range of pH changes in the GI lumen. The pH-effect predicted from the in vitro dissolution assay was correlated to the clinical pH-effect. Next, using the dissolution data, in silico methods such as maximum absorbable dose (MAD) calculations, Sugano predicted fraction absorbed, and physiologically based pharmacokinetic models were compared for their correlation to clinical pH-effect. To further probe the role of physicochemical properties and molecular descriptors, statistical analyses were conducted to identify correlations to the clinical pH-effect. Finally, a developmentstage appropriate strategy is proposed that combines the suite of biopharmaceutical tools (physicochemical analysis, in vitro, in silico calculations, and in vivo data) to identify and assess the pH-effect risk for a weak base NME entering clinical development.
discuss this negative impact for approved commercially available drugs when coadministered with pH-modifying agents.2,3,5−11 Some notable examples include ketoconazole (Nizoral), where the AUC is reduced by 81% when dosed in combination with a gastric pH-modifier;12 dipyridamole (Persantine), where the Cmax and AUC are reduced 79% and 37% respectively;4 and gefitinib (Iressa), where the AUC is reduced by 48% and Cmax by 71%.13 Weak base drugs that show a steep pH-dependent solubility profile across the physiological pH range are likely to be sensitive to pH-modifying agents or achlorhydria disease states. By virtue of their ionization characteristics, this class of drugs can show reduced solubility, slower dissolution rates, and/or faster precipitation kinetics at high pH, thus significantly affecting the driving force for absorption in the small intestine. In this article the term “pH-effect” is used to describe the drop in pharmacokinetic performance between the high gastric pH state (pH-modifier cotreated) and a normal gastric pH state (untreated). This is expressed as a ratio between the two treatment conditions based on AUC or Cmax. A ratio of 1 signifies no pH-effect. Early identification of a pH-effect risk for a weak base new molecular entity (NME) is very valuable.3,14 Lack of knowledge of this liability can significantly delay a drug’s clinical development path by adding increased costs and clinical studies to bridge changes in mitigating the pH-effect. Several in vitro, in silico, and in vivo methods for evaluating and mitigating pH-effect have been described in the literature. This includes a number of biorelevant in vitro dissolution methods,15−18 each with unique strengths, limitations, and level of complexity.19,20 In silico approaches commonly utilize dissolution−absorption simulation and modeling with inputs from in vitro biorelevant dissolution, pH−solubility data, and pharmacokinetic data to either predict the risk of pH-effect or retrospectively understand the mechanism of the pH-effect.14,21,22 Animal PK studies, comparing the plasma profiles from animals pretreated with H2-antagonists or antacids and untreated (no pHmodifier) animals, are used to confirm the risk in vivo.21 In most cases, the biopharmaceutical tools mentioned above have been used to explain the pH-effect liability of a specific drug and identify approaches to resolve it. To our knowledge, very little published information examines this pH-effect risk across several classes of chemical entities with an emphasis on
■
MATERIALS AND METHODS Materials. Dipyridamole and enoxacin were purchased from Sigma-Aldrich (St. Louis, MO). Ketoconazole and gefitinib were purchased from AK Scientific (Union City, CA). Erlotinib
4064
dx.doi.org/10.1021/mp400426f | Mol. Pharmaceutics 2013, 10, 4063−4073
Molecular Pharmaceutics
Article
HCl was purchased from LC Laboratories. All BMS compounds were synthesized by Bristol-Myers Squibb Company (New Brunswick, NJ). SIF (Simulated Intestinal Fluid) powder used to make fasted state simulated intestinal fluid (FaSSIF) was purchased from Biorelevant.com (Croydon, Surrey, United Kingdom). Sodium phosphate dibasic was purchased from EM Science (Gibbstown, NJ). Hydrochloric acid (1 N) was purchased from J.T. Baker (Phillipsburg, NJ). Methocel A4M was purchased from Dow Chemical Company (Midland, MI). Water was purified by Milli-Q UV plus systems (Millipore Co., Bedford, MA). All other chemicals were of analytical grade and purchased from commercial vendors. Microdissolution pH-Shift Test. The microdissolution pH-shift test consists of two stages: a gastric phase and an intestinal phase. As shown in Figure 1, drug is introduced into simulated gastric fluid (SGF) and dissolution monitored for 20 min, after which concentrated simulated intestinal fluid (FaSSIF) is introduced into the same vial and dissolution monitored for up to 180 min. Such a setup eliminates the need to pump fluid between compartments, or eliminate flow issues due to undissolved drug clumps or drug loss due to adsorption to tubing. The SGF used was either 0.01 N HCl (SGFpH2) or pH 6 phosphate buffer (SGFpH6). The concentrated FaSSIF was adjusted for bile and buffer concentrations such that, when diluted with SGF, the concentration was equivalent to the Dressman fasted state simulated intestinal fluid (FaSSIFpH6.5). The pH of the combined fluid (FaSSIF) was 6.5. Each test compound was exposed to a normal acidic gastric environment (SGFpH2→FaSSIFpH6) and high gastric pH environment (SGFpH6→FaSSIFpH6.5) that mimics the pH-shift induced by pH-modifying agents. Real time absorbance of drug in solution was collected throughout the experiment by UV fiber optic probes with the pION μDiss Profiler (pION Inc., Billerica, MA). All fluids were preheated to 37 °C and degassed for 30 min prior to use. Test compounds were presuspended in 0.5% MethocelA4M to minimize wetting issues and dosed to the vial within 2 min of preparation. Dosing suspension (typically BMS-D > BMS-E. This rank order was predicted with all the computational approaches as well. As expected, the MAD and Sugano predicted ratios were generally closer to the in vitro AUC ratio. The Biomodel Fa ratio and GastroPlus AUC ratio appear to be more predictive of clinical AUC ratio. Physicochemical Properties and Clinical pH-Effect. The ability to predict the clinical pH-effect from physiochemical properties was investigated using linear sequential feature selection analysis. The relevant pKa, log D at pH 7, molecular weight (MW), melting point, intrinsic solubility, clinical dose, polar surface area (PSA), freely rotatable bonds (FRB), hydrogen donors, and hydrogen acceptors were the physicochemical properties used as the independent variables to correlate with the clinical pH-effect or the in vitro AUC ratio as the dependent variable. In feature selection test 1, molecular properties were correlated with in vitro AUC ratio; in feature 4068
dx.doi.org/10.1021/mp400426f | Mol. Pharmaceutics 2013, 10, 4063−4073
Molecular Pharmaceutics
Article
Table 3. Statistical Analysis model
Y, correlated variable
X, input variables
feature molecular properties selection 1 feature molecular properties selection 2 feature molecular properties; in vitro selection 3 data and calculations
in vitro AUC ratio clinical AUC ratio clinical AUC ratio
statistically significant variables melting point none in vitro AUC ratio
other molecular property in predicting clinical pH-effect (Table 3). PLS regression confirmed the linear sequential feature statistical findings. Only one principal component was fit to the data based on 5-fold cross-validation method (data not shown). Additional components were tested, although the rootmean-squared error against the cross validation data sets increased. Variables with the highest VIP values were melting point, those that are based on the in vitro dissolution test (MAD predicted ratio, Sugano model predicted ratio) and estimated permeability. In the PLS analysis of molecular descriptors melting point, polar surface area, freely rotatable bonds, molecular weight and solubility had VIP >1, although they were not statistically significant. A larger data set is needed to explore the statistical significance of these contributions. One would expect pKa to show some trend in correlating with a clinical pH-effect, since it has a direct influence on pHdependent dissolution and the maximum solubility a compound can achieve at a given pH. However, this was not the case as seen in the plot of all compounds in the database with clinical AUC ratio (Figure 5A) and clinical Cmax ratio (Figure 5B). This plot includes salts and free base compounds, with the latter highlighted as open symbols. Generally, salt forms tend to result in transient supersaturated states that mask the true sensitivity of free base drugs to pH-effects. Besides, their disproportionation and precipitation kinetics vary depending on the counterion. The behavior of salt forms versus free base forms can be different in the in vitro test and in vivo. In all cases the free base or salt forms of drugs were used in the in vitro test that was the same as that studied in clinical studies. Salts are prone to showing different supersaturation and precipitation kinetics depending on the type of counterion used. For example, BMSH HCl salt has an in vitro AUC ratio of 0.37 while its free base counterpart exhibits a ratio of 0.28 at the same dose. The presence of a similar ion in the dissolution medium to that for the salt form can potentially influence dissolution characteristics (the common-ion effect). BMS-G HCl salt is a likely example of this scenario.
Figure 4. Correlation of in vitro pH-effect with clinical pH-effect based on AUC (panel A) and Cmax (panel B).
selection 2, they were compared with clinical AUC ratio; in feature selection 3 the in vitro AUC ratio as well as MAD and Sugano model Fa were correlated with clinical AUC ratio (Table 3). No physiochemical properties (or combination thereof) were predictive of the clinical AUC ratio using linear modeling given that the bootstrap aggregated root-mean-squared error between the model and clinical AUC ratio increased with each term entered into the model (Table 3). Melting point was the only significant physicochemical predictor of in vitro AUC ratio, but was not predictive of the clinical AUC ratio. The in vitro AUC ratio was the most significant predictor of clinical pH-effect. In other words, the in vitro microdissolution test outperforms any
Table 2. pH-Effect Ratios of in Vitro and Various Computation Models for Approved Commercially Available Compounds and Representative NMEs clinical ratio BMS-A BMS-D BMS-E dipyridamole gefitinib ketoconazole
in vitro ratio
predicted ratio
AUC
Cmax
AUC
MAD
Sugano
Biomodel
GastroPlus (AUC)
GastroPlus (Cmax)
0.253 0.4 1.04 0.63 0.527 0.19
0.069 0.26 0.82 0.21 0.29 0.19
0.07 0.09 0.72 0.12 0.22 0.22
0.08 0.25 0.89 0.27 0.25 0.12
0.14 0.26 0.88 0.29 0.39 0.22
0.08 0.48 1.00 0.86 0.79 0.25
0.31 0.44 0.80 0.66 0.51 0.64
0.06 0.20 0.72 0.24 0.50 0.64
4069
dx.doi.org/10.1021/mp400426f | Mol. Pharmaceutics 2013, 10, 4063−4073
Molecular Pharmaceutics
Article
bin the risk categories as high, medium, or low risk. In silico methods and physicochemical properties were also probed for correlations to the clinical data as additional factors to flag this risk. In Vitro pH-Shift Microdissolution Test. A novel, efficient, practical, and API-sparing in vitro dissolution test was developed to generate dissolution and precipitation kinetics in an in vitro dissolution system that simulates GI conditions for normal, untreated and high gastric (pH-modifying agent, treated) subjects. The uniqueness of the in vitro test relative to previously published models16−18 is that it mimics clinically relevant dosing conditions, provides efficient, continuous, realtime UV fiber optics assessment of drug dissolution− precipitation without the use of HPLC analysis (HPLC vials, filtration supplies and organic solvent sparing), and throughput that enables head-to-head comparison of multiple API forms or formulations in the same run. The time course of drug dissolution and precipitation kinetics is measured in simulated gastric and intestinal fluids under SGFpH6 →FaSSIF pH6.5 conditions and SGFpH2→FaSSIFpH6.5 conditions (Figure 1). The ratio between the dissolution profiles (represented as AUC of SGFpH6→FaSSIFpH6.5/AUC of SGFpH2→FaSSIFpH6.5) was used as the measure of the in vitro pH-effect risk. The dissolution profiles under each pH condition varied dramatically between compounds as illustrated for gefitinib, erlotinib and ketoconazole in Figure 2. This is driven by the inherent physiochemical properties of the compound, the gastric solubility at low and high pH, the dissolution rate in FaSSIF, and its ability to remain supersaturated in FaSSIF at the dose studied. In this test the gastric to intestinal transition is instantaneous, and therefore the intestinal supersaturation (if present) is reached immediately upon transfer from SGFpH2. As expected, dose plays a critical role in dissolution/precipitation kinetics as illustrated by gefitinib (Figures 3A and 3B). If the dose tested is above the solubility in FaSSIF (almost always the case in these tests), the SGFpH6→FaSSIFpH6.5 intestinal AUC does not change with dose (Figure 3B, dotted blue line). In contrast, for the SGFpH2→FaSSIFpH6.5 dissolution, the intestinal AUC increases with dose up to a threshold for sustainment of supersaturation for that drug, beyond which precipitation kinetics predominate, resulting in a dramatic drop in AUC (Figure 3B). At the highest dose, high supersaturation drives rapid precipitation and a drop in FaSSIF concentration down to its crystalline solubility (Figure 3A, black triangles). Therefore, selecting an in vitro dissolution dose that is clinically relevant is essential to appropriately gauge dissolution in the GI lumen responsible for the clinical pH-effect. The clinical relevance of the dose dependency was reported for ketoconazole where the pH-effect increases when the dose is increased from 200 mg (clinical AUC 0.19) to 400 mg (clinical AUC 0.06), although it must be pointed out that these are from independent studies that used different pH-modifying agents.7,12 For the 14 drugs in Table 1 for which in vitro data was generated, the in vitro AUC ratio was plotted against the clinical AUC ratio (Figure 4A) or Cmax ratio (Figure 4B). The correlation to clinical Cmax ratio is virtually linear (R2 = 0.83) suggesting that the in vitro dissolution is a strong predictor of the clinical Cmax ratio. For the comparison to clinical AUC ratio (Figure 4A), an apparent non-linear relationship is observed with steeper response over a narrow range of in vitro dissolution ratio from 0 to 0.4, which underscores the tendency of the test to overpredict the clinical pH-effect. The direct correlation to
Figure 5. Correlation of pKa with AUC ratio (panel A) and Cmax ratio (panel B) for all salt and free base compounds. Free base compounds are denoted as open red symbols, salts as closed symbols.
■
DISCUSSION The ability to assess the risk of a clinical pH-effect for a NME during the early stages of formulation and clinical development is vital.3,14 The inadvertent selection of a drug physical form (crystalline salt, free form, or amorphous) that shows a strong pH-dependent absorption or a formulation strategy14,18,21,22 that fails to mitigate the risk can significantly delay drug development plans and add significant costs associated with formulation changes and additional clinical bioavailability studies. Moreover, knowledge generated from early pH-effect risk assessment is necessary to proactively steer clinical study design to include a pH-modifier treatment arm and obtain a clinical readout on the pH liability for high risk compounds. In this article, we examined various methods to assess the pHeffect risk, highlight the ones that are predictive of the clinical pH-effect, and propose a systematic strategy for implementation toward future NMEs. To understand why some weak base compounds exhibit a stronger pH-effect than others, we developed a novel in vitro dissolution test to study the kinetics of drug dissolution, supersaturation and precipitation under conditions a molecule experiences in the GI lumen. The test was used to evaluate a wide variety of chemotypes (BMS and commercially available compounds) for their correlation to the clinical pH-effect. A good in vitro−in vivo clinical pH-effect correlation enabled us to 4070
dx.doi.org/10.1021/mp400426f | Mol. Pharmaceutics 2013, 10, 4063−4073
Molecular Pharmaceutics
Article
Cmax is not surprising given that most of the compounds in this data set are BCS II compounds, where their dissolution characteristics and ability to maintain supersaturation directly impact the absorption profile. In general, the AUC ratio tends to be more muted than the Cmax ratio, as many other factors come into play, including metabolism, distribution and excretion (clearance), which confounds an in vitro−in vivo relationship built on physicochemical attributes such as dissolution, supersaturation maintenance and precipitation tendency. Of the compounds studied, BMS-G is an intriguing case as it stands out as the only drug with a reverse or positive pH-effect (in vitro AUC ratio of 2.51). Importantly, the in vitro test for the HCl salt predicted a dramatically higher dissolution at SGFpH6→FaSSIFpH6.5 that was confirmed in clinical studies with a Cmax ratio of 1.58, and a clinical AUC ratio of 1.14. The reason for BMS-G’s enhanced dissolution at SGFpH6→ FaSSIFpH6.5 transfer is not fully understood but can perhaps be explained by a common-ion effect. BMS-G HCl salt shows a significant increase in solubility when pH was raised from pH 2 to 4. The chloride content in SGFpH2 may reduce the solubility of the HCl salt under normal gastric conditions, thereby resulting in a positive pH-effect. Moreover, BMS-G appears to strongly supersaturate on transfer to FaSSIF reaching a solubility of 26 μg/mL, which is 3.5 times higher than the equilibrium FaSSIF solubility on SGFpH2→FaSSIFpH6.5 transfer. This indicates bile micellar solubilization as a possible contributing factor. The role of bile solubilization is noted in the fasted and fed state equilibrium solubility, 8 μg/mL (FaSSIF) and 24 μg/mL (FeSSIF), and the 2-fold positive food effect observed in clinical studies (data not shown). The example of BMS-G also corroborates the previous argument on the direct correlation of in vitro dissolution to clinical Cmax pHeffect, as opposed to the AUC. Binning the pH-effect risk as high risk, moderate risk or low risk is a practical way to quantify the pH-effect liability based on the in vitro dissolution test. Table 4 describes the three risk
replicated. Being a microdissolution method, this test does not capture disintegration of dosage form (tablet breakup). The inclusion of a dosage form crushed into a powder or suspension is the closest substitute. Another distinction between in vitro test and in vivo is that the SGF→FaSSIF transfer is instantaneous, whereas in the body it occurs over the gastric emptying time (several minutes). An experimental limitation is that the UV fiber optic measurements sometimes dictate dose selection and detection sensitivity of poorly soluble compounds. High dose compounds require large amount of drug in the dissolution vial which can saturate the UV signal to the detector or result in a large amount of undissolved material that can interfere with the signal. The UV interference from undissolved material could be minimized using the second derivative UV method up to a critical amount of undissolved solid above which the UV signal gets saturated and loses its characteristic shape.34 On the flip side, very low doses could yield very low drug solubility that can challenge the lower limits of detection. For these dose-related limitations, the use of probes with different path lengths can increase signal reliability, and as a last resort dose adjustments may be warranted. Predicting Clinical pH-Effect with Physicochemical Properties. Our original hypothesis was that there would be a predictive correlation between key physiochemical properties of the compounds and their clinical pH-effect. By utilizing linear sequential feature selection and PLS analyses with a number of molecular descriptors and physicochemical properties for the data set, we found no significant linear correlation with any parameter or combination of parameters (Table 3). When in vitro microdissolution data was included along with dissolution derived parameters, namely, MAD Fa ratio and Sugano predicted Fa ratio calculations, the in vitro AUC ratio was the most predictive parameter of the clinical pH-effect based on the root-mean-square of the prediction (Table 3). Taken together, both the statistical analyses are in agreement in pointing out that, for this data set, in vitro dissolution is the most reliable predictor of clinical pH-effect and that physicochemical properties and molecular descriptors, or any combination thereof, show weak correlations. It was expected that pKa and dimensionless parameters such as dose number and dissolution number would also be parameters of relevance as they directly impact dissolution and the maximum solubility at a given pH, but that was not the case with both statistical methods. A plot of pKa with clinical pH-effect for AUC (Figure 5A) and Cmax (Figure 5B) for salt and free base compounds verifies this lack of correlation. When only weak base compounds are considered in the plot (open symbols), a trend emerges for both AUC and Cmax plots. It suggests that weak base compounds in the pKa range 3.5−6 tend to show the greatest clinical pH-effect, while those outside this range show ameliorated pH-effects. This situation would apply to high dose, free base compounds that have poor intrinsic solubility at high pH with a relatively narrow window of absorption. It is conceivable that potent, low dose compounds in this pKa range with a wide window of absorption could show reduced sensitivity as their solubility may be adequate to dissolve the dose delivered. This hypothesis on the pKa relationship for free base drugs cannot be confirmed or refuted due to the lack of data points within this range in the data set. In general, high-risk free base drugs are less likely to be studied in clinical trials in favor of salt forms or solubility enhancing formulations that have a reduced propensity for a
Table 4. In Vitro−in Vivo Correlation: pH-Effect Risk Categories risk category
in vitro AUC Ratio
clinical AUC or Cmax ratio
high moderate low
0.5
0.8
categories: in vitro AUC ratio of >0.5 as low risk (equivalent to clinical Cmax or AUC ratio >0.8); 0.2−0.5 as moderate risk (equivalent clinical Cmax and AUC ratio 0.5−0.8); and
