Drug Permeability - Best Practices for Biopharmaceutics Classification System (BCS)-Based Biowaivers: A workshop Summary Report.

The workshop "Drug Permeability - Best Practices for Biopharmaceutics Classification System (BCS) Based Biowaivers" was held virtually on December 6, 2021, organized by the University of Maryland Center of Excellence in Regulatory Science and Innovation (M-CERSI), and the Food and Drug Administration (FDA). The workshop focused on the industrial, academic, and regulatory experiences in generating and evaluating permeability data, with the aim to further facilitate implementation of the BCS and efficient development of high-quality drug products globally. As the first international permeability workshop since the BCS based biowaivers was finalized as the ICH M9 guideline, the workshop included lectures, panel discussions, and breakout sessions. Lecture and panel discussion topics covered case studies at IND, NDA, and ANDA stages, typical deficiencies relating to permeability assessment supporting BCS biowaiver, types of evidence that are available to demonstrate high permeability, method suitability of a permeability assay, impact of excipients, importance of global acceptance of permeability methods, opportunities to expand the use of biowaivers (e.g. non-Caco-2 cell lines, totality-of-evidence approach to demonstrate high permeability) and future of permeability testing. Breakout sessions focused on 1) in vitro and in silico intestinal permeability methods; 2) potential excipient effects on permeability and; 3) use of label and literature data to designate permeability class.

Exploring Canine-Human Differences in Product Performance. Part II: Use of Modeling and Simulation to Explore the Impact of Formulation on Ciprofloxacin In Vivo Absorption and Dissolution in Dogs.

This study explored the in vivo performance of three oral ciprofloxacin formulations (oral solution, fast, or slow dissolving tablets) in beagle dogs. The in vivo absorption and dissolution behaviors, estimated with in silico mechanistic models, were compared to the results previously published in human volunteers. Six normal healthy male beagle dogs (five to completion) received three oral formulations and an intravenous infusion in a randomized crossover design. Plasma ciprofloxacin concentrations were estimated by tandem mass spectrometry detection. A mechanistic absorption model was used to predict the in vivo dissolution and absorption characteristics of the oral formulations. Canine ciprofloxacin absorption was constrained to the duodenum/jejunum. This absorption window was far narrower than that seen in humans. Furthermore, while substantial within-individual variability in drug absorption was seen in human subjects, a greater magnitude of variability was observed in dogs. For three sets of data, a lag time in gastric emptying was necessary to improve the accuracy of model-generated in vivo blood level profile predictions. In addition to species-associated dissimilarities in drug solubilization due to human versus canine differences in gastrointestinal fluid compositions, the far more rapid intestinal transit time and potential segmental differences in drug absorption needed to be considered during human-canine extrapolation of oral drug and drug product performance. Through the use of mechanistic models, the data generated in the human and canine studies contributed insights into some aspects of the interspecies differences to be considered when extrapolating oral bioavailability/formulation effect data between dogs and humans.

Biowaiver monographs for immediate release solid oral dosage forms: diclofenac sodium and diclofenac potassium.

Literature data are reviewed regarding the scientific advisability of allowing a waiver of in vivo bioequivalence (BE) testing for the approval of immediate release (IR) solid oral dosage forms containing either diclofenac potassium and diclofenac sodium. Within the biopharmaceutics classification system (BCS), diclofenac potassium and diclofenac sodium are each BCS class II active pharmaceutical ingredients (APIs). However, a biowaiver can be recommended for IR drug products of each salt form, due to their therapeutic use, therapeutic index, pharmacokinetic properties, potential for excipient interactions, and performance in reported BE/bioavailability (BA) studies, provided: (a) test and comparator contain the same diclofenac salt; (b) the dosage form of the test and comparator is identical; (c) the test product contains only excipients present in diclofenac drug products approved in ICH or associated countries in the same dosage form, for instance as presented in this paper; (d) test drug product and comparator dissolve 85% in 30 min or less in 900 mL buffer pH 6.8, using the paddle apparatus at 75 rpm or the basket apparatus at 100 rpm; and (e) test product and comparator show dissolution profile similarity in pH 1.2, 4.5, and 6.8.

Effect of common excipients on Caco-2 transport of low-permeability drugs.

The Biopharmaceutics Classification System (BCS) allows waivers of in vivo bioequivalence for rapidly dissolving immediate-release (IR) formulations of drugs with high solubility and high permeability. One potential issue in possibly extending BCS biowaivers to low-permeability drugs is the potential for excipients to modulate the intestinal permeability of the drug. The objective of this work was to evaluate the effect of nine individual excipients on the Caco-2 permeability of seven low-permeable compounds that differ in their physiochemical properties. Generally, most excipients had no influence on drug permeability. With the exception of sodium lauryl sulfate, no excipient affected Caco-2 monolayer integrity. Sodium lauryl sulfate moderately increased the permeability of almost all the drugs. Tween 80 significantly increased the apical-to-basolateral direction permeability of furosemide, cimetidine, and hydrochlorothiazide, presumably by inhibiting their active efflux, without affecting mannitol permeability. Additionally, docusate sodium moderately increased cimetidine permeability. Other excipients did not have significant effect on the permeability of these three drugs. Further work is needed to interpret the in vivo consequences of these observations from cell culture.

Novel direct curve comparison metrics for bioequivalence.

PURPOSE: The object of this work was to devise four new direct curve comparison (DCC) metrics and examine each metric's distribution properties and performance characteristics. METHODS: DCC metrics, Cmax, and AUCi were calculated from two bioequivalence studies of three sustained release carbamazepine formulations, where a range of profile similarity was observed. DCC metric values and their confidence intervals were compared to Cmax and AUCi. RESULTS: The DCC metrics rho, rhom, deltaa, and deltas, exhibited more favorable distributions than Cmax and AUCi ratios, which were frequently skewed. The DCC metrics performed differently than Cmax and AUCi ratios in profile comparisons due to the nature of the DCC metrics. Unlike Cmax and AUCi, the DCC metrics utilize all data points to directly compare entire profiles. Each DCC metric appears to measure "exposure" in a single assessment. Possible bioequivalence acceptance criteria are: p < or =1.40, rhom, < or =0.35, deltaa, < or =0.27, and deltas < or =0.102. CONCLUSIONS: These DCC metrics, particularly rhom, are promising bioequivalence metrics for "exposure."

Transport of poly amidoamine dendrimers across Madin-Darby canine kidney cells.

The objective of this study was to determine the permeability of a series of poly amidoamine (PAMAM) dendrimers of generations 0-4 (G0-G4) across MDCK (Madin-Darby Canine Kidney) cell line. PAMAM dendrimers with incremental increase in size and molecular weight were labeled by fluorescein isothiocyanate (FITC) and the least polydisperse fractions were collected by size exclusion chromatography. MDCK cells were grown on Transwell filters for four days. The conjugates were detected by HPLC equipped with fluorescence detector. The permeability of the dendrimers across MDCK cells was determined in the apical to basolateral direction. The rank-order permeability of the PAMAM dendrimers was G4 >> G1 approximately G0 > G3 > G2. The permeability of mannitol in the presence of G4 increased by nine-fold. Results suggest that the transepithelial transport of PAMAM dendrimers is effected by both the polymer size, and the modulation of the cell membrane by the cationic dendrimers.

Development of a more rapid, reduced serum culture system for Caco-2 monolayers and application to the biopharmaceutics classification system.

The objectives were: (1) to develop a more rapid, reduced serum culture system for Caco-2 monolayers, relative to the traditional 21-day, 10% fetal bovine serum (FBS) system; and (2) to determine the biopharmaceutical drug classification of an oral therapeutic agent using this new system. Caco-2 cells were grown in the six well format on polycarbonate filters, in medium containing 2% iron supplemented calf serum (sCS) and a combination of growth factors and hormones. After 4 days in culture, permeabilities of three marker compounds (metoprolol, mannitol, and taurocholate) across monolayers were determined, and compared to permeabilities from the traditional 21-day, 10% FBS system, using cells at similar passage number. Cell morphology, degree of cell differentiation, and the presence of two efflux pumps were assessed. The 2% sCS model was also used to classify the permeability of an oral therapeutic agent as high or low. No difference in permeability was observed for metoprolol transport (P = 0.38) between the two culture methods, and the values obtained were independent of passage number of the cells. Mannitol permeability was about 2-fold higher from the 2% sCS system, as compared to the 10% FBS system. Taurocholate permeability was low indicating the 2% sCS culture at 4 days lacked this particular active transporter capability. Electron micrographs of cells grown in the 2% sCS system at 4 days revealed the presence of microvilli and tight junctions. P-glycoprotein and an efflux pump for furosemide were functionally present. The 2% sCS system indicated the oral therapeutic agent as highly permeable, which agreed with the 10% FBS system. This new system provides a rapid, accurate, and economical option for passive permeability determination, and appears to be applicable to the proposed Biopharmaceutics Classification System (BCS).

Influence of passive permeability on apparent P-glycoprotein kinetics.

PURPOSE: The objectives of this work were to evaluate the importance of moderate passive permeability on apparent P-glycoprotein (P-gp) kinetics, and demonstrate that inspection of basolateral to apical and apical to basolateral (BL-AP/AP-BL) permeability ratios may result in a compound being overlooked as a P-gp substrate and inhibitor of another drug's transport via P-gp inhibition. METHODS: The permeability ratios of nicardipine, vinblastine, cimetidine, and ranitidine were determined across Caco-2 monolayers that express P-gp, in the presence and absence of the specific P-gp inhibitor, GF120918. In addition, the permeability ratio of vinblastine was studied after pretreatment of Caco-2 monolayers with nicardipine, ranitidine, or cimetidine. Similar studies were repeated with hMDRI-MDCK monolayers. RESULTS: The permeability ratios for cimetidine and vinblastine were >2. The permeability ratios for nicardipine and ranitidine were close to unity, and were not affected by the addition of GF120918. Based solely on ratios, only compounds with moderate transcellular permeability (vinblastine and cimetidine) would be identified as P-gp substrates. Although the permeability ratios appeared to be unity for nicardipine and ranitidine, both compounds affected the permeability of vinblastine, and were identified as substrates and inhibitors of P-gp. Studies performed in hMDR1-MDCK cells confirmed these experimental results. Data were explained in the context of a kinetic model, where passive permeability and P-gp efflux contribute to overall drug transport. CONCLUSIONS: Moderate passive permeability was necessary for P-gp to reduce the AP-BL drug permeability. Inspection of the permeability ratio after directional transport studies did not effectively identify P-gp substrates that affected the P-gp kinetics of vinblastine. Because of the role of passive permeability, drug interaction studies with known P-gp substrates, rather than directional permeability studies, are needed to elucidate a more complete understanding of P-gp kinetics.

Prediction of dissolution-absorption relationships from a dissolution/Caco-2 system.

While the analysis of in vitro dissolution-in vivo absorption relationships from oral solid dosage forms provides biopharmaceutical insight and regulatory benefit, no well developed method exists to predict dissolution-absorption relationships a priori to human studies. The objective was to develop an integrated dissolution/Caco-2 system to predict dissolution-absorption relationships, and hence the contributions of dissolution and intestinal permeation to overall drug absorption for fast and slow formulations of piroxicam, metoprolol, and ranitidine. Dissolution studies were conducted on fast and slow dissolving immediate-release formulations of piroxicam, metoprolol tartrate, and ranitidine HCl. Dissolution samples were treated with concentrated buffers to render them suitable (i.e., isotonic and neutral pH) for Caco-2 monolayer permeation studies. The dissolution/Caco-2 system yielded a predicted dissolution-absorption relationship for each formulation which matched the observed relationship from clinical studies. The dissolution/Caco-2 system's prediction of dissolution or permeation rate-limited absorption also agreed with the clinical results. For example, the dissolution/Caco-2 system successfully predicted the slow piroxicam formulation to be dissolution rate-limited, and the fast piroxicam formulation to be permeation rate-limited. Moreover, the system predicted this change from dissolution rate-limited absorption for slow piroxicam to permeation rate-limited absorption for fast piroxicam, in spite of piroxicam's high permeability and low solubility. The dissolution/Caco-2 system may prove to be a valuable tool in formulation development. Broader evaluation of such a system is warranted.

Dependence of in vitro-in vivo correlation analysis acceptability on model selections.

The objectives of this study were to assess the influence of pharmacokinetic and dissolution model selection on the success of in vitro-in vivo correlation (IVIVC) analysis of fast-, medium-, and slow-dissolving metoprolol tartrate immediate-release tablet formulations. Several different compartmental models were fit to the fast formulation plasma data. Three candidate models with the best fits were ranked 1, 2, and 3 and used to predict AUC and Cmax of the medium and slow formulations. Acceptability of each model to predict the medium and slow formulations was determined using +/- 20% as the limit for acceptable relative prediction error. When the best dissolution models were used, models 1 and 2 each failed to adequately predict Cmax for slow formulation (-26.8% error for model 1 and -20.4% error for model 2). However, the less appropriate model 3 adequately predicted Cmax for slow formulation (-15.1% error). The selection of the dissolution model also determined the outcome of IVIVC analysis, again with a less appropriate model resulting in successful prediction. When the Weibull function was used to characterize dissolution, model 2 failed to adequately predict Cmax for slow formulation (-20.4% error); however, model 2 adequately predicted Cmax for slow formulation when dissolution was characterized using the poorer fitting first-order model (-14.4% error). These results indicate that the success or failure of external validation of these metoprolol tartrate tablets was dependent on the pharmacokinetic and the dissolution models employed. Considering the role of subjectivity in identifying pharmacokinetic and dissolution models, these findings suggest a need to develop objective criteria to identify models a priori to IVIVC analysis.

Prediction of dissolution-absorption relationships from a continuous dissolution/Caco-2 system.

The objectives were 1) to design a continuous dissolution/Caco-2 system to predict the dissolution-absorption relationships for fast and slow dissolving formulations of piroxicam, metoprolol tartrate, and ranitidine HCl, and compare the predicted relationships with observed relationships from clinical studies; 2) to estimate the effect of croscarmellose sodium on ranitidine dissolution-absorption relationships; and 3) to estimate the effect of solubilizing agents on piroxicam dissolution-absorption relationships. A continuous dissolution/Caco-2 system was constructed from a dissolution apparatus and a diffusion cell, such that drug dissolution and permeation across a Caco-2 monolayer would occur sequentially and simultaneously. The continuous system generally matched observed dissolution-absorption relationships from clinical studies. For example, the system successfully predicted the slow metoprolol and slow ranitidiine formulations to be permeation-rate-limited. The system predicted the slow piroxicam formulation to be dissolution-rate-limited, and the fast piroxicam formulation to be permeation-rate-limited, in spite of piroxicam's high permeability and low solubility. Additionally, the system indicated croscarmellose sodium enhanced ranitidine permeability and predicted solubilizing agents to not modulate permeability. These results suggest a dissolution/Caco-2 system to be an experimentally based tool that may predict dissolution-absorption relationships from oral solid dosage forms, and hence the relative contributions of dissolution and permeation to oral drug absorption kinetics.

Human drug absorption kinetics and comparison to Caco-2 monolayer permeabilities.

PURPOSE: This study aims to assess the drug absorption kinetics of three drugs and compare their resulting first-order intestinal permeation rate constants to their Caco-2 monolayer permeabilities. METHODS: In vitro dissolution--in vivo absorption analysis was conducted on four formulations of each ranitidine HCl, metoprolol tartrate, and piroxicam to yield apparent and "true" human clinical permeation rate constants. Drug permeability coefficients through Caco-2 monolayers were also determined. RESULTS: In vitro dissolution--in vivo absorption analysis revealed different relative and absolute contributions of dissolution and intestinal permeation to overall drug absorption kinetics for various drug formulations and yielded estimates of each drug's true and apparent human intestinal permeation rate constant [kp = 0.225 hr-1, 0.609 hr-1, and 9.00 hr-1 for ranitidine, metoprolol, and piroxicam, respectively]. A rank order relationship was observed for both the apparent and true permeation rate constant with Caco-2 monolayer permeability. The decrease in the true permeation rate constant relative to the apparent permeation rate constant was most significant (almost three-fold) for the least permeable compound, ranitidine. CONCLUSIONS: There were marked differences in the permeation kinetics of ranitidine, metoprolol, and piroxicam. The possibility of an association between absorption kinetics from dosage forms in humans and Caco-2 monolayer permeability may allow for a direct kinetic interpretation of human oral absorption from Caco-2 monolayer permeability values.

Evaluation of direct curve comparison metrics applied to pharmacokinetic profiles and relative bioavailability and bioequivalence.

PURPOSE: The intent was to investigate three direct curve comparison metrics, the Rescigno Index, f1, and the Chinchilli Metric as tools to assess relative bioavailability (BA) and bioequivalence (BE). The specific objectives were to 1) estimate relative bioavailability and bioequivalence and 2) compare detection of profile shape differences with typical (i.e. AUC and Cmax) criteria. METHODS: Product bioequivalence was estimated and the degree of concordance with typical criteria in detecting profile differences was determined from the eighteen bioequivalence studies (390 subjects). Descriptive statistical analysis was carried out on the concordant and discordant profile subsets. RESULTS: 1) Three of the eighteen studies failed typical criteria (AUC and Cmax). Of the curve metrics, 12 studies failed the Chinchilli metric criteria and 13 failed f1 criteria. Using three different exponents in the Rescigno Index calculation, 17 (exponent = 3), 13 (exponent = 1), and 11 (exponent = 1/3) failed the criteria for bioequivalence. The frequency of profiles found to be different was comparable across the metrics, but the specific profiles found to be different or not different varied across the metrics. The Chinchilli Metric and f1 agreed 71% and 72% with typical criteria in judging profiles to be different or not different. Descriptive evaluation suggested that the direct curve metrics more effectively detect differences in absorption time lags but less effectively detect differences in Cmax. The Rescigno Index showed dependence on the direction of the difference between test and reference concentrations. CONCLUSIONS: With the limits used here, the direct curve metrics represent a more conservative approach to evaluate product bioequivalence. With further investigation and development, the direct curve approach may be used effectively to evaluate relative BA and BE.

Methods to compare dissolution profiles and a rationale for wide dissolution specifications for metoprolol tartrate tablets.

The objectives of this work were to apply several profile comparison approaches to dissolution data of four different but bioequivalent metoprolol tartrate tablet formulations to (1) identify the advantages and disadvantages of each approach, (2) quantify the metric for comparing dissolution profiles of each method, (3) determine metric limits that are consistent with the observed bioequivalence, and (4) rationalize the observed metric limits with respect to the role of dissolution in overall metoprolol absorption. Dissolution was performed by the USP monograph method on four formulations of metoprolol tartrate tablets (Lopressor plus fast, medium, and slow dissolving test formulations). Three general approaches to compare dissolution profiles were examined; they were ANOVA-based, model-independent, and model-dependent approaches. It is concluded that model-independent approaches and several model-dependent approaches yielded numerical results that can serve as objective and quantitative metrics for comparing entire dissolution profiles of the four metoprolol tartrate formulations. However, these methods presented complications. Some metrics were dependent on the length of the dissolution profile and the sampling scheme. Results from the pairwise procedures also depended on the pairing assignment of individual profiles. In spite of complications, these methods suggested wide dissolution specification limits. Wide dissolution specifications were rationalized through an analysis of in vitro-in vivo relationships, which indicated metoprolol dissolution from these formulations was not the rate-limiting step; hence, a range of dissolution profiles can be expected to yield equivalent plasma profiles.

In vitro-in vivo relationships of several "immediate" release tablets containing a low permeability drug.

The objective of this work was to gain insight into the biopharmaceutical performance of four different but bioequivalent ranitidine hydrochloride tablet formulations. This analysis employed a recently described method1 to relate in vitro and in vivo data and aimed to facilitate an understanding of oral drug product performance. For each ranitidine formulation, dissolution was performed using the USP procedure. A four-way, single dose bioequivalence study (n = 14) was performed. The fraction of the total amount of dose absorbed at each plasma sample time was determined by the Wagner-Nelson method. Equation 1 (see below) was fitted to the in vitro vs. in vivo data. For all four formulations, this analysis suggests absorption was permeation-rate limited, where ranitidine exhibited a low permeation rate constant of 0.01/min.

Novel approach to the analysis of in vitro-in vivo relationships.

The objective of this study was to quantify the dependence of degree of in vitro-in vivo correlation on the relative rates of dissolution and intestinal permeation and on the fraction of dose absorbed. The following equation was derived assuming first-order dissolution and permeation after oral drug administration: Fa = fa-1(1 - alpha(alpha - 1)-1 (1 - Fd) + (alpha - 1)-1(1 - Fd)alpha), where Fa is the fraction of the total amount of drug absorbed at time t, fa the fraction of the dose absorbed at t = infinitive, alpha is the ratio of the first-order permeation rate constant to the first-order dissolution rate constant, and Fd is the fraction of dose dissolved in vitro at time t. This equation was examined in order to pursue a theoretical treatment of in vitro-in vivo correlation. The degree of in vitro-in vivo correlation between Fa and Fd was measured by r2. alpha was varied between 1000 and 0.001. fa was varied between 0.1 and 1.0. Points employed in the linear regression were geometrically balanced about the derived equation. r2 values decreased as alpha decreased for all values of fa. r2 values were virtually independent of fa for all values of alpha, except for 0.01 < alpha < 1.0. The slope of the regression was modulated by both alpha and fa; larger alpha and smaller fa each increased slope. Application of the equation to a piroxicam data set demonstrated the equation's utility relative to the USP Level A correlation approach. It is concluded that the degree of in vitro-in vivo correlation depends on the relative rates of dissolution and intestinal permeation and on the fraction of dose absorbed and that the derived model merits further study.

Mathematical model and dimensional analysis of glycocholate binding to cholestyramine resin: implications for in vivo resin performance.

In large doses, cholestyramine resin lowers blood serum cholesterol by binding bile salts in the intestinal lumen and thus increases the fecal excretion of bile salts. In order to gain a better understanding of the low in vivo potency of cholestyramine, mathematical models estimating the amount of glycocholate bound per gram of cholestyramine and the free glycocholate concentration were derived and employ the capacity-corrected molar selectivity coefficient. Predictions of the quantity of glycocholate bound per gram of cholestyramine and of the free glycocholate concentration matched observed values (r2 = 0.993 and r2 = 0.998, respectively) over a wide range of conditions. Simulated binding studies indicated the relative importance of several biopharmaceutical parameters for improved resin in vivo performance. Increasing resin selectivity of glycocholate over chloride has greatest therapeutic impact if bile salt sequestering is most important in the upper portion of the intestines. Furthermore, ion exchange phenomena was subjected to dimensional analysis and revealed the controlling factors as components of two dimensionless numbers, GC* and Cl*. Placing physiologic limits on values of GC* and Cl* suggests requisite selectivity properties of more potent bile acid sequestrants and dosing strategies to optimize current resin therapy.

In vitro characterization of sodium glycocholate binding to cholestyramine resin.

Cholestyramine resin in a bile acid sequestrant which binds with bile salts in the intestinal lumen to increase the fecal excretion of bile salts and, thus, lower blood serum cholesterol. In order to gain a better understanding of the low in vivo potency of cholestyramine, in vitro equilibrium binding studies, water sorption studies, and resin capacity measurements were performed using cholestyramine and the bile salt sodium glycocholate. Equilibrium binding and water sorption studies entailed equilibrating cholestyramine (1.0-20 mg/mL) with solutions which varied in glycocholate anion concentration (0.20-16.5 mM) and chloride anion concentration (15-150 mM). The resin's practical specific capacity for glycocholate was lower than the practical specific capacity for chloride. This difference suggests that the rigid, bulky bile salt was pore excluded from 10% of the resin's ionogentic sites. A fundamental parameter called the capacity-corrected molar selectivity coefficient, KGC-Cl-, was postulated to describe the underlying binding phenomena and was determined by measuring the free glycocholate and chloride anion concentrations; KGC-Cl- ranged from 9.8 (+/- 0.7) to 18.6 (+/- 0.2) and depended on the square of the free chloride concentration. The capacity-corrected molar selectivity coefficient was larger than the molar selectivity coefficient due to pore exclusion of glycocholate. A more simple method to calculate the capacity-corrected molar selectivity coefficient which required less data gave similar values to the more rigorous method (r2 = 0.955).