Predictive Potential of Acido-Basic Properties, Solubility and Food on Bioequivalence Study Outcome: Analysis of 128 Studies

All data from in-house bioequivalence studies sponsored by Sandoz¹ that satisfied the inclusion criteria were included in the analysis. Criteria for inclusion of the study into the database were as follows:

Information collected or calculated for each study and/or API is presented in Table 1. Sandoz in-house or, where not available, literature solubility data were used. GastroPlus^® v9.8 (Simulations Plus Inc., Lancaster, CA, USA) was used to retrieve calculated effective permeability (cPeff) and clogP for each API using the absorption, distribution, metabolism, excretion and toxicity (ADMET) Predictor^® and .mol file of the API [25].cPeff was used to classify APIs into high and low cPeff group with a cut-off value of 3 for metoprolol (considered as a standard for high permeability). Zhou et al. concluded that when an API’s logP was < 3, the solubility of poorly soluble APIs was less impacted by (bio)surfactant variations [26]. Thus, our hypothesis was that when clogP is < 3 (low clogP), the risk for non-bioequivalent results will be lower compared with a high-clogP group, due to potentially lower variability of solubility limited absorption and/or lower impact of composition (surfactant) differences between compared treatments.

Our database included bioequivalence studies under fasting and fed conditions. A bioequivalence study was considered successful when the bioequivalence criteria defined in the study protocol were satisfied and the compared products were concluded to have bioequivalence. In case of fixed-dose combination products, the success of bioequivalence study for each API was evaluated separately. Thus, the analysis was performed on 145 observations of 128 bioequivalence studies. For example, it could happen that for API 1 of a fixed-dose combination product, bioequivalence was concluded and for API 2, bioequivalence was not concluded. Such a product cannot be considered bioequivalent, but for the purpose of this analysis the outcome for API 1 was still considered as bioequivalent. The term non-bioequivalence is used intentionally since failure to conclude bioequivalence does not lead to conclusion of bioinequivalence.

Do for a specific pH was calculated with Equation 1:where Do_pH is Do at a specific pH, D is the dose used in a bioequivalence study (mg) and S (pH) is solubility at specific pH (mg/mL).

Drugs were classified as weak base, weak acid, amphoteric or neutral-like based on Do at gastrointestinal tract (GIT)-relevant pH (1.2, 3, 4.5 and 6.8). See Table 2 for classification criteria. If Do was ≤ 1, API was considered as highly soluble and if Do was ≤ 2 at any condition, the drug was considered borderline soluble under that condition. Any Do > 2 led to conclusion of low solubility. The initial idea for such classification came from the BCS subclasses proposed by Tsume et al. [15]. However, we have not differentiated between high and low permeability and introduced solubility at pH 4.5 into classification criteria. Namely, based on BCS-based biowaver guidelines, API is considered highly soluble if the highest dose is soluble in 250 mL of three buffers in a pH range of 1.2–6.8, preferably 1.2, 4.5 and 6.8 [6]. This leads to creation of an additional class not defined by Tsume et al., that is, amphoteric poorly soluble molecules. Note that many drugs classified as neutral-like also had acidic or basic functional groups. However, they were not ionizing in the GIT pH range or their ionization did not impact solubility.

We have also attempted to create acido-basic classification solely on the predicted pKa values. This classification of APIs to similar weak-acid, weak-base, amphoteric and neutral-like classes considered molecule structure, relevant GIT pH range (1.2–6.8) and pKa of the most acidic (pKa_ma) and the most basic (pKa_mb) functional groups of the API. With this information, proportion of ionization was calculated by H–H equation [27]. If < 1% of ionization was expected in the pH range, then the API was considered neutral. If ionization was > 1% anywhere in the GIT range (1.2–6.8), then, depending on the ionizable functional group(s), the APIs were classified as weak base, weak acid or amphoteric. For example, if a weak acid’s most acidic pKa was > 8.8, then this molecule was considered and classified as neutral in the GIT pH range, since < 1% would be ionized in the GIT pH range. For the same reason, amphoteric molecules were classified as weak bases if the most basic pKa of the drug was > 8.8.

When the two acido-basic classifications were compared, we noticed discrepancies, minor in classification of weak acids and weak bases and major in classification of neutral and amphoteric molecules. The reason could be in the error made by using predicted pKa. The solubility-based classification was considered more relevant since it was based on the measured solubility which plays a crucial role in the bioavailability of API. Thus, this solubility-based classification (as defined in Table 2) was used as a primary analysis.

Descriptive statistical analysis was performed to summarize API characteristics. Descriptive statistics were calculated on a subset of poorly soluble APIs for the bioequivalent and non-bioequivalent group for comparison purposes. Descriptive statistics were reported as median and interquartile ranges since the distributions departed from normal (see Tables 3, 4).

A set of univariate tests was performed to test how different variables are associated with the bioequivalence outcome. A Fisher’s exact test was utilized to test the association between categorical variables (Table 4) and bioequivalence outcome. Association between the numeric variables and bioequivalence outcome was assessed by non-parametric Kruskal–Wallis tests due to deviations from the normal distribution. p Values of < 0.1 and < 0.05 were set for conclusion of association and strong association, respectively. As this was an exploratory analysis there was no correction for multiple comparisons. Tests used for specific parameters are presented in Table 4.

To assess how well each significant parameter alone distinguishes between bioequivalence and non-bioequivalence outcome, we have created receiver operating characteristics (ROC) curves, calculated area under the ROC curve (AUROC) and its 95% confidence interval limits to assess if AUROC is significantly different from 0.5 (better than random guess). Youden index analysis was performed to find the optimal cutoff value (Youden threshold) providing the best tradeoff between sensitivity and specificity. The specificity and sensitivity at this threshold were also reported.

Data were analyzed using Minitab^® v19.2020.3 (Minitab, Inc., State College, PA, USA) and R software v4.1.2 (R Core Team 2021, Vienna, Austria) using packages pROC v1.16.2 and corrplot v0.92.

The highest incidence of non-bioequivalence outcome (53% non-bioequivalence rate) in the weak-acid group correlates with higher Do_12, Do_3 and significantly lower pKa value for the most basic functional group of the non-bioequivalent API group. In an attempt to explain this higher incidence of non-bioequivalence, it is important to note that 90% of these weak acids were also highly permeable (had cPeff higher than high permeability standard metoprolol and all had high clogP [≥ 3]) and would be classified as Class II-a according to the extended BCS classification [15]. These APIs are practically insoluble in gastric pH (without the solubility or dissolution enhancing excipients) and are subjected to highly variable gastric emptying. When they do reach the intestine, the conditions for solubility/dissolution are favorable due to ionization of the weak-acid APIs in the basic environment. Since these APIs are also highly permeable (due to passive or transporter-mediated absorption [28‐31]), they are readily absorbed, which means that the dissolution rate and potential formulation differences directly influence bioavailability [9]. This can explain the higher non-bioequivalence rate but at the same time points to the risk mitigation strategy. Namely, highly permeable weak-acid APIs are good candidates for building a Level A IVIVC, which was also the reason a BCS-based biowaiver extension was proposed for BCS Class II-a APIs [32]. Case-by-case research is needed to find biorelevant media to support such biowaivers [14]. There are several published successful product development cases and biowaiver applications of Level A IVIVC for weak-acid APIs [16‐18, 20].

The group of molecules with the second highest non-bioequivalence occurrence (24%) were neutral molecules, which contributed to higher median clogP values for the non-bioequivalent group. Nevertheless, 16% of the neutral molecules also had clogP < 3, which could have contributed to the lower non-bioequivalence occurrence compared with the weak-acid group. As hypothesized, solubility (and potentially absorption) of neutral APIs with lower clogP could be less impacted by variations in (bio)surfactants. In a related aspect, this group of neutral APIs included two thirds of highly permeable and one third of poorly permeable APIs and cPeff was significantly associated with the bioequivalence outcome. This cPeff difference may again help us in interpreting the results. Similar to a weak acid, if the permeability of a neutral API is high, the solubility/dissolution rate (and potential differences in formulations) would directly reflect the extent and rate of bioavailability. If the permeability is low, it limits the absorption and the potential difference in formulations might not impact the extent or rate of bioavailability. Thus, presence of poorly permeable neutral APIs (BCS subclass IVc) and APIs with clogP < 3 may contribute to a lower non-bioequivalence rate when compared with the weak-acid API group. Regardless, the 24% risk for non-bioequivalence needs to be mitigated during development, and examples of IVIVC for riskier BCS subclass IIc drugs include introduction of a biphasic system (absorptive compartment) into a gastrointestinal simulator and United States Pharmacopeia (USP) dissolution apparatus II [21].

It is worth adding that APIs in our database classified as weak acid or neutral were in larger proportion subject to first-pass metabolism and P-gp mediated efflux compared with weak bases and amphoteric molecules, which could also contribute to challenges with bioavailability [7] and bioequivalence of the neutral and weak-acid group of poorly soluble APIs.

In a weak-base API group, we could observe only one non-bioequivalent result. This phenomenon could again be explained by the fact that all APIs in a weak-base group had high permeability (cPeff more than those of metoprotol) and would be classified as BCS Class IIb APIs. Unlike weak acids, weak bases are completely ionized at gastric pH and they have a high potential to be completely dissolved (although incomplete dissolution may still occur [24]) from immediate-release formulations. However, when reaching high intestinal pH, there is a potential for precipitation, which can impact the absorption processes [33]. There is an alternative scenario where the API is retained in a supersaturated state (or with minimal precipitation) when passing through intestines. This could be retained by the presence of bile salts, excipients and/or fast absorption of the API from the intestine—and the higher the permeability, the higher the absorption rate of the API. Our observations support the hypothesis that highly permeable weak bases pose fewer challenges for bioequivalence due to lower probability of the precipitation phenomenon. On a separate note, there are successful IVIVC approaches available for BCS Class IIb APIs which could decrease the risk for non-bioequivalent results. These entail specific in vitro methods (e.g., using a multi-compartment GIT simulator) and in silico models for assessing complex dissolution, supersaturation and precipitation mechanisms [22‐24, 34] and are encouraged for use when developing products with weak-base APIs.

Another reason for observing low occurrence of non-bioequivalent results in the weak-base API group may be the smaller sample size.

The last group of poorly soluble APIs were amphoteric. There were no non-bioequivalent results for these APIs in the 13 fasting bioequivalence studies. These APIs had high solubility at gastric and intestinal pH. As such, they would be dissolved in the stomach and stay dissolved when reaching the intestine. However, when assessing their potential for a BCS biowaiver they would be considered as poorly soluble APIs (due to poor solubility at pH 3 or 4.5), even though they would resemble highly soluble APIs under fasting conditions. Thus, they could potentially be considered for BCS biowaiver approaches. Based on our results, the required testing of solubility across the whole pH range to classify an API as highly soluble for a BCS biowaiver may be too strict. Although we have not seen any non-bioequivalence in three fed bioequivalence studies, more research is needed to assess the risk there, since fed gastric conditions may increase pH to where these amphoteric APIs would behave like poorly soluble weak acids. However, it seems that even in this case the risk is lower, since gastric emptying may be the rate-limiting step for absorption in fed conditions, not the dissolution rate.

Our analysis showed no association between non-bioequivalence outcome and fasting/fed conditions (see Table 4). The 2% difference between percent of non-bioequivalence outcome within fasting and fed bioequivalence studies was far less than the approximately 15% difference observed by Tanguay et al. [4]. They reported 10–12% non-bioequivalence studies in fed conditions and 26.7% non-bioequivalence studies under fasted conditions (N = 1200 bioequivalence studies). The difference we observed is, however, closer to that observed by Lamouche et al., who reported a 14% failure rate within fasted bioequivalence studies and a 10% failure rate within fed bioequivalence studies within BCS Class I APIs (N = 918 bioequivalence studies) [5].

The fed subgroup analyses showed loss of significant differences for Do_12, cPeff and clogP between bioequivalent and non-bioequivalent groups, but the difference remained for the most basic pKa. The loss of significance can be caused by reduced power due to the lower sample size of the fed subgroup. However, it can also be explained by different GIT conditions caused by ingestion of a meal. There was less difference in median Do_12 for the bioequivalent and non-bioequivalent groups in fed compared with fasting conditions. Namely, the pH in the stomach increases to around 3 (or 3.5) in fed conditions [35] and the API solubility at pH 1.2 could become less relevant for prediction of in vivo behavior. But surprisingly, Do_3 and Do_45 also do not differ significantly between non-bioequivalence and bioequivalence in the fed condition, as well as in fasting. On the other hand, there was almost no difference in median clogP between the bioequivalent and non-bioequivalent groups in fed conditions, whereas it was significant in fasting. The ingestion of a meal triggers higher secretion of solubilizing bile salts, prolongs gastric emptying time and increases GIT fluid volumes, etc. [35] and all these effects could make solubility (e.g., Do) and lipophilicity (e.g., clogP) related parameters less relevant for assessing bioequivalence outcome in fed conditions. Interestingly, what remains unexplained is the significant difference of most basic pKa between the non-bioequivalent and bioequivalent groups, where the absence of basic functional groups in APIs (weak acids) still seems important for bioequivalence outcome even after meal ingestion, but apparently not by impacting solubility.

Fed bioequivalence studies in our database never had more discriminatory power to conclude non-bioequivalent results compared with fasted studies, that is, they always concluded the same outcome as studies in fasted conditions or passed bioequivalence when fasting study concluded non-bioequivalent results. This is acknowledged by most regulatory agencies, that is, that studies in fasted conditions are more discriminatory to detect differences between the test and reference product. Some agencies still request bioequivalence in fasted and fed conditions when a medicinal product is to be taken regardless of food and the product is not considered complex (e.g., solid dispersion). To avoid potentially non-discriminatory in vivo fed studies, alternatives are already suggested in the literature (e.g., PBPK modeling [36]).