Food Effect on Oral Bioavailability: Old and New Questions.

Abstract

Food and drug interaction or “food effect” has been broadly acknowledged to be a complicated topic that deserves more research effort so that it can be more optimally addressed during drug research and development. The effect of ingested food on oral bioavailability of a drug is the result of a complex interplay of active pharmaceutical ingredients, formulation, gastrointestinal (GI) physiology, and meal composition, which is usually categorized by fat content, ie, high, moderate, or low fat. Differences in fat content of a meal can have varying impacts on the magnitude of a food effect; an insignificant impact of fat content on food effect is more likely for compounds with high solubility and low permeability, where food would significantly reduce the rate and extent of drug absorption. Food effect can be caused by various factors with the following commonly accepted attributes: for “positive” food effect (ie, food increases drug absorption and exposure), food enhances drug absorption by increasing drug product dissolution and active pharmaceutical ingredient solubility, stimulating bile flow, delaying gastric emptying, increasing splanchnic blood flow (with hepatic blood flow upsurge likely playing a most significant role), enhancing lymphatic uptake, or inhibiting metabolic enzymes and efflux transporters in the first-pass route. For “negative” food effect (ie, food decreases drug absorption and exposure), food decreases drug absorption by stimulating secretion of gastric acid that breaks down acid-labile drugs, interacting with drugs physically or chemically (nonspecific binding, sequestration), or increasing intestinal motility.1-4 The effect of food on hepatic blood flow has been appreciated by scientists for many years. Although the exact mechanism by which food increases hepatic blood flow in man remains to be elucidated,5-7 it has been reported that some endogenous substances, such as cholecystokinin-pancreozymin, glucagon, and secretin, increase hepatic blood flow in animals.8 It is possible that the increased secretion of digestive hormones after meal intake causes a transient elevated hepatic blood flow.9 During the absorption of oral drugs, faster hepatic blood flow would accelerate the uptake of drug molecules from the portal vein into systemic circulation just as would escaping first-pass metabolism, which may enhance the systemic exposure for some drugs with low to moderate extraction ratio. It has been suggested that the food-induced increase of hepatic blood flow may reach a maximum 20% to 60% elevation in 40 to 80 minutes postmeal and last for up to 3 hours.9 It is also likely that the increased hepatic blood flow would expedite the elimination of drugs with high extraction ratios during and after the absorption, resulting in reduced systemic exposure. On the other hand, a food-induced increase of overall splanchnic blood flow could improve drug absorption by shunting blood flow around the liver during absorption, which may apply to highly extracted drugs.10 Hence, given the dynamic process in the drug absorption phase, the net result of food effect on bioavailability is a function of the combination of the aforementioned relevant factors plus other elements such as absorption rate and duration as well as hepatic extraction ratio. The food effect is statistically defined using 90% confidence interval (CI) of the ratio of fed state to fasted state population geometric means of drug systemic exposure (AUC and peak plasma concentration); ie, if the 90%CI falls outside the bioequivalence criterion (0.80-1.25), a nominal pharmacokinetic (PK) food effect is designated.11 However, drug exposure alterations elicited by food may or may not have clinical significance depending on a combination of factors such as efficacy and safety endpoints and the therapeutic window. Drug systemic exposure and the pharmacodynamic (PD) effect of most drugs are correlated, making the change in bioavailability an important parameter of food and drug interactions.12 The relationship between PK parameters and pharmacodynamic effects is often not straightforward and may only be evaluated if the impact of the change in PK on pharmacodynamics can be quantified. The hemodynamic response of the calcium channel antagonist nifedipine significantly correlates to its plasma concentration. Food increases the bioavailability of the nifedipine sustained-release formulation by 28% to 31%, which is reflected in a significantly increased hypotensive effect.13-15 Grapefruit juice increases the bioavailability of felodipine by 284% by inhibiting its first-pass metabolism, and the clinical effect on blood pressure and heart rate is doubled.16 Food intake decreases the bioavailability of perindopril by 35%, resulting in a clinically significant reduction in the inhibition of angiotensin-converting enzyme.17 However, not all food-induced significant changes in bioavailability have meaningful clinical significance. The bioavailability of pravastatin is reduced by 31% when taken with food, yet its lipid-lowering efficacy is unchanged.18 Food increases the bioavailability of the antimalarial agent mefloquine by 33% to 40%, but this change in the drug exposure has little consequence in the treatment of malaria due to its overall high oral bioavailability and likely a relatively wide therapeutic window.19 Because of the complex interplay of different factors mentioned above, it is not simple to determine if a food-induced change in bioavailability will have clinically meaningful consequences for efficacy and safety. Predicting clinical food effect in new drug development is challenging, especially in quantitative terms. Nevertheless, progress in this field has been made in recent years in parallel with the publication and application of the Biopharmaceutics Classification System (BCS)20 and Biopharmaceutics Drug Disposition Classification System (BDDCS).21 Fleisher et al analyzed various marketed drugs of different BCS classes based on rate-limiting factors for absorption including drug properties and GI physiology. They suggested that the general direction of food effect on absorption may be predicted based on BCS class.22 Wu and Benet further summarized the relationship between this effect on drug absorption and BCS class as described in Table 1.21 BCS was a significant step in setting up rational science for allowing waivers of in vivo bioavailability and bioequivalence testing (“biowaiver”) of immediate-release dosage forms for high-solubility and high-permeability drugs when such drug products also exhibit rapid dissolution.19, 20, 23 Based on its scientific theory and corresponding categorization of various marketed drugs, the utility of BCS has been extended to the prediction of direction of food effect on drug absorption, ie, positive, negative, or neutral (no clinically meaningful change in drug absorption and exposure), as depicted in Table 1. Wu and Benet noted that the major uncertainty in the application of BCS related to the permeability assignment, although there were also some difficulties in differentiating solubility classes. Therefore, they suggested it might be more useful to replace the permeability criteria with the major route of drug elimination by assigning drugs to BDDCS classes, as shown in Table 2.21 They believed that it would be easier and less ambiguous to determine the assignment of BDDCS classes based on the extent of metabolism than it would by using permeability in BCS classes. They further noted that, in general, BCS class 1 and 2 drugs are highly metabolized, whereas BCS class 3 and 4 drugs are primarily excreted unchanged via the biliary or renal routes. The research data indicated that changing the permeability element to route of elimination in BDDCS markedly expanded the number of class 1 drugs eligible for biowaiver of in vivo bioavailability/bioequivalence studies, which was supposed to further reduce the regulatory burden as compared with the application of BCS.21 Although the application of BDDCS to the prediction of food effect has not been published so far, based on the aforementioned observations (ie, BCS class 1 and 2 drugs are highly metabolized, whereas BCS class 3 and 4 drugs are primarily excreted unchanged via the biliary or renal routes), it is reasonable to say that the prediction on the direction of food effect would be, in general, similar in BDDCS classes as in BCS classes. Progress in the prediction of food effect has also been acknowledged with other literature published in the past few years. Singh used a quantitative methodology to probe the dependence and correlation of food effect with aqueous solubility, dose/solubility ratio, and log10 partition coefficient for immediate-release orally administered drug products.24 The study compared these physicochemical parameters with fed/fasted AUC obtained from clinical studies for over 100 structurally diverse compounds. Its results pointed out the estimated range within which a drug is not expected to have significant food effect is 0.148 to 89.39 mg/mL for aqueous solubility and is 0.23 to 624 mL for dose/solubility ratio. The corresponding range of log10 partition coefficient is –1.13 to 2.98. It was also clear that food effect may be correlated to these physicochemical properties, but it was difficult to predict the magnitude of AUC change between fasted and fed treatments.24, 25 Gu et al used published clinical food effect data of 90 marketed drugs (92 total entries due to duplicate use of 2 drugs either at 2 different doses or as 2 different formulations) and a logistic regression approach to investigate and establish the relationship between the effect of food on AUC and physicochemical parameters, including solubility, dose number (dose/solubility), log10 distribution coefficient, and maximum absorbable dose, which is a parameter that combines information on dose, solubility, and permeability.26 In these 92 data entries there was a reasonably good probability (80%) of correct predictions on the category of food effect, ie, positive, negative, or neutral. It was also found that there was a marked difference in accuracy of prediction among these different kinds of food effect. The correct predictions were 97%, 79%, and 68% for compounds with positive, negative, and neutral effects, respectively. Overall, these methods may well predict food effect direction, but unfortunately they are unable to predict the magnitude in the change of PK exposure parameters (AUC and peak plasma concentration) in the presence of food. In addition, these methods will not be able to compare different formulations, nor will they provide a mechanistic understanding of the cause of the food effect.25 Nevertheless, these methods still offer an advantage in the early stage of drug discovery to screen compounds for food effect, as physicochemical parameters such as solubility and permeability are relatively easy to obtain with a minimum amount of drug substance. It has also been reported more recently that physiologically based PK modeling and simulation may quantitatively predict a compound's PK profiles under the fed or fasted condition, integrating the data generated during preclinical and clinical research and development such as permeability, biorelevant solubility, dissolution of a compound, and (if needed or available) in vivo preclinical and clinical PK data or corresponding simulated values.1, 27 Parrott et al applied physiologically based PK modeling for prediction of food effect of theophylline and aprepitant, using the computer program GastroPlus™ (SimulationsPlus, Buffalo, New York). The effect of food for immediate- and controlled-release formulations of theophylline, a BCS class 1 compound, was well simulated by the dog and human models in comparison with observed in vivo study data. However, the prediction for aprepitant, a BCS class 2 compound, was more challenging and less accurate because of uncertainties such as translation of aqueous solubility into in vivo solubility, relevant permeability value, and changes in permeability and solubility in different regions of the GI tract.27 It was also concluded that direct prediction of the clinical food effect for aprepitant without prior verification against preclinical in vivo data was not advisable. Jones et al developed physiologically based PK models in GastroPlus™ and simulated the impact of food on pharmacokinetics of 6 Roche compounds using permeability, biorelevant solubility, elimination, and distribution data. The models were able to capture the magnitude of the food effect, and, for the majority of compounds in the study, they well predicted the observed plasma exposure in various food states.1 Animal in vivo studies are often used to address the difficulty in food effect prediction. The dog is the most studied species for understanding and predicting human food effect because it is easy to administer various formulations and test meals.25, 28-30 There have been cases in which dog models worked well for human food effect prediction. Lentz et al reported development and validation of a dog food effect model. A number of different BCS class compounds were studied in pentagastrin-pretreated dogs. Fed/fasted peak plasma concentration and AUC ratios using a 50-g aliquot of the US Food and Drug Administration (FDA) high-fat meal11 in the dog were found to be in the closest qualitative agreement with human data. The model predicts changes in human exposure in the presence of food, which will be helpful for determining food effect as a liability during the discovery stage.31 Merck's Emend® (Kenilworth, New Jersey), the antiemetic agent aprepitant (BCS class 2), provides another example of using dog as an animal model to predict drug exposure change under the fed condition in humans and to assist formulation development in order to diminish food effect. The significant food effect found in humans was first confirmed in the dog at a similar magnitude using the suspension of micronized bulk drug for the tablet formulation of aprepitant. A clear correlation between particle size and in vivo exposure was further demonstrated in dog studies. A nanoparticle colloidal dispersion formulation offered the highest exposure and also eliminated the food effect on oral absorption in the dog. Results from human studies using the same formulation showed excellent correlation with the findings in the dog, which led to a drug label stating that aprepitant could be taken with or without food.32 However, effects observed in dogs are not always translatable into humans and may under- or overestimate. The interspecies differences in bioavailability may be attributed to factors such as loose epithelial junctions in the dog, inhibition of intestinal P-glycoprotein by high bile salt levels in fed dogs, the volume of test meal administered to the dog, differences between dogs and humans in presystemic metabolism, GI physiology, test meal, and formulation.25, 33-35 Although all the aforementioned methodologies for the prediction of food effect in humans, including in silico approaches using physicochemical and biopharmaceutical properties, physiologically based PK modeling and simulation, and in vivo animal studies, may predict the direction and, under certain circumstances, the approximate magnitude of food effect in humans, a universally reliable approach that may quantitatively predict such effects in various scenarios is still lacking. This further reflects the complex nature of the food effect and the underlined challenges in addressing it. In addition, prediction of a clinically meaningful food effect is even less practical given that the relevant clinical significance of PK changes is unknown at the time of such exercises. Due to the complexity of mechanisms of food effect and the variation in interpretation of clinical significance by different health authorities, the label for the same drug with regard to food intake can vary across countries and regions. Some examples of such label differences between the United States and the European Union are listed in Table 3. The causes of the inconsistency in global regulations may be multiple, but lack of timely communications and data sharing among the stakeholders likely play an important role in this scenario. Because of the differences in food effect evaluation among health authorities across the world, sponsors oftentimes need to engage in extensive interactions on this topic during global market authorizations. The regional inconsistency in drug labels in relation to food effect is a reflection of the nature of its potentially disputable and uncertain clinical implications and also a suggestion for conversations among regulators for future opportunities to better harmonize drug labels globally for physicians and patients. Increased interaction among global regulators to align assessment and opinions may result in more consistent labeling, which will ultimately benefit patients and caregivers in our global health care systems. Because the FDA and European Medicines Agency are the 2 major world health-care regulators, frequently referenced by other counterparts, harmonization between the FDA and the European Medicines Agency on this topic of interest would have profound benefits. Interestingly, there have also been calls for using food effect as a tool to aid drug development and clinical use of medications in certain circumstances. These thoughts include taking advantage of positive food effect to reduce the drug dose in oncology therapy, which may possibly lead to lower treatment cost,36, 37 and improving poor oral bioavailability of some anti-HIV drugs such as Invirase® (saquinavir mesylate),38, 39 or dosing the drug with food to reduce adverse effects in the GI tract, eg, for Cognex® (tacrine) and Gleevec® (imatinib mesylate).40 Research findings have also been published on a systematic inconsistency between oncology and nononcology drugs in how food drug interactions are applied in their labels. It was found that nononcology drugs tended to take advantage of marked positive food effect in the labels on most occasions, ie, to be dosed in a fed state, but the opposite labeling pattern was observed for oncology drugs reviewed in that study.41 It was argued that enhancement of oral drug bioavailability via the dosing strategy offered a number of benefits including reduced GI toxicity (from unabsorbed drugs), decreased intra- and interindividual variability in drug exposure, and improved pharmacoeconomic efficiency.41 In the meantime, other scholars have clearly expressed their different opinions on the proposal to change the labels in order to take advantage of positive food effects to lower drug dose for some marketed oncology medications on the grounds that, for cancer patients, either disease or concurrent medication can induce nausea, anorexia, or other GI symptoms that may result in variations in daily tolerance of fatty foods, leading to significant intradose variation in bioavailability and intrapatient variability in effectiveness and safety of the drug.42, 43 Despite the different views in this public discussion on applying certain types of food effect to clinical drug administration to maximize oral drug delivery, it is generally agreed that drug developers should conduct food effect studies early in product development to guide the mode of drug administration in relation to food intake in late-phase (eg, phase 3) clinical development.41, 42 This practice is consistent with the potential use of certain types of food effect to improve drug delivery and absorption and would address the concerns over compromised effectiveness and safety of the therapy if such a use of food effect is not based on the data from a full development program. Thus, the question now is if this is a practical and beneficial approach. Although there may be a trend to avoid a marked food effect in drug development where possible, certain investigational drugs with positive food effects may well be developed into drug products utilizing the effect for enhanced bioavailability. This practice in drug development needs to be advocated and discussed further for the greatest benefits involving stakeholders from all parties, including industry, regulators, and academia. Active discussions were held recently on the topic at a workshop (“Evaluating and Modernizing our Approaches for Food-Effect Assessment”) cosponsored by the FDA, American College of Clinical Pharmacology, American Association of Pharmaceutical Scientists, and American Society for Clinical Pharmacology and Therapeutics in February 2015, and at a symposium (“Food Effect Confrontation: Exploring Clinical Significance”) cochaired by the author of this article at the ACCP Annual Meeting in September 2015. As mentioned, certain food effects may be used to aid the development of an optimal drug treatment regimen, but it should also be acknowledged that there are many different factors that may have significant impacts on benefit/cost calculations and decision making. Taking this beneficial potential into account for drug candidates early in the drug discovery and development stage is very important. The physicochemical and biopharmaceutical properties of a candidate would inform drug developers of the likely beneficial use of food before the clinical phase. For some drug candidates with very poor solubility and anticipated low bioavailability, achieving clinically efficacious exposure may be a daunting challenge or sometimes impractical via formulation work alone. Dosing such drug candidates in the fed condition may significantly improve the dissolution and absorption to achieve target therapeutic exposure and also very possibly help reduce variabilities. To avoid the dilemma of additional significant work at a late stage, the assessment of this beneficial potential should be started in first-in-human trials with a design tailored for case-based scenarios. Depending on target patient populations, the type of meals to be used is of importance in the design, knowing that this beneficial effect can be associated with meals of different fat content; eg, a light meal, which is usually more appropriate and popular in the clinic, may work to achieve the goal. Right after the first-in-human trial, the chosen treatment regimen should be continuously employed in subsequent clinical trials throughout development phases and eventually be included in the drug label if market approval is achieved. Additional dedicated registration food effect studies may or may not be conducted in parallel with late-phase clinical development depending on the needs and agreements with regulatory agencies. For this type of drug candidate with aforementioned physicochemical and biopharmaceutical properties, formulation and dose change in late development phase would usually not have a significant impact on drug dissolution and absorption, alleviating the concerns over translating clinical data from early to late phases. A reasonable amount of additional investment and effort in early clinical trials, to define an optimal choice of the regimen involving food for downstream clinical investigations, is justified for long-term returns. Although it is deemed an option to address practical challenges in handling with poor dissolution and bioavailability for some investigational drugs, the approach and vision discussed here may not practically apply to all drug development projects across various disease areas. For example, in highly unmet medical need areas such as oncology, the pace of development may dominate over other considerations, especially for drug candidates that cannot be tried in healthy subjects. Under the circumstances any encouraging response rates in a fast-running phase 1 trial would lead to a rapid entry into a pivotal trial without delay. In this scenario, food effect assessment and evaluation on its potential beneficial use for dosing regimen may present real challenges or even be unrealistic in the course of development. However, it should be pointed out that dosing a drug with a meal has also been used in oncology drug development and labels; eg, the Stivarga® (regorafenib) US label instructs patients to take it with a low-fat meal, which is based on clinical development trials that used the same regimen. Food effect is complex and an inevitable topic in the development of new oral medicines. Consistent regulations on this subject across the world will benefit the improvement of global health-care systems. Given the advancement in science and technology, our world is becoming smaller and more reachable with ever-increasing mobility for all global citizens. The harmonization of global regulations on this matter becomes more relevant and important, although there may need to be a great deal of joint effort to achieve the goal. To facilitate the global harmonization, timely publication of food effect studies should be encouraged in mainstream journals with easy access to all readers. Awareness of food-drug interactions could help to improve patients’ compliance with some drugs; eg, administration of erythromycin and nonsteroidal anti-inflammatory drugs with food reduces local adverse GI symptoms that may otherwise lead to noncompliance.12 However, it should also be mentioned that given the complexity of food and drug interaction, adverse GI effects may be caused by either local or systemic determinants. These adverse effects may also sometimes be attributable to formulation or the mass of dosage form during the dissolution/peri–gastric emptying period, so it is difficult to discern their root cause without actual clinical studies. In the early development stage it would be a challenge to choose a dosing regimen with a meal in order to ameliorate adverse GI effects because a high-strength final market image formulation to be used in late clinical trials may alone address the issue if it is solely formulation rather than an active pharmaceutical ingredient-related one. Timely sharing of all food-drug interaction study data during development would help global communications on food effect discussion and enable more consistent regulatory decisions. Given that these study results are not usually published in easily accessible literature, the sponsors and mainstream journals will need to work together to promote and implement this initiative. Additionally, regulators may also support this type of timely data publication by encouraging/urging sponsors to achieve the goal in their early-stage communications. There is the potential to make good use of certain food effects in clinical drug development so oral drugs with relevant physicochemical and biopharmaceutical properties may be used with maximized bioavailability, optimal safety, better compliance, and possibly improved pharmacoeconomic efficiency. Such use of food effect in clinics should be based on proven beneficial results from a product's full development program. A great amount of work needs to be done to promote the good use and global harmonization of drug labels for food effect in drug development and regulations. Global joint efforts and active participations by industry, regulators, and academia are equally important in this endeavor. The author thanks Dr Gangadhar Sunkara and Dr Orin Tempkin from Novartis Pharmaceuticals Corporation for their review, valuable comments, and advice on the manuscript. The author is an employee and stockholder of Novartis Pharmaceuticals.