“Natural” is not Synonymous with “Safe”: Toxicity of Natural Products Alone and in Combination with Pharmaceutical Agents

Abstract

During the 25 years since the US Congress passed the Dietary Supplement Health and Education Act (DSHEA), the law that transformed the US Food and Drug Administration’s (FDA’s) authority to regulate dietary supplements, the dietary supplement market has grown exponentially. Retail sales of herbal products, a subcategory of dietary supplements, have increased 83% from 2008 to 2018 ($4.8 to $8.8 billion USD). Although consumers often equate “natural” with “safe”, it is well recognized by scientists that constituents in these natural products (NPs) can result in toxicity. Additionally, when NPs are co-consumed with pharmaceutical agents, the precipitant NP can alter drug disposition and drug delivery, thereby enhancing or reducing the therapeutic effect of the object drug(s). With the widespread use of NPs, these effects can be underappreciated. We present a summary of a symposium presented at the Annual Meeting of the Society of Toxicology 2019 (12 March 2019) that discussed potential toxicities of NPs alone and in combination with drugs.

The Society of Toxicology 2019 Annual Meeting included a symposium entitled “‘Natural’ is not Synonymous with ‘Safe’: Toxicity of Natural Products Alone and in Combination with Pharmaceutical Agents”. This symposium, which included speakers from the FDA, drug development, and academic institutions, was created to inform participants about the potential risks associated with the use of common NPs, including dietary supplements, botanicals, and herbal products. Cannabis use and novel methods to evaluate NP-drug interactions were presented.

Introduction

Since implementation of the Dietary Supplement Health and Education Act (DSHEA), passed by the US Congress approximately 25 years ago, the dietary supplement market has grown from approximately 4,000 to more than 80,000 products today. The FDA’s priorities in the broad landscape of dietary supplements and other natural products (NPs) are safety, product integrity and informed decision-making (https://www.fda.gov/NewsEvents/Newsroom/PressAnnouncements/ucm631065.htm). Throughout this discussion, terminology from the National Center for Complementary and Integrative Health is used, considering NPs as an umbrella term that includes a large and diverse group of substances produced from a variety of sources, including marine organisms, bacteria, fungi, and plants (https://nccih.nih.gov/grants/naturalproducts), and includes dietary supplements, botanicals, and herbal products. The safety of NPs is often un- or understudied prior to market entry, creating the potential for adverse reactions with the NP alone and with concomitant drugs and/or other NPs. Standardized protocols and in vitro assays for predicting and mitigating NP-drug interactions and adverse events have not been established and dedicated clinical studies may be prohibitively expensive. Additionally, some NPs contain unknown and uncharacterized constituents or adulterants that can place consumers at great risk.

In April 2019, the FDA issued several warning letters about unapproved products that contained cannabidiol (CBD). The FDA analyzed some of these products and determined that the reported content of CBD was incorrect (https://www.fda.gov/news-events/public-health-focus/warning-letters-and-test-results-cannabidiol-related-products). In 2018, a law was passed that exempted hemp (i.e., Cannabis)-derived products from classification as controlled substances if they contained less than 0.3% Δ9-tetrahydrocannabinol (THC) on a dry weight basis. However, the bill “preserved FDA’s authority to regulate products containing cannabis or cannabis-derived compounds” (https://www.fda.gov/news-events/public-health-focus/fda-regulation-cannabis-and-cannabis-derived-products-questions-and-answers#farmbill). However, Epidiolex^®, a pharmaceutical CBD product, was approved by the FDA for specific pediatric epilepsy disorders. FDA has concluded that CBD may not be sold as a dietary supplement due to provisions in the Food, Drug, and Cosmetic Act that state if a “substance (such as THC or CBD) is an active ingredient in a drug product that has been FDA approved or had been authorized for investigation as a new drug for which substantial clinical investigations have been instituted and for which the existence of such investigations has been made public, then products containing that substance are excluded from the definition of a dietary supplement” (https://www.fda.gov/news-events/public-health-focus/fda-regulation-cannabis-and-cannabis-derived-products-including-cannabidiol-cbd#dietarysupplements). The FDA has taken the following steps to address the use of CBD (https://www.fda.gov/news-events/press-announcements/statement-fda-commissioner-scott-gottlieb-md-new-steps-advance-agencys-continued-evaluation):

1. There was a public hearing on May 21, 2019 for stakeholders to provide input.

2. The FDA is convening an internal working group to identify means to lawfully market CBD products as dietary supplements and in foods.

Thus, a major concern with cannabinoid products, as well as other NPs, is the lack of safety information and the potential to precipitate drug interactions inherent in products composed of hundreds of constituents with seasonally and regionally variable composition. The objective of this summary is to provide a discussion of the current state of affairs regarding NP regulatory guidelines and novel methods for the study and evaluation of NP-drug interactions.

Cannabidiol Pharmacology and NP-Drug Interactions with Anti-Epileptic Drugs

Cannabis is an herb/botanical that can be grown with various composition/ratios of over 100 constituents termed phytocannabinoids, which are pharmacologically active.^1,2 Two of the most well-known and studied phytocannabinoids are the psychoactive THC and non-psychoactive CBD. Although there has been longstanding interest in the use of Cannabis products for the treatment of various medical conditions, rigorous scientific studies to support such uses were lacking until recently, when a pharmaceutical grade, highly purified CBD demonstrated efficacy in the treatment of seizures associated with Lennox-Gastaut and Dravet syndromes via randomized, placebo-controlled trials (RCTs).^3–5 Results from these trials led to FDA approval and rescheduling of this particular product for prescription by healthcare providers.⁶ While the reported efficacy data are encouraging for highly purified CBD in the treatment of specific types of epilepsy, the RCTs and other human studies also investigated safety, tolerability, and potential CBD-drug interactions. These investigations are of utmost importance, as overall public perception is that Cannabis products are completely safe and void of any side effects or drug interactions, largely due to it being a NP.⁷

Another important distinction is the difference between pharmaceutical grade and artisanal Cannabis products. Pharmaceutical grade products are grown and manufactured under strict conditions that provide appropriate quality control and consistency of the products, whereas artisanal products do not have standardized quality control standards (but may be subject to certain standards under jurisdictions where these products can be legally sold) and could be inconsistent or incorrectly labeled. Possible labeling inconsistency with artisanal products has been studied. In a study of 75 edible Cannabis products that listed a specific THC content from US cities where medical Cannabis use and sale is legal,⁸ only 17% were correctly labeled for THC content. Of the incorrectly labeled products, 23% contained less THC than indicated on the label (under-labeled), while 60% contained more THC than indicated on the label (over-labeled). Additionally, 59% of the products had detectable CBD, but only 13% listed CBD content in the label. A separate study investigated the CBD content in 84 products purchased online from 31 different manufacturers⁹ that were tested blindly for content and compared with labeling. Regarding CBD content, 43% were under-labeled, 26% were over-labeled, and 40% were correctly labeled. Further, 18/84 of the products tested had detectable THC but were not labeled with THC content. The samples in the aforementioned studies were not analyzed for contaminants.

Although THC and CBD (and likely other phytocannabinoids) have anticonvulsant activity, the mechanisms likely differ. THC primarily has agonist activity at endocannabinoid receptors (CB₁ and CB₂); when activated in the nervous system, these G-protein coupled receptors promote inhibitory neuronal activity.¹⁰ However, CBD’s anticonvulsant activity is likely independent of direct action on endocannabinoid receptors, as CBD has low affinity for these receptors.^2,11 While the exact mechanism of action is not completely elucidated, it most likely involves indirect action or modulation of the endocannabinoid system. Possibilities include blockade of uptake and hydrolysis of anandamide (ANA), an endogenous ligand of endocannabinoid receptors.¹² CBD also activates and desensitizes transient receptor potential vanilloid 1 (TRPV1) channels, which are targets for ANA and may modulate calcium channels (via increasing calcium influx).¹³ Another likely contributing mechanism is CBD’s antagonist action on G-protein coupled receptor GPR55, antagonism of which enhances inhibitory neural activity.^14,15 Common side effects associated with CBD (Epidiolex) include drowsiness, decreased appetite, gastrointestinal effects (namely diarrhea), and fatigue/weakness/lethargy.¹⁶

Awareness of CBD’s metabolism and action on the cytochrome P450 (CYP) system provides insight into possible CBD-drug interactions. CBD (and THC) are highly lipophilic compounds with low oral bioavailability (<20%).² CBD undergoes substantial first-pass metabolism with extensive involvement of the CYPs; CBD and its metabolites are highly bound to plasma proteins (>94%).¹⁶ CBD is metabolized to 7-hydroxy-CBD via CYP2C19 and CYP3A4. CYP1A1, CYP1A2, CYP2C9, and CYP2D6 may also contribute.^2,11 It is noteworthy that many of the involved CYP enzymes are also polymorphic in humans, leading to further complications in the prediction of drug interactions. CBD also has effects on certain CYPs. For example in mouse models, a single high dose of CBD (120 mg/kg body weight) led to inactivation of Cyp2c and Cyp3a protein.¹⁷ When this dose was administered again, mRNA expression of both enzymes increased, but catalytic activity remained unchanged. Further, CBD was reported to decrease total Cyp hepatic content after repeated administration; these findings are pertinent in beginning to understand possible CBD-drug interations.^17,18

Data from the RCTs and other open label studies of pharmaceutical grade, highly purified CBD have provided some insight on potential CBD-drug interactions, particularly with other anti-seizure drugs. The most well recognized interaction is with clobazam; several studies have demonstrated an increase in the systemic exposure to the active metabolite, N-desmethylclobazam, with increasing CBD dose.^19,20 This interaction is largely due to inhibition of CYP2C19 by CBD and is clinically significant as it leads to increased sedation in patients.^19,20 There has been some suggestion that the presence of concomitant CBD with clobazam may lead to enhanced seizure control;²¹ however, this question has not been answered directly. One open label study demonstrated significantly increased serum concentrations of rufinamide, topiramate, zonisamide, and eslicarbazepine in certain patients with increasing CBD dose, but the absolute changes for these anti-seizure drugs were still within the normal therapeutic range.¹⁹ Finally, although not a pharmacokinetic interaction, an interaction between CBD and valproate has been identified, manifested as an increased incidence of elevated liver transaminases observed in patients taking concomitant valproate.^4,19,22

In summary, convincing data are available suggesting that one pharmaceutical grade highly purified CBD product (Epidiolex) is efficacious in treating seizures associated with two specific epilepsy syndromes; however, like with any medication, side effects and CBD-drug interactions have been identified. It is noteworthy that while these data exist for one product, safety and efficacy data cannot be extrapolated to all other Cannabis products due to differences in content, consistency, and method of delivery. More in vitro models and testing strategies are needed to determine efficacy, safety, and drug interactions with different methods of delivery and different ratios of phytocannabinoids.

In vitro Models and Testing Strategies to Assess NP-Drug Interactions

The scientific literature includes numerous reports of potential of NP (e.g. botanical) extracts and/or specific phytochemical constituents to inhibit CYPs. These studies are typically conducted with isolated in vitro systems such as liver microsomes or baculovirus-infected insect cells containing isolated enzymes. Many of these studies indicate potent inhibition of CYPs, particularly when individual phytochemical constituents are used in these assays. These in vitro systems are suitable as early screening assays and have served the pharmaceutical industry well for years. Unfortunately, for NP-based studies, there is rarely follow-up work conducted in more physiologically relevant systems, including human clinical studies, to further translate findings from these early screening studies.

These preliminary reports have obvious shortcomings; for example, enzyme inhibition is the interaction mechanism typically studied, as these systems cannot fully capture induction potential. Also, these systems do not capture transporter activity (uptake and efflux), which can be key factors in assessing interaction potential. Other shortcomings may not be immediately obvious, but potentially more impactful. Once published in the scientific literature, these data are incorporated into various public databases with little curation of study quality. Consequently, this may cause undue concern for consumers and healthcare professionals regarding the concomitant use of certain NPs with prescription medicines. These individuals then have difficulty making informed decisions about the use of NPs.

For the few NP extracts of which interaction potential has been studied in both in vitro and clinical studies, the correlation has been poor. In a recent review, the authors compared results from available preclinical data (largely in vitro) with clinical results across at least 16 different botanical-based NPs or phytochemical constituents.²³ Only for a few examples, such as Echinacea purpurea, garlic (Allium sativum), goldenseal (Hydrastis canadensis) and St John’s wort (Hypericum perforatum), did in vitro studies predict similar findings on CYPs in human subjects. However, as mentioned previously, the in vitro systems used in these studies, although appropriate for screening-level work, are known to have poor direct extrapolation to clinically meaningful predictions.

The authors of the aforementioned review acknowledged that there are a number of explanations for the poor correlation of in vitro data to clinical findings, including the in vitro model(s) used, and the lack of including CYP induction in addition to inhibition assessments, which can impact the ability to predict clinical relevance. Other factors leading to a lack of correlation between in vitro studies and clinical findings include difficulties in comparing botanical materials across studies to ensure similarity of the test article, and the solubility of the complex mixtures represented by botanical extracts can be challenging to study in certain in vitro systems.

Alternative in vitro models and testing strategies are necessary to assess NP-drug interactions.^24–28 One such model includes sandwich-cultured human hepatocytes (SCHH) and an in vitro clearance approach that treats the complex botanical mixture as a single entity, regardless of the constituent profile.²⁹ Hepatocytes in sandwich culture provide a fully integrated hepatic cell system that maintains drug clearance pathways (metabolism and transport) and key regulatory pathways necessary for quantitative assessments of NP-drug interaction potential.^30,31 Because the system integrates biliary excretion, intracellular concentrations are likely more physiologically relevant, resulting in more accurate recapitulation of in vivo conditions compared to other systems. Thus, this in vitro approach includes both mechanistic (e.g., inhibition and induction) and net effect (e.g., intrinsic clearance) studies to demonstrate the effectiveness of this model to make predictions that are translatable to the clinic. Using this approach with NP extracts, it is possible to evaluate the competing effects of enzyme induction or inhibition among the constituents within these complex mixtures. This ability to capture an overall effect on drug clearance by complex NP mixtures cannot be observed with a traditional deconstructionist approach (i.e. studying individual phytochemical constituents in isolation).

In recently published work, the SCHH model was used to predict clinically relevant NP-drug interaction potential using two species of Schisandra: S. sphenanthera (SSE) and S. chinensis (SCE).²⁹ SSE was selected as a model NP with proven clinical drug interaction potential.^29,32,33 The concomitant administration of SSE with both midazolam and tacrolimus in healthy human volunteers has been shown to significantly alter the pharmacokinetics of both drugs due to inhibition of CYP3A4. In contrast, SCE, a closely related and reasonably popular NP, had only been studied using in vitro systems.³⁴ In fact, a number of individual phytochemical constituents of SCE were reported to cause CYP inhibition using baculovirus-infected insect cells, but no clinical follow-on data are available. Hence, SSE and SCE provide the ideal examples illustrating the dilemma as discussed within this section.

To assess the inhibition potential of both Schisandra spp. on CYP3A4/5 activity using midazolam (MDZ) as the probe substrate, SCHH were exposed to MDZ in the absence or presence of SSE (0.30, 3.0, 30.0 and 300.0 ĝ/ml) and SCE (0.32, 3.2, 32.0 and 320.0 μg/ml), and with and without preincubation with SSE or SCE, to monitor the rate of 1’-hydroxymidazolam formation.²⁹ Results from these incubations demonstrated that both SSE and SCE have the potential to inhibit CYP3A4/5 activity as direct inhibitors, with estimated IC₅₀ values of 4.46 μg/ml for SCE and 1.23 μg/ml for SSE. The IC₅₀ values were left-shifted ≥3-fold for both SSE and SCE when studied with preincubation, confirming that both extracts exhibited the potential to inhibit CYP3A4/5 activity in a time-dependent manner.

As mentioned previously, evaluation of the ability of NPs and/or phytochemical constituents to induce CYPs and transporter activity is rare. Human hepatocytes are the “gold standard” for assessing induction of CYP and transporter activities.³⁵ We assessed induction potential of similar concentrations of SSE and SCE by monitoring effects on CYP3A4 mRNA in SCHH. For comparison purposes, we used known potent inducers of CYP3A4 mRNA, including rifampin and the well-studied botanical St. John’s wort (SJW).^36,37 Induction responses stimulated by SSE (30 mg/ml) and SCE (32 mg/ml) after 72 hours of exposure were greater than the response produced by St John’s wort (SJW) treatment and were ≥73% of the rifampicin treatment (Figure 1).²⁹ The decrease in mRNA content at higher concentrations may be related to cytotoxicity at those concentrations of Schisandra spp. (data not shown).

Figure 1.

Relative-fold change of CYP3A4 mRNA content in sandwich-cultured human hepatocytes following 72 hours of exposure to rifampin (10 mM), SJW (20 mg/ml), SSE, or SCE. Error bars represent 95% confidence intervals.

To further use the SCHH model to predict the overall clinical effect of Schisandra spp., MDZ intrinsic clearance was evaluated following 72 hours of exposure to 3 or 30 μg/ml SSE or SCE. MDZ intrinsic clearance in SCHH treated with SSE or SCE was reduced (by ~50% and ~40%, respectively) compared to solvent (DMSO) control (Figure 2). Results were consistent with inhibition of metabolism, and results with SSE were in agreement with clinical findings that showed a similar reduction (50%) of MDZ clearance in healthy subjects.³³ Likewise, in vitro results observed with SCE indicated that SCE treatment has the potential to cause clinically relevant effects on MDZ clearance. In contrast, the percentage of MDZ remaining versus time profiles demonstrated that MDZ intrinsic clearance was increased in SCHHs treated with the known CYP3A4 inducers SJW (2.36-fold) and rifampin (11.8-fold) compared to solvent control. Thus, the in vitro results for SSE were in remarkable agreement with a decrease in clearance observed in clinical midazolam interaction studies, and in vitro results were in good agreement with clinical interaction studies of well-known precipitants of in vivo drug interactions involving SJW and rifampin (Table 1).^36,37

Figure 2.

Intrinsic clearance of MDZ was calculated from linear regression analysis of the percentage of parent remaining (log transformed) versus time profile to estimate the elimination rate. MDZ intrinsic clearance in SCHH exposed to SSE or SCE (A), and St. John’s wort (B), or rifampin (B), for 72 hours. *P value ≤0.05 compared with solvent control.

Table 1.

Comparison of changes in clearance from in vitro and human clinical natural product-drug interaction studies

Extract	Midazolam Clearance (in vitro)	Midazolam Clearance (in vivo)	Tacrolimus Clearance (in vitro)	Tacrolimus Clearance (in vivo)
Schisandra sphenanthera	49%↓	55%↓	65%↓	50%↓
St. John’s wort	198%↑	209%↑	131%↑	Not Available
Rifampin	1180%↑	2400%↑	Not Available	Not Available

Our retrospective case study demonstrated that an intrinsic clearance approach in SCHH was feasible to evaluate net effect and relative strength of NP-drug interactions. We believe that this in vitro model approach is an appropriate strategy to screen complex mixtures for drug interactions and can effectively predict in vivo NP-drug interactions at a lower cost than clinical studies.

In summary, simplistic in vitro metabolism systems, while useful for screening-level purposes, lack cellular processes (uptake and efflux transport) and adaptive response (induction) function. These inadequacies contribute to the lack of accuracy of microsomes to predict clinically meaningful interactions (NP-based or drug-based). Our results with Schisandra spp. were in good agreement with clinical findings; however, improvements to the study design, and standardized guidelines for the study of all NPs, should be considered, including expanding in vitro assays to predict NP extract constituent bioavailability. One future direction may be to encourage the use of in vivo metabolism and transporter cocktail studies once in vitro assays have identified specific liabilities.

Development of Recommendations for the Evaluation of NP-Drug Interactions Using Green Tea as a Case Study

Recognizing the lack of harmonized guidelines with respect to NP-drug interaction studies, the National Institutes of Health National Center for Complementary and Integrative Health established the Center of Excellence for Natural Product Drug Interaction Research (NaPDI Center) in September 2015.³⁸ The overarching goals of the NaPDI Center are to 1) develop recommended approaches to guide researchers in the proper conduct of NP-drug interaction studies, and 2) apply the recommended approaches while evaluating four carefully selected NPs as precipitants of interactions with clinically relevant object drugs. The NaPDI Center consists of four synergistic cores: Administrative, Pharmacology, Analytical, and Informatics. As implied, responsibilities of the Administrative Core include 1) coordinating administrative and fiscal services for the scientific cores and participating institutions and 2) liaising with the oversight Steering Committee in monitoring progress on completion of each Core’s annual milestones. Responsibilities of the Pharmacology Core include 1) selecting high priority NPs as precipitants of potential clinically significant pharmacokinetic NP-drug interactions and 2) designing and completing human in vitro and clinical NP-drug interaction studies (collectively termed interaction projects) that address existing gaps in the scientific literature and assess clinical relevance. Responsibilities of the Analytical Core include 1) sourcing, acquiring, and characterizing NP study materials in sufficient quantities for the interaction projects and 2) analyzing human clinical samples (plasma, urine) for the object drug/metabolite and NP constituents of interest. Responsibilities of the Informatics Core include creating and maintaining a 1) repository to house the data generated by the Pharmacology and Analytical Cores and 2) public portal that summarizes the NaPDI Center’s interaction projects and provides access to the data repository (https://NaPDIcenter.org/). These activities of the NaPDI Center are expected to facilitate improved design of future NP-drug interaction research.

The Pharmacology Core developed an innovative method to select and prioritize candidate NPs as potential precipitants of clinically significant pharmacokinetic NP-drug interactions.³⁹ Beginning with a list of 47 candidates, compiled from the top-40 selling botanical NPs published by Herbalgram and the University of Washington’s Drug Interaction Database (www.druginteractionsolutions.org, now Drug Interactions Solutions), guided information-gathering tools were used to score, rank, and triage the NPs. Triaging was based on the presence and/or absence of a clinical NP-drug interaction, defined as a ≥20% or <20% change in the object drug area under the plasma concentration vs. time curve (AUC), respectively, as well as relevant mechanistic (e.g., inhibition or induction of drug metabolizing enzymes or transporters) and descriptive (e.g., NP formulation, phytoconstituent composition) data. A qualitative, conceptual decision-making tool, termed the fulcrum model, was applied to the remaining 11 NPs posing clinical NP-drug interaction risk. The following high priority NPs were ultimately selected for the interaction projects: cannabinoids, goldenseal, green tea, and licorice. Since publication of this recommended approach, licorice has been replaced with the emerging NP kratom in a strategic effort to keep pace with public health needs in the face of an ever-changing NP market.

Green tea (Camellia sinensis) was the first of the NaPDI Center’s high priority NPs to be advanced to an interaction project.^40,41 Green tea is one of the most commonly consumed beverages worldwide and is also available as an herbal product, ranking fifth in sales in the US mainstream multi-outlet channel in 2018.⁴² Green tea products are used for cardioprotection, chemoprevention, and particularly weight loss. In vitro and clinical studies suggest green tea has low potential to precipitate NP-drug interactions mediated by several CYPs and hepatic transporters.⁴³ One clinical study showed a canned green tea beverage to decrease the AUC of the beta blocker nadolol by 85% relative to baseline.⁴⁴ Because nadolol undergoes minimal metabolism, coupled with the lack of change in nadolol half-life, the decrease in nadolol AUC was attributed to inhibition of an intestinal uptake transporter by green tea, specifically an organic anion transporting polypeptide (OATP). A subsequent clinical study suggested the major catechin in green tea, (−)epigallocatechin gallate (EGCG), to be a key contributor to this interaction.⁴⁵ These studies, along with the well-known effects of grapefruit juice and other fruit juices on intestinal CYP3A and/or OATP, highlight the intestine as a major site for NP-drug interactions.

Like many botanical NPs, green tea is rich in polyphenolic constituents, including several catechins that are prone to extensive glucuronidation. Accordingly, the UDP-glucuronosyltransferases (UGTs), particularly those expressed in the intestine, represent potential targets for green tea-mediated NP-drug interactions. Indeed, EGCG was shown to be a strong inhibitor of UGT activity (4-methylumbelliferone glucuronidation) in human intestinal microsomes and human embryonic kidney cell lysates overexpressing the gut-relevant UGT isoforms UGT1A1, −1A8, and −1A10.²⁵ Based on these observations, a biochemometrics approach, which combines bioassay and chemometric data,⁴⁶ was used to identify candidate intestinal UGT inhibitors in green tea. Extracts and fractions prepared from four widely consumed green teas (provided by the Analytical Core) were tested as inhibitors of 4-methylumbelliferone glucuronidation in human intestinal microsomes at concentrations ranging from 20–180 μg/mL.⁴⁶ All extracts and fractions showed concentration-dependent inhibition. One biochemometrics-identified fraction from a representative tea that was rich in UGT inhibitors was purified further and subjected to second-stage biochemometric analysis. Five catechins – (+)-catechin, (−)-epicatechin, (−)-epicatechin gallate (ECG), (−)-epigallocatechin, and EGCG –were identified as major constituents in the bioactive subfractions and prioritized for further evaluation. Of these catechins, only ECG and EGCG demonstrated concentration-dependent inhibition, with IC₅₀ values (105 and 59 μM, respectively) that were near or below concentrations measured in 240 ml of brewed tea (66 and 240 μM, respectively).

Based on the compelling screening data, the inhibition kinetics of ECG and EGCG towards the clinically relevant intestinal UGT substrate raloxifene were determined using human intestinal microsomes.⁴⁶ Both constituents were relatively potent inhibitors of raloxifene glucuronidation, with inhibition constant (K_i) values ranging from ~1.0–2.0 μM. These parameters and others were used to populate the following mechanistic static model to predict the increase in raloxifene plasma AUC in the presence of green tea:

$$\frac{{\text{AUC}}_{\text{i}}}{\text{AUC}}\text{=}\frac{\text{1}}{\left({\text{1-F}}_{\text{g}}\right)\text{×}\frac{\text{1}}{\left(\text{1+}\frac{\left({\text{1}}_{\text{g}}{\text{×f}}_{\text{u,g}}\right)}{\left({\text{K}}_{\text{i}}{\text{×f}}_{\text{u},\text{mic}}\right)}\right)}{\text{+F}}_{\text{g}}}$$

where AUC_i denotes the AUC of the object drug (raloxifene) in the presence of inhibitor (ECG or EGCG); F_g denotes the fraction of the object drug that escapes intestinal extraction; f_u,g denotes the unbound fraction of the inhibitor in the gut; and f_u,mic denotes the unbound fraction of the inhibitor in human intestinal microsomes. I_g denotes the inhibitor concentration in the intestine, which was calculated using two methods. Method 1 used the following conventional equation:⁴⁷

$${\text{I}}_{\text{g}}=\frac{{\text{F}}_{\text{a}}\times {\text{k}}_{\text{a}}\times \text{Dose}}{{\text{Q}}_{\text{ent}}}$$

where F_a denotes the fraction of the oral raloxifene dose absorbed into enterocytes; k_a denotes the first-order absorption rate constant; and Q_ent denotes blood flow through enterocytes. Method 2 used average simulated maximum enterocyte concentration for ECG and EGCG in duodenum, jejunum, and ileum using the population-based simulator Simcyp (v15.1; SimCYP, Sheffield, United Kingdom). The model predicted an increase in raloxifene plasma AUC of up to 6.1- and 1.3-fold using I_g values based on Method 1 and Method 2, respectively, prompting clinical testing.

A commercially available green tea product, sourced and characterized by the Analytical Core using their recommended approach,⁴⁰ was used to test the effects of green tea on the pharmacokinetics of raloxifene in healthy adult volunteers (eight men and eight nonpregnant, non-lactating women). The sample size was based on 80% power to detect a 25% change in the primary endpoint [log-transformed raloxifene AUC ratio (green tea/baseline)] with a Type I error of 0.05. A single dose of raloxifene (60 mg) was administered orally alone (baseline) or with green tea (240 mL x 3, evenly spaced over the ~14-hour inpatient study day) in a crossover, fixed-sequence fashion. Plasma was collected (0–96 h) and analyzed for raloxifene, raloxifene-4’-glucuronide, and raloxifene-6-glucuronide by UPLC/MS/MS as described.⁴⁸ The pre-defined no effect range was 0.75–1.33. Preliminary data involving the first eight subjects showed the geometric mean treatment/baseline ratio for both the maximum concentration (C_max) and AUC of raloxifene lay below 0.75 (Table 2);⁴⁹similar trends were noted for both glucuronides. These observations held for all 16 subjects (MF Paine, Department of Pharmaceutical Sciences, Washington State University, personal communication).

Table 2.

Mean change in pharmacokinetic endpoints of raloxifene precipitated by green tea

Geometric mean AUC and Cmax for raloxifene (n=8 subjects)
Raloxifene	Geometric Mean (CV %)		Treatment/Baseline Ratio. [90% CI]
	Baseline	Green Tea
Cmax (nM)	0.57 (59)	0.22 (42)	0.39 [0.29–0.52]
AUC0–96h (nM•h)	11.2 (44)	7.12 (51)	0.64 [0.50–0.80]

Maximum concentration (C_max) and area under the concentration-time curve (AUC) were determined via non-compartmental analysis using Phoenix WinNonlin (v7.0). The pre-defined no effect range was 0.75–1.33.

Results suggested a green tea-raloxifene interaction was evident, but the magnitude of change (~30% decrease in raloxifene AUC) was in the opposite direction predicted by the mechanistic static model (≥30% increase in raloxifene AUC). Raloxifene C_max also decreased by a greater extent (~60%) than AUC (Table 2), which, combined with the minimal change in terminal half-life (<15%), suggested that green tea altered one or more processes in the intestine. Potential alternate or additional mechanisms include inhibition of an uptake transporter in the intestine and/or inhibition of reabsorption of raloxifene via downregulation of β-glucuronidase-producing gut bacteria. These alternate mechanisms are currently under investigation by the NaPDI Center and include 1) testing an extract prepared from the green tea clinical product as an inhibitor, and raloxifene as a substrate, of multiple clinically relevant transporters using established in vitro systems and 2) assessing the influence of the gut microbiota on the raloxifene-green tea interaction in a follow-up pilot clinical study involving six healthy volunteers (MF Paine, Department of Pharmaceutical Sciences, Washington State University, personal communication).

In summary, the NaPDI Center was created to provide leadership and guidance on how best to conduct NP-drug interaction studies. The recommended approaches developed by the Center, based on the aggregate data obtained from studies involving four high-priority NPs, ideally will lead to consensus on evidenced-based guidance to address the many challenges unique to NP-drug interaction research.³⁸ Green tea was the first of these NPs to be advanced to an interaction project. Using established procedures, including those detailed in FDA guidance documents regarding CYP-mediated drug interaction predictions,³⁵ green tea was predicted to increase the AUC of the object drug raloxifene by at least 30%. However, the subsequent clinical study showed the same green tea product to decrease raloxifene AUC by 30%. These observations highlight that current methods used routinely to predict CYP-mediated drug interactions may not be applicable to UGT-mediated drug interactions and/or may require additional tools. As such, studies are ongoing to address this in vitro-in vivo disconnect involving parallel in vitro and clinical studies to address the influence of transporters and the gut microbiota on the green tea-raloxifene interaction. The knowledge gained from this interaction project informed the forthcoming recommended approaches focused on the design and conduct of in vitro and clinical NP-drug interaction studies.

Concluding Remarks

The presenters of this symposium agreed more research is needed to evaluate the safety and efficacy of NPs. CBD, which is now available as a purified and standardized FDA approved isolate, is a powerful example of the potential for clinically relevant interactions between a NPs and conventional drugs. Fortunately, in vitro tools such as SCHH are available to predict these complex interactions, particularly if CYP-mediated, aiding healthcare providers and consumers in avoiding adverse events (such as those observed with CBD) in the absence of dedicated clinical interaction studies. However, standardized guidelines for the use of SCHH, other in vitro tools, and clinical studies have not been established, leading to difficulties in translating and comparing published data. As such, the NaPDI Center was created to develop recommended approaches to guide researchers on the proper conduct of NP-drug interaction studies by evaluating four high priority NPs as precipitants of interactions with clinically relevant object drugs.

During the discussion following the formal presentations, participants raised several concerns regarding research strategies and regulatory guidance for NPs. One participant raised the concern that testing modalities (both in vivo and in vitro) did not test different preparation methods (e.g., hot vs. cold brewed teas) which vary by country and culture, leading to misleading or regionally spurious recommendations. Others encouraged a balanced approach that considered the possible benefits of NPs and wanted an improved framework for determining risk-vs-benefit of NPs. Finally, there was significant discussion regarding federal regulation of NPs; many participants believed that federal guidelines for efficacy and quality would support more widespread and safer use of NPs in the community, while others were concerned that access to NPs would be hindered by increased regulation and would render NPs unaffordable.

We believe that a harmonized approach to study NPs, along with novel in vitro technologies and the development of standardized products with well characterized drug interaction potential, will optimize the safety and efficacy of NPs administered alone or with concomitant medications to improve health outcomes.