Toward a More accurate, Evidence‐Based Classification of CYP Inhibitors: A Critical Appraisal of Current Systems

Abstract

In this issue of CPT, Malavé et al. deliver a comprehensive reassessment of CYP2C19 inhibitor classifications through a systematic review of clinical drug–drug interaction (DDI) studies. While their findings bring clarity to specific discrepancies, this commentary explores the origins of such inconsistencies, particularly focusing on methodological, pharmacogenetic, and substrate‐specific complexities. These issues prompt renewed consideration of how CYP inhibitor categories are defined, communicated, and used in clinical and regulatory settings.

The classification of CYP substrates and inhibitors by the US Food and Drug Administration (FDA) has been increasingly used in drug–drug interaction (DDI) management after its introduction about two decades ago. ¹ It defines sensitive substrates as drugs that demonstrate an increase in area under the concentration‐time curve (AUC) of ≥ 5‐fold with strong inhibitors of a given metabolic pathway in clinical DDI studies, while it specifies moderately sensitive substrates as drugs that demonstrate an increase in AUC of ≥ 2‐ to < 5‐fold with strong inhibitors. The system employs a straightforward categorization of inhibitors into three levels: strong, moderate, and weak. Strong and moderate inhibitors increase the AUC of sensitive substrates by ≥ 5‐fold and ≥ 2‐ to < 5‐fold, respectively. A similar classification is used also in the well‐known Flockhart DDI Table™, as well as in the recently adopted harmonized ICH M12 Guideline on Drug Interaction Studies.

CYP INHIBITOR CLASSIFICATION: PRACTICALITY VS. SCIENTIFIC RIGOR

This approach can be lauded for its simplicity and practicality, offering clear guidance in clinical recommendations and aiding healthcare professionals in DDI management. It is particularly useful from a clinical perspective, enabling management of interactions for substrate drugs without requiring specific trial data for every inhibitor–substrate pair. For example, separate dosing recommendations can be given for the substrate drug with strong, moderate, and weak CYP inhibitors. Additionally, the classification aids in selecting suitable inhibitors for clinical DDI trials.

However, this same simplicity becomes its Achilles’ heel. The classification relies primarily on the fold changes in area under the curve (AUC) of a sensitive substrate, where small differences in effect size can shift an inhibitor from “moderate” to “strong” or to “weak,” potentially influencing clinical decisions. The moderate category alone includes a wide range, spanning from 50% inhibition up to 80% inhibition (Figure 1 ). For instance, an inhibitor causing a 4.9‐fold increase in AUC is labeled moderate, whereas anything above fivefold is strong, although in such studies some individuals are typically above and some below fivefold, and the 95% confidence interval crosses fivefold. These outcomes can depend on the substrate’s sensitivity, which varies by population (e.g., due to genetics) or substrate pharmacokinetics (e.g., autoinhibition).

Figure 1.

Relationship between fold increase in CYP substrate AUC and the fraction of the substrate metabolized by the CYP enzyme when 50%, 80%, 90%, or 95% of CYP activity is inhibited. The dashed horizontal line indicates unity, that is, no increase in AUC, and the light gray area displays the moderate inhibition range, that is, an AUC increase ranging from two‐ to fivefold. Of note, a greater than fivefold increase is possible only if the fraction metabolized is at least 80%.

Accordingly, an optimal index substrate should display linear pharmacokinetics and be as sensitive and specific as possible so that sensitivity would not be easily compromised by individual factors. ¹ Greater emphasis should be placed on verifying substrate sensitivity. Beyond requiring an AUC increase of ≥ 5‐fold with a selective inhibitor in clinical DDI studies (a criterion that may involve circular reasoning), in vitro data, pharmacogenetic studies, and PBPK modeling should be considered. Of note, orally administered CYP substrates are generally more sensitive to AUC change than parenterally given drugs, since they can be subject to first‐pass metabolism in the liver and at least in the case of CYP3A4 also in the intestinal wall.

Specificity issues also arise when other pathways, such as membrane transporters, affect the AUC of the substrate drug, and the inhibitor simultaneously inhibits both the CYP and the transporter. In such cases, plasma (or sometimes urine) metabolite to parent ratios can supplement the parent drug AUC‐based approach when they provide better specificity and sensitivity. For example, an 80% reduction in an enzyme selective metabolite to parent drug AUC ratio would be equivalent to an 80% reduction in the enzyme specific clearance, akin to a fivefold AUC increase. Guidelines recommend the use of metabolite to parent ratios in challenging cases when omeprazole is used as a CYP2C19 substrate and in the case of CYP2B6, since the recommended index substrate bupropion is not sensitive in terms of AUC change while the hydroxybupropion/bupropion plasma concentration or AUC ratio is both selective and sensitive. Even when the formation of a metabolite is highly selectively catalyzed by a single enzyme, metabolic ratios can have sensitivity and variability issues, for example, due to factors affecting metabolite elimination. Urinary metabolic ratios are not exempt from such issues and may be more difficult to interpret than plasma AUC‐based ratios. Due to these complexities, guidance on the acceptability of metabolic ratios would be needed. Thereby, such ratios could more often be recommended as an alternative approach in the evaluation of inhibitor strength. Furthermore, between DDI studies, there is often variability in AUC increases due to population differences in substrate pharmacokinetics or methodological variability, which may cause inconsistencies in inhibitor classification. This underlines the need for systematic and standardized approaches.

THE CHALLENGE OF CLASSIFYING CYP2C19 INHIBITORS

Malavé and collaborators noted that there are discrepancies in the classification of inhibitors of cytochrome P450 2C19 (CYP2C19) between FDA and the Flockhart Table. ² Therefore, they set out to carry out a comprehensive literature review of DDI studies on the effects of CYP2C19 inhibitors on CYP2C19 substrates and thereafter meticulously analyze concordance between literature‐reported classifications and those listed in the FDA and Flockhart tables. Based on a systematic literature review, Malavé and collaborators identified a total of 50 publications/records (including FDA documents) that included clinical trial data on the effects of FDA and Flockhart database listed CYP2C19 inhibitors on the pharmacokinetics of CYP2C19 substrate drugs. They prioritized data with sensitive substrates and carefully considered CYP2C19 genotype data (when data were available, IM and PM phenotypes were excluded), inhibitor drug doses, and whether steady‐state for the inhibitor was achieved. As there were several clinical DDI studies for many of the substrate‐inhibitor pairs, they adopted a “predominant” weak/moderate/strong classification based on the class of greatest frequency with literature‐reported fold increases in AUC. This method exposed critical discrepancies, highlighting the challenges of maintaining accuracy in classifications, and proposed recommendations to refine inhibitor classifications. Their conclusions included for example that at typical clinical doses fluoxetine and fluconazole are strong CYP2C19 inhibitors and omeprazole and esomeprazole are weak inhibitors. With the evidence, they were also able to suggest footnotes describing dose‐dependent CYP2C19 inhibition for fluvoxamine and fluconazole.

CYP2C19 is one of the key drug‐metabolizing enzymes, characterized by extensive genetic variability. Its substrates include a diverse range of drugs such as abrocitinib, amitriptyline, brivaracetam, citalopram, clobazam, clomipramine, clopidogrel, doxepin, escitalopram, imipramine, lansoprazole, mavacamten, omeprazole, pantoprazole, sertraline, trimipramine, and voriconazole. ³ Genetic variation in CYP2C19 can result in phenotypes ranging from poor metabolizers (PMs) to ultra‐rapid metabolizers (UMs), affecting drug exposure and DDI potential of these drugs. ⁴ CYP2C19 can also be subject to DDIs mediated by inhibition and induction of the enzyme. A notable example is clopidogrel, whose active thiol metabolite formation depends on two metabolic steps, partly mediated by CYP2C19. This makes its antiplatelet effect highly susceptible to inhibition of CYP2C19. For example, the CYP2C19 inhibitory drugs fluoxetine, fluvoxamine and omeprazole can increase the risk of ischemic events in cardiac patients treated with clopidogrel. ⁵ , ⁶

With regard to inhibitor strength classification, CYP2C19 is probably one of the most challenging CYP enzymes, because there is substantial genetic variability in CYP2C19 activity, spanning from lack of activity to ultrarapid metabolism, and because there is a paucity of optimal sensitive substrates. Genetic variability can cause variability in the AUC increase by affecting both the index substrate and inhibitor drug. The most common index substrate omeprazole is partly metabolized by CYP3A4, whose role becomes greater when CYP2C19 activity is poor. ¹ Omeprazole is also an inhibitor of CYP2C19 leading to autoinhibition and nonlinearity at higher omeprazole concentrations. ⁷ , ⁸ Although Malavé et al. ² concluded that normal daily doses of omeprazole or esomeprazole (40 mg or below) cause weak rather than moderate CYP2C19 inhibition, already a four times higher daily dose (160 mg esomeprazole) can cause almost a fivefold increase in sensitive substrate AUC, ⁷ highlighting the possibility that autoinhibition can contribute to omeprazole AUC increase when it is used as the index drug with moderate or strong CYP2C19 inhibitors. Of note, there is strengthening evidence that significant amounts of CYP2C19 protein are expressed in the small intestinal wall, ⁹ which may further add to variability in CYP2C19‐mediated DDIs. These complexities help explain divergent reports in fold increase in AUC for CYP2C19 inhibitors. ²

Voriconazole, in particular, illustrates the classification challenge. There were almost equal numbers of reports in EM subjects with voriconazole causing an < 5‐fold and an > 5‐fold increase in omeprazole AUC. ² The authors decided to give weight to an intravenous omeprazole microdose study, ⁸ to mitigate the autoinhibition issue with omeprazole. This study did not quite reach the AUC threshold for strong inhibition, and the classification was therefore kept as moderate. ² As additional complexities, voriconazole is both a substrate and inhibitor of CYP2C19, but also a strong inhibitor of CYP3A4. In such a case, the 5‐hydroxyomeprazole to omeprazole AUC ratio approach can be preferred over the omeprazole AUC approach. In fact, even in the intravenous microdose study, as well as in most studies with oral omeprazole, the 5‐hydroxyomeprazole/omeprazole AUC ratio approach is usually consistent with an almost 90% inhibition of CYP2C19 after a single 400 mg dose of voriconazole. ⁸ , ¹⁰ Accordingly, given the lack of steady‐state studies, the available evidence supports classifying typical voriconazole doses as causing strong CYP2C19 inhibition.

REFINING THE FRAMEWORK: FUTURE DIRECTIONS

The findings by Malavé et al. ² underscore the need for improved inhibitor classification systems that accommodate the complexities of enzyme activity, genotypic differences, inhibitor doses, and substrate characteristics. Their recommendations—focused on improving accuracy and clinical utility—pave the way for future efforts to develop a more dynamic and reliable framework for classifying metabolic enzyme inhibitors. Such a framework should adapt a comprehensive scientific approach that better incorporates in vitro data, pharmacogenetic data, physiologically based modeling approaches and details from clinical DDI studies, perhaps in a similar systematic manner as the Clinical Pharmacogenetics Implementation Consortium does in pharmacogenetics. Ultimately, such efforts will benefit clinical decision‐making and patient safety by offering more precise and evidence‐based guidance on managing DDIs.

FUNDING

The laboratory of the author is supported by grants from the Sigrid Jusélius Foundation (Grant number 8037; Helsinki, Finland), and by State funding for university‐level health research (TYH2025308 and TYH2024301; Hospital District of Helsinki and Uusimaa, Finland).

CONFLICTS OF INTEREST

The author declares no competing interests for this work.