Introduction

The genetically polymorphic human cytochrome P450 enzyme CYP2C19 is involved in the metabolism of several frequently prescribed drugs, such as omeprazole [1] and diazepam [2]. CYP2C19*1 is considered to represent the wild-type allele, with 26 variant alleles having been described to date [3]. There are pronounced differences in the frequency of polymorphisms between ethnic groups. The CYP2C19 alleles *2 [4] and *3 [5], which are almost exclusively responsible for truncated CYP2C19 proteins with absent functional activity, are present in 30 and 5% of the Asian, 17 and 0.4% of the African-American and 15 and 0.04% of the Caucasian population, respectively. The remaining defective alleles are rarely encountered throughout all ethnicities [6].

Fast and reliable genotyping assays have been developed, enabling a clear assignment of individuals to the poor (PM) or extensive metaboliser (EM) phenotype [7]. Individual enzyme activity, however, is also influenced by various non-genetic factors, such as the intake of drugs that inhibit or induce CYP2C19, age or different disease states [8, 9]. In contrast to a genetic approach, phenotyping determines actual enzyme activity in vivo; as such, it can take the non-genetic factors into account and quantify their effects. There is therefore an essential demand for reliable, selective and reproducible CYP2C19 phenotyping metrics.

Mephenytoin is the standard probe drug for the assessment of CYP2C19 activity in humans [10] and has been successfully employed in several clinical studies in which the cocktail phenotyping method has been used to evaluate the effects of drugs or disease states on CYP2C19 activity in vivo [11, 12]. Omeprazole and proguanil are used as alternative probe drugs for CYP2C19 phenotyping. However, in addition to the CYP2C19-mediated hydroxylation of these probe drugs, there is also CYP3A4-mediated sulfoxidation of both omeprazole and 5-hydroxyomeprazole, the metabolite formed by CYP2C19 [1]. Interactions with CYP3A4 (phenotyping) substrates and/or CYP3A4 inhibitors could be the consequence when these are administered concomitantly, which may be one reason why CYP3A4 substrates were not administered simultaneously with omeprazole in published phenotyping cocktails [13, 14]. Additionally, due to the short half-life of omeprazole, plasma sampling has to take place quite rapidly post-administration, and results may therefore be influenced by the dissolution or enteral absorption of the probe drug. As a result, the within-subject variability of omeprazole metrics is substantial (approx. 35–40%) [15] and cannot be attributed to variation in CYP2C19 activity only. All of these data suggest that omeprazole is not an optimal alternative CYP2C19 probe. Proguanil metabolism seems to be affected significantly by CYP3A4 and is no longer recommended for CYP2C19 phenotyping [10].

Mephenytoin is administered as a racemic mixture and undergoes a complex metabolism (see Fig. 1). Only EM, who express functional CYP2C19, are able to hydroxylate (S)-mephenytoin to (S)-4′-hydroxymephenytoin, which is subsequently conjugated almost completely to the (S)-4′-hydroxymephenytoin glucuronide and excreted renally [16, 17]. In CYP2C19-deficient PM, (S)-mephenytoin is biotransformed via slow N-demethylation, which in vitro is preferentially mediated by CYP2B6 (with a minor contribution of CYP2C9), resulting in the formation of (S)-nirvanol [18–20]. (R)-mephenytoin is preferably demethylated to (R)-nirvanol and hydroxylated only to a small extent [16].

Fig. 1
figure 1

Supposed stereoselective metabolism of mephenytoin (MEP) to nirvanol (NIR) and 4′-hydroxymephenytoin (4OHM). Phase I enzymes mediating the metabolism of (S)-mephenytoin are indicated; those for (R)-mephenytoin are unknown in vivo [16–20]

Full size image

Based on this highly stereoselective metabolism, either the enantiomeric S/R ratio of unchanged mephenytoin in urine collected 0–8 h after administration [21] or the amount of 4′-hydroxymephenytoin recovered in the urine [22] are used as CYP2C19 phenotyping metrics. The shortcomings of the S/R ratio have been reported elsewhere [23, 24]. Therefore, we chose to assess the molar urinary recovery of 4′-hydroxymephenytoin in our study.

The urinary recovery of 4′-hydroxymephenytoin was widely used to phenotype for CYP2C19 activity during the last 20 years, but the validation of this metric is incomplete in the literature. The optimal duration of urine collection remains to be defined, and it is still unknown whether cumulative or fractioned urine collections yield the better results. Finally, it is unclear whether enzymatic hydrolysis of 4′-hydroxymephenytoin glucuronide prior to analysis is inevitable, or whether the small fraction of free 4′-hydroxymephenytoin found in untreated urine satisfactorily reflects individual CYP2C19 enzyme activity when a sensitive liquid chromatography tandem mass spectrometry (LC-MS/MS) assay is used.

An increasing number of compounds, such as bupropion, cyclophosphamide or efavirenz, have been identified as substrates of the highly polymorphic enzyme CYP2B6 [25, 26]. As (S)-mephenytoin demethylation to (S)-nirvanol is mediated mainly by CYP2B6 in vitro, it is tempting to use mephenytoin as a dual phenotyping drug for CYP2C19 and CYP2B6. However, it is unknown whether the urinary excretion of nirvanol represents an appropriate metric for CYP2B6 in vivo [20].

The main objective of the study reported here was to identify the optimal CYP2C19 phenotyping metric of urinary 4′-hydroxymephenytoin excretion. To decrease the risk of sedation, particularly in PM, [27], we also assessed the feasibility of phenotyping with 50 mg of mephenytoin in a Caucasian population. Finally, we examined whether the phenotyping of CYP2B6 can be accomplished with mephenytoin in vivo.

Methods

Volunteers and general procedures

A total number of 52 male, healthy Caucasians participated in four studies (A, B, C, D) in which phenotyping cocktails containing mephenytoin as a CYP2C19 probe were used. Two of the volunteers took part in two studies; to avoid confounding, only the data from the first study in which these two individuals participated were used for evaluation. The demographics of study participants is shown in Table 1.

Table 1 Demographic characteristics and urinary collection periods in all studies
Full size table

All studies were approved by the Ethics Committee of the Medical Faculty of the University of Cologne, Germany, and were conducted in accordance with the laws of Germany. Written informed consent was obtained from each subject before participation in the studies. All subjects were in good health as indicated by pre- and post-study screening examinations.

To rule out possible interactions with the cytochrome P450 system, all study participants were advised to avoid the consumption of food or beverages containing methylxanthines, grapefruit or alcohol from at least 48 h prior to the drug administration until 48 h thereafter. In the case of studies C and D, the dietary restrictions were extended to beverages containing quinine and maintained until the end of the post-study examinations. Therapeutic or illicit drug intake 2 weeks preceding the trials was an exclusion criterion. All subjects were regularly asked for the occurrence of adverse events, their wellbeing was continuously surveyed and blood pressure and pulse rate were recorded regularly.

In all four studies the subjects had to empty their bladder completely before administration of the phenotyping cocktail. Half a tablet of Epilan Gerot (Gerot Pharmazeutika, Vienna, Austria), corresponding to 50 mg racemic mephenytoin, was administered to each study participant as part of a phenotyping cocktail for important drug-metabolising enzymes and transporters. The individual dose was determined by weighing the tablet halves. Food and fluid intake as well as body position (supine position until 6 h post-dose) during the studies was standardised.

Individual study designs and objectives

Study A was a one-period, single-dose cocktail phenotyping pilot study; studies B, C and D were cocktail drug interaction studies in which the effect of an investigational drug in a test period was compared to a reference period. These three studies had a randomised, placebo-controlled cross-over design. The phenotyping drugs used in each of the four studies are shown in Table 2.

Table 2 Other phenotyping drugs co-administered with mephenytoin in studies A to D
Full size table

Studies A and B were primarily used to assess β-glucuronidase pre-treatment of the samples and fractioned versus cumulative urine collection. The procedures of these studies have been described in more detail in earlier studies [11, 28].

The primary phenotyping objectives of study C with regard to mephenytoin were the generation of confirmatory data for variability and selectivity evaluations of the CYP2C19 metrics as well as a more thorough characterisation of the slow nirvanol excretion. In this study, urine was collected in-house until 48 h post-dose and completed by two additional overnight in-house urine collection periods from 86 to 94 and 158 to 166 h post-dose. There was a washout phase of 4 weeks between both study periods.

The objective of study D for this evaluation was further confirmation of intra-individual variability data with respect to CYP2C19 phenotyping. In this study, a washout phase of 2 weeks was used between study periods.

Urine and blood sampling

In all studies urine was collected in 2-L plastic containers. The containers were cooled at +4°C during the collection. The exact duration of urine collection (for scheduled periods in the individual studies, see Table 1) and the pH value were recorded for each sample. The volume of urine was determined by weighing the containers assuming a mean density of 1 g/mL for urine. Aliquots of 10 mL were stored in plastic tubes at –80°C until analysis.

Whole blood samples for genotyping were drawn into tubes containing ethylene diamine tetraacetic acid (EDTA tubes; Sarstedt, Nümbrecht, Germany) before the first drug administration in all studies.

Urine analysis by LC-MS/MS

Mephenytoin, nirvanol and 4′-hydroxymephenytoin were analysed using a validated assay for LC-MS/MS as described earlier [29]. Both the fraction of free, unconjugated 4′-hydroxymephenytoin and total 4′-hydroxymephenytoin after β−glucuronidase pre-treatment [29] were measured in samples of study A and B. For study C and D, only total 4′-hydroxymephenytoin concentrations up to 12 h post-dose were determined, while nirvanol concentrations were determined for all collection periods of study A, B and C.

Genotyping

Genotyping for CYP2C19 was performed for all subjects participating in studies A, B and C. DNA was extracted from blood samples using a standard phenol–chloroform extraction method. Analysis for allele CYP2C19*2 was performed by a PCR/restriction fragment length polymorphism (RFLP) method as described by de Morais et al. [4] The volunteers of studies A, B and C were also genotyped for CYP2B6. Single nucleotide polymorphisms (SNPs) accounting for the CYP2B6 alleles *2, *3, *4, *6 and *7 were determined as described by Lang et al. [30].

Data analysis

Calculation of urinary recovery

For each cumulative and each fractioned collection period, the urinary recovery of metabolites was calculated as the molar excretion of 4′-hydroxymephenytoin or nirvanol in urine and expressed as a percentage of the accurate dose of mephenytoin administered. Recovery for cumulative collection periods was calculated by summing up the results of the fractioned collection periods. Urinary excretion rates were calculated as molar excretion divided by sampling time.

To estimate the amount of nirvanol excreted during periods without urine collection, the following formula was used in study C:

$$A_{e\left( {1 - 2} \right)} = \sqrt {R1 \cdot R2} \cdot \left( {t1 - t2} \right)$$

In this formula, A e reflects the amount excreted, R is the mean excretion rate in the collection interval before and after the period in which no urine was collected, respectively, and t is the corresponding time where the adjacent collection intervals ended or started.

CYP2C19 genotype and 4′-hydroxymephenytoin metrics

To identify the CYP2C19 metric which reflects the CYP2C19 genotype best, we calculated the urinary recovery of 4′-hydroxymephenytoin (%) for fractioned and cumulative collection intervals of the respective reference periods, each with and without a previous deconjugation of glucuronic acid, in studies A, B and C. The Kruskal-Wallis one-way analysis of variance (ANOVA) with CYP2C19 genotype as the independent parameter and the various metrics as dependent parameters was used for statistical analysis.

CYP2B6 genotype and nirvanol metrics

For a comparison of nirvanol excretion between CYP2B6 genotypes, we calculated fractioned and cumulative urinary recovery of nirvanol (%) for study A, for the respective first study periods of study B (assuming that propiverine, the drug characterised in this interaction study, had no effect on nirvanol excretion) and for the respective reference periods of study C. For nirvanol metrics in study B, it was not possible to use the reference period only because after 2 weeks of wash-out, there was still a pronounced excretion of nirvanol prior to administration of the mephenytoin test dose in the second period. Our comparison of nirvanol excretion in relation to the CYP2B6 genotype is descriptive only.

Intra-subject variability of CYP2C19 and CYP2B6 metrics

To determine the intra-subject variability of 4′-hydroxymephentoin and nirvanol excretion parameters, respectively, we calculated the intra-individual coefficients of variation after the logarithmic transformation of urinary recovery data from the residual variance obtained by the standard ANOVA approach used for the bioavailability assessment (software BIAV; Stephan Rietbrock, Lemgo, Germany, 1994). For 4′-hydroxymephentoin metrics, data from both the test and reference periods of studies B, C, and D were used, and all calculations were performed for fractioned and cumulative collection periods, each with and without previous deconjugation of glucuronic acid. In the case of nirvanol metrics, intra-individual coefficients of variation for fractioned and cumulative excretion could only be calculated for study C because the wash-out period was too short in studies B and D. Based on the intra-individual coefficients of variation, the sample size needed for a cross-over drug interaction study with the CYP2C19 and CYP2B6 metrics identified was assessed assuming a power of 80%, a range of “no drug interaction“ from 0.8 to 1.25 and a significance level of 5% by means of confidence intervals, as described by Diletti et al. [31].

All other analyses were performed with SPSS ver. 12.00 (SPSS, Chicago, IL) and Excel 2000 (Microsoft, Redmond, WA). For all statistical procedures, P < 0.05 was considered to be significant. The results are given as arithmetic means ± standard deviations (SD) unless stated otherwise.

Results

In general, all phenotyping cocktails were well tolerated. No severe adverse effects related to mephenytoin were observed throughout all studies. Since analytical selectivity had been validated previously, no interfering peaks were observed in any of the chromatograms.

4′-Hydroxymephenytoin metrics for CYP2C19 activity

Analysis of the CYP2C19 genotype in studies A, B and C resulted in the identification of 24 homozygous CYP2C19*1/*1 carriers, 12 heterozygous CYP2C19*1/*2 EM and one poor CYP2C19*2/*2 metaboliser. (Volunteer B11, who was a carrier of the CYP2C19*1/*2 genotype, also took part in study C. Therefore, the results obtained for this subject in study C were excluded from further evaluation.) Respective numbers for individual studies A, B and C were 5, 5, 0; 12, 3, 1; 7, 4, 0. The effect of the CYP2C19 genotype on urinary excretion of 4′-hydroxymephenytoin was statistically significant for cumulative urine collections from the administration of mephenytoin to 6 or more h post-dose when combined with β-glucuronidase pre-treatment of the samples (see Table 3). A significant genotype effect was also seen for fractioned collection of (deglucuronidated) urine from 2 to 8 h post-dose (data available for studies A and B; n = 26) with a P value of 0.036. Apart from these results, neither the cumulative collection of urine without the previous deconjugation of glucuronic acid nor all other fractioned urine sampling periods provided significant data (data not shown). The one CYP2C19*2/*2 homozygote PM was, however, identified in all cumulative and fractioned collection intervals because the excretion of 4′-hydroxymephenytoin was roughly 100-fold lower than in this individual than in carriers of at least one *1 allele. This is also indicated by a comparison of urinary excretion rates, as depicted in Fig. 2. Without deglucuronidation, 4′-hydroxymephenytoin concentrations were below the lower limit of quantification in all samples of the PM, whereas all post-dose concentrations were above the lower limit of quantification in the EM group.

Table 3 Urinary recovery (mean±SD) of 4′-hydroxymephenytoin relative to the mephenytoin dose for individual genotype groups
Full size table
Fig. 2
figure 2

Mean urinary excretion rates of 4′-hydroxymephenytoin for different genotype groups after deglucuronidation in the reference periods of studies A, B and C. Note the different scaling on the second ordinate for the homozygous carrier of the CYP2C19*2 allele

Full size image

In the drug–drug interaction studies where we used the residual error to estimate intra-individual variability (studies B, C, D), there was no relevant effect of the potentially interacting drugs on CYP2C19 metrics (data not shown). As evident from Fig. 3, the lowest intra-individual variability of 4′-hydroxymephenytoin recovery was achieved by cumulative collection periods combined with β-glucuronidase pre-treatment. Because of the findings on deglucuronidation in studies A and B, only β-glucuronidase pre-treated samples were analysed for studies C and D. Overall intra-subject variability from these samples was 67, 57, 22, 11 and 7% for urine collection periods 0–2 h (study included: B), 0–4 h (B/D), 0–6 h (B/D), 0–8 h (B/C) and 0–12 h (B/C/D) following oral administration of 50 mg mephenytoin, respectively. As expected, for the longer sampling periods these values were clearly lower than inter-subject variation (CV 72, 45, 32, 25 and 23% for the respective sampling periods). A further extension of the collection periods up to 0–16 h (B) or 0–24 h (B) post-dose led to coefficients of variation of 6 and 11%, respectively, and thus did not provide clearly better results. On the basis of these values for intra-individual variability, sample sizes of 24, eight or six subjects were calculated for cross-over studies to assess an effect on CYP2C19 phenotype if collection periods of 0–6 h, 0–8 h or 0–12 or more hours post-dose, respectively, are used.

Fig. 3
figure 3

Study B (n = 16). Intraindividual variability of 4′-hydroxymephenytoin urinary recovery for fractioned and cumulative collection of urine, with and without previous β-glucuronidase pre-treatment

Full size image

Nirvanol metrics for CYP2B6 activity

Nirvanol excretion was very slow, with elimination rates being almost unchanged during the first week after mephenytoin administration. Even after the 4-week wash-out period in study C, concentrations above the lower limit of quantification (30 ng/mL) were found in seven of 12 samples. Mean ± SD urinary excretion of nirvanol within 166 h after cocktail administration in the reference periods of study C was 3.2 ± 1.4 mg and thus reached 6.4 ± 2.7% of the molar mephenytion dose.

The excretion of nirvanol as a function of the CYP2B6 genotype in studies A, B and C is shown in Fig. 4. In this three studies, the following CYP2B6 genotypes were determined: *1/*1, n = 10; *1/*2, n = 2; *4/*4, n = 1; *1/*5, n = 6 (subject B11 was excluded from further evaluation, thus *1/*5, n = 5); *5/*5, n = 2; *1/*6, n = 8; *6/*6, n = 3; *1/*7, n = 2. For these genotypes, mean urinary excretion up to 24 h post-dose was 0.72-, 1.94-, 1.07-, 0.63-, 0.82-, 1.64-, and 0.63-fold that of homozygote carriers of the wild-type allele. Four volunteers could not be assigned clearly to one of the genotypes examined.

Fig. 4
figure 4

Arithmetic means of cumulative urinary nirvanol recovery by carriers of different CYP2B6 genotypes in studies A, B and C (n = 37; subject B11 excluded from further evaluation, thus *1/*5, n = 5). For study B, only the results of the first period are shown irrespective of co-administration with the test drug in this interaction study. For study C, only the results of the reference period are shown (for detailed explanation, see Methods section). Legend: The genotypes arranged in descending order of mean urinary recovery

Full size image

For studies B and D, a wash-out phase of 2 weeks between both study periods resulted in a considerable carry-over of nirvanol into the second period; therefore, intra-individual variability of nirvanol excretion could only be calculated for study C. Figure 5 shows that intra-individual variability for cumulative urine collections declined with increasing collection time, reaching 27, 23, 22, 22 and 19% for collection periods up to 24 h, 36 h, 48 h, 94 h and 166 h and thereafter, respectively, from mephenytoin administration. On the basis of these intra-individual coefficients of variation, sample sizes of 28, 24, 24, 20 or 14 individuals were calculated to be required for cross-over studies to assess the effect of CYP2B6 phenotype, if collection periods of 0–24, 0–36, 0–48, 0–94 or up to 166 h, respectively, were to be used.

Fig. 5
figure 5

Intra-individual variability of nirvanol urinary recovery for fractioned and cumulative collection of urine in study C (n = 12). Data points at midpoint sampling time of each collection period; the duration of the collection periods is indicated in the lower part of the figure. The figure includes extrapolated collection intervals 48–86 h and 94–158 h post-dose

Full size image

Discussion

Our results indicate that the quantification of 4′-hydroxymephenytoin with prior deglucuronidation of samples in a cumulated urine collection 0–12 h after administration of 50 mg mephenytoin is the most suitable mephenytoin-based phenotyping metric for CYP2C19. Under these conditions, we obtained an intra-individual coefficient of variation as low as 7% and a significant relationship between CYP2C19 genotype and phenotyping metric. A small sample size of n = 6 would be sufficient to assess a potential inhibitory or inductive effect of a drug on CYP2C19 activity, if this phenotyping metric is used in a cross-over study. In contrast, mephenytoin was not useful as a phenotyping metric for assessing CYP2B6 activity in vivo.

4′-Hydroxymephenytoin metrics for CYP2C19 activity

The characteristics of a probe drug for metabolic phenotyping should include good tolerability and availability, lack of drug–drug interaction with other probes and high sensitivity and reproducibility [32]. The use of mephenytoin as a phenotyping probe for CYP2C19 activity in humans meets most of these criteria. Among the valued advantages of mephenytoin are a high reproducibility between two or more study periods (Table 4), a significant gene–dose relationship, [33, 34] and a validated lack of a drug–drug interaction between mephenytoin and numerous other phenotyping probes, including the other components of the cocktails used here [35–37].

Table 4 Intra-individual variability for different CYP2C19 phenotyping metrics
Full size table

However, the use of mephenytoin as a phenotyping probe for CYP2C19 activity is also associated with a number of disadvantages. Severe sedation following the administration of 100 mg mephenytoin has been reported in PM and in subjects with a low body mass index [27]. Additionally, the urinary recovery of 4′-hydroxymephenytoin is, as is any urine-based metric, susceptible to incomplete or deficient urine collection [21]. Finally, mephenytoin has lost its clinical relevance and is no longer commercially available in the United States and most European countries. Nevertheless, in terms of CYP2C19 phenotyping, if available, mephenytoin seems to be the current best choice.

In comparison to the S/R ratio, the application of the urinary recovery of 4′-hydroxymephenytoin appears to be favourable as the reliability of the S/R ratio is compromised by the presence of a pH- and storage-labile (S)-mephenytoin cysteine conjugate that can leads to an increased risk of false phenotype classification by 1.5% [23, 24, 38]. Such problems do not occur with 4′-OH-mephenytoin [29].

However, no systematic efforts aimed at optimising 4′-hydroxymephenytoin recovery in urine as a CYP2C19 metric have been reported since this metric was introduced more than 20 years ago [22]. Only a few studies with urinary collection intervals deviating from the “traditional” 8 h have been conducted [33, 39]. Therefore, we tested various urinary collection periods of different duration to assess an optimal collection interval.

To facilitate the analytical sample processing method we first examined whether the small fraction of free 4′-hydroxymephenytoin present in untreated urine would satisfactorily reflect individual CYP2C19 activity. However, despite the applied sensitive the LC-MS/MS assay, our results clearly support the necessity that the urine be subjected to deglucuronidation before it is analysed since the omission of β-glucuronidase pre-treatment resulted in both an overall loss of reproducibility and the phenotype–genotype relationship. Hence, the well-recognised intra- and inter-individual variability of UDP-glucuronosyltransferase activity seems to be an important confounder [40].

Urinary recovery of 4′-hydroxymephenytoin allowed the PM phenotype to be clearly identified in all fractioned and cumulative collection periods. The quantification of 4′-hydroxymephenytoin in cumulated urine collections of at least 6 h post-dose was statistically significantly affected by CYP2C19 genotype (Table 3). However, the overlapping of the 95% confidence intervals (see Table 3) for 4′-hydroxymephenytoin in homozygous and heterozygous EM indicate that no clear-cut differentiation between these genotypes can be made with this phenotyping metric. This finding is in accordance with published results for 4′-hydroxymephenytoin urinary recovery [34].

Our studies also demonstrate that optimal reproducibility is reached after 12 h of cumulative urine collection. A generally smaller susceptibility towards minor sampling errors in longer collection intervals as well as a minor interference of 4′-hydroxymephenytoin excretion with mephenytoin absorption and distribution are likely reasons for this result. Consequently, a lower sample size of n = 6 is needed when urine is collected for a cumulative period of 12 h in comparison to a sample size of n = 8 when the “traditional” 8-h sampling interval is chosen. There is remarkable concordance between literature data and our results (Table 4). Hence, an intra-individual coefficient of variation of below 10% appears to be consistent for 4′-hydroxymephenytoin urinary recovery for both the 100 mg and the 50 mg mephenytoin dose.

We compared our results to published data on other established CYP2C19 metrics. In analogy to our results, a clear-cut discrimination between the CYP2C19*1/*1 homozygote and *1/*2 heterozygote EM was not achieved with either the S/R-ratio of mephenytoin [33, 41] or the omeprazole metabolic ratio [42, 43]. A similar finding for several other metrics indicates that although homozygous EM may have a slightly higher CYP2C19 activity than heterozygous EM, this difference is probably not relevant for any drug metabolised by CYP2C19 [44]. As presented in Table 4, urinary recovery of 4′-hydroxymephenytoin yields the lowest intra-individual coefficients of variation throughout all phenotyping metrics. This was not achieved by a generally low variability of this metric (as compared to metabolic ratios); indeed, inter-subject variability remained threefold higher for the 0- to 12-h sampling period. The principal reason for this superior performance seems to be caused by the analytical assay. Mass spectrometric quantification of 4′-hydroxymephenytoin following deglucuronidation takes place at concentrations that are more than 200-fold above the lower limit of quantification. In contrast, many of the samples used for determining the omeprazole metabolic ratio and the S/R ratio of mephenytoin, respectively, exhibit concentrations close to or below the lower limit of quantification of the high-performance liquid chromatography-based analytical method used in the respective studies.

In conclusion, we propose to carry out CYP2C19 phenotyping with a low dose of 50 mg mephenytoin and cumulative urinary collection of 12 h, instead of 8 h, combined with the prior deglucuronidation and LC-MS/MS detection of 4′-hydroxymephenytoin.

Nirvanol metrics for CYP2B6 activity

Although in vitro studies have proven that (S)-nirvanol is formed from (S-mephenytoin), [18, 19, 30], the respective parameters have not yet been established as in vivo phenotyping metrics. The only study currently available is one is which tentatively assessed the autoinduction by artemisinin using the (S)-nirvanol to (S)-mephenytoin ratio [45]. In the present study, we carried out only a limited evaluation of the suitability of nirvanol metrics to estimate CYP2B6 activity. Because of the diversity of the CYP2B6 gene, much higher sample sizes would be required to assess effects of CYP2B6 genotypes on nirvanol excretion. With this limitation, the high urinary excretion of nirvanol of the single homozygous carrier of the CYP2B6*4 allele is indeed in accordance with previous findings for CYP2B6-mediated bupropion hydroxylation [46], and the low excretion rates for carriers of the *5 or *7 allele correspond to in vitro findings [30]. However, the relative high nirvanol excretion of homozygote carriers of the *6 allele does not reflect previous findings in vivo. For those subjects, a lower CYP2B6-mediated clearance of efavirenz and methadone was reported [47, 48].

Nevertheless, our investigations clearly show that mephenytoin is not optimal to phenotype for CYP2B6 activity, at least for cross-over studies. First, adequate intraindividual variability with coefficients of variation of 23% resulting in a sample size of 24 subjects were achieved no earlier than 36 h after cocktail administration. While cumulative urinary collection of 36 h is still feasible, a wash-out phase of more than 6 weeks between two consecutive study periods seems to be necessary to rule out the carry-over of nirvanol. Phenotyping under such premises is not practicable. Secondly, in addition to CYP2B6 as a low-affinity/high-capacity component of (S)-mephenytoin demethylation, the influence of CYP2C9 as a high-affinity/low-capacity component cannot be dismissed, especially at low plasma concentrations that occur after a single phenotyping dosage [20]. For these reasons, it seems more reasonable to validate other metrics, such as buproprion hydroxylation, as a phenotyping tool for CYP2B6 in vivo, as has been proposed elsewhere [25].