Meta-Analysis of Noncompartmental Pharmacokinetic Parameters of Ertugliflozin to Evaluate Dose Proportionality and UGT1A9 Polymorphism Effect on Exposure.

Ertugliflozin, a sodium-glucose cotransporter 2 inhibitor, is primarily metabolized via glucuronidation by the uridine 5'-diphospho-glucuronosyltransferase (UGT) isoform UGT1A9. This noncompartmental meta-analysis of ertugliflozin pharmacokinetics evaluated the relationship between ertugliflozin exposure and dose, and the effect of UGT1A9 genotype on ertugliflozin exposure. Pharmacokinetic data from 25 phase 1 studies were pooled. Structural models for dose proportionality described the relationship between ertugliflozin area under the plasma concentration-time curve (AUC) or maximum observed plasma concentration (Cmax ) and dose. A structural model for the UGT1A9 genotype described the relationship between ertugliflozin AUC and dose, with genotype information on 3 UGT1A9 polymorphisms (UGT1A9-2152, UGT1A9*3, UGT1A9*1b) evaluated as covariates from the full model. Ertugliflozin AUC and Cmax increased in a dose-proportional manner over the dose range of 0.5-300 mg, and population-predicted AUC and Cmax values for the 5- and 15-mg ertugliflozin tablets administered in the fasted state demonstrated good agreement with the observed data. The largest change in ertugliflozin AUC was in subjects carrying the UGT1A9*3 heterozygous variant, with population-predicted AUC (90% confidence interval) values of 485 ng·h/mL (458 to 510 ng·h/mL) and 1560 ng·h/mL (1480 to 1630 ng·h/mL) for ertugliflozin 5 and 15 mg, respectively, compared with 436 ng·h/mL (418 to 455 ng·h/mL) and 1410 ng·h/mL (1350 to 1480 ng·h/mL), respectively, in wild-type subjects. Overall, the mean effects of the selected UGT1A9 variants on ertugliflozin AUC were within ±10% of the wild type. UGT1A9 genotype did not have any clinically meaningful effects on ertugliflozin exposure in healthy subjects. No ertugliflozin dose adjustment would be required in patients with the UGT1A9 variants assessed in this study.

Ertugliflozin is a selective inhibitor of sodium‐glucose cotransporter 2 approved in the United States, Europe, and other countries as an adjunct to diet and exercise to improve glycemic control in adults with type 2 diabetes mellitus (T2DM). It is recommended for use as monotherapy or in combination with other diabetes therapies at single daily doses of 5 and 15 mg.,

The pharmacokinetics (PK) of ertugliflozin have been well characterized across a number of different study populations. Oral absorption of ertugliflozin is rapid, with a median time to maximum plasma concentrations (Tmax) of 1.0 hour (single doses; fasted) to 2.0 hours (multiple doses; fed) postdose. Ertugliflozin has a mean terminal‐phase half‐life (t1/2) of ∼10 to 17 hours. No clinically meaningful effect on ertugliflozin exposure is observed following administration with food or following concomitant administration with commonly coprescribed drugs such as metformin, sitagliptin, glimepiride, or simvastatin, and no adjustment of ertugliflozin dose is required in patients with renal impairment or mild to moderate hepatic impairment based on ertugliflozin PK. A mass‐balance study in humans revealed that ∼35% of the administered dose is recovered in urine and feces as unchanged ertugliflozin. O‐glucuronidation is the major biotransformation pathway for ertugliflozin, with 2 pharmacologically inactive glucuronide metabolites—ertugliflozin‐2‐O‐β‐glucuronide and ertugliflozin‐3‐O‐β‐glucuronide—representing the primary metabolites of ertugliflozin in plasma., In vitro metabolism studies have shown that formation of these metabolites is catalyzed primarily by the uridine 5′‐diphospho‐glucuronosyltransferase (UGT) enzyme isoform UGT1A9 (81%); the UGT2B7 and UGT2B4 isoforms play a minor role in ertugliflozin glucuronidation (19%)., ,

Sequence variation in UGT genes has been shown to affect enzyme expression and activity levels, with the potential to profoundly affect target drug exposures. As such, current regulatory guidance recommends that for drugs in which the primary biotransformation pathway is governed by an enzyme that is genetically polymorphic, the effect of this variation on the PK of the active substance is assessed., Three UGT1A9 variants were chosen for this analysis based on their allelic frequency across different racial groups and the potential for a clinical effect on drug disposition. The UGT1A9‐2152(C>T) single‐nucleotide polymorphism (SNP), occurring in the UGT1A9 promoter with a minor allele frequency (MAF) in whites of 0.06, results in increased UGT1A9 expression and higher rates of substrate glucuronidation., The UGT1A9*3 98(T>C) nonsynonymous SNP has a minor allele frequency in whites of 0.02‐0.04 and encodes an M33T substitution that leads to reduced enzyme activity for certain substrates., The UGT1A9*1b‐118(dT)9>10 allele, which results from a 1‐bp insertion in the UGT1A9 promoter and leads to increased UGT1A9 expression and higher substrate clearance, has a minor allele frequency in whites of ∼0.4.,

As the allelic frequency of UGT1A9 variants is generally low, a pooled analysis of ertugliflozin exposure values from phase 1 trials was prospectively planned to assess the impact of UGT1A9 genotype on ertugliflozin PK. In addition, although the dose proportionality of ertugliflozin exposure when administered as a solution or suspension was evaluated in the first‐in‐human studies,, , , dose proportionality of ertugliflozin tablets was not formally assessed in a dedicated clinical study during phase 1 development. Instead, a model‐based approach was taken to evaluate dose proportionality of ertugliflozin using data from the phase 1 development program (including solution, suspension, and tablet formulations, and fasted or fed conditions) to support the conclusion of dose‐proportional increases in ertugliflozin exposure in the product label. Hence, the objectives of this analysis of pooled data from 25 ertugliflozin phase 1 studies were (1) to assess the dose proportionality of ertugliflozin area under the plasma concentration‐time curve (AUC) and maximum observed plasma concentration (Cmax) across various ertugliflozin doses, regimens, and formulations, and (2) to evaluate the impact of UGT1A9 genotype on ertugliflozin AUC in sufficient numbers of participants to allow a meaningful analysis across UGT1A9 variants that have relatively low minor allele frequencys in global populations.

Methods

Ethical Conduct of the Studies

The final protocols and informed consent documentation for the studies included in this analysis were reviewed and approved by an institutional review board. A list of study sites and locations is provided in Table S1. All subjects provided signed and dated informed consent. All studies were compliant with the ethical principles of the Declaration of Helsinki and all International Conference on Harmonisation Good Clinical Practice guidelines.

Genotyping

K2‐ethylenediaminetetraacetic acid blood samples were taken prior to dosing for each subject and frozen until time of analysis. DNA was extracted from whole‐blood samples using a QIAsymphony DNA Mini Kit (Qiagen, Germantown, Maryland), quantified by NanoDrop spectroscopy (Thermo Fisher Scientific, Waltham, Massachusetts), and normalized to a concentration of 8 ng/μL.

DNA samples from the study participants were genotyped for 3 UGT1A9 polymorphisms (2 SNPs and 1 insertion‐deletion mutation), chosen for analysis based on their population frequency and their potential for a clinical effect on drug disposition., , , , None of these polymorphisms represent complete loss of function of the UGT1A9 isozyme, and, as such, they do not constitute a traditional poor metabolizer status, as found with the cytochrome P450 (CYP) 2D6 isozyme. The 2 SNPs were UGT1A9‐2152(C>T) (rs17868320) and UGT1A9*3 98(T>C) (rs72551330), and the insertion‐deletion mutation was UGT1A9*1b‐118(dT)9>10 (rs3832043; previously referred to as UGT1A9*22). The specified UGT1A9 polymorphisms were assessed by polymerase chain reaction (PCR). Commercially available TaqMan assays were used to genotype the 2 UGT1A9 SNPs, rs17868320 (C__34418857_10) and rs72551330 (C__64627083_10), which were analyzed using a QuantStudio 12K Flex Real‐Time PCR system (Applied Biosystems, Thermo Fisher Scientific, Waltham, Massachusetts). The insertion‐deletion mutation, rs3832043, was detected by amplifying a region around the base insertion (forward primer, ACTTAACATTGCAGCACAGG; reverse primer, 6FAM‐CAGAGAACTGCAGCTGAGAGCA) and sizing the fragment using an Applied Biosystems 3730xl DNA Analyzer (Thermo Fisher Scientific, Waltham, Massachusetts).

Data Sources

Study Selection and Data Sets

PK data from 25 phase 1 studies of ertugliflozin were included in this pooled analysis (Table S1). Selected studies were phase 1 studies with dense PK sampling and appropriate informed consent for DNA sampling to allow a pharmacogenomic analysis. Studies consisted of single‐ and multiple‐dose regimens of ertugliflozin as well as crossover study designs that evaluated the effects of dose regimen, formulation, food, and concomitant medications on the PK of ertugliflozin across a range of study populations.

The pooled data set included subject identification, dosing information, and the reported PK parameters. Two analytical data sets (1 for AUC values and 1 for Cmax values) were created from the pooled data set. Both the AUC and Cmax data sets were used in the dose‐proportionality analysis; only the AUC data set was used in the UGT1A9 genotype analysis. Of the 25 studies in the data set, data from 17 studies contributed to the dose‐proportionality analysis and data from 20 studies contributed to the UGT1A9 genotype analysis (Table S1).

Data Inclusion Criteria for the Dose‐Proportionality and UGT1A9 Genotype Analyses

All AUC and Cmax data reported using noncompartmental analysis methods were included in the dose‐proportionality and UGT1A9 genotype analyses. For bioequivalence, drug‐drug interaction (DDI) and twice‐daily versus once‐daily regimen studies, only PK parameters from the reference treatment arms were included (see Table S1). For single‐dose studies, AUC from time zero extrapolated to infinite time (AUCinf) and Cmax were included. For multiple‐dose studies, AUC from time zero to time tau, the dosing interval, where tau is 24 hours (AUCtau) at steady state and Cmax following the first dose (day 1 data, if available) were included. For the food‐effect study on the commercial ertugliflozin tablet, PK parameters from the fasted arm were included. For the renal or hepatic impairment studies, only PK parameters from the healthy subjects with normal renal or hepatic function were included.

Data Exclusion Criteria for the Dose‐Proportionality and UGT1A9 Genotype Analyses

Data excluded from statistical analyses because of vomiting or protocol deviations in the original studies were also excluded from the current analyses.

Specific Data Inclusion/Exclusion Criteria for the Dose‐Proportionality Analysis

For the dose‐proportionality analysis only, PK parameters from studies with fasted oral administration of ertugliflozin or administration of ertugliflozin with a light meal in healthy subjects in the absence of other drugs were included. Bioequivalence studies evaluating fixed‐dose combinations of ertugliflozin and metformin or ertugliflozin and sitagliptin versus coadministration of individual components were excluded.

Specific Data Inclusion/Exclusion Criteria for the UGT1A9 Genotype Analysis

For the UGT1A9 genotype analysis only, AUC values from studies with fasted oral administration of ertugliflozin tablets in healthy subjects in which UGT1A9 genotype data were collected were included; studies with no UGT1A9 genotype data available were excluded. PK parameters from the reference arms of bioequivalence studies evaluating fixed‐dose combinations of ertugliflozin and metformin or ertugliflozin and sitagliptin were included because clinical drug‐drug interaction studies have shown that there are no meaningful differences in ertugliflozin PK between coadministration of ertugliflozin with metformin or sitagliptin versus ertugliflozin alone.

Model Development for the Dose‐Proportionality Analysis

Structural Model

The data sets were used to develop regression models that described the relationships between dose and exposure parameters (AUC or Cmax). The models were constructed using AUC or Cmax as the dependent variable and dose as the independent variable. The model structures are shown below: where DOSE is the ertugliflozin dose in milligrams, SLP (slope) is the increase of AUC or Cmax per milligram increase in dose, and INT (intercept) is the AUC or Cmax at a dose of zero. The requirement for an intercept was assessed using a likelihood ratio test (based on differences in the NONMEM objective function values, ΔOFV) between the model with and without an intercept to determine the significance. The test was performed at a significance of α = 0.05, and the intercept was included in the structural model if the P value was less than α (ΔOFV > 3.841). An intercept was included in the AUC model for fitting purposes (ΔOFV > 3.841) but was not included in the Cmax model (ΔOFV < 3.841). Power models were tested during structural model development. However, the power model did not provide appropriate fitting to the AUC data with increase of OFV (ΔOFV, 41). The Cmax power model seemed to have a good fit to the data, but the estimate of 95% confidence interval (CI) of the power parameter included 1 after covariates were added to the model, which further supported that the linear model is more appropriate for the data.

Random‐Effects Model

The equations shown below provided the structure for random effects on slope and base model parameterization: where i indexes subjects and j indexes dose for each ith subject, SLPі represents the individual model‐predicted slope, θSlope is the mean value of the slope, and η1,i denotes the interindividual error around the slope, accounting for the ith subject's deviation from the mean value having zero mean and variance ω2.

The residual variability was modeled using an additive residual error model shown below: where AUC(or Cmax)ij is the observed AUC or Cmax of the ith subject at the jth dose, AUC(or Cmax)(tij) represents the model‐predicted AUC or Cmax, and εij denotes the normally distributed intraindividual (residual) error assumed to have a mean of zero and variance σ2.

Full Model

The full models for AUC and Cmax were generated by the addition of the covariates of interest to the base model multiplicatively as a factor. The covariates of interest for the dose‐proportionality analysis were formulation (tablet as reference) and food status (light meal versus fasted).

The method of addition of covariates onto the slope parameter is shown below: where TVSLP is the typical individual (mean) value of slope with covariates, and θSlope is the population central tendency of slope with the tablet formulation and the fasted state. The theta for each covariate describes the fold change in θSlope when the formulation is a suspension or solution (θSusp,slope and θSolu,slope, respectively) or when the drug is administered with a light meal (θFood,slope). FLG1, FLG2, and Food are all indicator variables. FLG1 and FLG2 are the flags for formulation, and values were derived based on the FORM (formulation) variable in the data set, which was coded as 2, 3, and 4 for tablet, suspension, and solution, respectively. When FORM = 2 (tablet), then FLG1 = 0 and FLG2 = 0; when FORM = 3 (suspension), then FLG1 = 1 and FLG2 = 0; when FORM = 4 (solution), then FLG1 = 0 and FLG2 = 1. Food was coded as 0 and 1 for fasted and fed (light meal) states, respectively.

The full model approach was implemented for model building; no stepwise inclusion or exclusion of covariates for model building was performed. Rather, all covariates were retained in the full model regardless of significance. The purpose of the covariate analysis was not to identify the covariate effect size, but to account for confounding covariate effects and better understand dose proportionality. If a linear model was successfully fitted to the AUC or Cmax versus dose data, then dose proportionality could be concluded for AUC or Cmax.

Model Development for the UGT1A9 Genotype Analysis

The same methods from the dose‐proportionality analysis were used to develop the structural model for the UGT1A9 genotype analysis. Briefly, the relationship between AUC and dose was described with a structural model, and then the covariates of interest were added to the base model to develop the full model.

The covariates of interest for the UGT1A9 genotype analysis consisted of 3 allelic variants of UGT1A9: (1) UGT1A9‐2152(C>T) — rs17868320 — abbreviated as RS20 in the model, with the wild type of rs17868320 coded as “C/C” and the heterozygous variant coded as “C/T”; (2) UGT1A9*3 98(T>C) — rs72551330 — abbreviated as RS30 in the model, with the wild type of rs72551330 coded as “T/T” and the heterozygous variant coded as “T/C”; and (3) UGT1A9*1b−118(dT)9>10 — rs3832043 — abbreviated as RS43 in the model, with the wild type of rs3832043 coded as “9/9,” the heterozygous variant coded as “9/10,” and the homozygous variant coded as “10/10.”

The covariate model development focused on UGT1A9 genotype only because all data were from fasted subjects who received the tablet formulation. The 3 allelic variants were included in the model as categorical variables. As the slope is the determinant of the change in exposure as a function of dose, covariates were included on the slope parameter. The method of addition of covariates onto the slope parameter is shown below: where TVSLP is the typical individual (mean) value of slope with covariates and θSlope is the population central tendency of slope in UGT1A9 wild‐type subjects. The theta of each covariate describes the fold change in θSlope for each type of UGT1A9 variant carriers: θRS43Hom,slope and θRS43Het,slope for the UGT1A9*1b homozygous and heterozygous variants, respectively; θRS20,slope for the UGT1A9‐2152 heterozygous variant; and θRS30,slope for the UGT1A9*3 heterozygous variant. There were no homozygous UGT1A9‐2152 or UGT1A9*3 variants in the data set. In the final data set, [RS]20 (UGT1A9‐2152) and [RS]30 (UGT1A9*3) were coded as 0 and 1 for wild‐type and heterozygous variants, respectively; and [RS]43 (UGT1A9*1b) was coded as 0, 1, and 2 for wild‐type, homozygous, and heterozygous variants, respectively.

All 3 variants (4 covariates) were added to the model simultaneously, so the effect of each variant against the “true” wild type (the wild type for all 3 alleles) could be defined. The full model was used for all inferences regarding variance in drug exposure attributable to the covariates. The effects of the UGT1A9 genotype on the ertugliflozin AUC were estimated by the covariate effect as a fold change relative to the AUC of wild‐type subjects.

Assessment of Model Adequacy (Goodness of Fit)

Goodness of fit was evaluated using change in the log likelihood, visual inspection of diagnostic plots, precision of the parameter estimates, and impact on between‐subject variability or residual variability. At both the base‐ and full‐model stages, diagnostic plots were examined to assess model adequacy, possible lack of fit, or violation of assumptions. Plots of AUC and Cmax versus dose and population‐predicted and individual‐predicted versus dose were evaluated for base‐ and full‐model appropriateness. Furthermore, observed versus predicted and observed versus individual‐predicted values were evaluated for concordance with the line of unity. Similarly, plots of residuals versus predicted values were evaluated for randomness around the zero line.

Assessment of Model Predictive Performance (Evaluation)

The full model was used in bootstrap analysis (stratified by protocol) with 1000 replications to evaluate parameter uncertainty. The resultant parameter estimates for slope and intercept (for AUC model only) from the bootstrap analysis were used to derive AUC and Cmax with 95%CI at ertugliflozin doses of 5 and 15 mg. The 5th and 95th percentiles of the predicted exposure data were then quantified to determine the predicted mean AUC and Cmax following administration of a 5‐ or 15‐mg oral dose. The results were displayed graphically and overlaid with the observed data to allow for a visual predictive check of the model fit for the dose‐proportionality analysis. The predicted mean AUC at 5‐ and 15‐mg doses of ertugliflozin for different UGT1A9 genotypes was then evaluated. For simulated parameters, 90%CIs were reported.

Modeling Software

Modeling was performed in the Pfizer repository ePharmacology (version 4.4.1) using the software program NONMEM (version 7.3; ICON plc, Gaithersburg, Maryland). All analyses were conducted using the first‐order conditional estimation method with interaction. Preanalysis and postprocessing diagnostic plots and summary statistics were generated using R software (version 3.0.2 or higher; https://www.r‐statistics.com/).

Results

Dose‐Proportionality Analysis

Observed Data

For the dose‐proportionality analysis, 344 records from 309 subjects and 307 records from 260 subjects were used in the analysis for AUC and Cmax, respectively. The number of records by dose for the fasted‐versus‐fed and formulation covariates is shown in Table 1. The doses contained within the dose‐proportionality data set ranged from 0.5 to 300 mg, with the largest number of records obtained following administration of a 15‐mg dose, which is the highest approved dose for ertugliflozin. The majority of subjects received the tablet formulation administered under fasted conditions.

Table 1

Number of Records by Formulation, Fed Status, Regimen, and Ertugliflozin Dose in the Dose‐Proportionality Analysis

	Ertugliflozin Dose (mg)
Covariate	0.5	1	2.5	5	10	15	25	30	100	300	Total
Food
Fed (light meal)	0	8	0	8	0	0	14	0	8	0	38
Fasted	8	0/12a	8	34/12a	20	155/128a	18	8	48	7	306/269a
Formulation
Tablet	0	0/12a	0	34/12a	12	155/128a	18	0	40	0	259/222a
Suspension	0	0	0	8	8	0	14	8	16	7	61
Solution	8	8	8	0	0	0	0	0	0	0	24
Regimen
Single doseb	8	0	8	12	20	127	18	8	48	7	256
Steady stateb	0	8	0	30	0	28	14	0	8	0	88
Total	8	8/20a	8	42/20a	20	155/128a	32	8	56	7	344/307a

AUC, area under the plasma concentration‐time curve; Cmax maximum observed plasma concentration.

Values are for the AUC data set/Cmax data set.

Single‐dose and steady‐state values are for the AUC data set only. All Cmax data were obtained after administration of the first dose.

Across the ertugliflozin dose range included in the data set, exposure data ranged from 37.3 to 32 600 ng·h/mL for AUC and 6.1 to 5160 ng/mL for Cmax. Ertugliflozin dose‐normalized AUC and Cmax values demonstrated an approximately linear relationship between observed AUC and dose and between observed Cmax and dose (Figure 1).

Figure 1

Observed ertugliflozin dose‐normalized (A) AUC and (B) Cmax values by dose from the dose‐proportionality analysis. Red circles represent dose‐normalized AUCinf or dose‐normalized Cmax following a single dose; black triangles represent dose‐normalized AUCtau at steady state; geometric mean values for each dose group are represented by blue asterisks. AUC, area under the plasma concentration‐time curve; AUCinf, AUC from time zero extrapolated to infinite time; AUCtau, AUC from time zero to time tau, the dosing interval, for which tau is 24 hours; Cmax, maximum observed plasma concentration; SD, single dose; SS, steady state.

Modeled Data

For the dose‐proportionality base models, linear models relating AUC to ertugliflozin dose and Cmax to ertugliflozin dose were fitted and displayed a positive linear relationship between dose and AUC and between dose and Cmax. The parameter estimates from the base model fits for AUC and Cmax are shown in Table 2.

Table 2

Base‐ and Full‐Model Parameter Values and Precision Estimates for AUC and Cmax as a Function of Dose From the Dose‐Proportionality Analysis

	Base Model	Full Model
	Estimate (RSE, %a)	Estimate (RSE, %a)	95%CIb
AUCc
Slope (ng·h/mL/mg)	93.2 (1.50)	94.1 (1.76)	91.1‐97.4
Intercept (ng·h/mL)	–26.5 (28.8)	–33.7 (33.2)	–57.3 to –11.6
Suspension	—	1.04 (3.44)	0.976‐1.11
Solution	—	1.16 (6.35)	1.02‐1.31
Food (light meal)	—	0.896 (4.98)	0.813‐0.985
Cmax d
Slope (ng/mL/mg)	16.2 (2.20)	17.7 (1.92)	17.1‐18.4
Suspension	—	0.897 (4.77)	0.830‐0.978
Solution	—	0.821 (7.08)	0.719‐0.946
Food (light meal)	—	0.539 (8.09)	0.462‐0.613

AUC, area under the plasma concentration‐time curve; CI, confidence interval; Cmax, maximum observed plasma concentration; RSE, relative standard error.

Percent relative standard error was calculated as (100 × standard error/estimate).

95%CIs were generated from bootstrap.

AUC residual error was 44.9 ng·h/mL for the base model and 44.5 ng·h/mL for the full model (square root sigma values).

Cmax residual error was 25.2 ng/mL for the base model and 24.5 ng/mL for the full model (square root sigma values).

The covariates of formulation and food status (light meal) were added into the base models to develop full models for dose proportionality, and a positive linear relationship between dose and predicted AUC and Cmax was again observed. Population‐predicted, individual‐predicted, and observed values for AUC and Cmax fits using the full model are shown in Figure 2, including the model fit for the lower ertugliflozin dose range of 0.5 to 15 mg. The parameter estimates from the full models for AUC and Cmax and the 95%CIs generated from a nonparametric bootstrap utilized for model performance qualification are shown in Table 2. Administration of ertugliflozin as a suspension formulation did not have a significant effect on ertugliflozin AUC or Cmax relative to administration as a tablet. When administered as a solution, ertugliflozin AUC was higher by 16% (95%CI, 2%‐31%), but Cmax was lower by 18% (95%CI, 5%‐28%) versus the tablet form. AUC and Cmax were lower by 10% (95%CI, 1%‐19%) and by 46% (95%CI, 39%‐54%), respectively, when ertugliflozin was administered with a light meal relative to fasting administration (Table 2).

Figure 2

Population‐predicted, individual‐predicted, and observed ertugliflozin (A, B) AUC and (C, D) Cmax values by dose from the full model in the dose‐proportionality analysis. Results are shown across the full dose range (A, C) and across the 0.5‐ to 15‐mg dose range (B, D). AUC, area under the plasma concentration‐time curve; Cmax, maximum observed plasma concentration.

The predicted mean (90%CI) AUC values following administration of the 5‐ and 15‐mg tablet doses in the fasted state were 437 ng·h/mL (422‐451 ng·h/mL) and 1380 ng·h/mL (1350‐1410 ng·h/mL), respectively. For Cmax, these values were 88.7 ng/mL (86.0‐91.4 ng/mL) and 266 ng/mL (258‐274 ng/mL), respectively. The density plots of the predicted mean AUC overlaid with the mean observed AUC and the predicted mean Cmax overlaid with the mean observed Cmax are shown in Figure S1 and demonstrated good agreement between the predicted and observed data. The observed mean Cmax at 5 mg fell at the boundary of the predicted range of mean Cmax values. This could be because of the very small number of subjects (n = 12) in the 5‐mg dose group for the observed data.

UGT1A9 Genotype Analysis

For the UGT1A9 genotype analysis, 417 subjects (1 record per subject) were included in the analysis for AUC. The number of subjects by dose for the UGT1A9 genotype covariates is shown in Table 3. The doses contained within the UGT1A9 genotype data set ranged from 2.5 to 100 mg, with the largest number of records obtained following administration of a 15‐mg dose. The majority of subjects in this data set were carriers of UGT1A9*1b allelic variants (Table 3).

Table 3

Number of Subjects by UGT1A9 Genotype and Ertugliflozin Dose in the UGT1A9 Genotype Analysis

	Ertugliflozin Dose (mg)
Genotype	2.5	5	7.5	15	100	Total
WT for all SNPs	7	14	24	51	4	100
UGT1A9‐2152_het	1	5	4	6	0	16
UGT1A9*3_het	0	2	1	15	13	31
UGT1A9*1b_hom	10	10	17	33	0	70
UGT1A9*1b_het	12	26	47	62	0	147
UGT1A9‐2152_het, UGT1A9*1b_het	2	1	2	5	0	10
UGT1A9*3_het, UGT1A9*1b_het	0	0	2	15	16	33
UGT1A9*3_het, UGT1A9*1b_hom	0	0	0	3	0	3
UGT1A9‐2152_het, UGT1A9*3_het	0	0	0	0	6	6
UGT1A9‐2152_het, UGT1A9*3_het, UGT1A9*1b_het	0	0	0	0	1	1
Total	32	58	97	190	40	417

het, heterozygous; hom, homozygous; SNPs, single‐nucleotide polymorphisms; WT, wild type.

Across the ertugliflozin dose range included in the data set, exposure data ranged from 115 to 22 200 ng·h/mL for AUC. Ertugliflozin dose‐normalized AUC values are shown in Figure 3A; dose‐normalized AUC values by UGT1A9 genotype are shown in Figure 3B. An approximate linear relationship between observed AUC and dose was observed when AUC values were plotted by ertugliflozin dose (Figure 3A).

Figure 3

Observed ertugliflozin dose‐normalized AUC values by (A) dose and (B) UGT1A9 genotype from the UGT1A9 genotype analysis. (A) Red circles represent dose‐normalized AUCinf following a single dose; black triangles represent dose‐normalized AUCtau at steady state; geometric means for each dose group are represented by blue asterisks. (B) Open circles represent individual subject dose‐normalized AUC values. Box plot provides median and 25%/75% quartiles with whiskers to the last point within 1.5× interquartile range. AUC, area under the plasma concentration‐time curve; AUCinf, AUC from time zero extrapolated to infinite time; AUCtau, AUC from time zero to time tau, the dosing interval, for which tau is 24 hours; het, heterozygous; hom, homozygous; SD, single dose; SS, steady state; WT, wild type.

For the UGT1A9 genotype base model, a linear model relating AUC to ertugliflozin dose was successfully fitted to the observed data. The parameter estimates from the base model for AUC are shown in Table 4.

Table 4

Base‐ and Full‐Model Parameter Values and Precision Estimates for AUC as a Function of Dose From the UGT1A9 Genotype Analysis

	Base Model	Full Model
	Estimate (RSE, %a)	Estimate (RSE, %a)	95%CIb
Slope (ng·h/mL/mg)	96.7 (1.89)	97.1 (3.25)	91.8‐104
Intercept (ng·h/mL)	–63.9 (16.3)	–50.1 (22.8)	–72.9 to –30.6
UGT1A9–2152_het	—	0.988 (3.72)	0.914‐1.05
UGT1A9*3_het	—	1.10 (3.35)	1.03‐1.17
UGT1A9*1b_hom	—	0.945 (3.53)	0.879‐1.01
UGT1A9*1b_het	—	0.938 (2.89)	0.887‐0.987

CI, confidence interval; het, heterozygous; hom, homozygous; RSE, relative standard error.

Percent relative standard error was calculated as (100 × standard error/estimate).

95%CIs were generated from bootstrap.

Residual error was 0.244 ng·h/mL for the base model and 0.238 ng·h/mL for the full model (square root sigma values).

The covariates of the 3 allelic variants of UGT1A9 were added into the base model to develop the full model for UGT1A9 genotype, and a positive linear relationship between dose and predicted AUC was observed. Population‐predicted and ‐observed values for the AUC fit using the full model are shown in Figure S2. The parameter estimates from the full model for AUC and the 95%CIs generated from a nonparametric bootstrap utilized for model performance qualification are provided in Table 4. Based on the final parameter estimates, the UGT1A9‐2152 heterozygous variant did not have a significant effect on ertugliflozin AUC (the 95%CI included 1). The UGT1A9*3 heterozygous variant increased ertugliflozin AUC by 10% (95%CI, 3%‐17%). The UGT1A9*1b heterozygous variant decreased ertugliflozin AUC by 6% (95%CI, 1%‐11%), whereas homozygosity for UGT1A9*1b did not have a significant effect on ertugliflozin AUC (the 95%CI included 1). The fold change of ertugliflozin AUC for each UGT1A9 variant relative to the wild‐type variant is shown in Figure 4.

Figure 4

UGT1A9 polymorphic effects on ertugliflozin AUC from the UGT1A9 genotype analysis. The 90th percentiles of the bootstrap confidence intervals for AUC are provided. Effects are reported relative to the wild‐type subjects in the analysis. A value of 1 represents no change relative to the wild type. AUC, area under the plasma concentration‐time curve.

The predicted mean (90%CI) AUC values for the different UGT1A9 genotypes following administration of ertugliflozin 5‐ and 15‐mg tablets in the fasted state were derived using the full model and are shown in Table S2. The largest change in ertugliflozin AUC was with the UGT1A9*3 heterozygous variant: the mean (90%CI) AUC values following administration of the 5‐ and 15‐mg doses were 485 ng·h/mL (458‐510 ng·h/mL) and 1560 ng·h/mL (1480‐1630 ng·h/mL), respectively, compared with mean (90%CI) AUC values in UGT1A9 wild‐type subjects of 436 ng·h/mL (418‐455 ng·h/mL) and 1410 ng·h/mL (1350‐1480 ng·h/mL), respectively.

For multiple variants, the combined effect on ertugliflozin AUC was determined by multiplication of the effect from each individual variant. The maximum combined effect from the genotypes observed in the study participants would be for subjects carrying both UGT1A9‐2152 heterozygous and UGT1A9*1b heterozygous variants. The combined effect of this genotype (0.988 × 0.938 = 0.927) corresponds to a 7.3% decrease in ertugliflozin AUC compared with subjects carrying wild‐type UGT1A9.

Although some of the covariates were statistically significant, the magnitude of the effect of the allelic variants on ertugliflozin AUC were within ±10% of the wild type and were not considered clinically relevant (discussed below).

Discussion

This noncompartmental meta‐analysis of ertugliflozin PK parameters from 25 phase 1 studies evaluated the relationship between ertugliflozin exposure (AUC and Cmax) and dose and the effect of UGT1A9 genotype on ertugliflozin exposure (AUC). The outcomes of this analysis provided confirmation of a linear dose‐exposure relationship between ertugliflozin AUC and Cmax and the 5‐ and 15‐mg tablet formulation. Most importantly, because of the numbers of subjects and data in the analysis, precise estimates of UGT1A9 genotype impact on exposure were generated, affording the ability to indicate that there was no significance of the UGT1A9 variants assessed on ertugliflozin exposure. If a small study exploring the impact of genotype on the dose‐exposure relationship had been conducted, the ability to tease out the effect of genotype on exposure would have been diminished because of the relatively low allelic frequency of UGT1A9 variants in the general population.

Within the data set of subjects with ertugliflozin AUC values and UGT1A9 genotype information, 100 subjects were wild type for all 3 UGT1A9 allelic variants examined, 33 subjects carried heterozygous variants of UGT1A9‐2152, 74 subjects carried heterozygous variants of UGT1A9*3, and 264 subjects carried homozygous or heterozygous variants of UGT1A9*1b (Table 3). The relative proportions of subjects carrying each allelic variant within this data set are generally consistent with the known population frequencies of these variants,, , , with the majority of subjects carrying the common allelic variant UGT1A9*1b (minor allele frequency of 40% to 60%).,

Dose proportionality of ertugliflozin exposure was assessed via a meta‐analysis of ertugliflozin AUC and Cmax parameters across phase 1 studies to evaluate the relationship between ertugliflozin exposure and dose in healthy subjects. This analysis established that ertugliflozin AUC and Cmax values increased in a dose‐proportional manner across an ertugliflozin dose range of 0.5 to 300 mg. The effect on AUC and Cmax values when ertugliflozin was administered with a light meal relative to fasting administration was consistent with the results of the clinical study examining the effect of food on ertugliflozin PK, where administration of the ertugliflozin 15‐mg commercial tablet with a high‐fat meal resulted in no meaningful effect on ertugliflozin AUC. The effect of food on ertugliflozin Cmax is not considered clinically relevant, as ertugliflozin efficacy is driven by total exposure (AUC) rather than peak concentration.

The AUC data from these trials in healthy subjects were also used to conduct a pooled analysis evaluating the impact of the UGT1A9 genotype on ertugliflozin exposure. UGT‐mediated glucuronidation is an important biotransformation pathway in humans and facilitates the inactivation and excretion of a number of therapeutic drug targets. Indeed, the primary clearance pathway for ertugliflozin is metabolism, with glucuronidation accounting for nearly 90% of ertugliflozin biotransformation, primarily mediated by the UGT1A9 isoform., Polymorphic variation in the UGT gene family has the potential to affect UGT enzyme expression and activity levels, and ultimately target drug exposures, which may require compensatory modifications of drug dosages. Allelic variations that result in decreased glucuronidation activity or enzyme expression may lead to lower target drug clearance, necessitating the use of a reduced dose. Conversely, variants that result in increased enzyme activity or expression may lead to accelerated drug clearance, requiring administration of an increased dose. The Pharmacogenomics Knowledgebase lists nearly 90 reported associations between ∼40 separate UGT1A9 variants (SNPs or haplotypes) and a drug phenotype, with around half these associations classified as significant. However, functional evidence of an effect on UGT enzyme activity or expression with the potential to affect drug metabolism exists for only a handful of these variants (reviewed in reference ) and appears to be compound specific. For example, the immunosuppressant mycophenolic acid (MPA) is metabolized by several enzymes, including UGT1A9, with the major metabolic pathway being glucuronidation. A systematic meta‐analysis indicated that heterozygous white carriers of the UGT1A9*3 variant (associated with lower clearance of MPA) might benefit from receiving only about 70% of the average dose, whereas carriers of UGT1A9‐2152 (associated with higher clearance of MPA) may need higher than average doses. This current analysis found that the mean effects on ertugliflozin AUC of UGT1A9 polymorphic variation, as represented by the 3 UGT1A9 variants assessed in this study, were within ±10% of the wild‐type variant. Based on the dose‐linear PK of ertugliflozin (this study and references and ) and the dose‐safety relationships observed in phase 1‐3 clinical studies,, , , , combined with dose‐response modeling of clinical efficacy using phase 2/3 data, no adjustments in ertugliflozin posology are proposed for any extrinsic or intrinsic factors that were found to contribute to observed or predicted changes in ertugliflozin exposure., This included an assessment of potential DDIs, in which coadministration of the UGT and CYP inducer rifampin resulted in about a 39% reduction in ertugliflozin AUC, and physiologically based PK modeling demonstrated that coadministration of the UGT inhibitor mefenamic acid increased ertugliflozin AUC by ∼1.5‐fold. In the context of changes to ertugliflozin exposure of these magnitudes, the observed effects on ertugliflozin AUC of the UGT1A9 polymorphisms assessed in this study (within ±10%) will not have a clinically meaningful effect on ertugliflozin exposure, and no adjustments of the approved doses are required for patients carrying these UGT1A9 variants.

A meta‐analysis of pooled data from several phase 1 clinical studies was used in this analysis, which offers a number of advantages over a small, dedicated clinical study. Pooling data from several studies and conducting a meta‐analysis using this larger data set increases the statistical power of the analysis to detect if there is any impact on drug exposure because of the presence of specific enzyme variants. As the frequency of UGT variants is generally low, a dedicated clinical study would require the screening of a large number of individuals to find those few carriers of the specific variants. Often, dedicated genotyping clinical studies are relatively small and therefore lack the statistical power to detect the impact of genotype on drug exposure. In addition, there is no reported evidence of the UGT phenotype being impacted by T2DM or any other disease state; hence, the results of this meta‐analysis of pooled data from healthy subjects is applicable to T2DM patients.

Although a population PK analysis using ertugliflozin concentration‐versus‐time data could have been used for these analyses, the preferred approach was to not make any assumptions regarding ertugliflozin disposition but rather perform a meta‐analysis of noncompartmentally derived PK parameters. Therefore, this meta‐analysis was performed on individual subject AUC and Cmax values, precluding the parameterization of a population PK model, which would have required a number of additional binary covariates to accommodate the many patient populations and study design features with what, most likely, would have generated imprecise parameter estimation.

Conclusions

This meta‐analysis of healthy subjects from phase 1 trials showed that ertugliflozin AUC and Cmax increased in a dose‐proportional manner over the ertugliflozin dose range of 0.5 to 300 mg. No dose adjustments are required for patients with the UGT1A9 allelic variations examined in this study.

Conflicts of Interest

J.‐C.M., V.S., T.T., D.J.F., V.K.D., L.S.W., and K.S. are employees of Pfizer Inc., New York, New York, and may own shares/stock options in Pfizer Inc., New York, New York. Y.L. was an employee of Pfizer Inc., New York, New York, at the time the study was conducted. S.Z. is an employee of Merck Sharp & Dohme Corp., a subsidiary of Merck & Co., Inc., Kenilworth, New Jersey, and may own stock in Merck & Co., Inc., Kenilworth, New Jersey. R.K. was an employee of Merck Sharp & Dohme Corp., a subsidiary of Merck & Co., Inc., Kenilworth, New Jersey, at the time the study was conducted and may own stock in Merck & Co., Inc., Kenilworth, New Jersey.

Funding

This study was sponsored by Pfizer Inc., New York, New York, in collaboration with Merck Sharp & Dohme Corp., a subsidiary of Merck & Co., Inc., Kenilworth, New Jersey. Medical writing support was provided by Shirley Smith, PhD, of Engage Scientific Solutions (Horsham, UK) and was funded by Pfizer Inc., New York, New York, and Merck Sharp & Dohme Corp., a subsidiary of Merck & Co., Inc., Kenilworth, New Jersey.

Supplementary information

Click here for additional data file.