Results and evaluation of a first‐in‐human study of RG7342, an mGlu5 positive allosteric modulator, utilizing Bayesian adaptive methods

Abstract

Aim

The objectives of this first‐in‐human study were to evaluate the safety and tolerability, pharmacokinetics and pharmacodynamics, and maximum tolerated dose (MTD) of single ascending oral doses of RG7342, a positive allosteric modulator (PAM) of the metabotropic glutamate receptor 5 (mGlu5) for the treatment of schizophrenia, in healthy male subjects.

Methods

This was a single‐centre, randomized, double‐blind, adaptive study of 37 subjects receiving single ascending oral doses of RG7342 (ranging from 0.06–1.2 mg, n = 27) or placebo (n = 10). A modified continual reassessment method, with control for the probability of overdosing based on the occurrence of dose‐limiting events (DLEs), was applied to inform the subsequent dose decisions for RG7342.

Results

DLEs consisted of dizziness, nausea and vomiting, and the incidence and severity of these adverse events increased in a concentration‐dependent manner. RG7342 doses of 1.2 mg under fasting conditions, which reached a mean maximum plasma concentration (C_max) of 10.2 ng ml^–1, were not tolerated (four out of six subjects experienced DLEs). RG7342 showed dose‐proportional pharmacokinetics, with rapid absorption and a biphasic decline, and a mean terminal half‐life estimated to be >1000 h.

Conclusions

Single oral doses of RG7342 were generally tolerated up to 0.6 mg under fasting and 0.9 mg under fed conditions in healthy subjects. Bayesian adaptive methods describing the probability of DLEs were applied effectively to support dose escalation. MTDs (fasting, fed) were associated with a C_max of 6.5 ng ml^–1. The development of RG7342 was discontinued owing to the potential challenges associated with a long half‐life in context of the observed adverse events.

What is Already Known about this Subject

• Activation of the metabotropic glutamate receptor 5 (mGlu5) is a proposed novel mechanism of action for the treatment of schizophrenia.

• Bayesian adaptive designs are used to guide dose escalations and dose decisions in order to optimize the efficiency in terms of safety assessments.

What this Study Adds

• RG7342 is the first positive allosteric modulator of the mGlu5 receptor investigated in humans. Tolerability of higher RG7342 doses was limited by adverse events of dizziness, nausea and vomiting.

• Bayesian adaptive methods were applied successfully. Dose‐limiting events and maximum tolerated doses could be linked to plasma concentrations of RG7342.

Introduction

Activation of the metabotropic glutamate receptor 5 (mGlu5) has been proposed as a novel mechanism of action for the treatment of schizophrenia, with the potential to address symptom domains which are poorly alleviated with current treatments 1. The therapeutic potential of this mechanism is also supported by current glutamatergic associations with schizophrenia (glutamatergic dysfunction in patients is related to treatment resistance, clinical severity and clinical outcome) 2. The mGlu5 receptor is located on neurones postsynaptically as well as in glial cells, and expression is most prominent in brain areas considered to be relevant to schizophrenia, including the cortex, hippocampus, amygdala and striatum. mGlu5 receptors are also expressed peripherally, and drugs acting on these receptors in the pituitary gland or the islet of Langerhans have been shown to modulate levels of prolactin and to affect the homeostasis of D‐glucose, which were recognized as potential side effects 3.

RG7342 represents a novel, potent and selective, orally bioavailable mGlu5 positive allosteric modulator (PAM) which potentiates the activity of glutamate, with a half maximal effective concentration (EC₅₀) of 11.2 nM. In nonclinical behavioural paradigms, RG7342 shows a pro‐social, pro‐cognitive and antipsychotic drug‐like profile, in line with data reported for other mGlu5 PAMs 4, 5.

The nonclinical safety data consisted of repeat‐dose studies (daily up to 4 weeks) in rats, minipigs and cynomolgus monkeys. Full genotoxicity and safety pharmacology studies, as well as studies to assess the potential of phototoxicity (in vitro), skin sensitization, and skin and eye irritation, were also included. Effects noted after dosing of RG7342 included emesis in monkeys and minipigs, and increases in body temperature and respiratory rate in rats. In 4‐week multiple‐dose good laboratory practice (GLP) toxicity studies in rats and minipigs, RG7342 was well tolerated up to high dose levels. The no observed adverse effect levels (NOAELs) were 30 mg kg^–1 day^–1 in rats and 10 mg kg^–1 day^–1 in minipigs, which were well above the predicted efficacious exposure range in humans (corresponding to a safety margin of at least 70‐fold). Adverse events (AEs) were only observed at the highest doses tested and included: body weight gain or loss (in both species), frequent emesis, liquid faeces with related electrolyte changes and liver toxicity in the minipig, and cardiomyopathy, marked clinical signs and convulsions (one animal) in the rat.

Collectively, the pharmacological properties, preclinical toxicology and safety profile supported this first‐in‐human (FIH) study.

Bayesian adaptive designs are used to guide dose escalations and dose decisions, in addition to conventional approaches based on safety, tolerability and exposure in order to optimize the efficiency in terms of safety assessments. Bayesian adaptive designs have shown very good performance in the precise and accurate estimation of maximum tolerated dose (MTD), guiding dose selection during escalation steps, and in reducing the total number of subjects exposed 6. The modified continual reassessment method (mCRM) with overdose control has been shown specifically to reduce the risk of exposing subjects to doses greater than the MTD, compared with more traditional rule‐based designs 7.

The objectives of the present study, registered at http://clinicaltrials.gov/ (NCT02196636), were to evaluate the safety and tolerability, pharmacokinetics (PK) and pharmacodynamics (PD) of single ascending oral doses of RG7342 in healthy male subjects in the absence or presence of food, and to determine the MTD.

A strategy to limit risks was implemented in the study; this included careful starting dose selection, thorough planning and conduct of the study by applying sentinel dosing, and use of a diligent dose escalation procedure using Bayesian adaptive methods. To our knowledge, this is the first published clinical study with a PAM of the mGlu5 receptor.

The study demonstrated how AEs, including dose‐limiting events (DLEs), can be described quantitatively and how they can be related to study drug concentrations. DLEs from previous dose levels guided dose escalations using Bayesian model‐based methods. Such an approach in healthy subjects is novel, and might also be used for the development of other central nervous system drugs.

Methods

The study followed the principles of the Declaration of Helsinki and good clinical practice, and was conducted in full conformance with local laws and regulations. It was approved by the Independent Ethics Committee in Assen (METC St Bebo), the Netherlands, and the health authorities in Den Haag (CCMO Central Committee on Research Involving Human Subjects), the Netherlands. Healthy male subjects were enrolled at PRA Health Sciences in Groningen. Dose escalation steps were approved by the Independent Ethics Committee in Assen, the Netherlands.

Subjects – inclusion and exclusion criteria

Healthy male subjects aged 18–45 years with a body mass index (BMI) of 18–30 kg m^–2 were included in the study, after giving written informed consent. The main exclusion criteria included major illness within 1 month before screening, any prescribed or over‐the‐counter medication taken within 4 weeks prior to study drug administration until follow‐up, and subjects who, in the investigator's judgement, posed a suicidal or homicidal risk.

Subjects were healthy, as determined by prestudy medical history, physical examination, 12‐lead electrocardiogram (ECG), vital signs and clinical laboratory tests.

Study design

In this single‐centre, randomized, double‐blind, placebo‐controlled, adaptive, parallel group study, single ascending doses of RG7342 were administered as an oral solution.

A total of 37 subjects received single oral doses of RG7342, ranging from 0.06 mg to 1.2 mg, or placebo. The oral solution was administered to each subject under fasting conditions (for at least 10 h) or under fed conditions (30 min after starting a standardized normal breakfast).

The expected therapeutic dose range of RG7342 in humans was estimated to be between 1.1 mg and 4.3 mg once a day. A starting dose of 0.06 mg, which corresponded to 1/500 of the maximum recommended starting dose [calculated from the NOAEL in the most sensitive GLP species (rat) as per Food and Drug Administration (FDA) guidance 8], was selected as this dose was expected to be safe and associated with no or minimal PD effects.

The anticipated receptor occupancy for the 0.06 mg dose of RG7342 was 12% at maximum plasma concentration (C_max) and 1% 24 h after administration. The predicted C_max and area under the plasma concentration–time curve (AUC) after this dose were associated with substantial safety margins (>3500‐fold), based on the estimated NOAEL in the toxicology studies and toxicokinetic data. Further, the predicted C_max was much lower (>9‐fold) than the estimated EC₅₀ in the preclinical rodent in vivo models and also in relation to plasma concentrations at which safety‐related effects have been observed preclinically.

The study consisted of six successive groups (cohorts) of four to eight healthy subjects receiving a single oral dose of 0.06–1.2 mg of RG7342 or placebo (Figure 1). In the first cohort, three subjects were dosed with 0.06 mg under fasting conditions, and two subjects received placebo. In order to avoid simultaneous exposure of all subjects on the same day, cohort 1 was split into two groups: a sentinel group of two subjects were dosed on one day (one on active treatment and one on placebo) and three subjects (two on active treatment and one on placebo) were dosed on the following day.

Figure 1.

Disposition of subjects [sample size (receiving single doses of RG7342 or placebo)]

In both cohorts 2 and 3, three subjects were dosed with 0.2 mg and 0.6 mg under fasting conditions, respectively, whereas one subject per cohort received placebo.

In cohort 4, six subjects were dosed with 1.2 mg under fasting conditions, and two subjects received placebo. Dose escalation stopping rules were reached with cohort 4 when severe drug‐related AEs of the same type were observed in ≥50% of subjects.

In the subsequent cohorts 5 and 6, RG7342 was administered under different feeding conditions to test tolerability: six subjects per cohort were dosed with 0.6 mg and 0.9 mg under fed conditions, whereas two subjects per cohort received placebo.

In cohorts 4 to 6, the eight subjects were dosed on two different days (three subjects on active treatment and one on placebo on each day).

The subjects remained in the clinical unit from 2 days preceding study drug administration until 3 days (72 h) after study drug administration.

Safety and tolerability assessments

The safety of RG7342 was assessed by postdose monitoring of vital signs (supine and standing systolic and diastolic blood pressure, pulse rate and respiratory rate, body temperature), laboratory tests (haematology, clinical chemistry, coagulation tests and urinalysis) and ECGs, and comparison with baseline assessments. AEs that occurred after drug administration were rated on a three‐point severity grading scale as mild, moderate or severe. Intensity and causality of AEs including DLEs were assessed by the investigator. A DLE was defined as any treatment‐related adverse reaction that would prevent another drug administration at the same dose in a given subject.

PK assessments

Samples were collected to establish RG7342 plasma and urine concentrations and analysed using a specific and validated liquid chromatography–tandem mass spectrometry method (Swiss BioAnalytics AG, Birsfelden, Switzerland). Plasma and urine concentrations were measured before and at time points up to 110–115 days and 48 h, respectively, after drug administration. The plasma samples were analysed within the validated stability period of at least 33 days at −25°C, and the urine samples within 90 days at −20°C. The calibration ranges were from 0.05 ng ml^–1 to 50 ng ml^–1, and 0.1 ng ml^–1 to 100 ng ml^–1 for plasma samples and urine samples, respectively.

Noncompartmental analysis of PK parameters was performed using Phoenix WinNonlin (Version 6.4; Certara, Princeton, NJ, USA).

Statistical analysis

The mCRM, with control for the probability of overdosing, provided an updated estimation of the MTD based on the observed pattern of DLEs, using a Bayesian framework after each new cohort to support dose selection. It used a two‐parameter logistic regression to model the DLE rate by dose, AUC or C_max and to estimate the MTD.

The MTD was defined as the dose (or equivalent C_max level) for which the probability of being within the target safety interval (of 20–35% of the DLE rate) is maximized and the probability of being within the excessive toxicity interval (above 35% of the DLE rate) is below 25%. The algorithm used here was different to the original version of the continual reassessment method (CRM) introduced by O'Quigley et al. 9, and this is why the wording ‘modified’ CRM has been used 7. A minimally informative prior, as defined by Neuenschwander et al. 7, was used. Given the logistic regression model equation:

$$\text{logit}\left(P\right)=\alpha +\beta \cdot log\left(\frac{{\text{C}}_{\text{max}}}{{\text{C}}_{\text{ref}}}\right)$$

with P being the probability of a DLE at a given C_max level and Cref being a normalizing C_max level set, in this case, to 2 ng ml^–1; the prior for the two model parameters α and β was then defined based on the following bivariate normal distribution:

$$\left(\begin{array}{c}\alpha \\ log\left(\beta \right)\end{array}\right)=N\left(\mu =\left(\begin{array}{c}-2.5\\ -0.28\end{array}\right),\mathrm{\Sigma }=\left(\begin{array}{cc}6.34& ‐0.696\\ ‐0.696& 0.173\end{array}\right)\right)$$

The model based on C_max is only reported here, given that C_max was the strongest predictor of DLE occurrence (better than AUC and dose). The posterior distribution for the probability of DLE was calculated/updated after each new cohort and used to guide subsequent dose selection.

The R package crmPack, publicly available in the Comprehensive R Archive Network (CRAN; https://cran.r‐project.org/) was used to implement the mCRM. A detailed description of the package can be found in the corresponding package's vignette.

All PK and PD parameters were subjected to descriptive analysis, including arithmetic mean values, standard deviations, geometric mean values, medians, coefficients of variation and ranges.

A statistical analysis of PK parameters (AUC and C_max) was performed to explore dose proportionality under fasting or fed conditions. A linear model was applied to the log‐transformed, dose‐normalized PK study variables. Least‐square means and 90% confidence intervals were derived for each dose level and plotted to evaluate if there were obvious trends for normalized PK parameters (e.g. decreasing/increasing with dose). A formal analysis of variance was also applied to test the null hypothesis that the means across dose groups for a given normalized PK parameter were all equal.

An exposure–response analysis, as described by Darpo et al. 10 and recently mentioned by the FDA 11, was applied to investigate the effect of RG7342 concentration on the time‐matched change from baseline in Fridericia's correction for QTc Measurement (QTcF: QT (ms) = QTcF (ms)·(RR(ms) /1000)^1/3)). The following mixed‐effect linear model was used:

$$Y_{\mathit{ij}}=\left(\alpha +{\alpha }_i\right)+\left(\beta +{\beta }_i\right)\cdot c_{\mathit{ij}}+{\lambda }_k+{\vartheta }_s+{\epsilon }_{\mathit{ij}}$$

where Y_ij and c_ij are, respectively, time‐matched change from baseline in QTcF and RG7342 concentration at time j for subject i; α is an overall mean effect; α_i values are subject‐specific random intercepts (with a mean of zero and constant variance); β is the overall slope describing the effect of exposure; β_i values are subject‐specific random slopes (with a mean of zero and constant variance); λ_k is a two‐level categorical variable for treatment (with two levels: placebo and active); ϑ_s is a categorical variable for time (with levels according to the scheduled time points at which concentration/QTcF pairs are measured); and ε_ij values are random error terms (with a mean of zero and constant variance, assumed to be independent of the random effects).

PD to assess for safety and tolerability

PD samples to measure prolactin and glucose homeostasis parameters were collected following RG7342 oral dose administration and compared with time‐matched baseline samples (day −1). Prolactin, D‐glucose, insulin and proinsulin C‐peptide were measured using a specific and validated assay (Clinical Chemistry Laboratory, PRA Health Sciences, Zuidlaren, the Netherlands).

Dose escalation decision criteria and dose escalation stopping rules

Planned dose escalation steps were set to not exceed a 3.3‐fold increase. The study treatment was adaptive in nature, and the decision to escalate to the next dose was made following review of all safety information up to 48 h, PK data over at least 24 h postdose, and the available PD data in at least four subjects in each cohort. Doses could be repeated or adjusted downward on the basis of safety, tolerability, PK and/or PD observations at each dose level, as well as on the model‐based prediction of the anticipated MTD. Intermediate doses could also be proposed to be investigated.

Dose escalation was to be stopped if severe or clinically significant drug‐related changes in vital signs, ECGs, laboratory abnormalities or AEs of the same type occurred in 50% or more subjects receiving RG7342.

Nomenclature of targets and ligands

Key protein targets and ligands in this article are hyperlinked to corresponding entries in http://www.guidetopharmacology.org, the common portal for data from the IUPHAR/BPS Guide to PHARMACOLOGY 12, and are permanently archived in the Concise Guide to PHARMACOLOGY 2017/18 13.

Results

Thirty‐seven subjects received single ascending oral doses of RG7342 ranging from 0.06 mg to 1.2 mg (n = 27), or placebo (n = 10) (Figure 1). The demographics of the study subjects per treatment group are summarized in Table 1. Subjects were 18–42 years old, 78% were white and subjects had a BMI of 18.4–29.1 kg m^–2. All enrolled subjects were included in the analyses of safety, PK and PD. Five subjects took concomitant medications during the course of the study. One subject received chlorhexidine and lidocaine, and four subjects received ondansetron to treat AEs of nausea which were severe in intensity. One of these four subjects also received influenza virus vaccine.

Table 1.

Demographic characteristics of healthy male subjects per treatment group and feeding status [means (standard deviation)]

	Fasting					Fed
Dose	Placebo	0.06 mg	0.2 mg	0.6 mg	1.2 mg	Placebo	0.6 mg	0.9 mg
Male, N (%)	6 (100)	3 (100)	3 (100)	3 (100)	6 (100)	4 (100)	6 (100)	6 (100)
Race, N (%)
White	5 (83)	2 (67)	3 (100)	6 (100)	5 (83)	2 (50)	5 (83)	4 (67)
Black	–	–	–	–	–	–	–	2 (33)
Mixed race	1 (17)	1 (33)	–	–	1 (17)	2 (50)	1 (17)	–
Age (years)	19.8 (2.5)	22.0 (1.0)	19.7 (0.6)	22.3 (1.1)	24.2 (2.2)	26.3 (6.7)	28.8 (6.9)	25.5 (4.4)
Weight (kg)	71.6 (5.5)	70.5 (6.8)	82.1 (8.3)	77.1 (15.0)	75.2 (11.0)	87.2 (6.6)	81.7 (9.6)	68.4 (9.9)
BMI (kg m –2 )	22.5 (2.2)	22.6 (2.2)	25.3 (3.3)	22.9 (3.8)	23.1 (3.7)	25.3 (1.8)	23.7 (2.5)	21.3 (2.4)

BMI, body mass index

Safety and tolerability

No deaths, serious AEs, or AEs that led to study withdrawal occurred during the study. Single oral doses of RG7342 were generally tolerated up to 0.6 mg under fasting and 0.9 mg under fed conditions.

The most frequently affected system organ classes were gastrointestinal disorders, followed by nervous system disorders and then general disorders and administration conditions. The most frequently observed treatment‐emergent AEs were dizziness, nausea, vomiting, headache and fatigue (Table 2). The vast majority of AEs (~80%) started and resolved within 24 h postdose.

Table 2.

Summary of reported treatment‐emergent adverse events (AEs) by system organ class, trial treatment and feeding status

	Fasting					Fed
RG7342 (mg)	Placebo	0.06	0.2	0.6	1.2	Placebo	0.6	0.9
Total number of subjects	6	3	3	3	6	4	6	6
Total subjects with at least one AE (%)	3 (50)	3 (50)	2 (67)	3 (100)	6 (100)	3 (50)	3 (50)	6 (100)
Total number of AEs	7	3	2	17	31	2	18	20
Dose‐limiting events (%)	0	0	0	1 (33)	4 (67)	0	0	1 (33)
Gastrointestinal disorders (%)	1 (17)	0	1 (17)	3 (50)	6 (100)	0	3 (50)	6 (100)
Nausea	1	0	1	3	6	0	3	5
Vomiting	0	0	0	1	6	0	0	2
Abdominal discomfort	0	0	0	1	0	0	1	2
Nervous system disorders (%)	2 (33)	0	0	3 (100)	5 (83)	0	4 (67)	5 (83)
Dizziness	0	0	0	3	4	0	3	5
Headache	2	0	0	1	2	0	2	0
Paraesthesia	0	0	0	0	2	0	1	0
General disorders and administration conditions (%)	1 (17)	2 (33)	0	3 (50)	1 (17)	0	2 (33)	1 (17)
Fatigue	0	1	0	3	0	0	2	0
Infections and infestations (%)	0	0	0	0	0	0	2 (33)	1 (17)
Nasopharyngitis	0	0	0	0	0	0	2	0

Individual AEs were listed with their incidence if they appeared in at least two subjects in any of the treatment groups. AE, adverse event

Six subjects on RG7342 experienced DLEs at doses ≥0.6 mg, one subject in the 0.6 mg fasting group, four subjects in the 1.2 mg fasting group and one subject in the 0.9 mg fed group. DLEs, which were considered as severe in intensity, were dizziness, nausea and vomiting.

The incidence and severity of these AEs increased in a dose‐ and concentration‐dependent manner. Dose escalation stopping rules were reached after a single dose administration of 1.2 mg under fasting conditions when a severe drug‐related AE of nausea was observed in four out of six subjects.

There were six sensory, motor AEs (three AEs of paraesthesia, dyskinesia, tremor and muscle twitching), which were considered to be treatment related, reported in six subjects who received ≥0.6 mg RG7342 under fasting or fed conditions. All of these AEs were mild in intensity, except for one AE of dyskinesia, which was of moderate intensity. The onset and offset of these AEs were concentration dependent and all AEs resolved without sequelae within 24 h.

Doses of RG7342 administered under fed conditions were better tolerated compared with those under fasting conditions. A dose of 0.6 mg under fed conditions was well tolerated in 50% and moderately tolerated in 50% of the subjects (three out of six subjects on active treatment), whereas the same dose under fasting conditions was well tolerated in only 33% of the subjects (one out of three subjects on active treatment), moderately tolerated in 33% of the subjects (one out of three) and not tolerated in 33% of the subjects (one out of three).

No clinically relevant changes in laboratory parameters, physical examination, body temperature or respiratory rate were observed.

A trend towards an increase in systolic and diastolic blood pressure, in both the supine and standing position, was observed at 0.6 mg and 1.2 mg under fasting conditions. Mean increases in supine systolic and diastolic blood pressure up to +11 mmHg and +6 mmHg, respectively, compared with baseline were observed within the first few hours after drug administration.

There was an apparent trend for RG7342 to cause a dose‐dependent increase in QTc and heart rate on day 1 within the first few hours after administration, compared with the day −1 baseline (mean increase up to +12 ms and +9 bpm for QTc and heart rate, respectively, at a dose of 1.2 mg). Results from an exposure–response analysis revealed that the time‐matched change from baseline in QTcF significantly increases with RG7342 concentration [the corresponding slope, from the model, was 1.32 (95% confidence intervals: 0.48, 2.15); P = 0.002]. Figure 2 shows the prediction from the model of the effect of RG7342 concentration on the ΔQTcF (the black line is the mean, and the grey shaded area the 90% confidence interval), together with the observed ΔQTcF (red squares).

Figure 2.

The predicted effect on time‐matched change‐from‐baseline QTcF (ΔQTcF) using a concentration–QTcF effect model. The solid black line with the grey shaded area denotes the model‐predicted ΔQTcF with 90% confidence interval (CI) as a function of plasma concentration. The horizontal red line shows the range of plasma concentrations divided into deciles. The red squares with vertical bars denote the observed arithmetic means with 90% CIs for ΔQTcF within each plasma concentration decile (the x‐axis locations of the red squares represent the median concentration for all samples within each decile)

PK

Following oral administration after an overnight fast, RG7342 was absorbed rapidly, with a median time to C_max (T_max) achieved at 1 h (Table 3). When administered 30 min after the start of a normal breakfast, C_max was reached at a later time point, with a median T_max of 2.5–3.5 h achieved postdose. Food had no effect on overall plasma exposure (AUC_0–24 h) but reduced the C_max by 37%. Plasma concentration vs. time profiles of RG7342 then showed a biphasic decline (Figure 3). The terminal phase, representing the major portion of the total AUC, was characterized with a mean terminal half‐life estimated at >1000 h for all dose levels ≥0.6 mg. The terminal half‐life of RG7342 at doses of 0.6–1.2 mg could not be estimated accurately as time points were spread over only approximately two half‐lives [blood samples collected up to 16 weeks (>2500 h) postdose]. At lower doses, the terminal half‐life could not be estimated owing to plasma concentrations falling below the limit of quantification (0.05 ng ml^–1). Similarly, Area under the plasma concentration versus time curve extrapolated to infinity (AUC_0–∞) could not be reliably determined in the present study. However, it was estimated that at doses ≥0.6–1.2 mg, the AUC_0–24 h accounted for less than 5% of the total AUC of RG7342. Plasma exposure (AUC_0–24 h) and C_max increased in a dose‐proportional manner (the means of the dose‐normalized PK parameters across the dose groups were not statistically different) following single oral administration from 0.06 mg to 1.2 mg of RG7342. Although no statistically significant deviation from dose proportionality was observed for the C_max of RG7342, there was a visual trend indicating that the dose‐normalized C_max decreased (by approximately 25%) between the 0.2 mg and the 0.6 mg dose level.

Table 3.

Summary of pharmacokinetic parameters of RG7342 after single‐dose administration of 0.06 mg, 0.2 mg, 0.6 mg and 1.2 mg RG7342 under fasting conditions, and 0.6 mg and 0.9 mg RG7342 under fed conditions

	Fasting				Fed
Dose	0.06 mg	0.2 mg	0.6 mg	1.2 mg	0.6 mg	0.9 mg
N	3	3	3	6	6	6
Cmax (ng ml–1)	0.677 (15.7)	2.22 (15.0)	4.93 (2.52)	10.2 (20.6)	3.12 (34.4)	4.04 (29.2)
Tmax a (h)	1.00 [0.50‐1.00]	1.00 [1.00–1.00]	1.00 [0.50–1.50]	1.00 [0.50–1.50]	3.50 [0.50–6.00]	2.50 [1.00–4.00]
AUC0–24 h (ng.h ml–1)	3.58 (8.5)	10.4 (18.7)	34.0 (15.8)	60.0 (23.7)	34.2 (14.3)	46.5 (31.6)

All data geometric mean (%CV) unless stated otherwise. AUC_0–24 h, area under the plasma concentration–time curve from time zero to 24 h; C_max, maximum plasma concentration; %CV, coefficient of variation; T_max, time to reach C_max

^a Median values [range]

Figure 3.

Arithmetic mean plasma concentrations vs. time profiles of RG7342 after single‐dose administration of 0.06 mg, 0.2 mg, 0.6 mg and 1.2 mg RG7342 under fasting conditions, and 0.6 mg and 0.9 mg RG7342 under fed conditions (A) up to 48 h postdose, and (B) up to the follow‐up visit, on a semi‐logarithmic scale (error bars are standard deviations)

After oral dosing, no quantifiable amount of RG7342 was found in the urine samples collected up to 48 h postdose.

Bayesian methods

A Bayesian model‐based approach, an mCRM, based on the occurrence of a DLE with control for the probability of overdosing, was applied to determine the MTD and to guide the dose escalation steps.

As there were no DLEs observed and RG7342 was safe and well tolerated in cohorts 1 and 2, the dose was escalated by approximately threefold to 0.2 mg and 0.6 mg for cohorts 2 and 3, respectively, under fasting conditions. In cohort 3 (total of four subjects), RG7342 was moderately tolerated in one subject and poorly tolerated in another subject. It was decided to proceed to cohort 4 with 1.2 mg under fasting conditions (increasing the dose by twofold) and with a cohort size of eight subjects (six on active treatment and two on placebo). Dose escalation stopping rules were reached with cohort 4 when severe drug‐related AEs of the same type were observed in ≥50% subjects. In cohorts 5 and 6, the safety and tolerability of 0.6 mg and 0.9 mg of RG7342 under fed conditions were assessed in eight subjects (six on active treatment and two on placebo). As the onset and offset of AEs which resulted in DLEs could be linked to RG7342 plasma concentrations, and both the incidence and severity of AEs correlated with maximum plasma concentrations in fasting and fed conditions (Figure 4), the MTD of RG7342 was estimated as the plasma C_max. With this approach, data under fasting and fed conditions could be combined. Figure 5 shows the probability of DLEs at different values of C_max based on the posterior estimates from the mCRM model (median and 95% credible intervals are shown). Figure 6 shows that the probability of a true DLE rate in the targeted safety interval of 20–35% was maximized around a C_max of 6.5 ng ml^–1, for which the probability of overdosing (above 35% of the DLE rate) was below 25%. At this level of exposure, the DLE rate (and corresponding 90% credible intervals) was estimated to be 27% (14%, 44%), while the probability of being in the target safety window was 58%.

Figure 4.

Probability of a moderate and severe adverse event (AE) as a function of maximum plasma concentrations (C_max) [solid line indicates the median, and dashed lines the 90% credible intervals (CI)]

Figure 5.

Probabilities of dose‐limiting events (DLEs) vs. maximum plasma concentrations (C_max) based on the continual reassessment method (medians and 90% credible intervals for each maximum plasma concentration). Solid horizontal line for the modified continual reassessment method indicates a target toxicity interval of 20% and 35% of the DLE rate

Figure 6.

(A) Probability of overdosing (dose‐limiting event (DLE) rate above 35%) at different maximum plasma concentration (C_max) levels. The solid horizontal line indicates a 25% probability of overdosing (above 35% of the DLE rate). (B) Probability of being at target (DLE rate between 20% and 35%. (C) Probability of underdosing (DLE rate between 0% and 20%). These probabilities were derived based on the modified continual reassessment method

Therefore, doses of RG7342 reaching C_max of 6.5 ng ml^–1 were considered to be the MTD, which corresponds approximately to a dose of 0.7 mg under fasting and 1.3 mg under fed conditions.

PD to assess for safety and tolerability

At a dose of 1.2 mg of RG7342, mild acute changes in fasting D‐glucose (mean increase of approximately 1–2 mmol l^–1 compared with a baseline of approximately 5 mmol l^–1), and prolactin levels (mean increase of up to 24 ng ml^–1 at 1 h postdose compared with a baseline of approximately 8 ng ml^–1) were observed in healthy male subjects. After meal intake (for both fasting and fed cohorts), a trend towards a decrease in insulin and proinsulin C‐peptide compared with the day −1 baseline was observed (mean decreases of up to 27 uU ml^–1 and 0.8 nmol l^–1 compared with baselines of approximately 50 uU ml^–1 and approximately 2 nmol l^–1 for insulin and proinsulin C‐peptide, respectively).

Discussion

The RG7342 starting dose of 0.06 mg was appropriately selected and targeted low receptor occupancy levels at the predicted plasma concentrations. No or minimal PD effects, including AEs, were observed at this dose level. Electroencephalograms (EEGs) were assessed as a PD outcome in the study, and these data will be reported separately. The first DLE was observed in cohort 3, at a dose of 0.6 mg. In cohort 4, a dose of 1.2 mg, which was already above the MTD, was investigated. In summary, with only four cohorts and a total of 19 subjects, the doses investigated ranged from producing no or minimal PD effects to not being tolerated. Estimating the starting dose based solely on the NOAEL in the most sensitive species (rat), conversion to the human equivalent dose and applying a safety factor of 10 8 would have resulted in a dose of 29 mg based on the rat and minipig GLP toxicity studies. Such a dose would have been >40 times above the MTD and could have resulted in serious complications. This demonstrates that it is of utmost importance that the starting dose is considered as not, or minimally, pharmacologically active, especially for novel mechanisms of action without prior experience in humans 8.

A Bayesian adaptive design was applied to improve the way that this FIH study with RG7342 was conducted. It utilized the mCRM with overdose control to reduce the risk of exposing subjects to doses greater than the MTD and to improve the precision and accuracy of the MTD estimation. The mCRM was combined with the strategy of having smaller cohorts at the beginning of the study (3 + 1; i.e. three subjects on active treatment and one on placebo) and expanding the cohort size to 6 + 2 at projected therapeutic levels. This expansion occurred in cohort 4 after observing the first DLE with RG7342 in cohort 3 at a dose of 0.6 mg. A total of 37 subjects received study medication (RG7342 or placebo) to assess the MTD and to investigate the effect of food. Twenty‐eight out of 37 subjects were enrolled to cohorts in which doses ≥0.6 mg RG7342 were administered. Compared with a non‐Bayesian study design, this represents an improvement in efficiency by increasing the relative number of subjects exposed to more informative doses.

Dose escalation stopping rules were reached with cohort 4 at a dose of 1.2 mg under fasting conditions owing to the severe drug‐related AE of nausea in ≥50% subjects. We hypothesized that the administration of food might be beneficial to the safety and tolerability profile because of the expected reduction in the absorption rate. Utilizing physiologically based PK modelling, it was predicted that food would lower the C_max of RG7342, whereas the AUC would not be affected, compared with fasting conditions. Indeed, administering RG7342 under fed conditions was better tolerated compared with fasting conditions, and AEs such as nausea and vomiting could be reduced in both intensity and occurrence. The administration of food reduced the absorption rate and RG7342 C_max, but did not have an effect on the overall extent of exposure. An exploratory analysis revealed that the improved safety and tolerability profile of RG7342 under fed conditions could be attributed to the reduction in C_max rather than the reduced absorption rate, and the frequency and intensity of AEs were similar for fasting and fed conditions at comparable values of C_max.

Therefore, the overall analysis to define the MTD of RG7342 was performed utilizing C_max, which allowed the safety and tolerability data of AEs under fasting and fed conditions to be combined, avoiding the need to define the MTD separately for fasting and fed conditions. The MTD was associated with a C_max of 6.5 ng ml^–1, and at this level of exposure the DLE rate was maximized to be in the target safety window with a good probability of 58%.

Based on preclinical data, the estimated mGlu5 receptor occupancy at this concentration of 6.5 ng ml^–1 (C_max) is approximately 50–80%. While preclinical behavioural models have suggested that occupancy data in this range might be clinically relevant, there would be virtually no therapeutic or safety window. It is also unclear how long these concentrations would need to be maintained to provide a clinical benefit. Due to the extremely long half‐life, a multiple‐dose study to explore higher receptor occupancies was not supported as the risk of prolonged AEs was considered unacceptable.

The DLEs in the present single ascending dose study were dizziness, nausea and vomiting. Higher tolerated concentrations could potentially be achieved under a multiple‐dose setting, as suggested by preclinical data. In a repeat‐dose study in cynomolgus monkeys, it was shown that tolerance to RG7342, with regard to emesis, was developed after a few days with daily treatment. Exposure to RG7342 (C_max and AUC) could be increased by 7–9‐fold, and the occurrence of emesis disappeared over the study period. However, it remains unknown if tolerability in humans would improve after repeated administration, allowing for sustained receptor occupancy. Nausea and vomiting are more pronounced for fast‐acting drugs and have been shown to diminish over time 14. In general, higher concentrations than those anticipated to be tolerated can be achieved from either accumulation under multiple dosing or through an up‐titration schedule. For RG7342, significant accumulation is expected following multiple‐dose treatment. Based on a mean estimated terminal half‐life of >1000 h, it was predicted that it would take longer than 4 months to reach steady‐state concentrations, resulting in an >10‐fold increase in C_max. By contrast, for compounds with a short half‐life, gradual up‐titration might mitigate the observed AEs.

Increases in QTc, heart rate and blood pressure were observed in subjects on day 1 after administration of RG7342, and correlated with exposure to RG7342. While mGlu5 receptors are expressed in the heart 3, including the conduction system, no cardiovascular safety signals were observed with RG7342 in vitro or in the minipig. It is not known whether AEs such as dizziness, nausea and vomiting, which were also exposure dependent, had an impact on the observed changes. Though, increases in QTc were still apparent in an exploratory analysis when data of subjects experiencing vomiting, as well as data of subjects receiving ondansetron as concomitant medication (known to cause QTc prolongation 15) were removed.

In addition, mild acute changes in blood D‐glucose (increases in fasting D‐glucose), insulin and proinsulin C‐peptide (decreases in postprandial insulin and proinsulin C‐peptide), and prolactin levels were observed in healthy male subjects receiving RG7342 in a concentration‐dependent manner. It is not clear whether these mild and acute effects on endocrine parameters, including blood D‐glucose, insulin and prolactin, are target mediated. Observed AEs such as nausea, vomiting, feeling unwell and stress are all known to have a clinically significant impact on these parameters. Stress is known to increase cortisol levels, and this could explain the slight elevations in blood D‐glucose. In addition, RG7342 might have an effect on gastric emptying 3, resulting in slight changes in insulin and proinsulin C‐peptide levels under fed conditions 16. Elevations in prolactin levels are well known for antipsychotic treatments, and are most pronounced in the first few hours after administration. The maximum observed prolactin concentrations observed with RG7342 (up to 30 ng ml^–1 at 1.2 mg) were observed within the first few hours after administration, and are also within the range for other antipsychotic drugs 17.

While some cardiovascular and endocrine changes were observed after treatment with RG7342, it cannot be ruled out that these PD effects were confounded by observed AEs, study‐related stress, concomitant medication and autonomic effects. Further limitations of the study were that it had a small overall sample size and a limited number of data points at higher RG7342 concentrations. An adequate sample size, under a multiple‐dose regimen would be needed to enable a conclusive assessment to be made about the clinical relevance of the observed changes and the potential link to the mechanism of action. As tolerance to the potential confounding AEs (e.g. nausea and vomiting) are expected on repeated RG7342 administrations, a better assessment of mGlu5‐related effects on these systems could be conducted.

The present study demonstrated that a Bayesian adaptive design, when applied successfully, can provide significant advantages. Overall, single oral doses of RG7342 were tolerated up to 0.6 mg under fasting, and 0.9 mg under fed conditions. The tolerability of higher doses under fasting conditions was limited by the observed DLEs of dizziness, nausea and vomiting, and doses reaching a C_max of 6.5 ng ml^–1 were considered to be the MTD, regardless of feeding status.

Although mGlu5 PAM remains a promising mechanism for the treatment of schizophrenia warranting further exploration, the development of RG7342 was discontinued owing to the challenges associated with an unfavourable half‐life, in view of the observed AEs of the drug. Nevertheless, the present study illustrates a novel approach to describing AEs quantitatively. This could also be applied to other central nervous system drugs.

Competing Interests

This study was sponsored by F. Hoffmann‐La Roche Ltd. All authors have completed the Unified Competing Interest form at www.icmje.org/coi_disclosure.pdf (available on request from the corresponding author) and declare that, at the time of the study, S.S., M.D., S.S., L.L., R.W., G.J., M.D. and G.P. were employees with F. Hoffmann‐La Roche and had no other relationships or activities that could appear to have influenced the submitted work.

The authors thank Núria Bech for the operational management of the study.

Sturm, S. , Delporte, M.‐L. , Hadi, S. , Schobel, S. , Lindemann, L. , Weikert, R. , Jaeschke, G. , Derks, M. , and Palermo, G. (2018) Results and evaluation of a first‐in‐human study of RG7342, an mGlu5 positive allosteric modulator, utilizing Bayesian adaptive methods. Br J Clin Pharmacol, 84: 445–455. doi: 10.1111/bcp.13466.