CNS Trial Failures: Using the Fragile X Syndrome-mGluR5 drug target to highlight the complexities of translating preclinical discoveries into human trials

Abstract

Purpose/Background:

Drug trials of the central nervous system (CNS) have been plagued with uninformative failures, often due to the difficulties of knowing definitively whether dosing achieved was sufficient to modulate the intended CNS target at adequate concentrations to produce pharmacodynamic (PD) or dose-related changes in readouts of brain function. Key design elements can be introduced into early-stage trials to get at this issue.

Methods/Procedures:

This commentary builds on a review of earlier clinical studies in Fragile X Syndrome (FXS) to explore the extent to which the chain of evidence is in place to allow for interpretation of the results as ruling in or out the utility of modulating one or another molecular target to treat this disorder. Recent and current biomarker studies in FXS occurring subsequent to the clinical studies are reviewed to see if they might address any chain of evidence gaps.

Findings/Results:

Despite the strong preclinical basis for targeting molecular mechanisms, the lack of efficacy seen in clinical studies remains uninterpretable, with regard to ruling in or out the utility of targeting the mechanism in a clinical population, given the absence of studies which address whether doses of administered drug impacted the targeted brain mechanism.

Implications/Conclusions:

The value of pursuing clinical studies of compounds targeted to novel mechanisms in the absence of clinical pharmacological evidence of some anticipated mediating pharmacokinetic/pharmacodynamic (PK/PD) signals is questionable. One or more biomarkers of a drug effect on brain function are needed to establish dose dependent CNS effects that allow one to interpret clinical results as ruling in or out a mechanism and providing a firm basis for continuing or not, as well as informing dose selection in any clinical efficacy trials. Initiatives to address this general need in pediatric psychopharmacology are highlighted.

Advances in understanding possible pathologic mechanisms typically drive drug development, and Fragile X Syndrome (FXS) is a striking example of this approach. However, in this case, the speed with which researchers moved from a postulated mechanism to clinical study was very quick, while the development of tools and methods needed to successfully test the hypothesis clinically did not move in parallel. Thus, it is unclear whether the early negative trials disproved the hypothesis. On the face of it, the hypothesized FMR protein in FXS is very compelling, yet both Novartis and Roche have recently abandoned their FXS R&D programs focused on targeting the glutamatergic metabotropic 5 (mGluR5) receptor (Novartis’ AFQ056/mavoglurant and Roche’s RG7090/basimglurant) due to failed Phase II studies in adults and adolescents (1–3). Additionally, the small biotech Seaside Therapeutics had to close down when results of primary outcome measures in Phase II ASD and FXS trials testing STX209 (arbaclofen - a GABA-B receptor agonist) were negative in both trials (4). There were sufficient technical or methodological problems with these studies that prevented the companies from knowing whether the drug candidate’s mechanism of action was clinically tested or not. This paper focuses on the possibility that these studies involve type 2 errors, failing to find differences when there is one. An extended earlier review by Berry-Kravis and colleagues (5) on “lessons learned” from some of these trials highlights a range of other issues with trial design and outcome measures as well as limits on the utility of rodent models and translating their results into clinical efficacy designs. The problem of large placebo effects is also raised (5) and based on the troubling trend of increased placebo responses in schizophrenia over time (6), the problem is not likely to improve as long as the field depends solely on clinical measures subject to expectation bias. The Berry-Kravis review (5) focuses on most issues that could lead to a signal being missed but not on the possibility that the dose of drug administered might have been inadequate to test engagement of the target in question. The present commentary focuses on what ideally would have been in place in terms of clinical pharmacology studies to provide an adequate understanding of the PK/PD relationships needed to best interpret clinical efficacy studies.

The FXS studies cited above were implemented based on the organizing principle of the fragile X mental retardation protein (FMRP; a translational repressor protein) and its potential balancing act with group 1 mGluR receptor activation, where FMRP suppresses protein synthesis at excitatory synapses, while the group 1 mGluRs enhance it. FXS is a disorder resulting from the silencing of FMR1 which normally encodes FMRP; loss of fMRP results in excessive protein synthesis and altered neuronal dendritic morphology and may underlie the cognitive and behavioral deficits seen in animals (7). The availability of mGluR5 antagonists (mGluR5 receptors are part of the group 1 mGluR receptor family) or genetic suppression of mGluR5 expression in rodents with FXS loss of function allowed for preclinical studies that demonstrated rescue of morphological, functional and behavioral deficits (7). The use, in turn, of GABA-B receptor agonists is intended to indirectly target the mGluR5 receptors by decreasing pre and postsynaptic glutamate release (8). Both targets are intriguing from a drug development perspective and have received highly publicized attention based on the human genetic findings and the preclinical “rescue” studies. Is there now enough evidence to rule out these targets?

Trials in conditions such as FXS for which no treatment exists are especially challenging in terms of ruling in or out benefit from a novel intervention. First, is the difficulty in establishing that the dose used in a trial “engages” the molecular target in the brain which ideally can be answered in clinical pharmacological studies with recent methodological advances. Second, and more generally, in the absence of a positive control, one is not sure of “assay sensitivity” in trials where numerous factors related to signal detection may have made it impossible to detect a real signal. Unfortunately, as already noted, classic primary outcome measures in psychopharmacologic trials in developmental disorders such as ASD and FXS are subject to caregiver, patient or clinician-based ratings that are often laden with expectation bias which, to date, has not been addressed by methodological advances. Indeed, the ability to measure a signal beyond the placebo effect, is especially challenging given an average effect size of 0.5 in a meta-analysis of placebo effects in FXS drug trials (9). An additional consideration for assay sensitivity is the typically large age range of subjects enrolled in the FXS studies, due to rareness of the illness, as in the case of the arbaclofen Phase II FXS trial which included subjects 6–39 years old. These drug agents may have different PK/PD relationships across neurodevelopmental age, an important clinical pharmacological consideration which remains an open question. Finally, although FXS subjects are already stratified based on a genetic mutation, patients express the symptoms such as social deficits, cognitive deficits, and irritability, to different extents. Therefore, early trials may consider which FXS subjects are most likely to respond to the drug, based on specific symptom expression but much better would be the use of quantitative CNS measures that objectively define more homogeneous subpopulations. Such measures should also, if available, be incorporated into one’s Phase 1 PK/PD brain effect studies. At the time the cited FXS studies were designed and carried out, however, relevant CNS measures were not available and may still need to be developed to fully test the mGluR5 mechanistic hypothesis.

Subsequent to the aforementioned negative trials, objective measures of brain biochemistry (mGluR5 receptor occupancy using positron emission tomography (PET) radioligands (10, 11)) and measures of brain function (electroencephalography, EEG (12, 13)) have been developed with a goal of utilizing them in clinical trials. The PET studies observed less displacement of mGluR5 receptors in various subcortical regions of the brains of FXS subjects compared to typically developing controls, although the regional differences were not consistent across the studies. The results raise the possibility that individual subjects can be identified with lower regional uptake of the radioligand indicating that fewer unoccupied receptors are available which, in turn, could be used to objectively select subjects for FXS treatment trials targeting the mGluR5 pathway as well as to consider regional differences in the analysis of any clinical pharmacology stage studies that include receptor occupancy measures.

In addition to PET receptor occupancy studies, various EEG approaches have been applied to FXS, in an attempt to uncover pathophysiological CNS processes. At least one has been indirectly replicated by Ethridge et. al. (12) involving an increased amplitude of an early (N1) peak in the EEG waveforms observed following a stimulus. Whether this response is influenced by altering mGluR5 function remains to be determined. Another application of EEG-based biomarkers of brain function is raised by a recent study of polysomnography in children with FXS (without a history of epilepsy) to assess sleep architecture abnormalities and possible interictal epileptiform discharges (IEDs) (13). In addition to altered polysomnography parameters, about half of the children with FXS displayed IEDS. These findings raise the question of both different PD and therapeutic effects in FXS patients with and without the IEDS detected sleep recordings as well as the question of whether brain effect PK/PD studies in healthy volunteers can substitute for doing these in the target patient population. The unifying feature among these studies and others using various biomarker modalities is the ability of objective measures to differentiate FXS subjects into subgroups that may better associate with pathophysiological alterations as well as inform the interpretation of PK/PD studies. There are obviously many gaps, with regard to selecting the best measures to characterize participants and drug effects in their brains, but these examples present a basis for designing studies which will allow us to be more confident in ruling in or out the efficacy of a particular molecular intervention.

The issues raised by this brief discussion of trials in FXS have been at the forefront of the National Institute of Mental Health (NIMH) with the recent establishment of early-stage clinical trial funding programs (Phase Ia, Ib and Phase IIa) to test investigational drugs through an experimental medicine (EM) approach(14) followed by a 2020 National Advisory Mental Health Council Workgroup on Drug Development (15). The overall premise of EM is to design the trial by incorporating objective measures of the drug’s mechanism of action into the trial to enable one to evaluate pharmacodynamically: 1) whether the drug dose-dependently modulates its molecular target without significant side effects, and 2) how the extent of target engagement associates with symptom or functional changes. This approach should better predict CNS drug actions in humans by relying less on subjective clinical outcome data, and instead use PD measures such as EEG, functional magnetic resonance imaging (fMRI), magnetic resonance spectroscopy (MRS), eye tracking, etc. based on knowledge of the drug target and which CNS measure can best detect target engagement. In using these PD measures, if the drug does not modulate the target, the dose range cannot be validated, and efficacy studies should not be pursued. If the drug does modulate the target, but only at doses where significant safety or tolerability issues are observed, efficacy studies should also not be pursued.

Importantly, for FXS and other developmental CNS disorders, pediatric bridging studies lend themselves nicely to the EM paradigm, where pharmacokinetic/pharmacodynamic (PD/PK) dose responses are measured to determine whether acute drug dosing in pediatric subjects can achieve drug plasma levels that are comparable to adults (PK measure), whether adequate acute drug exposure correlates with a CNS functional response (PD measure), and whether adequate dose exposure is safe and tolerable. Early-stage pediatric trials necessitate clinical pharmacologist involvement to provide guidance on dosing strategy, PK/PD, yet use of this expertise has been limited in NIMH funded pediatric drug trials. In an effort to build this expertise, NIMH has established a pilot T32 program aimed at training the next generation of pediatric psychiatry trialists on pediatric clinical psychopharmacology approaches to trial design (16).

Based on the issues we presented regarding assay sensitivity, let’s re-examine the mavoglurant FXS trials and consider how an EM design based on solid PK/PD studies of brain effect(s) might address the issues: two double blind placebo-controlled studies were performed (1 in adults, 1 in adolescents) using three doses (25, 50 and 100 mg BID) over 12 weeks (1). How was dosing established? There is no information publicly available on whether the doses for these formulations reflect receptor occupancy (RO) although investigators appeared to have believed this data was imminent. Not until 2021, however, did relevant RO data emerge and that was only for mavoglurant showing that doses higher than those used in trials would be needed to explore the full range of RO (17). In the clinical trials, dizziness and insomnia were described at the highest dose and appear to be distinct from placebo AEs and were thus taken as evidence of brain effects presumably consistent with mechanisms in question rather than being “off target”. A few subjects were described as having agitation and visual hallucinations at that dose suggesting that dose selection was based on the upper limit of tolerability. Two Parkinson’s disease studies testing mavoglurant for treating L-Dopa induced dyskinesias, were designed to determine if slower titration (up titrating every two weeks until target dose was reached) could improve tolerability of the drug. The first study used 100 mg dose immediate release for the same trial duration (12 weeks) as the FXS trials but a second 12-week Parkinson’s disease study used modified release and increased doses to 150mg and 200 mg (18). In both trials, AEs that separated from placebo included: abnormal dreams, confusional state, and visual hallucinations, indicating tolerability was still an issue with the modified titration approach. This is an important consideration in trials: how does RO associate with side effects as well as with any postulated brain effect? An interesting case study whereby a clinical brain mediated effect (weight loss) is provided by three different cannabinoid antagonists which turned out to produce severe dysphoric states, which included suicidal ideation in a subset of patients. As argued in a study undertaken years after the trials in question, since a relatively modest degree of RO appeared needed for weight loss, had such data been available at the outset it would have been possible to answer whether or not there was a separation between the degree of RO needed for efficacy vs that associated with serious adverse psychiatric effects offering a better means of establishing a therapeutic index (19).

In the case of mavoglurant, one would explore the dose, blood concentration and brain RO relationships to establish whether the level of occupancy needed to get beneficial behavioral effects is the same as that associated with unacceptable mechanism related side effects (side effects that are so serious that from a regulatory viewpoint, they would never be an acceptable margin). With this data in hand, one would consider this a “failed drug” which should preclude this compound from further development unless response could be shown at lower RO in some other condition. If more than one mgluR5 antagonist produced similar psychiatric symptoms or other intolerable side effects at the same RO, that should be sufficient to call this a “failed target.” A further refinement using an EM approach would involve some sort of brain state characterization of subjects which possibly could identify, in the case of an undesirable CNS side effect, two subpopulations (such as with and without IEDs detectable on polysomnography) with only one showing side effects which could rescue the failed compound.

The NIMH commitment to funding studies that incorporate the most informative measures of target engagement and brain effect with PK/PD clinical pharmacological studies including pediatric populations is based on the conviction that these trial designs are essential for ruling in or out whether a particular drug mechanism has value for any particular disorder (15). Given a growing number of potential drug targets generated by the fields of genetics and neuroscience, an EM strategy for ruling in or out any mechanism among so many offers an alternative path to the ongoing practice of multiple clinical studies targeting a mechanism, often with different compounds, without generating interpretable data on the worth of the mechanism. Such a strategy depends heavily on the field of clinical psychopharmacology to incorporate state of the art PK/PD studies ideally of both the primary interaction of the drug with its target(s) in the brain and a desired downstream functional effect which, once established, can be related or not to a clinical outcome measure. There have been sufficient advances in the needed technologies to put this strategy in place for many compounds. A key factor in these studies is identifying the appropriate biomarker that reflects modulation of the molecular target, (but does not require direct links with preclinical morphological effects). Prioritizing compounds for which such tools already exist or can be developed with certainty over those many for which seductive but untestable hypotheses are generated, should be key.

List of abbreviations

CNS

Central nervous system

FXS

Fragile X Syndrome

mGluR5

metabotropic glutamate receptor 5

FMRP

Fragile X mental retardation protein

EEG

electroencephalography

IEDs

interictal epileptiform discharges

MRS

Magnetic Resonance Spectroscopy

NIMH

National institute of Mental Health

EM

Experimental medicine

PD

Pharmacodynamic

PK

Pharmacokinetic

RO

Receptor occupancy