Research ArticleFragile X Syndrome

Mavoglurant in fragile X syndrome: Results of two randomized, double-blind, placebo-controlled trials

See allHide authors and affiliations

Science Translational Medicine  13 Jan 2016:
Vol. 8, Issue 321, pp. 321ra5
DOI: 10.1126/scitranslmed.aab4109

The mGluR theory of fragile X, put to the test

People with the genetic disorder fragile X syndrome exhibit a variable constellation of debilitating physical and cognitive problems. Promising evidence from mouse models had raised hopes that an overactive glutamate signaling pathway (mGluR) was a smoking gun at the heart of the disease and that it could be successfully repaired. A pilot study in patients supported the mouse work: Down-regulation of mGluR improved behavioral problems, at least in patients carrying a certain genetic methylation marker. Here, in a larger, well-powered clinical trial, these results are put to the test and come up short. In adolescent or adult fragile X patients, whether they have the methylation marker or not, the glutamate antagonist mavoglurant had no effect on patient behavior. The authors discuss what further trials will be required, however, before permanently putting the mGluR theory of fragile X syndrome out to pasture.


Fragile X syndrome (FXS), the most common cause of inherited intellectual disability and autistic spectrum disorder, is typically caused by transcriptional silencing of the X-linked FMR1 gene. Work in animal models has described altered synaptic plasticity, a result of the up-regulation of metabotropic glutamate receptor 5 (mGluR5)–mediated signaling, as a putative downstream effect. Post hoc analysis of a randomized, placebo-controlled, crossover phase 2 trial suggested that the selective mGluR5 antagonist mavoglurant improved behavioral symptoms in FXS patients with completely methylated FMR1 genes. We present the results of two phase 2b, multicenter, randomized, double-blind, placebo-controlled, parallel-group studies of mavoglurant in FXS, designed to confirm this result in adults (n = 175, aged 18 to 45 years) and adolescents (n = 139, aged 12 to 17 years). In both trials, participants were stratified by methylation status and randomized to receive mavoglurant (25, 50, or 100 mg twice daily) or placebo over 12 weeks. Neither of the studies achieved the primary efficacy end point of improvement on behavioral symptoms measured by the Aberrant Behavior Checklist—Community Edition using the FXS-specific algorithm (ABC-CFX) after 12 weeks of treatment with mavoglurant. The safety and tolerability profile of mavoglurant was as previously described, with few adverse events. Therefore, under the conditions of our study, we could not confirm the mGluR theory of FXS nor the ability of the methylation state of the FMR1 promoter to predict mavoglurant efficacy. Preclinical results suggest that future clinical trials might profitably explore initiating treatment in a younger population with longer treatment duration and longer placebo run-ins and identifying new markers to better assess behavioral and cognitive benefits.


Fragile X syndrome (FXS) is an X-linked genetic syndrome that results in a spectrum of intellectual disabilities, as well as physical and behavioral characteristics. The major symptoms of FXS include moderate to severe intellectual disability, attention deficit and hyperactivity, anxiety with mood lability, and, in some cases, aggression, as well as perseverative and autistic behaviors (1). With a prevalence estimated at 1 in 4000 males and 1 in 8000 females, it is generally regarded as the most common inherited cause of intellectual disability and autistic spectrum disorder, and second only to Down syndrome as a disease of intellectual disability with known genetic cause (24). Females tend to be less severely affected and may not show intellectual disability, a result of compensation by the normal X chromosome (5). Currently, there are no symptomatic or disease-modifying treatments in FXS that have received regulatory approval.

FXS is typically caused by a trinucleotide repeat (CGG) expansion of >200 repeats (full mutation) in the promoter of the X-linked FMR1 (fragile X mental retardation 1) gene (6). This mutation is associated with complete or partial methylation at the FMR1 promoter, resulting in the loss or significant reduction of expression of the gene product FMRP (fragile X mental retardation protein) (7). FMRP is an RNA binding protein that modulates the dendritic localization and translation of several hundred RNA ligands (8). Its absence alters dendritic morphology and synaptic plasticity and is associated with abnormalities of long-term depression (LTD) and long-term potentiation (911).

In particular, preclinical results indicated that metabotropic glutamate receptor (mGluR)–dependent LTD signaling of translation is enhanced in the hippocampus of Fmr1 knockout (KO) mice lacking FMRP (10). This led to the “mGluR theory of FXS,” which proposes that the absence of FMRP can cause overactivation of mGluR signaling, leading to enhanced hippocampal LTD and contributing to the features of the FXS phenotype (12).

Selective antagonists of mGluR5, a group I mGluR, have been evaluated in a number of preclinical and clinical studies. Mavoglurant (AFQ056), a structurally novel, noncompetitive mGluR5 inhibitor rescues molecular, neuronal spine, and behavioral phenotypes in the mouse model of FXS (Fmr1 KO mice) (1315). The efficacy and safety of mavoglurant were further evaluated in a phase 2, randomized, double-blind, placebo-controlled, crossover design trial in 30 adult male FXS patients treated for 28 days (NCT00718341). The patients were classified on the basis of the level of methylation detected at the FMR1 promoter. They were considered completely methylated (CM) if only methylated DNA was detected and partially methylated (PM) if both methylated and unmethylated DNA were detected. Although no significant treatment differences between mavoglurant and placebo groups were found in the overall study population, results of a post hoc analysis suggested improvement in maladaptive behavior on the Aberrant Behavior Checklist—Community Edition (ABC-C) score, Clinical Global Impression (CGI)—Improvement scale (CGI-I), and the Repetitive Behavior Scale (RBS) in the CM subgroup of patients (16).

These results provided the rationale for our phase 2b, multinational, double-blind, placebo-controlled, and parallel-group 3-month trial evaluating the effects of multiple doses of mavoglurant in adult patients with FXS stratified by methylation status (NCT01253629). This was followed by a similar trial conducted in adolescents (NCT01357239), since preclinical evidence, as well as outcomes from early intervention studies in other genetic disorders with impaired neurocognitive development, indicated that the younger the patient, the greater the potential benefit of the therapeutic intervention (17). In addition, selective mGluR5 inhibitors have shown greater effect in younger animal models (18).

The results of the two double-blind, controlled trials conducted in adults (NCT01253629) and in adolescents (NCT01357239) are presented here. The aim of both studies was to assess the efficacy and safety of three doses [25 mg, 50 mg, and 100 mg twice daily (bid)] of mavoglurant versus placebo in reducing maladaptive behavior after 12 weeks of treatment in patients with FXS as measured by the ABC-C total score [using the FXS-specific algorithm (ABC-CFX)]. The patients were divided into two strata depending on the extent of methylation of their FMR1 gene, CM and PM. The efficacy of mavoglurant was also assessed with a variety of other behavioral measures, including CGI-I, individual subscales of the ABC-CFX, the RBS-revised (RBS-R) total score, and the Social Responsiveness Scale (SRS). A few cognitive parameters were also assessed.


Demographic and background characteristics

A total of 175 adult and 139 adolescent patients were randomized into the four treatment arms (Figs. 1 and 2) of each trial, respectively. Overall, 162 patients completed the adult study and 135 patients completed the adolescent study with a total of 13 patients discontinuing because of adverse events (AEs), including 11 in the adult study and 2 in the adolescent study (Fig. 2, A and B). A protocol amendment was made in the adolescent study, setting the primary objective to assess the efficacy and safety of the highest dose of mavoglurant (100 mg bid) compared with placebo. Because all the PM patients, but not the CM patients, were already recruited, this resulted in an imbalance in the CM stratum toward more patients randomized to 100 mg bid mavoglurant compared to the PM stratum. However, the number of patients in total (“all patients”; that is, both strata) and in each methylation stratum (CM and PM) is well balanced in the placebo and 100 mg bid mavoglurant treatment groups.

Fig. 1. Study design.
Fig. 2. The flow of patients through the studies (CONSORT diagram).

In each study, patients’ demographic and background characteristics were generally comparable across the treatment groups within the combined patients group, as well as within the two strata (CM and PM) (Table 1). The mean age ranged from 24.2 to 26.9 years in the adult population and 14.4 to 14.6 years in the adolescent population across the treatment groups (Table 1). In both trials, most of the randomized patients were Caucasian (>87%), which is expected given the racial distribution of the populations covered by the investigational sites. The very high proportion of men (>83%) is a result of the reduced penetrance and expression of fragile X syndrome in women, due to presence of a normal FMR1 gene on the unaffected X chromosome. In the adult study, the baseline ABC-CFX total score in the mavoglurant 50 mg bid group was higher than that in the other treatment groups (Table 1).

Table 1. Demographics and other baseline characteristics, by treatment (randomized set), in all patients.

Data are presented as mean (SD) unless specified otherwise. Percentage (%) is calculated on the basis of patients in the randomized set within the specified stratum. CGI-S, CGI—Severity.

View this table:

ABC-CFX total score

Neither study met the primary or the key secondary objective of showing efficacy in reducing the ABC-CFX total score after 12 weeks of treatment with any of the three doses of mavoglurant versus placebo in either CM or PM patients (Fig. 3, A to D).

Fig. 3. ABC-CFX least square (LS) mean change from baseline (after placebo run-in).

(A) CM stratum in adult population. (B) PM stratum in adult population. (C) CM stratum in adolescent population. (D) PM stratum in adolescent population. *, statistical significance at 5% level. (A) 50 mg bid, week 12: P = 0.018; (C) 100 mg bid, week 2: P = 0.022; week 4: P = 0.013; week 8: P = 0.002; week 12: P = 0.004; 25 mg bid, week 2: P = 0.003. Data presented as LS means (±SEM).

CM stratum. In adult CM patients, the average improvement in ABC-CFX total score from baseline to week 12 was similar in the placebo group (−11.4) and in the mavoglurant 25 mg bid treatment group (−14.3), whereas there was almost no change from baseline in the mavoglurant 50 mg bid (1.8) and mavoglurant 100 mg bid (−1.8) treatment groups (Fig. 3A). A small but statistically significant deterioration by 1.8 points in ABC-CFX total score was seen in the CM mavoglurant 50 mg bid treatment group compared with the placebo group (P = 0.018).

In the CM stratum of the adolescent population, there was also an improvement in the ABC-CFX total score with placebo (−9.4) and mavoglurant 25 mg bid (−11.8), and to a lesser extent with mavoglurant 50 mg bid (−3.4) at week 12 (Fig. 3C). These improvements with mavoglurant 25 and 50 mg bid were not significantly different from placebo. However, a deterioration was observed after multiplicity adjustment in the 100 mg bid treatment group (8.6), which was statistically significant compared to placebo.

PM stratum. In both studies, the results in the PM population were comparable to those observed in the CM population. In the adult population of PM patients, improvements from baseline to week 12 in the ABC-CFX total score were observed in all arms, with the largest improvement shown in the placebo arm (−8.9 points), but the difference was not significant compared with the mavoglurant dose groups (25 mg bid, −1.9; 50 mg bid, −5.1; and 100 mg bid, −4.6) (Fig. 3B).

In the adolescent PM stratum, all arms also showed improvements in ABC-CFX total score, with the greatest improvement observed in the mavoglurant 25 mg bid arm (−6.8) (Fig. 3D). As per the adult population, changes from baseline after 12 weeks for all the three mavoglurant treatment groups were not significantly different from placebo (placebo, −3.5; mavoglurant 50 mg bid, −2.8; and mavoglurant 100 mg bid, −5.7).

The overall improvements in the ABC-CFX total score for the pooled (all patients) group in the adult and adolescent populations compared to placebo are given in fig. S1, A and B, respectively.

Placebo run-in. Improvements were also observed during the 4-week placebo run-in period in all arms for both stratums and in both studies (fig. S2, A to D). In the adult study, they reached maximum scores of −12.9 and −12.1 (CM stratum) and −15.6 and −11.2 points (PM stratum) in the groups of patients later randomized to placebo and mavoglurant 100 mg bid, respectively (fig. S2, A and B). In the adolescent study, the maximum improvement (−27.3) was observed in CM patients later randomized to mavoglurant 100 mg bid group. When compared with their pre-placebo run-in ABC-CFX scores, all the adolescent treatment groups had improved at week 12 (fig. S2C). Improvements during the 4-week placebo run-in period were also observed for these adolescent patients in the PM stratum, with a maximum of −12.5 and −18.9 observed in the placebo and 50 mg bid groups, respectively (fig. S2D).

Other secondary outcomes

Behavioral measures. Efficacy of any of the three doses of mavoglurant versus placebo on the CGI-I, ABC-CFX subscale scores, RBS-R total score and subscale scores, and SRS scores after 12 weeks of treatment was not demonstrated in either stratum (CM or FM) in either population (adult or adolescent). Similarly to total scores, change from baseline to week 12 in the ABC-CFX subscale scores for the mavoglurant treatment groups was similar to placebo in the pooled (all patients) group (Table 2), as well as in both methylation strata on nearly all subscales (table S1). In the adult population, the comparison of each mavoglurant dose with placebo in the all-patients group reached statistical significance in favor of placebo in the 100 mg bid mavoglurant–treated subgroup for inappropriate speech and in the 50 mg bid mavoglurant–treated subgroup for irritability (Table 2). In the adolescent population, comparisons of the mavoglurant 100 mg bid and placebo treatment groups showed significance in favor of placebo in the CM stratum for the following subscales: irritability, lethargy/withdrawal, stereotypic behavior, and inappropriate speech (table S1). However, all the changes were numerically small.

Table 2. Change from baseline to week 12 for ABC-CFX subscale scores in all patients.

Data are presented as LS mean (SE) unless specified otherwise.

View this table:

At week 12, the CGI-I mean scores for the mavoglurant treatment groups were similar to the placebo group for all patients in both the CM and PM strata, in both the adult and the adolescent populations (Table 3). Any observed differences between treatment groups and placebo were not significant.

Table 3. CGI-I score at week 12 by treatment in CM, PM, and all patients.

PBO, placebo; CI, confidence interval; Mav, mavoglurant.

View this table:

Similarly, the changes from baseline to week 12 in RBS-R and SRS total scores did not show any statistically significant improvement when the three mavoglurant treatment groups were compared with placebo in the pooled (all patients) group (tables S2 and S3, respectively) or in either strata (CM and PM), regardless of the population studied.

Cognitive measures. Results from the computerized cognitive test battery CNS Vital Signs (CNS VS), conducted in the adult population for the key domains, emotion, recognition, and reaction time (table S4), were difficult to interpret. The same statement can be made for the results of the three Test of Everyday Attention in Childhood (TEA-Ch) battery subtests conducted in the adolescent population (table S5). The small number of patients able to perform these two tests suggested that the components of the test battery may have been too challenging for many participants, likely a result of low cognitive functioning and behavior disturbances. Therefore, no meaningful conclusions on the effects of mavoglurant on cognitive function could be made in either of the populations studied.

Effect of age. In a post hoc analysis, the potential impact of age on the efficacy of mavoglurant was evaluated by pooling data from the two study populations (that is, adults and adolescents). ABC-CFX total scores did not change significantly from baseline for any age group (Fig. 4) or in an exploratory analysis when age was considered as a continuous variable. Relatively younger age (adolescent population) did not predict better efficacy for mavoglurant.

Fig. 4. Effect of age on ABC-CFX total score LS mean change from baseline to week 12.

Data presented as LS means (±SE).

Effect of methylation. Finally, another post hoc analysis was conducted in the adult population to assess the ABC-CFX total scores as a function of methylation levels (as a continuous variable). The level of methylation was found not to be correlated to the behavioral response to the drug for any of the doses used (fig. S3).

Safety results

The incidence of any AEs experienced by patients during the double-blind treatment period was highest in the mavoglurant 100 mg bid treatment group (adults, 82.2%; adolescents, 87.2%). In the adult population, this was followed by the mavoglurant 50 mg bid, placebo, and mavoglurant 25 mg bid treatment arms (69.0%, 61.4%, and 56.8%, respectively). In the adolescent study, the mavoglurant 25 mg bid treatment group experienced more AEs (80.6%) than the placebo (61.9%) and mavoglurant 50 mg bid (51.9%) treatment groups. Most reported AEs were mild in severity. Incident rates of AEs were similar in both methylation (CM and PM) strata. AEs during the double-blind treatment period are shown in Table 4 (and table S6). In both studies, the four most commonly experienced AEs by primary system organ class were psychiatric disorders, infections and infestations, gastrointestinal disorders, and nervous system disorders. Dizziness and insomnia were the most frequently reported AEs in the mavoglurant 100 mg bid treatment arm (17.8% each) of the adult population, whereas in the mavoglurant highest dose arm of the adolescent population, headache was reported most frequently (25.6%), followed by insomnia (15.4%) and vomiting (15.4%).

Table 4. AEs from baseline to week 12 by treatment (safety population).
View this table:

Four serious AEs (SAEs) were reported in the adult population. One patient on placebo experienced lobular pneumonia; one patient experienced upper respiratory tract infection, headache, and agitation (mavoglurant 50 mg bid treatment group); one patient reported agitation, visual hallucination, and insomnia and another reported agitation (both from the mavoglurant 100 mg bid group). In the adolescent group, two patients experienced SAEs: One patient on placebo experienced convulsion after an accidental overdose of clonidine given for insomnia, whereas another experienced appendicitis leading to hospitalization.

In the adult population, discontinuations because of AEs were higher in the mavoglurant 100 mg bid treatment group (13.3%) compared to the 50 mg bid group (4.8%) and the 25 mg bid group (4.5%), versus 2.3% in the placebo arm. Psychiatric disorders were the most common AEs leading to discontinuation, with 11.1% discontinuing in the mavoglurant 100 mg bid treatment group, mostly because of insomnia (11.1%) and agitation (4.4%), compared to none on placebo. There were only two discontinuations because of AEs in the adolescent population. One patient in the placebo group with a history of seventh nerve paralysis developed this condition and was given prednisone (a medication prohibited within the study). One patient in the 100 mg bid group experienced hyperhidrosis, inappropriate affect, insomnia, psychomotor hyperactivity, and tachycardia.

On the basis of the number of discontinuations because of AEs, the number of patients with SAEs, as well as the profile of AEs in each stratum, we conclude that there were no clinically meaningful differences among groups in either study. The 100 mg dose was, however, more tolerated in the adolescent population than in the adult population, in which it led to a higher occurrence of AEs.

There were also no clinically relevant changes in vital signs, weight, electrocardiograms (ECGs), or laboratory test results in any of the four treatment groups of each study population. Overall, changes in the laboratory parameters intended to assess activation of the HPA (hypothalamic-pituitary-adrenocortical) axis were small across all treatment groups, and there were no clinically meaningful differences between treatment groups. Large fluctuations in oxytocin levels observed in a small number of adult and adolescent patients are not unexpected because emotional events can affect oxytocin levels. Fluctuations in prolactin levels, also observed in some patients in each population, were not clinically meaningful and may be due to the concomitant use of antipsychotics, which are known to affect prolactin levels. The changes from baseline in the total Neuropsychiatric Inventory—Questionnaire (NPI-Q) scores measured in the adolescent population were small. No deaths occurred in either study.


We performed two well-powered, multinational, double-blind, placebo-controlled studies to assess the safety and efficacy of mavoglurant, a selective mGluR5 antagonist, for treating behavioral symptoms characteristic of adult and adolescent patients with FXS. The studies were adequately powered to detect an effect size of 1.15 (corresponding to a treatment difference of 17 points in ABC-CFX) for the primary end point. Contrary to what was predicted by a decade of studies in animal models, we did not find any evidence to support efficacy of mavoglurant (25 mg, 50 mg, or 100 mg bid) versus placebo on any of the behavioral outcome measures used: ABC-CFX total score and subscale scores, CGI-I, RBS-R total score, and SRS total score.

Comparisons of each mavoglurant dose with placebo reached statistical significance in favor of placebo in a few ABC-CFX total score and subscale measures. However, the differences were small and not consistent across scales and study populations, and may be due to chance, especially as we made no adjustment for multiplicity in analyzing multiple ABC-CFX subscales. Results of the cognitive measures proved inconclusive because of the small number of subjects who were able to perform the task.

A previous phase 2 study suggested that the extent of methylation of the FMR1 gene promoter region may predict response to treatment with mavoglurant in patients with FXS (16). However, in our studies, methylation status did not demonstrate utility in predicting response to mavoglurant with any of the measures used. A post hoc analysis showing the ABC-CFX total scores as a function of methylation levels (as a continuous variable) confirmed that the level of methylation was not correlated to the behavioral response to the drug, regardless of the dose. Because of the very small number of female patients, it was not possible to draw any meaningful conclusions about the effect of gender on the efficacy of mavoglurant.

Together, our results suggest that mavoglurant is not likely to be effective for behavior in FXS patients, irrespective of methylation status. Overall, the safety and tolerability profile of mavoglurant was similar to that seen in earlier studies. Mavoglurant was associated with a dose-related increase in AEs, but few subjects discontinued participation as a result of AEs. The observed dose-dependent central nervous system (CNS) AEs (insomnia, agitation, and hallucinations) were expected on the basis of previous studies.

One of the most plausible explanations for the negative results is that the mGluR theory on which the studies were based may not be valid or may manifest itself differently in humans compared to rodents. Recent studies suggest that, although some proteins are overexpressed in the absence of FMRP (which could account for the increase in mGluR5 activation–related LTD), others appear to be underexpressed or misexpressed (12). If aspects of FXS in humans are attributable to such a decreased translation of a pool of proteins in response to mGluR5 activation, mGluR antagonists would not likely be effective. In addition, the mGluR5 pathway is only one pathway disrupted in FXS; others are dysregulated in the absence of FMRP (19). A combination of targeted treatments may be needed to optimally treat FXS.

It is possible that the age of the patients in our studies was not optimal. Our analysis did not reveal an influence of age on mavoglurant treatment effects even when the adult and adolescent populations were combined in a post hoc analysis. However, preclinical evidence from Fmr1 KO mice treated with selective mGluR5 inhibitors suggested a greater treatment effect in younger animals. For instance, 2-methyl-6-(phenylethynyl)pyridine (MPEP), a short-acting mGluR5 inhibitor, reduced average neuronal spine length and density at early developmental ages but was less effective in adult mice (20). Our youngest patients were likely equivalent to a young adult age group in mouse studies, animals that responded less to mGluR5 inhibition; in future clinical trials, it would be desirable to include patients even younger than those in the studies reported here to fully test the effect of age on responsiveness to mGluR5 inhibitor treatment.

Finally, the duration of treatment may be another factor that was not fully explored by our trial design. A study conducted with 2-chloro-4-((2,5-dimethyl-1-(4-(trifluoromethoxy)phenyl)-1H-imidazol-4-yl)ethynyl)pyridine, a selective long-acting mGluR5 inhibitor with a greater potency than MPEP, indicates that very long-term treatment (potentially equivalent to multiple years in a human) can produce incremental improvements, in some outcome measures, with some restoration of spine density (18). It would be informative to use a study duration longer than the 12-week treatment used here in future clinical trials.

The outcomes reported here raise a number of issues that should be considered in studies in FXS or similar neurodevelopmental disorders. First, substantial improvements in ABC-CFX total scores were observed during the placebo run-in in all groups, indicating that this measure is subject to a strong placebo effect (which may be amplified by family and caregiver involvement in rating). Improvements in ABC-CFX scores were also observed in both placebo- and drug-treated individuals, suggesting that a longer follow-up of patients might be required to capture any effect of treatment compared with placebo.

The selected outcome measures may also not cover all the FXS symptomatic domains. The ABC was chosen largely because there was a regulatory precedent in the use of this measure for approval of treatments for autism, and because the ABC-C scale had been refactored to ABC-CFX by selecting items and subscales to focus on behavioral domains commonly associated with FXS. The hope was that behavior would be an adequate surrogate for the improvements in synaptic plasticity and disease pathophysiology seen in the mouse. However, as suggested by preclinical data, part of the drug effect may actually be in the domain of learning or cognition. In this respect, better predictors of treatment response might include biological indicators of pathway function such as S6 kinase (21), the mRNA 5′ cap–binding protein eukaryotic initiation factor 4E (eIF4E) (22), and matrix metalloproteinase 9 (MMP9) (23) in the periphery, but also markers of central response, such as eye tracking, prepulse inhibition (PPI), or event-related potentials (ERPs) (24). ERP, in particular, which measures changes in information processing of the brain, has shown utility as a marker of CNS function that demonstrated significant improvement in a controlled trial of minocycline in FXS (24). Such markers should be used in the future to help establish whether a drug has an effect on neuronal functioning that could lead to positive cognitive outcomes.

The results of the studies presented here highlight the difficulty of translating information on molecular mechanisms identified in animal models to humans. FXS is the first developmental disorder for which there has been such a large amount of information gathered on pathways in animal models. Although animals are much easier to manipulate, it is nevertheless difficult to relate animal age to human age accurately and even more difficult to gather information on synaptic plasticity and CNS mechanisms in humans. Animal models may have more innate plasticity than humans, allowing for correction of phenotypes more easily at relatively later periods of life. Challenges to translation of results in model organisms to humans for FXS thus include uncertainties around optimal patient selection, age of treatment onset, dosages and durations of treatment, differences in pharmacokinetics and pharmacodynamics, dose-limiting side effects, and biomarkers of CNS improvement. The most challenging issue for translational medicine in general, however, is the a priori identification of the species-specific outcome measure that is most relevant to a patient treatment response.

If the understanding of targeted treatment effects in FXS is to progress, and if the mGluR theory of FXS is to be fully tested, it will be necessary to design new trial paradigms to investigate effects of mGluR agents in young children and to incorporate measures of learning into the protocol. It will also be critical to validate new quantitative and objective measures of behavioral and cognitive performance that are better adapted to assess function and treatment response in patients with moderate to severe intellectual disability.



The participants were male and female patients aged 18 to 45 years (adult study) and aged 12 to 17 years (adolescent study) who had a diagnosis of FXS that had been confirmed by genetic testing. Patients for whom informed consent was obtained from a legal guardian or legally acceptable representative and who met the inclusion criteria were enrolled in the trial.

Patients were eligible to participate if they had a CGI-S score of ≥4 (moderately ill), an ABC-C score of >20, and an IQ score during screening that was lower than 2 SDs below the IQ test median. They also had to have a caregiver(s) who spent at least 6 hours per day with them to supervise treatment and assess outcomes. All caregivers watched a training video explaining the ABC-C. In addition, female patients must have been following an acceptable method of birth control according to the protocol.

The major exclusion criteria were unacceptable laboratory values according to the protocol; past medical history of clinically significant ECG abnormalities; history and/or presence of schizophrenia, bipolar disease, psychosis, confusional states, and/or repeated hallucinations; and history of uncontrolled seizure disorder within the past 2 years. Patients concomitantly treated with more than two psychoactive medications, excluding antiepileptics, or who used any potent inhibitors or inducers of CYP3A4 or glutamatergic agents within 6 weeks of randomization, were also excluded from the study.

Study design

These studies were two phase 2b, randomized, double-blind, placebo-controlled, parallel-group trials of mavoglurant in adult (NCT01253629) and adolescent patients (NCT01357239) with FXS. The adult study was conducted at 31 centers across 10 countries including Australia, Canada, Denmark, France, Germany, Italy, Spain, Switzerland, UK, and the United States. The adolescent study was conducted at 38 centers across 16 countries, which, in addition to the countries listed above, included Belgium, Indonesia, Israel, Netherlands, Sweden, and Turkey. The adult study was initiated in November 2010 and completed in August 2013, whereas the adolescent study was initiated in May 2011 and completed in January 2014.

After a 4-week, single-blind placebo run-in period, eligible patients were randomized in a ratio of 1:1:1:1 using a centralized Interactive Response Technology system (IRT) to one of the following four treatment arms: mavoglurant 25, 50, and 100 mg bid or placebo over a 12-week double-blind period (Fig. 1). Patients allocated to the mavoglurant arm initiated therapy with 25 mg bid, and the dose was up-titrated in weekly increments as required until the target dose was reached. Down-titration was not allowed; patients unable to tolerate their assigned dose were discontinued.

Subjects who completed the 12-week double-blind studies or who discontinued because of intolerability to their assigned dose were then eligible for enrollment in two open-label extension studies in adult (NCT01348087) or adolescent (NCT01433354) patients with FXS. These two long-term studies are now completed, and the results will be published separately.

Both study populations were stratified by gender, methylation status, and region during randomization via the IRT. Gender and region stratifications were only used to ensure balanced distribution across treatment groups. Sites in the United States, Canada, and Australia, where a large proportion of patients are treated with antipsychotics, stimulants, or selective serotonin reuptake inhibitors, were assigned to the first region, whereas sites in European countries were assigned to the second region. The stratification into CM and PM was considered necessary because of the differential efficacy observed in the two populations in the previous study (NCT00718341). Methylation status of the patient’s FMR1 gene was determined via a blood sample collected at the screening visit.

Each study enrolled approximately equal numbers of CM and PM patients to maintain statistical power. The methylation ratio between PM and CM strata was higher than the expected 2:1, which resulted in more patients being screened to fill the CM stratum. In the adolescent study, because of the challenges associated with recruiting adolescent patients with FXS, the protocol was modified to focus on the evaluation of the efficacy and safety of the highest dose of mavoglurant versus placebo. Patients were thus subsequently assigned to two treatment arms, 100 mg bid mavoglurant or placebo bid, in a ratio of 1:1. The dose of patients randomized before the amendment remained unchanged.

Efficacy assessments

The primary outcome of each study was the change from baseline in behavioral symptoms of FXS as assessed by the ABC-CFX total score after 12 weeks of treatment in FXS patients with CM FMR1 gene (CM stratum). The key secondary outcome was the change from baseline in behavioral symptoms of FXS as assessed by the ABC-CFX total score after 12 weeks of treatment in FXS patients with PM FMR1 gene (PM stratum).

The other secondary outcomes were applied to both CM and PM strata. These included the safety and tolerability of mavoglurant, as well as a number of other efficacy assessments including the following: the global improvement of symptoms in FXS patients using the CGI-I scale, the change from baseline in the six individual subscales of the ABC-CFX, and the improvement of repetitive behavior as measured by changes in the RBS-R total scores. All secondary outcomes were also assessed after 12 weeks of treatment. In addition, exploratory efficacy assessments at selected sites included effect on social interactions, as measured by changes in the SRS-A (in adults) or SRS (in adolescents) score after 12 weeks of treatment, as well as effects on cognition as assessed by the CNS VS cognitive battery in adults and the TEA-Ch battery in adolescents.

The ABC-C and CGI-I were performed at weeks 2, 4, 8, and 12. In addition, ABC-C was performed at weeks −8 (screening), −4 (start of run-in period), −2, and 0 (baseline). RBS-R, SRS-A, and SRS were performed at weeks −4, 0 (baseline), 4, and 12. The cognitive test battery was conducted in adults at weeks 0 (baseline), 4, and 12 and only at a limited number of sites (n = 3) in the United States. TEA-Ch battery was conducted in adolescents at baseline and week 12.


The ABC-C is a caregiver-rated symptom checklist for assessing problem behaviors via a 58-item and 5-subscale questionnaire. Each item is attributed a score from 0 (“not at all a problem”) to 3 (“problem is severe in degree”) and the total score ranks from 0 to 174 (25, 26). The ABC-C should have been completed, as far as possible, by the same rater throughout the study for each subject. The raw data collected from the full 58-item ABC-C were analyzed according to the modified ABC-CFX (27). In this modified version, 55 items and 6 subscales (irritability, lethargy/withdrawal, stereotypic behavior, hyperactivity, inappropriate speech, and social avoidance) are considered, and the total score ranks from 0 to 165. A negative change from baseline indicates improvement.


The CGI scale, used to assess treatment response in psychiatric patients, is divided into two parts, with one part (CGI-S) assessing the severity of illness and the other part (CGI-I) assessing improvement (28). The CGI-I is a clinician-rated scale that reports the global changes of the symptoms rated on a seven-point scale (with 1 being “very much improved,” 4 being “no change,” to 7 being “very much worse”) (28).


The RBS-R is a caregiver-rated 43-item questionnaire that captures the breadth of repetitive behavior across six domains (ritualistic behavior, sameness behavior, stereotypic behavior, self-injurious behavior, compulsive behavior, and restricted interests), with every behavior rated from 0 (behavior does not occur) to 3 (behavior is a severe problem) and a total score ranking from 0 to 129. A negative change from baseline indicates improvement (29).


The SRS (and the adult version SRS-A) is a 65-item questionnaire filled by the caregiver, which aims at identifying the presence and extent of autistic social impairment. Every behavior is rated from 0 (not true) to 3 (almost always true), and the total score ranges from 0 to 195. A negative change from baseline indicates improvement (30).


The CNS VS is a computerized testing battery focused on working memory; the following four modules were used: Visual Memory Test (VIM), Symbol-Digit Coding (SDC), Perception of Emotions Test (POET), and three-part Continuous Performance Test (CPT) [Working Memory Test (WMT)] (31).


The TEA-Ch (32) battery was used to test for every day attention and inhibition. The battery is primarily a clinical tool to identify children with possible impairments in attention and has been used on children with different neurogenetic disorders including FXS and Down syndrome. Only three subtests from the battery were used: Map Mission, Walk/Don’t Walk, and Opposite Worlds.

Safety assessments

Safety assessments consisted of collecting all AEs, SAEs, with their severity and relationship to study drug, and pregnancies. Vital signs, physical examination, body height and weight, as well as ECG were regularly assessed, and monitoring of hematology (including assessment of coagulopathy), blood chemistry, and urine was performed at the central laboratory. The self-administered questionnaire NPI-Q (33) assessed potential neuropsychiatric events in adolescent patients. Because of the well-established involvement of excitatory amino acids including glutamate in the control of adrenocorticotropic hormone release (stimulatory effect), nonstandard clinical evaluations included endocrinologic investigations to assess the activation of the HPA axis. The assessment of safety was based primarily on the frequency of AEs, SAEs, and laboratory abnormalities.

Statistical analysis

For each study, a sample size of 20 patients per arm per stratum was estimated to provide at least 80% power to detect an effect size of 1.15 for the pairwise comparisons of each mavoglurant dose to placebo in the CM patients, assuming a dropout rate of 10% at a significance level of 1.42% (two-sided). The effect size of 1.15 corresponded to a treatment difference of 17 points in ABC-CFX total score between mavoglurant and placebo with an SD of 15 [based on results from the previous study of 30 adult male patients with FXS (NCT00718341)]. A similar calculation estimated that 20 patients per arm would also be sufficient to achieve 80% power to detect the effect size of 1.4 at the significant level of 0.25% (two-sided) in the key secondary analysis.

The null hypothesis, no difference in the change from baseline to week 12 in ABC-CFX total score between any mavoglurant dose and placebo, was tested among CM patients for the primary objective and among PM patients for the key secondary objective. The primary and key secondary analyses, which included change from baseline in ABC-CFX total score, were based on a mixed-effects model for repeated-measures (MMRM) including region, treatment, week, treatment by week interaction as fixed effects, and baseline ABC-CFX total score as covariate, with an unstructured covariance structure. From the model, the contrast between each mavoglurant dose versus placebo after 12 weeks of treatment was estimated and presented together with a two-sided 95% CI and P value. As agreed with health authorities, a multiplicity adjustment was applied to the primary and the key secondary analysis to control the overall type I error rate, leading to a required level of significance smaller than the ordinal significant level of 0.05.

All secondary efficacy analyses were performed separately in stratum I (CM), stratum II (PM), and overall (combined strata I + II), and comparisons were made between each mavoglurant dose group and placebo at the two-sided 5% type I error rate with no adjustment for multiplicity. The CGI-I score and the RBS-R total score were analyzed using the similar MMRM model as the primary analysis.


Fig. S1. ABC-CFX LS mean change from baseline in all patients (CM and PM) after placebo run-in.

Fig. S2. ABC-CFX mean change from baseline including placebo run-in.

Fig. S3. Change in ABC-CFX versus methylation levels at screening (all patients).

Table S1. Change from baseline to week 12 for ABC-CFX subscale scores.

Table S2. RBS-R total score at baseline and week 12 and change from baseline to week 12 in all patients (CM and PM).

Table S3. SRS total score at baseline and week 12, and change from baseline to week 12 in all patients (CM and PM).

Table S4. Number and percentage of patients scoring better on each module of the CNS VS than results achieved by chance (probability) alone in all patients (CM and PM) adult study.

Table S5. TEA-Ch change from baseline in raw scores for the three subsets in all patients (CM and PM) adolescent study.

Table S6. AEs (≥5% in any group) from baseline to week 12 by treatment (safety population).

Table S7. Mavoglurant principal investigators (ordered alphabetically by country).


  1. Acknowledgments: We thank M.-C. Mousseau and C. Kelly (Novartis Ireland Limited, Dublin, Ireland), as well as H. K. Mittal (Novartis Healthcare Private Limited, Hyderabad, India) for medical writing and editorial assistance in the preparation of this manuscript. Writing support was funded by the studies’ sponsor. We also thank patients, investigators, and site staff who participated in these studies, including the mavoglurant principal investigators (table S7). Funding: This study was supported by Novartis Pharma AG, Basel, Switzerland. Author contributions: E.B.-K., R.H., V.D.P., B.K., G.M.B., L.Z., T.J., G.A., and F.v.R. participated in the study design. E.B.-K., R.H., V.D.P., P.C., J.V., M.B., K.R., and G.M.B. performed the experiments. L.Z. analyzed the data, and all authors contributed to data interpretation. B.K., E.B.-K., R.H., S.J., and V.D.P. helped draft the report. All authors critically revised the manuscript and approved the final version of the report. Competing interests: S.J., R.H., and E.B.-K. have served on the Novartis Fragile X Advisory Board and have consulted for Novartis. R.H. has also consulted for Roche on FXS. M.B., K.R., B.K., L.Z., T.J., G.A., and F.v.R. are employees of Novartis. All of the other authors declare that they have no competing interests. Novartis has a patent on the drug tested in this mavoglurant study. The clinical trials are registered in with the identifiers NCT01253629 and NCT01357239. Data and materials availability: The reader can obtain the raw data (anonymized) by connecting to and signing a data sharing agreement with Novartis.
View Abstract

Stay Connected to Science Translational Medicine

Navigate This Article