Riluzole, a Glutamate Modulator, Slows Cerebral Glucose Metabolism Decline in Patients With Alzheimer's Disease

Dawn C. Matthews; Xiangling Mao; Kathleen Dowd; Diamanto Tsakanikas; Caroline S. Jiang; Caroline Meuser; Randolph D. Andrews; Ana S. Lukic; Jihyun Lee; Nicholas Hampilos; Neeva Shafiian; Mary Sano; P. David Mozley; Howard Fillit; Bruce S. McEwen; Dikoma C. Shungu; Ana C. Pereira


Brain. 2022;144(12):3742-3755. 

In This Article

Materials and Methods

Study Design and Participants

Patients with a clinical diagnosis of probable Alzheimer's disease based on neurological and neuropsychological evaluation (National Institute on Ageing-Alzheimer's Disease Association, NINCDS-ADRDA criteria),[45,46] Mini-Mental State Examination (MMSE) score of 19 to 27, and 50 to 95 years were enrolled in this pilot phase 2 double-blind, randomized, placebo-controlled study. For inclusion, FDG-PET baseline scans were also evaluated to confirm a lack of a frontotemporal dementia or Lewy body disease pattern of hypometabolism. All participants were stable on acetylcholinesterase (AChE) inhibitors for at least 2 months before starting the trial and continued to take AChE throughout the study with the exception of one participant who had never been on AChE therapy. The study was conducted at two sites (Rockefeller University Hospital and Icahn School of Medicine at Mount Sinai, both in New York City), with the approval of the Institutional Review Boards (IRB) of both Institutions. All neuroimaging was performed at Citigroup Biomedical Imaging at Weill Cornell Medicine under an IRB protocol separately approved by that Institution. Memantine, which acts on the glutamatergic system through a different mechanism than riluzole, was not allowed for 6 weeks before study entrance nor over the study duration. Other exclusion criteria were: abnormal liver function [>2 times the upper limit of normal for alanine aminotransferase (ALT) or aspartate aminotransferase (AST); or bilirubin >1.5 times the upper limit of normal, positive hepatitis serology (Hep. B antigen+ or Hep. C antibody+)], uncontrolled diabetes mellitus (Hba1c > 7), chronically uncontrolled hypertension, MRI contraindication, history of brain disease, current smoker or user of nicotine-containing products, currently taking medications with evidence of glutamatergic activity or effects on brain glutamate levels such as lamotrigine, lithium, opiates, psychostimulants such as amphetamines and methylphenidate, tricyclic antidepressants, benzodiazepines and any other drug that the investigators judged might interfere with the study (participants on those medications could be included in the study but without MRS measurements) and others (full criteria at NCT01703117).

Participants were randomly assigned in a double-blind fashion to receive riluzole at a dose of 50 mg twice a day or placebo for 6 months, with age-matched cohorts of 50–74 and 75–95 years old. Written informed consent was obtained from participants or their legally authorized representative before initiation of study procedures. Data were periodically reviewed by the study Data Safety and Monitoring Board. Two participants had a delay in end point due to the COVID-19 pandemic (see 'Statistical analysis' section).

Randomization and Blinding

Random codes were generated by the hospital pharmacy before study initiation, using fixed seed numbers and validated randomization software, and used in sequence. In each of the two age groups, 24 participant numbers were randomized into balanced blocks of either two or four, which were randomly assigned. Study capsule dosage forms (active and placebo) were prepared by pharmacy staff in a blinded manner using over-encapsulation, and opaque (size 3 capsule shells with Lactose NF as an excipient at Rockefeller University Hospital and size 0 capsule shells with microcrystalline cellulose as an excipient at Mount Sinai Hospital). The active drug product contained FDA-approved 50 mg riluzole tablets. For ease of use and compliance, the pharmacy packaged the blinded capsules into medication bottles or organizer trays. Bottles or trays were labelled in a blinded manner, and included patient name, visit and per protocol dosing instructions. Returned trays/bottles were collected by the pharmacy and patient returns, including capsule counts, were recorded by the pharmacy. All encapsulation, packaging and labelling procedures were double verified by pharmacy staff before dispensing.


All study personnel had training on study procedures and assessments. A board-certified neurologist made a neurological assessment and administered the MMSE to all participants. FDG-PET scans were acquired at baseline and at 6 months. 1H MRS was performed at baseline, 3 months and 6 months. A neuropsychological testing battery was performed by a licensed neuropsychologist at Rockefeller University and supervised by one at Mount Sinai at baseline, 3 months and 6 months. Patients were seen once a month in clinic for clinical assessment and blood samples were obtained at every visit for safety laboratory exams; blood test results were evaluated by a physician not directly involved in the study to maintain physician-investigators blind.

Outcome Measures

Primary end points were: (i) change from baseline to 6 months in cerebral glucose metabolism measured with FDG-PET in posterior cingulate cortex, hippocampus, precuneus and medial temporal, lateral temporal, inferior parietal and frontal lobes, referred to collectively as our prespecified regions of interest; and (ii) changes in 1H MRS measures of NAA in posterior cingulate (PC) at 6 months. Secondary outcome measures were neuropsychological testing [Alzheimer's Disease Assessment Scale-Cognitive Subscale (ADAS-cog[47,48]), Alzheimer's Disease Cooperative Study (ADCS) Activities of Daily Living (ADL Inventory[49]), Neuropsychiatry Inventory (NPI[50]) total and other measures of memory, executive, visuospatial, attention and language functions for correlation with neuroimaging biomarkers. Another secondary outcome measure was in vivo measurement of glutamate with 1H MRS in PC as a marker of target engagement at 3 and 6 months compared to baseline. 1H MRS measures were obtained in bilateral hippocampi as a predefined exploratory outcome. Each FDG-PET image was also analysed using a previously developed Alzheimer's disease Progression Classifier [Figure 1A(ii)] that quantifies the degree to which a pattern of hypometabolism and preservation relative to whole brain is expressed.[51] Increases in classifier score correspond to increased expression of a pattern of hypometabolism that corresponds to the progression of Alzheimer's disease as validated using over 500 ADNI participants.[51]

Figure 1.

Neuroimaging measures for the study: FDG-PET (left) and 1 H MRS (right). [A(i)] Prespecified regions of interest, which were masked with each participant's grey tissue segment. [A(ii)] Alzheimer's disease progression classifier pattern, in which increasing progression scores reflect increasing expression of the pattern (subset shown) of hypometabolism (blue) and preservation (red) relative to whole brain. The progression scores of 517 test participants from amyloid negative cognitively normal status through amyloid-positive early mild cognitive impairment (EMCI), late MCI (LMCI) and Alzheimer's dementia (AD) are shown, with mean and standard error, illustrating the correspondence between increased score and worsening clinical severity (data derived using FDG-PET scans from ADNI,, as described in Matthews et al. 51). [B(i)] Axial, (ii) sagittal and (iii) coronal magnetic resonance images of a human brain, with depiction of the size and placement of the voxel of interest in the posterior cingulate cortex (PC). PC voxel dimensions: 2.0 cm (anterior–posterior) × 2.0 cm (left–right) × 2.0 cm (superior–inferior) or 8 cm3. [B(iv)] Sample CT-PRESS MRS data from the PC voxel, showing (a) an experimental spectrum with a clearly resolved C-4 glutamate (Glu) resonance at 2.35 ppm, as well as the resonances for NAA, total creatine (tCr), total choline (tCho) and combined resonances of C-2 glutamate and C-2 glutamine (Glx); (b) model fitting of spectrum in a to obtain the metabolite peak areas of interest; (c) individual components of the model-fitted spectrum in a; (d) residuals of the difference between spectra in a and in b.

As post hoc exploratory analyses, relationships between the FDG-PET and MRS measures and the cognitive end points were examined. NPI scores and FDG-PET in the orbitofrontal cortex, a region associated with disinhibition, apathy and other neuropsychiatric attributes,[52,53] were examined for potential association. MRS outcomes in NAA were evaluated at 3 months, and total creatinine (tCr), and other brain metabolites were examined at 3 and 6 months. A composite subregion of posterior cingulate and inferior precuneus was measured on post hoc basis to maximize spatial overlap with the boundaries defined in the MRS scans. Subgroup analyses were conducted, stratified by APOE ɛ4 carrier status, age group and sex. Age was of interest due to differences in clinical rates of decline and in the distribution of tau pathology in younger versus older Alzheimer's disease patients.[54]

FDG-PET Methods

For each FDG-PET scan, 5 mCi of FDG was administered followed by a 40-min uptake period during which the participant was in a resting state with eyes and ears open, without activity or audiovisual distraction. Images were acquired on a Siemens Biograph 64mCT scanner as a series of four frames of 5 min each. In some initial cases, a full dynamic scan was performed and late timeframes were extracted for processing and analysis.

All PET images were inspected for motion or artefact. Using SPM12 (Wellcome Trust), motion correction was performed and frames averaged into a static image. Each 6-month scan was co-registered to the baseline FDG scan, which was co-registered to the participant's T1-weighted MRI scan. MRI scans were segmented into grey, white and CSF tissue and spatially transformed to a template in MNI space, and the spatial transformations applied to the PET scans. Regions of interest [Figure 1A(i)] adapted from Freesurfer[55,56] atlases were thresholded with a smoothed grey participant-specific segment and average intensities within each region of interest were measured. A reference region for calculation of standardized uptake value ratios (SUVRs) was defined based on preserved voxels in the Alzheimer's disease Progression Classifier, most pronounced in the paracentral region. Longitudinal changes in SUVRs using this reference were compared to SUVRs referenced to (separately) centrum semi-ovale white matter, cerebellum, pons and whole brain. While these regions tend to be more variable due to technical factors (cerebellum, pons), progressive hypometabolism effects (whole brain) or potentially affected by riluzole (cerebellum),[57] consistency in results could help to confirm the robustness of findings. Images were also evaluated (scored) using the FDG Alzheimer's disease Progression Classifier.

MRI and 1H MRS Methods

All MRS neuroimaging studies were conducted on a multinuclear 3.0 T GE SIGNA HDx or Discovery MR750 system. Each enrolled participant underwent high resolution axial T1-, T2- and spin density-weighted scans. These images were used to prescribe the voxels of interest for the 1H MRS scans. A T1-weighted volumetric scan was acquired using a spoiled gradient-recalled echo sequence (SPGR, repetition time 12.21 ms, echo time 5.18 ms, flip angle = 7°, voxels 0.94 × 0.94 × 1.5 mm) on the GE HDx system or a magnetization-prepared rapid gradient-echo sequence (MPRAGE, repetition time 8.34 ms, echo time 1.7 ms, flip angle = 7°, voxels 0.94 × 0.94 × 1.5 mm) on the GE Discovery MR750, along with an axial fast FLAIR scan for brain tissue segmentation and use in PET image co-registration and region of interest definition, and to rule out exclusionary focal brain lesions.

In vivo brain levels of glutamate, NAA, tCr and other major metabolites were obtained using 1H MRS and a 2 × 2 × 2-cm3 PC cortex voxel of interest [Figure 1B(i–iii)], in ~6.5 min using the constant-time point-resolved spectroscopy (CT-PRESS) technique[58,59] with echo time 30 ms, 129 constant-time increments (t1) of 0.8 ms and repetition time 1500 ms, and a receive-only eight-channel phased-array head coil, as we recently described.[60] The distinguishing feature of CT-PRESS is that it enables MRS measurement of glutamate uncontaminated by glutamine.[58,59] Figure 1B(iv) presents a sample PC CT-PRESS spectrum acquired in 6 min, and its processing to derive the levels of the metabolites of interest.

Methods for 1H MRS of hippocampi obtained for exploratory analysis are in the Supplementary material.

1H MRS Data Processing and Quantification

Using previously described spectral quality assessment criteria,[61] the areas of the individual spectral peaks, which are proportional to their respective concentrations, were obtained by frequency-domain fitting each resonance to a Gauss–Lorentz (i.e. pseudo-Voigt) line-shape function using the Levenberg–Marquardt non-linear least-squares algorithm as implemented in 1H MRS data processing software written in IDL[61] and illustrated in Figure 1B(iv) and Supplementary Figure 2A(d) for CT-PRESS and J-edited spectra, respectively. The levels of NAA, glutamate, GABA, Glx and other metabolites were expressed semi-quantitatively as ratios of peak areas relative to that of the unsuppressed water signal (W) from the same voxels, as previously described.[61] For consistency with earlier MRS literature, levels of the same metabolites were also expressed as peak ratios relative to tCr area in the same voxel. To estimate the proportions of grey matter, white matter and CSF contained in the voxels of interest, SPM8 ( and in-house program in MATLAB ( were used to generate the proportions of grey matter, white matter and CSF for each voxel.

Statistical Analysis

Placebo and treatment groups were compared to identify potential baseline differences in region of interest SUVRs and in age, sex, APOE ɛ4 dose and carrier status, and MMSE score. The 6-month change in SUVR in each region of interest was compared across groups using a one-way analysis of covariance model with the change in SUVR value as the dependent variable, study arm as the categorical independent variable, and baseline SUVR value as a continuous variable covariate (JMP v.15, SAS software) (results were consistent with use of post-treatment SUVR as the independent variable). Age, gender, APOE ɛ4 carrier status and baseline MMSE were investigated as covariates. Assumptions including normal distribution, homogeneity of variance and linear correlation between baseline and post-treatment SUVR were verified, and the number of covariates in a given parametric model was limited to 1–3. Non-parametric tests were applied depending on the number of participants per analysis group and other assumption tests. Effect sizes were calculated using Cohen's d. For the two participants who received their FDG-PET scan 2 and 3 months after the 6-month time point due to restrictions arising from COVID-19, the change in value was adjusted using a linear proportional reduction (e.g. value × 6/8 or 6/9). Groups were evaluated post hoc without these two participants. 1H MRS data, with three time points, were analysed using linear mixed effects models with group, time point and group × time point interaction as fixed effects and participant as a random effect (SAS Studio v.3.8). In this exploratory study, a P-value of less than 0.05 was considered significant and correction for multiple comparisons was not prespecified. However, results using a Bonferroni correction for multiple comparisons were also reported for significant primary end points.

Non-prespecified FDG and MRS data were evaluated in the same manner as prespecified outcomes, without correction for multiple comparisons. The study was not statistically powered for clinical end points or these additional MRS measures, but directional trends were examined for potential effect.

Data Availability

Anonymized data will be shared by request from a qualified academic investigator for the sole purpose of replicating procedures and results presented in the article and as long as data transfer is in agreement with IRB of the involved Institutions, which should be regulated in a material transfer agreement.