Cognitive and imaging markers in non-demented subjects attending a memory clinic: study design and baseline findings of the MEMENTO cohort

The MEMENTO cohort is a clinic-based study of patients presenting with a large variety of cognitive symptoms and subjective cognitive complaints (SCCs) that will be followed over a 5-year period. Between April 2011 and June 2014, among the 2449 participants screened as meeting the inclusion criteria, 2323 patients consented to participate in the study. The recruitment took place within the French national network of university-based memory clinics (Centres de Mémoires de Ressources et de Recherche [CMRR]). The 28 CMRRs comprising the network were approached because their clinical research centres had (1) the potential to include a substantial number of participants, (2) access to neuroimaging (1.5- or 3-T magnetic resonance imaging [MRI]) and (3) biobank facilities. Twenty-six CMRRs agreed to participate. The number of included subjects per CMRR ranged from 17 to 305, and Fig. 1 shows the distribution of inclusions numbers by centre.

The participants were screened for either very mild to mild cognitive impairment or isolated cognitive complaints, and they were recruited consecutively. Very mild to mild cognitive impairment was defined as (1) performing 1 SD worse than the subject’s own age, sex and education-level group mean in one or more cognitive domains, this deviation being identified for the first time through cognitive tests performed recently (less than 6 months preceding screening phase), and (2) having a Clinical Dementia Rating (CDR) [10] ≤0.5 and not being demented. The neuropsychological tests battery and the references for age, sex and education-level norms used for each test are detailed in Additional file 1. A participant was eligible for inclusion in the isolated SCCs stratum if he or she had SCCs (assessed through visual analogue scales) without any of objective cognitive deficit as defined above and was aged 60 years or older.

At memory clinics, baseline data collection included socio-demographic characteristics; personal and familial medical history; neurological and physical examination (including anthropometric measurements and three measures of blood pressure after 2 minutes of rest in sitting position using the Omron M6 monitor, OMRON Healthcare, Kyoto, Japan); current medication intake with detailed recording of doses as well as treatment onset date; cognitive and non-cognitive subjective complaints (domains assessed were memory, attention, language, physical health, mood, sensory organs, general health) measured using a visual analogue scale ranging from 0 to 10; a full neuropsychological test, which consisted of the Mini Mental State Examination (MMSE) [11], digit span (forward and backward) [12], Free and Cued Selective Reminding Test [13], Delayed Matching to Sample 48 (DMS48) [14], Verbal Fluency (animals and letter p) [15], Image Naming [16], praxis assessment [17], Rey-Osterrieth Complex Figure Test [18], Trail Making Test (TMT) A and B [19, 20], Frontal Assessment Battery (FAB) [21, 22] and Clinical Dementia Rating that was completed with an informant in person if he or she was accompanying the participant, by phone otherwise; rating of neuropsychiatric symptom presence and severity using the Neuropsychiatric Inventory–Clinician (NPI-C) [23]; lifestyle indicators such as self-report of current and past alcohol consumption and smoking habits, physical activity using International Physical Activity Questionnaire [24] and leisure activity frequency; autonomy using the Instrumental Activities in Daily Living Scale and Activities of Daily Living Scale [25, 26]; and motricity using the Short Physical Performance Battery (SPPB) [27]. For CDR, NPI-C and quoting of neuropsychological tests, training sessions were organised to optimise standardisation across centres. All data were recorded via secured online electronic data capture.

The percentages of missing data entered in the electronic case report forms ranged from 0.09% to 2.12% across centres. Twenty centres had less than 1% of missing data. These numbers illustrate the high compliance rate of the centres with the MEMENTO study, as well as the efficiency of data-monitoring procedures (both on-site and centralised).

As part of the inclusion criteria, participants had to agree to undergo brain MRI, and 86% of participants had a 3.0-T MRI scan (1.5 T otherwise). ¹⁸F-fluorodeoxyglucose positon emission tomography (FDG-PET) was optional and was performed in 60% of participants. All neuroimaging acquisition was coordinated by the Center for Automated Treatment of Images (CATI; cati-neuroimaging.com), a platform dedicated to multicentre neuroimaging [28]. For the MEMENTO study, CATI harmonised MRI and PET imaging across a network of facilities associated with the network of French memory centres. Physicists and engineers standardised acquisitions according to a systematic qualification procedure, ensuring parameter uniformity and image quality. They had frequent contact with each acquisition site to track any hardware or software upgrades. Detailed information on scanner manufacturers and acquisition parameters for each study’s site can be found on the CATI website.

The MRI protocol harmonised by CATI is made up of the sequences described in Table 1. It guarantees compatibility with the Alzheimer's Disease Neuroimaging Initiative (ADNI) protocol to a large extent. The order of the MRI sequences corresponds to priority. Diffusion magnetic resonance imaging (dMRI) scan acquisition leads to a final sequence of 15–60 homogeneous directions according to the comfort and compliance of the subject. Non-clinical acquisitions (functional MRI, dMRI) can be considered as options to be triggered according to the compliance of the subject. Acquisition parameters were tuned for each scanner to minimise inter-site variability.

PET centres participating in the MEMENTO study were equipped with systems set-up between 2003 and 2014 and consisted of 10 GE Healthcare, 6 Philips and 12 Siemens Healthcare systems, as well as 15 different models of scanners. No brain-dedicated high-resolution system or 2D tomography systems were included. Therefore, we chose to harmonise patient data acquisition using phantom experiments prior to patient studies to ensure good-quality imaging data among centres [29]. Our objective was to optimise contrast recovery while preserving, if possible, spatial resolution. Two phantom studies were acquired for the qualification process of the 22 centres that agreed to perform FDG-PET: the Jaszczak phantom, composed of cold rods and six hollow spheres, and the Hoffman 3D brain phantom. Standardised uptake value measurements were also checked to assess the cross-calibration between the PET scanner and the dose calibration system. Computed tomographic (CT) image acquisition parameters were set such that the effective dose would be low (≤0.3 mSv), in accordance with the recommendations of the European Association of Nuclear Medicine [30]. Different reconstruction parameters were compared on the basis of recovery coefficients (RCs) computed for each sphere and spatial resolution estimated from the cold rods. RCs were calculated from measurements in volumes of interest (VOIs) defined on each sphere and on the background. In-house software was developed for this purpose. We also computed the image spatial resolution (ISR) using a method developed by Prieto et al. [31] to estimate the full width at half maximum of each PET scanner from the three largest hot spheres. Hoffman 3D brain phantom scans were segmented in VOIs to compute right-to-left and anterior-to-posterior cortical ratios, as well as caudate-to-white matter, putamen-to-white matter and grey matter-to-white matter ratios. For each centre, the set of optimal reconstruction parameters was chosen as the one maximising ISR and the RC without a noticeable decrease in signal-to-noise ratio.

After the set-up visit, we provided recommendations to each centre for acquisition and reconstruction parameters to be used for the MEMENTO cohort. In addition, centres were qualified after the analysis of the images of a first test patient. They received a technical manual describing in detail the procedures pertaining to patient preparation, injected dose, PET-CT image acquisition and data transfer to CATI.

Brain FDG-PET scans were obtained 30 minutes after injection of 2 MBq/kg of 2-deoxy-2-¹⁸F-fluoro-d-glucose. All acquisitions consisted of 3 × 5-minute frames. Images were then reconstructed using and iterative algorithm, and last, frames were realigned, averaged and quality-checked by the CATI team.

The data flow between the acquisition network and the CATI centralised analysis team relied on a secured web service. A team of clinical research assistants performed quality control of the incoming data in the days following the data transfer to provide rapid feedback when required. Quality control relied on a dedicated software programme supporting the check for protocol consistency (e.g., scanner type, software version, reception coil, sequences acquired, order of sequences, sequence parameters, reconstruction parameters) and the generation of a documented series of indices characterising, for instance, acquisition slab positioning; movement; spikes and other artifacts and their localisation; and overall quality of the image through contrast, noise or intensity non-uniformity.

Validated data were moved to a centralised database at the disposal of several teams in charge of analysis. Each analysis of CATI’s portfolio was performed following a systematic procedure providing quality control indices. Below is a summary of the measurements included in MEMENTO database:

A selection of these measurements is described in this paper.

From baseline blood intake, standard biological measurements (including glycaemia, triglycerides, high-density lipoprotein, and low-density lipoprotein) were performed at local biochemistry departments. Study-specific blood sampling included serum (12 tubes of 0.25 ml), plasma ethylenediaminetetraacetic acid (EDTA; 8 tubes of 0.25 ml), total blood heparin (2 tubes of 1 ml), plasma heparin (4 tubes of 500 μg), blood EDTA without plasma (1 tube of 0.25 ml), blood heparin without plasma (1 tubes of 3 ml) and Tempus (2 tubes of 3 ml). Samples were stored in a centralised biobank (Genomic Analysis Laboratory-Biological Resource Centre [LAG-CRB], Pasteur Institut Lille, BB-0033-00071).

LAG-CRB extracted genomic DNA from peripheral blood samples using Gentra Puregene blood kits (QIAGEN, Hilden, Germany). Apolipoprotein E (APOE) genotypes were determined by KBiosciences (Hoddesdon, UK; www.​kbioscience.​co.​uk), using their own system of fluorescence-based competitive allele-specific polymerase chain reaction. Two APOE single-nucleotide polymorphisms, rs429358 and rs7412, allowed identification of the three major APOE alleles (ε2, ε3 and ε4).

Lumbar puncture was optional and was performed in 17% of participants at baseline. Cerebrospinal fluid (CSF) was collected in polypropylene tubes following standardised conditions and using an atraumatic needle. Each CSF sample was transferred to the CSF bank within 4 h after collection and was centrifuged at 1000 × g at 4 °C for 10 minutes. CSF samples were aliquoted in polypropylene tubes (16 tubes of 250 μl) and stored at −80 °C. All tubes were shipped for storage in a centralised biobank (LAG-CRB, Pasteur Institut Lille, BB-0033-00071). Measurements of CSF amyloid-β 42 peptide (Aβ₄₂), CSF Aβ₄₀, total tau, and phosphorylated tau (p-tau181) levels are ongoing using the standardised commercially available INNOTEST sandwich enzyme-linked immunosorbent assay (Fujirebio, Ghent, Belgium).

Longitudinal follow-up took place every 6 months. Table 2 describes the schedule of overall evaluations. During follow-up, all incident cases of dementia (DSM-IV criteria for dementia and National Institute of Neurological and Communicative Disorders and Stroke/Alzheimer’s Disease and Related Disorders Association criteria for AD) [45, 46] were reviewed by an independent committee.

Baseline characteristics of participants were summarised by cognitive categories (isolated SCC or MCI staging according to Petersen criteria [47]; i.e., single-domain amnestic mild cognitive impairment [aMCI], multi-domain aMCI, single-domain non-amnestic mild cognitive impairment [naMCI], multi-domain naMCI), age categories in years (≤ 60; 60 – 69, 70 – 79, ≥ 80), sex and APOE genotype. An isolated SCC category comprised participants without impairment at any of the cognitive tests of the screening battery (impairment defined as a score ≥1.5 SD worse than a participant’s own age, sex or education-level group mean in a cognitive domain).

Definitions used were as follows: for highest diploma, at least baccalaureate degree (yes/no); for cardiovascular burden, diabetes (self-reported diabetes or anti-diabetic drug intake of glycaemia > 7 mmol/L), hypertension (anti-hypertensive drug intake or mean of three blood pressure measurements either ≥ 140 mmHg for systolic blood pressure or ≥ 90 mmHg for diastolic blood pressure), dyslipidaemia (plasma cholesterol > 6.24 mmol/L or use of any lipid-lowering drugs), history of cardiovascular disease (self-reported history of myocardial infarction, surgical bypass, stroke, peripheral artery disease, angina pectoris); for neuropsychiatric symptoms, apathy, depression, anxiety based on NPI-C; for physical impairment and motricity, number of limitations in instrumental activities of daily living (none, 1, 2 or more) and SPPB score; for genetic AD risk, number of copies of the APOE ε4 allele (0, 1 or 2); for cognitive level, performance on all cognitive tests; for brain MRI biomarkers, hippocampal volume (by hemisphere), brain parenchymal fraction (grey matter + white matter volumes divided by total intracranial volume), total white matter lesion (WML) volume and mean cortical thickness by hemisphere; and for FDG-PET biomarkers, mean FDG uptake values normalised to the pons in five AD-specific regions derived from the ADNI cohort (angular and parietal inferior right, parietal inferior left, precuneus and cingulum posterior left, temporal inferior left, temporal inferior right) [48].

Analyses were performed using SAS version 9.3 software (SAS Institute, Cary, NC, USA) and Stata release 14 software (StataCorp LP, College Station, TX, USA). Results are reported following the Methods in Longitudinal Research on Dementia guidelines [49].