Cerebrospinal fluid flow dynamics in Huntington's disease evaluated by phase contrast MRI

2.1 Study design and participants

We conducted a prospective cross-sectional pilot study named “Phase1-HD” involving participants enrolled simultaneously in the Enroll-HD study (CHDI Foundation, 2012) an international multicentre, prospective registry, and in the HD-CSF study (Byrne, Rodrigues, Johnson, De Vita, et al., 2018; Byrne, Rodrigues, Johnson, Wijeratne, et al., 2018), a single-centre prospective cohort based at the UCL Huntington's Disease Centre, London, UK.

The main goal of Phase1-HD was to establish the methods for and feasibility of applying PCMRI to people with Huntington's disease and to generate exploratory data to inform current practice and design future definitive studies.

Ten healthy volunteers were recruited before study start for scanner optimization, and their data were not used in any of the analyses. All gave written informed consent.

Ten people with early manifest Huntington's disease and 10 healthy controls were recruited through the Huntington's Disease Multidisciplinary Clinic at the National Hospital for Neurology & Neurosurgery in London, United Kingdom, from July 2016 to April 2017. All volunteers had to be 18 year of age or older, and to have no known contraindication to MRI or venepuncture. Early manifest Huntington's disease participants were defined as gene expansion carriers (i.e. 40 or more CAG repeats in the longer HTT allele) with unequivocal signs of Huntington's disease (Unified Huntington's Disease Rating Scale [UHDRS] Diagnostic Confidence Level of 4) and a good functional status implying early stage disease (UHDRS Total Functional Capacity [TFC] between 13 and 7). Healthy controls were free from significant comorbidities, and those with a family history of Huntington's disease, had all had a negative test for the Huntington's disease mutation. They were age and gender matched to the Huntington's disease participants.

Exclusion criteria were as follows: inability to tolerate any of the procedures involved, chiefly MRI scanning or phlebotomy; any clinically significant and relevant history that could affect the conduct of the study and evaluation of the data, as ascertained by the investigators, or through detailed medical history and screening assessments; inability or unwillingness to participate in an Enroll-HD core assessment, or for data obtained from Enroll-HD to be accessed.

The study was compliant with the principles of the Declaration of Helsinki, and applicable components of the International Conference on Harmonisation (ICH) Good Clinical Practice (GCP). It was approved by the London-Queen Square ethics committee, and all participants gave written informed consent.

2.2 Clinical and laboratory assessments

All participants, excluding the optimization controls, were subject to a detailed clinical evaluation before undergoing the MRI scan, including the Enroll-HD core assessment, which includes the UHDRS-99 (Huntington Study Group, 1996; Siesling, van Vugt, Zwinderman, Kieburtz, & Roos, 1998). Overall, the core assessment contains: demographic data; height and weight; alcohol, tobacco and drug history; family history; characteristics of the disease; genetic testing history; past medical history; comorbid conditions; pharmacotherapy, non-pharmacological therapy, nutritional supplements; the UHDRS TMS; the UHDRS TFC; the UHDRS Functional Assessment (FA); the UHDRS Independence Scale (IS); the Symbol Digit Modality Test (SDMT); the Categorical Verbal Fluency Test (CVFT); the Stroop Color Naming Test; the Stroop Word Reading Test; and the Problem Behaviors Assessment Short (PBA-s; Craufurd, Thompson, & Snowden, 2001). The disease burden score (DBS) – age × (CAG-35.5) – was calculated for all people with early manifest Huntington's disease (Paulsen et al., 2006, 2008).

A blood sample (approx. 10 ml) was collected in EDTA tubes (BD, NJ, USA) for osmolality measurement. Analytes’ concentrations were quantified by freezing point depression, and osmolality was computed according to the following formula:

2.3 Structural brain imaging

T1 and T2-weighted MRI data were acquired on a 3T scanner, using a protocol designed for this study (Supporting Information: Imaging Methods). From T1-weighted scans we segmented whole-brain, ventricles, and total intracranial volume (TIV) with Medical Image Display Analysis Software (MIDAS); and volumes of grey and white matter, and intracranial CSF volumes with Statistical Parametric Mapping 12 (SPM12). T2-weighted scans were stitched and total CSF volumes were calculated using MIDAS, using a protocol that was validated internally. All segmentations underwent visual quality control to ensure accurate delineation of the regions. No scans failed processing. Brain and ventricle volumes are expressed as a percentage of total intracranial volume, to account for overall head size.

2.4 CSF flow and velocities quantification

PCMRI was used to quantify CSF velocity and flow at three levels of interest (LOI): the cerebral aqueduct, and the spinal subarachnoid space at the level of the first (T1), and eighth (T8) thoracic vertebral bodies (Figure 1a,b, and c, respectively). Peripheral cardiac gating was performed via an MRI-compatible pulse oximeter. Oblique-axial slices for PCMRI were prescribed based on T1-weighted brain images and T2-weighted spine images, both perpendicular to the expected direction of flow as assessed by an experienced radiographer. Flow velocity encoding was then performed in the through-slice direction (perpendicular to the expected direction of flow). The maximum encoding velocity was 20 cm/s. For the cerebral aqueduct, 32 cardiac phases were recorded, with repetition time (TR) = 31 ms, echo time (TE) = 10 ms, field of view 135 × 135 mm (100% phase oversampling), in-plane spatial resolution 0.53 mm × 0.70 mm and 5 mm slice thickness (acquisition time approx. 6.5 min). For T1 level, 25 cardiac phases were recorded with TR = 31 ms, TE = 10 ms, field of view 135 mm × 135 mm (100% phase oversampling), in-plane spatial resolution 0.53 mm × 0.70 mm and 5 mm slice thickness (acquisition time approx. 6.5 min). For T8 level, same number of cardiac phases, TE, TR, and slice thickness as for T1, field of view 150 mm × 150 mm (100% phase oversampling), in-plane spatial resolution 0.59 mm × 0.65 mm (acquisition time approx. 7.6 min). Flow scans underwent quality control at the time of acquisition and prior to processing. This protocol was optimized using the scans of 10 healthy volunteers.

A single set of T8 level PCMRI images from a male manifest HD participant was excluded due to the presence of motion artifact.

Two independent investigators (FBR and RIS), blinded to participant's disease status, analysed all datasets in duplicate. The investigators were instructed to draw manually one region of interest (ROI) per LOI per participant, only selecting pixels with flow during the cardiac cycle, while maximizing the number of pixels included in the analyses. For each ROI, CSF velocity and flow for each cardiac phase were calculated using a built in commercial flow analysis package (Argus Flow, Siemens Healthcare, Erlangen, Germany) as shown in Figure 1; our analyses then used the averages over the complete cardiac cycle. Flow values were calculated in millilitres per minute; mean velocity and peak velocity were calculated in centimetres per second.

2.5 Statistical analysis

Statistical analyses were performed using Stata v14.2 (College Station, TX). All analysis were carried out blind to subject group. The threshold for statistical significance for all analyses was p < 0.05.

Inter-group differences in demographic variables were tested using two-sample Wilcoxon rank-sum (Mann–Whitney) test for continuous variables, or Fisher's exact test for categorical variables.

To quantify the agreement between measurements, individual interclass correlations (ICCs) for flow and velocity measurements were calculated between the two raters (FBR and RIS) using a two-way mixed effect model. Only measures with an ICC over 0.75 were included in the analysis.

For all the analyses below, the velocity and flow values used were a mean of the values generated independently by each investigator.

Non-parametric Spearman's rank correlation analyses were done to test the effect of age and serum osmolality on CSF flow and velocities in healthy controls, and CAG and DBS in gene expansion carriers. The same was done for brain volumes in healthy controls.

Inter-group comparisons of CSF flow parameters were performed using a two-sample Wilcoxon rank-sum (Mann–Whitney) tests, and as a sensitivity analysis, a t test for equal (Student's) or unequal (Welch's) variances according to the variable.

Additional models were constructed to examine the relationship among velocity and flow parameters and clinical data, CAG, DBS, serum osmolality and brain volumes if there was a statistically significant effect on these covariates on the study variables.

Finally, we used the largest magnitude of effect and the most clinically significant variable, as calculated according to the Hedges’ g, to produce sample size computations for future comparisons of two independent means, using an alpha (type I error) of 0.05, and a beta (type II error) of 0.20 and 0.10.

2.6 Data availability

Access to the full dataset used in this work will be granted upon request to the lead author.