Advantages and challenges of using arterial spin labelling MRI to monitor cerebral blood flow in multi-centre clinical trials of neurodegenerative disease: Experience from the RADAR study

Abstract

Arterial spin labelling (ASL) enables non-invasive quantification of regional brain perfusion using MRI. ASL was used in the Reducing Pathology in Alzheimer's Disease through Angiotensin TaRgeting (RADAR) multi-centre trial to pilot the assessment of the effects of the anti-hypertension drug losartan on cerebral blood flow (CBF). In the multi-centre setting, disparities in ASL implementation on scanners from different manufacturers lead to inherent differences in measured CBF and its associated parameters (e.g. spatial coefficient of variation (sCoV) of CBF, a proxy of arterial arrival times). In addition, differences in ASL acquisition parameter settings can also affect the measured quantitative perfusion values. In this study, we used data from the RADAR cohort as a case study to evaluate the site-dependent systematic differences of CBF and sCoV, and show that variations in the readout module (2D or 3D) and the post-labelling delay acquisition parameter introduced artifactual group differences. When accounting for this effect in data analysis, we show that it is still possible to combine ASL data across sites to observe the expected relationships between grey matter CBF and cognitive scores. In summary, ASL can provide useful information relating to CBF difference in multi-centre therapeutic trials, but care must be taken in data analysis to account for the inevitable inter-site differences in scanner type and acquisition protocol.

Keywords: Alzheimer's disease; Angiotensin; Arterial spin labelling; Blood pressure; Cerebral blood flow; Hypertension; MRI; Multi-centre randomised controlled trials.

Fig. 1

Examples of arterial spin labelling (ASL) image artefacts leading to rejection from the final analysis. (A) Tagged blood seen in the vessels but not in the tissue, commonly referred to as arterial transit artefacts (ATA) (B) Fat suppression artefact visible in the centre of the CBF image and (C) in the temporal standard deviation of CBF. (D) Motion artefact in 2D EPI images acquired without background suppression. (E) Nyquist ghost artefact present in the CBF image and (F) emphasised in the temporal standard deviation of CBF.

Fig. 2

Representative CBF maps from the four sites with ASL scans used for analysis. For illustrative purposes, all images are shown in MNI space. The display window is from 0 to 100 mL/100 g/min.

Fig. 3

Site comparison of global grey matter brain perfusion. Circles show mean CBF_GM for each participant; solid black lines represent group mean values. Statistical testing by one-way ANOVA with post-hoc testing by Tukey; * p < 0.05, ** p < 0.01 (for specific p values, see text in Results).

Fig. 4

Multiple linear regression results for significant brain perfusion predictors: (A) ADAS-Cog score and (B) ASL readout module type.

Fig. 5

Flow territories where perfusion was linked with ADAS-Cog score. Flow territories perfused by bilateral proximal middle cerebral arteries (green), proximal posterior cerebral arteries (yellow) and intermediate posterior cerebral arteries (blue). Coordinates for sagittal, coronal and axial slices are in MNI152 space. Outlines of the anatomical masks of hippocampi, defined using the Harvard-Oxford cortical structural atlas, are overlaid in white. After applying Bonferroni adjustment for multiple comparisons, only the associations between blood flow values in territories perfused by the bilateral posterior cerebral arteries (proximal and intermediate; yellow and blue) remained statistically significant in relation to ADAS-Cog scores.

Fig. 6

Multiple linear regression results for sCoV of CBF, demonstrating associations with (A) age, (B) sex and (C) PLD/TI of the ASL sequence. (D) Visual representation of the relative contributions of each predictor to the R².