Longitudinal magnetic resonance imaging in progressive supranuclear palsy: A new combined score for clinical trials

Abstract

Background: Two recent, randomized, placebo-controlled phase II/III trials (clinicaltrials.gov: NCT01110720, NCT01049399) of davunetide and tideglusib in progressive supranuclear palsy (PSP) generated prospective, 1-year longitudinal datasets of high-resolution T1-weighted three-dimensional MRI.

Objective: The objective of this study was to develop a quantitative MRI disease progression measurement for clinical trials.

Methods: The authors performed a fully automated quantitative MRI analysis employing atlas-based volumetry and provide sample size calculations based on data collected in 99 PSP patients assigned to placebo in these trials. Based on individual volumes of 44 brain compartments and structures at baseline and 52 weeks of follow-up, means and standard deviations of annualized percentage volume changes were used to estimate standardized effect sizes and the required sample sizes per group for future 2-armed, placebo-controlled therapeutic trials.

Results: The highest standardized effect sizes were found for midbrain, frontal lobes, and the third ventricle. Using the annualized percentage volume change of these structures to detect a 50% change in the 1-year progression (80% power, significance level 5%) required lower numbers of patients per group (third ventricle, n = 32; midbrain, n = 37; frontal lobe, n = 43) than the best clinical scale (PSP rating scale total score, n = 58). A combination of volume changes in these 3 structures reduced the number of required patients to only 20 and correlated best with the progression in the clinical scales.

Conclusions: We propose the 1-year change in the volumes of third ventricle, midbrain, and frontal lobe as combined imaging read-out for clinical trials in PSP that require the least number of patients for detecting efficacy to reduce brain atrophy. © 2017 International Parkinson and Movement Disorder Society.

Keywords: clinical trials; magnetic resonance imaging; power calculation; progressive supranuclear palsy; volumetry.

Figure 1

Image processing and volume determination shown exemplarily for the left caudate nucleus: 1) Segmentation of gray matter (GM), white matter (WM) and cerebrospinal fluid (CSF) compartments by applying the ‘unified segmentation’ algorithm of SPM12 to a T1-weighted 3D image. 2) Normalization of the resulting GM image in native space using diffeomorphic anatomical registration through exponentiated Lie algebra (DARTEL) with predefined templates in Montreal Neurological Institute (MNI) space. 3) Multiplication of the modulated GM image derived by DARTEL normalization with a mask of the left caudate nucleus. In this example the masking image is derived from the Harvard-Oxford probabilistic brain atlas of subcortical structures. ^– The voxel-wise multiplication results in a modulated GM image with the caudate nucleus isolated. Due to ‘modulation’ of the grey matter image, the effect of normalization (i.e. extension or shrinkage of the investigated structure) is compensated for so that the sum of the residual voxels in the final image represents the volume of the original structure in native space. For volume measurements of white matter structures or CSF compartments the same image processing steps are based on normalized and modulated white matter or CSF images, respectively.