Although the mechanisms underlying the accumulation of disability in primary progressive (PP) multiple sclerosis (MS) are still unclear, a major role seems to be played by 'occult' tissue damage. We investigated whether conventional and magnetization transfer (MT) MRI may provide complementary information for the assessment of PPMS severity. Conventional and MT MRI scans from 226 PPMS patients and 84 healthy controls were collected for centralized analysis. The expanded disability status scale (EDSS) score was rated at the time of MRI acquisition. T2 lesion volume, normalized brain volume (NBV) and cervical cord cross-sectional area (CSA) were measured. Magnetization transfer ratio (MTR) histograms from whole brain tissue, normal-appearing white matter and grey matter (NAGM) were also obtained. Mean NBV, CSA and MTR histogram-derived metrics showed significant inter-centre heterogeneity. After correcting for the acquisition centre, pooled average MTR and histogram peak height values were different between PPMS patients and controls for all tissue classes (P-values between 0.03 and 0.0001). More severe brain and cord atrophy and MT MRI-detectable NAGM damage were found in patients who required walking aids than in those who did not (P-values: 0.03, 0.001 and 0.016). A composite score of NBV, CSA, whole brain and NAGM MTR histogram peak height z-scores was correlated with patients' EDSS (r = 0.37, P 0.001). Magnetization transfer MRI might provide information complementary to that given by conventional MRI when assessing PPMS severity. Sequence-related variability of measurements makes the standardization of MT MRI acquisition essential for the design of multicentre studies.