BACKGROUND: The diagnostic criteria for primary-progressive multiple sclerosis (PPMS) have undergone revision over the last 20 years. Cerebrospinal fluid oligoclonal bands (CSFOBs) have received less emphasis in recent revisions of the McDonald criteria. The aim of this study was to examine the sensitivity of the diagnostic criteria for PPMS with particular reference to spinal cord criteria and examine the utility of CSFOBs in a cohort of PPMS patients. METHODS: All new PPMS diagnoses between 1990 and 2011 were identified. Baseline clinical details and paraclinical evaluations including MRI of the brain, spinal cord, CSF and visually evoked responses (VERs) were assessed. The proportion of patients who met the requirements for diagnosis of PPMS on the basis of Thompson's and the McDonald Criteria (2001, 2005, 2010) were determined. RESULTS: There were 88/95 PPMS patients who had at least two diagnostic investigations. The sensitivity of Thompson's and the McDonald 2001 criteria was 64%; the McDonald 2010 revisions gave the highest sensitivity (77%); the McDonald 2005 criteria had intermediate sensitivity (74%). The combination of CSFOBs and MRI of the brain yielded the greatest number of patients demonstrating dissemination in space (DIS) on only two investigations. VERs did not aid diagnosis. Reducing requirements for the number of spinal cord lesions (symptomatic or not) to one increased diagnostic sensitivity to 84%. CONCLUSION: An alternative criterion requiring two of: i) MRI of the brain with one or more lesions in two of three regions typical for demyelination; ii) the presence of one T2-weighted spinal cord plaque (typical for demyelination); iii) CSFOBs; would increase the diagnostic sensitivity for PPMS.