Assessing the newly proposed MRI criteria for diagnosing sporadic Creutzfeldt-Jakob disease.

Sporadic Creutzfeldt-Jakob disease (sCJD) is a frequent differential diagnostic consideration in patients with rapidly progressive dementia (RPD). Fortunately, in the last 2 decades there has been substantial cumulative improvements in sCJD biomarkers, particularly those based on imaging and cerebrospinal fluid (CSF) interrogation. Brain MRI is a very frequently employed investigation in patients with RPD, often utilized quite early in the evaluation and thereby offering a potentially key role in prompting initial concerns for sCJD. Extant conventional MRI criteria for sCJD diagnosis are relatively stringent, requiring fluid attenuated inversion recovery (FLAIR) or diffusion weighted imaging (DWI) high signal changes in 2 or more cortical regions (excluding frontal) or in both the caudate and putamen. Challenging these conventional criteria, a recent publication described improved sensitivity and unchanged specificity if MRI criteria were arguably less rigorous, requiring DWI high signal changes in only 1 or more of 7 discrete brain regions: frontal, parietal, occipital or temporal cortices, as well as the caudate, putamen or thalamus. The aim of the current study was to test the diagnostic performance of this proposed change in MRI criteria in the Australian context and compare it with conventional criteria, as well as 2 other stringent sets of criteria, predicting that a similar improved sensitivity with unchanged specificity would be observed when the proposed criteria were utilized. Sixty-five definite sCJD cases were compared with 63 age- and sex-matched controls. Radiological review of all MRIs applying the different sets of MRI criteria was undertaken by a blinded neuroradiologist, very experienced in CJD interpretation, with independent assessment of 71 MRIs performed by a second blinded neuroradiologist less experienced in sCJD imaging findings. Our study found the sensitivity of the recently proposed MRI criteria (92.3%) to be comparable to that originally reported (90-95%) and also equivalent to the conventional MRI diagnostic criteria (92.3%), while the specificities were also quite similar between the conventional MRI criteria (87.3%) and proposed criteria (85.7%), with the latter lower than previously reported. Negative predictive values and positive predictive values were also very similar between the conventional and proposed MRI criteria. Other MRI criteria assessed were associated with unacceptably low sensitivity for clinical use. Inter-rater reliability as assessed by intra-class correlation coefficients (ICC) revealed moderate reliability for the conventional and proposed MRI criteria, modestly better in the former and when the frontal lobe was retained versus excluded in comparisons.