Premium
P1‐166: A prospective validation study of the predictnd tool: A diagnostic decision support tool—rationale and design of the study
Author(s) -
Bruun Marie,
Frederiksen Kristian Steen,
Waldemar Gunhild,
Soininen Hilkka,
Flier Wiesje M.,
Mecocci Patrizia,
Rhodius-Meester Hanneke F.M.,
Herukka Sanna-Kaisa,
Baroni Marta,
Remes Anne,
Urhemaa Timo,
Mattila Jussi,
Lötjönen Jyrki,
Hasselbalch Steen Gregers
Publication year - 2015
Publication title -
alzheimer's and dementia
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 6.713
H-Index - 118
eISSN - 1552-5279
pISSN - 1552-5260
DOI - 10.1016/j.jalz.2015.06.365
Subject(s) - medical diagnosis , dementia , medicine , memory clinic , disease , population , prospective cohort study , gold standard (test) , neuropsychology , cohort , medical physics , cognition , pathology , psychiatry , environmental health
HMPAO SPECT brain imaging and AD CSF biomarkers were previously collected in diagnostically ambiguous cases were analysed. Multiple regions of interest (ROIs) were created using the automatic anatomical labelling atlas and regional measures of perfusion were obtained from HMPAO SPECT images using SPM8. Patients were subsequently grouped into abnormal and normal based on their CSF total tau, amyloid beta 42 (ab-42), p-tau protein levels and tau to ab-42 protein ratio using recommended normal reference ranges. A stepwise discriminant analysis was then performed to obtain regions on the perfusion SPECT which best predicted AD CSF group. Results:Discriminant analysis showed voxel maps of regional perfusion were highly accurate predictors of CSF biomarker status for ab-42 (79.3% correctly classified) and p-tau (89.9% correctly classified). Classification accuracy was lower for t-tau and tau-ab ratio. The left angular gyrus was the best regional predictor of biomarker status, with a positive predictive value (PPV) of 73% and 81%, and negative predictive value (NPV) of 69% and 76% for ab-42 and p-tau respectively. Other regions including the left and right medial temporal lobe, left superior occipital gyrus, left cerebellum, right supplementary motor area and the right precuneus also contributed to classification accuracy, producing a ROI group PPVof 84% and 90%, and NPVof 76% and 90% for ab-42 and p-tau respectively. Conclusions: Our findings show HMPAO SPECT can predict AD CSF results with reasonable accuracy, and has good positive and negative predictive value in a clinically relevant setting.