Defining A Centiloid Scale Threshold Predicting Long-Term Progression To Dementia In Patients Attending The Memory Clinic: An [F-18] Flutemetamol Amyloid Pet Study

Purpose To evaluate cerebral amyloid-beta(A beta) pathology in older adults with cognitive complaints, visual assessment of PET images is approved as the routine method for image interpretation. In research studies however, A beta-PET semi-quantitative measures are associated with greater risk of progression to dementia; but until recently, these measures lacked standardization. Therefore, the Centiloid scale, providing standardized A beta-PET semi-quantitation, was recently validated. We aimed to determine the predictive values of visual assessments and Centiloids in non-demented patients, using long-term progression to dementia as our standard of truth. Methods One hundred sixty non-demented participants (age, 54-86) were enrolled in a monocentric [F-18] flutemetamol A beta-PET study. Flutemetamol images were interpreted visually following the manufacturers recommendations. SUVr values were converted to the Centiloid scale using the GAAIN guidelines. Ninety-eight persons were followed until dementia diagnosis or were clinically stable for a median of 6 years (min = 4.0; max = 8.0). Twenty-five patients with short follow-up (median = 2.0 years; min = 0.8; max = 3.9) and 37 patients with no follow-up were excluded. We computed ROC curves predicting subsequent dementia using baseline PET data and calculated negative (NPV) and positive (PPV) predictive values. Results In the 98 participants with long follow-up, Centiloid = 26 provided the highest overall predictive value = 87% (NPV = 85%, PPV = 88%). Visual assessment corresponded to Centiloid = 40, which predicted dementia with an overall predictive value = 86% (NPV = 81%, PPV = 92%). Inclusion of the 25 patients who only had a 2-year follow-up decreased the PPV = 67% (NPV = 88%), reflecting the many positive cases that did not progress to dementia after short follow-ups. Conclusion A Centiloid threshold = 26 optimally predicts progression to dementia 6 years after PET. Visual assessment provides similar predictive value, with higher specificity and lower sensitivity.