Diagnostic Z-score mapping: Use of younger normal database to improve discrimination of dementia

Abstract

17

Objectives: Use of normal reference databases to create Z-score maps for diagnostic brain PET/SPECT is widespread. However, occult neurodegenerative and cerebrovascular pathology in “normal” elderly controls may reduce the sensitivity of Z-score mapping for dementia diagnosis. We hypothesized that using a reference group of younger normals would improve discriminatory power owing to less metabolic variance (less comorbidity) and outweigh confounding age-related changes.

Methods: Brain [F-18]FDG PET (10mCi) was performed in 51 normals recruited from prospective aging study cohorts and 10 dementia patients. Metabolic activity was extracted using 3D-SSP. Two normal databases from old (OC, age 70-82, n=23) and younger (YC, age 42-69, n=23) controls were established to generate Z-score maps for the dementia patients (n=10, age 61-81) and comparison normals (n=5, age 83-89).

Results: Normalized metabolic activity declined with age: -8.0% in medial frontal (mF) vs. -2.6% in fronto-tempo-parietal (FTP) cortices. However, coefficients of variation in the same regions were 21% smaller using the YC (2.5%-4.0%) vs. OC (2.8%-4.9%) (p<0.05 in parietal and mF; variance ratio). Averaged Z-scores in FTP cortices discriminated dementia vs. old normals 23% better using the YC vs. OC database (YC: -2.01 & 0.06, differential Z=2.07 vs. OC: -3.40 & -0.85, differential Z=2.55).

Conclusions: Despite presence of age-related changes, use of a younger instead of age-similar normal database yielded more robust discrimination of dementia owing to smaller metabolic variances. This approach, similar to T-score evaluation of DEXA for osteoporosis, improves the diagnostic workup of prevalent aging disorders such as dementia.

Research Support: RO1NS045254 U01AG06781