Predicting survival in patients with early Alzheimer's disease

Abstract

We investigated whether an index based on clinical features, electroencephalogram and computed tomography is useful to predict survival in early Alzheimer's disease. One hundred and sixty-three consecutively referred patients to an outpatient memory clinic and first diagnosed with Alzheimer's disease (105 'probable' and 58 'possible', NINCDS-ADRDA criteria) were studied and outcome measure was death. Cox proportional hazards regression analysis and Kaplan-Meier survival curves were used to investigate relations between baseline parameters and survival. Eighty-four patients (51. 5%) died during the follow-up period that extended to 5.8 years, with a median duration of survival after entry of 4.3 years. Baseline factors that were statistically significant and independently related to increased risk of mortality were high age, male sex, poor cognitive function as measured with the CAMCOG, low alpha and beta power on electroencephalogram, and temporoparietal atrophy on computed tomography scan. These results were independent of the diagnosis probable or possible Alzheimer's disease. Based on the coefficients from the regression equation, we computed a survival index for each patient and we constructed three groups according to tertiles of this index. After 5.2 years of follow-up, survival curves showed a low mortality group with 81.7% patients alive (median survival at least 5.7 years), an intermediate mortality group with 35.9% patients alive (median survival 3.8 years), and a high mortality group with no patients alive (median survival 2.3 years). Log rank tests were statistically significant for comparisons between all three groups. We conclude that an overall index combining demographic, cognitive, electroencephalogram and computed tomography features is a strong predictor of survival in early Alzheimer's disease.