A multivariate model of time to conversion from mild cognitive impairment to Alzheimer's disease

Abstract

The present study was aimed at determining which combination of demographic, genetic, cognitive, neurophysiological, and neuroanatomical factors may predict differences in time to progression from mild cognitive impairment (MCI) to Alzheimer's disease (AD). To this end, a sample of 121 MCIs was followed up during a 5-year period. According to their clinical outcome, MCIs were divided into two subgroups: (i) the "progressive" MCI group (n = 46; mean time to progression 17 ± 9.73 months) and (ii) the "stable" MCI group (n = 75; mean time of follow-up 31.37 ± 14.58 months). Kaplan-Meier survival analyses were applied to explore each variable's relationship with the progression to AD. Once potential predictors were detected, Cox regression analyses were utilized to calculate a parsimonious model to estimate differences in time to progression. The final model included three variables (in order of relevance): left parahippocampal volume (corrected by intracranial volume, LP_ ICV), delayed recall (DR), and left inferior occipital lobe individual alpha peak frequency (LIOL_IAPF). Those MCIs with LP_ICV volume, DR score, and LIOL_IAPF value lower than the defined cutoff had 6 times, 5.5 times, and 3 times higher risk of progression to AD, respectively. Besides, when the categories of the three variables were "unfavorable" (i.e., values below the cutoff), 100% of cases progressed to AD at the end of follow-up. Our results highlighted the relevance of neurophysiological markers as predictors of conversion (LIOL_IAPF) and the importance of multivariate models that combine markers of different nature to predict time to progression from MCI to dementia.

Keywords: Alzheimer’s disease; Delayed recall; Episodic memory; Individual alpha peak frequency; Medial–temporal volume; Mild cognitive impairment; Time to progression.

Fig. 1

Survival curves showing all possible combinations of favorable and unfavorable categories of the three variables included in the final model. These curves were estimated by means of Cox regression analyses. a Survival curves classified according to DR categories when the LIOL_IAPF and LP_ICV categories are favorable. b Survival curves classified according to DR categories when the LIOL_IAPF category is unfavorable and the LP_ICV category is favorable. c Survival curves classified according to DR categories when the LIOL_IAPF category is favorable and the LP_ICV category is unfavorable. d Survival curves classified according to DR categories when the LIOL_IAPF and LP_ICV categories are unfavorable. IAPF, individual alpha peak frequency; ICV, intracranial volume; LIOL, left inferior occipital lobe; DR, delayed recall; LP, left parahippocampal volume

Fig. 2

Proportion of stable MCI cases (sMCI) at the end of follow-up considering all the possible combinations of favorable (green color) and unfavorable (red color) categories of LIOL_IAPF, LP_ICV, and DR. The cutoff values defined for each variable are also displayed. IAPF, individual alpha peak frequency; ICV, intracranial volume; LIOL, left inferior occipital lobe; DR, delayed recall; LP, left parahippocampal volume