2 Early onset APOE Ɛ 4-negative Alzheimer ’ s disease patients show faster cognitive decline on non-memory domains Submitted

Objective: Age-at-onset and APOE Ɛ4 genotype have been shown to influence clinical manifestation of Alzheimer’s disease (AD). We investigated rate of decline in specific cognitive domains according to age at onset and APOE Ɛ4 genotype. Methods: 199 patients with probable AD underwent at least two annual neuropsychological assessments. Patients were classified according to age-at-onset (≤65 years vs >65 years) and APOE Ɛ4 genotype (positive vs negative). The neuropsychological battery compromised tests for memory, language, attention, executive and visuo-spatial functioning. For each domain compound z-scores were calculated, based on baseline performance of patients. Average duration of follow-up was 1.5±1 years. We used linear mixed models (LMM) to estimate effects of age, APOE and age*APOE on cognitive decline over time. Results: At baseline, patients were 65±8 years, 98(49%) were female and MMSE was 22±4. LMM showed that early onset patients declined faster on executive functioning (β±SE:-0.09±0.06) than late onset patients, but age was not related to decline in other cognitive domains. APOE Ɛ4-negative patients declined faster on language than APOE Ɛ4-positive patients (β±SE:-0.1±0.06). When we took age and APOE genotype into account simultaneously, we found that compared to late onset-Ɛ4 positive patients, early onset-Ɛ4 negative patients declined faster on language (β±SE:-0.36±0.1), attention (β±SE:-0.42±0.1), executive (β±SE:-0.41±0.1) and visuospatial functioning (β±SE:-0.43±0.1). Late onset-Ɛ4 negative and early onset-Ɛ4 positive patients showed intermediate rates of decline. We found no differences in decline on memory. Conclusion: We found that patients who develop AD despite absence of the two most important risk factors, show steepest cognitive decline on non-memory cognitive domains. Introduction Alzheimer’s disease (AD) is increasingly being considered a disease in which the clinical presentation between patients may differ [1,2]. One way to look at heterogeneity of AD is by analysing rate of decline. By definition, patients with AD show cognitive decline over time, but the rate of deterioration is highly variable between individuals. Age and the apolipoprotein (APOE) Ɛ4 gene are risk factors for AD. Moreover, both factors have been shown to modify clinical presentation and rate of cognitive decline in AD [2]. Lower age at onset and absence of the APOE Ɛ4 allele have been associated with different rates of cognitive decline [2,3]. In general, a lower age at onset seems to be associated with faster progression [2,4]. The influence of APOE Ɛ4, however, is less clear. Some studies found that presence of the Ɛ4 allele was associated with faster cognitive decline, while others found a slower cognitive decline, or concluded that there was no influence of APOE Ɛ4 on rate of cognitive decline at all [5-10]. Studies investigating both age at onset and APOE Ɛ4 in AD are rare. A recent longitudinal study assessing both age and APOE Ɛ4 had a limited follow up duration of one year and focused mainly on memory performance [11]. They found that younger patients who were APOE Ɛ4-positive showed steeper decline. By contrast, in an earlier study by our group we showed more rapid global cognitive decline in early onset than in late onset AD and this was most prominent in APOE Ɛ4-negative patients [3]. It remains unclear, however, how age at onset and APOE Ɛ4 status influence decline of specific cognitive domains other than memory. In this longitudinal study in AD patients, we therefore aimed to investigate the rate of decline in memory, language, attention, executive and visuo-spatial functioning, according to age at onset and APOE Ɛ4 status. Methods Subjects We included 199 patients with a diagnosis of probable AD and a minimum of two neuropsychological evaluations (at least one year apart) from the Amsterdam Dementia Cohort between January 2008 and December 2011 [12]. All patients underwent a standardized one-day assessment including medical history, informant-based history, physical and neurological exam including Clinical Dementia Rating (CDR), neuropsychological assessment, laboratory tests, magnetic resonance imaging (MRI) of the brain and electroencephalogram (EEG). Age at diagnosis of 65 years or younger was considered as early onset AD. The duration of the cognitive complaints as reported by the patient and/or caregiver was recorded to estimate the disease duration at time of diagnosis. Diagnoses were made in a multidisciplinary consensus meeting using international diagnostic consensus criteria. The diagnosis of probable AD patients was made using the criteria of McKhann [13,14]. Level of education was classified according to the system of Verhage ranging from 1 to 7 (low to highly educated) [15]. The local Ethics Review Board approved the study and all patients gave written informed consent for their clinical data to be used for research purposes. Neuropsychological assessment Cognitive functions were assessed with a standardized test battery. We used the MMSE as a measure for global cognitive decline [16]. For memory, we used the Visual Association Test (VAT) and total immediate recall and delayed recall of the Dutch version of the Rey auditory verbal learning task (RAVLT) [17-19]. To examine language, we used VAT naming, category fluency (animals), the Dutch version of Controlled Oral Word Association Test (COWAT) (letter fluency), comparative questions and naming of the Arizona Battery for Communication Disorders (ABCD) [17,20-22]. For attention we used Trail Making Test (TMT) A and the forward condition of Digit Span (extended version) [23,24]. We used TMT B, the backwards condition of Digit Span (extended version) and the Frontal Assessment Battery (FAB) to examine executive functioning [23-25]. We used three subtests of the Visual Object and Space Perception Battery (VOSP) to assess visuo-spatial functioning, namely (i) incomplete letters, (ii) dot counting and (iii) number location [26]. Additionally, the Geriatric Depression Scale (GDS) was assessed [27]. TMT A and B scores were log-transformed because they were not normally distributed. TMT A and B scores were logtransformed due to non-normal distribution, and inverted by computing the score by -1, because higher scores imply a worse performance. Follow up At follow-up, all subjects underwent physical and neurological exam, and a repeated neuropsychological evaluation. For the total sample, the median number of neuropsychological assessments was 2 (range 2-4) and the mean duration of follow-up was 1.5±1 years. APOE DNA was isolated from 10 ml blood samples in ethylenediaminetetraacetic acid (EDTA). APOE Ɛ4 genotype was determined at the Neurological Laboratory of the Department of Clinical Chemistry of the VUmc with the LightCycler APOE mutation detection method (Roche Diagnostics GmbH, Mannheim, Germany). APOE Ɛ4 data were available for 181 patients (early onset: N=100; late onset: N=81) and were analysed according to the presence or absence of an APOE Ɛ4 allele. Statistical analysis PASW Statistics 20.0 for Mac was used. For baseline demographics and raw neuropsychological data, χ-tests, independent samples T-test and Univariate Analysis of Variance (ANOVA) were performed when appropriate. ANOVA’s were conducted with age at onset (≤65 vs > 65) or APOE Ɛ4 status (negative versus positive) as between-subjects factor and neuropsychological test as dependent variable. Sex, education and age (when appropriate) were entered as covariates. To obtain unbiased estimation of cognitive domain scores, we imputed missing neuropsychological test scores by multiple imputation of individual test scores in PASW. The method we used was predictive mean matching, because of the non-Gaussian distribution of some of the tests. Predictors for imputation were age (at time of neuropsychological assessment), gender, education, diagnosis, CDR-score, GDS-score and all available neuropsychological tests. Fifteen imputed data sets were created. Neuropsychological baseline tests were standardized into z-scores and based on these z-scores, we calculated z-scores for each follow up for each patient, relative to baseline z-scores. Next, we computed compound z-scores for memory, language, attention, executive functioning and visuo-spatial functioning for each imputed dataset. We report pooled statistics over 15 imputed data sets. Linear mixed models with an unstructured covariance pattern were used to assess associations between diagnosis and baseline cognition and cognitive performance over time. First, we performed a model including terms for age, time and the interaction between age and time. Second, we performed a model with terms for APOE Ɛ4 status, time and the interaction between APOE Ɛ4 status and time. In both models random effects were subject-ID and time and outcome measures were compound z-scores of the five cognitive domains. Beta±SE for diagnosis represents the association between age or APOE Ɛ4 status and baseline cognitive performance, whereas the interaction between age or APOE Ɛ4 status and time represents the association between group membership and cognitive performance over time. Next, we combined age and APOE Ɛ4 status to create a new four-level variable ‘age*APOE’. The levels were 1) late onset-Ɛ4 positive patients, 2) late onset-Ɛ4 negative patients, 3) early onset-Ɛ4 positive patients and 4) early onset-Ɛ4 negative patients. In an additional analysis, we evaluated the combined effect of age and APOE by including the newly constructed four-level variable Age*APOE as categorical term in the model (late onset-Ɛ4 positive patients as reference). We recoded the variable in order to estimate beta’s and SE’s for all levels of this variable. All analyses were corrected for gender and education, and age (when age at onset was not a factor in the analysis). Data of linear mixed models are presented as uncorrected beta±SE with p-values of the corrected models. Fo