Supplemental Materials Eikelboom et al. Neuropsychiatric and cognitive symptoms across the Alzheimer's disease clinical spectrum

The objective of this study was to investigate the prevalence and trajectories of neuropsychiatric symptoms (NPS) in relation to cognitive functioning in a cohort of amyloid-β positive individuals across the Alzheimer's disease (AD) clinical spectrum. In this single-center observational study, we included all individuals who visited the Alzheimer Center Amsterdam and had 1) a clinical diagnosis of subjective cognitive decline (SCD), mild cognitive impairment (MCI), or probable AD dementia, and 2) were amyloid-β positive (A+). We measured NPS with the Neuropsychiatric Inventory (NPI), examining total scores and the presence of specific NPI domains. Cognition was assessed across five cognitive domains and with the MMSE. We examined trajectories including model based trends for NPS and cognitive functioning over time. We used linear mixed models to relate baseline NPI scores to cognitive functioning at baseline (whole-sample) and longitudinal time-points (subsample n=520, Mean=1.8 [SD=0.7] years follow-up). We included 1,524 amyloid-β positive individuals from the Amsterdam Dementia Cohort with A+ SCD (n=113), A+ MCI (n=321), or A+ AD dementia (n=1,090). NPS were prevalent across all clinical AD stages (≥1 NPS 81.4% in SCD, 81.2% in MCI, 88.7% in dementia; ≥1 clinically relevant NPS 54.0% in SCD, 50.5% in MCI, 66.0% in dementia). Cognitive functioning showed an uniform gradual decline; while in contrast, large intra-individual heterogeneity of NPS was observed over time across all AD groups. At baseline, we found associations between NPS and cognition in dementia that were most pronounced for NPI total scores and MMSE (range β=0.18 to 0.11, FDR-adjusted p0.05). There were no associations between baseline NPS and cognitive functioning over time in any clinical stage (range β=-0.13 to 0.44, all FDR-adjusted p>0.05). NPS and cognitive symptoms are both prevalent across the AD continuum, but show a different evolution during the course of the disease.Data not provided in the article and additional information on methods and materials can be shared upon reasonable request. Funding provided by: Alzheimer NederlandCrossref Funder Registry ID: http://dx.doi.org/10.13039/501100010969Award Number: 733050823Funding provided by: The Amsterdam UMC Alzheimer Centre *Crossref Funder Registry ID: Award Number: Funding provided by: WF*Crossref Funder Registry ID: Award Number: Funding provided by: The Amsterdam UMC Alzheimer CentreCrossref Funder Registry ID: Funding provided by: WFCrossref Funder Registry ID:We included all patients who visited Alzheimer Center Amsterdam between June 2002 and December 2017 and had 1) a clinical diagnosis of subjective cognitive decline (SCD), mild cognitive impairment (MCI), or probable Alzheimer's disease (AD) dementia, 2) were amyloid-β positive, and 3) had NPI and neuropsychological assessment available at baseline. This resulted in a total of 1,524 individuals of which 113 participants had a 64 clinical diagnosis of SCD, 321 participants with MCI, and 1,090 participants with AD 65 dementia at baseline. A subsample of the participants had follow-up assessments available: n=53 (46.9%) with 68 A+ SCD at baseline, n=142 (44.2%) with A+ MCI at baseline, and n=325 69 (29.8%) with A+ AD dementia at baseline. We compared baseline clinical characteristics, neuropsychiatric symptoms (NPS) prevalence, and cognitive performance across the diagnostic groups using ANOVA (with Tukey's HSD post hoc test) or Chi-squared tests where appropriate. We plotted individual trajectories of NPS and cognitive functioning over time according to disease stage and added model based trends with 95% confidence intervals to the graphs for descriptive purposes. In addition, we investigated the extent to which NPS and cognitive functioning changed over time within individuals (intra-individual variance) and between individuals (inter-individual variance). To quantify the variation within and between individuals, we conducted multilevel null models to obtain the percentage variance explained by intra-variance and inter-variance for neuropsychiatric measures and cognitive measures over time. For these analyses, the continuous severity x frequency scores (0-12) of specific Neuropsychiatric Inventory (NPI) domains were used. To study associations between baseline NPS and cognitive functioning at baseline and over time, we performed linear mixed models (LMMs) including random intercepts and fixed slopes that were corrected for age, sex, and education. Determinants included the NPI total score and the presence of specific NPI domains. Outcomes were performance on the MMSE and the five predefined cognitive domains. LMMs were run separately for the clinical stages at baseline (i.e. SCD, MCI, dementia). We tested non-linear associations using LMMs with quadratic and cubic splines and selected linear LMM for all models based on the Likelihood-Ratio Chi-Squared Test and Akaike information criterion. We checked assumptions by visual inspection of standardized residuals scatterplots and Q-Q plots. As normality of cognitive scores slightly deviated in language and visuospatial abilities most pronounced in A+ SCD, we conducted sensitivity analyses using a bootstrap procedure with 200 bootstrap samples to calculate confidence intervals. This approach did not change the initial findings. Level of significance was set at p