Subjects with AD (by National Institute of Neurological and Communicative Disorders and Stroke - Alzheimer's Disease and Related Disorders Association (NINCDS-ADRDA)  criteria, n = 78), amnestic MCI (aMCI) (by Petersen criteria, n = 59), and cognitively healthy "normal controls" (n = 66) were recruited at the Memory Clinic, Department of Geriatrics, University Hospital, Basel, Switzerland, and underwent detailed neuropsychological, clinical, biomarker and imaging assessments at baseline and 12 months post-baseline. All study participants were aged ≥ 50 and had between 7 and 20 years of education. Subjects with AD and aMCI were current clinic patients. Controls were identified from the participants of the Basel Study on the Elderly (BASEL Project) described elsewhere; recruitment was stratified to ensure an even distribution of controls across each decade of age (50 to 59, 60 to 69, 70 to 79 and 80 to 89) and across genders . The study protocol was approved by the University of Basel Institutional Review board and written informed consent was obtained from each patient. The current study is a cross-sectional analysis of data collected at the baseline study visit.
Plasma aliquots, extracted from baseline blood samples and stored at -80°C, were transferred to Quest Diagnostics Inc. for Lp-PLA2 enzymatic activity measurement using an established colorimetric activity method (CAM) . The upper limit of valid measurement for the CAM assay is 300 nmol/min/ml and subjects (n = 6) with values greater than or equal to 300 nmol/min/ml were assigned a value of 300 nmol/min/ml. Markers of AD in CSF -- total Tau (T-Tau), phosphorylated Tau, 181P-epitope (P-Tau), and amyloid beta protein 42 (Aß42) -- were collected at the baseline visit and measured by ELISA at the University of Basel using the manufacturer's recommended protocols (Innogenetics NV, Ghent, Belgium).
Differences in mean Lp-PLA2 activity between diagnosis groups (AD, aMCI and normal controls) were explored initially using ANOVA. The primary comparison of interest was between AD and normal controls. The study was powered to detect a difference of 20 nmol/min/ml with 90% power and a two-sided alpha of 0.05, assuming a mean (SD) of 144 (36) nmol/min/ml in the normal controls .
Potential confounders or modifiers of the relationship between AD and aMCI and Lp-PLA2 were explored in a multiple linear regression model using backwards elimination with a retention criterion of P < 0.1. Covariates included in the models were statin use (yes/no), age, gender, body mass index (BMI), European Cardiovascular Society (ESC) cardiovascular risk score  of > 5%, history of diabetes type 1 or 2, history of heart disease, Hachinksi ischaemia score , and white matter changes (Scheltens  and Wahlund scores ). Statin use, age and gender were forced to remain in the model given demonstrated associations between these factors and dementia or Lp-PLA2 activity [18–20].
Lp-PLA2 is largely bound to LDL in the circulation, possibly through apolipoprotein B (apoB) 100 , and whether or not to adjust analytically for apoB and/or LDL in studies of Lp-PLA2 and cardiovascular outcomes is a matter for current scientific debate [3, 22]. While it is important to assess whether any observed associations of Lp-PLA2 with dementia may simply be proxies for an effect of LDL, controlling for LDL analytically could result in over-correction of the LpPLA2 values, obscuring a true association. To address this, the effect of LDL was explored by adjusting the final linear regression model arrived at through backwards elimination for LDL to assess whether this improved the model (assessed by comparison of model R2, BIC and AIC). Reporting models both with and without adjustment for LDL is an approach used in the cardiovascular field . Ten subjects with data missing for at least one covariate were dropped from the backwards elimination modelling.
The secondary analyses of the correlates of Lp-PLA2 activity were exploratory, and were not adjusted for multiple comparisons. Spearman's correlation coefficients were used to explore the association between Lp-PLA2 activity and (i) CSF markers of AD (Aß42, T-Tau and P-Tau), (ii) white matter changes (Scheltens score) and (iii) markers of cardiovascular disease and diabetes (LDL, high-density lipoprotein (HDL), total cholesterol:HDL ratio, homocysteine and haemoglobin A1c (HgbA1C)). Student's t-tests were used to test for differences in mean Lp-PLA2 by APOE ε4 genotype (positive (1 or 2 ε4 alleles) versus negative (0 ε4 alleles)). Multiple linear regression was used to adjust the APOE ε4 comparison in the AD group for covariates (statin use, heart disease and LDL; explored in separate models).
All analyses were performed using SAS software, Version 9.1 for Windows. Copyright, SAS Institute Inc. SAS and all other SAS Institute Inc. product or service names are registered trademarks or trademarks of SAS Institute Inc., Cary, NC, USA.