Search

Additional file 1: eFigure 1. Percentage of peak power for each month of the intervention for the high-intensity and moderate-intensity exercise groups. Each data point represents a participant’s mean peak power (from all attended exercise sessions) for that month. A VO2peak test following the third month was used for re-calculation of power. Corresponding months are different between groups (p < 0.001); whereby the high-intensity group had higher percentage peak power during all months of the intervention. Abbreviations: M1-6, Month 1 – 6; W, watts. eFigure 2. Linear relationship between change in cardiorespiratory fitness (residuals) and change in global cognition (residuals) from pre- to immediately post-intervention (6-months) in APOE ε4 carriers and non-carriers Abbreviations: APOE, Apolipoprotein E; VO2peak, peak aerobic capacity (fitness measurement). eTable 1. Beta coefficients and standard error of time*group from linear mixed models, and projected sample size for each group to detect a significant effect from longpower function in R statistical package. Supplementary methods.

Threshold determination using mixture modelling for the biomarkers (A, B) Aβ42, (C, D) pTau and (E, F) tTau.

Supplementary Methods. Information on handling of Aβ42 values above the measurement range and additional statistical information.

Threshold determination using mixture modelling for the biomarkers (A, B) Aβ42/Aβ40, (C, D) pTau/Aβ42 and (E, F) tTau/Aβ42.

CSF biomarkers versus PET SUVR for: (A) Aβ42 versus FLUTE SUVR, (B) Aβ42/Aβ40 versus FLUTE SUVR, (C) pTau/Aβ42 versus FLUTE SUVR, (D) Aβ42 versus FBP SUVR, (E) Aβ42/Aβ40 versus FBP SUVR and (F) pTau/Aβ42 versus FBP SUVR.

Results of ROC-AUC analysis – CSF biomarkers to predict Aβ-PET status, by PET tracer.

Results of ROC-AUC analysis – CSF biomarkers to predict Aβ-PET status in CN individuals.

Results of ROC-AUC analysis – CSF biomarkers to predict Aβ-PET status in CN individuals.

Results of ROC-AUC analysis – CSF biomarkers to predict Aβ-PET status, by PET tracer.

Background:The brain-derived neurotrophic factor (BDNF) Val66Met polymorphism Met allele exacerbates amyloid (Aβ) related decline in episodic memory (EM) and hippocampal volume (HV) over 36–54 months in preclinical Alzheimer's disease (AD). However, the extent to which Aβ+ and BDNF Val66Met is related to circulating markers of BDNF (e.g. serum) is unknown. We aimed to determine the effect of Aβ and the BDNF Val66Met polymorphism on levels of serum mBDNF, EM, and HV at baseline and over 18-months.Methods:Non-demented older adults (n = 446) underwent Aβ neuroimaging and BDNF Val66Met genotyping. EM and HV were assessed at baseline and 18 months later. Fasted blood samples were obtained from each participant at baseline and at 18-month follow-up. Aβ PET neuroimaging was used to classify participants as Aβ– or Aβ+.Results:At baseline, Aβ+ adults showed worse EM impairment and lower serum mBDNF levels relative to Aβ- adults. BDNF Val66Met polymorphism did not affect serum mBDNF, EM, or HV at baseline. When considered over 18-months, compared to Aβ– Val homozygotes, Aβ+ Val homozygotes showed significant decline in EM and HV but not serum mBDNF. Similarly, compared to Aβ+ Val homozygotes, Aβ+ Met carriers showed significant decline in EM and HV over 18-months but showed no change in serum mBDNF.Conclusion:While allelic variation in BDNF Val66Met may influence Aβ+ related neurodegeneration and memory loss over the short term, this is not related to serum mBDNF. Longer follow-up intervals may be required to further determine any relationships between serum mBDNF, EM, and HV in preclinical AD.

FFQ are commonly used to examine the association between diet and disease. They are the most practical method for usual dietary data collection as they are relatively inexpensive and easy to administer. In Australia, the Cancer Council of Victoria FFQ (CCVFFQ) version 2 and the online Commonwealth Scientific and Industrial Research Organisation FFQ (CSIROFFQ) are used. The aim of our study was to establish the level of agreement between nutrient intakes captured using the online CSIROFFQ and the paper-based CCVFFQ. The CCVFFQ and the online CSIROFFQ were completed by 136 healthy participants. FFQ responses were analysed to give g per d intake of a range of nutrients. Agreement between twenty-six nutrient intakes common to both FFQ was measured by a variety of methods. Nutrient intake levels that were significantly correlated between the two FFQ were carbohydrates, total fat, Na and MUFA. When assessing ranking of nutrients into quintiles, on average, 56 % of the participants (for all nutrients) were classified into the same or adjacent quintiles in both FFQ, with the highest percentage agreement for sugar. On average, 21 % of participants were grossly misclassified by three or four quintiles, with the highest percentage misclassification for fibre and Fe. Quintile agreement was similar to that reported by other studies, and we concluded that both FFQ are suitable tools for dividing participants’ nutrient intake levels into high- and low-consumption groups. Use of either FFQ was not appropriate for obtaining accurate estimates of absolute nutrient intakes.