2.1 Participants
The used study population was acquired from the Alzheimer's Disease Neuroimaging Initiative (ADNI) cohort (http://adni.loni.usc.edu). Data for this longitudinal study is acquired across 50 different sites within the United states of America and Canada. The principal investigator Michael W. Weiner, MD, VA Medical Centre and University of California – San Francisco initiated this cohort in 2003. The goal of this study has been mapping out the progress of AD by using different forms of imaging, such as MRI and PET, other forms of biological markers, and clinical and cognitive testing.
A description of the clinical aspect of the ADNI study has been published in other studies (REFERENTIES 3). A list of inclusion and exclusion criteria can be found the ADNI webpage describing the full protocol (http://adni.loni.usc.edu/wp-content/uploads/2010/09/ADNI_GeneralProceduresManual.pdf). Amongst the inclusion criteria were fluency in English or Spanish, the ability to be scanned in a MRI scanner, no psychiatric history etc.
For this particular study, we selected the participants who had IQ scores and neuropsychological testing available. In total, 269 participants of ADNI were included in this research with 147 mildly cognitive impaired (MCI) subjects and 122 cognitively normal (CN) subjects. All of these participants have MRI data, however and 182 participants have PET amyloid data. However, only 129 subjects of the 269 participants have PET tau data and 130 subjects have amyloid and tau cerebrospinal fluid(CSF) data.
For subjects to be qualified as MCI, memory concerns had to be present but should show no significant functional impairment. As far as cognitive testing, the Mini-Mental State Examination (MMSE) (REFERENTIE) score had to be between 24 and 30, a CDR score of 0.5, and mild impairments on objective cognitive measures (Wechsler Memory Scale – Logical Memory II (REFERENTIE)). CN had the same MMSE score range and a CDR score of 0. Furthermore, they did not qualify for either MCI or AD.
2.2. Neuropsychological assessment
To estimate neuropsychological performance, memory, language, visual-spatial, executive function and MMSE scores were looked at. Composit scores were created for each of the domains, except for MMSE. (link to ADNI protocols page is sufficient?? Or you can look there and type up a summary)
2.3. Cognitive reserve assessment
2.3.1. Cognitive Resilience
The proxy measures used for cognitive resistance were sex, age, and pre-morbid verbal IQ. The American National Adult Reading Test intelligence quotient (AMNART IQ) was used for the IQ measurement (Bright et al., 2002). The test consised of 50 words, and the difficulty increased the further the participants got on the list. All the words were irregular and therefore the participant must have had a former knowledge of the words, otherwise the participants would not be able to know how to say the word. An example word is, hiatus. The IQ score was calculated as followed, Predicted full-scale IQ = 128 – 0.83 x NART error score (S.E. est. = 7.6). The AMNART test is often used as a proxy measure for cognitive resistance (Stern, 2012).
2.3.2. Cognitive Resistance
The proxy measures used for cognitive resilience were occupation and years of education. The occupations were categorized by using the Educational and Occupational Level Questionnaire. The questionnaire categorized the occupations into 7 categories: 7) higher executives/ proprietors of large concern, 6) Business managers of medium-sized business, 5) Administrative personnel small business owners minor professional, 4) Clerical and sales workers technicians owners of little business, 3) Skilled manual employee, 2) Machine operator, and 1) Other employee. The occupations were manually categorized.
2.4. Disease burden
As discussed before, the A/T/N classification system was used to determine disease burden. Different measurements can be used for the classification, and the measurements that were chosen were, for A, CSF amyloid beta and PET amyloid beta, for T, CSF tau and PET tau, and for N, we used structural MRI data to determine the neurodegeneration.
• Protein Deposition
o CSF
For a full and elaborate explanation of the collection and processing of the CSF data, the ADNI protocol can be advised (http://adni.loni.usc.edu/wp-content/uploads/2010/09/ADNI_GeneralProceduresManual.pdf). As part of the study, a lumbar puncture (LP) was obtained by a subset of participants. The CSF was extracted by using 24 or 25 gauge Sprotte needles and a total of 21 to 22 ml of CSF was collected and kept in polypropylene tubes (Sarstedt; Numbrecht, Germany). The INNO-BIA AlzBio3 (Innogenetics; Ghent, Belgie) method was used to analyse the extracted CSF, and provided measures for Aβ142, total tau and phosphorylated tau.
♣ Amyloid
♣ Tau
o PET
♣ Amyloid
♣ Tau (early vs. late)
• Atrophy
The chosen measurement for atrophy was cortical thickness for every subject. A volumetric 3D T1-weighted magnetization – prepared rapid gradient-echo (MP RAGE) was collected for each subject. This was done on a 1.5T or a 3T scanner, with a voxel size of 1.1 X 1.1 X 1.2mm3 (?)(Ekman, Ferreira & Westman, 2018). For a more elaborate explanation, the ADNI1 protocol on MRI acquisition can be consulted (http://adni.loni.usc.edu/wp-content/uploads/2010/09/ADNI_MRI_Tech_Proc_Manual.pdf).
2.6. Statistical analysis
The data was analysed with Statistical Package for Social Science (SPSS) 24.0 (International Business Machines Corp., Armonk, NY, USA). The distribution of the demographic characteristics was assessed with
To address our first hypothesis, we checked the relationship between neuropathology and neurocognitive performance. As we expected neuropathology to play a predictive role in one's performance on neurocognitive tests, a linear regression analysis was executed. Firstly, correlations coefficients (Pearson) between all the variables were addressed, with neuropathology having three variables (tau, amyloid, and cortical thickness) and with neurocognitive performance having five variables (memory, language, visual-spatial, executive functioning, and MMSE scores). By paring them all up, 15 different models were created and the linear regression was performed on all the models.
To complete the first hypotheses, we added proxy measures of CR as a mediation factor. Both proxy measures for resistance (IQ, sex, and age) and for resilience (occupation and years of education) will be used as the mediating factors. To do the mediation analysis, we look at the linear regressions models. The linear regression consisted of two variables, on the dependent variable end, there is a neurocognitive score and on the independent end, we have a form of neuropathology. Within SPSS a second "block" can be added to a linear regression analysis, which allowed us to create a second model within the existing model. The first model will explain the variance explained by solely the neuropathological variable, and the second model will explain the variance explained after adding in the mediators. Again, 15 different linear regressions were done and all the regressions existed of two models. This analysis was done twice, once with the CSF measures for Tau and Amyloid and the second time it was done with the imaging measures for Tau and Amyloid. In both analyses, cortical thickness was used from MRI data acquired on corresponding dates.
To assess the second hypothesis, we simply added a third model within the linear regression analysis. The first model will account for the variance explained by a measure of neuropathology, the second model will account for the added variance explained by the proxy measures for resistance (IQ, sex, and age), and the third model will account for the added variance explained by the proxy measures for resilience (occupation and years of education). By distinguishing CR into two different models, we can look separately at the mediating effect of resistance and resilience on the predictive role neuropathology plays on neurocognitive performance.