To eliminate the effect of psychology, Spearman's correlations were calculated between cognitive tests and psychology tests, and we did not find any correlation between cognitive tests and severity of psychological impairment.
To compare the correlations between cognitive tests and angular deviation, we computed t statistics for the differences between correlation coefficients.
This design allowed adjustment for the high inter-correlation between cognitive tests and between waves whilst allowing investigation of effects specific for each test or at each wave after adjustment for the general effect.
Multivariate linear regressions models were performed at baseline separately in DM and NGT participants to test the relationship between cognitive test scores CCS and MMSE) and tertiles of total fat mass, total fat mass after adjusting multiple confounders.
There were no significant associations between cognitive test performance and treatment with lithium and anticonvulsants.
To identify associations between cognitive test scores and MetS, HOMA-IR, adiponectin, and CRP, we performed regression analyses using the SOLAR software version 4.1.0 [ 33].
Examining the patient group, there were no significant associations between cognitive test performance and MADRS score, history of psychosis or ADHD diagnosis.
In addition, stepwise regression analysis was performed to determine correlations between cognitive test performances and regional cerebral glucose metabolism as measured by F-FDG PET in various brain areas during the progression of disease (age 53 58 years).
Because the association between cognitive test scores and BC exposure estimates appeared log-linear, we log-transformed BC for use in all analyses and report associations for a doubling in BC concentration on the natural scale, or approximately a 0.69 unit change in ln BC).
