Estimating and testing variance components in a multi-level GLM

Most analysis of multi-subject fMRI data is concerned with determining whether there exists a significant population-wide ‘activation’ in a comparison between two or more conditions. This is typically assessed by testing the average value of a contrast of parameter estimates (COPE) against zero in a general linear model (GLM) analysis. However, important information can also be obtained by testing whether there exist significant individual differences in effect magnitude between subjects, i.e. whether the variance of a COPE is significantly different from zero. Intuitively, such a test amounts to testing whether inter-individual differences are larger than would be expected given the within-subject error variance. We compare several methods for estimating variance components, including a) a naïve estimate using ordinary least squares (OLS); b) linear mixed effects in R (LMER); c) a novel Matlab implementation of iterative generalized least squares (IGLS) and its restricted maximum likelihood variant (RIGLS). All methods produced reasonable estimates of within- and between-subject variance components, with IGLS providing an attractive balance between sensitivity and appropriate control of false positives. Finally, we use the IGLS method to estimate inter-subject variance in a perfusion fMRI study (N = 18) of social evaluative threat, and show evidence for significant inter-individual differences in ventromedial prefrontal cortex (VMPFC), amygdala, hippocampus and medial temporal lobes, insula, and brainstem, with predicted inverse coupling between VMPFC and the midbrain periaqueductal gray only when high inter-individual variance was used to define the seed for functional connectivity analyses. In sum, tests of variance provides a way of selecting regions that show significant inter-individual variability for subsequent analyses that attempt to explain those individual differences.     

Maternal-fetal blood incompatibility and neuromorphologic anomalies in schizophrenia: Preliminary findings

Prior research has shown that maternal-fetal Rhesus (Rh) and ABO blood incompatibility increase the risk for schizophrenia. In the present study, the relationship between blood incompatibility and volumes of brain structures previously implicated in schizophrenia was assessed in schizophrenia cases and controls from a large birth cohort. Rh/ABO incompatible cases had significantly reduced cortical gray matter volume compared to compatible cases, a finding which appears to be driven by significant volume reductions in the dorsolateral prefrontal cortex and inferior frontal cortex. Larger hippocampal and putamen volumes were also observed in exposed controls compared to unexposed controls. Although the sample size is small and replications are required, these data suggest that maternal-fetal blood incompatibility may increase the risk for altered brain morphology in both schizophrenia and in controls. The findings also suggest that the larger hippocampal volume in exposed controls may indicate a mechanism of adaptive resilience which diminishes the risk that controls will develop schizophrenia.     

The epidemiology of community acquired bacteremic pneumonia, due to Streptococcus pneumoniae, in the Top End of the Northern Territory, Australia--over 22 years

Background

Diseases caused by Streptococcus pneumoniae continue to cause substantial morbidity and mortality throughout the world. Furthermore, detrimental outcomes are more pronounced in some populations—such as those living in third world poverty, and Indigenous people who live in developed nations.

Methods

This study describes the epidemiology of blood culture positive S. pneumoniae community-acquired pneumonia (CAP) in the Top End of the Northern Territory of Australia. Demographics, indigenous status, medical risk factors, serotype and outcomes were collected from adults presenting to hospital with blood culture positive S. pneumoniae CAP, from 1987 to 2008.

Results

We report 205 cases, with a median age of 40 years. The average overall incidence rate ratio was 10.3 for indigenous adults compared with non-indigenous adults. There was no statistical difference between incidence rates pre and post-23-valent pneumococcal polysaccharide vaccine (23vPPV) introduction. Serotypes in presenting cases were predominantly (84.7%) 23vPPV types. The whole-population logistic regression model identified significant adjusted relative risks: 95% CI, for age 45 and older 1.6: 1.1, 2.2, indigenous 5.9: 3.7, 9.5, diabetes 2.3: 1.6, 3.3, excess alcohol 4.8: 2.8, 8.3, smoking 2.7: 1.9, 3.7 with indigenous + excess alcohol 18.5: 17.3, 19.7 as predictive for bacteremic S. pneumoniae CAP presentation.

Conclusions

Our results suggest that, the national 23vPPV program appears to be under-utilized. An integrated Public Health approach vigorously targeting indigenous adolescents, before substances such as alcohol and smoking are habitual, together with increased vaccine coverage, will reduce the burden of pneumococcal disease in this population.     

Chemical characterization of chestnut cultivars from three consecutive years: Chemometrics and contribution for authentication

Four Castanea sativa Miller cultivars (Aveleira, Boaventura, Judia and Longal) belonging to the Protected Designation of Origin “Castanha da Terra Fria”, from the Northeast of Portugal, were selected in 2006, 2007 and 2008. Their nutritional, fatty acids, triacylglycerols and tocopherols profiles were evaluated. Water was the major component, followed by carbohydrates, protein and fat, with energetic values lower than 190 kcal/100 g of fresh fruit. Oleic, linoleic and palmitic were the major fatty acids, 1-oleoyl-2-linoleoyl-3-linoleoyl-sn-glycerol, 1-linoleoyl-2-linoleoyl-3-palmitoyl-sn-glycerol, 1-oleoyl-2-linoleoyl-3-oleoyl-sn-glycerol and 1-linoleoyl-2-oleoyl-3-palmitoyl-sn-glycerol were the prevalent triacylglycerols and γ-tocopherol was the most abundant tocopherol. In each parameter, differences between cultivars, harvest year and the possible cultivar × year interaction were screened through a two-way analysis of variance. Differences among cultivars have been attenuated by the variability among years, leading, in general, to a significant interaction effect, which resulted in a relative homogeneity regarding chemical parameters, showing that nutritional and chemical composition was influenced by seasonal variability. A stepwise linear discriminant model, based on 10 (α-tocopherol, γ-tocotrienol, LLL, OLLn, δ-tocopherol, γ-tocopherol, δ-tocotrienol, PLLn, protein and OOO) of the 38 initial evaluated variables was also established. The model allowed the complete discrimination of cultivars with overall sensibilities and specificities of 100%, for both original grouped data and leave-one-out cross-validation procedures. Furthermore, similar results were also obtained using only tocopherols data, showing their usefulness as a discriminant factor for chestnut cultivars.     

Temporally anticorrelated brain networks during working memory performance reveal aberrant prefrontal and hippocampal connectivity in patients with schizophrenia

Functional neuroimaging studies on cognitive dysfunction in schizophrenia have suggested regional brain activation changes in the dorsolateral prefrontal cortex and the medial temporal lobe. However, less is known about the functional coupling of these areas during cognitive performance. In this study, we used functional magnetic resonance imaging, a verbal working memory (WM) task and multivariate statistical techniques to investigate the functional coupling of temporally anticorrelated neural networks during cognitive processing in patients with schizophrenia (n = 16) compared to healthy controls (n = 16). Independent component analysis identified 18 independent components (ICs) among which two ICs were selected for further analyses. These ICs included temporally anticorrelated networks which were most highly associated with the delay period of the task in both healthy controls and patients with schizophrenia. Functional network abnormalities in patients with schizophrenia were detected within a “task-positive” lateral frontoparietal network, where increased functional connectivity was found in bilateral dorsolateral prefrontal regions. In addition, aberrant functional coupling of the hippocampal cortex in patients with schizophrenia was detected within a “task-negative” medial frontotemporal network. In patients with schizophrenia, functional connectivity indices in the left dorsolateral prefrontal cortex and the right hippocampal cortex were positively correlated with accuracy during the WM task, while the connectivity strength in the right dorsolateral prefrontal cortex was negatively correlated with measures of symptom severity. These data suggest that within two temporally anticorrelated network states, patients with schizophrenia exhibit increased and persistent dorsolateral prefrontal and hippocampal connectivity during WM performance.     

Interparticipant correlations: a model free FMRI analysis technique

FMRI analysis techniques can be broadly divided into model based and data driven techniques. The most widely used approach assumes an explicit temporal hemodynamic model based upon the experimental paradigm. Such an approach has proven very useful and powerful even though it is limited by the accuracy of the prespecified model. An alternative approach is to use data driven techniques like independent component analysis or fuzzy cluster analysis. These approaches have proven useful for exploratory analysis in a multivariate sense; however, they can present additional difficulties in the interpretation of the results. An alternative to these approaches is to take advantage of similarities in the patterns of the hemodynamics between participants [i.e., interparticipant correlation (IPC)]. This FMRI analysis technique enjoys the parsimony of the general linear model (GLM) but does not assume a specific FMRI time course. The technique consists of calculating voxel-wise correlations between participants resulting in IPC maps, which indicate the activated regions the participants have in common. We applied the IPC approach to data collected from healthy controls in an auditory oddball task. As expected, high inter-participant correlations were detected in auditory cortical regions in the temporal lobes where highest correlations were evident. In addition, areas that appear to be involved in the task were detected using IPC's but not the GLM regression. This technique, designed to have increased sensitivity to inter-subject correlations that are not necessarily task-related, may potentially be useful as a compliment to model-based approaches.     

Baseline activity predicts working memory load of preceding task condition

The conceptual notion of the so‐called resting state of the brain has been recently challenged by studies indicating a continuing effect of cognitive processes on subsequent rest. In particular, activity in posterior parietal and medial prefrontal areas has been found to be modulated by preceding experimental conditions. In this study, we investigated which brain areas show working memory dependent patterns in subsequent baseline periods and how specific they are for the preceding experimental condition. During functional magnetic resonance imaging, 94 subjects performed a letter‐version of the n‐back task with the conditions 0‐back and 2‐back followed by a low‐level baseline in which subjects had to passively observe the letters appearing. In a univariate analysis, 2‐back served as control condition while 0‐back, baseline after 0‐back and baseline after 2‐back were modeled as regressors to test for activity changes between both baseline conditions. Additionally, we tested, using Gaussian process classifiers, the recognition of task condition from functional images acquired during baseline. Besides the expected activity changes in the precuneus and medial prefrontal cortex, we found differential activity in the thalamus, putamen, and postcentral gyrus that were affected by the preceding task. The multivariate analysis revealed that images of the subsequent baseline block contain task related patterns that yield a recognition rate of 70%. The results suggest that the influence of a cognitive task on subsequent baseline is strong and specific for some areas but not restricted to areas of the so‐called default mode network. Hum Brain Mapp 34:3010–3022, 2013. © 2012 Wiley Periodicals, Inc.