Task functional MRI measurements—good for understanding the average human brain but may not be reliable for predicting individual differences

In a recent publication, researchers funded by NIA, NIH Blueprint for Neuroscience Research, NSF, and others cast doubt on the usefulness of task functional MRI (fMRI) for between-subject effect predictions and biomarker development. Recently, there has been an increasing interest in identifying brain biomarkers of disease risk for personalized medicine. One limitation to identifying meaningful biomarkers is measurement reliability, such that a measure provides consistent results under similar circumstances. Measures with low reliability are not suitable as biomarkers or for predicting clinical health outcomes. Task fMRI has been used for decades for measuring brain activity to understand the “average” brain; however, the reliability of task fMRI has not been systematically evaluated. In this study, the researchers performed a meta-analysis of 90 experiments and a test-retest of reliability of brain activity in a priori regions of interest (ROIs) collected from two large imaging studies—the Human Connectome Project and the Dunedin Study.

The researchers reexamined 56 published papers and conducted a meta-analysis of the test-retest reliability of regional activation in task fMRI across 90 experiments (N = 1,008). In the second experiment, they used two recently collected datasets and analyzed the test-retest reliability of brain activation in a priori ROIs across 11 commonly used fMRI tasks collected by the Human Connectome Project (N = 45) and the Dunedin Multidisciplinary Health and Development Study (N = 20). Test-retest reliability is widely quantified using the intraclass correlation coefficient (ICC), which represents the proportion of a measure’s total variance that is accounted for by variation among individuals. Measures with ICCs of less than 0.40 are thought to have poor reliability, between 0.40 and 0.60 have fair reliability, 0.60 and 0.75 have good reliability, and greater than 0.75 have excellent reliability. An ICC greater than 0.80 is considered a clinically required standard for reliability in the field of psychology.

The results of the meta-analysis showed that across the 90 experiments, there was poor overall reliability (mean ICC = 0.397). Additionally, the test-retest reliabilities of activity in a priori regions of interest across 11 common fMRI tasks collected by the Human Connectome Project and the Dunedin Study were poor. For the Human Connectome Project, six out of seven measures of brain function, the correlation between tests taken about 4 months apart with the same person, was weak (mean ICC = 0.251). The seventh measure studied—language processing—was only a fair correlation (ICC = 0.485). In the Dunedin Study there was also low reliability for each of the four tasks from one test to the next in a person (mean ICC = 0.309). To provide a benchmark for evaluating the test-retest reliability of task fMRI, the researchers looked at the reliability of three commonly used structural MRI measures: cortical thickness, surface area, and subcortical gray-matter volume. Consistent with prior research, they found that the structural MRI phenotypes have excellent reliability (ICCs > 0.9).

These data indicate that commonly used task fMRI measures do not currently meet the minimal reliability standards needed for brain biomarker discovery or for individual differences research. The researchers suggest this is likely due to the current task fMRI paradigms being largely descended from the experimental discipline, where the paradigms are intentionally designed to reveal how the “average” human brain responds to stimuli, while reducing between-subject variance. Using this robust, group-level paradigm and converting it into tools for assessing individual differences can lead to the misinterpretation that robust within-subjects effects imply between-subjects reliability. This is similar to what has been seen in other areas of research. Behavioral measures that elicit robust within-subjects (i.e., group) effects have been shown to have low between-subjects reliability; for example, previous research has shown that the Stroop test’s mean test-retest reliability (ICC = 0.45) is similar to the mean reliability of this study’s task fMRI meta-analysis (ICC = 0.397). The researchers further explain that it is not that MRI measures are unreliable since both structural MRI measures in the present study demonstrate high test-retest reliability. There are potential solutions to the task fMRI measures reliability problem, such as increasing the duration of data collection (e.g., for 1 hour or longer in the scanner instead of 5 minutes), or the development of new tasks with the explicit purpose of reliably measuring individual differences in brain activity.

Citation:
Elliott ML, Knodt AR, Ireland D, Morris ML, Poulton R, Ramrakha S, Sison ML, Moffitt TE, Caspi A, Hariri AR. 2020. What is the Test-Retest Reliability of Common Task-fMRI Measure? New Empirical Evidence and a Meta-Analysis. Psychol Sci. doi: 10.1177/0956797620916786. Online ahead of print.

Children from disadvantaged neighborhoods have altered regulation of genes related to chronic inflammation, tobacco smoke, air pollution, and lung cancer as adults

What is the connection between your childhood neighborhood, gene regulation, and later health outcomes? A study funded by the NICHD, NIEHS, and others examined the different ways DNA methylation may be influenced by one's childhood social environment. Previous studies have shown that children raised in socioeconomically disadvantaged neighborhoods have worse health as adults when compared with their peers from more affluent communities. Environmentally induced alterations to the epigenome, such as differential DNA methylation, have been proposed as one potential mechanism linking early-life environments to later-life health outcomes. Although previous studies have linked individual-level socioeconomic factors and differential DNA methylation patterns, little is known about how characteristics of the wider neighborhood environment are associated with epigenetic differences.

In this study, the researchers replicated and expanded previous findings about neighborhood characteristics and DNA methylation, using data from the Environmental Risk (E-Risk) Longitudinal Twin Study, a nationally representative birth cohort of children born between 1994 and 1995 in England and Wales and followed up from age 5 to 18 years (N = 1,619; 806 female). Multiple aspects of the participants’ neighborhoods were measured across childhood and adolescence, indexing neighborhood deprivation, dilapidation, disconnection, and dangerousness. The neighborhood assessments data was then integrated with measures of DNA methylation in whole blood drawn at age 18 to test the hypothesis that children raised in more socioeconomically disadvantaged neighborhoods show differential methylation patterns in young adulthood compared with their peers raised in more advantaged neighborhoods. Associations between neighborhood socioeconomic disadvantage and methylation were tested using three prespecified approaches: (1) testing probes annotated to candidate genes involved in stress reactivity–related and inflammation-related genes; (2) polyepigenetic scores indexing differential methylation in phenotypes of obesity, inflammation, and smoking associated with growing up in disadvantaged neighborhoods; and (3) a more exploratory epigenome-wide association study.

Their findings indicate that children raised in socioeconomically disadvantaged neighborhoods exhibit differential DNA methylation in genes connected to inflammation (β = 0.12; P < .001) and smoking (β = 0.18; P < 0.001) but not obesity (β = 0.05; P = 0.12). The epigenome-wide association study identified multiple sites at an array-wide significance (P < 1.16 × 10^−7) in genes involved in the metabolism of hydrocarbons. Associations between neighborhood disadvantage and methylation were small but robust to family-level socioeconomic factors and to individual-level tobacco smoking.

This study indicates that children raised in more socioeconomically disadvantaged neighborhoods appeared to enter young adulthood epigenetically distinct from their less-disadvantaged peers. Potentially, these children may enter adulthood wired at the cellular level for different future health outcomes than their more affluent peers and may help explain how long-term health disparities among communities emerge. This finding suggests that epigenetic regulation may be a mechanism by which the childhood neighborhood environment can impact adult health. Further research is still needed to determine whether these differences are lasting or modifiable.

Citation:
Reuben A, Sugden K, Arseneault L, et al. 2020. Association of Neighborhood Disadvantage in Childhood with DNA Methylation in Young Adulthood. JAMA Netw Open. 3(6):e206095.

A healthier lifestyle may lead to a 60% lower risk of Alzheimer’s dementia

Within the United States, approximately 5.7 million people are living with dementia. In a recent study funded by the NIA, researchers sought to determine the behavioral correlates that may help to delay the onset of this devastating disease. Currently, there are no effective pharmacotherapies that modify the course of dementia; however, epidemiologic studies and clinical trials suggest that primary prevention, typically through increasing healthy behaviors, may be able to delay the onset of the disease. The goal of the current study was to comprehensively assess the specific healthy lifestyle factors that are connected to the onset or delay of Alzheimer’s dementia.

The researchers used two prospective longitudinal studies: the Chicago Health and Aging Project (CHAP; n = 1,845) and the Rush Memory and Aging Project (MAP; n = 920). More specifically, the authors defined a healthy lifestyle score on the basis of five healthy lifestyle factors:(1) nonsmoking status, (2) exercise (≥ 150 minutes per week of moderate or vigorous-intensity physical activity), (3) alcohol consumption (low-moderate), (4) diet (high-quality Mediterranean-DASH Diet Intervention for Neurodegenerative Delay diet [upper 40 percent]), and (5) engagement in late-life cognitive activities (upper 40 percent), giving an overall score ranging from 0 to 5. They then estimated Cox proportional hazard models for each cohort to estimate the hazard ratio (HR) and the 95 percent confidence interval (CI) of the lifestyle score with Alzheimer dementia, and a random-effect meta-analysis was used to pool the results.

During a median follow-up of 5.8 years in CHAP and 6.0 years in MAP, 379 and 229 participants, respectively, had incident Alzheimer’s dementia. In multivariable-adjusted models, the pooled HR of Alzheimer’s dementia across two cohorts was 0.73 per each additional healthy lifestyle factor. Compared to participants with zero to one healthy lifestyle factor, the risk of Alzheimer’s dementia was 37 percent lower (pooled HR = 0.63) in those with two to three healthy lifestyle factors and 60 percent lower (pooled HR = 0.40) in those with four to five healthy lifestyle factors.

This study provides compelling evidence that an increased number of healthy lifestyle behaviors are associated with a lower risk of developing Alzheimer’s dementia. Specifically, individuals that adhered to four or all five of the specified healthy behaviors had a 60 percent lower risk of Alzheimer's. However, the causal mechanisms underlining the protective effects of healthy lifestyles in Alzheimer’s dementia are not entirely understood. Some theories include that positive behaviors may initiate a chain of metabolic and molecular alterations that may reduce inflammation processes and oxidative stress, and, in turn, could reduce Alzheimer’s disease–associated changes in the brain. That said, specific insights into the precise pathways involved are limited and warrant further investigation.

Citation:
Dhana K, Evans DA, Rajan KB, Bennett DA, Morris MC. 2020. Healthy lifestyle and the risk of Alzheimer dementia Findings from 2 longitudinal studies. Neurology. 95:1-10. doi:10.1212/WNL.0000000000009816