“We controlled for race and ethnicity...” Recommendations for the responsible use and communication of race and ethnicity in neuroimaging research

Abstract

The growing availability of large population neuroimaging datasets provides researchers with unique opportunities to conduct rigorous and impactful studies on brain and behavioral development. Increasing sample sizes and participant diversity provide researchers with a more comprehensive view of neurodevelopmental processes among more diverse populations than those that have traditionally been studied in more homogeneous samples. However, patterns observed in these types of large datasets are also more likely to be impacted by, and thereby reproduced in their dissemination, upstream forces of inequities (i.e., structural racism) that contribute to disparities in health behaviors and outcomes based on race, ethnicity, and social class. Despite their importance, there is limited guidance for neuroscientists on conceptualizing, contextualizing, and communicating about topics of race and ethnicity. This paper aims to discuss considerations regarding use and communication of race and ethnicity in research and serve as a starting point for researchers to evaluate their own practices around race and ethnicity topics. We provide recommendations for researchers to reflect on race and ethnicity choices in study design, model specification, statistical analysis, and communication of results, implement practices to avoid the further stigmatization of historically minoritized groups, and engage in research practices that counteract existing harmful biases.

Neighborhood Disadvantage and Neural Correlates of Threat and Reward Processing in Survivors of Recent Trauma

“…Ethnoracial group was not included in any analyses given that the variable serves as proxy for racism-related stress and inequitable exposures, including income and neighborhood disadvantage (eFigure 2 and eFigure 3 in Supplement 1). 37,38 General linear models (GLMs) were used to test whether ADI was associated with task-based activity in ROIs (5 threat related and 5 reward related) and included income, lifetime trauma, sex, and age as covariates. A Holm-Bonferroni correction was applied for each set of models (eg, models examining threat-related ROIs).…”

“…Ethnoracial group was not included in any analyses given that the variable serves as proxy for racism-related stress and inequitable exposures, including income and neighborhood disadvantage (eFigure 2 and eFigure 3 in Supplement 1 ). 37 , 38 …”

Transparency and Reproducibility in the Adolescent Brain Cognitive Development (ABCD) Study

“…More thoughtful consideration of variable relationships (e.g., using directed acyclic graphs) can help determine whether statistical adjustment is necessary (51). The habitual inclusion of sociodemographic variables (e.g., race and/or ethnicity) in statistical models has also been criticized for perpetuating stigmatization and inequity (52). We strongly encourage that researchers provide support for the inclusion of all covariates in their statistical models.…”

“…Researchers should determine whether inclusion, omission, or some other method (e.g., effect measure modification) is necessary when considering the use of race or ethno-racial variables. Researchers should consult recommended best practices (51)(52)(53) prior to using race and/or ethnicity variables in their analyses.…”

Building Towards an Adolescent Neural Urbanome: Expanding Environmental Measures using Linked External Data (LED) in the ABCD Study

“…For example, researchers studying neighborhood deprivation should also discuss the role that systemic racism may play in their associations. Further discussion of these issues can be found in Webb, Cardenas-Iniguez, andDouglas (2022), andGonzalez (2023).…”