3
Addressing Sample Representativeness to Enhance Study Reproducibility
Objective:
Across a range of scientific disciplines, the replicability and reproducibility of scientific findings are in question. One potential contributor to our failures in the behavioral sciences are related to the demographic homogeneity of our samples. In a paper by Henrich et al., (2010) the authors provide important reminder that much of what we have learned in the behavioral sciences is from a very small portion of the population from white educated, industrialized, and democratic (WEIRD) countries. This talk focuses on how team science can help address this issue by diversifying the voices involved in leading the science and addressing a critical issue regarding the make-up of the participants contributing to our knowledge base. Sample representativeness must be addressed in the study of TBI (and other neurological disorders) for advancement of interventions and to optimize outcome prediction. The ENIGMA-TBI working group provides a model for increasing the breadth of collaboration around the world and the opportunity to increase the representativeness of our sample with respect to social and biological determinants of health.
We summarize a recent collaborative effort through ENIGMA-TBI including over 300 individuals with moderate and severe TBI to examine risk for neurodegeneration and to directly replicate the "accelerated aging hypothesis". Given the team science approach we were able to also examine directly the effects of sex, SES and genetic risk (e.g., APOE status) on advancing brain age assessed using structural MRI scans.
We observed an accelerated aging affect but with important caveats only discernable with a sample this size. First, there were clear survivor biases, with older cohorts being more educated. Second, sex was a moderator of outcome, with men showing a more advanced brain-age gap after controlling for education. Finally, genetic risk defined as being at least heterozygous for the Apoe-4 allele did not influence outcome, contrary to predictions.
In summary, through collaboration and team science, this talk demonstrates how large scale data sharing can help to shine light on less studied characteristics in study samples such as sex and genetic risk. Open sciences offers unique opportunities to address the demographic and clinical heterogeneity in TBI (and many clinical samples) and to move the clinical neurosciences toward robust, reliable research.