Reproducible brain-wide association studies require thousands of individuals.

Area of Science:

• Neuroimaging
• Cognitive Neuroscience
• Psychiatric Neuroscience

Background:

• Magnetic resonance imaging (MRI) has advanced brain structure and function mapping but faces challenges in mental health research.
• Replicating associations between individual brain differences and complex phenotypes (brain-wide association studies, BWAS) is difficult with small sample sizes.

Purpose of the Study:

• To quantify brain-wide association study (BWAS) effect sizes and reproducibility based on sample size.
• To identify factors influencing the reproducibility of neuroimaging-based brain-behavior associations.

Main Methods:

• Utilized three large neuroimaging datasets totaling approximately 50,000 individuals.
• Quantified BWAS effect sizes and replication rates as a function of increasing sample size.
• Compared reproducibility for functional MRI vs. structural MRI, cognitive tests vs. mental health questionnaires, and multivariate vs. univariate methods.

Main Results:

• BWAS associations are smaller than anticipated, leading to underpowered studies, inflated effect sizes, and replication failures at typical sample sizes.
• Replication rates improve and effect size inflation decreases as sample sizes increase into the thousands.
• Functional MRI, cognitive tests, and multivariate methods showed more robust BWAS effects.

Conclusions:

• Small effect sizes and population variability contribute to widespread BWAS replication failures.
• Reproducible BWAS require substantially larger sample sizes (thousands of individuals) compared to classical brain mapping.
• Future mental health research using MRI necessitates large-scale collaborations to achieve reliable brain-behavior findings.