Test-retest reliability of fMRI-measured brain activity during decision making under risk
This study aimed at investigating long-term test-retest reliability (TRR) of regional brain activations related to decision making under risk using the Balloon Analogue Risk Taking task (BART) based on the magnitude and significance of Intraclass correlations (ICCs), the latter being estimated with the use of permutation statistics. We also identified regions with strong familial influences (indicated by intrapair twin correlations), which must be demonstrated before brain correlates of decision making are used in genetic studies. Our results revealed that regions activated during decision period rather than outcome period of risky decisions showed the greatest reliability and familiality, including the right fusiform, right rostral anterior cingulate and left superior parietal regions (ICCs ranging from .66 to .69).
Neural correlates of decision making under risk are being increasingly utilized as biomarkers of risk for substance abuse and other psychiatric disorders, treatment outcomes, and brain development. This research relies on the basic assumption that fMRI measures of decision making represent stable, trait-like individual differences. However, reliability needs to be established for each individual construct. Here we assessed long-term test-retest reliability (TRR) of regional brain activations related to decision making under risk using the Balloon Analogue Risk Taking task (BART) and identified regions with good TRRs and familial influences, an important prerequisite for the use of fMRI measures in genetic studies. A secondary goal was to examine the factors potentially affecting fMRI TRRs in one particular risk task, including the magnitude of neural activation, data analytical approaches, different methods of defining boundaries of a region, and participant motion. Across regions, reliabilities ranged across brain regions from poor to good (0.0-0.8, with a mean ICC of 0.17) and highest reliabilities were observed for parietal, occipital, and temporal regions. Among the regions that were of a priori theoretical importance due to their reported associations with decision making, the activation of left anterior insula during the decision period and the right caudate during the outcome period showed the highest reliabilities (ICCs of 0.54 and 0.63, respectively). Among the regions with highest reliabilities, the right fusiform, right rostral anterior cingulate and left superior parietal regions also showed high familiality as indicated by intrapair monozygotic twin correlations (ranging from 0.66 to 0.69). Overall, regions identified by modeling the average BOLD response to a specific event type (rather than its modulation by a parametric regressor), regions including significantly activated vertices (compared to a whole parcel), and regions with greater magnitude of task-related activations showed greater reliabilities. Participant motion had a moderate negative effect on TRR. Regions activated during decision period rather than outcome period of risky decisions showed the greatest TRR and familiality. Regions with reliable activations can be utilized as neural markers of individual differences or endophenotypes in future clinical neuroscience and genetic studies of risk-taking.
- Ozlem Korucuoglu
- Michael P Harms
- Serguei V Astafiev
- James T Kennedy
- Semyon Golosheykin
- Deanna M Barch
- Andrey P Anokhin
- Washington University in St. Louis