Altered representation of expected value in the orbitofrontal cortex in mania

Abstract

Objective: Increased responsiveness to appetitive and reduced responsiveness to aversive anticipatory cues may be associated with dysfunction of the brain reward system in mania. Here we studied neural correlates of gain and loss expectation in mania using functional magnetic resonance imaging (fMRI).

Method: Fifteen manic patients and 26 matched healthy control individuals performed a monetary incentive delay task, during which subjects anticipated to win or lose a varying amount of money. Varying both magnitude and valence (win, loss) of anticipatory cues allowed us to isolate the effects of magnitude, valence and expected value (magnitude-by-valence interaction).

Results: Response times and total gain amount did not differ significantly between groups. FMRI data indicated that the ventral striatum responded according to cued incentive magnitude in both groups, and this effect did not significantly differ between groups. However, a significant group difference was observed for expected value representation in the left lateral orbitofrontal cortex (OFC; BA 11 and 47). In this region, patients showed increasing BOLD responses during expectation of increasing gain and decreasing responses during expectation of increasing loss, while healthy subjects tended to show the inverse effect. In seven patients retested after remission OFC responses adapted to the response pattern of healthy controls.

Conclusions: The observed alterations are consistent with a state-related affective processing bias during the expectation of gains and losses which may contribute to clinical features of mania, such as the enhanced motivation for seeking rewards and the underestimation of risks and potential punishments.

(c) 2009 Wiley-Liss, Inc.

Figures

Figure 1

Experimental Design. (a) Monetary incentive delay task (example trial). In each trial an incentive cue indicates the amount of money that can be gained or lost. Subjects had to react to the white target square in order to gain or avoid losing money. Subsequently, feedback was given. (b) FMRI data were modeled using magnitude, valence, and expected value (magnitude‐by‐valence interaction) as parametric regressors of interest. Illustrated are the weights that were used to model magnitude (black line), valence (red line) and expected value (blue line). Parametric analysis of cued incentive magnitude identifies brain regions that respond according to the following model: expectation of 3.00 [euro] gain or loss weighted with +3, 0.60 [euro] gain or loss with +2, 0.10 [euro] gain or loss with +1, and neutral outcome with 0 (analogously for valence and expected value). The dotted green line illustrates the negative magnitude‐by‐valence interaction. [Color figure can be viewed in the online issue, which is available at www.interscience.wiley.com.]

Figure 2

Effect of magnitude. (a) Effects of cued incentive magnitude (SPM parametric modulation analysis). One‐sample t‐test; healthy control subjects and manic patients pooled to one group. SPM map presented in section views through the left ventral striatal peak voxel, T = 7.14; x = −15, y = 3, z = −9 (MNI space). Peak voxel in the right ventral striatum (T = 6.18) at x = 21, y = 9, z = −9 (MNI space). For illustration purposes the significance level was set at P < 0.05, FWE‐corrected. R, right. (b) Parameter estimates for cued incentive magnitude in the left and right ventral striatum (SPM parametric modulation analysis). Effects are shown for healthy control subjects (Ctls), manic patients (Man), remitted patients (Rem), and the subgroup of manic patients that was re‐tested after remission (SGM). Values refer to the functional regions of interest identified in Figure 2a (mean of ROI). Error bars show the standard error of the mean. VS, ventral striatum. (c) Parameter estimates for anticipatory cues in the right ventral striatum (standard SPM analysis, modeling the anticipatory cues separately as explanatory variables). Expectation of 3.0 [euro] gain (++), expectation of 0.1 [euro] gain (+), expectation of neutral outcome (0), expectation of 3.0 [euro] loss (–), expectation of 0.1 [euro] loss (−). Values refer to the functional region of interest identified in Figure 2a (mean of ROI). Error bars show the standard error of the mean. Very similar activations were found in the left ventral striatum. [Color figure can be viewed in the online issue, which is available at www.interscience.wiley.com.]

Figure 3

Effect of expected value. (a) Effect of expected value representation in manic patients compared to healthy controls (SPM parametric modulation analysis). Two‐sample t‐test: manic patients > healthy controls. SPM map presented in section views through the peak voxel (T = 3.77) at x = −30, y = 42, z = −9 (MNI space). P < 0.001, uncorrected. R, right. (b) Parameter estimates for expected value in the left lateral OFC (SPM parametric modulation analysis). Effects are shown for healthy control subjects, manic patients, remitted patients, and the subgroup of manic patients that was retested after remission. Values refer to the functional region of interest identified in Figure 3a (mean of ROI). Error bars show the standard error of the mean. Manic patients activate according to expected value, while healthy controls and remitted patients tend to show an inverse effect. (c) Parameter estimates for incentive cues in the left lateral OFC (standard SPM analysis, modelling the anticipatory cues separately as explanatory variables). Expectation of 3.0 [euro] gain (++), expectation of 0.1 [euro] gain (+), expectation of neutral outcome (0), expectation of 3.0 [euro] loss (–), expectation of 0.1 [euro] loss (−). Values refer to the functional region of interest identified in Figure 3a (mean of ROI). Error bars show the standard error of the mean. [Color figure can be viewed in the online issue, which is available at www.interscience.wiley.com.]