Neuropsychology 2014, Vol. 28, No. 6, 870 – 880

© 2014 American Psychological Association 0894-4105/14/$12.00 http://dx.doi.org/10.1037/neu0000120

This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

Functional Neuroimaging of the Iowa Gambling Task in Older Adults Kameko Halfmann

William Hedgcock

University of Iowa

University of Iowa Tippie College of Business

Antoine Bechara

Natalie L. Denburg

University of Southern California

University of Iowa

Objective: The neural systems most susceptible to age-related decline mirror the systems linked to decision making. Yet, the neural processes underlying decision-making disparities among older adults are not well understood. We sought to identify neural response patterns that distinguish 2 groups of older adults who exhibit divergent decision-making patterns. Method: Participants were 31 healthy older adults (ages 59 – 88, 53% female), defined as advantageous or disadvantageous decision-makers based on Iowa Gambling Task (IGT) performance, who completed an alternate version of the IGT while undergoing functional MRI. The groups were indistinguishable on neuropsychological testing. We contrasted the BOLD signal between groups during 3 phases of the decision-making process: Prechoice (preselection), Prefeedback (postselection), and Feedback (receipt of gains/losses). We further examined whether BOLD signal varied as a function of age in each group. Results: We observed greater activation among the IGT-Disadvantageous relative to -Advantageous older adults in the prefrontal cortex during the early phases of the decision-making process (Prechoice), and in posterior brain regions (e.g., the precuneus) during the later phases (Prefeedback and Feedback). We also found that with increasing age, IGTAdvantageous older adults showed increasing activation in the prefrontal cortex during all phases and increasing activation in the posterior cingulate during earlier phases of the decision process. By contrast, the IGT-Disadvantageous older adults exhibited a reduced or reversed trend. Conclusions: These functional differences may be a consequence of altered reward processing or differing compensatory strategies between IGT-Disadvantageous and -Advantageous older adults. This supports the notion that divergent neurobiological aging trajectories underlie disparate decision-making patterns. Keywords: aging, decision making, Iowa Gambling Task, reward, ventromedial prefrontal cortex

nances, long-term care, and medical treatment that older adults face, reductions in decision-making skills will have a substantial impact on individuals and society alike. Characterizing neurobehavioral phenotypes of decision-making patterns among healthy older adults (aged 56 years and older in this context) will help in the development of decision aids to circumvent some of these problems. The neurobiological system that governs decision making, namely the frontostriatal system (for a review, see Glimcher & Rustichini, 2004), parallels the system that declines most readily in older adulthood. For example, the prefrontal cortex (PFC) is one of the first regions to deteriorate in older adulthood (e.g., Raz et al., 1998). Declines in neuropsychological performances such as executive functioning (West, Murphy, Armilio, Craik, & Stuss, 2002), decreases in PFC volume (Resnick, Pham, Kraut, Zonderman, & Davatzikos, 2003), and functional differences in the PFC (Cabeza, Anderson, Locantore, & McIntosh, 2002) exemplify the early deterioration of the PFC. Moreover, PFC decline occurs at a greater rate than other brain regions, such as the hippocampus (Raz et al., 1998). One sector of the PFC, the ventromedial PFC (VMPC), exhibits a particularly strong negative correlation with age (Raz et al., 1997). Of note, the VMPC serves as a convergence zone for multiple sensory association systems, including the viscera (Damasio, 1996). In addition, the VMPC provides output to motor cortex and

Aging progresses along diverse paths that often include considerable anatomical and physiological changes in the brain (Cohen, Janicki-Deverts, & Miller, 2007). These neural changes impact cognitive abilities. For example, even within seemingly healthy populations, the elderly often experience diminishing decisional capacity (Bruine de Bruin, Parker, & Fischhoff, 2012). Considering the vast array of decisions regarding fi-

This article was published Online First July 28, 2014. Kameko Halfmann, Division of Cognitive Neuroscience, Department of Neurology, University of Iowa; William Hedgcock, Department of Marketing, University of Iowa Tippie College of Business; Antoine Bechara, Institute for the Neurological Study of Emotion and Creativity, Department of Psychology, University of Southern California; Natalie L. Denburg, Division of Cognitive Neuroscience, Department of Neurology, University of Iowa. Preparation of this article was supported by a DANA Foundation Program in Brain and Immuno-Imaging grant to Natalie L. Denburg. We thank Daniel Tranel for valuable ongoing feedback on this project and manuscript. We would also like to thank Joel Bruss and Nicholas Jones for technical assistance, and Julie Gudenkauf, Jonathan Yuksa, and Joshua Hartman for assistance with data collection. Correspondence concerning this article should be addressed to Kameko Halfmann, 2007 RCP, Neurology, UIHC, 200 Hawkins Drive, Iowa City, IA 52242-1053. E-mail: [email protected] 870

This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

NEUROIMAGING OF GAMBLING TASK IN OLDER ADULTS

is intricately connected with limbic structures (Damasio, Anderson, & Tranel, 2011). With such links, the VMPC combines various inputs and integrates sensory, visceral, and emotional information in a coherent manner to direct action and behavior. Consequently, the VMPC is positioned to guide emotion-related decisions (Öngür & Price, 2000), and age-related deterioration of this brain region may have an adverse impact on decision-making abilities in emotional contexts. Consistent with this idea, human lesion studies have shown the critical role the VMPC plays in value-based decision making using assessments such as the Iowa Gambling Task (Gläscher et al., 2012). The Iowa Gambling Task (IGT; Bechara, 2007) was developed as a measure of complex decision making, involving cognitive and emotional elements such as ambiguity, risk, reward, and punishment. The IGT involves 100 trials in which the participant must select from one of four card decks, with the ultimate goal of gaining money. Two of the four decks are disadvantageous and lead to long-term losses. The other two card decks are advantageous and lead to long-term gains. Patients with damage to the VMPC show both impaired behavioral performance and diminished anticipatory skin conductance responses (SCRs) prior to making selections on the IGT (Bechara, Damasio, Tranel, & Damasio, 1997). In other words, patients with VMPC damage do not adopt an optimal strategy of selecting primarily from the advantageous decks over the course of the task. Damasio (1996) proposed that dysfunction in the neural circuitry linked to representing somatic or visceral states causes deficits in anticipatory SCRs, with the VMPC playing a key role in this circuitry. The amygdala, striatum, insula, posterior cingulate, and several brainstem nuclei form the auxiliary circuitry underlying somatic states (Li, Lu, D’Argembeau, Ng, & Bechara, 2010), as these regions correlate with specific emotion states (Damasio et al., 2000). Although most young adults perform advantageously on the IGT, approximately one third of older adults perform disadvantageously in a manner not unlike patients with focal damage to the VMPC (Denburg, Tranel, & Bechara, 2005). Older adults who perform disadvantageously on the IGT (hereafter called IGTDisadvantageous) also have lessened discriminating anticipatory SCRs. By contrast, older individuals with advantageous performance on the IGT (hereafter called IGT-Advantageous) exhibit discriminating anticipatory SCRs (Denburg, Recknor, Bechara, & Tranel, 2006). The variability among older adults’ IGT performance aligns with the notion that divergent, neurobiologically based aging trajectories arise with explicit behavioral outcomes. For example, performance on the IGT among older adults predicts performance on other decision-making tasks, such as temporal discounting and consumer decision making. Specifically, IGT-Disadvantageous, compared to IGT-Advantageous, older adults exhibit higher discount rates (Halfmann, Hedgcock, & Denburg, 2013), in addition to lower comprehension scores when evaluating deceptive advertisements and a greater intent to purchase products in deceptive advertisements (Denburg et al., 2007). Despite the likelihood that differing IGT performance in older adults arises from divergent neural activation patterns, little work has directly investigated this idea. Among younger adults, neuroimaging work on the neural correlates of the IGT supports earlier behavioral and physiological research. For example, in a positron emission tomography study in younger adults, Ernst and col-

871

leagues (2002) found greater activation during the IGT relative to a control task in the VMPC, anterior cingulate, dorsolateral PFC (DLPFC), insula, and parietal cortex—primarily regions that have previously been linked to IGT performance based on behavioral lesion work (Bechara, 2004). More recently, using functional MRI (fMRI), Li and colleagues (2010) found similar regions were activated during the IGT compared to a control task, including the VMPC, anterior cingulate, DLPFC, insula, posterior cingulate cortex (PCC), medial orbitofrontal cortex, and ventral striatum. Among older adults, Rogalsky, Vidal, Li, and Damasio (2012) used fMRI to examine activation in the VMPC during the IGT. They discovered that older adults engaged the right VMPC during the IGT relative to the control task. Moreover, activity in the left VMPC positively correlated with IGT performance level. This study provides initial evidence that PFC functioning contributes to age-related disparities in decision making. But the findings have notable limitations due to small sample size (N ⫽ 14), and a study design that did not dissect the various phases of the decisionmaking process. This means it is not possible to know whether activation differences occurred during the decision process (prior to choice) or after the decision was made while the individuals were receiving feedback about losses and gains. It very well could be that aging influences one or many phases of the decisionmaking process in distinct ways. Here, we set out to expand on previous research by testing whether IGT-Disadvantageous and -Advantageous older adults would demonstrate neural processing differences using fMRI. We specifically examined the different phases of the decision process (Prechoice, Prefeedback, and Feedback) while they performed the IGT. Based on the earlier finding that IGT performance in older adulthood predicts anticipatory SCRs (Denburg et al., 2006), we predicted that IGT-Disadvantageous and -Advantageous older adults would exhibit activation differences in convergence zones, such as the VMPC, and/or emotion systems, such as the posterior cingulate or striatum, as these are regions that have been implicated in emotional states (Damasio, 1996; Li et al., 2010). Further, we predicted that IGT-Advantageous older adults would show increases in neural activation in these regions relative to the IGT-Disadvantageous older adults, akin to Rogalsky et al. (2012). Lastly, because the VMPC is hypothesized to play a central role as a convergence zone that guides emotion-related decisions (Damasio, 1996; Öngür & Price, 2000), we specifically predicted that the IGT-Advantageous older adults would show increases in VMPC activation during the Prechoice phase and increases in activation in emotion-related regions during the Feedback phase.

Method Participants Thirty-three cognitively and psychologically well-characterized, healthy older adults (59 – 88 years of age; Mdn ⫽ 77.0 years) were recruited from the principal investigator’s participant database. These participants were selected such that our sample would consist of approximately 50% females and 50% IGTAdvantageous performers. Participants were consented according to the institutional review board’s policy. All participants had been administered a comprehensive neuropsychological battery (see Table 1 for a subset of the battery used) and health interview. The

HALFMANN, HEDGCOCK, BECHARA, AND DENBURG

872

Table 1 Means (SDs), T and Cohen’s D for Demographic and Neuropsychological Data Category N Demographics

Intellect Language

This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

Memory Attention Motor/Construction Executive function Complex decision making Other

Measure

IGT-Disad-vantageous

IGT-Advantageous

t-statistic

Cohen’s d

— Age Education Sex Handedness WRAT Reading Full Scale IQ BNT COWA ReyO Delay AVLT Delay WMI BVRT Trails Part A ReyO Copy Trails Part B WCST PE IGT A’B’C’D’ⴱⴱⴱ K’L’M’N’ (1st Half)⫹ K’L’M’N’ (2nd Half)⫹ⴱ BDI Lipkus Numeracy BMI Smokers

15 74.3 (8.5) 16.3 (3.0) 53%F 14 RH, 1 LH 50.7 (4.4) 117.6 (12.8) 19.3 (0.7) 43.0 (15.3) 15.1 (5.6) 9.1 (2.5) 111.0 (14.1) 7.2 (2.2) 35.8 (10.8) 31.3 (3.4) 80.5 (25.6) 19.7 (17.0) ⫺23.9 (16.2) 54% (2%) 49% (2%) 4.2 (4.9) 8.6 (2.2) 29.0 (4.9) 0%

16 77.4 (5.5) 16.1 (2.9) 50% F 15 RH, 1 Ambi 53.6 (4.8) 123.8 (9.1) 19.6 (0.6) 46.1 (11.8) 17.5 (6.4) 9.9 (3.4) 112.9 (11.7) 6.8 (1.8) 34.5 (10.6) 32.7 (3.6) 79.6 (26.9) 11.3 (8.5) 35.4 (15.6) 54% (2%) 55% (2%) 4.6 (3.7) 8.3 (2.7) 29.6 (5.8) 0%

— ⫺0.56 0.19 — — ⫺1.47 ⫺1.56 ⫺1.51 ⫺0.64 ⫺1.11 ⫺0.80 ⫺0.40 0.34 0.34 ⫺1.12 0.10 1.78 ⫺10.3 0.02 4.44 ⫺0.27 0.33 ⫺0.21 —

— ⫺0.43 0.07 — — ⫺0.63 ⫺0.56 ⫺0.46 ⫺0.23 ⫺0.40 ⫺0.27 ⫺0.15 0.20 0.12 ⫺0.40 0.03 0.63 ⫺3.73 ⫺0.09 0.12 ⫺0.11 —

Note. Demographic and neuropsychological means, standard deviations, t-values and Cohen’s d effect sizes for the IGT-Disadvantageous and IGTAdvantageous older adults group differences. WRAT ⫽ Wide Range Achievement Test; BNT ⫽ Boston Naming Test; COWA ⫽ Controlled Oral Word Association; AVLT ⫽ Auditory Verbal Learning Test; ReyO ⫽ Rey Osterrieth Complex Figure; WMI ⫽ Working Memory Index from the Wechsler Adult Intelligence Scale–IV; BVRT ⫽ Benton Visual Retention Test; Trails Part A and B ⫽ Trail Making Test Parts A and B; WCST PE ⫽ Wisconsin Card Sorting Test Perseverative Errors; IGT ⫽ Iowa Gambling Task; BDI ⫽ Beck Depression Inventory; BMI ⫽ body mass index. ⫹ Indicates rows in which GEE estimates are reported, along with Wald chi square values (rather than t-statistics), and means are reported with standard error. ⴱ p ⬍ 0.05. ⴱⴱ p ⬍ 0.01. ⴱⴱⴱ p ⬍ 0.001.

neuropsychological data in Table 1 were collected, on average, 5 years ago (SD ⫽ 3.2 years). Since initial testing sessions, individuals in our database have been evaluated for cognitive decline annually by a neuropsychologist who knows the participants well. Individuals were excluded from the database if they had any outstanding medical or psychiatric conditions (i.e., stroke, head injury, Type I diabetes, neurosurgery, seizure disorder, demyelinating disorder, substance abuse, uncontrolled medical condition, vision/hearing loss, psychiatric illness necessitating inpatient treatment, self-reported depression and/or anxiety exceeding mild severity). In addition, the following inclusion criteria were used for this study: ability and willingness to provide consent, living independently, normal (age-related) hearing and vision, cognitive performance within the normal range (1.5 standard deviations) for age and educational attainment (Lezak, Howieson, Bigler, & Tranel, 2012), and no MRI contraindications. In addition to the above criteria, all participants had previously completed the computerized A’B’C’D’ version of the IGT in the standard fashion, selecting 100 cards, one at a time, from four decks. Each selection was followed by feedback displaying how much money the participant won on that trial. A subset of trials also included losses, which were displayed following the gains. Overall, two of the decks yielded large gains, but even larger losses, leading to a net loss across the task. Two of the decks yielded small gains, but even smaller losses, leading to a net gain across the task. In order to perform advantageously, participants

must choose more frequently from the latter two decks, which yield net gains. After completing the A’B’C’D’ IGT, participant’s scores were computed by counting the total card selections from the advantageous decks and then subtracting the total card selections from the disadvantageous decks, for each block of 20 cards (Bechara, Tranel, & Damasio, 2000). They were subsequently categorized as IGT-Advantageous (A’B’C’D’ scores ⬎16) or IGTDisadvantageous (A’B’C’D’ scores ⬍5; see Table 1 for mean scores). These scores were used as selection criteria for participation in the current study. In previous work by our lab, cut-off scores for IGT-Disadvantageous performance has been more stringent (e.g., Denburg et al., 2006), but with the added exclusionary criteria for fMRI research we were unable to attain a sample with the stricter cut-off. Participants completed the A’B’C’D’ version of the IGT an average of 5 years prior to this study (SD ⫽ 3.2 years). It is important to note that once participants complete the IGT, practice effects become a considerable concern. For example, healthy participants show significant improvement on the A’B’C’D’ version of the IGT when retested anywhere between 6 hours and 5 years later (Bechara, Damasio, & Damasio, 2000; Waters-Wood, Xiao, Denburg, Hernandez, & Bechara, 2012). Test–retest effects have not been examined beyond 5 years; however, considering that half of our sample completed the A’B’C’D’ IGT in the last 5 years, we did not choose to reclassify participants.

NEUROIMAGING OF GAMBLING TASK IN OLDER ADULTS

This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

Two individuals who completed the fMRI study were excluded from subsequent analyses because they responded on fewer than 80% of the trials while undergoing fMRI, resulting in 31 participants. Of those, 15 were IGT-Disadvantageous and 16 were IGTAdvantageous (note that this grouping refers to participant’s performance on the A’B’C’D’ IGT throughout the manuscript). Independent samples t tests were used to determine whether the groups differed in neuropsychological performances across multiple cognitive domains. The groups did not differ on any demographic, neuropsychological, or health status variables (ps ⬎ 0.05), with the exception of IGT scores (p ⬍ 0.001), as shown in Table 1.

Task Participants completed the K’L’M’N= version of the IGT while undergoing fMRI scanning. This K’L’M’N= alternate version was used because repeated testing with the IGT can lead to practice effects (Xiao et al., 2013), and all individuals had already completed the original A’B’C’D’ version of the IGT. Additionally, scores on the K’L’M’N= version and A’B’C’D’ version of the IGT tend to be fairly consistent within participants (Xiao et al., 2013). Compared to the A’B’C’D’ version, the K’L’M’N= version of the IGT has a reduced percentage of times that advantageous selections yield lower initial rewards than disadvantageous selections, and the K’L’M’N= version has a reduced magnitude of increases in net gains and losses within the good and bad decks, respectively. This reduction leads to smaller group differences and mitigates functional activation confounds related to performance differences. The IGT program was written in Matlab based on the Psychtoolbox extensions (Brainard, 1997) as described by Li et al. (2010). Participants were given instructions (as described in Bechara et al., 2000) and time to answer any questions they might have. Because older adults have a difficult time enduring long scanning sessions, we used a truncated version of the IGT where participants completed 80 (rather than 100) trials that were divided into four (rather than five) 20-trial blocks. By 80 trials most participants reach what is termed the “conceptual” period, where they are able to distinguish good versus bad decks (Bechara et al., 1997). Participants had a predetermined, jittered interval of 3–7 s to make a choice (“Prechoice”; M ⫽ 4 s), after which time the computer randomly chose a deck to keep the task within a certain time limit. After participants made their selections, they waited for approximately 1 s in addition to the remainder of the abovementioned predetermined interval in which they had to make a choice (“Prefeedback”). Finally, participants were shown how much money they won and subsequently, if relevant, lost (“Feedback”). The feedback was displayed for 3 s total. If the trial was a win/loss trial, the win feedback was displayed for 1.5 s followed by the loss feedback displayed for 1.5 s. The intertrial interval (ITI) varied between 1 and 3 s. Participants responded using a fiber-optic response pad. The buttons used to make selections were spatially aligned with the card decks on the screen (e.g., the index finger was used to select from the first deck). The stimuli were backprojected onto a screen through a mirror that was built into the head coil.

873

Imaging MRI scans were obtained using a standard 12-channel head array on a Siemens 3T TIM Trio. Participants lay supine on a scanner bed. Foam pads were used to minimize head motion. A high-resolution full brain 256 ⫻ 256 ⫻ 256 mm T1-weighted 3D MPRAGE sequence (voxel size ⫽ 1 mm3, TI ⫽ 909 ms, TR ⫽ 2,530 ms, TE ⫽ 2.8 ms, Flip angle ⫽ 10 degrees, slice thickness ⫽ 1.0 mm, scan time ⫽ 5 min and 36 s) was collected for each participant in the coronal plane. During the task, blood oxygen level dependent (BOLD) signal was measured with a T2ⴱ weighted echo-planar imaging sequence. Standard scan parameters were used (TR ⫽ 2,000, TE ⫽ 30, flip angle ⫽ 90 degrees, FOV ⫽ 220 mm ⫻ 220 mm, in plane resolution ⫽ 64 ⫻ 64 pixels, slice thickness ⫽ 4 mm, no gap). Thirty-one interleaved slices were acquired in an oblique transverse orientation of approximately 20 degrees to the anterior commissure/posterior commissure line. The slice orientation helped minimize potential VMPC susceptibility problems. Scan time was 12 minutes and 14 seconds during which 367 volumes were collected. The first two volumes were discarded to eliminate scans with saturation effects.

Analysis Behavior Overall scores were calculated by counting the total card selections from the disadvantageous decks and the total card selections from the advantageous decks, for each block of 20 cards, then subtracting the disadvantageous selections from the advantageous selections (Bechara et al., 2000). Scores on the K’L’M’N= version completed in this study tended to regress toward the mean. Importantly, no IGT-Disadvantageous older adults performed advantageously on the K’L’M’N= IGT. Mean scores are reported in Table 1. To conduct our group analyses, we employed the Generalized Estimating Equations (GEE) technique using SPSS 19.0.0.1 (IBM SPSS, 2010). GEE is an alternative generalized regression method to the generalized linear mixed model (Liang & Zeger, 1986) that can take into account repeated measurements, dichotomous choices, and correlations between choices across the task. Decks were categorized as “Good” and “Bad.” Then, the dependent measure, choice, was modeled using a binary logistic distribution. Trials were handled as repeated measurements with IGT-Status (IGT-Disadvantageous vs. IGT-Advantageous) as the betweensubject factor. We separately analyzed the data for the first half (Trials 1– 40) and the second half (Trials 41– 80) of the task. Lastly, we examined whether there were reaction time (RT) differences between groups in the first and second half of the task using a linear GEE model with IGT-Status as a between-subjects factor. Missed trials were excluded from the GEE analyses.

Imaging All fMRI analyses were conducted using BrainVoyager QX 2.6 (Brain Innovation, Maastricht, Netherlands). The anatomical data were corrected for image intensity inhomogeneities and transformed into Talairach space (Talairach & Tournoux, 1988) using a cubic spline interpolation method for the initial ACPC transfor-

This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

874

HALFMANN, HEDGCOCK, BECHARA, AND DENBURG

mation and then a sinc interpolation method for the final Talairach transformation. All data were inspected for poor registration. The functional data were preprocessed to correct for slice timing using cubic spline interpolation and head motion in six directions using a trilinear/sinc interpolation approach. We applied high-pass temporal filtering using Fast Fourier Transform to remove lowfrequency drifts and spatially smoothed images using a 6 mm full-width-half-maximum Gaussian kernel. Functional images were registered to each individual’s high-resolution structural image collected in the same session. Then images were registered into standard Talairach space. Statistical analyses were performed on a participant-averaged brain in Talairach space (See Figure 1 for the functional coverage map). Statistical analyses were performed using a multistage approach. In the first stage, each part of the decision phase (ITI, Prechoice, Prefeedback, and Feedback) was convolved using a two-gamma hemodynamic response function corrected for temporal serial correlation using an AR(2) model. In the second stage, we used a multisubject random-effects model with separate subject predictors. At this stage in the analysis, we used analysis of covariance (ANCOVAs) to determine group differences while controlling for age. More specifically, we computed group differences by separately contrasting IGT-Disadvantageous relative to -Advantageous activation for each decision phase (Prechoice, Prefeedback, and Feedback) after subtracting the ITI to control for between subject

Figure 1. A probabilistic functional coverage map, where green (dark grey in printed version) regions indicate voxels where there is 100% functional coverage, meaning all of the subject files contain functional data in those voxels. We observed full coverage of our functional data across all of our ROIs, and we specifically demonstrate that we have good coverage in problem areas such as the VMPC. Pictured slices were chosen to correspond with the results depicted in Figure 4. See the online article for the color version of this figure.

differences. We note that the IGT is a complex decision task, and, although our analyses aim to parse activation during different phases of the decision process that differ between groups, we do not aim to interpret the precise mechanisms occurring during each decision phase and we cannot determine normal task-related activation. Given the complexity of the task, determining task-related activation would be difficult regardless of the baseline chosen. Our analyses mitigate potential problems with task complexity and interpretation of activation by contrasting two groups, the IGTDisadvantageous relative to -Advantageous. Therefore, our results show relative activation between groups and care should be taken when interpreting absolute changes since that will depend on choice of baseline. To test the robustness of our results, we redid the analyses without subtracting ITI from each decision phase. Between group results remained the same.

ROI Analysis We assessed activation differences between the IGTDisadvantageous and -Advantageous older adults in predefined regions of interest (ROIs) using ANCOVAs to compare average beta weights within each ROI, controlling for the age of each subject. We focused our analysis on predefined ROIs that have previously been associated with performing the IGT relative to a control task. To elaborate, Li et al. (2010) found that the IGT required the interaction between several systems: a memory system (i.e., DLPFC and the parahippocampal gyrus), an emotional system (i.e., the striatum, insula, and PCC), a convergence zone for the aforementioned systems (i.e., the VMPC/ ACC and OFC), and a motor/behavioral system. We expected to observe group differences in the emotion system and in convergence zones because we suspect these systems are affected by aging and previous research indicates that IGT performance in older adults is associated both with a psychophysiological marker (e.g., Denburg et al., 2006) and with differences in knowledge of gains and losses rather than learning and memory (e.g., Wood, Busemeyer, Koling, Cox, & Davis, 2005). For these analyses, we used the predefined Brodmann areas (BAs) in Brainvoyager and an anatomically defined spherical ROI in each the left and right striatum that correspond to significant results in Li et al. (2010). See Table 2 for specific ROIs. After examining IGT-Disadvantageous relative toAdvantageous contrasts, we next used ANCOVAs to examine whether IGT-Status predicted differences in the relationship between age and the BOLD signal (e.g., whether age interacted with IGT-Status). The age-activation slopes for each group are reported in Table 2. A significant interaction between age and IGT-Status indicates that as age advances, the two groups show a different trajectory in their activation patterns. The slopes represent the relationship between age and BOLD signal in each respective ROI. For example, a positive slope (e.g., 0.13 for IGT-Advantageous older adults in BA 24 [Prechoice]) indicates that as age advances greater levels of activation emerge for this group.

Whole-Brain Analyses To supplement these specified analyses, we also performed whole-brain analyses contrasting the IGT-Disadvantageous ver-

This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

NEUROIMAGING OF GAMBLING TASK IN OLDER ADULTS

sus IGT-Advantageous older adults (controlling for age). Whole-brain analyses allow us to 1) capture areas of activation that were not included in the ROI analysis, 2) demonstrate the specificity of our ROI analysis, and 3) address multiple comparison issues commonly observed in fMRI research in a second way. We used ANCOVAs to test the effect of IGT while controlling for age in the following analysis. The reported whole-brain results were thresholded with a cluster-corrected significance level of ␣ ⬍ .05, according to a three-dimensional extension of a randomization procedure first described by Forman et al. (1995). To achieve this, we first set the voxel-level threshold to p ⫽ 0.01, uncorrected. We then calculated cluster level false positive rates based on an estimate of spatial smoothness (Prechoice ⫽ 2.05 voxels, Prefeedback ⫽ 2.23 voxels, Feedback ⫽ 2.08 voxels), which depends on the signal-to-noise ratio, and on 1,000 Monte Carlo simulations using the Cluster Threshold plugin in Brainvoyager. The reported results are based on the simulated images for each decision phase. The minimum cluster size for the estimated false positive rate of ␣ ⬍ .05 fell between 819 and 1,070 mm3 depending on the decision phase. We lastly used the results from the whole-brain group comparison analysis to create ROIs (see Table 3 for ROIs) and examine whether there was an interaction between IGT-Status (IGT-Disadvantageous vs. IGT-Advantageous) and half (Trials 1– 40 vs. Trials 41– 80). We created contrasts subtracting activation during the first half of the task from the second half. Then we ran ANCOVAs on these contrast maps in each ROI to analyze the interaction between IGT and activation during each half (Half Interaction). We predicted that IGTAdvantageous older adults would show a greater change in activation between halves, as accumulated experience over the course of the task would change the selection strategy in the Advantageous group.

875

␹2 ⫽ 19.4, p ⬍ 0.001, OR ⫽ 1.20, 95% C.I. [1.11, 1.31]), such that participants were faster in the second half than the first. The interaction between IGT Status and task half neared significance (Wald ␹2 ⫽ 2.7, p ⫽ 0.10, OR ⫽ 0.90, 95% C.I. [0.80, 1.02]), such that the IGT-Advantageous older adults trended toward having a greater reduction in RT from the first to second half relative to the IGTDisadvantageous older adults (see Figure 2). The IGT-Advantageous older adults had an average RT of 1.33 s (SE ⫽ 0.03 s), and the IGT-Disadvantageous older adults had an average RT of 1.27 s (SE ⫽ 0.03 s) during the first half. By contrast, during the second half of the task, the IGT-Advantageous older adults had an average RT of 1.15 s (SE ⫽ 0.03 s) and the IGT-Disadvantageous older adults had an average RT of 1.18 s (SE ⫽ 0.03 s).

Region of Interest Results We first assessed activation differences between the IGTDisadvantageous and -Advantageous older adults in specified ROIs (see Table 2), using an ANCOVA to control for subject age. Reported statistics are average t and p values across the voxels within each ROI, along with the effect size, ␩2, for the group comparison controlling for age. We found group differences during the Prechoice phase in VMPC (e.g., BA 11, tave ⫽ 2.23, pave ⫽ 0.03, ␩2 ⫽ 0.15), PCC (tave ⫽ 2.77, pave ⫽ 0.01, ␩2 ⫽ 0.21), and right striatum (tave ⫽ 2.09, pave ⫽ 0.05, ␩2 ⫽ 0.14), with the IGT-Advantageous older adults showing less relative activation in these regions than the IGTDisadvantageous older adults. A similar pattern was observed in VMPC (tave ⫽ 2.12, pave ⫽ 0.04, ␩2 ⫽ 0.13) and PCC (e.g., tave ⫽ 2.74, pave ⫽ 0.01, ␩2 ⫽ 0. 21) during the Prefeedback phase, and in the VMPC (tave ⫽ 2.01, pave ⫽ 0.05, ␩2 ⫽ 0.13) during the Feedback phase. Notably, we did not observe group differences in the parahippocampal gyrus or DLPFC (ts ⬍ 1.01, ps ⬎ 0.32, ␩2 ⬍ 0.03). Results are summarized in Table 2.

Results Behavioral Results IGT-Advantageous older adults had a mean total score of 6.5 (SE ⫽ 4.3) on the K’L’M’N= version of the IGT, and IGTDisadvantageous older adults had a mean total score of 1.5 (SE ⫽ 1.3). We utilized the GEE technique to assess group differences in IGT performance during the first (Trials 1– 40) and second half (Trials 41– 80) of the K’L’M’N= IGT. There was not a significant difference between the IGT-Disadvantageous versus IGTAdvantageous older adults during the first half of the IGT (Wald ␹2 ⫽ 0.02, p ⫽ 0.88). In other words, the odds of the IGTAdvantageous relative to -Disadvantageous older adults picking from a good deck was not different from one (OR ⫽ 1.02, 95% C.I. [0.81, 1.28]). By contrast, there was a significant difference between the groups during the second half (Trials 41– 80) of the IGT, such that the IGT-Advantageous older adults were 1.28 (95% C.I. [1.02, 1.61]) times more likely to choose from a good deck than the IGT-Disadvantageous older adults (Wald ␹2 ⫽ 4.44, p ⫽ 0.04, Estimated Mean Difference ⫽ 6%, SE ⫽ 3%). See Table 1 for mean advantageous selections for each group in the first and second half of the task. Next, we examined RTs (in seconds). We did not find a significant difference between groups (Wald ␹2 ⫽ 0.22, p ⫽ 0.64, OR ⫽ 1.04, 95% C.I. [0.95, 1.13]), but there was a main effect of task half (Wald

Figure 2. Average RT (in seconds) for IGT-Disadvantageous and -Advantageous older adults depicting a trend toward an interaction between IGT Status and half (Wald ␹2 ⫽ 2.7, p ⫽ 0.10), such that the IGT-Advantageous older adults trended toward having a greater reduction in RT from the first half to second half relative to the IGT-Disadvantageous older adults. Error bars represent 95% confidence intervals.

HALFMANN, HEDGCOCK, BECHARA, AND DENBURG

876

Table 2 Predefined Region of Interest (ROI) Analysis ROI Region Prechoice VMPC/ACC

This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

PCC VS PHc DLPFC Prefeedback VMPC/ACC

PCC VS PHc DLPFC Feedback VMPC/ACC

PCC VS PHc DLPFC

Beta weights

Group differences

Test for equivalent age slopes

Age slope

BA

Dis.

Adv.

Ave t

Ave p

Ave F

Ave p

Dis.

Adv.

11 24 25 32 23 29 31 R. L. 36 46

⫺0.18 ⫺0.65 ⫺0.49 ⫺0.97 ⫺0.25 ⫺0.06 ⫺0.29 ⫺1.07 ⫺1.23 ⫺0.13 ⫺0.43

⫺0.39 ⫺0.43 ⫺1.32 ⫺0.84 ⫺0.95 ⫺1.13 ⫺0.84 ⫺2.72 ⫺2.40 ⫺0.30 ⫺0.56

2.23 ⫺0.50 2.22 ⫺0.26 1.47 2.77 1.17 2.09 1.58 0.82 0.73

0.03 0.62 0.03 0.80 0.15 0.01 0.25 0.05 0.13 0.42 0.47

0.87 16.08 0.06 16.05 4.37 8.05 7.89 0.01 0.004 3.13 0.66

0.36 0.0004 0.82 0.0004 0.05 0.009 0.009 0.91 0.95 0.09 0.42

0.01 ⫺0.09 0.01 ⫺0.08 ⫺0.04 ⫺0.04 ⫺0.03 0.02 ⫺0.03 ⫺0.02 0.01

0.02 0.13 ⫺0.001 0.16 0.11 0.11 0.15 0.01 ⫺0.04 0.04 0.03

11 24 25 32 23 29 31 R. L. 36 46

⫺0.22 ⫺1.53 ⫺0.54 ⫺1.85 ⫺1.13 ⫺0.93 ⫺0.86 ⫺1.75 ⫺1.58 ⫺0.53 ⫺0.68

⫺0.42 ⫺1.74 ⫺1.36 ⫺1.98 ⫺2.20 ⫺2.24 ⫺2.05 ⫺3.07 ⫺2.49 ⫺0.80 ⫺0.84

2.12 0.38 1.94 0.24 1.95 2.74 2.09 1.37 1.06 1.01 0.75

0.04 0.70 0.06 0.81 0.06 0.01 0.05 0.18 0.30 0.32 0.46

2.08 17.55 0.004 18.07 6.12 7.14 8.89 0.04 0.17 15.04 1.62

0.16 0.0003 0.95 0.0002 0.02 0.01 0.006 0.84 0.68 0.0006 0.21

⫺0.01 ⫺0.11 ⫺0.02 ⫺0.07 ⫺0.07 ⫺0.07 ⫺0.07 ⫺0.01 ⫺0.08 ⫺0.06 ⫺0.004

0.01 0.16 ⫺0.01 0.21 0.12 0.11 0.16 0.02 ⫺0.03 0.07 0.04

11 24 25 32 23 29 31 R. L. 36 46

⫺0.19 ⫺1.27 ⫺0.60 ⫺1.46 ⫺0.65 ⫺0.10 ⫺0.38 ⫺1.86 ⫺1.73 ⫺0.43 ⫺0.39

⫺0.44 ⫺1.30 ⫺1.45 ⫺1.27 ⫺1.42 ⫺0.93 ⫺1.30 ⫺3.1 ⫺2.23 ⫺0.5 ⫺0.53

2.01 0.07 1.99 ⫺0.34 1.32 1.66 1.67 1.18 0.53 0.24 0.63

0.05 0.95 0.06 0.74 0.20 0.11 0.11 0.25 0.60 0.81 0.53

5.13 16.40 0.31 13.37 2.37 3.01 4.97 1.29 1.02 10.07 2.10

0.03 0.0004 0.59 0.001 0.14 0.09 0.03 0.27 0.32 0.004 0.16

⫺0.01 ⫺0.12 ⫺0.02 ⫺0.1 ⫺0.08 ⫺0.11 ⫺0.09 ⫺0.08 ⫺0.14 ⫺0.08 ⫺0.01

0.02 0.15 0.01 0.16 0.05 0.02 0.09 0.10 0.002 0.05 0.04

Note. Region of Interest (ROI) analysis in which we examined activation in IGT-Disadvantageous (Dis.) relative to IGT-Advantageous (Adv.) older adults in the Prechoice phase, the Prefeedback phase, and the Feedback phase of the decision. Depicted on the left are the brain regions and beta weights for each group, showing the relative strength in activation, controlling for age. Second, depicted on the right side of the table are the IGT-Status ⫻ Age interaction results, including the age-activation slopes for each group. All p-values are averaged across the voxels within each ROI (uncorrected). VMPC ⫽ ventromedial prefrontal cortex; ACC ⫽ anterior cingulate cortex; PCC ⫽ posterior cingulate cortex; VS ⫽ ventral striatum; PHc ⫽ parahippocampal gyrus; DLPFC ⫽ dorsolateral prefrontal cortex.

We next examined the interaction between age and IGT-Status to determine whether age was differentially associated with activation in each group. During the Prechoice phase, as age increased, the IGTAdvantageous older adults showed increasing activation across several ROIs (see Table 2). By contrast, as age increased among the IGT-Disadvantageous older adults, activation tended to decrease. In other words, the two groups showed opposing age-activation slopes in several regions, including the VMPC (F28, ave ⫽ 16.08, pave ⬍ 0.001). Among the IGT-Advantageous, age positively correlated with activation in the VMPC, radv ⫽ 0.60. Among the IGT-Disadvantageous, age negatively correlated with activation, rdisadv ⫽ ⫺0.64 (see Figure 3). Similarly, age interacted with IGT-Status in the PCC (e.g., BA 29, F28, ave ⫽ 8.05, pave ⫽ 0.009, radv ⫽ 0.54, rdisadv ⫽ ⫺0.37). A similar pattern emerged in both the Prefeedback and Feedback phase, such that increasing age showed a positive slope with activation among the IGT-Advantageous older adults, but increasing age

showed a negative slope with activation among the IGTDisadvantageous older adults. This occurred in the VMPC and PCC, as well as in the parahippocampal gyrus. See Table 2 for a summary of the results.

Whole-Brain Results We next tested whole-brain BOLD signal differences between the IGT-Disadvantageous and IGT-Advantageous older adults across each phase of the decision-making process (Prechoice, Prefeedback, and Feedback). For all of the following results, IGT-Disadvantageous older adults had greater activation relative to IGT-Advantageous older adults in the indicated regions at a cluster corrected threshold of ␣ ⬍ .05. Specifically, the IGT-Disadvantageous older adults had greater activation relative to IGT-Advantageous older adults in the VMPC (␩2 ⫽ 0.32– 0.43) and PCC (␩2 ⫽ 0.32) during Prechoice phase, in

This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

NEUROIMAGING OF GAMBLING TASK IN OLDER ADULTS

877

Figure 3. IGT-Advantageous older adults show greater activation with increasing age (r ⫽ 0.60) whereas IGT-Disadvantageous older adults show less activation with increasing age (r ⫽ ⫺0.64). Beta weights indicate the strength of the activation in the ventromedial prefrontal cortex (VMPC), specifically in Brodmann Area (BA) 24.

the Precuneus (␩2 ⫽ 0.46), middle frontal gyrus (␩2 ⫽ 0.40), and superior temporal gyrus (␩2 ⫽ 0.48) during PreFeedback phase, and in the occipital lobe (␩2 ⫽ 0.32– 0.35), parietal lobe (␩2 ⫽ 0.36), and superior frontal gyrus (␩2 ⫽ 0.45) during Feedback phase. These results are summarized in Table 3 and Figure 4. Finally, we examined the main effect of task half, as well as the effect of IGT status on the difference between task halves (Half Interaction) to analyze the interaction between IGT and activation during each half of the task. We contrasted early trials (1– 40) to later trials (41– 80) for the ROIs summarized in Table 3. There was a main

effect of task half in the left superior temporal gyrus during the Prefeedback phase (t30 ⫽ 2.1, p ⫽ 0.04, d ⫽ 0.38). We did not find any significant effects of IGT on the contrast between halves (TaskHalf Interaction; ps ⬎ 0.05, ␩2 ⬍ 0.12). Results are summarized in Table 3.

Discussion We report on the neural correlates of IGT performance during distinct phases of the decision-making process (e.g., Prechoice,

Table 3 Whole-Brain Results Contrasting IGT-Disadvantageous (Dis.) and IGT-Advantageous (Adv.) Older Adults, Controlling for Age Clusters significant at ␣ ⫽ .05 (corrected) Region Prechoice L. PFC/BA 25 L. MFG/BA 10 L. PCC/BA 31 L. MFG/BA 11 L. SOG/BA 19 Prefeedback L. Precuneus/BA 31 L. MFG/BA 6 L. STG/BA 13 Feedback L. Cuneus/BA 19 R. SFG/BA 6 L. SFG/BA 6 L. Cuneus/BA 17 L. Parietal/BA 1

Beta weights

Half interaction

Dis.

Adv.

␩2

tdis ⬎ adv

␩2

941 1003 854 998 917

0.21 ⫺1.37 0.29 ⫺0.19 0.42

⫺2.09 ⫺4.03 ⫺1.50 ⫺2.57 ⫺1.59

0.43 0.32 0.32 0.34 0.34

⫺0.86 ⫺0.93 1.10 0.34 1.18

0.03 0.03 0.04 0.005 0.05

27 57 21

23,267 1219 3625

⫺0.23 ⫺0.95 0.40

⫺2.70 ⫺3.73 ⫺1.96

0.46 0.40 0.48

⫺1.98 ⫺1.29 ⫺1.53

0.12 0.06 0.08

33 60 63 9 52

2346 1054 2289 1060 1040

0.88 ⫺0.35 ⫺0.58 0.81 ⫺0.62

⫺1.75 ⫺3.38 ⫺3.52 ⫺1.78 ⫺3.08

0.35 0.39 0.45 0.32 0.36

⫺1.00 ⫺0.51 ⫺0.19 ⫺0.55 ⫺0.89

0.03 0.01 0.001 0.01 0.03

x

y

z

⫺4 ⫺4 ⫺13 ⫺22 ⫺34

4 55 ⫺59 28 ⫺80

⫺12 6 18 ⫺12 30

⫺13 ⫺7 ⫺43

⫺59 1 ⫺44

⫺7 8 ⫺7 ⫺22 ⫺52

⫺77 13 7 ⫺77 ⫺17

size (mm3)

Note. Results from whole-brain analysis showing the brain region, Brodmann Area (BA), Peak Talairach coordinates, and cluster size for each cluster in which there is a relative activation difference between the IGT-Disadvantageous and IGT-Advantageous older adults (controlling for age) during the Prechoice, Prefeedback and Feedback phases of the Iowa Gambling Task. All results are significant at the ␣ ⫽ .05 (corrected) level. IGT-Disadvantageous older adults showed greater relative activation in each cluster. Beta weights indicate the relative strength of activation for each group. The right-most two columns depict results from subtracting activation during early trials (1– 40) from later trials (41– 80) for each cluster resulting from the whole-brain analysis. We then tested the effect of IGT-Status on this contrast, controlling for age (Half Interaction). None of the half interaction results met significance. PFC ⫽ prefrontal cortex; MFG ⫽ middle frontal gyrus; PCC ⫽ posterior cingulate cortex; SOG ⫽ superior occipital gyrus; STG ⫽ superior temporal gyrus; SFG ⫽ superior frontal gyrus.

This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

878

HALFMANN, HEDGCOCK, BECHARA, AND DENBURG

Figure 4. IGT-Disadvantageous older adults had greater activation relative to IGT-Advantageous older adults in the (A) VMPC and PCC during Prechoice, (B) precuneus, middle frontal gyrus, and superior temporal gyrus during Prefeedback, and (C) occipital lobe, parietal lobe, and superior frontal gyrus during Feedback. All results shown are controlling for age and at a cluster corrected threshold of ␣ ⬍ .05. See the online article for the color version of this figure.

Prefeedback, and Feedback) among older adults. Moreover, we report on the divergent relationship between age and activation for IGT-Disadvantageous relative to IGT-Advantageous older adults. We have three key findings: (1) IGT-Disadvantageous older adults showed relatively greater activation than IGT-Advantageous older adults in more anterior brain regions (e.g., the VMPC) during the Prechoice phase; (2) IGT-Disadvantageous older adults showed relatively greater activation than the IGT-Advantageous older adults in more posterior brain regions (e.g., the precuneus) during the Prefeedback and Feedback phases; and (3) Increasing age corresponded with reduced activation in the aforementioned regions among the IGT-Disadvantageous older adults but with greater activation among the IGT-Advantageous older adults. The observed findings are consistent with the notion that IGTDisadvantageous and -Advantageous older adults show divergent aging trajectories. These trajectories are characterized by differing activation patterns in neural circuits involved in reward processing, as well as in related convergence zones that help guide decisions (e.g., the VMPC). Relative activation patterns between groups differed as a function of the decision phase. Focusing on the whole-brain results, we found

that the IGT-Disadvantageous older adults showed greater activation in several prefrontal regions, including the VMPC, and in the PCC during the Prechoice phase. The VMPC acts as a convergence zone, and the PCC functions as part of an emotion-based system (Damasio et al., 2000); therefore, we may be observing differing activation related to these processes. Consistent with previous work in older adults (e.g., Rogalsky et al., 2012), we, too, observed that relative activation difference between IGT groups was most pronounced in the left PFC during the Prechoice phase. Unlike Rogalsky et al. (2012), we observed greater activation among disadvantageous performers. This may be explained, in part, by the fact that they used a nongambling control task whereas we focused on relative activation levels between groups. In addition, we separated the phases of the decision-making process and examined activation specifically during card selection (i.e., the Prechoice phase). Lastly, demographics such as age and sex may play a role, and although Rogalsky et al. (2012) focused on sex, we only focused on age in our analyses. During the Prefeedback phase, we observed activation differences in the PCC, the middle temporal gyrus, and the dorsomedial PFC. During the Feedback phase, we observed a similar relative activation pattern but to a lesser extent. PCC activation differences align with the notion that IGT groups differ in affect-based processing in anticipation of and response to the gain/loss feedback. Further work needs to be conducted to more specifically parse out activation differences between gains and losses in IGTDisadvantageous and -Advantageous older adults. Age correlated with activation differently between the IGTDisadvantageous and -Advantageous older adults, aligning with the idea that these two groups of older adults age along divergent trajectories. These correlations emerged across decision phases, primarily in the VMPC and PCC, but also in the parahippocampal gyrus during the Prefeedback and Feedback phases. The VMPC is one of the first regions to deteriorate in older adulthood (Raz et al., 1998), and with this decline, the IGT-Advantageous older adults may initially rely on additional resources (e.g., sensory, memory, attention regions) to compensate for structural deficits. The compensation strategies employed across the IGT-Advantageous older adults might be variable enough such that no consistent activation patterns emerge in the VMPC. IGT-Advantageous older adults may rely more heavily on the most “accessible” decision-making resource (e.g., the VMPC and related structures) as they advance in age and as the additional resources (e.g., sensory, memory) deteriorate. The IGTDisadvantageous older adults may rely on more accessible resources (i.e., the VMPC) early on in the aging process and then experience dedifferentiation with increasing age (Eyler, Sherzai, Kaup, & Jeste, 2011), leading to reduced activation in the VMPC with age. Similarly, IGT-Disadvantageous older adults could have greater “noise” in their neural processing (Reuter-Lorenz & Park, 2010). For example, IGT-Disadvantageous older adults might not disengage the default mode network during the IGT. However, this does not explain the emerging pattern of activation across the three phases of the decision-making process. To elaborate, incongruent brain regions activated during the Prechoice and Feedback phases as opposed to the entirety of the default mode network, suggesting that specific functions linked to decision-making rather than general default mode activation contribute to the observed disparities. Instead, we reason differing compensatory mechanisms (e.g., VMPC) during the Prechoice phase, and differing reward sensitivity (e.g., PCC and precuneus) during the Feedback phase, likely contrib-

This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

NEUROIMAGING OF GAMBLING TASK IN OLDER ADULTS

utes to the observed activation patterns. In fact, this latter proposition is consistent with the notion that older adulthood leads to a positivity bias as proposed in Socioemotional Selectivity Theory (Carstensen, Isaacowitz, & Charles, 1999). Thus, increased attention to positive (e.g., gain) relative to negative (e.g., loss) information may lead to or be caused by the greater activation levels we observed in the IGTDisadvantageous older adults, but this remains to be tested. Atrophy could contribute to the observed relative activation patterns. To test this hypothesis, we conducted a comparison between the IGT-Disadvantageous and -Advantageous older adults PFC volumes, parsed out into white and gray matter in the dorsomedial, dorsolateral, orbitofrontal, ventromedial, and ventrolateral prefrontal cortices. Using a student’s t test, we did not observe any group differences in volume (ps ⬎ 0.14), suggesting that changes in structure are not driving the functional group effect. Relatedly, the finding that IGT-Disadvantageous older adults showed greater relative activation, on average, than the IGT-Advantageous older adults dissociates our participants from patients with focal damage to the VMPC because, even though they are engaging this brain region during the task more so than the IGT-Advantageous older adults, they perform worse on the IGT. This suggests that the behavioral decision-making deficits we observed are qualitatively different than those in the lesion patients (Bechara et al., 1997). Although we can largely rule out the contribution of structural atrophy, we cannot fully rule out the possibility that the groups differ in vascular health status. For instance, one subset of older adults might be experiencing greater levels of neurovascular disease or have differing rates of basal cerebral blood flow. We think this is unlikely because, although we do not have detailed vascular health data on this specific sample, we tested the vascular health (e.g., cardiovascular and neurological scales, among others, from the ACE-27) between IGTDisadvantageous and -Advantageous older adults in an overlapping and demographically similar sample (N ⫽ 50) and found a) no differences between groups and b) no risk for vascular problems greater than mild risk. Within our sample, we observed no differences between groups in smoking status or body mass index (see Table 1). In addition, Moser et al. (2012) found that cerebral blood flow related to neuropsychological functioning, and we do not observe differences in neuropsychological functioning between our groups. Relatedly, if basal blood flow or vascular disease caused the detected differences, we would have expected to observe effects in a greater array of brain regions and across decision phases (D’Esposito, Deouell, & Gazzaley, 2003). In fact, the differences we observed between IGT-Disadvantageous and -Advantageous older adults largely occurred in the absence of differences in brain regions that serve working memory or motor functions (e.g., dorsolateral PFC). Also, the relative activation effects occurred in the absence of neuropsychological differences between groups. Thus, the behavioral disparities in decision-making on the IGT are also unlikely to be secondary to vascular health or cognitive and neuropsychological performance alone. Our study differs from many others because we contrast two groups of cognitively intact older adults rather than comparing older adults to younger adults. Comparing older adults to each other, rather than younger adults, precludes the confounding factors of cohort effects and age-related disparities in brain volume and hemodynamics commonly encountered in fMRI studies on aging (Samanez-Larkin & D’Esposito, 2008). This design, however, does not allow us to determine activation during normal task

879

performance on the IGT. To mitigate this concern, we used ROIs based on Li et al.’s (2010) findings regarding normal activation during the IGT compared to a control task. This study is not without its limitations. First, the behavioral performance on the K’L’M’N= alternate version of the IGT was not as differentiating between the advantageous and disadvantageous performers as we observed on the A’B’C’D’ version. This could be because we had 80 trials in our task, rather than 100. Alternatively, it could be because the K’L’M’N= version is considerably more challenging than the original A’B’C’D’ version (Xiao et al., 2013). However, reduced behavioral differences while undergoing fMRI actually help in the interpretation of our findings: the functional differences we observed are not secondary to how the groups performed on the task and the feedback they received. In addition, our analyses focused on several ROIs that tend to be susceptible to signal drop out, such as VMPC, which can be a limitation. There are several ways to address this problem, including increasing the bandwidth to reduce the signal-to-noise ratio, decreasing slice thickness, and increasing the amount of brain tissue acquired by using oblique prescriptions (Buxton, 2002). We chose to use the third approach and were able to demonstrate adequate functional coverage in our ROIs (see Figure 1). Examining functional connectivity during resting state could shed additional light on the functional differences between these groups. Also, given the complexity of the IGT, it would of interesting to combine data from imaging during the IGT with tasks that engage each of the component processes (e.g., ambiguity, risk, reward, and punishment) uniquely. This would allow us to better disentangle the processes that may contribute to declines in decision making at a more complex level.

References Bechara, A. (2004). Disturbances of emotion regulation after focal brain lesions. International Review of Neurobiology, 62, 159 –193. doi: 10.1016/S0074-7742(04)62006-X Bechara, A. (2007). Iowa Gambling Task professional manual. Lutz, FL: Psychological Assessment Resources. Bechara, A., Damasio, H., & Damasio, A. R. (2000). Emotion, decision making and the orbitofrontal cortex. Cerebral Cortex, 10, 295–307. Bechara, A., Damasio, H., Tranel, D., & Damasio, A. R. (1997). Deciding advantageously before knowing the advantageous strategy. Science, 275, 1293–1295. doi:10.1126/science.275.5304.1293 Bechara, A., Tranel, D., & Damasio, H. (2000). Characterization of the decision-making deficit of patients with ventromedial prefrontal cortex lesions. Brain: A Journal of Neurology, 123, 2189 –2202. doi:10.1093/ brain/123.11.2189 Brainard, D. H. (1997). The Psychophysics Toolbox. Spatial Vision, 10, 433– 436. doi:10.1163/156856897X00357 Bruine de Bruin, W., Parker, A. M., & Fischhoff, B. (2012). Explaining adult age differences in decision-making competence. Journal of Behavioral Decision Making, 25, 353–360. Buxton, R. B. (2002). Introduction to functional magnetic resonance imaging: Principles and techniques. Cambridge, UK: Cambridge University Press. doi:10.1017/CBO9780511549854 Cabeza, R., Anderson, N. D., Locantore, J. K., & McIntosh, A. R. (2002). Aging gracefully: Compensatory brain activity in high-performing older adults. NeuroImage, 17, 1394 –1402. doi:10.1006/nimg.2002.1280 Carstensen, L. L., Isaacowitz, D. M., & Charles, S. T. (1999). Taking time seriously: A theory of socioemotional selectivity. American Psychologist, 54, 165–181. doi:10.1037/0003-066X.54.3.165

This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

880

HALFMANN, HEDGCOCK, BECHARA, AND DENBURG

Cohen, S., Janicki-Deverts, D., & Miller, G. E. (2007). Psychological stress and disease. Journal of the American Medical Association, 298, 1685– 1687. doi:10.1001/jama.298.14.1685 Damasio, A. R. (1996). The somatic marker hypothesis and the possible functions of the prefrontal cortex. Philosophical Transactions of the Royal Society of London B: Biological Sciences, 351, 1413–1420. doi: 10.1098/rstb.1996.0125 Damasio, A. R., Anderson, S. W., & Tranel, D. (2011). The frontal lobes. In K. M. Heilman & E. Valenstein (Eds.), Clinical neuropsychology (5th ed., pp. 417– 465). New York, NY: Oxford University Press. Damasio, A. R., Grabowski, T. J., Bechara, A., Damasio, H., Ponto, L. B. L., Parvizi, J., & Hichwa, R. D. (2000). Subcortical and cortical brain activity during the feeling of self-generated emotions. Nature Neuroscience, 3, 1049 –1056. doi:10.1038/79871 Denburg, N. L., Cole, C. A., Hernandez, M., Yamada, T. H., Tranel, D., Bechara, A., & Wallace, R. B. (2007). The orbitofrontal cortex, realworld decision making, and normal aging. Annals of the New York Academy of Sciences, 1121, 480 – 498. doi:10.1196/annals.1401.031 Denburg, N. L., Recknor, E. C., Bechara, A., & Tranel, D. (2006). Psychophysiological anticipation of positive outcomes promotes advantageous decision making in normal older persons. International Journal of Psychophysiology, 61, 19 –25. doi:10.1016/j.ijpsycho.2005.10.021 Denburg, N. L., Tranel, D., & Bechara, A. (2005). The ability to decide advantageously declines prematurely in some normal older persons. Neuropsychologia, 43, 1099 –1106. doi:10.1016/j.neuropsychologia .2004.09.012 D’Esposito, M., Deouell, L. Y., & Gazzaley, A. (2003). Alterations in the BOLD fMRI signal with ageing and disease: A challenge for neuroimaging. Nature Reviews: Neuroscience, 4, 863– 872. doi:10.1038/nrn1246 Ernst, M., Bolla, K., Mouratidis, M., Contoreggi, C., Matochik, J. A., Kurian, V., . . . London, E. D. (2002). Decision making in a risk-taking task: A PET study. Neuropsychopharmacology, 26, 682– 691. doi: 10.1016/S0893-133X(01)00414-6 Eyler, L. T., Sherzai, A., Kaup, A. K., & Jeste, D. V. (2011). A review of functional brain imaging correlates of successful cognitive aging. Biological Psychiatry, 70, 115–122. doi:10.1016/j.biopsych.2010.12.032 Forman, S. D., Cohen, J. D., Fitzgerald, M., Eddy, W. F., Mintun, M. A., & Noll, D. C. (1995). Improved assessment of significant activation in functional magnetic resonance imaging (fMRI): Use of a cluster-size threshold. Magnetic Resonance in Medicine, 33, 636 – 647. doi:10.1002/ mrm.1910330508 Gläscher, J., Adolphs, R., Damasio, H., Bechara, A., Rudrauf, D., Calamia, M., . . . Tranel, D. (2012). Lesion mapping of cognitive control and value-based decision making in the prefrontal cortex. Proceedings of the National Academy of Sciences of the United States of America, 109, 14681–14686. doi:10.1073/pnas.1206608109 Glimcher, P. W., & Rustichini, A. (2004). Neuroeconomics: The consilience of brain and decision. Science, 306, 447– 452. doi:10.1126/science .1102566 Halfmann, K., Hedgcock, W., & Denburg, N. L. (2013). Age-related differences in discounting future gains and losses. Journal of Neuroscience, Psychology, and Economics, 6, 42–54. doi:10.1037/npe0000003 Lezak, M. D., Howieson, D. B., Bigler, E. D., & Tranel, D. (2012). Neuropsychological assessment (5th ed.). New York, NY: Oxford University Press. Li, X., Lu, Z.-L., D’Argembeau, A., Ng, M., & Bechara, A. (2010). The Iowa Gambling Task in fMRI images. Human Brain Mapping, 31, 410 – 423.

Liang, K., & Zeger, S. L. (1986). Longitudinal data analysis using generalized linear models. Biometrika, 73, 13–22. doi:10.1093/biomet/73.1.13 Moser, D. J., Boles Ponto, L. L., Miller, I. N., Schultz, S. K., Menda, Y., Arndt, S., & Nopoulos, P. C. (2012). Cerebral blood flow and neuropsychological functioning in elderly vascular disease patients. Journal of Clinical and Experimental Neuropsychology, 34, 220 –225. doi:10.1080/ 13803395.2011.630653 Öngür, D., & Price, J. L. (2000). The organization of networks within the orbital and medial prefrontal cortex of rats, monkeys, and humans. Cerebral Cortex, 10, 206 –219. doi:10.1093/cercor/10.3.206 Raz, N., Gunning, F. M., Head, D., Dupuis, J. H., McQuain, J., Briggs, S. D., . . . Acker, J. D. (1997). Selective aging of the human cerebral cortex observed in vivo: Differential vulnerability of the prefrontal gray matter. Cerebral Cortex, 7, 268 –282. Raz, N., Gunning-Dixon, F. M., Head, D., Dupuis, J. H., & Acker, J. D. (1998). Neuroanatomical correlates of cognitive aging: Evidence from structural magnetic resonance imaging. Neuropsychology, 12, 95–114. doi: 10.1037/0894-4105.12.1.95 Resnick, S. M., Pham, D. L., Kraut, M. A., Zonderman, A. B., & Davatzikos, C. (2003). Longitudinal magnetic resonance imaging studies of older adults: A shrinking brain. The Journal of Neuroscience, 23, 3295–3301. Reuter-Lorenz, P. A., & Park, D. C. (2010). Human neuroscience and the aging mind: A new look at old problems. The Journals of Gerontology, Series B: Psychological Sciences and Social Sciences, 65B, 405– 415. doi:10.1093/geronb/gbq035 Rogalsky, C., Vidal, C., Li, X., & Damasio, H. (2012). Risky decisionmaking in older adults without cognitive deficits: An fMRI study of VMPFC using the Iowa Gambling Task. Social Neuroscience, 7, 178 – 190. doi:10.1080/17470919.2011.588340 Samanez-Larkin, G. R., & D’Esposito, M. (2008). Group comparisons: Imaging the aging brain. Social Cognitive and Affective Neuroscience, 3, 290 –297. doi:10.1093/scan/nsn029 Talairach, J., & Tournoux, P. (1988). Co-planar stereotaxic atlas of the human brain. New York, NY: Thieme Medical Publishers, Inc. Waters-Wood, S. M., Xiao, L., Denburg, N. L., Hernandez, M., & Bechara, A. (2012). Failure to learn from repeated mistakes: Persistent decision-making impairments as measured by the Iowa Gambling Task in patients with ventromedial prefrontal cortex. Journal of the International Neuropsychological Society, 18, 927–930. doi:10.1017/S135561771200063X West, R., Murphy, K. J., Armilio, M. L., Craik, F. I. M., & Stuss, D. T. (2002). Lapses of intention and performance variability reveal agerelated increases in fluctuations of executive control. Brain and Cognition, 49, 402– 419. doi:10.1006/brcg.2001.1507 Wood, S., Busemeyer, J., Koling, A., Cox, C. R., & Davis, H. (2005). Older adults as adaptive decision makers: Evidence from the Iowa Gambling Task. Psychology and Aging, 20, 220 –225. doi:10.1037/ 0882-7974.20.2.220 Xiao, L., Wood, S. M., Denburg, N. L., Moreno, G. L., Hernandez, M., & Bechara, A. (2013). Is there a recovery of decision-making function after frontal lobe damage? A study using alternative versions of the Iowa Gambling Task. Journal of Clinical and Experimental Neuropsychology, 35, 518 –529. doi:10.1080/13803395.2013.789484

Received August 8, 2013 Revision received May 23, 2014 Accepted May 27, 2014 䡲

Functional neuroimaging of the Iowa Gambling Task in older adults.

The neural systems most susceptible to age-related decline mirror the systems linked to decision making. Yet, the neural processes underlying decision...
572KB Sizes 2 Downloads 4 Views