The present findings of ACC and insula abnormality comport with emerging models of HD that emphasize problems in decision-making processes3,35
that contribute to patients’ difficulty discarding items. The region of ACC in the present findings is commonly associated with error monitoring under conditions of uncertainty,16,17
and the mid-to-anterior insula regions in the present findings are thought to be associated with interoception, perception of unpleasant feeling states, and the salience of stimuli,36,37
and emotion-driven decisions.39
Together, these regions are thought to be part of a functionally connected network of structures used to identify the emotional significance of a stimulus, generate an emotional response, and regulate affective state.36,37,40–42
The apparent biphasic pattern (ie, hypofunction to EPs but hyperfunction to PPs) of ACC and insula activity in patients with HD merits further study. These regions, considered to be at the core of a “salience network,”42
(p656) were hypoactive in patients with HD when not making decisions about their own possessions, a finding that is consistent with the baseline positron emission tomography results of Saxena et al.21
Low activity in this network, a pattern reminiscent of that seen in patients with autism spectrum disorders,43
may result in attenuated response to salient stimuli (including disgust-eliciting stimuli44,45
) and may contribute to the diminished motivation and poor insight frequently observed in patients with HD.14,15
When deciding about personal possessions, however, a shift is seen in which these same regions become hyperactive, a pattern seen among patients with anxiety disorders.46,47
Hyperactivity in these regions may hamper the decision-making process by leading to a greater sense of outcome uncertainty,47
which would be consistent with the present correlations of subjective indecisiveness and not just right feelings, contributing to a subjective sense that the wrong decision is being made,48
increasing response conflict, and leading to exaggerated risk judgments.49
The fact that the obtained interaction effect remained significant when controlling for VAS ratings, however, suggests that the emotional responses alone cannot account for the findings. Rather, it appears that hyperactivity in the ACC-insula network during decision making is characteristic of HD and may contribute to subjective indecisiveness and decisions to save.
Notably, among patients with HD, anxiety, indecisiveness, sadness, and not just right feelings were associated with inferior frontal gyrus activation, possibly indicating an inhibitory signal to accept a risky option.50,51
Findings from behavioral economics, such as the endowment effect52
in which ownership of an item can increase its perceived worth, may be informative for the study of hoarding. The endowment effect has been demonstrated to be associated with inflated estimates of the desirability of objects53,54
and may be mediated by activity in the inferior frontal gyrus and insula,55
regions implicated in the present study, and the ventral striatum and nucleus accumbens,56
which could be reflected in the obtained striatal activity. Contrary to previous research,22–24
we did not find evidence that PP decisions were associated with excessive activity in the ventromedial prefrontal cortex or OFC, perhaps reflecting the fact that PP and EP decisions taxed these decision-making regions equally.
Perhaps the most striking finding for the patients with OCD is the apparent absence of differential responding to PP vs EP items. The patients with OCD experienced less change in hemodynamic activity between these 2 item types than did HCs or patients with HD (for regions other than ACC and insula, the attenuated differentiation for patients with OCD drove the significant interaction effects). A possible explanation for this lack of differentiation is that the patients with OCD simply might have been less engaged with the task, perhaps due to ongoing obsessive ruminations or anxiety, although the higher rate of selective serotonin reuptake inhibitor use in that group may have a role. In addition, the present results for patients with HD differ from those seen in patients with OCD under symptom provocation conditions, in which greater activity is noted in OFC and prefrontal regions22,57–60
and the amyg-dala57,61
Furthermore, the HD-specific findings remained significant even while controlling for non-hoarding OCD symptoms (as well as depression).
Several study limitations should be noted. The first of these concerns the sampling procedures. Because an aim of the present study was to compare patients with primary HD to patients with primary OCD, potential participants with comorbid HD and OCD were excluded from the study. Although only one participant was excluded for this reason, previous research suggests that 18% of patients with HD meet diagnostic criteria for nonhoarding OCD.63
The self-report measures suggest some degree of subclinical hoarding (difficulty discarding) in patients with OCD and some degree of subclinical OCD symptoms (checking, neutralizing, obsessions, and/or ordering) in patients with HD. The equivalent not just right experiences during the fMRI task also may be indicative of an OCD-like process in patients with HD. The HD sample was predominantly female and white and thus may not represent the larger population of individuals with HD. Previous research has suggested that HD may be more common in men.64
The ethnic/racial composition of the HD population is not known. The present sample also likely did not include the least insightful (and perhaps most severely impaired) patients with HD who would not identify themselves as such and thus would not volunteer for the present study. Because of the younger age of the OCD group, an age effect cannot be definitively ruled out (although there was no age difference between the HD and HC groups and covarying for age did not alter the findings).
The second limitation involves the examination of the decision-making process. Decision making is a complex process that includes assessing available options, selecting the appropriate action, and evaluating outcomes associated with the action. The longer response time for patients with HD vs other participants introduces the possibility that the activation in the insula and ACC derives from greater time on task in regions associated with attentional control.64–66
Indeed, activity in the inferior frontal gyrus and insula was correlated with self-reported indecisiveness; however, results remained significant even after controlling for indecisiveness, suggesting that HD processing may be both longer than and qualitatively different from that in participants without HD. We considered controlling for response time; however, we ultimately decided not to because slowed decision making may be a central feature of impaired decision making in hoarding. Future studies might investigate the temporal dynamics of decision making to examine changes in neural activity from earlier to later stages, using behavioral tasks that are well suited for that purpose.67
The third limitation involves the task structure itself. The stimuli in the present study were limited to paper items. Although paper items are among the most commonly hoarded possessions in HD,1
other items, such as clothing, craft supplies, sentimental items, and food, are also commonly hoarded. It is possible that a different set of stimuli would have elicited a different neural response, and future research is needed to explore this possibility. Furthermore, it cannot be conclusively determined that the words “yours” and “ours,” which preceded the stimuli, or the red and black margins surrounding the stimuli affected neural response. An examination of neural response to these words and colors alone would have facilitated interpretation of the results. In addition, because the VASs were administered after each run (which contained both PP and EP decisions), the specific emotional reaction to decisions about PPs cannot be pinpointed. The addition of a true neutral baseline condition (rather than a comparison of 2 kinds of decision making) would also help clarify the full range of neural processes involved in the decision-making process.
Finally, although α levels were corrected for searching within an ROI (ie, the number of voxels) using small volume correction, the large number of ROIs searched underscores the need for replication in future studies. We considered whole-brain analyses as an alternative to ROI analysis; however, given the amount of statistical correction that would be needed for such analysis, the sample would have lacked sufficient statistical power to detect differences at the whole-brain level with the appropriate corrections.