The aims of this study were twofold: (1) to use meta-analysis to characterize the ERN in OC-affected individuals delineated by ERN reflections of task type, and (2) to use pooled analyses of original raw data to examine if individuals with a specific OCD/S subtype, i.e., OCD/S with hoarding, exhibit specific ERN abnormalities relative to those without hoarding. We hypothesized that the ERN would be larger (i.e., more negative) in OC-affected participants than in controls, and further, that this response would be specific to those studies that used response conflict tasks. This hypothesis was based in part on the findings reported by Grundler et al., who found a hyperactive ERN in a response conflict task and a hypoactive ERN in a probabilistic learning task. Specifically, they predicted that hyperactive error monitoring circuits would lead to repetition and over-learning in OC-affected individuals (to avoid making choices that lead to errors) in response conflict tasks. It was further predicted by Grundler et al. that in uncertain environments requiring adaptive responses, such as in reinforcement or probabilistic learning tasks, OC-affected individuals would show enhanced ERN mechanisms as a reflection of rigidity and avoidance learning (learning to avoid maladaptive choices). However, Grundler et al. reported that in his study, OCD patients ultimately demonstrated hypoactive ERN during reinforcement learning tasks.
The present findings from our meta-analysis confirm the prediction of hyperactive ERN responses for those studies using a response conflict task, and in fact show that there is a large effect size (−0.86). All of the studies showed effects in the same direction (larger ERN for OC-affected individuals) and there was no significant evidence of between-study heterogeneity, suggesting that the results of the meta-analysis are quite robust. In contrast, the meta-analysis did not support the presence of ERN abnormalities when tested with probabilistic learning tasks. Due to the high level of heterogeneity between studies for this task type, the failure to find ERN abnormalities in OCD/S may be due to specific weaknesses of study design in one or more of the studies, or greater differences between these studies due to subject recruitment, task type, or the point in which the ERN, rather than the FRN, is elicited over the course of the task. Alternatively, it may be that probabilistic learning tasks are not an appropriate tool to detect ERN alterations in OCD/S if the source of the ERN abnormality cannot be defined; that is, group differences in ERN responses that arise in probabilistic learning tasks may be due to differences in the underlying physiology, but may also be due to differences in how well each group learned the task. In any case, the result from probabilistic learning tasks was not statistically significant, and the high level of heterogeneity between studies suggests that the results of the meta-analysis for this condition are not entirely reliable. While it is possible that variability in task design across studies have led to inconsistencies in the literature, or that the inclusion of OC-affected individuals who do not necessarily meet clinical criteria for OCD biases the heterogeneity among probabilistic learning tasks, these task and sample properties were present in the response conflict condition as well, but still led to conclusive findings. Therefore, at the least, evidence for enhanced error signals in OCD/S appears to be convergent across response conflict studies, and is suggestive of a neural basis for the pathological repetitious behavior observed in these individuals.
In addition to the meta-analysis of ERN differences across studies examining error-monitoring dysfunction in OCD/S, we conducted a critical analysis of the ERN in OC-affected individuals with hoarding symptoms vs.
OC-affected individuals without hoarding in a subgroup of studies (N
= 4, total of 7 independent conditions) for which we had symptom-level data. In contrast to our hypothesis, we found no significant differences between those with hoarding and those without hoarding in the ERN for the overall sample (not separated by task type), or for the response conflict task. However, we did find a trend for an enhancement in the ERN for those with hoarding symptoms compared to those without among the OCD/S participants in the probabilistic learning task (). As suggested previously (Frank et al., 2005
; Holroyd and Coles, 2002
), a larger ERN in the probabilistic learning task has been associated with a greater tendency to avoid an error, i.e.
, the “NoGo bias.” As evidenced by individuals exhibiting hoarding behaviors, and considering the strong reluctance with which OCD patients part with belongings, even if dangerous, non-sentimental, or unhygienic, avoidance behaviors are necessarily the rule, and not the exception, in defining hoarding. Accordingly, our findings suggest that OC-affected individuals with hoarding symptoms may express neural circuitry consistent with a tendency toward an avoidance bias.
Interestingly, our ERN results in individuals who hoard are consistent with Grundler’s original hypothesis predicting an enhanced ERN in reinforcement or probabilistic learning tasks, as a reflection of avoidance learning (learning to avoid maladaptive choices) (Grundler et al., 2009
). While there was no evidence of a hypoactive ERN for OCD/S participants compared to healthy controls in probabilistic learning tasks in the overall meta-analysis, the results of the analysis of the original pooled symptom data suggest that OC-affected individuals with hoarding behaviors have a hyperactive
rather than a hypoactive
ERN when compared to OCD/S individuals without hoarding behaviors. This result was unexpected, and the results merely trending toward significance, may be due to stochastic variation within the sample. Alternatively, individuals with hoarding behaviors may have more difficulty with the probabilistic learning task, have slower learning rates, or may take longer to complete the task. Nevertheless, despite the heterogeneity across these scenarios, OC-affected individuals who hoard appear to have distinctive pathophysiology underlying error monitoring compared to OC-affected individuals without hoarding behaviors. Under this hypothesis, individuals who hoard may generate an enhanced ERN in the context of a reinforcement environment, such that the NoGo bias would result in an over-activation of avoidance and rigidity.
Compulsive hoarding is currently characterized as a symptom subtype of OCD in the DSM-IV-TR (APA, 2000
). However, as mentioned previously, several researchers have suggested that hoarding behaviors are part of a discrete clinical syndrome comprising additional symptoms and abnormalities suggestive of frontally mediated neurocognitive dysfunction (Grisham et al., 2007
; Hartl et al., 2004
; Lawrence et al., 2006
; Mataix-Cols et al., 2010
; Pertusa et al., 2008
; Saxena, 2007
; Steketee and Frost, 2003
). Our studies of the neurocognitive function of older individuals with and without hoarding behaviors support this hypothesis (Mackin et al., 2010
), and suggest that those with hoarding behaviors have several abnormalities of executive function, most prominently, impairments in speed of information processing, as well as deficits in categorization and attention/working memory. Consistent with this hypothesis and with the results from imaging and other studies, the results of the current study suggest that individuals with compulsive hoarding may have an overlapping, yet distinct, neural circuitry from that of individuals with OCD/S but without hoarding. The putative neurocognitive profile of individuals with compulsive hoarding, which awaits confirmation in larger samples, consists of (1) abnormalities in speed of information processing, categorization, attention and working memory, and (2) a hyperactive ERN in probabilistic learning tasks but not in response conflict tasks. As more information is acquired on the task-related pathophysiology underlying the discrepant symptom clusters of OCD, clarification of subtype-specific cognitive impairments may subserve targeted interventions.
The primary limitation of this study is that it uses previously collected data from a variety of research groups, and thus, data on specific severity of symptoms, and clinical or functional impact of symptoms are not available. This is particularly relevant for the hoarding analysis; in this analysis, we were only able to examine the relationship of the ERN in the context of the presence or the absence of hoarding symptoms among individuals with OCD or high OCS. All individuals with hoarding symptoms also had additional obsessive–compulsive symptoms, and we do not have information about whether the hoarding symptoms were the primary symptoms, nor about whether they were clinically impairing. Future work with individuals currently diagnosed with OCD, yet primarily defined by their hoarding behaviors, is important to reduce sample heterogeneity, and to gain further understanding of the distinct neurocognitive and neurophysiological abnormalities associated with the hoarding subtype; examination of the ERN within participants with primary hoarding is ongoing in our laboratory.
The use of secondary data leads to variance caused by differences in task type, phenotype definition, and measurement of the ERN, among other factors. We were able to measure the degree of heterogeneity between studies in the meta-analysis, and in fact, for the response conflict task, where there was little evidence of heterogeneity, the results were quite robust. To control for cross-study variance in the hoarding analysis, we used within-study z-scores rather than raw scores to capture the degree of difference from each study’s healthy control sample means; nevertheless, the remaining heterogeneity, particularly for the probabilistic learning tasks, may have contributed either to an under-estimate of the ERN differences in those with and without hoarding symptoms, or alternatively, an over-estimate. Neurophysiological and neuroimaging data contributing to the clarification of hoarding behaviors are currently underway.