To our knowledge, there are no published accounts of the neurofunctional effects of autism interventions. Our goal in this proof-of-principle case study was to demonstrate the feasibility and potential promise of investigations of neurofunctional alterations in the context of specific behavioral changes in response to particular treatments in autism. We employed a visual target-detection task that has been shown previously to elicit abnormal prefrontal activation in individuals with autism (Shafritz et al., 2008
Case 2 demonstrated marked improvements in RBs as assessed by the primary outcome measure, the CYBOCS-PDD, whereas Case 1 did not show CYBOCS-PDD improvement. Specifically, CYBOCS-PDD scores of Case 2 decreased from 16 to 10, a clinically significant decrease. In a sample of 172 medication-free children with pervasive developmental disorder, CYBOCS-PDD score averages were 14.4, with a standard deviation of 3.86 (Scahill et al., 2006
). Thus, CYBOCS-PDD scores of Case 2 dropped approximately 1.5 standard deviations, and changed from approximately one-half standard deviation above the mean to approximately one standard deviation below the mean. Of note, CYBOCS-PDD change in Case 2 did not parallel his CGI-I scores, suggesting that the effects of citalopram in Case 2 may have been constrained to changes in RBs, but not of sufficient global impact on overall functioning to reflect a large change in his CGI-I score.
Functional brain imaging data indicated that improvements in RB symptoms in Case 2 were accompanied by increased activation in a number of prefrontal regions, with greatest increased activation in the ACC (Broadmann’s Area 32). In contrast, there were no changes in the activity of the ACC or other relevant prefrontal regions in Case 1 who showed no reduction in RBs. We interpret increased activation of the ACC and other prefrontal regions in Case 2 potentially to reflect compensatory mechanisms that were engaged during performance of the target detection task after effective treatment. We interpret the lack of increased activation in these regions in Case 1 to reflect ineffective treatment. The increase in supplementary motor cortex activation in Case 1 may reflect the observed increase in RBs after citalopram treatment in this individual.
The direction of fMRI effects in the present context (i.e., greater prefrontal activation that corresponded with RB reductions in Case 2) warrants replication, particularly in light of a wealth of data demonstrating that autism is associated with relatively increased
prefrontal activation during cognitive control (see Dichter et al., 2009
for a review). In particular, Thakkar et al. (2008)
reported a significant correlation between increased ACC activation during an antisaccade task and increased RBs. As indicated above, in the present context, prefrontal activation increases may be conceptualized to reflect compensatory mechanisms that emerge soon after initiation of effective treatment that are not necessarily inconsistent with a pattern of general prefrontal hyperactivation during executive tasks in autism. Manoach (2003)
, in a review of an analogous issue in the schizophrenia literature, has suggested that variability in behavior and brain activation may best be regarded as intrinsic to heterogeneous disorders. Thus, we interpret the present findings within the framework of dysregulated and inefficient frontostriatal recruitment in autism during cognitive control tasks that is altered with citalopram-induced reductions in RB symptoms.
The ACC has a number of functions, but most specifically mediates overcoming prepotent responses (e.g., Devinsky, Morrell, & Vogt, 1995
) and online monitoring of errors in the presence of response conflict (MacDonald, Cohen, Stenger, & Carter, 2000
). In the present context, reductions in RBs, a real-world form of overcoming prepotent behavioral tendencies, may coincide with increased ACC activity while overcoming prepotent responses in the context of a target detection task.
More broadly, these neuroimaging findings offer unique information about the potential mechanisms of action of citalopram-induced RB symptom reductions. Changes in questionnaire-based measures of RBs in the context of a clinical trial may be due to what Braunholtz, Edwards, and Lilford (2001)
have summarized as ‘trial effects’, that is, the effects of participating in a clinical trial rather than the direct effects of a treatment. Such effects may include treatment effects, placebo effects, Hawthorne effects, and observer effects. Although such effects cannot definitely be ruled out in the present context, neuroimaging data may indicate that the benefits of a treatment are due to changes in functioning of relevant brain regions rather than such other factors.
We note a number of methodological features which warrant interpretive caution. The final citalopram dosage differed between participants. Case 1 experienced adverse effects (i.e., amotivation and mild fatigue) that may have compromised his performance and patterns of brain activation during the post-treatment oddball task, whereas Case 2 did not report adverse effects. Thus, it is possible that the differences in brain activation may reflect changes in RBs as well as differences in motivation. Differences in reaction times between pre-and post-treatment scans (i.e., Case 1, but not Case 2, showed increased reaction times) is a potential confound as well that may be addressed in future trials with adequate sample sizes that allow for including reaction time as a covariate in imaging analyses. The difference in RB type may have moderated treatment response as well, a possibility that will be evaluated in future, larger-scale studies. Additionally, the relatively long intervening time period between scans suggests that fMRI changes may not have been due solely to citalopram.
Regarding the analysis of fMRI data, we used a relatively liberal statistical threshold for change in these case reports, and future studies with larger samples will use more conservative approaches. Moreover, analyses in case studies are descriptive and qualitative. Firm inferences about linkages between changes in prefrontal activation and changes in RBs in response to treatment will require an examination of group-level data from the larger trial. We also note that the case study approach does not suggest conclusively a linkage between changes in RBs and changes in fMRI activation, and we interpret the present findings as supporting but not sufficient evidence of a causal link between changes in fMRI activation and changes in RB symptom severity.
Additionally, analyses focused on fMRI response to target events without controlling for responses to novel events, an approach that has been employed in group-level fMRI studies of set shifting in autism (i.e., Gomot et al., 2008
, Shafritz et al., 2008
). In this regard, it cannot be firmly concluded that neural responses were target-specific. This approach was used because the case study design did not yield sufficient statistical power to detect significant interaction effects (i.e., Time X [target vs. novel] effects), even at a liberal statistical threshold. Thus, responses to target events only were examined. However, we note that contrasting Time 2 and Time 1 scans controlled for each participant’s pre-treatment responses to target events, though this across-scan correction is clearly less optimal than a within-scan correction.
Finally, the poor accuracy of both patients restricted the number of trials analyzed. We note that Shafritz et al. (2008)
reported results from a highly similar visual oddball task where participants with autism spectrum disorders viewed 55 target events and achieved only 42% accuracy (i.e., an average of 23 correct responses). These rates are roughly comparable to the present context, where Case 2 achieved accuracies of 38% and 52% at timepoints 1 and 2 (i.e., 23 and 31 correct responses to target events, respectively) and Case 1 achieved accuracies of 35% and 21% at timepoints 1 and 2 (i.e., 21 and 13 correct responses to target events, respectively). Additionally, we note that event-related fMRI psychiatry studies have utilized far lower numbers of events than analyzed in the present study (e.g., Forbes et al., 2009
These two cases clearly demonstrate that fMRI studies are feasible and can be utilized longitudinally in individuals with autism who are treated with medication. These cases also demonstrate the importance of distinguishing between treatment responders and non-responders when assessing potential mechanisms of treatment action. These results also raise the intriguing possibility that increases in prefrontal brain activation during a visual oddball task may be a sensitive biomarker of reductions in RBs in autism. Though the utility of brain imaging is clearly limited in clinical settings, these data suggest that brain imaging is an invaluable research tool to elucidate neurobiological mediators of autism interventions. In summary, although by no means a definitive assessment of potential treatment mechanisms, the present case study results are consistent with the hypothesis that reductions in RBs in autism in response to treatment may be related to compensatory increases in activation of the ACC and other pre-frontal cortical brain regions, a pattern of findings that holds promise as a potential biomarker of treatment response in autism.