In this study, we examined how cortical thickness abnormalities in patients with BP vary according to the presence or absence of ADHD. Existing structural and functional brain imaging studies of adults with BP have, for the most part, neglected the potential importance of accounting for ADHD comorbidity when examining neural abnormalities. Several of our key findings support the importance of accounting for comorbid ADHD in structural neuroimaging studies of patients with BP. Interaction and pairwise analyses between groups revealed that the effect of BP on cortical thickness was different in patients with and without ADHD comorbidity in the right lateral OFC (BA47) and the left subgenual cingulate (BA25). The effects of BP and ADHD in these regions were found to be not additive, but rather interdependent, resulting in a unique phenotypic signature for the comorbid diagnostic group. However, in other subregions of the PFC (BA 8, 9, 10, and 11) and ACC (BA 24, 32, and 33), the effect of a BP diagnosis on cortical thickness was not changed by ADHD comorbidity status.
In the right lateral OFC (BA47), BP was associated with significant cortical thinning only in the absence of an ADHD diagnosis; however, in the presence of ADHD no such cortical thinning was detected. Results from prior studies of BP examining cortical thickness of the lateral OFC of the prefrontal cortex are variable (9
). Our findings suggest the possibility that a varying proportion of subjects with a comorbid ADHD diagnosis in each of these studies could account for this inconsistency. One recent study showed increases in cortical thickness in adults with ADHD in the PFC (24
) which, when combined with the cortical thinning associated with BP, may explain in our study the non-significant difference in the comorbid BP-ADHD relative to controls.
The interaction findings in left BA25 showed that BP-only patients did not differ from healthy controls in this region, but that the presence of ADHD and BP together was associated with cortical thinning relative to the ADHD-only group. Similar to the findings in BA47, BA25 has been previously reported to be anatomically and functionally abnormal in some studies of patients with BP (14
), but not in others. The interaction effect in BA25, and subsequent comparisons demonstrating no significant difference between the BP-only group and healthy controls in this region, again suggests that the discrepancy in results from prior studies could be due to differences in the proportion of patients in each study with an ADHD comorbidity. We would like to note, however, that the thinning associated with BP in the presence of ADHD was only trend-level significance, a finding that may or may not remain significant with larger population samples. Other studies using larger samples report reduced gray matter volume in BA25 in patients with BP (14
), suggesting that the trend level significance in our findings may simply be due to a lack of power. Interestingly, a follow-up analysis that included K-SADS scores for ADHD and comorbid patients (calculated by combining the hyperactive and inattentive scores) revealed that the initial trend level (p = 0.07) thinning of comorbid patients relative to ADHD-only patients reached significance (p = 0.04) when ADHD severity was accounted for in the model. Regressing the values out of the model gave us the important comparison of cortical thickness measurements in ADHD and comorbid patients with the same symptoms.
Our findings of significant interactions show that ADHD has non-uniform effects on the cortical thickness of BP patients. In BA47, the presence of ADHD eliminates the cortical thinning associated with BP relative to controls. On the other hand, BP is only associated with cortical thinning in BA25 when ADHD is present. This again suggests that a comorbid diagnosis is not merely represented by the overlaying of the cortical abnormalities associated with BP and ADHD separately, but rather that individual brain regions are differentially affected by the comorbid diagnosis.
Mood disorders are typically associated with dysfunction in a network such as the corticolimbic circuit, which includes both the subgenual cingulate and the lateral orbitofrontal cortex. It is striking that both regions of interaction are found in the corticolimbic circuit. Future research is necessary to further investigate the nature of these interactions. It's is possible that the interaction of BP and ADHD in BA25 disrupts the corticolimbic circuitry in a way such that cortical thinning in BA47 associated with BP is strengthened. For example, given that both regions are involved in inhibition, perhaps the weakening of the corticolimbic projections from one region strengthens those from another as a compensatory mechanism. We acknowledge that this is speculative and clearly more work needs to be done, with larger populations samples, to investigate this relationship. In addition, the differential interaction of ADHD and BP on brain structure raises questions about the problems associated with strictly phenotypically defined diagnoses. Although outside of the scope of this paper, the non-uniform effect of ADHD on the cortical structure of patients with BP may be representative of the diagnostic classification problems currently facing the scientific community.
Main effects analyses revealed cortical thinning associated with a BP diagnosis, regardless of ADHD comorbidity, in certain PFC regions (BA10, BA11, and left BA9) and the left ACC (BA24 and BA32). Our finding of cortical thinning in these areas is consistent with previous studies in BP (9
). Of particular interest is the robust main effect found in the medial OFC (BA11) since a recent fMRI study by our group examining the effects of a comorbid diagnosis on BOLD signal also revealed a BP main effect (in this case, of hypoactivation) in BA11 (29
). This may serve as evidence that functional and structural abnormalities in BA11 are hallmark features of BP.
To our knowledge, only one other study has examined how brain structure in adults with BP varies in the presence of ADHD. That study found that ADHD and BP contributed additively and selectively to brain structure, resulting in a comorbid phenotype comprised of the abnormalities found separately in each individual disorder (27
). Our maps extend these prior findings by suggesting that, in some regions, the impact of a BP/ADHD comorbidity is not simply additive but may be reflective of an interactive effect, resulting in a distinct neural signature. Two previous fMRI studies evaluating BP patients with ADHD comorbidity similarly have found differences in neural signatures between ADHD/BP patients and patients with BP-only (28
). Adler et al. reported that a BP/ADHD diagnosis in adolescents is associated with hyperactivation of the posterior parietal cortex and middle temporal gyrus, as well as hypoactivation of the PFC and ACC, when compared to BP-only patients (28
). In another study by our group, Townsend et al. (currently under review
) found interaction effects in the anterior and posterior cingulate, left medial and middle frontal gyri, left inferior parietal lobule, precuneous and striatum, suggesting that the neural effects of BP vary in relation to the presence or absence of ADHD (29
). These studies together suggest that ADHD comorbidity must be considered in neuroimaging analyses.
The etiology of cortical thinning in patients with BP remains to be determined. It may reflect fewer neurons, a reduction of glia without a loss of neurons, or an increase in white matter myelination rather than gray matter reduction (57
). Any of these structural abnormalities may affect the function of not only the structurally abnormal brain region (59
), but also those regions to which it projects (60
). Many brain regions implicated in our interaction and main effects analyses are part of an anterior limbic network. Functional deficits in portions of this network are associated with the emotional dysregulation characteristic of BP (32
). For example, the lateral OFC (BA47) and medial OFC (BA11) (where abnormalities were associated with BP in our analyses) have direct and indirect reciprocal projections to regions within the limbic circuit, including the amygdala, anterior temporal cortex (BA38/20), ACC (BA24/32) and subgenual cingulate (BA25) (61
Although not included in our a priori
hypothesis, the left fusiform gyrus (BA37) and the bilateral superolateral cortex (BA20) in the inferior temporal lobe also showed an interaction between BP and ADHD (). Strakowski et al. (64
) recently found blunted fusiform gyrus responses in manic patients with BP during an emotional response task when compared to healthy control subjects. Nomura et al. (67
) reported a negative correlation between activation in the fusiform gyrus and amygdala. To our knowledge, there are no studies directly examining the relationship between the fusiform gyrus and amygdala in BP, although this may be useful to evaluate in future studies. In further analysis of the interaction effect found in BA20, we found that the presence of ADHD was associated with an increase in the magnitude of the BP effect in this region. BA20 projects to the lateral OFC (BA47), suggesting that abnormalities BA20 may contribute to the dysfunction of the anterior limbic network. The effect of BP on cortical thinning in this region may explain why in this study the presence of ADHD was not associated with any BA47 cortical thinning in patients with BP relative to controls.
The current study has several strengths in its design. First, no participants were currently taking lithium. Second, all BP patients were euthymic at the time of scanning. In prior structural studies of BP, only four of these studies controlled for patient use of lithium (10
), and only two of these studies explicitly controlled for mood state and BP subtype (10
). Gray matter volume may increase in as little as four weeks post lithium treatment (69
). Further, significant differences have been found in gray matter volume between patients with BP treated with lithium and those who were not (37
). The impact of mood state on gray matter measures has also been reported, with a depressed mood state being associated with gray matter deficits relative to patients in a euthymic state (40
There are some limitations, however, in the current study. First, while none of the patients in this study were currently taking lithium, many were on other medications, including anticonvulsants, antipsychotics, and stimulants. The effects of these medications on brain structure are not clearly known. One study has reported that antipsychotics do not affect cortical thickness (70
), although others have found a medication effect (71
). Some studies have reported that stimulants have no effect (72
) while others have reported an association with a deficit in gray matter (74
). Second, it has been suggested that reductions in gray matter may occur as a consequence of affective episodes rather than as a result of aging (77
). As the previous number of episodes correlates with illness duration and age, we could not disentangle these effects. However, the number of episodes did not differ significantly between BP patients with and without ADHD and therefore could not by itself account for our findings. In addition, there has been some evidence that gray matter volume scales with brain size. We find it unlikely that brain size significantly contributes to our findings since the aforementioned study reported the exponent factor of cortical thickness scaling to brain size as less than one third, identifying a much stronger relationship between cortical surface area and brain size (78
). In addition, a postmortem study found no correlation between brain size and cortical thickness, again reporting that increases in gray matter volume with brain size are attributable to cortical surface area (79
). However, as an extra precaution, we re-examined regions identified as having either a significant interaction or main effect and found that while controlling for total brain volume could shift the p-value slightly in either direction (depending on the region in question), it did not affect the statistical significance of any findings. Last, a larger subject pool for each of our patient groups would have been ideal; it is possible that there are other regions of interaction or main effects that may not have been detected here due to insufficient power. None-the-less, the findings here underscore the importance of properly distinguishing between BP and ADHD diagnoses, and we hope that they will be expanded upon with larger subgroups in future studies.
In conclusion, this is one of the first studies, to assess interactions of BP and ADHD diagnoses on brain structure. Interactions, which were present in the left subgenual cingulate and right orbitofrontal cortex, suggest that the effect of BP on cortical thickness in these regions varies according to the presence or absence of ADHD. An accurate depiction of the underlying neural phenotype of BP, as opposed to that of a combined BP/ADHD diagnosis, is essential for the understanding of the pathophysiology of BP and developing targeted treatment approaches.