Neuroimaging studies in adult BD have documented a range of morphometric abnormalities in brain regions known to be involved in affect regulation such as prefrontal and anterior cingulate cortices, striatum, thalamus, and amygdala.⁵ However, a recent meta-analysis of regional morphometry in adult bipolar disorder suggests significant heterogeneity across studies for several of these brain structures, including the amygdala, left subgenual prefrontal cortex, and thalamus.⁶ For example, both increases and decreases in gray matter (GM) volume have been reported in the amygdala, striatum, anterior cingulate gyrus, and regions of the prefrontal cortex. Some studies have indicated amygdalar volume increases in adult individuals with BD type I (BD-I)⁷; other studies have found decreased volume and GM density in anterior cingulate and subgenual gyri and ventral prefrontal cortex,^6,8,9 whereas others found no significant differences in prefrontal cortical volumes between patients with BD-I versus healthy controls.¹⁰ These inconsistencies may be due to variations in methodology, age at onset, presence of comorbid disorders, exposure to stress, or medication effects.

Few neuroimaging studies have examined volumetric changes in cortical and subcortical structures in pediatric BD.¹¹ Most of these studies found a decreased GM volume. For instance, some studies reported decreased amygdalar¹² and hippocampal volumes.¹³ Others have found decreased GM volume in orbital medial prefrontal cortex (OMPFC),^14,15 left superior temporal gyrus,¹⁶ and dorsolateral prefrontal cortex¹⁷ in children and adolescents with BD. Some, however, have reported increased GM volume bilaterally in the basal ganglia, thalamus, and left temporal cortices.¹⁵ Findings from these studies suggest that abnormalities in distinct neural regions, most consistently GM volume reductions, within the amygdala, hippocampus, and OMPFC, may be implicated in the pathophysiology of pediatric BD.

In a recent review of MRI studies conducted in individuals genetically at-risk for BD (i.e., affected and unaffected family members, first-episode patients, and familial bipolar patients), Hajek et al.¹⁸ showed that abnormalities in GM volume represent potential candidate neuroanatomical risk factors for BD. These factors may involve GM volume abnormalities in the subgenual prefrontal cortex, striatum, amygdala, and hippocampus. With regard to research conducted in unaffected relatives of subjects with BD, one study found that genetic liability for BD was associated with GM deficits in the right anterior cingulate gyrus and ventral striatum,¹⁹ whereas another study did not find any differences in GM volume between unaffected relatives of BD patients and healthy controls.²⁰ These contradictory findings may be attributed to the uncertainty of whether unaffected relatives of bipolar patients in such studies were free of other psychopathology (e.g., depression) at the time of the MRI scan. Moreover, it is difficult to ascertain to what extent differences in GM volume indeed represent neuroanatomical risk factors because neuroimaging studies conducted in individuals at genetic risk (particularly offspring of BD parents) often included those diagnosed with BD.²¹ Furthermore, studies that included healthy family members (e.g., discordant twin pairs, first-degree relatives) have been inconsistent in their exclusion criteria (e.g., some participants either met criteria for another Axis I diagnosis or had a history of major depressive disorder).²² To delineate neuroanatomical risk factors of BD, it is important to control not only for the presence of BD but also for other psychopathology, particularly mood disorders. One way to achieve this goal is to focus examination on offspring of parents with BD yet free of any psychopathology.

Examination of neuroanatomical abnormalities in such healthy bipolar offspring offer several advantages, including the possibility of identifying neuroanatomical abnormalities, potentially predating the onset of BD, that are not confounded by the presence of psychopathology or medication; identifying neuroanatomical abnormalities that may confer risk for, or are protective against, BD to ultimately inform evaluation of risk for subsequent development of BD and subsequent therapeutic intervention; and increasing understanding of the developmental course of BD.

We used voxel-based morphometry (VBM) to examine whole-brain GM volumetric MRI changes in a group of young healthy offspring of parents diagnosed with BD. VBM provides an objective means of examining the brain voxel by voxel in an automated fashion, determining differences in tissue volume and avoiding the potential for operator bias that may be present in the more traditional MRI, region of interest—based analytical methods.²³ VBM is particularly useful for the examination of neuroanatomical abnormalities in at-risk populations that may have structural abnormalities in distributed neural systems. Based on the most consistent findings from previous neuroimaging studies conducted in pediatric BD and unaffected relatives of adults with BD, we hypothesized that healthy bipolar offspring would have decreased GM volume in subcortical and prefrontal cortical neural regions involved in affect regulation, including the amygdala, hippocampus, and OMPFC. Because previous findings suggest that some neural regions change in GM volume as a function of age,²⁴ whereas others change with puberty,²⁵ we also wished, in exploratory analyses, to examine the relationship between age and pubertal development in neural regions in which GM volume abnormalities were observed in healthy offspring of parents with BD.

Our findings of increased parahippocampal/hippocampal GM volume in HBO are in sharp contrast to findings from MRI studies in adult and pediatric BD. Although some studies have found hippocampal differences between adult BD patients relative to healthy controls,³⁴ most studies typically have not found any significant differences in parahippocampal or hippocampal GM volume.³⁵ Decreases in hippocampal GM volume have been reported in adolescent BD,^12,13 suggesting the involvement of the hippocampus in the pathophysiology of adolescent BD that may represent a particular characteristic of early-onset BD. Our findings are congruent, however, with previous data from studies in adults at risk for BD; a study that compared adult monozygotic twins discordant for BD found a smaller right hippocampus in the twin with BD relative to the twin without BD.³⁶

The finding of increased GM volume in the parahippocampus/hippocampus in HBO is particularly interesting because of the potential role of these regions in the regulation of stress and emotional responses. The hippocampus has been implicated in inhibition of the stress response via inhibitory connections with many of the subcortical structures involved in this response.³⁷ As part of the limbic system, the parahippocampal gyrus has multiple direct connections with the hippocampus and amygdala,³⁸ and it has been suggested that a dynamic relationship between the amygdala and parahippocampal gyrus may confer a protective effect against potentially harmful experiences.³⁹ Previous findings indeed indicate that individuals with parahippocampal gyral, but not amygdalar, resections show abnormal appraisal of dissonant musical cues in that, unlike healthy controls, they label these cues as pleasant rather than highly unpleasant.⁴⁰ A recent study reported significant decreases in parahippocampal gyral activity to emotional words in adults with BD relative to age-matched healthy controls.⁴¹ These findings therefore support a protective role for the parahippocampal gyrus in the normal appraisal of emotional information, which may become dysfunctional in adult BD.

Although it is possible that the pattern of increased GM volume in the left parahippocampus/hippocampus in HBO represents a potential neuroanatomical risk marker for BD, this interpretation is problematic given that decreased rather than increased GM volume in the hippocampus has been reported in youths with BD.^12,13 Moreover, it is unlikely that sudden morphological changes occur specifically with BD onset. Elucidating the role of this region in the subsequent development of BD in genetically at-risk pediatric populations can be addressed only by longitudinal studies examining subcortical development before and after the onset of the disorder.

Another interpretation of the observed GM increase in HBO is that it may have a potential role in protecting against or delaying subsequent development of BD, given that HBO in our study were completely free of any Axis I disorder. Pediatric or adult BD onset is often preceded and/or accompanied by other psychiatric disorders such as disruptive behavior disorders or anxiety disorders.^1,42 It is therefore possible that the HBO in our sample represent a potentially emotionally resilient group despite being at risk for BD. The pattern of increased rather than decreased GM volume observed in our HBO group may be interpreted as being associated with successful affect regulation that acts as a compensatory mechanism against the development of mood or anxiety disorders or because of being exposed repeatedly to situations (e.g., family conflict) that require these bipolar offspring to regulate their affect. The latter interpretation would be consistent with recent evidence demonstrating that cognitive activity itself can alter brain morphometry in certain relevant brain structures.⁴³ Thus, examining behavior or neural systems associated with affect regulation in these bipolar offspring would allow us to begin addressing this question.

The significant positive correlation between left parahippocampal/hippocampal GM volume and pubertal maturation in HBO, but not CONT, may be further support of the protective role of this GM volume increase in HBO. Although there is no evidence directly linking BD with puberty onset, large-scale studies examining BD age at onset (e.g., STEP-BD) report a high frequency of first episodes occurring between 13 and 18 years (i.e., during puberty).⁴⁴ If structural neural abnormalities reported in youths with BD are potential mediators for the development of BD, then it is plausible that the increased parahippocampal/hippocampal GM volume associated with pubertal maturation in HBO could play a neuroprotective role by providing additional neural resources in affect regulation. Further-more, this increase in parahippocampal/hippocampal GM volume could not be attributed to an effect of age because there was no significant correlation between parahippocampal/hippocampal GM volume and age in either HBO or CONT.

The interpretation that this parahippocampal/hippocampal GM volume increase in HBO plays a neuroprotective role remains speculative until longitudinal follow-up studies are conducted to determine who will go on to develop BD. Moreover, future studies examining regional neural activity during emotional information processing or affect regulation tasks in HBO without Axis I disorder will allow us to examine the extent to which abnormal patterns of parahippocampal/hippocampal activity during these tasks may accompany the increase in GM volume in these regions in HBO.

Our findings of increased GM volume in the parahippocampus/hippocampus are in contrast with some MRI studies conducted in individuals who are at high risk for BD.¹⁸ However, it is important to note that most of these studies included healthy control participants who, although not meeting criteria for BD, either had a history of mood disorder or met criteria for other psychiatric disorders.^21,22 Our findings in HBO allow us to address questions not only regarding the nature of neuroanatomical abnormalities that may confer risk for BD but also how such abnormalities may protect against the development of BD. They also provide a neuroanatomical focus for future study of the neurodevelopmental trajectory of BD in youths at genetic risk for BD.

The left-sided localization of the parahippocampal/hippocampal GM volume increase in HBO is additionally interesting. The left more than the right hemisphere has been linked with positive emotion processing.⁴⁵ Previous findings from functional neuroimaging studies in adults with BD have also indicated abnormally increased left amygdalar activity to positive emotional stimuli.⁴⁶ However, our findings are inconsistent with the reported decreased GM volume in bilateral hippocampus in adolescent BD.¹³ Nevertheless, these findings suggest that it is possible that increased leftsided parahippocampal/hippocampal GM volume in HBO may, in particular, protect against abnormal positive emotion processing that may be associated with the development of BD.

One important point that merits discussion is with regard to the potential limitation of the VBM technique. First, given that the computation of VBM statistical parametric map requires several thousands of independent statistical comparisons, the likelihood of type I error is increased. As such, it is important to include appropriate corrections to protect against type I error to ensure that results reflect statistically significant differences. The recommended approach to control for type I error in VBM is to include voxel-wise corrections such as the FWE correction, which we used in this study.²³ However, these corrections are rather conservative, thereby potentially increasing type II error. The latter point is particularly relevant given our sample size. One way to circumvent this potential limitation could have been to conduct analyses on hand-traced data. However, such an approach encompasses its own set of limitations.⁴⁷ For instance, hand tracing is subjective to tracer position of anatomical landmarks and is restricted to predefined regions of interest, at the expense of detecting differences observed in whole-brain analyses. In this regard, hand tracing encompasses its own source of error, particularly given the level of accuracy, reliability, and training of the research staff. A methodological strength of VBM in this regard is that VBM enables researchers to conduct whole-brain analyses and is sensitive to systematic differences in GM volume.⁴⁷ Furthermore, by using automated anatomical templates, it is possible to replicate findings in a different sample or to investigate the same regions in different populations or research centers.

In addition, the findings from the present study are limited by certain factors, including the relatively small sample size, although such sample sizes do provide adequate power to detect medium to large effect sizes.⁴⁸ The stringent criteria we used in recruiting our sample, namely, the absence of any Axis I disorder, necessarily limited the number of participants in the present study and the ability to match on sex or exclude the few offspring with a parent with BD-II. Another limitation is related to the cross-sectional design, which prevented examination of neurodevelopmental trajectories underlying BD. The next stage, therefore, is to follow HBO participants in the present study to allow examination of the extent to which increased GM volume in the parahippocampus/hippocampus may be associated with the subsequent development of BD or another mood disorder.