The research agenda for DSM-V emphasizes the need for translating basic and clinical neuroscience research findings into a new classification system for psychiatric disorders based upon pathophysiologic and etiological processes (1, 2). It also supports the recent call for such research to help create “rational treatment advances” in disorders such as bipolar depression (3) and the National Institute of Mental Health recommendation for the need to translate basic science discoveries into biomarkers, diagnostic tests, and new treatments for individuals with psychiatric disorders. Biological evidence of pathophysiologic processes can help meet critical challenges in psychiatric research by aiding the construction of diagnostic and treatment response groupings. This is particularly relevant to bipolar disorder, which is frequently either misdiagnosed or diagnosed late, often as unipolar depression in individuals without a clear previous history of manic episodes (4).

Studies employing neuroimaging techniques, and functional neuroimaging techniques in particular, provide direct measures of neural system abnormalities that may be associated with different domains of pathology in bipolar disorder (2). These domains include abnormal emotion regulation and impaired cognitive control. A major issue for studies of psychiatric populations, however, is the potential confound of psychotropic medication upon neuroimaging measures. It remains unclear how different psychotropic medications affect neuroimaging measures, such as the proxy measures of blood flow used in functional magnetic resonance imaging (fMRI). This is especially problematic for studies of bipolar populations, in which the majority of individuals, particularly individuals with bipolar disorder type 1, may be receiving psychotropic medication.

Restricting study to medication-naive individuals with bipolar disorder is one strategy to avoid any potential confound of medication upon neuroimaging measures. This strategy would likely limit recruitment to small numbers of participants who may not be representative of the chronically ill bipolar populations managed in most clinical settings. The inclusion of only unmedicated individuals in neuroimaging studies in bipolar disorder necessarily involves withdrawing individuals from psychotropic medication, since maintenance treatment is usually required to decrease the probability of a relapse because of the recurrent nature of the disorder. For the individuals who are able to tolerate medication withdrawal and washout, there are clear benefits. These include not only the potential of identifying biomarkers of bipolar illness but also the opportunity provided by medication washout to identify biomarkers that may predict individual response to subsequent treatments without confounds of current psychotropic medication. There are other justifications for implementing medication withdrawal and washout. The process may decrease copharmacy, whose rational use is hampered by a lack of systematic research on its safety and efficacy (5). Furthermore, there are instances in which psychiatric symptoms or physical side effects improve upon discontinuation of ongoing medications.

There are two difficulties inherent in medication withdrawal. First, withdrawing individuals from medication may increase the probability of relapse for individuals in remission and a worsening of clinical status for individuals experiencing an acute episode (6). This may therefore make medication withdrawal clinically unfeasible for some individuals with bipolar disorder. A related problem is that the medication withdrawal period may need to be long to include both withdrawal itself and a subsequent medication-free washout period to ensure the absence of potential withdrawal phenomena that may confound neuroimaging measures. This may further increase the risk of relapse or a worsening of symptoms in individuals with bipolar disorder. It takes five half-lives to eliminate 97% and six half-lives to eliminate 98% of residual drug (7). For medications with a long half-life, this translates into a washout period of several months. The clinical setting in which the washout and withdrawal are performed clearly plays an important role in the evaluation of the risk-benefit ratio of the procedure, and careful examination of the history of each individual is necessary to determine the level of risk and potential complications that the procedure would entail.

A second potential difficulty with medication withdrawal and washout is that it may carry the risk of selection of a subpopulation that is not representative of the bipolar population as a whole. Individuals with a history of severe manic or psychotic episodes, individuals with comorbid anxiety disorders, and those with a high suicide risk or a high risk of dangerous behaviors may be underrepresented in subpopulations that are able to tolerate the washout procedure and remain medication free. Conversely, individuals with less severe symptoms are likely to be overrepresented in these subpopulations. For the development of new treatments for bipolar disorder, the most at-need group includes individuals with correctly diagnosed bipolar disorder yet who are poorly responsive or nonresponsive to existing treatments. These individuals with bipolar disorder are likely to be medicated and, as discussed above, may be unable to tolerate medication withdrawal. Biomarkers of bipolar illness and treatment response identified from neuroimaging studies of unmedicated individuals only may not, therefore, be generalizable to medicated groups. This limitation may reduce the usefulness of any putative investigations.

Some neuroimaging studies have been able to recruit unmedicated individuals with bipolar disorder type II (8), small numbers of unmedicated individuals with bipolar disorder type 1 (9), or unmedicated individuals in a first episode (10) or with pediatric bipolar disorder (11). Recruitment of large numbers of unmedicated individuals with bipolar disorder type 1 remains difficult, however. Indeed, the actual proportion of individuals with bipolar disorder type 1 excluded from neuroimaging studies because of an inability to tolerate medication withdrawal and washout has not been previously reported, probably because individuals with the disorder at higher risk of experiencing problems with withdrawal are usually not referred for study.

We performed an Ovid MEDLINE literature search on studies in bipolar disorder from 1996 to 2007 using the key words “bipolar disorder” and “neuroimaging.” The majority of these studies examined adults with bipolar disorder type 1. Many were studies that employed functional and structural neuroimaging to measure abnormalities in neural regions associated with important domains of pathology in bipolar disorder: impaired cognitive control, abnormal emotion processing, and mood instability. These neural regions include the prefrontal cortical and subcortical limbic regions (12). Inclusion of all existing neuroimaging studies of bipolar disorder was clearly beyond the scope of this review. Furthermore, not all existing neuroimaging studies in bipolar disorder examined the potential effects of psychotropic medication. Since our main focus was to examine potential medication effects on direct measures of activity in neural regions implicated in the pathophysiology of bipolar disorder, we included in this review all existing functional neuroimaging studies of adult and pediatric bipolar disorder that employed cognitive control and emotion-processing paradigms and examined potential effects of psychotropic medication on neuroimaging measures (see Table 1 for specific details of these studies). To complement these studies, we also include a brief overview of volumetric structural neuroimaging studies that examined the potential effects of psychotropic medication on regional cerebral volumes in adult and pediatric bipolar disorder. We additionally note the importance of spectography and ligand studies in bipolar disorder to the examination of potential confounding effects of psychotropic medication on neurochemical and neuroreceptor systems of interest in bipolar disorder, although a full description of all existing spectrography and ligand studies was beyond the scope of this review.

Functional Neuroimaging Studies

Functional neuroimaging studies in bipolar adults and adolescents demonstrate abnormally decreased activity in prefrontal cortical regions during cognitive control and abnormally increased activity in subcortical limbic regions during emotion processing that may persist during remission, depression, and mania (14, 15, 27). More recent studies have examined neural activity in remitted bipolar populations, many of whom were medicated (13, 21–23). Reports of the effects of medication include 1) no significant association between psychotropic medication and magnitude of activity in anterior cingulate gyral and prefrontal cortical regions during cognitive control tasks, e.g., selective attention (16, 24), 2) increased activity in the dorsolateral prefrontal cortex during selective attention (17), and 3) increases and decreases in anterior cingulate gyral and dorsolateral prefrontal cortical activity, respectively, during executive control (23) in medicated relative to unmedicated remitted individuals with bipolar disorder. A significant positive correlation has also been reported between antipsychotic medication dose (in chlorpromazine equivalents) and activity in the rostral anterior cingulate gyrus and the dorsolateral prefrontal cortex during a cognitive control task in remitted individuals (22). Medicated more than unmedicated bipolar individuals demonstrate levels of prefrontal cortical activity during cognitive control that are similar to those observed in healthy individuals. Similarly, other studies of remitted individuals with bipolar disorder report ameliorative effects of different types of psychotropic medication upon the magnitude of subcortical limbic activity to emotional stimuli such that abnormally elevated subcortical limbic activity is decreased in bipolar individuals (19, 21). A specific effect of antipsychotics in reducing amygdala activity has been demonstrated in bipolar men (26). As was found for cognitive tasks, medicated more than unmedicated individuals with bipolar disorder demonstrate levels of subcortical limbic activity to emotional stimuli that are similar to those observed in healthy individuals. Inconsistent with the above findings is a report of decreased prefrontal cortical activity in bipolar individuals during executive control (word generation) after taking lithium (18).

The few studies that have examined the effects of psychotropic medication on neural activity in bipolar individuals during depression or mania report a positive correlation between levels of chlorpromazine equivalents and decision-making task-related activity in the rostral prefrontal cortex during mania (25) but no effect of such medication on the magnitude of ventral prefrontal cortical activity in mania during attentional control (20). In pediatric bipolar populations, studies indicate no significant effect of the number of psychotropic medications taken upon the magnitude of neural activity to emotional stimuli (15). One study demonstrated an ameliorative effect of psychotropic medication in reducing abnormally elevated striatal and ventral prefrontal cortical activity on a motor inhibition task (14). To our knowledge, no studies have examined the potential effects of psychotropic medication on neural regions implicated in cognitive control in pediatric bipolar disorder.

Volumetric Structural Neuroimaging Studies

Volumetric structural neuroimaging studies in adults and adolescents with bipolar disorder have shown gray matter volume abnormalities in different neural regions subserving cognitive control and emotional processing. The main findings include decreased ventral/orbital medial prefrontal cortical gray matter volume and amygdala and striatal gray matter volume increases in adults with bipolar disorder (see reference 28 for a review of this literature), although amygdala gray matter volume decreases in bipolar youth, adolescents, and young adults (29, 30). Studies of individuals in first-episode psychosis, including both schizophreniform and nonschizophreniform psychoses, have not revealed any significant effect of antipsychotic medication upon regional gray matter volume (31). In non-first-episode individuals with schizophrenia, however, there are associations between antipsychotic medication and enlargement of basal ganglia volumes (32), and in non-first-episode individuals with bipolar disorder, there are associations between concurrent use of psychotropic medication and both decreased (33) and increased (34) ventral prefrontal cortical gray matter volume. Lithium has been associated with significantly increased total and anterior cingulate gyral (35) and hippocampal (36) gray matter volume, although see reference 37. Valproate and valproate-plus-quetiapine combination have been associated with significantly increased anterior and posterior cingulate gyral volumes (38). In pediatric and adolescent bipolar disorder, findings suggest no effect of psychotropic medication exposure upon regional cerebral volumes (29, 39, 40). Other findings indicate an association between past lithium or valproate exposure and greater amygdalar gray matter volume in children and adolescents with bipolar disorder type 1 (41), although lifetime antidepressant exposure has been associated with reduced amygdala gray matter volume in these populations (42). The most consistent finding from volumetric structural neuroimaging studies is an association between lithium and increased gray matter volume in key neural regions, such as the anterior cingulate gyrus, amygdala, and hippocampus, implicated in cognitive control and emotion processing, indicating a neurotrophic—and potentially neuroprotective—effect of lithium (43).

Magnetic Resonance Spectography (MRS) and Ligand Studies

Some MRS studies have examined the extent to which psychotropic medication is associated with changes in abnormal concentrations of metabolites that are measures of neuronal integrity, including N-acetyl-aspartate and myoinositol, in previously unmedicated individuals with bipolar disorder after treatment with different psychotropic medications (44). Findings from these studies include either no changes in such measures after treatment with lithium (45) or increases in N-acetyl-aspartate concentration over several neural regions after 4 weeks of lithium treatment (46). The few receptor ligand studies in bipolar disorder that have examined the effects of psychotropic medication upon neuroreceptor availability have focused on the serotonin transporter, the 5-HT_1A, and dopamine D₂ receptors. Findings indicate no change in cortical 5-HT_2A receptor availability observed in seven manic individuals successfully treated with sodium valproate, one of whom had also received lithium (47). Similarly, sodium valproate monotherapy over 2–6 weeks for nonpsychotic mania had no significant effect on the availability of striatal D₂ receptors (48). These preliminary findings need to be expanded with further research but suggest either ameliorative or no significant effects of mood-stabilizing medications upon neurochemical systems and neuroreceptor availability so far studied in bipolar disorder.

There is a convergence of findings from the functional neuroimaging studies described above, predominantly in adult bipolar populations, that indicates that abnormal frontal cortical and subcortical limbic activity during cognitive control and emotion-processing paradigms may reflect pathophysiologic processes of bipolar disorder (13, 16–26). While the strength of these studies is the consistency of these main findings, there are inconsistencies regarding the potential effects of psychotropic medication and associated methodological limitations that make it difficult to draw firm conclusions. Some studies find no significant effects of psychotropic medication on functional neuroimaging measures (14–16, 20, 24). In many of these studies, however, the small sample sizes inherent in examining individuals taking versus those not taking each individual class of psychotropic medication make it possible that findings of no significant effect of psychotropic medication resulted from type II error, as authors of these studies have emphasized in discussions of their findings (e.g., reference 24). Other studies found significant ameliorative effects of psychotropic medications upon functional neuroimaging measures (19, 21, 22, 26).

Specifically, both antipsychotics and lithium increased prefrontal cortical activity during cognitive control and reduced abnormally elevated subcortical limbic activity during emotion processing in bipolar adults so that activity in these regions resembled that observed in healthy individuals participating in the studies. These findings suggest that abnormalities reflect pathophysiologic processes that may be partly ameliorated by, rather than abnormalities that are secondary to, psychotropic medication. The majority of studies to date, however, either did not include examination of unmedicated individuals or included only small numbers of unmedicated individuals (13, 16, 20–23, 25, 26), making definitive conclusions about the effects of psychotropic medication upon neuroimaging variables in bipolar disorder difficult. There is supporting evidence of either no significant effects or ameliorative effects of psychotropic medication upon regional cerebral volumes from the larger number of volumetric structural neuroimaging studies in adult and pediatric bipolar disorder, but these studies do not allow measurement of the functional integrity of neural regions that may be implicated in the pathophysiology of bipolar disorder. There are also similar findings of no significant effects or ameliorative effects of psychotropic medication upon measures of neural integrity and neuroreceptor availability from the more preliminary MRS and ligand studies described above, but these latter studies require replication.

Although there are no definitive methods for examining the potential effects of medication in studies of psychiatric populations, such as individuals with bipolar disorder who often receive polypharmacy, there are several potential options. The first option is to limit study to individuals taking certain numbers and categories of medications only, but this may not always be feasible in real world contexts, where individuals with bipolar disorder are typically treated with a number of different psychotropic medications and medication combinations. The second option, therefore, is to employ specific strategies for resolving the issue of potential effects of psychotropic medications in individuals taking a variety of different medications. One strategy is to compare individuals taking versus those not taking each class of psychotropic medication, including mood stabilizers, antipsychotics, antidepressants, and anxiolytics (13, 21), but this strategy may have the limitation of examination of multiple small subgroups. Other studies have converted doses of each medication within a specific class into dose equivalents of prototypical medications for each class, e.g., converting all antipsychotics into chlorpromazine dose equivalents, followed by correlational analyses to examine associations between medication dose and relevant dependent variables in larger subgroups of individuals taking each class of psychotropic medication (22, 25, 26).

Since different classes of psychotropic medications may exert effects by different mechanisms, testing the effects of each medication class separately is important. A problem with this strategy is that it does not account for interactions between different psychotropic medications. Composite measures of total psychotropic medication load, reflecting the dose, variety, and duration of different medications taken, may therefore be preferable to the separate examination of individual classes of psychotropic medication. The computation of a single index of psychotropic medication load for each medicated individual will also allow the inclusion of larger numbers of medicated individuals taking different types of psychotropic medication and may be the most feasible option for the examination of psychotropic medication effects in large neuroimaging studies of individuals with the disorder, particularly for studies of individuals with bipolar disorder type 1. Ideally, comparisons of medicated with unmedicated individuals should be included in cross-sectional functional neuroimaging studies of bipolar disorder.

Future studies can also include longitudinal designs in which the potential effects of individual psychotropic medications upon neuroimaging variables are examined over time. These studies clearly have the advantage of powerful, within-subject designs and can recruit individuals with bipolar disorder (either in remission or in an acute illness episode) into carefully controlled clinical treatment platforms. This type of design is especially important for MRS and ligand studies measuring the potential confounding effects of psychotropic medication on neurochemical and neuroreceptor systems of interest in bipolar disorder. Longitudinal treatment designs examining the within-group effects of specific psychotropic medications over time have, for example, been previously employed in major depressive disorder (49) but rarely in bipolar disorder (18).

The interpretation of findings regarding potential effects of psychotropic medication in individuals with bipolar disorder will necessarily be informed by the extensive preclinical literature that has examined the effects of psychotropic medication upon neural regions and behaviors relevant to domains of pathology in bipolar disorder. The emerging field of pharmacological fMRI that allows examination of the effects of different psychotropic medications upon task-related neural activity in healthy individuals will also inform interpretation of likely effects of psychotropic medication in bipolar groups studied in neuroimaging studies. Here it has been demonstrated, for example, that different categories of psychotropic medication can reduce amygdala activity to emotional stimuli (50). The development and employment of strategies to examine potential confounding effects of psychotropic medication in functional neuroimaging studies of bipolar disorder should continue be a major focus of research in this disorder.