Although significant advances have been made in the treatment of mood disorders, how to individualize treatment is still not understood. Diagnoses in psychiatry are based on symptoms rather than on the causes of the disorder; thus, treatment decisions are based on choosing an intervention that works in patients with a particular group of symptoms and not necessarily on the pathology specific to an individual patient. This is not necessarily a problem except that a substantial percentage of patients with mood disorders do not achieve remission, even with optimal treatment.1,2 In fact, complete symptom remission is rare and recurrences are frequent—finding a successful treatment often takes multiple trials of various drugs.
CT, MRI, diffusion tensor imaging (DTI), functional MRI (fMRI), magnetic resonance spectroscopy, positron emission tomography (PET), single photon emission CT (SPECT), and functional near-infrared spectroscopy (fNIRS) offer a noninvasive way to investigate the neurobiology associated with mood disorders. Unfortunately, because there are no neuropathological criteria that define mood disorders, no criteria exist to compare directly with neuroimaging findings. Thus, the neurobiology to be examined has to be defined in other ways.
One method by which neuroimaging could potentially help in clinical decision making would be to increase our understanding of the brain correlates of treatment. Identifying the neurobiology that can be used to predict a patient’s response to a particular treatment, as well as what changes occur with a successful treatment, would have profound clinical and scientific implications. A number of different strategies have been used to understand the brain characteristics that predict response and that correlate with successful outcome. (For a more detailed coverage of the literature, reviews of research using neuroimaging in mood disorders from several perspectives are available.4-8)
Brain correlates of treatment are often investigated by acquiring neuroimaging data on a group of patients (and sometimes healthy controls), treating the patients, and clinically determining which patients respond. In addition, neuroimaging data are often acquired after a specified treatment period. Responders and nonresponders are defined on the basis of clinical criteria. To determine neural correlates of treatment response, the pretreatment data are compared with the posttreatment data to determine changes in the 2 groups. These brain differences are then considered predictors of treatment response. This method has been used with various imaging modalities to investigate predictors of response and correlates of treatment response in patients whose condition is being treated with a variety of interventions (Table).
Multiple reports of neurobiological differences in responders versus nonresponders, as well as brain changes with successful treatment, have been reported. Various researchers, however, have focused on different regions of the brain. In their study on depression, Mayberg and colleagues9 concentrated on the anterior cingulate, especially the subgenu portion (Brodmann area 25).
Saxena and colleagues10 differentiated predictors of treatment response for major depressive disorder (MDD) and obsessive-compulsive disorder and found the amygdala and midline prefrontal regions to be important markers.
Siegel and colleagues11 used fMRI to predict outcome from cognitive-behavioral therapy and found that the subgenu of the anterior cingulate and amygdala were important structures.
Although brain differences associated with clinical response are regularly reported, the specific brain regions implicated are not consistent. Thus, a clear understanding of the neural correlates of predictors of response or the correlates of response has not emerged. The consistency of finding a significant difference in the brain of those who do and those who do not respond may be related to publication bias, because significant findings are more likely to be published than nonsignificant findings. The large number of studies that have found a difference, however, suggests that there are clear brain differences that predict treatment response and brain changes that are correlated to successful treatment.
Variable methodologies produce inconsistent results
Brain findings associated with predictors of treatment and with change with successful treatment may be related to the treatment used (eg, medications, psychotherapy, brain stimulation). In addition to treatment variability, the imaging techniques for measuring predictors of treatment and correlates of response are also quite varied. There is no universally agreed on method for acquiring and analyzing imaging data. Even within a particular imaging modality, the method of analysis can be quite different. The treatment modality used and the way the brain structure and/or function is measured could account for some of the variability in results.
■ Currently, the only clinical role for neuroimaging in the evaluation of a patient with a mood disorder is to determine whether a neurological condition (eg, tumor, stroke, normal pressure hydrocephalus) may be contributing to the clinical picture. There is no scientifically validated role for neuroimaging to assist in choosing a particular treatment or to make a particular diagnosis in a patient who has a mood disorder.
■ If a neuroimaging test can differentiate whether a patient with a depressive episode would respond to treatment A versus treatment B, it could have a significant effect on the patient’s evaluation, treatment, and outcome.
■ Neuroimaging holds great promise in the evaluation of neurobiological predictors of treatment response and correlates of treatment outcome and thus in helping us to understand mood disorders in a clinically meaningful way.
Another problem is the fundamental assumption that each mood disorder—as diagnosed in DSM-IV using a syndrome-based approach—represents similar neuropathology. This is especially pertinent in neuroimaging research because the vast majority of neuroimaging studies obtain results from a “group analysis.” A group analysis involves averaging the data from multiple participants to have enough power to distinguish a difference. This is a consequence of the generally low signal to noise ratio of neuroimaging data. The essential assumption of a group analysis is that all the members of a group share the same neurobiology. There is growing evidence, however, that the neuropathology of mood disorders assessed with neuroimaging is not consistent.12 With the small sample sizes in neuroimaging studies, the group result can be especially sensitive to a small number of participants within that group (ie, a few participants can drive the result for the whole group). Thus, differences in results across studies may be related to differences in samples studied by the investigators.
Measuring the connectivity of brain circuits
Although there is a lack of consistency for specific brain regions, a majority of the findings associated with mood disorders involve the prefrontal and limbic regions. Initially, the focus was on the level of activation of a brain region either at rest or during a particular task (eg, is the dorsolateral prefrontal cortex more or less active?). There is, however, a growing interest in how different brain regions work together. Because the brain is made up of circuits, connectivity between different regions can be measured in various ways.
Connectivity can be directly measured intraoperatively during neurosurgery. Such measurements, however, are only available for a very small, restricted population with mood disorders (eg, patients undergoing neurosurgery for epilepsy, deep brain stimulation, or epidural brain stimulation). To study a broader sample of patients with a mood disorder in an ethical manner, several noninvasive methods of measuring connectivity in brain circuits are emerging.
One involves focal stimulation of the brain, measuring changes in multiple regions. These include interleaved repetitive transcranial magnetic stimulation (rTMS)/fMRI,13 rTMS/PET,14 rTMS/SPECT,15 and simultaneous rTMS/fNIRS.16 Each method has its own strengths and weaknesses, but all offer a unique opportunity to test the circuitry of the brain as it relates to predictors of treatment and correlates of outcome. These technologies are, however, still at the research stage of development.