Brain Imaging in Mood Disorders
Klaus P. Ebmeier and Dina Kronhaus
Depression, as the reversible psychiatric condition par excellence, is clearly an ideal object for functional neuroimaging studies. In theory, patients return to their initial (healthy) brain state so that any image changes observed during an affective episode should mark the brain structures and circuits involved in the expression of symptoms and signs. Authors have imaged patients when ill and after recovery and used a number of strategies to exploit short-term fluctuations of symptoms. Such fluctuations occur naturally, as in the typical diurnal variations of mood (Moffoot et al, 1994b), or they can be provoked by interventions, such as mood induction (Baker et al, 1997), sleep deprivation (Ebert et al., 1994a) or tryptophan depletion (Smith et al., 1999b). If, on the other hand, anatomical changes did exist in depression, they would be predicted in cases of treatment resistance (Shah et al., 1998), in secondary or late-onset depression (Ebmeier et al., 1998), or possibly a priori in certain patients with a genetic predisposition for the illness (Drevets et al., 1998).
Anatomical systems involved are likely to be medial limbic, with the anterior cingulate cortex and orbitofrontal cortex playing a prominent role (Ebert and Ebmeier, 1996). There is also the well-rehearsed hypothesis of hypercortisolaemia, which occurs frequently in depression and, at least in animal models, leads to hippocampal damage. Hippocampal damage, in turn, would release the pituitary secretion of adrenocorticotropic hormone (ACTH) from hippocampal suppression and result in a positive feedback loop (Sapolsky et al., 1986). This mechanism may not be specific to depression (Welberg et al., 2001), as some authors have also used it to explain cognitive impairment or dementia (Hibberd et al., 2000) and the sequelae of severe psychological trauma (Bremner et al., 1995).
Neuropsychological tasks have been employed in imaging studies to activate brain systems thought to be implicated in depression, in particular using 'frontal' (e.g. word-generation) or 'temporal' (memory) tasks. In such experiments, limited task performance may be responsible for group differences. Attempts to control for such performance differences include pacing tasks at a speed that all patients can manage, and post-hoc correlation of brain activity with task performance, e.g. by using analysis of covariance. A further complication of functional imaging protocols is that it is now very difficult to recruit untreated patients in a psychiatric setting. Primary-care physicians have usually already treated their patients with a standard antidepressant (e.g. a selective serotonin reuptake inhibitor, SSRI) by the time of referral. The cost and effort required to recruit patients at the primary-care level is usually seen as prohibitive. For this reason, many studies contain samples of medicated patients and have to be interpreted with caution. It also cannot be excluded that changes in brain activity or even brain anatomy may be caused by medication (DelBello et al., 1999).
Medication is, of course, a particular problem for receptor ligand studies. Based on effective pharmacological treatment, there are a variety of hypotheses, particularly involving the serotonergic and noradrenergic transmitter systems, which are theoretically amenable to in vivo testing with neuroimaging (Delgado et al., 1990). Not only the availability of untreated patients but also the availability of receptor ligands has limited such research. The latter may be partially responsible for the dearth of noradrenaline ligand studies. Not all ligands are suitable; their use may be limited by their specificity for the receptor concerned, their affinity (i.e. the likelihood to be displaced by endogenous ligand) and their nondisplaceable (non-specific) binding fraction. In spite of these limitations, first results are now emerging that test some of the extant pharmacological hypotheses in depression.
Rather than giving a balanced review of all studies carried out in the field, we will focus on certain themes and future prospects that appear to be emerging. Our selection will no doubt be idiosyncratic, but we hope that we have captured the important paradigms and paths of current research. In order to limit the size of the chapter, we will focus on key publications of the last 5 years (at the time of writing), as earlier literature has been summarized well in a number of other reviews (Davidson et al., 1999; Drevets, 1998; Kennedy et al., 1997; Norris et al., 1997; Stoll et al., 2000; Videbech, 1997; Videbech, 2000). Rather than systematically dividing the imaging literature by image modality or diagnosis, we will attempt to present a logical narrative, proceeding from simple (e.g. neurochemical) hypotheses, such as the dopamine theory of psychomotor retardation in depression, to more complex models. Hypotheses that are, in a sense, post-hoc, i.e. exploit the natural history of depressive symptoms and their treatment, will be followed by experimental approaches, which imply complex neuronal systems and attempt to activate selectively such systems that are thought to be implicated in the expression of depressive symptoms. Mania is a rare condition that is very difficult to study with neuroimaging techniques, and reports are rare (Al-Moussawi et al., 1996). This illness will, therefore, not be discussed, except when included in studies of bipolar depressed patients.
Although dopamine is not thought to be involved primarily in the treatment and the experience of symptoms of depression, it may play an important role in the brain reward systems and in movement control. It has been implicated in retarded depression, both by the reduction of the dopamine metabolite homovanillic acid in cerebrospinal fluid (Jimerson, 1987) and by increased D2 receptor binding, particularly in psychomotor retarded patients (Ebmeier and