Individual variability and “vulnerability”
The complexity of stress–diathesis models of depression and of the clinical heterogeneity of depression have been laid bare by the work of Nemeroff and colleagues. In an imaginative study designed to test hypotheses derived directly from laboratory animal studies of the influence of early social experience on the development of the HPA system, Heim and colleagues studied 49 healthy women with or without current depression, with or without a history of childhood sexual or physical abuse. They found that women with a history of childhood abuse exhibited increased pituitary–adrenal and autonomic responses to a standard laboratory stressor compared with non-abused controls (Heim et al., 2000). This effect was largest in women with current symptoms of depression (and anxiety). Women with a history of childhood abuse and a current major depressive episode exhibited a more than six-fold greater ACTH response to stress than did controls. In a subsequent study, with a similar design, Heim and colleagues used standard neuroendocrine challenge tests to further probe HPA-axis responsiveness in the same clinical groups. Women with childhood abuse histories, but without major depressive disorder, exhibited enhanced ACTH responses to CRF administration, whereas those with both childhood abuse and major depressive disorder, and with depression alone, demonstrated blunted ACTH responses, suggesting some adaptive downregulation of pituitary sensitivity to hypothalamic CRF (Heim et al., 2001). Consistent with this, in an ACTH stimulation test, abused women without major depressive disorder exhibited lower baseline and stimulated plasma cortisol concentrations. These data prompted the authors to speculate that childhood abuse generates a sustained upregulation of CRF production, which leads to adaptive downregulation of CRF receptors in the pituitary, but that under subsequent conditions of stress, perturbed regulatory function throughout the axis generates symptoms of depression and anxiety. Further, dichotomising a population of depressed female subjects, the same group have gone on to demonstrate that hippocampal atrophy, a previously inconsistent, but much heralded, finding in clinical studies of depressed patients, may be attributable to the effects of early environmental trauma, and not to depression (Vythilingam et al., 2002). These studies, although originally prompted by specific hypotheses partially derived from (and refined by) animal models, serve to highlight the lack of confidence that we ought to have in the clinical neuroscience of depression. The inconsistency of reports from different studies can be attributed to many influences. Early experience is clearly one of the most potent. These issues are not a matter purely for academic activities. Building on the above, a recent reanalysis of the largest well-conducted clinical trial of treatment efficacy in chronic depression (described in greater detail below) has suggested that antidepressant drug efficacy is profoundly diminished (rendered virtually undetectable) by the presence of a prior history of childhood abuse, whereas the efficacy of structured psychotherapy is not (Nemeroff et al., 2003). Why are such observations relevant for the development of animal models of depression? Let us consider our state of knowledge concerning the bedrock of both the clinical management of depression and the pharmacologically based models used for drug discovery—the efficacy of the group of drugs called antidepressants
Clinical trials and antidepressant efficacy
There is extensive literature documenting the apparent efficacy of antidepressant drugs, dating back to the discovery of the antidepressant effects of monoamine oxidase inhibitors (MAOIs) in the late 1950s (Kline, 1958). There are a number of systematic meta-analyses examining aspects of antidepressant drug efficacy. The question is not so much whether antidepressants work, since meta-analysis data supports efficacy claims of all classes for antidepressants (Joffe et al., 1996). The more important question which has not yet been answered satisfactorily is how large are the effects of antidepressants? Moncrieff et al. (2004), in their analysis of antidepressants versus active placebos (a placebo containing an active drug, but not one known to have an effect in the disorder being examined) concluded that differences between antidepressants and active placebos were small. Furthermore, Kirsch et al. (2002) have recently reported their controversial analysis of the antidepressant drug efficacy data submitted to the US Food and Drug Administration (FDA), between 1987 and 1999, regarding the six most commonly prescribed antidepressant drugs, including the selective serotonin re-uptake inhibitors fluoxetine, paroxetine, sertraline and citalopram, plus the atypical antidepressants venlafaxine and nefazodone. The FDA dataset includes analyses of data from all patients who attended at least one evaluation visit during a clinical trial, even if they subsequently withdrew from the study prematurely. These data are derived from all of the wellcontrolled efficacy trials of the use of these medications for the treatment of depression. FDA medical and statistical reviewers access the raw data and evaluate the trials independently, and the findings of the primary medical and statistical reviewers are verified by at least one further reviewer. Finally, the analysis is reviewed by an independent advisory panel. These processes and decisions form the basis upon which such medications are approved. Such studies are virtually always short-term, rarely exceeding 12 weeks of treatment. Hence, the ‘real world’ clinical value of any information derived from such studies remains relatively restricted. To try to achieve a degree of ‘generalisability’, the FDA restrict their analyses to trials where over 70% of participants complete the study. According to Kirsch and colleagues, only 4 of the 45 studies included in their analysis met this criterion. As is frequently the case, the relevant trials were virtually all restricted to the study of short-term drug efficacy in a population of outpatients with mild to moderate depression. Kirsch and colleagues concluded that only 18% of the observed change during these short-term studies in patients with non-severe forms of depression could be attributed to the active effects of medication. Active medication and placebo both shared 82% of the maximum clinical changes observed. The authors therefore concluded that, assuming the effects of antidepressant medication and placebo are additive, the effect of medication, even in this clinically ‘favourable’ group (very few or no psychotic or suicidal participants, very little comorbidity), is extremely modest, perhaps negligible, and potentially of little clinical significance. In a similar vein, Posternak and Miller analysed data from 221 depressed patients who were allocated to ‘waiting list-control groups’ for the purposes of clinical trials of psychological treatment methods. They found that almost one in five of these ‘controls’ demonstrated a short-term reduction in depression scores that would merit description as a clinically significant response within the context of a typical acute pharmaceutical treatment study (Posternak and Miller, 2001). Hence 20% of depressed trial participants achieve a reduction in symptoms equivalent to a positive response in drug trials without even placebo pill administration. Such responses presumably also occur in drug studies, but trial design does not usually permit such observations. Clearly, such assertions are highly controversial and strike at the heart of psychopharmacological understanding. There has been considerable debate over the nature of these analyses and how these data can best be interpreted (Keitner et al., 2003; Zimmerman et al., 2002) However, the single most comforting suggestion for psychopharmacology is, perhaps, that our drugs may have powerful antidepressant effects that are actually masked by current clinical trial designs and that we ought to rethink our research designs. What might this mean for our belief that manipulation of monoamine metabolism and function represents a key approach to the development of antidepressant medications? Which novel compounds have been tested and considered to fail because of inherently flawed clinical trial design? Again on a positive note, we must consider the possibility that psychiatrists’ faith in antidepressant drug efficacy may, in part, derive from robust clinical experiences whereby the most severely ill derive the greatest benefit but these patients are rarely, if ever, studied. These criticisms are as much a reflection of incomplete clinical research as they are a criticism of animal models of the disorder. However, such a focus on the unknowns in the treatment of depression should provoke both clinical researchers and basic scientists to strive towards more robust methods to test our hypotheses.