Animal Models of Depression:
A Diathesis/Stress Approach
Paul Willner and Paul J. Mitchell
Animal models of depression are used both as screening tests to discover and develop novel antidepressant drug therapies, and as simulations for investigating various aspects of the neurobiology of depressive illness, including the neuropharmacological mechanisms mediating the effects of antidepressant treatments. These different functions of animal models have different and, to some extent, conflicting requirements.
A simulation of depression aims to mimic aspects of the clinical situation, and should embody a degree of complexity, to permit investigation of the validity of the model. In addition, if a model is to be used to investigate antidepressant actions, a slow onset comparable to the clinical time course is highly desirable, and the model should therefore exhibit differences (either in direction of response, or in response magnitude) between single (acute) and repeated (chronic or continuous) treatment regimes. By contrast, the only essential requirement for antidepressant screening tests is that they make accurate predictions of antidepressant activity. For practical reasons, they should also be cheap, robust, reliable and easy to use (Danysz et al., 1991; Willner, 1991a), and for all of these subsidiary reasons, a screening test should in principle be as simple as possible. However, the view that such screening tests should also respond acutely has changed during the past decade, in line with the drive from the clinic to identify rapidonset antidepressant treatments. By necessity this approach involves the assessment of drug action associated with chronic/continuous drug treatment regimes and an acute response is thus of little value. Appropriate screening tests to be used relatively early during drug development should therefore have the ability to identify the time course of drug action associated with repeated treatment schedules.
The present chapter is concerned primarily to evaluate the suitability of the available animal models as research tools, and we therefore focus initially on issues relating to the validity of the available models as simulations of depression.
In line with current thinking, our assessment of the validity of animal models of depression addresses the three dimensions of predictive, face and construct validity (Willner, 1984, 1991a). The concept of predictive validity implies that manipulations known to influence the pathological state should have similar effects in the model. Face validity refers to a phenomenological similarity between the model and the disorder being modelled. Construct validity implies that the model has a sound theoretical rationale. Some reviewers have advocated the primacy of one of these approaches (predictive validity: Geyer and Markou, 1995; face validity: Weiss and Kilts, 1998; construct validity, Sarter and Bruno, Chapter III). In principle, we share Sarter and Bruno's position, that construct validity is the most important of the three dimensions. In practice, however, the construct validity of animal models of depression is difficult to determine, and therefore we favour a balanced approach in which a view of the validity of a model is formed only after considering all three sources of evidence. We therefore begin by reviewing briefly the three sets of validation criteria, as they apply to animal models of depression.
In practice, the predictive validity of animal models of depression is determined solely by their response to antidepressant drugs. A valid test should be sensitive and specific: it should respond to effective antidepressant treatments ('true positive' effects), including electroconvulsive shock (ECS), and should fail to respond to ineffective agents ('true negative' effects). A model with high predictive validity should therefore maximize identification of both 'true positives' and 'true negatives', but should minimize identification of 'false positives' and 'false negatives'. Furthermore, positive responses should occur at behaviourally selective doses (i.e. those which do not generally disrupt behaviour, or induce motor impairment) that are within or close to the clinical range, and should be demonstrable with a range of structurally diverse compounds. It should be recognized that no animal model has a 100% prediction rate although some complex experimental paradigms have approached this level of predictive ability. Part of the problem lies not so much with the preclinical model but with the fact that there are several grey areas in the clinical literature where it is not known whether certain drugs (e.g. anticholinergics) possess antidepressant activity or not. It is generally agreed that the most effective treatment for depressive illness is electroconvulsive therapy (ECT). A suitable starting point to test the validity of an animal model should therefore be to demonstrate a positive response to repeated ECS. Failure to respond appropriately to ECS would severely question the predictive validity of the model.
About 30% of depressed patients fail to respond to antidepressant drug treatment, while the response rate for ECT is slightly higher. The occurrence of refractory patients probably reflects the heterogeneity of depressive illness. Nevertheless, a model that did not respond to the benchmark tricyclic antidepressants (TCAs), selective serotonin reuptake inhibitors (SSRIs), or the recently developed serotonin–noradrenaline reuptake inhibitors (SNRIs)