Animal models of human disease serve many different purposes and their utility is critically dependent upon the explicit purpose of the model (Willner, 1984). For the purposes of this article, we will define relevant animal models as those that have been developed overtly to enhance our understanding of the pathophysiology of the common clinical disorder known as depression and to guide the development of novel and more effective treatments. Within this special journal issue devoted to animal models of depression and antidepressant activity, the reader will doubtless be impressed by the range and sophistication of the putative models currently being explored. Having been asked to provide a ‘clinical perspective’ on the status of animal models of depression, we have inevitably reflected upon our state of knowledge of the clinical neuroscience of depression and upon our knowledge of mechanisms of antidepressant treatment activity. This has been an interesting process. As clinicians from different professional backgrounds, we are struck by the continuing poverty of high quality neuroscience derived from clinical studies in psychiatry and psychology, and the resultant chasm that separates basic laboratory science from the assumed ‘knowledge’ that is applied on a daily basis in clinical practice. As a consequence, we must continue to raise questions about the validity of existing animal models in the advance of psychiatric neuroscience. In this article, we seek to highlight the difficulties that require to be addressed before clinicians can view animal modelling as a robust approach to elucidate the neuroscience of depression. Our comments should not be interpreted as criticisms of specific models or of the work of specific groups. Rather, our intention is to encourage a revisiting of relevant clinical investigation to assist in the development of more valid models within which to test those many hypotheses that can never be accessed through direct clinical study.
The complexity of depression as a clinical disorder
Many animal models of human disease have proven useful in elucidating the basic mechanisms of disease processes (Guay-Woodford, 2003); others have assisted in the development of novel and more effective treatments (Hyde et al., 1993; Le Fichoux et al., 1998; McDonald et al., 1999). Within many areas of clinical medicine, compared with psychiatry, both clinicians and basic scientists enjoy the relative luxury of working within a coherent pathophysiological taxonomy. Sometimes, disease classification can be considered at a molecular level and, therefore, justifiable confidence exists with respect to the aetiological and construct validity of the relevant animal models. However, psychiatry remains devoid of such a pathophysiological or molecular taxonomy and is ultimately reliant upon the unwieldy and imperfect syndromal classification tools provided by the Diagnostic and Statistical Manual of Mental Disorders—Fourth Edition (DSM-IV) (American Psychiatric Association, 1994) and The International Classification of Diseases and Related Health Problems, Tenth revision (ICD-10).(World Health Organisation, 1992) A detailed critique of this approach to classification falls out with the scope of this article, although it is the authors’ contention that such nosological approaches have successfully advanced our understanding of other forms of human disease. However, the interested reader is directed towards reviews of classification and diagnostic systems which are relevant to the present debate (Berrios and Chen, 1993; van den Brink et al., 1989).
Current classification systems
Both DSM-IV and ICD-10 operationalise psychiatric diagnoses, including depression, according to the coexistence of dominant clinical features over specified periods of time and they assume a capacity to exclude other potential causes. Both systems permit individual variability in the presentation of depression as long as essential clinical features are shared and thus they acknowledge the heterogeneous nature of the disorder. However, these nosological systems are categorical in nature, and as stated above, do not imply any particular aetiology to the symptoms presented. Therefore, the validity of the diagnosis depends on the reliability of the category to predict responses to treatment and the course of the disorder. Contemporary animal models of depression usually attempt to relate behavioural or neurochemical observations (induced by a variety of means) in the laboratory animal to one or more features described within the DSM or ICD framework. However, due to the clinical heterogeneity of the disorder, we must exercise caution in assuming that the nosological entity called ‘depression’ is a specific disease state, with a specific pathophysiology and predictable outcome. Rather, contemporary thinking moves us towards a ‘dimensional’ or transnosological approach to diagnostic categorisation where the emphasis is on reducing complex behaviours or clinical syndromes into their component parts, or ‘endophenotypes’. Thus, a focus on the generation of robust, discrete pathophysiological changes would seem to represent the best way forward for depression modelling. Even then, however, valid modelling is entirely dependent upon valid, reliable clinical observation.
What is depression?
Surely this represents the first issue upon which we must reach consensus? Most animal studies will, correctly, pin their clinical relevance and scientific value upon the extent to which the model in question is considered to show construct validity (the accuracy with which the model replicates the key abnormalities or phenomena under study within the clinical condition) or predictive validity (a given model’s ability to correctly identify effective antidepressant treatments, usually drugs), in terms of the diagnostic features of Major Depression as defined by DSM-IV, or depressive episode under ICD-10. Is there, however, general recognition that many of the ‘core’ constructs upon which depression models are based, are in fact relatively unreliable clinical observations? Is the profound heterogeneity of phenotypic expression, clinical course and response to treatment of the clinical disorder known as depression recognised? One of the difficulties is that the phenotypic heterogeneity of depression mitigates strongly against the development of ‘single phenomenon’ animal models (e.g. learned helplessness, behavioural despair) with significant construct or predictive validity. Without a cogent and accepted pathophysiological explanation for the depressed state in humans, we must accept that depression may represent the final common pathway of many different disorders of brain function. How would one then set about developing a model with construct and predictive validity where the presumed disease entity in humans could encompass multiple primary pathologies?
“Core” features of depression
There are no symptoms, nor any other clinical features, that are pathognomonic for depression. Assumed ‘core’ psychopathological phenomena such as a blunting, or absence of the capacity to experience pleasure (anhedonia), can also present as a common clinical feature in substance misuse (Uslaner et al., 1999) and schizophrenia (Loas et al., 1996). Indeed, each of the diagnostic features listed within DSM for major depressive episode and ICD for depressive disorder can be found within other medical and psychiatric disorders. The diagnostic criteria for DSM-IV Major Depressive Disorder are given below. The requirement for co-existence of a range of different features within a specified time period does confer strength of precision and reliability of diagnosis for depression that is to be welcomed. This approach has served scientific advance well in other branches of medicine and remains the bedrock of high quality clinical investigation. However, an experienced clinician can sit in front of a patient using the most rigorous structured interview schedule available, derive a research quality diagnosis according to a major classification system, and still be wrong. Amongst many possibilities would be depressive symptoms generated within the context of undeclared Bipolar Disorder, thyroid disease, undiagnosed brain tumour, or the earliest stages of neurodegeneration with Parkinson’s or Huntington’s Disease. Each has very different genetics, clinical features, natural history, treatment responsiveness and indications for treatment. Of course, usually the clinician’s diagnosis will be correct, but this serves to illustrate the multiple pathophysiological routes to the expression of depressive symptoms within the depression phenotype. Furthermore, the same clinician can work through the same process with a group of patients, arriving at exactly the same diagnosis for each—yet the range of clinical features, the clinical course, presumed aetiology, and the responsiveness to treatment can vary dramatically.