KAREN L. OVERALL

Biology Department, University of Pennsylvania, Philadelphia, Pennsylvania, USA and Atlantic Veterinary College, University of Prince Edward Island, Charlottetown, Prince Edward Island, Canada

Some of the more reductionistic mechanistic or causal levels can be tested to some extent using treatment (the rare, very specific pharmacologic agents), but few phenotypic diagnoses can be specifically tested using behavior modification. Regardless, the logic for using very specific phenomenological diagnoses is to (i) assess and identify the particular behavioral manifestation that needs to be altered or assessed, and (ii) to identify areas where specific behavioral intervention can be useful (Overall, 1997a,b, 2013).

13.2 What is a Diagnosis?

Cautious and discrete use of terminology will lead to clear thinking. Phenotypic diagnoses are useful because they cluster together patients whose pathologies are more similar than they are different, but that differ from other pathologies in some major aspect of form or context. Phenotypic diagnoses may comprise a varying mechanistic population. For example, fear can be caused by neurodevelopmental delays that affect amygdala and hippocampal volume (Kim and Diamond, 2002; Mahar et al., 2014; Schoenfeld et al., 2017) or by some genetic change in a neurotransmitter that putatively affects how receptors and neurotransmitters interact (Campbell et al., 2019).

The main problem in understanding behavioral conditions involves the fact that the act of exhibiting the behavior alters other mechanistic levels that then go on to alter the phenotype (Table 13.1; Fig. 13.1). The dynamism of behavior means that the pathology one sees now may not be the pathology that was seen last week. We know that many behavioral conditions in humans, rodents, and dogs progress if left untreated. How these conditions progress may depend on mechanism and endophenotypes. As we move toward precision medicine, these differences will become apparent and should shift how we group diagnoses. For us to benefit from emergent science we need to acknowledge that clear use of terminology helps to make apparent the parts of phenotypic diagnoses that are consistent, so that we can understand and separate them from those that are more complex.

Separate clusters of phenotypes that are characterized by shared nonspecific signs may be ‘endophenotypes’ (Gottesman and Shields, 1972). For example, once the definitional criteria are met, condition A could sort into two phenotypic groups based on treatment response. In human psychiatry, assessments for the number of categories to which the patient responds and/or the intensity of the response have been used as one form of biomarker, particularly as they respond to treatment (Perlis, 2011). In the simplest scenario, group 1 responds only to drug 1 and group 2 responds only to drug 2, although behaviorally the groups are indistinguishable. A pattern like this would hint that two underlying mechanisms are functioning (Fig. 13.2). Condition B may also have a group that responds only to drug 2 and although these are different conditions, these two variants may be endophenotypes, sharing an underlying mechanism.

In another variant of this example, the definitional criteria are met, but group 1 most commonly displays signs 1–3 and group 2 displays signs 3–5. The question now becomes whether shared or separate mechanisms contribute to these clusters (Fig. 13.2). If these clusters are truly wholly separate at all levels of mechanism, one could rationally argue that these are two truly phenotypically separate diagnostic conditions, and that sign 3 is a truly nonspecific, noninformative sign for this level of inquiry.

When we know little about the mechanisms linking all levels of diagnosis we have broader, more inclusive diagnoses. To take the example of cognitive impairment in humans and cognitive dysfunction in dogs and cats, the list of potential pathological changes at levels below the phenotypic include plaque and tangle formation, vascular pathologies, neurochemical deficits, cellular injury, inflammation, oxidative stress, mitochondrial changes, changes in genomic activity, synaptic dysfunction, disturbed protein metabolism, and disrupted metabolic homeostasis (Stephan et al., 2012). Some of these factors drive the behavior we see – e.g., the confusion, the decrease in problem solving ability – while some drive the accompanying physical factors we see – e.g., the locomotor and elimination changes. Finally, some of these factors may drive the neurodegeneration affecting these other changes. The situation is further complicated because some of what we see are simply derivative signs of damage due to disease progression. Such complexity renders an understanding of how behavioral signs interact with each other essential and also how tests of mechanisms are important. Few somatic diagnostic domains are so plastic and dynamic.

The genetic study of these behavioral and biological intermediate phenotypes/endophenotypes can identify patterns of behaviors that may act as markers for later pathology. For example, a group of ‘high reactive’ human male infants can be characterized by specific patterns of reactivity to visual, olfactory, and auditory stimuli. When these individuals are followed through time, subsequent complex social behaviors including shyness or fearful interactions with strangers become apparent. There must be a mechanistic link between the earlier ‘reactivity’ and the later social fears, and in this example, the reactivity affects functioning of the amygdala (Schwartz et al., 2012). Yet we do not know if the altered functioning of the amygdala is due to changes in migration of neurons, altered pruning or arborization of the neurons, or altered synaptic spine density that then leads to decreases in synaptic density (Brennand et al., 2012). All of these processes have been implicated in autism spectrum disorder (ASD) in humans, and there is evidence for all of them causing one or more morphs of ASD. If these link to an intermediate functional endophenotype, we can more finely parse our phenotypic diagnosis in a way that short cuts how we seek treatment. If not, we may be able to develop a further series of diagnostic tests that suggests some treatment or prevention strategies are better than others for some subsets of patients with the diagnosis.

The implementation of ‘necessary and sufficient’ criteria, using the terms as they are used in logical and mathematical applications, is a refinement over descriptive definitions of terms. These act as qualitative, and potentially quantitative, exclusion criteria, allowing for uniform and unambiguous assessment of aberrant, abnormal, and undesirable behaviors. A necessary criterion or condition is one that must be present for the listed diagnosis to be made. A sufficient criterion or condition is one that will stand alone to singularly identify the condition. Sufficiency is an outcome of knowledge: the more we learn about the genetics, molecular responses, neurochemistry, and neuroanatomy of any condition as well as its behavioral correlates, the more succinctly and accurately we will be able to define a sufficient condition.

These criteria are not synonymous with a compendium of signs associated with the condition, as discussed. The number of signs present and their intensity may be a gauge for the severity of the condition, or act as a flag when there can be variable, nonoverlapping presentations of the same condition. The pattern by which the signs cluster will help in defining heterogeneity of the underlying afflicted population, may identify endophenotypes, and will permit epidemiological studies and tests of underlying mechanisms to be conducted.

Implicit in this approach is that there is no known underlying somatic (physical/physiological) reason for the behavioral problem and that somatic ‘causes’ have been ruled out. It is also important to remember that classifications as discussed here represent diagnoses of problem behaviors, not just descriptions of a behavioral event (i.e., impulse control aggression can only be a diagnosis for an abnormal behavior, but ‘protective aggression’ can be both a diagnosis and description). The proposed terminology represents an attempt to create a terminology that is internally consistent, easily used because of its descriptive utility, and informative because of the manner in which it allows data (e.g., demography, associated nonspecific signs, etc.) to be collected and used to test ideas about various levels of mechanism, while concurrently avoiding psychological jargon. This is harder than it sounds.

This approach is actually similar to that taken by the American Psychiatric Association for the Diagnostic and Statistical Manual, and for the World Health Organization and International Classification of Diseases diagnostic guidelines. The required criteria for human psychiatric diagnoses are embedded within the descriptions of the conditions. Because large numbers of patients are seen in human psychiatry, subgroups of patients can be characterized by nonspecific signs, demography, treatment responses, etc., often within the diagnostic criteria. These groupings have increasingly become the basis for diagnosis, without consideration of the original criteria. In other words, many diagnoses in human psychiatry are now actually based on nonspecific signs, and then assigned a label that may not reflect the biological reality. This failure is due, at least in part in the USA, to the need to have a diagnostic code to receive payment for health care or disability, and it is one reason that genome scans utilizing diagnostic codes have produced so little useful information. This failure in human psychiatry has led many researchers to suggest that we refocus on precision medicine, endophenotypes (Castellanos and Tannock, 2002), stratified diagnoses, and a research domain criteria (RDoC) approach that focuses on integrated response and presentation clusters (Insel et al., 2010; Kapur et al., 2012; Insel, 2014)

In the world view presented here, diagnosis and treatment is about both understanding the neurochemical changes that occur with learning and repeated exposure, and about becoming humane. To do this, we must begin to see the world from our animal patient’s point of view, which requires that we understand normal ethology and behavioral ontogeny of that species. Heuristically, this approach minimally requires that we let go of labels which may say more about us and our needs than they do about the behavior. As the field of veterinary behavioral medicine advances we should become more mindful of terminology, issues, and approaches which can inadvertently do more harm than good.

The unclear distinction between normal aggression (the aggressor is truly threatened and aggressive behavior is adaptive) and abnormal aggression (there is no realistic threat to the aggressor) exists, and is a function of our lack of knowledge about how behavioral conditions develop. The extent to which an animal deviates from ‘normal’ in aggression or any other suite of behaviors may depend on ontogeny, multiple gene effects, and pleiotropic environmental effects. If anxiety-based aggression has a causal pattern similar to other anxiety-based conditions like obsessive-compulsive disorder (OCD), both a familial or genetic ‘predisposition’ and a social stressor play roles in the development of the aggression (Overall and Dunham, 2002).

13.3 Understanding Different Levels of Mechanistic Interaction

Identification of a diagnosis using definitional criteria represents an algorithmic approach that clusters behaviors of patients that are more similar to each other and separates them from those less similar. This clustering, or labeling as a diagnosis, does not mean that patients will be equally afflicted, or that they are all exhibiting the same underlying pathology even if their behavior is the same (see Table 13.2). This approach acknowledges variability in cause, variability in presentation, and that there may not be a unitary causal mapping.

The value of a phenotypic diagnosis should be to help the clinician and client alike to understand the provocative circumstances that can induce a worsening or an improvement in the behavior and the distress that goes with it. Examples for four sets of conditions – canine impulse control aggression, canine separation anxiety, canine noise reactivity and phobia, and canine and feline OCD – will make clear the value of these approaches (Table 13.3). There are now sufficient data for these conditions so that we can examine patterns of pathology for at least one other level than the phenotypic one, and for some putative endophenotypes.

13.3.1 The evolving story of impulse control aggression

Aggression is best defined as an appropriate or inappropriate, interspecific or intraspecific challenge, threat, or contest resulting in deference or in combat and resolution (Overall, 1997a). The importance of context cannot be over emphasized in any evaluation of aggression.

Most abnormal aggressions are the result of underlying anxiety (Overall, 1997a, 2000; King et al., 2000). Some of the best data for aberrant or abnormal aggression involve one of the most controversial canine behavioral diagnoses: impulse control aggression (formerly called ‘dominance aggression’ and sometimes now called ‘conflict aggression’ or ‘aggression to humans’; [Luescher and Reisner, 2008]). Impulse control aggression is about control, or access to control, in direct social situations involving humans. This discrete definition has the advantage of not coupling the challenge to food (food-related aggression), toys (possessive aggression), or space (territorial aggression). These types of aggression can all be correlates of impulse control aggression and when associated with it may be indicative of a more severe situation. This diagnosis cannot be made on the basis of a one-time event.

This approach is a radical departure from the common descriptions of this aggression that specify that the dog will often react to being pushed on, to being corrected with a leash, or to being pushed from a sofa or a person. The number of situations in which the dog reacts inappropriately or the intensity with which he or she reacts do not affect the necessary and sufficient conditions, although these factors may affect the ability to treat the condition, the risk to people, and the prognosis.

The range of behaviors manifest in this condition includes postural threats and stares to sudden stiffening and bites. This is the primary category of canine aggression in which little to no obvious warning is given (Borchelt, 1983). Careful observation may reveal pupil dilation and a slight stiffening immediately prior to the aggression. The classically afflicted dog growls, lunges, snaps, or bites if they are stared at, physically manipulated – often when reaching over their head to put on a leash, physically disrupted, or moved from a resting site – no matter how gently this is done, and when they are physically or verbally ‘corrected’.

Within the population of dogs manifesting the behavior at social maturity, at least two phenotypic groups have been identified: (i) those dogs that are not able to function using the social cues in the human environment and become explosive when they reach their stimulus threshold (the truly impulsive dogs); and (ii) those dogs that are uncertain of the human social environment and provoke it to gain information about what expected social responses and consequences could be (the dogs who use control as a tool) (Overall, 1997a). Both phenotypes of this condition are forms of internalized rule structures that have gone wrong. Keys to treatment include replacement with rule structures that clearly and humanely specify expectations, as well as the use of anti-anxiety medications.

As for most other behavioral conditions, this aggression commonly fully develops during social maturity when neurochemistry undergoes changes that will result in the individual’s adult neurochemical profile. However, dogs exhibiting this behavioral abnormality at social maturity tend to be male, whereas affected females exhibit the behavioral pathology in puppyhood, suggesting that this is a multi-factorial disorder with different underlying mechanisms leading to similar phenotypes (Overall, 1995; Overall and Beebe, 1997).

Using either ante-mortem imaging or post-mortem neuroanatomical or cytoarchitectural studies, little work has been done on impulse control aggression or impulsivity, per se, although limbic system structures in general have been related to impulsive risk-taking, behavioral timing, and time judgments (Nedergaard et al., 2002).

The serotonin system has been implicated in both canine impulse control aggression and in human impulsivity. Affected dogs in one study (Reisner et al., 1996

Stay updated, free articles. Join our Telegram channel

Join