The Kraepelinian Dichotomy

[Note: An editorial on this topic appears on pp 297–299.]

Schizophrenia and affective disorder are both characterized by disturbances of thought, behavior, and mood, but for much of the century since Kraepelin¹ distinguished the two, investigators approached them as distinct disorders with quite separate etiologies. However, dissatisfaction with the Kraepelinian dichotomy has been growing,^2–4 and recent evidence from developmental, genetic, epidemiological, and outcome studies suggests that the two have much more in common than was previously thought.

INTRODUCTION

Family, High-Risk, Twin, and Molecular Genetic Studies

Several recent family, high-risk, and twin studies provide evidence that the solution to the old controversy^5,6 regarding genetic overlap is eclectic: the data support both shared and unique liabilities to nonaffective and affective psychosis. For example, in the Roscommon family study it was shown that the familial liability to schizophrenia was at least in part a liability to develop psychosis in general,⁷ and in the Mainz family study an increased risk of unipolar depression was shown in the relatives of schizophrenia probands.⁸ However, the Roscommon family study also showed that the relatives of probands with nonpsychotic bipolar disorder did not have an increased risk for schizophrenia.⁷

Similar results were reported in a recent high-risk follow-up study.⁹ The risk of schizophrenia-related psychosis and affective psychosis was elevated both in subjects at high risk for schizophrenia and subjects at high risk for affective disorder. However, the rate of nonpsychotic affective illness in either high-risk group was not higher than in the control subjects.⁹ A study of all psychotic twins seen at the Maudsley Hospital over four decades also suggested that such individuals inherit a predisposition to psychosis in general, as well as specific genetic factors that determine more exactly the form of the psychosis.¹⁰ A recent family study examined the prevalence of psychiatric disorders in the first-degree relatives of a sample of 103 inpatients with severe and chronic bipolar disorder and in a matched control sample of 84 healthy individuals. The morbid risks for both bipolar disorder and schizophrenia were higher in relatives of patients (4.9% and 2.8%, respectively) than in relatives of control subjects (0.3% and 0.6%, respectively). The relative risks were 14.2 (95% confidence interval [CI] 3.1–64.2) for bipolar disorder and 4.9 (95% CI 1.3–18.8) for schizophrenia. Furthermore, the presence of more than one patient with bipolar disorder in a family increased the risk for schizophrenia nearly fourfold (relative risk=3.5, 95% CI 1.2–10.2). These results suggest that bipolar illness characterized by a high familial loading is associated with increased risk of schizophrenia in the relatives.¹¹

Molecular genetic studies suggest it is possible that a broad phenotype including unipolar depression, bipolar disorder, schizoaffective disorder, and schizophrenia (when accompanied by significant affective symptoms) can result from mutations within certain gene regions.¹² Other evidence raises the possibility of location of genes involved in bipolar disorder and schizophrenia in similar regions on chromosome 18.¹³ Although all of the above studies defy simple summary, it is increasingly clear that a simple dichotomy of disorders is not compatible with the data. We have suggested that it may be more appropriate to think in terms of genetic influences on continuous variation in symptom dimensions of, for example, positive, negative, depressive, and manic symptoms.¹⁴ Rather than patients having to be grouped into dichotomous categories of disorder, each patient can be thought of as having a unique mixture of symptoms from various domains, which are the result of the effect of various risk factors operating across a psychosis continuum.⁴

Developmental Perspectives

One of the most productive lines of research in the past decade has been the study of schizophrenia and affective disorder from a life-course perspective. It is now evident that many psychiatric syndromes can be conceived as the adult outcomes of processes that have their origin in early development and even fetal life. As with the genetic evidence, it is likely that at least some of these early processes are shared by disorders that continue to be classified as separate. For example, factors associated with fetal environmental exposures such as prenatal maternal nutritional deficiency,¹⁵ winter birth,¹⁶ birth in an urban environment,¹⁷ and birth and pregnancy complications^18,19 have been associated with schizophrenia. However, recent studies have shown that fetal exposure to maternal food deprivation later in pregnancy is associated with an increased risk for severe affective disorder requiring hospital admission.²⁰ Similarly, urban birth was associated not only with schizophrenia, but also with affective disorder in a nationwide study, although the effect size was smaller in the latter diagnostic group.²¹ In population-based prospective studies, associations have been shown between adversity during pregnancy and labor on the one hand and affective psychosis on the other.^22,23 Winter birth has also been shown to be associated not only with schizophrenia, but also with affective psychosis.²³

The existence of population-based birth cohort samples in the United Kingdom have made it possible to study prospectively the postnatal life of individuals destined to develop schizophrenia and affective disorder. Especially the 1946 birth cohort (N=5,362) has provided important evidence, as it was possible to identify the individuals who developed schizophrenia and individuals who developed affective disorder with case detection independent of whether they had ever received treatment. In two separate studies, three groups of patients were identified: those who had developed schizophrenia (n=30); those who had developed significant affective symptomatology as children ages 13–15 years (hereafter: CAD; n=195); and those who had developed affective disorder as adults between ages 36 and 43 years (hereafter: AD; n=270).^24,25 Schizophrenia cases were twice as likely to be male, whereas AD and CAD cases were about twice as likely to be female. In the schizophrenia cases, development of motor milestones was delayed. Similar delays were present in the affective disorder sample, but only in the CAD cases with illness expression in childhood and the AD cases whose illness had already been evident in childhood. Educational test scores at ages 8, 11, and 15 years were lower in children destined to develop schizophrenia. Similar differences were seen in children who would develop CAD or AD, but compared with schizophrenia cases, the differences from normal control subjects were smaller. Children who had been described as tired, gloomy, or timid had a higher risk of developing both schizophrenia and AD. However, effect sizes were larger for schizophrenia, and pre-schizophrenia children also had abnormalities on other personality variables, especially those associated with social anxiety. The findings therefore suggest that developmental deviance is more pronounced and more generalized in children destined to develop schizophrenia. However, such developmental problems do not appear to be specific, and especially severe affective disorder that expresses itself at an early age appears to be continuous with schizophrenia.^24,25

Other Evidence

Elsewhere, we have reviewed other evidence suggesting correlated liabilities for affective disorder and schizophrenia.²⁶ For example, long-term follow-up studies have shown that the prognosis for those patients diagnosed as schizophrenic is not invariably bad,²⁷ whereas that for affective psychosis is not always as benign as previously thought.²⁸ There is considerable overlap in outcome with, as always, schizoaffective patients taking up the middle ground.

Models of Psychosis

How can we integrate such findings into a plausible model of psychosis? Crow³ opted for a unitary concept of psychosis, but such a view ignores the obvious differences between patients with severe schizophrenia and those with classic manic depression. We²⁹ initially suggested a dichotomy, but a different one from that described by Kraepelin. We proposed that Schneiderian schizophrenia should be divided into 1) those cases that are neurodevelopmental in origin; such cases, we suggested, are associated with poor childhood function, early onset, negative symptoms, and poor outcome; and 2) adult-onset cases associated with good childhood function, prominent affective features, and a relapsing and remitting course; we concluded that such cases have much in common with affective psychoses and should be classified with them.

However, it soon became obvious that such a system has a similar fault to the Kraepelinian dichotomy; that is, some cases cannot be readily categorized because they seem to manifest characteristics of both conditions. We have therefore moved to a dimensional approach in which continuous symptom dimensions are seen to be under the influence of two major factors.^4,14,26 The first, neurodevelopmental impairment, has its maximal effect at the poor-outcome, predominantly male pole with prominent negative symptoms, and is subject to risk factors such as family history of schizophrenia, obstetric complications, and poor childhood function. The second, the affective–socially reactive factor, has its maximal influence at the opposite good-outcome, more frequently female pole with prominent affective symptoms and is associated with a family history of affective disorder and social adversity. However, both factors operate across the psychosis continuum so that many cases are subject to both; such a nosological system copes better with so-called schizoaffective cases than do dichotomous systems. Indeed, we have shown that such a dimensional system predicts the needs and outcome of patients with psychosis better than orthodox diagnostic systems such as DSM-IV and ICD-10.^30,31

However, the validation or refutation of such a theory must await a more detailed knowledge of the precise relationship of brain structure, function, and psychopathology that underlies psychosis. Fortunately, technical advances have allowed us to make some recent progress. In particular, recent developments in brain imaging techniques have enabled us to study psychotic subjects not only with structural scans, but with also functional scans; this has provided a better understanding of the neural networks subserving cognitive tasks and underlying symptoms such as hallucinations. Below, we provide an overview of some of the neuroimaging approaches that have been applied to the study of these disorders and what these have to tell us about the similarities and differences between schizophrenia and affective psychosis.

STRUCTURAL STUDIES

The development of magnetic resonance imaging has facilitated the study of cerebral anatomy in vivo. Despite differing methods of image acquisition and data analysis, studies of schizophrenic subjects have consistently shown enlarged ventricles,³² with additional findings of reduced whole-brain volume;³³ reduced volume of temporal lobe structures, particularly the hippocampus;³⁴ and reduction of the corpus callosum.³⁵ A recent meta-analysis of 58 structural studies encompassing 1,158 schizophrenic subjects has found a reduction in mean cerebral volume together with increase in ventricular volume (126%) and a bilateral reduction of medial temporal lobe structures (amygdala, left 94% of normal volume, right 95%; hippocampus, left 98%, right 97%; parahippocampus, left 93%, right 95%) to be the most consistent features.³⁶

The incorporation of voxel-based morphometry (a technique derived from the analysis of positron emission tomography [PET] images that enables the construction of a three-dimensional map of regional changes on a voxel-by-voxel basis) to the study of patient groups enables investigators to search for “brain-wide” pathology rather than focusing on preselected candidate regions. Wright et al.³⁷ were among the first to apply this to groups of schizophrenic patients and have found evidence of significant reductions in regional gray matter in the temporal lobe bilaterally and the left insula and prefrontal cortex.

Bipolar disorder has been less intensively studied, and the findings have been less consistent. Ventricular enlargement is again the most consistent finding, although the effect size is smaller than in schizophrenia.³⁸ Apart from this, structural imaging studies of bipolar disorder have suggested that mood disorders are associated with regional structural abnormalities in areas involved in mood regulation rather than global atrophy. The literature is fully reviewed by Soares and Mann,³⁹ who conclude that in bipolar disorder the most replicable findings are white matter and periventricular hyperintensities, large third ventricle, and small cerebellum. In a recent study, Strakowski et al.⁴⁰ found an increase in the size of the amygdala, thalamus, globus pallidum, and striatum, which was not associated with duration of illness, prior medication, or duration of illness.

The possible role of medication as a confounder cannot be ignored in any study involving a patient group, and there is some evidence that the striatal (rather than cortical) volume may be increased in subjects receiving typical neuroleptic medication;⁴¹ interestingly, when patients are switched from a typical neuroleptic to clozapine, these volumes tend to decrease again. To the authors' knowledge there have been no studies of the long-term effects of mood stabilizers on brain structure.

The role of a possible genetic state/trait marker has been investigated in both groups. In a large study of families multiply affected with schizophrenia, Sharma et al.⁴² demonstrated that in such families lateral ventricular enlargement and loss of the normal cerebral asymmetry distinguishes people with schizophrenia from unrelated control subjects. Most interestingly, those relatives who appear to be transmitting liability to the disorder showed these same abnormalities, suggesting that these may be a possible marker for genetic liability to schizophrenia. The finding of cerebral asymmetry may represent a failure of normal neurodevelopment and has been described in postmortem,⁴³ structural magnetic resonance,⁴⁴ and functional studies.⁴⁵ Only a few structural studies have directly compared schizophrenic and bipolar subjects using the same methodology. The first of these, by Harvey et al.,⁴⁶ found no differences between control subjects and bipolar subjects but smaller cerebral and cortical volumes in the schizophrenia group, suggesting that these abnormalities are a reflection of schizophrenia rather than psychosis per se. Pearlson et al.⁴⁷ reported that patients with schizophrenia differed from patients with bipolar disorder on measures of bilateral entorrrhinal cortex, left amygdala, and right superior temporal gyrus volumes (all greater in bipolar subjects) but not on measures of total brain gray and white matter, hippocampus, and parahippocampal gyrus and temporal horn volumes. Using a region-of-interest approach, Roy et al.⁴⁸ replicated the finding that both patient groups had enlargement of the temporal horn of the lateral ventricles. Altshuler et al.⁴⁹ reported amygdalar enlargement in bipolar relative to schizophrenic subjects, and hippocampal reduction in schizophrenic relative to bipolar subjects, and suggested that these relative volume differences might have diagnostic specificity.

Most recently, Lim et al.⁵⁰ compared age-matched bipolar (n=9), schizophrenic (n=9), and normal control (n=16) subjects. The schizophrenic subjects had the most severe gray matter volume deficits as well as greater sulcal and lateral ventricular enlargement than the bipolar patients, but the bipolar subjects still had gray matter volume decrease compared with the control subjects. Thus, the bipolar subjects fell midway between the schizophrenic and control subjects.

To summarize, the structural imaging literature does demonstrate abnormalities within both schizophrenia and bipolar disorder. Both diagnostic groups are associated with increased ventricular volume, although the increase is generally greater in schizophrenia. The pattern of cortical abnormalities suggests relatively global changes in schizophrenia, and so far only the study of Lim et al. has reported gray matter deficits in bipolar patients. There is relatively good evidence for a decrease in schizophrenia in the hippocampus, parahippocampus, and amygdala. No such evidence has been forthcoming in bipolar disorder, and indeed some of the structures in the mood-regulating limbic system may be increased in volume in bipolar disorder.

FUNCTIONAL STUDIES: RESTING STATES

Studies of the resting state were among the first neuroimaging studies to be applied to psychiatric disorders and date back to Ingvar and Franzen's⁵¹ landmark paper demonstrating “hypofrontality” in schizophrenia. Resting-state studies have been frequently carried out in schizophrenia using both PET and single photon emission computed tomography (SPECT) and have yielded results that vary between hypofrontality,⁵² no difference in frontal activation,^53,54 and hyperfrontality.⁵⁵ In addition, these studies have also demonstrated increased activity within the basal ganglia, which is thought to be a sequela of typical neuroleptic treatment⁵⁶ and has been shown to distinguish treatment responders and nonresponders after a single dose of antipsychotic.⁵⁷

There have been relatively few functional imaging studies in bipolar disorder. Studies of the resting state in depression have also described hypofrontality, particularly in the dorsolateral prefrontal and anterior cingulate cortices.⁵⁸ However, there have also been reports of no differences in frontal function⁵⁹ and decreased activation in the caudate and basal ganglia.⁶⁰

Sackheim et al.⁵⁸ found elevated resting cortical blood flow in manic patients, and Baxter et al.⁶¹ reported relatively reduced left prefrontal metabolism in bipolar depressed compared with manic patients. Rubin et al.⁶² described reduced resting blood flow in anterior cortical areas in both depressed and manic patients relative to control subjects, whereas Silfverskiold and Risberg⁵⁹ observed no differences between manic patients and control subjects. In a study of familial mood disorder, Drevets et al.⁶³ used PET to identify a region of hypometabolism in the subgenual prefrontal cortex in depression that was not affected by improvement in mood or treatment. This functional change was reinforced by findings of significantly reduced mean gray matter volume in this region in familial major depressive disorder and bipolar disorder compared with control subjects.⁶⁴

Thus, for both schizophrenia and bipolar disorder, study of the resting state in isolation does not provide a consistent pattern of diagnosis-dependent results. In many respects, the diversity of resting blood flow findings reflects the diversity of the clinical presentations that are being compared. In an attempt to control for this, Dolan et al.⁶⁵ studied patients diagnosed with schizophrenia and patients diagnosed depression where both groups suffered from poverty of speech. Using resting-state PET, they demonstrated a hypofrontality that was related to symptoms but not to diagnosis, suggesting that a common mechanism may give rise to similar symptoms across the two disorders (although the symptoms may be interpreted as negative symptoms in schizophrenia and retardation in depression).

FUNCTIONAL STUDIES: COGNITIVE ACTIVATIONS

A common criticism of the resting-state design (and thus of all the studies described above) is that the interpretation of the results is clouded by the difficulty of ascertaining the subjects' cognitive activity during the resting state.⁶⁶ Conductors of functional studies have therefore moved toward more sophisticated designs that enable investigation of the neurovascular correlates of either symptoms or cognitive activity and have also sought tasks that might serve as state/trait markers for the underlying disease processes.

An example of a task that has been used extensively is verbal fluency, which involves the generation of words from letter cues. Deficits in verbal fluency performance are associated with frontal lobe lesions,^67,68 and, despite variability in study design, the results from verbal fluency experiments carried out using PET have been reasonably consistent.^69–72 This methodology has also been used to investigate cerebral blood flow changes during verbal fluency tasks in patients with schizophrenia. A series of studies in acutely ill, drug-free,⁷³ and chronically ill patients with distinct symptom profiles⁷⁴ has suggested that schizophrenic patients demonstrate abnormally correlated rCBF changes in frontal and temporal regions compared with control subjects. Specifically, when scanned under identical conditions, schizophrenic patients have demonstrated small, positive frontal-temporal correlations in rCBF, compared with relatively large, negative correlations in control subjects.

These findings have contributed to the notion that frontal-temporal dysconnectivity may be a central pathophysiological feature of schizophrenia.⁷⁵ By altering the nature of task demands, we have used functional magnetic resonance imaging (fMRI) to demonstrate within the same scanning session an attenuated frontal activation during an intrinsic generation task (verbal fluency)⁷⁶ and a normal frontal response during an extrinsic generation task (semantic decision making) in schizophrenia.⁷⁷ These findings suggest that any network dysfunction in schizophrenic patients is not fixed, but rather a reflection of task demands.

Studies of verbal fluency in bipolar patients have found either no differences in frontal activation relative to control subjects,⁷⁸ greater orbitofrontal activation in manic relative to depressed bipolar subjects,⁷⁹ or greater dorsolateral prefrontal activation in bipolar than control subjects.⁸⁰ In addition to our studies of verbal fluency in schizophrenia, we have used fMRI to study matched bipolar subjects performing the same tasks; we found no evidence of the same task-dependent hypofrontality, but rather a suggestion of hyperfrontality in the bipolar group.⁸¹ Thus, while there are inconsistencies in the bipolar literature, in contrast to neuroimaging studies of schizophrenia, there have been relatively few reports of “hypofrontality” (and one showing a trend to hyperfrontality); thus, there may be a diagnosis-dependent difference in the manner of processing this type of task.

SUMMARY

Research over the past decade from genetic and developmental studies suggests that schizophrenia and affective disorder share some, but not all, risk factors. In particular, birth cohort studies have shown that risk factors such as prenatal and perinatal hazards or childhood cognitive and social difficulties that were thought to be specific precursors of schizophrenia are also found in pre-affective patients; however, in the latter the effect size is generally smaller.

Unfortunately, brain imaging studies have too frequently assumed the truth of the Kraepelinian dichotomy and have not generally addressed the question of what schizophrenia and bipolar disorder share. Furthermore, compared with the large numbers of studies of schizophrenic subjects, there have been relatively few studies of affective disorder, especially of bipolar patients, and only a handful of studies directly comparing the two disorders.

Although there is no single brain lesion underpinning psychosis, there is evidence, given the caveats that apply to comparisons of different studies of subjects on medication, of altered pathology and brain function in both schizophrenia and bipolar disorder. Thus, structural imaging studies reveal increased ventricular volume and abnormalities of temporal lobe anatomy in both groups; the role of cerebral asymmetry is relatively well established in schizophrenia, and so far only one unreplicated study has extended this to bipolar subjects with psychosis. Functional studies of the resting state suggest that in both diagnoses there may be alterations in perfusion depending on clinical state and symptoms, and that certain symptoms (e.g., poverty of speech) cause similar deficits independent of diagnosis. In addition, functional studies using cognitive tasks demonstrate that both groups have (differing) abnormalities within the frontal and temporal lobes; for example, our own work has demonstrated the presence of task- and diagnosis-dependent hypofrontality in schizophrenia relative to both bipolar and control subjects.

Our conclusion from the brain imaging evidence is that Kraepelin was neither completely right nor completely wrong. Rather, the most parsimonious explanation is that suggested by the twin study of Cardno et al.¹⁰ quoted earlier: that there are some factors common to schizophrenia and bipolar illness, and others that are more specific to the individual condition. The major difference between the two conditions is the presence of severe global cortical abnormalities in schizophrenia but not in bipolar disorder. This may reflect the greater impact of neurodevelopmental impairment in causing dysplastic neural networks in schizophrenia. Such a view is compatible with the dimensional notions that we have put forward elsewhere implying that two major etiological factors are operating in psychosis: 1) neurodevelopmental impairment and 2) a genetic propensity to react to adverse circumstances with mixed affective and psychotic symptoms.^4,14,26