Comorbidity can be described broadly as the presence of more than one disorder in a person in a defined period of time (Wittchen & Essau, 1993). Stimulated by the introduction of explicit diagnostic criteria and operationalised diagnoses in the Diagnostic and Statistical Manual of Mental Disorders (DSM–III; APA, 1980) and the Diagnostic Criteria for Research in ICD–10 (WHO, 1991), numerous studies in the 1980s and early 1990s, have assessed the extent, the nature, and more recently, the implications of comorbidity for a better understanding of mental disorders. Most studies investigated the association of anxiety and mood disorders, but increasingly there are also studies looking into the association of mood disorders with other disorders (such as somatoform and substance use disorders (Wittchen et al, 1993, 1996)) as well as with somatic conditions (axis II) and personality disorders (axis III).

Comorbidity has become an increasingly popular theme in psychiatry and clinical psychology, although its heuristic value was recognised long ago. Frequently used in research and practice, no definition of comorbidity is uniformly accepted and it has no comprehensive and coherent theoretical framework. These factors have led to substantial variation in the magnitude of comorbidity across studies. The variability in the definition, assessment and design of comorbidity studies has led to an increasingly complex and confusing picture about the potential value of this concept. The full exploration of mechanisms of comorbidity requires an interdisciplinary approach to investigating nosology, assessment, and underlying models of comorbidity, as well as experimental study designs beyond the scope of clinical and epidemiological studies. A more precise specification of comorbidity patterns might help identify common biochemical and cognitive markers relevant in the aetiology of specific mental disorders as well as comorbid conditions. Critical issues that might help us understand and explain the variability of findings are described.

General population data are presented on the prevalence and correlates of comorbidity between DSM–III–R major depressive disorder (MDD) and other DSM–III–R disorders. The data come from the US National Comorbidity Survey, a large general population survey of persons aged 15–54 years in the non-institutionalised civilian population. Diagnoses are based on a modified version of the Composite International Diagnostic Interview (CIDI). The analysis shows that most cases of lifetime MDD are secondary, in the sense that they occur in people with a prior history of another DSM–III–R disorder. Anxiety disorders are the most common primary disorders. The time-lagged effects of most primary disorders on the risk of subsequent MDD continue for many years without change in magnitude. Secondary MDD is, in general, more persistent and severe than pure or primary MDD. This has special public health significance because lifetime prevalence of secondary MDD has increased in recent cohorts, while the prevalence of pure and primary depression has remained unchanged.

From the Zurich cohort study (n=591), the association of major depressive episodes and recurrent brief depression (RBD) with other psychiatric disorders is presented cross-sectionally at age 28 and 30 years, and over ten years (age 20 to 30 years). Longitudinally, the odds ratios of major depression are highest with dysthymia (4.4), generalised anxiety disorder (4.4), panic disorder (2.7), hypomania and agoraphobia (2.6), and social phobia (2.4). There is a significant association with cannabis consumption and smoking. Follow-up data over nine years are available for 41 patients with a major depressive disorder (MDD) and 62 with RBD: approximately 20% of MDD patients did not receive a diagnosis during follow-up. Major depression reoccurred in 32%, became bipolar in 24%, or developed into RBD in 24%. RBD remitted in 41 %, reoccurred in 35%, turned into major depression in 22%, and became bipolar in only 7%. Longitudinally, MDD and RBD show a symmetrical diagnostic change in a quarter of the cases. There is no substantial development of MDD or RBD into minor depression or generalised anxiety disorder. Thirteen per cent of those with RBD later developed panic disorder.

The World Health Organization collaborative study on “Psychological Problems in General Health Care” investigated the form, frequency, course and outcome of common psychological problems in primary care settings at 15 international sites. The research employed a two-stage case-finding procedure. GHQ–12 was administered to 25 916 adults who consulted health-care services. The second-stage assessment (n=5438) consisted of the Composite International Diagnostic Interview (GDI), the Social Disability Schedule, and questionnaires. Possible cases or borderline cases of mental disorder, and a sample of known cases, were followed up at three months and one year. Using standard diagnostic algorithms (ICD–10), prevalence rates were calculated for current disorder (one-month) and lifetime experience disorder. Well-defined psychological problems are frequent in all the general health-care settings examined (median 24.0%). Among the most common were depression, anxiety, alcohol misuse, somatoform disorders, and neurasthenia. Nine per cent of patients suffered from a “subthreshold condition” that did not meet diagnostic criteria but had clinically significant symptoms and functional impairment. The most common co-occurrence was depression and anxiety. Comorbidity increases the likelihood of recognition of mental disorders in general health care, and the likelihood of receiving treatment.

Although categorical diagnoses are valued by clinicians and those wishing to collect homogeneous groups of patients for research projects, in the field of non-psychotic mental disorders they correspond poorly to actual clinical syndromes. Indeed, patients often satisfy several sets of categorical criteria simultaneously. A more parsimonious representation of symptomatology can be obtained by using a dimensional model for symptoms – with two correlated axes corresponding to anxiety and depression for ratings derived from the Present State Examination. The Composite International Diagnostic Interview is likely to produce a third dimension – somatic symptoms – because of a radically different symptom content. Preliminary data analysis indicates that the somatic symptoms dimension correlates with both anxiety and depression. Symptomatic individuals in particular areas of dimensional space will satisfy several criteria for categorical diagnoses, and thus display comorbidity. However, evidence that these disorders are distinct is far from persuasive, as they appear to share several common causes.

Differences in rates of depression across and between populations can be considerable and are largely determined by psychosocial factors. Such findings have important implications for comorbidity. On the basis of multiplying independent probabilities, the proportion of comorbid conditions increases as base rates of the disorders increase in a population. If such a chance combination has clinical implications, it would appear to be a “fact” of significance irrespective of biological underpinnings. In a recent survey of 404 women living with at least one child in an inner-city area, the rate of both anxiety and depression was highly related to the childhood experiences of neglect and abuse. However, adversity in adult life (e.g. widowhood or divorce), which might be expected to relate to current stressors, was only related to the rate of depression. The two risk factors had a considerable impact on comorbidity by increasing the rate of each disorder and thus the probability of their occurring together. They had a lesser impact as “common antecedents”. If both influences are considered, such adversity explains around half of the comorbidity. This is a conservative estimate of the impact of psychosocial factors.

Associations between affective disorders, anxiety disorders, and substance use disorders were examined in epidemiological studies conducted in Germany, Switzerland, Puerto Rico, and the mainland US. There was a remarkable degree of similarity across studies in the magnitude and type of specific disorders associated with the affective disorders. Comorbidity with affective disorders was greater for the anxiety disorders than for substance misuse. Panic disorder was the subtype of anxiety that was most highly comorbid with depression. Social phobia was the specific phobic type with the strongest association with the affective disorders. The magnitude of associations between substance misuse and affective disorders generally was quite low and less consistent across sites. No major differences were found in the patterns of comorbidity by gender or age group, affective subtype or prevalence period. The onset of anxiety disorders generally preceded that of depression, whereas alcohol misuse was equally likely to pre- or post-date the onset of affective disorders. Finally, comorbidity was associated with an elevation in treatment ratesacross all sites, confirming Berkson's paradox on an international level.

In both clinical and epidemiological samples, major depression (MD) and generalised anxiety disorder (GAD) display substantial comorbidity. In a prior analysis of lifetime MD and GAD in female twins, the same genetic factors were shown to influence the liability to MD and to GAD. A follow-up interview in the same twin cohort examined one-year prevalence for MD and GAD (diagnosed using a one-month minimum duration of illness). Bivariate twin models were fitted using the program Mx. High levels of comorbidity were observed between MD and GAD. The best-fitting twin models, when GAD was diagnosed with or without a diagnostic hierarchy, found a genetic correlation of unity between the two disorders. The correlation in environmental risk factors was +0.70 when GAD was diagnosed non-hierarchically, but zero when hierarchical diagnoses were used. Our findings provide further support for the hypothesis that in women, MD and GAD are the result of the same genetic factors. Environmental risk factors that predispose to ‘pure’ GAD episodes may be relatively distinct from those that increase risk for MD.

The observation that people with an anxiety or depressive disorder also have an increased likelihood of having had other anxiety or depressive disorders in their lifetime means either that the separate causes of these disorders have aggregated, much more than chance would allow, or that some general vulnerability factor has made them liable to each and all of the disorders they report. In this paper, three separate sources of information – symptoms occurring in the general population, disorders occurring in the general population, and disorders occurring in patients who have sought treatment – are reviewed. In all three domains of information, a general vulnerability factor, associated with personality trait measures of high trait anxiety and poor coping, emerges as a principal cause of these symptoms or disorders, and accounts for the majority of the variation in the comorbidity of symptoms or disorders. This vulnerability factor is shown to be under substantial genetic control. Nevertheless it can be modified by appropriate treatment, and prevention strategies exist to reduce the incidence of the anxiety and depressive disorders.

In large families with affective illness, identification of a biological variable is needed that reflects brain dysfunction at an earlier point than symptom development. Eye movement disorder, a possible vulnerability marker in schizophrenia, is less clearly associated with affective illness, although a subgroup of affective disorders shows smooth-pursuit eye movement disorder. The auditory P300 event-related potential may be a useful marker for risk to schizophrenia, but a role in bipolar illness is less certain. The distribution of these two biological variables and their association with symptoms in two multiply affected bipolar families is described.

In a single, five-generation family identified for linkage studies through two bipolar I (BPI) probands, 128 members (including 20 spouses) were interviewed. The 108 related individuals had diagnoses of BPI (7), bipolar II (2), cyclothymia (3), or major depressive disorder (19). Eight others had generalised anxiety (1), minor depression (5), intermittent depression (1), or alcoholism (1). Sixty-nine subjects had no psychiatric diagnosis. P300 latency (81) and eye tracking (71) were recorded from a subgroup of relatives within the pedigree. Eye tracking was abnormal in 11 of 71 relatives (15.5%) and was bimodally distributed. In these 11 relatives, clinical diagnoses included minor depression (1), alcoholism (1) and generalised anxiety disorder (1). P300 latency was normally distributed and did not differ from controls. In a second family in which five of seven siblings have BPI illness, P300 latency and eye movement disorder were found in affected relatives and in some unaffected offspring. In these large families, clinical diagnoses of general anxiety, alcoholism and minor depression, when associated with eye tracking abnormality, may be considered alternative clinical manifestations of the same trait that in other relatives is expressed as bipolar illness.

The analysis of patterns of co-occurrence and cotransmission of affective disorders and alcoholism in families may provide clues for understanding the excess comorbidity between these conditions in clinical settings and in the general population. This paper reports the results of a family study of the relatives of patients with bipolar disorder, unipolar depression and alcoholism, and combinations thereof. Excess comorbidity between affective disorders and alcoholism was observed in all groups of relatives. However, the sharing of familial aetiological components was not a major contributor to the excess comorbidity between affective disorders and alcoholism. Unipolar depression and alcoholism segregated independently in families, whereas a modest correlation between familial components of alcoholism and bipolar disorder was observed.

Empirical results from epidemiological studies on pain–depression comorbidity in primary care and population samples have shown that: (a) pain is as strongly associated with anxiety as with depressive disorders; (b) characteristics that most strongly predict depression are diffuseness of pain and the extent to which pain interferes with activities; (c) certain psychological symptoms (low energy, disturbed sleep, worry) are prominent among pain patients, while others (guilt, loneliness) are not; (d) depression and pain dysfunction are evident early in the natural history of pain, but dysfunction and distress are often transient; and (e) among initially dysfunctional pain patients whose dysfunction is chronic, depression levels do not improve but neither do they increase over time with chronicity alone. These results seem consistent with these mechanisms of pain–depression comorbidity; (1) a trait of susceptibility to both dysphoric physical symptoms (including pain) and psychological symptoms (including depression), and a state of somatosensory amplification in which psychological distress amplifies dysphoric physical sensations (including pain); (2) psychological illness and behavioural dysfunction being interrelated features of a maladaptive response to pain evident early in the natural history of the condition, and often resolving during an early recovery phase; (3) pain constituting a significant physical and psychological stressor that may induce or exacerbate psychological distress. Thus, pain and psychological illness should be viewed as having reciprocal psychological and behavioural effects involving both processes of illness expression and adaption, as well as pain having specific effects on emotional state and behavioural function.

Half of all cancer patients have a psychiatric disorder, usually an adjustment disorder with depression. Anxiety about illness, such as cancer, often leads to delay in diagnosis, which has been estimated to reduce prospects of long-term cancer survival by 10% to 20%. Although earlier studies showed that depressed individuals were at higher risk for cancer incidence, later studies have not confirmed this predictive relationship. Nonetheless, effective psychotherapeutic treatment for depression has been found to affect the course of cancer. Psychotherapy for medically ill patients results in reduced anxiety and depression, and often pain reduction. In three randomised studies, psychotherapy resulted in longer survival time for patients with breast cancer (18 months), lymphoma, and malignant melanoma. The physiological mechanisms for these findings have not yet been determined, but four fundamental possibilities for psychotherapeutic effects on physiological change include health maintenance behaviour, health-care utilisation, endocrine environment, and immune function. Thus, effective treatment of depression in cancer patients results in better patient adjustment, reduced symptoms, reduced cost of care, and may influence disease course. The treatment of depression in these patients may be considered a part of medical as well as psychiatric treatment.

The association between the infection produced by the human immunodeficiency virus (HIV) and syndromal or subsyndromal depression has been the topic of several studies in recent years. The results of the WHO Neuropsychiatric AIDS Study, conducted in the five geographical areas predominantly affected by the HIV epidemic, suggest that the symptomatic stages of HIV infection are associated with an increased prevalence of depressive symptoms, and, at least in some contexts in which the spreading of the infection is more recent and the social rejection of HIV-seropositive subjects is harsher, may also be associated with an increased prevalence of a syndromal diagnosis of depression.

Of the various hypothalamic–pituitary–end organ axes, the thyroid and adrenal systems have been implicated most often in affective disorders. Patients with primary thyroid disease have high rates of depression, and patients with Addison's disease or Cushing's syndrome have relatively high rates of affective and anxiety symptoms. However, the major support for these endocrine axes in the pathophysiology of mood disorders comes from studies in which alterations in components of the hypothalamic–pituitary–thyroid (HPT) and the hypothalamic–pituitary–adrenal (HPA) axes have been documented in patients with primary depression. Concerning the HPT axis, depressed patients have been reported to have: (a) alterations in thyroid-stimulating hormone response to thyrotropin-releasing hormone (TRH); (b) an abnormally high rate of antithyroid antibodies; and (c) elevated cerebrospinal fluid (CSF) TRH concentrations. Moreover, tri-iodothyronine has been shown conclusively to augment the efficacy of various antidepressants. Concerning the HPA axis, depressed patients have been reported to exhibit: (a) adrenocorticoid hypersecretion; (b) enlarged pituitary and adrenal gland size; and (c) elevated CSF corticotropin-releasing factor concentrations. All of the HPA axis alterations in depression studied thus far are state-dependent, whereas the HPT axis alterations may be partially trait and partially state markers.

Many psychiatric patients seem to suffer from a multitude of psychiatric disorders. The term comorbidity' was introduced to characterise that situation. This term, however, is multi-interpretable and thus conceals more than it clarifies. Five interpretations of the concept of comorbidity are discussed, in conjunction with the consequences their validity would have for psychiatric research, in particular for the biological branch. If the term comorbidity continues to be used without further qualification, it will slow down the process of conceptualisation of novel research strategies in experimental psychiatry, which is so urgently needed.