Abstract
Obsessive-compulsive disorder (OCD) has enjoyed a resurgence of scientific interest since the Epidemiological Catchment Area (EGA) study in the USA [1] suggested that it is not at all a rare disorder; indeed, it became known as the ‘hidden epidemic’, testament to the alleged degree of hitherto unrecognised morbidity in the general population.
But how common is OCD really? As with any of the anxiety disorders, the symptoms associated with OCD are extremes on a normal continuum, and the stage at which they become a ‘disorder’ is a moot point. The DSM rubric of ‘significant distress’ or ‘impairment of functioning’ is useful enough, but fails to reflect the fact that subjective appraisal of these parameters is often incompatible with objective considerations. For example, an agoraphobia sufferer who is entirely confined to her house by her fears, might not consider herself to be significantly distressed (because the feared situations are always avoided), nor impaired in functioning (because her family has rallied around, support and cosset her, doing the shopping for her, etc.). However, any objective measure of functioning would undoubtedly rate her as impaired. The obverse of this is that some individuals rate themselves as very distressed and disabled by anxiety symptoms, while more objective appraisal does not bear this out.
A question of definition
This scenario also pertains to OCD, and is at the heart of any attempt to define incidence and prevalence of the disorder. The problems associated with definition are no more clearly evident than in a closer consideration of the ECA data themselves. Readers will be aware that the ECA study relied on the Diagnostic Interview Schedule (DIS), a lay-administered, structured interview, to define caseness. The lifetime prevalence of OCD according to the DIS, ranged from 1.9%% to 3.2%% across the five ECA sites. There are, however, immediate problems with acceptance of these figures on face value [2]. For example, the lifetime prevalence is incompatible with the reported 1-year's incidence rate of around 0.8%%, for a usually chronic disorder. Furthermore, the reported gender ratio of patients of 2:1 female: male, is higher than in most clinical samples. And in terms of symptoms, obsessions and compulsions reportedly occurred together in only 9%% of cases, discrepant with the close to 90%% co-occurrence in clinical samples.
In a study to assess the validity of the ECA data, Anthony and colleagues [3] investigated the extent to which DIS diagnoses of OCD (and other disorders) tallied with expert clinician diagnosis. Concentrating on the Baltimore ECA site, a subsample of the DIS-interviewed cohort were independently assessed by psychiatrists, and all available data were then subjected to a ‘clinical reappraisal’ (CR) by psychiatrists. For DSM-III defined OCD, the 1-month prevalence rate according to the DIS was 1.3%%, while the CR figure was 0.3%%. What is more worrying is that the proportion of DIS-positive OCD cases that were also considered cases by CR, was 0.04. In terms of statistical agreement between the two diagnoses, the kappa was only 0.05.
In a separate study, Nelson and Rice [2] re-interviewed individuals ascertained as ‘cases’ of OCD in the ECA study, after an interval of 1 year and reported a stability of diagnosis of only 0.2 (kappa statistic). Put another way, only 56 (19%%) of the 291 original OCD ‘cases’ were re-diagnosed at 1 year. Individuals with obsessions and compulsions were twice as likely to have a ‘stable’ diagnosis, as were those with an early onset of illness.
A further example of the problems associated with making a definitive diagnosis of OCD in a large-scale epidemiological study, is the investigation by Stein et al. from Canada [4]. In this study, lay interviewers used a telephone version of the Composite International Diagnostic Interview (CIDI), to interview a general population sample of 2261 people. Fully 26.2%% of the sample reported having experienced obsessions, and 22.2%% compulsions. However, few of these individuals considered their obsessions and compulsions to be unreasonable or excessive, repetitive or recurrent, or excessively time-consuming (> 1 h/day); thus, only 69 people (3.1%%) met DSM-IV criteria for OCD. However, when a sub-sample of these ‘cases’ was re-interviewed by research personnel, using the Structured Clinical Interview for DSM-IV (SCID), only one-quarter were considered truly to meet DSM-IV criteria, giving a revised 1-month prevalence of 0.6%%. The majority of discordance between the CIDI-positive and SCID-positive cases was due to a labelling of ordinary sources of worry or concern as ‘obsessions’, and a tendency to accept reports of ‘distress’ and ‘disability’ too readily, on the basis of patient report.
One way around some of these problems is to ascertain symptoms, rather than define ‘caseness’ as such. Some indication of the utility of this approach can be found in those epidemiological studies which have adopted the Present State Examination (PSE), which essentially determines presence or absence of symptoms, either currently, or over the individual's lifetime. In the study of Bebbington et al. [5], the PSE was used to interview a stratified subsample of 310 of an original 800 people randomly selected from the general population of Camberwell, southeast London. The PSE questions pertaining to OCD included items on checking, washing and ruminations; rates were 9.1%%, 1.6%%, and 0.9%%, respectively. Looked at another way, however, the actual number of ‘cases’ of OCD, on clinical review by a psychiatrist, was only one patient with OCD ‘severe enough to warrant treatment’; many other individuals with OCD symptoms were considered to have manifested these symptoms secondary to another disorder (mostly depression), or to have symptoms too mild to warrant treatment.
So, how common is OCD? The question is unanswerable, the answer dependent upon the definition of the disorder, and the internal ‘boundaries’ imposed by diagnostic systems. A range of anything from 0.05%% to 3.50%% lifetime prevalence is possible to defend, a close to one hundred-fold difference. Perhaps more useful is focusing our attention, as clinicians, on individuals, and their particular symptoms, and how these symptoms are perceived by them and how they impact on their lives.
If we are unclear about the ‘internal’ boundaries of OCD (i.e. what precise set of symptoms are required to be present for the diagnosis), we are equally unclear about the ‘external’ boundaries of the disorder. Thus, there are a number of other psychiatric and neuropsychiatric disorders whose symptomatology shows many similarities with OCD. This raises important questions with respect to whether some of these disorders might have aetiological links, and/or share similar pathogenetic mechanisms, with OCD. To inform these issues, we turn now to a consideration of the external boundaries of OCD, with a review of the so-called ‘OCD spectrum’.
Obsessive–compulsive spectrum disorders
Given the difficulties inherent in defining caseness of OCD, it is worth considering whether the symptoms are in themselves diagnostically useful. One approach [6] is to consider OCD per se to be a disorder with very permeable external boundaries and a great deal of symptomatic overlap with an array of other disorders, including those of ‘impulse control’ (e.g. pathological gambling, kleptomania, trichotilomania), neurological disorders with associated OCD features (e.g. autism, Tourette's syndrome) and disorders characterised by preoccupation with bodily appearance or sensation (e.g. hypochondriasis, anorexia nervosa, body dysmorphic disorder).
The case for the inclusion of each of these disorders in the OCD spectrum can be defended to a greater or lesser degree. For example, a strong case can be made for body dysmorphic disorder (BDD). The defining features of BDD are preoccupation with bodily appearance (intrusive, distressing thoughts recognised by the individual as being excessive; indeed, having all the hallmarks of OCD ruminations), and repeated attempts to reduce the anxiety associated with such intrusive thoughts (i.e. compulsions: mirror checking, ritualised application of make-up, etc.). Furthermore, there is some overlap in familial loading for OCD and BDD (an excess in family risk for OCD in patients with BDD). And of therapeutic importance, BDD is responsive to the same treatment parameters which have proven efficacy in OCD, namely those antidepressants with potent serotonergic activity (e.g. clomipramine, and the SSRIs), and the behavioural approach of exposure and response prevention [7].
The other psychiatric disorders which have been considered to fall within the OCD spectrum can also be considered to lie on an impulsive-compulsive continuum of ‘risk avoidance’ [6]. As Hollander [6] puts it, impulsive individuals are seen as ‘risk seekers who try to maximise pleasure, arousal or gratification’, while compulsive individuals ‘attempt to avoid harm or reduce anxiety or discomfort, associated with the rituals’. In this model, disorders such as sexual compulsions and impulsive personality disorders would be considered ‘impulsive’, and hypochondriasis, body dysmorphic disorder, and anorexia nervosa, ‘compulsive’. This model has more than heuristic attraction. Indeed, there is some evidence of different neurochemical substrates of disorders at either end of the continuum. For example, impulsive disorders show a response to serotonergic agents characterised by a rapid response which attenuates over time, while disorders at the compulsive end of the spectrum tend to have a lag time before onset of response, but tend to maintain their gains [6].
An overlap with psychosis?
We turn now to a consideration of OCD and psychosis, to explore whether a definite distinction can be made between psychotic processes, and those processes underpinning obsessive-compulsive symptoms. Although textbooks tend to draw an absolute distinction between obsessional thoughts on the one hand, and delusions on the other, it is increasingly recognised that obsessional thoughts can come to be believed, by their tormentee, with a strength of conviction that places them in the delusional spectrum. Indeed, Insel and Akiskal [8] have argued that OCD ‘represents a psychopathological spectrum varying along a continuum of insight’, and that at the extreme end of the spectrum, patients could be considered to have ‘obsessive-compulsive psychosis’. Of course, clinicians should be wary about accepting bizarre and firmly held beliefs as anything other than an indication of a psychotic process, but the recognition that OCD can become ‘delusional’ is important, as treatment should be directed at the ‘primary’ pathology.
Having said this, it has long been recognised that a patient with an evolving psychotic process can present with obsessive-compulsive symptoms; the unwary clinician can be misled and delay appropriate antipsychotic therapy. In more established cases of schizophrenia, OCD is also overrepresented. For example, Eisen et al. [9] found that 6 (7.8%%) of 77 patients with schizophrenia or schizoaffective disorder also met DSM-III-R criteria for OCD.
In an early consideration of this overlap, Stengel [10] explored the ‘interrelationship between neurotic manifestations and psychotic reaction types’, and concluded that ‘the excessive inclination of the obsessional neurotics to reality proving and doubt affected their attitude to psychotic experiences in a favourable way’. Thus, he foreshadowed the current vogue for cognitive-behavioural interventions for individuals with psychotic symptomatology.
An overlap with Axis II disorders?
In a further consideration of the external boundaries of OCD, we now explore the relationship between obsessive-compulsive personality disorder, and OCD. Conventional wisdom has it that individuals with obsessional traits are particularly prone to OCD, but it is often difficult to tease apart the DSM Axis I and Axis II pathology, expressly in early onset cases. Baer et al. [11] assessed 96 patients with OCD, for personality disorder diagnoses, using a standardised interview schedule for DSM-III Axis II disorders. They found that 52%% of the OCD subjects met criteria for at least one personality disorder, with the most common diagnosis being of mixed, histrionic and dependent types. Indeed, only six patients met criteria for obsessive-compulsive personality disorder, and 5 of these had experienced an onset of OCD before the age of 10 years. Thus, obsessive-compulsive personality disorder appears not usually to be the soil out of which OCD grows, and this challenges the current nosology. Furthermore, it is highly questionable whether one can apply a personality disorder label to patients with a very early onset of illness, and some individuals attracting a label of obsessive-compulsive personality disorder almost certainly merely have a severe early onset form of OCD.
Axis I comorbidity
A further complication in defining the external boundaries of OCD, lies in the extensive comorbidity associated with the condition. Indeed, it is more the rule than the exception to find at least one comorbid psychiatric disorder in individuals with OCD. Depression is particularly common in OCD subjects [12].
One of the difficulties here is that of trying to determine cause and effect, and to ascertain which ‘cases’ of OCD might be secondary to another disorder. An extreme approach is that adopted by the CATEGO algorithm, linked to the PSE, in that the symptoms of depression effectively ‘trump’ those of OCD, making it almost impossible to diagnose anyone as having OCD. Clearly, more flexibility is required in this regard, but clinicians will be aware of the difficulties inherent in teasing apart depressive and OCD symptoms, and will have seen patients whose OCD symptoms only come to the fore when they are depressed. Conversely, of course, OCD, with its associated disability and negative impact of family, leisure, and occupational functioning, provides fertile ground for the evolution of depression.
In terms of eating disorders, Rubenstein [13], using the SCID diagnostic interview, found that in 62 patients with OCD, the lifetime prevalence of anorexia nervosa or bulimia nervosa was 12.9%%, with an additional 17.7%% of subjects having had a subthreshold eating disorder at some time in their lives; these rates are far higher than those reported for the general population. Also of interest is that males with OCD were as likely as their females counterparts to have suffered from an eating disorder, in stark contrast to the uniform female excess of eating disorders in general population samples.
The high comorbid occurrence of OCD and anorexia/bulimia nervosa is worth considering further. Indeed, many of the features of these eating disorders could be construed as OCD symptoms. For example, the obsessional thinking about fatness and the often ritualised eating habits, calorie-counting and exercise, as well as the anxiety associated with the ‘fear of fatness’, would all be compatible with diagnoses of OCD or BDD, were it not for specific exclusion criteria in the DSM. But, it does beg the question, touched on above, of the relationship of these disorders to each other.
Having reviewed the internal and external boundaries of OCD, we now outline a putative model for subtyping OCD, on the basis of gender and age at onset. As a prelude, we examine these parameters in OCD, and also explore determinants of good and poor outcome of the disorder.
Obsessive-compulsive disorder in childhood
It is well recognised that many children, as part of normal development, go through a stage characterised by obsessions and compulsions. These often have a somewhat ‘magical’ or ‘superstitious’ quality, such as having to step over every crack in the pavement or touch every lamp post on the walk to school. However, the symptoms rarely cause much distress, nor impair functioning, and usually children ‘grow out of’ them.
In a minority of such children, the child or their parents become sufficiently concerned to seek professional help, and usually only brief intervention is required. However, in a small proportion of cases, the symptoms persist and become entrenched, such that they can be considered the beginning of OCD. Boys tend to be affected more often than girls.
Obsessive-compulsive disorder and gender
Gender effects have increasingly been seen as providing potential clues to the pathogenesis of a number of psychiatric disorders, and in this context a consideration of gender effects in OCD is appropriate. However, the field is once more bedeviled by methodological problems, not least of which is case ascertainment. Thus, while in treated samples of patients with OCD, the male: female ratio is usually around unity, in epidemiological samples, females tend to be more commonly represented than males. For example, Bebbington [5], in reviewing epidemiological studies of OCD, concluded that the male: female ratio is ‘roughly 1:1.5’. This discrepancy between treated and epidemiological samples is presumably a reflection, in part at least, of differences between men and women in terms of help-seeking behaviour, the degree to which the impairment associated with OCD can be tolerated and accommodated by families in male and female family members and the degree of occupational impairment and gender differences in occupational expectations.
The extent to which these factors can explain the difference in gender ratio between treated and untreated samples, is difficult to assess. What is also probable is that males are more likely to get into treatment because they are relatively more prone to a severe form of OCD (although formal attempts to assess gender differences in severity of illness is inevitably complicated by the very fact that those who are most likely to be included in any such study, are those who have sought treatment).
What is clear from both epidemiological and treated samples is that the symptoms of OCD tend to be somewhat different between the sexes. Thus, males tend to have more problems with ruminations, and females are more likely than males to be afflicted with cleanliness and checking rituals. It might be that this plays some part in determining the impact of the illness on the individuals' daily functioning. Certainly, ruminations without rituals are notoriously difficult to treat.
Another issue of potential importance in determining the cause of gender differences in OCD, is age at onset differences between males and females with the disorder. In general, males tend to have a somewhat earlier onset than their female counterparts, and males predominate in cases of childhood OCD, as detailed above. In a study specifically of gender differences in OCD, Castle et al. [14] assessed gender differences in a sample of 219 OCD patients, who had been referred to a tertiary treatment centre in London, UK. The male: female ratio was 1:1.35, and the mean age at onset was 22 years for males and 26 years for females (p = 0.003). Duration of illness prior to seeking help was 11.5 years for males and 9.2 years for females. Males exceeded females among those individuals whose onset of illness occurred before the age of 16 years, whereafter females predominated. This finding is compatible with the notion that there might be relatively discrete subtypes of OCD, to which males and females are differentially prone.
Long-term outcome
A further consideration in any attempt to subtype OCD, is what determines good and poor long-term outcome of the disorder. Although there are numerous studies of the treated outcome of OCD, these have mostly been conducted to determine the long-term efficacy of particular treatment interventions, and thus do not reflect the natural history of OCD. For that, we require studies which have determined the longitudinal course of unselected samples of OCD patients over an extended period. Regrettably (for the epidemiologist, at least), such studies are very difficult to do, and almost always have some methodological flaws.
However, the recent long-term follow-up study of Skoog and Skoog [15] is most enlightening. These authors personally followed up 122 of 251 OCD patients who had originally been admitted to hospital in Sweden between 1947 and 1953, over a mean period of 47 years from illness onset. For a further 22 patients, necessary information was obtained from informants and medical records, resulting in an overall 82%% follow up of surviving patients. Subjects had received an array of different interventions, including psychosurgery (6 patients) and clomipramine (17 patients).
Improvement was observed in 83%% of subjects, with complete recovery in 20%% and subclinical symptoms persisting in a further 28%%. However, 48%% of patients had had OCD symptoms for over 30 years. Associations with poor outcome included early illness onset, low baseline social functioning and having both obsessive and compulsive symptoms. This adds further to the notion that early onset of illness might be a characteristic of a particularly pernicious subtype of OCD. We now outline an integrative and interpretative model for subtyping of OCD.
A neurodevelopmental subtype of obsessive-compulsive disorder?
The parameters outlined above, of the impact of gender and age-at-onset, on the long-term outcome of OCD, reinforce the notion that some patients with OCD might have a neurodevelopmental disorder, in the sense that something goes awry with brain development at an early age (perhaps prenatally or perinatally). Of course, such an hypothesis is currently very much speculative, but a number of strands of evidence can be brought together to support it [16].
One of the parameters underpinning this hypothesis is that OCD has been shown, despite its historical ‘neurotic’ label, to be a disorder associated with brain dysfunction. The compelling evidence for serotonergic dysregulation in OCD, has been alluded to above. With the advent of modern neuroimaging techniques, further light is being shed on brain structural and functional abnormalities in OCD. In a recent review, Saxena et al. [17] noted that structural and brain imaging studies of OCD patients generally suggest abnormalities in orbitofrontal and anterior cingulate cortex, and in parts of thalamus and striatum. These structures have historically been implicated as linked in a functional neuranatomic circuit. Functional neuroimaging studies have shown evidence of hypometabolism in the orbitofrontal cortex of OCD patients, and have found that this correlates with the severity of symptoms and normalises with treatment [17].
Of particular current interest is that the immune sequelae of infection by group A beta-haemolytic streptococcus can result in a variety of neuropsychiatric conditions in children, including tic disorder and OCD. These so-called paediatric autoimmune neuropsychiatric disorders associated with streptococcal infections (PANDAS) are presumed to be consequent upon the formation of antibodies which cross-react with neural tissue [18]. Again, this underlines the importance of brain insult in the pathogenesis of at least some cases of OCD. A further example of brain insult associated with obsessive-compulsive symptomatology is the occurrence of such symptoms as a sequel to encephalitis lethargica [12].
Neurodevelopmental disorders are usually characterised by an early onset, severe symptoms and a poor longitudinal course; these disorders also tend to affect males more than females [16]. As detailed above, those individuals whose onset of OCD occurs in childhood or early adulthood do indeed tend to be male. They also tend to have severe symptoms, significant disability and a poor outcome, with a relatively poor response to treatment. Furthermore, such individuals tend to have certain associated features which are also compatible with the notion that they have a disorder of neurodevelopment. These include an excess of ‘soft’ neurological signs, an excess of motor tics and poor performance on neuropsychological tests of visuospatial functioning. These strands have been brought together in a proposal that a neurodevelopmental form of OCD can be delineated, distinct from a putative ‘primary’ type, characterised by a later onset, a milder/episodic course and a favourable response to serotonergic agents; females are relatively more prone to this latter form of the illness [16]. This subtyping has strong similarities with the models proposed for schizophrenia. Of course, much more research is required before accepting this proposed typology, and the mediating effect of gender and associated confounding factors must also be considered.
Conclusions
We have reviewed here some of the more controversial aspects of the epidemiology of obsessive-compulsive disorder. What is clear is that OCD is not uncommon, though the internal and external boundaries of the disorder are difficult to delineate definitively. The notion of an ‘OCD-spectrum’ of disorders has heuristic appeal, and can also inform aetiological models as well as therapeutic interventions. Of particular theoretical interest is the notion that some patients with OCD might have a neurodevelopmental disorder.