The course and outcome of schizophrenia and their prediction.
- Methodological issues
- Definitions and assessment of course and outcome variables
- Length of follow-up
- Cohort attrition
- Other aspects of study design
- Statistical analysis
- The ‘natural history' of schizophrenia before the neuroleptic era Long-term prognosis
- Recent course and outcome studies
- Patterns and stages of the course of schizophrenia
- Geographical and cultural variation
- First-episode psychosis
- Prognosis of specific clinical symptoms and syndromes
- Depression in schizophrenia
- Schneiderian first-rank symptoms
- Prognosis of schizophrenia subtypes
- Predictors of course and outcome
- Limitations of clinical prediction
- Medium-range predictors
- Short-term predictors
- Summary and conclusions
The course of schizophrenia is as variable as its symptoms. Systematic investigations of the courses of the psychoses were initiated by Kraepelin who believed that, in the absence of demonstrable brain pathology and aetiology, a common outcome into ‘psychic weakness' of the clinical syndromes he grouped together as dementia praecox would provide a validity test for the disease entity. Later, Kraepelin revised his claim that the prognosis of dementia praecox was invariably poor and noted that ‘permanent cures' had occurred in about 15 per cent of his cases. (1) Subsequent longitudinal studies have confirmed the striking variability of course as one of the salient characteristic of the ‘natural history' of schizophrenia.
The large number of studies on the course and outcome of schizophrenia published since the beginning of the twentieth century might suggest that the longitudinal aspects of the disorder are well established and exhaustively documented. Unfortunately, this is not the case since the methodological difficulties that accompany this type of research are complex and few studies have adequately dealt with all the major sources of error and confounding. (2)
The studies of the course and outcome of schizophrenia comprise statistical reports on admissions and discharges, long-term follow-back studies (in which the initial features of the cases and the course of the disorder are reconstructed retrospectively from admission records), and prospective investigations (in which patients are enlisted at an early stage of the disorder and followed up for a varying length of time). Each design is vulnerable to bias: admission and discharge statistics usually comprise patients at different stages of disease progression, follow-back studies rely on prevalence samples in which chronic cases tend to be over-represented, and prospective studies, though superior to other designs, tend to exclude patients who initially have diagnoses other than schizophrenia and are subsequently rediagnosed as schizophrenic. The methodological issues that need to be considered in interpreting the results from longidutinal research into schizophrenia include the following.
The use of either ‘broad' or ‘restrictive' definitions of schizophrenia may result in vastly different samples on which follow-up data are reported. Systems with an inbuilt illness duration criterion, such as DSM-III and DSM-IIIR which require at least 6 months of unremitting symptoms and a decline in functioning, are likely to overselect patients already developing a chronic course. The result would be a greater homogeneity of outcome at the cost of a compromised representativeness of the sample as regards the range of possible outcomes of schizophrenia. Diagnostic systems that emphasize the cross-sectional features of the disorder, such as ICD-10 (which requires 1 month's duration of clinically characteristic symptoms) avoid this limitation, possibly at the expense of including some cases of good prognosis that may be aetiologically or pathogenetically different from poor prognosis schizophrenia. However, until aetiology is elucidated, or a validating pathognomonic lesion is established, the decision as to what constitutes ‘true' schizophrenia will remain arbitrary. With regard to prognostic studies, less restrictive systems have the important advantage that a broad spectrum of outcomes would be available at the endpoint of prospective observation, allowing for subgroups to be identified and their characteristics related to the initial manifestations of the disorder and various risk factors.
Definitions and assessment of course and outcome variables
There is no single measure of course and outcome of a complex disorder such as schizophrenia, and blanket terms such as ‘recovery', ‘improvement', or ‘deterioration' tend to conflate substantially different aspects of the evolution of the disorder over time. Most investigators today agree that course (comprising the pathways or trajectories of the disorder) and outcome (the net balance of the clinical and functional descriptors at the endpoint of observation) are multivariate composites. As a minimum, three domains that need not covary over time should be independently assessed: symptom severity, functional impairments including cognitive deficits, and disablement in social and occupational role performance. Each one of these can be further articulated into a number of areas or dimensions. In addition, one must consider extrinsic variables such as measures of environmental and treatment-related influences on course and outcome, as well as subjective experiences commonly described as ‘quality of life'. Standardized reliable instruments (interviews, inventories, rating scales) are required for the assessment of most variables. It should not be forgotten, however, that some of the richest sources of information are the perceptive, in-depth case studies based on personal patient contact over many years. Collectively, such single case observations can generate hypotheses for testing in epidemiologically designed studies.
Length of follow-up
The evidence from previous research suggests that very different impressions of the course and outcome of schizophrenia would be gained depending on the duration of prospective observation and the degree of control over the inclusion of patients that are comparable in terms of age and length of previous illness.
In any follow-up study, a proportion of cases will be lost to observation because of death, migration, refusal of contact, or other reasons for untraceability. Since such loss of subjects is likely to correlate with particular patterns of course and outcome, it is essential to estimate its possible effect (e.g. by statistical modelling) on the interpretation of the final results, especially if cohort attrition is greater than 15 to 20 per cent of the original sample.
Other aspects of study design
Variation in the sources of recruitment of cases (e.g. any admission to a treatment facility or catchment area sampling), and of information regarding course and outcome variables (e.g. face-to-face interviews or collateral data from case notes or informants), can obviously influence the results of any follow-up study. In addition, subtle variations in study design, such as whether investigators assessing patients' symptoms and functioning at any point in time are ‘blind' to data from previous assessments, can influence the final results. Inclusion of a comparison group (e.g. patients with other psychotic disorders) would help evaluate the extent to which any observed patterns of course and outcome are specific to schizophrenia, whereas appropriate controls drawn from the general population can provide reference points for assessing social variables, such as occupational functioning, stressful life events, or habit-related behaviour such as substance use.
Longitudinal research poses a number of specific requirements with regard to data analysis. Thus, the problem of multiple comparisons is likely to arise when examining the data for significant associations; time series, survival, or path analysis may be required when observations are made and recorded at successive time points in the evolution of the disorder; and methods of unconfounding are called for at each step of the analysis of longitudinal data. While no single study to date has met all the rigorous methodological requirements, a number of studies have succeeded in controlling at least some of the sources of bias and confounding. The results from previous research are, therefore, not strictly comparable in specific detail, but are informative as regards the general trends and patterns.
The ‘natural history' of schizophrenia before the neuroleptic era
Since the great majority of schizophrenic patients are today receiving pharmacological treatment, current and recent studies may not reflect the ‘natural' course and outcome of the disorder. Two recent studies in Scotland (3) and India (4) estimated the proportions of never-hospitalized schizophrenic patients at 6.7 per cent and 28.7 per cent respectively. About half of the Scottish patients had been prescribed neuroleptics by their general practitioners while the Indian patients had been virtually untreated. However, in both settings the outcomes of these interesting samples (which presumably approximate the ‘natural' history of the disorder) were heterogeneous and, by and large, did not differ from the outcomes in the treated groups. In a historical study of 70 Swedish patients with first admissions in 1925, the lifetime records were retrieved and rediagnosed in accordance with DSM-III. (5) None of these patients had received neuroleptics. The final outcome was rated as good in 33 per cent (but no patient was considered as completely recovered), as ‘profoundly deteriorated' in 43 per cent, and as intermediate in 24 per cent.
Another, long-term perspective on the course of schizophrenia over successive generations is provided by a meta-analysis of 320 outcome studies on schizophrenia or dementia praecox published between 1895 and 1992 and including a total of 51 800 subjects. (6) Overall, about 40 per cent of the patients were reported as improved after an average length of follow-up 5.6 years. There was a significant increase in the rate of improvement during the period 1956 to 1985 compared with 1895 to 1955, clearly related to the introduction of neuroleptic treatment, (7) but a secular trend towards better outcomes with every successive decade had been present for much longer. Coupled with the virtual disappearance of the most malignant or ‘catastrophic' forms of schizophrenia resulting in a profound defect state after a first psychotic episode, or death (‘lethal catatonia'), these observations suggest that some transition to a less deteriorating course of the disorder had occurred prior to modern pharmacological treatment. (8) Among the factors explaining this shift one should consider improvements in general care, progressive changes in attitudes and hospital regime which occurred in a number of institutions on both sides of the Atlantic in the 1930 and 1940s, as well as heightened expectations that psychosocial measures such as psychotherapy or rehabilitation could result in a cure in some cases.
Three European (12,14,28) and two North American studies, (15,16) in which a total of nearly 1300 patients were followed up for 23 to 37 years, provide a global overview of the long-term course of schizophrenia. Although the studies differ in their design (prospective; (12,14,16) follow-back or retrospective (15,28), and include patients with onsets in the preneuroleptic era who were later treated with antipsychotic drugs for varying length of time, their results have much in common.
Manfred Bleuler's monograph (12) is the account of an intensive study of 208 patients first admitted in 1942 to 1943 and personally followed up by the author for 22 years or until death. Another 23-year follow-up of 504 patients admitted in 1945 to 1959 has been completed by Huber et al. (14) Ciompi (28) interviewed 289 surviving patients in Switzerland first admitted between 1900 and 1962 (median follow-up length 36.9 years).
Notwithstanding methodological constraints that apply to these studies, their findings are a unique record of what probably represents the closest approximation to the ‘natural history' of schizophrenia. In summary, they indicate the following:
- Lasting recovery (‘complete cure') occurred in 20 to 26 per cent of the cases—according to Ciompi (28) 43 per cent had either remitted or exhibited mild residual abnormalities which did not interfere with their living in the community.
- Forty-four per cent were still in hospital (28) and severe chronic states had developed in 14 to 24 per cent.
- In 50 to 75 per cent of the patients, a clinically stable state set in after the fifth year since onset, with no significant further deterioration.
- Remitting course with multiple episodes and full remissions characterized 22 per cent of the patients; (14) catastrophic course (rapid onset of chronic deterioration) was observed in 1 to 4 per cent.
- The 20-year suicide rate was 14 to 22 per cent.
The two American studies largely concur with these findings. In the Vermont study (16) no less than 62 per cent of the cohort had achieved significant improvement or recovery after an average length of follow-up 32 years; the corresponding proportion in the Iowa 500 study (15) was 46 per cent.
The most striking finding from the long-term follow-up studies is the high proportion of patients who recover, either completely or with mild residual abnormalities, after decades of severe illness. This contrasts with the ingrained image of schizophrenia as an intractable, deteriorating illness that many clinicians tend to adopt on the basis of a limited follow-up horizon and patient samples selected for unfavourable course and treatment response. It is unlikely that the high percentage of recoveries in the long-term studies could be explained by cases of affective illness or brief transient psychoses misdiagnosed as schizophrenia (the retrospective rediagnosis of cases according to Feighner's (29) or DSM-III criteria in the two American studies did not alter significantly the results). Similarly unlikely would be the attribution of all the good outcomes to the antipsychotic treatment many of these patients received in the later stages of their illnesses, since comparable proportions of improvement of recovery had been reported for patients who never received neuroleptics. (5) A tentative conclusion from such follow-up research would be that schizophrenia is not invariably a chronic deteriorating disorder and that the progression of the disease can be arrested, or even reversed, at any stage. The causes of such reversibility remain largely unknown, but research focusing specifically on the recovering cases will undoubtedly provide essential clues for understanding the nature of schizophrenia.
Recent course and outcome studies
Results of longitudinal studies published in the last decade do not add substantially new evidence and tend to corroborate the pattern of outcomes outlined by the earlier studies.
The reported rates of recovery or complete remission vary between 12 per cent (30) and 32 per cent. (27) If the rates of improvement or partial remission are added to those of complete remission, the general improvement rate over 5 to 15 years of follow-up would be no less than 30 per cent and may be as high as 50 per cent. On the other hand, the proportion of patients with an early deteriorating course of illness without clinical remissions is remarkably similar across different studies and is in the order of 20 to 35 per cent. There is a relationship between the length of the follow-up and the proportions of patients who are reported as recovered, remitting, or deteriorating, with a general trend towards higher improvement rates in long-term studies. For example, the proportion of good outcomes increased from 10 per cent at the 2.5-year follow-up to 17 per cent at 5 years in one of the studies; (19) it was 31 per cent in a study with a follow-up of 21 to 25 years (17) and nearly 60 per cent at the end of the 32-year follow-up of the Vermont study.(16)
Patterns and stages of the course of schizophrenia
The great heterogeneity of the course of schizophrenia can be reduced to a limited number of patterns into which cases tend to cluster over time. In the long-term follow-up studies referred to above, eight different categories of course were described by Bleuler (12) and Ciompi, (28) and 12 by Huber et al. (14) These classifications were derived from empirical observation, rather than statistical modelling, and conflated into single categories the mode of onset, longitudinal aspects such as psychotic episodes and remissions, and end states. Treating the various aspects of the longitudinal profile of the illness as independent axes in a multidimensional construct has been recommended. (31,32) However, the complexity of statistical modelling of the course of schizophrenia is such that the development of a classification of course that would be both useful in clinical practice and rigorous in a mathematical sense may not be feasible. Therefore, an heuristic compromise between these two requirements should, as a minimum, define operationally and assess separately the number and duration of discrete episodes of illness, the predominant clinical features of each episode (e.g. psychotic or affective), and the number and length of remissions and their quality (presence/absence of residual negative or deficit symptoms and signs). By combining these variables, several patterns of course have been derived that have found good empirical support in international follow-up studies: (33)
- single psychotic episode followed by complete remission;
- single psychotic episode followed by incomplete remission;
- two or more psychotic episodes, with complete remissions between episodes;
- two or more psychotic episodes, with incomplete remissions between episodes;
- continuous (unremitting) psychotic illness.
With some modifications, these longitudinal patterns have been incorporated into ICD-10 and DSM-IV as additional descriptors.
Although the components of the course patterns, such as episode, remission, residual symptomatology, etc. almost certainly do not represent ‘pure' dimensions, it is desirable to restrict the definition of pattern of course to clinical variables only, in order to be able to examine its correlations with risk factors and predictors, such as premorbid impairments or mode of onset, and with social outcomes. The assessment of social functioning independently of the clinical pattern of course is crucial to the study of illness–environment interactions and the causes of disablement in schizophrenia.
At present it does not seem possible to define with any precision discrete stages in the progression of schizophrenic illnesses using combined clinical and pathological criteria, as in cancer or cardiovascular disease. Nevertheless, a ‘softer' form of staging is feasible since there is on the whole a good agreement between the results of different studies on the general pattern of course in schizophrenia. On the basis of long-term follow-up studies, the lifetime course of schizophrenia can be articulated into a premorbid phase (from birth to the onset of psychosis), a phase of acute or positive schizophrenic symptomatology, and a residual phase. (34) Various substages have been proposed to describe in finer detail the pre-onset period, (35,36) and theorizing about the implications of the so-called neurodevelopmental model of the aetiology of schizophrenia has led to suggestions about backdating the premorbid period to include gestation. However, for most practical and research purposes, a three-stage classification of post-onset course has been proposed: (20)
- an early deteriorating phase (the first 5–10 years);
- a middle (stabilization) phase;
- a gradual improvement phase.
This model agrees well with the empirical evidence and could be useful in the collection and summarizing of data on individual risks and prognosis.
Geographical and cultural variation
Three prospective investigations initiated by the World Health Organization ( WHO)—the International Pilot Study of Schizophrenia, (33,37) the 10-country study on Determinants of Outcome of Severe Mental Disorders, (38) and the study on Assessment and Reduction of Psychiatric Disability (39)—provide a cross-cultural database for the evaluation of the course and outcome of schizophrenia which comprises extensive initial and follow-up information on a total of 2736 patients in 16 countries, diagnosed as schizophrenic according to strict and comparable criteria. The assessment procedures and instruments used in these studies were identical or similar. Results of the WHO 10-country study (pooled data on 1070 patients in all the research sites) are considered. The following general conclusions can be drawn from this research.
- The striking variability of the course and outcome of schizophrenia has been cross-culturally confirmed. Patients with similar clinical and diagnostic characteristics at the baseline assessment develop a spectrum of outcomes ranging from stable clinical and social recovery after a single psychotic episode to chronic unremitting psychosis and severe impairment. The intensive follow-up and the relatively low cohort attrition lend credibility to the finding of a high proportion (over 30 per cent) of favourable outcomes, a proportion which is in agreement with the results of other longitudinal studies reviewed above.
- The frequencies of both relapses and remissions tend to increase over time: while at 2-year follow-up of the International Pilot Study of Schizophrenia 11 per cent of the patients had experienced two or more psychotic episodes followed by complete remission and another 18 per cent had two or more episodes followed by residual symptoms and impairments, the corresponding proportions at 5-year follow-up were 15 per cent and 33 per cent. (37)
- Regardless of the increasing relapse rate, the cumulative proportion of follow-up time during which patients have psychotic symptoms (as a percentage of the total follow-up time), tends to remain stable or decrease. At the end of the 5-year follow-up, 57 per cent of the patients had experienced a total of less than 9 months of active psychosis; only 22 per cent had been psychotic for 45 to 60 months.
- The levels of social impairment assessed at 2 years changed very little during the subsequent years of follow-up. Overall, most of the observed change in the clinical state and social functioning of patients between the 2-year follow-up and the 5-year follow-up was towards improvement rather than deterioration.
- While the percentage of patients with continuous, deteriorating illness was similar across the study sites, there were significant differences in the proportions of patients who achieved symptomatic and social recovery. In this respect, outcome was better in the developing countries. This unexpected finding of the follow-up of the International Pilot Study of Schizophrenia patients, (33,37) who had been recruited from consecutive hospital admissions with the attendant possibilities of a selective bias, was subsequently replicated by the 10-country study which had an epidemiological design and recruited only first-contact patients from delimited populations.(38) The better course and outcome in the developing country areas could not be attributed to any particular subtype of the disorder, e.g. cases of acute onset, since it applied equally to the cases of slow, insidious onset. The main outcome difference across the study areas was in the occurrence and average length of symptom-free remissions. Such remissions tended to be more frequent and to last longer in the developing countries. No single factor accounting for this difference could be identified and it is likely that complex interactions between illness and environment are involved that may include, on one hand, population differences in predisposing genes, (40) and on the other hand, cultural differences such as the relative absence of an institutionalized role of the ‘schizophrenic' in traditional societies. (41)
Whether the better outcome of schizophrenia in the developing countries is ‘transportable' following migration to other settings, remains unclear. Data on immigrants treated for first episodes of schizophrenia in the United Kingdom suggest that while Asian patients have a considerably lower relapse and readmission rate than British-born white people, African-Caribbeans show a higher rate. (42) The marked social and family structure differences between the Asian and the African-Caribbean immigrant communities suggest that the likelihood of a more benign course in the new setting may depend on the degree to which the immigrant group has retained its traditional values and intragroup cohesion.
The recent resurgence of interest in the early detection and treatment of first episodes of psychosis, driven by theoretical considerations and clinical concerns, is supported by empirical evidence suggesting that the course and outcome of the earliest stages of a schizophrenic illness may have a pathoplastic effect on its subsequent course. Specifically, the period between the first onset of psychotic symptoms and the initiation of treatment (duration of untreated psychosis) has been shown to correlate with increased time to remission and poor response to treatment. (43) Plausible clinical considerations have been proposed in support of the view that the first episode of psychosis represents a critical developmental transition that may impact the subsequent course of schizophrenia, possibly by inducing irreversible changes in the connectivity between neural networks and thus preparing the ground for chronic illness. (44) An extension of this mode of thinking is the suggestion that a behavioural or pharmacological intervention prior to the onset of psychotic symptoms could delay or prevent the onset of schizophrenia. (45,46)
None of these hypotheses has been properly tested. However, a number of studies focusing on the earliest manifestations of psychosis have highlighted features such as a presymptomatic drop in cognitive performance and social functioning, (47) early co-occurrence of ‘positive' and ‘negative' symptoms, (48) as well as a general malleability of such dysfunction in response to appropriate behavioural interventions and low-dose, time-limited pharmacological treatment. (49) This suggests that clinical research bridging the gap between statistical investigations of risk factors or antecedents of disease and individual pathways to psychotic illness may have an important role to play in understanding and, ultimately, influencing the development and course of schizophrenia.
Prognosis of specific clinical symptoms and syndromes Longitudinal studies suggest that the characteristic symptoms of schizophrenia tend to ‘breed true', i.e. only a minority of patients are eventually reclassified into other disease categories because of a significant and lasting change in the predominant symptoms. However, the proportion of cases warranting a rediagnosis seems to increase with the length of follow-up.
Depression in schizophrenia
In the WHO International Pilot Study of Schizophrenia, (33) the proportion of patients with initial schizophrenic symptomatology who developed non-schizophrenic (mostly affective) episodes in the course of time increased from 3 per cent in the first 2 years to 17 per cent at the end of the 5-year follow-up. In contrast, subsequent episodes with schizophrenic features occurred in fewer than 10 per cent of the patients with an initial diagnosis of major depression. Depression is the most common non-schizophrenic syndrome intercurrent with schizophrenia also in those patients who retain the essentially schizophrenic character of their illnesses. The proportion who develop clear-cut episodes of major depression ranges from 15 per cent during a 5-year follow-up (50) to 24 per cent during a 12-year follow-up. (20) This is a much higher period prevalence rate than in the general population, which suggests that depression is part of the clinical spectrum of schizophrenia. Based on such data, a diagnostic rubric of post-schizophrenic depression has been added to the classification of schizophrenia in ICD-10.
Schneiderian first-rank symptoms
Schneider (51) explicitly disclaimed any particular prognostic value for the subjective thought disorder phenomena, passivity experiences, and particular type of auditory hallucinations to which he attributed a ‘first-rank' significance in the differential diagnosis between schizophrenic and affective psychoses. However, the first-rank symptoms have a marked tendency to recur in subsequent psychotic episodes. Patients with one or more first-rank symptoms on the initial examination had a threefold increase in the risk of experiencing such symptoms in subsequent episodes compared to patients with no first-rank symptoms on initial examination. (38)
Prognosis of schizophrenia subtypes
The evidence that each of the ‘classic' subtypes of schizophrenia is associated with a characteristic patterns of course is generally weak but surprisingly good for some of the subtypes. Thus, consistent differences have been reported between paranoid, hebephrenic, and undifferentiated schizophrenia (diagnosed according to DSM-III) on a long-term follow-up of 19 years. (52) Paranoid schizophrenia tended to have a remittent course, and to be associated with less disability, in contrast to hebephrenia which had an insidious onset and poor long-term prognosis. Undifferentiated schizophrenia occupied an intermediate position. In the WHO International Pilot Study of Schizophrenia, (33) four alternative groupings of the ICD-9 subtypes were tested by a discriminant function for differences with regard to six course and outcome measures. Clear discrimination was achieved between simple and hebephrenic schizophrenia grouped together on the one hand and the schizoaffective subtype on the other. However, the comparison of simple and hebephrenic schizophrenia with paranoid schizophrenia resulted in a considerable degree of overlap.
Better discrimination has been claimed for groups of patients diagnosed according to the criteria of Leonhard. (53) A 5- to 13-year follow-up of 178 patients admitted with a diagnosis of schizophrenia and rediagnosed according to the Leonhardian criteria as systematic schizophrenia, atypical (unsystematic) schizophrenia, cycloid psychosis, or reactive psychosis (54) resulted in marked outcome differences on blind assessment. While only 10 per cent of the cases in the two schizophrenia groups were judged to have ‘recovered', the corresponding proportion in the cycloid and reactive psychoses group was 38 per cent. Conversely, the proportions of ‘unimproved' cases were 49 per cent and 3 per cent.
The question of whether good-prognosis remitting schizophrenia of an acute onset is a separate subtype that could also be distinguished symptomatologically was addressed in the WHO 10-country study (38) by comparing 274 patients with an initial ICD-9 diagnosis of acute schizophrenic episode and 752 patients with other schizophrenia subtypes. The group of acute cases tended to be younger and had a lower male-to-female ratio, but was no different from the rest of the schizophrenic patients with regard to initial symptomatology. Similar conclusions were reached by Vaillant (55) in a 4- to16-year follow-up of schizophrenic patients who remitted after an acute episode. This argues against acute schizophreniform illness being a discrete syndrome, outside the clinical spectrum of schizophrenia.
The course and outcome data on schizoaffective disorders seem to support their placement within the broad category of schizophrenia. A retrospective and prospective study of 150 schizoaffective patients and 95 bipolar affective patients (56) established general similarities between the two groups but the schizoaffective cases were less likely to achieve a full remission and more likely to develop a residual state (in 57 per cent compared to 24 per cent for the bipolar group). An intermediate outcome between that of schizophrenia and bipolar affective disorder is a common finding. (22)
Predictors of course and outcome
A wide range of variables have been explored as possible predictors of course and outcome in schizophrenia: sociodemographic characteristics; characteristics of the premorbid personality and pre-onset functioning; family history of psychiatric disorder; history of past psychotic episodes and treatments; substance use; characteristics of the onset; features of the initial clinical state and treatment response; variables related to brain morphology and neurocognitive functioning.
Many predictors have been replicated independently by different investigators and there is reasonable agreement on the general direction of their effects. However, the definitions of both the independent (predictor) and the dependent (outcome) variable vary across studies, and the statistical methods employed range from simple descriptive statistics (e.g. x per cent of the patients with characteristic y developed outcome z) to proper statistical models with capacity to quantify the independent contribution of individual variables to a specified outcome.
Limitations of clinical prediction
The explanatory power of any predictor in schizophrenia (in terms of accounting for a proportion of the outcome variance) is likely to vary depending on sample size, setting, homogeneity of patient groups, and measurement error, but generally tends to be limited (rarely exceeding 30 per cent of the outcome variance). This suggests that no single background or premorbid characteristics of the person, and no clinical symptom or sign among the initial manifestations of the disorder, is in a particularly strong association with its prognosis in the longer term. Similarly to the genetic epidemiology of schizophrenia, where non-shared environmental influences account for a greater amount of variance than the shared environment, person-specific emergent life events or changes in the mental state may have a similar or greater impact on outcome than the initial or premorbid predictors. Indeed, variables such as negative symptoms have been shown to gain in predictive power if they are assessed 2 or more years after the onset, or after the patients have received adequate treatment. The predictive capacity of other variables, for example mode of onset (34) or a high index of expressed emotion,(66) tends to become attenuated in the course of time. Thus, there is no fixed set of predictors of the course and outcome of schizophrenia, but rather a number of prognostic indicators which allow a judgement to be made about the probability of one or another type of course over a limited time period (usually not exceeding 5 years).
In first-episode cases, male sex, single marital status, premorbid social withdrawal, and insidious onset have been shown by a number of studies to be relatively robust predictors of a poor outcome in the short to medium term (2–5 years), while female sex, being married, having social contacts outside the home, and acute onset predict a relatively good outcome. No consistent findings have been reported for age at onset as a predictor, and the long-term follow-up studies do not lend support to the view that an early onset is associated with a poor prognosis. Similarly, a history of psychotic illness (including schizophrenia) in a first-degree relative does not predict a worse prognosis. On the contrary, in some studies (61) patients with a high familial load were found on follow-up to have a better outcome than patients with no psychotic illness among their close relatives.
A consistent finding of many studies is that the clinical symptoms in either the early or the advanced stages of schizophrenia have a very limited capacity to predict future course and outcome. An exception is the modest predictive power of any clear-cut negative symptoms appearing early in the course of the disorder, or when assessed under the conditions referred to above.
The sociocultural setting, i.e. a developing country or a developed country, was the best predictor of 2-year and 5-year outcome in the WHO studies. (33,38) Exactly what factors may be underlying these marked cultural differences in the prognosis of schizophrenia remains an unresolved issue.
The predictive status of psychophysiological or neurophysiological variables (such as the electrodermal orienting response or rapid eye movement sleep latency), or neurocognitive measurements (sustained attention, event-related potentials or the pre-pulse inhibition of the startle reflex) has not been sufficiently investigated. At least one study (59) found that brain imaging (computed tomography) evidence of cortical atrophy on first admission was a predictor of poor 2-year clinical and social outcome.
There is good agreement between different studies on the factors that help predict a relapse of psychotic symptoms after a period of stabilization or remission. Stressful life events(67) and a high expressed emotion index(66) have attracted considerable interest in this regard but are technically difficult to assess. In addition, the expressed emotion measure is only applicable to patients interacting on a daily basis with family members or other persons in the household. In many settings, such patients would be a minority. By and large, the best predictor of relapse in the short term remains the withdrawal of antipsychotic medication, usually due to non-compliance.(68) Cannabis use on a daily basis has been shown to be associated with an increased risk of relapse in a dose–response relationship. (69) Short-term predictors of a good outcome for the episodes are good initial response to neuroleptics (20) and a positive response to placebo. (70)
Summary and conclusions
Studies conducted over many decades consistently demonstrate that schizophrenia presents a broad spectrum of possible outcomes and course patterns, ranging from complete or nearly complete recovery after acute episodes of psychosis to continuous, unremitting illness leading to progressive deterioration of cognitive performance and social functioning. Between these extremes, a substantial proportion of patients show an episodic course with relapses of psychotic symptoms and partial remissions during which affective and cognitive change becomes increasingly conspicuous and may progress to gross deficits. Although no less than one-third of all patients with schizophrenia have relatively benign outcomes, in the majority the illness has a profound, lifelong impact on personal growth and development. The initial symptoms of the disorder are not strongly predictive of the pattern of course but the mode of onset (acute or insidious), the duration of illness prior to diagnosis and treatment, the presence or absence of comorbid substance use, as well as background variables such as premorbid adjustment (especially during adolescence), marital status, and availability of a social network allow a reasonable accuracy of prediction in the short to medium term (2–5 years).
One of the most striking aspects of the longitudinal course of schizophrenia is the relatively high proportion of patients who improve substantially, or remit, with ageing. What determines the ultimate outcome is far from clear but the view of schizophrenia as an invariably progressive, deteriorating disorder is certainly too limited and does not accord well with the evidence. Similarly, a model of schizophrenia as a static neurodevelopmental encephalopathy decompensating post-adolescence under the influence of a variety of stressors fits only part of the spectrum of course patterns. In a significant proportion of cases, the disorder exhibits the unmistakable features of a shift-like process with acute exacerbations and remissions which may progress to severe deterioration or come to a standstill at any stage. Whether a single underlying pathophysiology can explain the variety of clinical outcomes, or several different pathological processes are at work, remains obscure. It has been suggested that the longitudinal course of schizophrenia should be seen as an open-ended dynamic life process (71) with multiple, interacting biological and psychosocial determinants. Obviously, such issues cannot be resolved by clinical follow-up studies alone, and require a strong involvement of biological research in prospective investigations of representative samples of cases spanning the entire spectrum of course and outcomes. No such studies have been possible until recently, both because of the technical complexity of such an undertaking and because of the tendency to recruit selectively for biological investigations patients from the severe deteriorating part of the spectrum. Overcoming such limitations will be essential to the uncovering of the mechanisms driving the ‘natural history' of schizophrenia.
1. Kraepelin, E. (1919). Dementia praecox and paraphrenia. Livingstone, Edinburgh.
2. van Os, J., Wright, P., and Murray, R.M. (1997). Follow-up studies of schizophrenia I: natural history and non-psychopathological predictors of outcome. European Psychiatry 12 (Supplement 5), 327S-41S.
3. Geddes, J.R. and Kendell, R.E. (1995). Schizophrenic subjects with no history of admission to hospital. Psychological Medicine, 25, 859–68.
4. Padmavathi, R., Rajkumar, S., and Srinivasan, T.N. (1998). Schizophrenic patients who were never treated—a study in an Indian urban community. Psychological Medicine, 28, 1113–17.
5. Jonsson, S.A.T. and Jonsson, H. (1992). Outcome in untreated schizophrenia: a search for symptoms and traits with prognostic meaning in patients admitted to a mental hospital in the preneuroleptic era. Acta Psychiatrica Scandinavica, 85, 313–20.
6. Hegarty, J.D., Baldessarini, R.J., Tohen, M., Waternaux, C. and Oepen, G. (1994). One hundred years of schizophrenia: a meta-analysis of the outcome literature. American Journal of Psychiatry, 151, 1409–16.
7. Wyatt, R.J. (1991). Neuroleptics and the natural course of schizophrenia. Schizophrenia Bulletin, 17, 325–51.
8. Ödegaard, O. (1964). Patterns of discharge from Norwegian psychiatric hospitals before and after the introduction of psychotropic drugs. American Journal of Psychiatry, 120, 772–8.
9. Rennie, T.A.C. (1939). Follow-up study of five hundred patients with schizophrenia admitted to the hospital from 1913–1923. Archives of Neurology and Psychiatry, 42, 877–91.
10. Astrup, C. and Noreik, K. (1966). Functional psychoses—diagnostic and prognostic models. Charles C. Thomas, Springfield, IL.
11. Brown, G.W., Bone, M., Dalison, B., and Wing, J.K. (1966). Schizophrenia and social care. Oxford University Press, London.
12. Bleuler, M. (1972). Die schizophrenen Geistesstörungen intramuscular Lichte langjähriger Kranken- und Familiengeschichten. Thieme, Stuttgart. English translation by S.M. Clemens (1978). The schizophrenic disorders. Long-term patient and family studies. Yale University Press, New Haven, CT.
13. Ciompi, L. and Müller, C. (1976). Lebensweg und Alter der Schizophrenen. Eine katamnestische Langzeitstudie bis ins Senium. Springer, Berlin.
14. Huber, G., Gross, G., Schüttler, R., and Linz, M. (1980). Longitudinal studies of schizophrenic patients. Schizophrenia Bulletin, 6, 592–605.
15. Tsuang, M., Woolson, R., and Fleming, J. (1979). Long-term outcome of major psychoses: I. Schizophrenia and affective disorders compared with psychiatrically symptom-free surgical conditions. Archives of General Psychiatry, 36, 1295–301.
16. Harding, C., Brooks, G.W., Ashikaga, T., et al. (1987). The Vermont longitudinal study of persons with severe mental illness: I. Methodology, study sample, and overall status 32 years later. American Journal of Psychiatry, 144, 718–26.
17. Ogawa, K., Miya, M., Watarai, A., Nakazawa, M., Yuasa, S., and Utena, H. (1987). A long-term follow-up study of schizophrenia in Japan—with special reference to the course of social adjustment. British Journal of Psychiatry, 151, 758–65.
18. Shepherd, M., Watt, D., Falloon, I., et al. (1989). The natural history of schizophrenia: a five-year follow-up study of outcome and prediction in a representative sample of schizophrenics. Psychological Medicine, Monograph Supplement 15. Cambridge University Press.
19. Carone, B.J., Harrow, M., and Westermeyer, J.F. (1991). Posthospital course and outcome in schizophrenia. Archives of General Psychiatry, 48, 247–53.
20. Breier, A., Schreiber, J.L., Dyer, J., and Pickar, D. (1991). National Institute of Mental Health longitudinal study of chronic schizophrenia. Archives of General Psychiatry, 48, 239–46.
21. Johnstone, E.C. et al. (1991). Background, method, and general description of the sample. British Journal of Psychiatry, 159 (Supplement 13), pp. 7–20.
22. Marneros, A., Deister, A., and Rohde, A. (1992). Comparison of long-term outcome of schizophrenic, affective and schizoaffective disorders. British Journal of Psychiatry, 161 (Supplement 18), 44–51.
23. Thara, R., Henrietta, M., Joseph, A., Rajkumar, S., and Eaton, W.W. (1994). Ten-year course of schizophrenia—the Madras longitudinal study. Acta Psychiatrica Scandinavica, 90, 329–36.
24. Mason, P., Harrison, G., Glazebrook, C., Medley, I., Dalkin, T., and Croudace, T. (1995). Characteristics of outcome in schizophrenia at 13 years. British Journal of Psychiatry, 167, 596–603.
25. Wieselgren, I.M. and Lindström, L.H. (1996). A prospective 1–5 year outcome study in first-admitted and readmitted schizophrenic patients: relationship to heredity, premorbid adjustment, duration of disease and education level at index admission and neuroleptic treatment. Acta Psychiatrica Scandinavica, 93, 9–19.
26. Wiersma, D., Nienhuis, F.J., Sloof, C.J., and Giel, R. (1998). Natural course of schizophrenic disorders: a 15-year follow-up of a Dutch incidence cohort. Schizophrenia Bulletin, 24, 75–85.
27. Ganev, K., Onchev, G., and Ivanov, P. (1998). A 16-year follow-up study of schizophrenia and related disorders in Sofia, Bulgaria. Acta Psychiatrica Scandinavica, 98, 200–7.
28. Ciompi, L. (1980). Catamnestic long-term study on the course of life and aging of schizophrenics. Schizophrenia Bulletin, 6, 606–18.
29. Feighner, J.P., Robins, E., Guze, S.B., Woodruff, R.A., Winokur, G., and Munoz, R. (1972). Diagnostic criteria for use in psychiatric research. Archives of General Psychiatry, 26, 77–83.
30. Thara, R., Henrietta, M., Joseph, A., Rajkumar, S., and Eaton, W.W. (1994). Ten-year course of schizophrenia—the Madras longitudinal study. Acta Psychiatrica Scandinavica, 90, 329–36.
31. Strauss, J.S. and Carpenter, W.T. (1972). The prediction of outcome in schizophrenia. I. Characteristics of outcome. Archives of General Psychiatry, 27, 739–46.
32. Marengo, J. (1994). Classifying the courses of schizophrenia. Schizophrenia Bulletin, 20, 519–36.
33. World Health Organization (1979). Schizophrenia. An international follow-up study. Wiley, Chichester.
34. Ciompi, L. (1984). Is there really a schizophrenia? The long-term course of psychotic phenomena. British Journal of Psychiatry, 145, 636–40.
35. McGlashan, T.H. and Johannessen, J.O. (1996). Early detection and intervention with schizophrenia: rationale. Schizophrenia Bulletin, 22, 201–22.
36. Hambrecht, M., Häfner, H., and Löffler, W. (1994). Beginning schizophrenia observed by significant others. Social Psychiatry and Psychiatric Epidemiology, 29, 53–60.
37. Leff, J., Sartorius, N., Jablensky, A., Korten, A., and Ernberg, G. (1992). The International Pilot Study of Schizophrenia: five-year follow-up findings. Psychological Medicine, 22, 131–45.
38. Jablensky, A., Sartorius, N., Ernberg, G., et al. (1992). Schizophrenia: manifestations, incidence and course in different cultures. A World Health Organization ten-country study. Psychological Medicine Monograph Supplement 20. Cambridge University Press.
39. Jablensky, A., Schwarz, R., and Tomov, T. (1980). WHO collaborative study of impairments and disabilities associated with schizophrenic disorders. Acta Psychiatrica Scandinavica Supplementum, 285, 152–63.
40. Goldman, D., Brown, G.L., Albaugh, B., et al. (1994). D2 receptor genotype and linkage disequilibrium and function in Finnish, American Indian, and US Caucasian patients. In Genetic approaches to mental disorders (ed. E.S. Gershon and C.R. Cloninger), pp. 327–44. American Psychiatric Press, Washington, DC.
41. Waxler, N.E. (1979). Is the outcome for schizophrenia better in non-industrial societies? The case of Sri Lanka. Journal of Nervous and Mental Diseases, 167, 144–58.
42. Bhugra, D., Leff, J., Mallett, R., Der, G., Corridan, B., and Rudge, S. (1997). Incidence and outcome of schizophrenia in Whites, African-Caribbeans and Asians in London. Psychological Medicine, 27, 791–8.
43. Loebel, A.D., Lieberman, J.A., Alvir, J.M.J., Geisler, S.H., Szymanski, S.R., and Mayerhoff, D.I. (1992). Duration of psychosis and outcome in first-episode schizophrenia. American Journal of Psychiatry, 149, 1183–8.
44. Hoffman, R.E. and McGlashan, T.H. (1993). Parallel distributed processing and the emergence of schizophrenic symptoms. Schizophrenia Bulletin, 19, 119–40.
45. Falloon, I.R.H. (1992). Early intervention for first episodes of schizophrenia: a preliminary exploration. Psychiatry, 55, 4–15.
46. McGorry, P.D., Edwards, J., Mihalopoulos, C., Harrigan, S.M., and Jackson, H.J. (1996). EPPIC: an evolving system of early detection and optimal management. Schizophrenia Bulletin, 22, 305–6.
47. Goldberg, T.E., Hyde, T.M., Kleinman, J.E., and Weinberger, D.R. (1993). Course of schizophrenia: neuropsychological evidence for a static encephalopathy. Schizophrenia Bulletin, 19, 797–804.
48. Gupta, S., Andreasen, N.C., Arndt, S., Flaum, M., Hubbard, W.C., and Ziebell, S. (1997). The Iowa longitudinal study of recent onset psychosis: one-year follow-up of first episode patients. Schizophrenia Research, 23, 1–13.
49. Szymanski, S.R., Cannon, T.D., Gallacher, F., Erwin, R.J., and Gur, R.E. (1996). Course and treatment response in first-episode and chronic schizophrenia. American Journal of Psychiatry, 153, 519–25.
50. Sheldrick, C., Jablensky, A., Sartorius, N., and Shepherd, M. (1977). Schizophrenia succeeded by affective illness: catamnestic study and statistical enquiry. Psychological Medicine, 7, 619–24.
51. Schneider, K. (1959). Clinical psychopathology (transl. M.W. Hamilton). Grune and Stratton, New York.
52. Fenton, W.S. and McGlashan, T.H. (1991). Natural history of schizophrenia subtypes. I. Longitudinal study of paranoid, hebephrenic, and undifferentiated schizophrenia. Archives of General Psychiatry, 48, 969–77.
53. Leonhard, K. (1979). The classification of endogenous psychoses (5th edn) (transl. R. Berman). Halsted Press, New York.
54. Stephens, J.H. and Astrup, C. (1963). Prognosis in ‘process' and ‘non-process' schizophrenia. American Journal of Psychiatry, 119, 945–53.
55. Vaillant, G.E. (1978). A 10-year follow up of remitting schizophrenics. Schizophrenia Bulletin, 4, 78–84.
56. Angst, J., Felder, W., and Lohmeyer, B. (1980). Course of schizoaffective psychoses: results of a follow up study. Schizophrenia Bulletin, 6, 579–85.
57. Harder, D.W., Strauss, J.S., Greenwald, D.F., et al. (1990). Predictors of outcome among adult psychiatric first-admissions. Journal of Clinical Psychology, 46, 120–8.
58. Jørgensen, P.M., Mortensen, P.B., and Machon, R.A. (1991). Hospitalization patterns in schizophrenia. A 13-year follow-up. Schizophrenia Research, 4, 1–9.
59. Vita, A. et al. (1991). CT scan abnormalities and outcome of chronic schizophrenia. American Journal of Psychiatry, 148, 1577–9.
60. Eaton W.W. et al. (1992). Long-term course of hospitalization for schizophrenia: Part I. Risk for rehospitalization. Schizophrenia Bulletin, 18, 217–27.
61. Johnstone, E.C., Frith, C.D., Lang, F.H., and Owens, D.G.C. (1995). Determinants of the extremes of outcome in schizophrenia. British Journal of Psychiatry, 167, 604–9.
62. Harrison, G., Croudace, T., Mason, P., Glazebrook, C., and Medley, I. (1996). Predicting the long-term outcome of schizophrenia. Psychological Medicine, 26, 697–705.
63. Goldman, M., Tandon, R., DeQuardo, J.R., et al. (1996). Biological predictors of 1-year outcome in schizophrenia in males and females. Schizophrenia Research, 21, 65–73.
64. Ho, B.C., Nopoulos, P., Flaum, M., Arndt, S., and Andreasen, N. (1988). Two-year outcome in first-episode schizophrenia: predictive value of symptoms for quality of life. American Journal of Psychiatry, 155, 1196–201.
65. Castle, D.J., Wessely, S., van Os, J., and Murray, R.M. (1998). Psychosis in the inner city: the Camberwell first episode study. Psychology Press, Hove.
66. Leff, J.P., Wig, N.N., Ghosh, A., et al. (1990). Relatives' expressed emotion and the course of schizophrenia in Chandigarh: a two-year follow-up of a first-contact sample. British Journal of Psychiatry, 156, 351–6.
67. Brown, G.W. and Birley, J.L.T. (1968). Crises and life changes and the onset of schizophrenia. Journal of Health and Social Behaviour, 9, 203–14.
68. Dencker, S.J., Malm, U., and Lepp, M. (1986). Schizophrenic relapse after drug withdrawal is predictable. Acta Psychiatrica Scandinavica, 73, 181–5.
69. Linszen, D.H., Dingemans, P.M., and Lenior, M.E. (1994). Cannabis abuse and the course of recent-onset schizophrenic disorders. Archives of General Psychiatry, 51, 273–9.
70. Johnstone, E.C., Macmillan, J.F., Frith, C.D., et al. (1990). Further investigation of the predictors of outcome following first schizophrenic episodes. British Journal of Psychiatry, 157, 182–9.
71. Harding, C.M., Zubin, J., and Strauss, J.S. (1992). Chronicity in schizophrenia revisited. British Journal of Psychiatry, 161 (Supplement 18), 27–37.