60464ournal ofNeurology, Neurosurgery, and Psychiatry 1996;61:604-613 J Neurol Neurosurg Psychiatry: first published as 10.1136/jnnp.60.6.604 on 1 June 1996. Downloaded from NEUROEPIDEMIOLOGY Schizophrenia Mary Cannon, Peter Jones Psychiatric epidemiology is the study of the studies of time trends and outcome in schizo- distribution and causes of mental disorder in phrenia. Also, the commitment to a phenotype the population. Over the past 30 years based on symptoms in adult life may have progress in psychiatric epidemiology has been ignored an important developmental or life- slower than in other areas-for example, car- long dimension to the disorder. diovascular disease or cancer epidemiology, because of certain methodological problems SCHIZOPHRENIA AS A "RARE DISEASE" that are just now being overcome. In particular The low incidence (10-40 cases per 100 000 the epidemiology of schizophrenia has suffered per year) and relatively low lifetime prevalence from two major difficulties: uncertainty about (0 5%-1 %) of schizophrenia in the population how to define a case, and the relative rarity of have led to a reliance on case-control study schizophrenia in the population. designs in research. Chronic patients recruited from hospital wards are compared with volun- teer controls from the community and conse- Special methodological problems in quent problems of bias and confounding have schizophrenia epidemiology often led to unreplicated and contradictory WHAT IS A CASE findings. It seemed for a time that schizophre- The lack of physical signs or laboratory tests nia, already known as the "graveyard of neu- means that a diagnosis of schizophrenia is ropathologists",7 would also prove to be the based on evaluation of patients' self reported, undoing of epidemiologists. Over the past two subjective experiences. Until the late 1960s, decades, however, the application of advances there was a relatively loose definition of schizo- in case-control methodology to psychiatric phrenia and a good deal of diagnostic latitude epidemiology has led to more robust results.8 was allowed to individual psychiatrists. How- Also, cohort designs, for long thought too http://jnnp.bmj.com/ ever, a report published in 1972,' showed that costly and time consuming to use in schizo- the diagnostic habits of psychiatrists in the phrenia research, have made a "comeback". United States and United Kingdom differed to For the purposes of this review schizophre- an unacceptable degree, and led to the intro- nia refers to a recent, operational definition duction of strict operational diagnostic criteria such as CATEGO,9 ICD-10, DSM-III, DSM- for schizophrenia. The Diagnostic and III-R, or DSM-IV. Robust findings are those Statistical Manual DSM-III,2 published in from studies with an epidemiological design- 1980, reduced the reliance on symptoms alone that is, population based studies with well on September 29, 2021 by guest. Protected copyright. by incorporating an element of chronicity-six defined samples and appropriate controls. We months prior duration of symptoms and an aim to show how the findings from robust epi- upper age limit of 45 years for a first diagnosis demiological research can help to unravel the of schizophrenia. Since 1980 there have been complex aetiology of schizophrenia. two more restrictive revisions of this diagnostic system, DSM-III-R3 and DSM-IV,4 and the International Classification of Diseases, Geography Department of (ICD), the diagnostic system favoured by In 1978, a large multicentre study of schizo- Psychological European psychiatrists, has also undergone a phrenia was initiated by WHO the 10 country Medicine, Institute of similar "narrowing" process with the change study (also known as the Determinants of Psychiatry, Denmark Hill, London SE5 8AF, from ICD-9 to ICD-10.5 Outcome of Severe Mental Disorders Study), UK But has this process gone too far? Although to provide information about the incidence, M Cannon the restrictive diagnostic criteria have certainly course, and outcome of schizophrenia in dif- Department of increased the reliability of the diagnosis, they ferent cultures.10 Two case definitions of schiz- Psychiatry, University have not improved the validity of "DSM-IV" ophrenia were used: a clinical of Nottingham, broad, Duncan Macmillan or "ICD-10 schizophrenia".6 Recent family definition comprising ICD-9 schizophrenia House, Porchester studies indicate that a broader concept of and paranoid psychoses, and a narrow, restric- Road, Nottingham schizophrenia actually fits genetic models tive definition including only cases classified as NG3 6AA, UK P Jones more readily than a restrictive definition. In "nuclear" schizophrenia using the CATEGO 9 Correspondence to: addition, the frequent changes in diagnostic computer programme Figures 1A and B Dr Peter Jones. criteria have affected the comparability of show the incidence rates for both definitions. Cannon, _rones 605 J Neurol Neurosurg Psychiatry: first published as 10.1136/jnnp.60.6.604 on 1 June 1996. Downloaded from Figure 1 (A) Incidence of 20 of non-affective psychosis was 0.7%.12 This CATEGO S narrow schiz- FA ophrenia with 95% confi- Cu discrepancy may be related to issues of sam- m pling, interview methodology, or actual dence intervals from eight 15 H centres in the WHO change over time and remains to be clarified Determinants of Outcome 0C) Study. (B) Incidence of 0 by future reports. CATEGO S, P, and 0 10i A major prevalence study of psychiatric broad schizophrenia with 0 morbidity carried out in the United Kingdom ._0a) 95% confidence limits from -0 between and found a eight centres in the WHO 5 April September 1993,13 determinants ofoutcome .5 prevalence rate of 0 4% for functional psy- study. Data from J7ablensky C- chosis among persons aged 16-64 living in pri- et. al.'0 vate households. This rate is even lower than 0 L that found in the national comorbidity survey 6oB but as 37% of those who screened positive for Cu B Cu co 5 psychosis refused a second diagnostic inter- 0 60_ view, the true prevalence may be higher.'4 At 0 4 0 present, it seems that the lifetime prevalence of 0 0 330 _ schizophrenia in the western world lies some- C- where between 0 4% and 1-4%, and may have C.) C 220 TI decreased over the past decade. T .5'a ,o- + i C "INCIDENCE" DATA v,\ Prevalence of schizophrenia can be influenced NX by factors such as change in treatment, change It 00A in mortality rate, and changes in the age and sex structure of the population, and therefore, examination of incidence rates may be more informative. Ideally, such studies should be based on community incidence samples but, as this is difficult to achieve, case registers and There is little variation between centres for hospital admission data are commonly used. narrowly defined schizophrenia with rates Eagles and Whalley'5 first reported a decline ranging only between 7 and 14/100 000, (fig in the diagnosis of schizophrenia among first 1A). Although the parsimonious conclusion admissions in Scotland between 1969 and would be that the rates for narrowly defined 1978. There have since been 14 papers exam- schizophrenia are the same, the confidence ining this issue in England,'6-'0 Scotland,"'3 intervals for these estimates were wide and Denmark,'425 New Zealand,'6 Canada,'7 there may not have been sufficient statistical Ireland,'8 the United States,'9 and the power to detect differences. Netherlands.30 Those based on national statis- The incidence rates for broadly defined tics have found a large (40%-50%) decline in schizophrenia do seem to vary between coun- first admission rates for schizophrenia during tries 16 to 42/100 000 per and 17 18 25 26 (range year), the 1970s and the 1980s. Findings http://jnnp.bmj.com/ the rates for centres in the developing world based on case register data have been less con- are about twice as high as those in the devel- sistent and indeed, two such studies, from oped world (fig 1B). Further studies of schizo- Camberwell"7 and Salford'8 in the United phrenia in the developing world are needed to Kingdom, actually found a slight increase in replicate and further investigate these interest- the incidence of schizophrenia during the ing findings, and the problem of mental health same period. Others have reported a decrease in the developing world can no longer remain a in first admission rates for schizophrenia low priority for funding agencies and govern- among female patients only." 2830 on September 29, 2021 by guest. Protected copyright. ments. On the whole, however, the variation The major question is whether the decrease in incidence rates for schizophrenia world noted in first admission rates corresponds to wide is very small compared with illnesses an actual decrease in the incidence of the con- such as ischaemic heart disease or cancer, dition. Many factors influence this apparent which are known to have major environmental "administrative decline" in schizophrenia: (1) risk factors. The introduction of more restrictive diagnos- tic criteria for schizophrenia may cause a shift to diagnoses such as "borderline states"24 or Time trends affective psychosis.'930 (2) The move to com- PREVALENCE DATA munity care over the past two decades could Two major prevalence studies of psychiatric have affected hospital admission rates.'7 (3) illness have been carried out in the United Clinicians have become more reluctant to States, which indicate a decrease in prevalence make a diagnosis of schizophrenia on the first of schizophrenia over one decade. The hospital admission,'4