Psychological Medicine, 2003, 33, f 2003 Cambridge University Press DOI: /S Printed in the United Kingdom

Transcription

1 Psychological Medicine, 2003, 33, f 2003 Cambridge University Press DOI: /S Printed in the United Kingdom Association between psychotic disorder and urban place of birth is not mediated by obstetric complications or childhood socio-economic position: a cohort study G. HARRISON, D. FOUSKAKIS, F. RASMUSSEN, 1 P. TYNELIUS, A. SIPOS AND D. GUNNELL From the Division of Psychiatry and Department of Social Medicine, University of Bristol; and Department of Epidemiology, Institute of Environmental Medicine, Karolinska Institute, Stockholm, Sweden ABSTRACT Background. Although urban place of birth has been identified as a risk factor for schizophrenia, the extent to which this association is mediated by socially patterned risk factors such as obstetric complications and childhood socio-economic position is unclear. The diagnostic specificity of the association within the clinical psychotic syndromes is also unclear. Method. A population cohort of males and females, born in Sweden between 1973 and 1980 and with linked birth and socio-economic data was followed up from age 16 for up to 9. 8 years. Hospitalized cases of schizophrenia and other non-affective psychosis were identified from the Swedish Inpatient Discharge Register. We examined associations of these disorders with a three-level measure of urbanicity of birthplace before and after controlling for measures of foetal nutrition, obstetric complications and level of maternal education. Results. Urban compared to rural birthplace was associated both with increased risk of adult onset schizophrenia (hazard ratio 1. 34, CI ) and other non-affective psychoses (hazard ratio 1. 63, CI ). None of these associations was greatly affected by adjustment for obstetric complications or maternal educational level. In the group of other non-affective psychoses urban rural differences in disease risk were strongest among those born in the winter months. Conclusion. Urbanization of birthplace is associated with increased risk of non-affective psychosis but this is not confined to narrowly defined cases. The magnitude of the association in Sweden is lower than that reported in other studies. Causal factors underlying this association appear to operate independently of risks associated with obstetric complications and parental educational status. INTRODUCTION Population-based studies indicate that people born in urban areas are at increased risk of schizophrenia and that the increased risk is independent of family history of psychosis. Studies in Denmark, a small homogeneous country with 1 Address for correspondence: Dr Finn Rasmussen, Division of Epidemiology, Karolinska University Hospital, Norrbacka, SE Stockholm, Sweden. 723 a population of 5. 3 million, report relative risks (RR) of between and (Mortensen et al. 1999; Pedersen & Mortensen, 2001a, b). The strength of associations seen in studies carried out in other social and cultural settings are generally lower. In the Netherlands, Marcelis et al. (1998) reported a RR of and in the USA (Torrey et al. 1997), analyses based upon 1980 Census data reported a RR of In the Untied Kingdom, Takei et al. (1995) showed a

2 724 G. Harrison and others marginally increased risk of schizophrenia (odds ratio (OR) 1. 12, 95% CI ) for persons born in city areas compared with non-city areas, although they used non-schizophrenic psychiatric admissions as their control population. A Finnish study of four birth cohorts spanning the period (Haukka et al. 2001) found that urban place of birth marginally increased the risk of schizophrenia in all cohorts, except for those born after 1965, when incidence was increased by about 25% in relation to urban residency. In Sweden, Lewis et al. (1992) reported a relative risk of (CI ) for those brought up in cities compared with rural areas, but their study was carried out on males only. The biological and social mechanisms that could mediate this association are unclear. The association reported by Lewis et al. (1992) remained after adjusting for several possible confounding factors known to be connected with urban residence, including family history, cannabis use and parental divorce. Machon et al. (1983) and Takei et al. (1995), report an interaction with season of birth (place of residence associations were stronger among cases who were born in winter) but this was not replicated by Mortensen et al. (1999) or by Pedersen et al. (2001 a, b). Obstetric complications are known to be socially patterned (Jonas et al. 1992; Wilcox et al. 1995; Gudmundsson et al. 1997) but the potential role of obstetric complications and childhood socio-economic position as possible mediating variables in relatively unexplored. Eaton et al. (2000) investigated obstetric complications (but not socio-economic position) in a record-linkage study involving a small sample of early onset cases of schizophrenia, and cases of other non-affective and affective psychoses. After controlling for maternal parity, the OR for those born in Copenhagen compared with rural areas was (95% CI ). The association was little different after controlling for other obstetric factors including birth weight and mode of delivery. In addition, the association with non-affective psychosis was stronger than that with schizophrenia (OR 5. 72, 95% CI ) suggesting that urbanization of birthplace may not be specific to schizophrenia. To investigate these questions further, we assembled a cohort based upon linked birth, socio-economic and psychiatric admission data from Sweden. We examined the association between urbanization of birthplace and the risk of adult onset (>16 years) schizophrenia and other non-affective psychoses. We then investigated the distribution of the potentially confounding risk factors of obstetric complications and childhood socio-economic position across a three-level measure of urbanicity. Finally, we examined the extent to which urban rural birthplace differences in schizophrenia are due to the social patterning of obstetric complications and childhood socio-economic position in urban areas. METHOD Population The risk set was based upon males and females, born in Sweden between 1973 and 1980, and still alive and resident in Sweden at the age of 16. We excluded all subjects diagnosed with schizophrenia or other non-affective psychosis before the age of 16. Overall, (6. 9%) subjects were excluded from the main analysis on account of missing data for one or more of the factors investigated. Our main analyses were therefore based on subjects. Information on the study sample was obtained from a linkage between the Swedish Medical Birth Registry (MBR), the Population and Housing Censuses of 1970 and 1990, and the Swedish Inpatient Discharge Register (up to December 1997). Our analysis was based on hospital admissions between 1989 and Admissions were coded using the International Classification of Diseases (ICD-9) (WHO, 1978) was used in Sweden from 1987 to 1996 and ICD-10 (WHO, 1993) has been used since We examined associations with all non-affective, nondrug related psychoses (ICD-9 295, ; ICD-10 F20 29), categorizing these disorders into two groups: (1) people admitted with a diagnosis of schizophrenia (ICD-9 295; ICD-10 F20) at any point in the period of follow-up; (2) all remaining non-schizophrenic non-affective psychoses (ICD , ICD-10 F21 29). Many subjects had more than one hospital admission, with different diagnoses. We therefore carried out an additional analysis based upon a more restrictive definition for schizophrenia: if

3 Association between urban birth and schizophrenia 725 subjects were admitted on more than one occasion, we used the latest diagnosis as we assumed this to be the most accurate. Using this method of case definition, the number of cases of schizophrenia fell from 363 to 303. Subjects were followed up for a mean of 5. 1 years (range: 1 day to 9. 8 years) after the age of 16. Risk factors Place of birth Place of birth was coded using a nine category classification: (1) main cities (Stockholm, Gothenburg, and Malmo); (2) suburbs to main cities; (3) large cities (population ); (4) medium size cities (population ); (5) industrial municipalities (where >40% of the population is employed in the industrial sector regardless of city size); (6) other larger municipalities (population ); (7) municipalities in rural areas; (8) sparsely populated areas; (9) other smaller municipalities (population <15 000). We classified place of birth according to the degree of urbanization based upon this nine-point classification. To reduce the number of categories, we grouped these nine categories (on the basis of a division decided a priori) into a three level measure (coded 1 to 3 in the following order): main cities and their suburbs (categories 1 and 2); large and medium size cities (categories 3, 4, 5 and 6) and rural areas (categories 7, 8 and 9). Obstetric complications and measures of foetal nutrition We extracted data on the following indicators: birth weight; birth length; gestational age; season of birth; age of mother (at childbirth); Caesarean section; APGAR score at 1 min; and parity. Childhood socio-economic position We used maternal education as an indicator of childhood socio-economic position. We had information on paternal education, but as this was missing for 10% of subjects this variable was not included in our main analysis. Education was categorized into four groups: <9 years; 9 10 years; full secondary education; higher education. Statistical methods All analyses were carried out in STATA (Stata- Corp, 2001). We used ANOVA and chi-square statistics as appropriate to test for association of potentially confounding factors with urbanicity. We used Cox s proportional hazards models to assess the influence of degree of urbanization on the risk of schizophrenia and other non-affective psychosis. We tested the validity of the proportional hazards assumption graphically with log log plots. We also compared hazard ratios for incidence of schizophrenia for each half of the time period and found these to be comparable. All analyses were controlled for age (by using it as the time axis in the Cox model) and for sex. We assessed the effects on hazard ratios of controlling for our specified obstetric complications, entering the following obstetric explanatory variables separately: birth weight (as five categories: <2500 g, g, g, g, and >4000 g); birth length (as a continuous term); gestational age (f36, and o41 weeks); season of birth (spring, summer, autumn, winter (defined as December, January and February)); APGAR score at 1 min (f6 or>7); maternal parity (1, 2 or o3); Caesarean section birth; maternal age (f20, 20 24, 25 29, 30 34, o35). Lastly, we assessed the effect of controlling for maternal education (a four level variable, full model). Tests for interaction were based on likelihood ratio tests comparing models with and without the relevant explanatory variables. Subjects were censored at the time of first admission for non-affective, non-drug-related psychotic illness, death or emigration. RESULTS Three hundred and sixty-three (0. 05%) subjects were admitted to hospital with a diagnosis of schizophrenia at least once in the period of follow-up and 590 (0. 08%) were admitted with a diagnosis of non-affective, non-schizophrenic psychosis. The time between the age of 16 and hospital admission for schizophrenia ranged from to years and for non-affective, non-schizophrenic psychoses from to years. The annual incidence rate for schizophrenia was per 1000 person years and for non-schizophrenic, non-affective psychosis was per 1000 person years.

4 726 G. Harrison and others Table 1. Characteristics (N (%) unless stated otherwise) of those who developed and those who did not develop non-affective psychoses and schizophrenia Schizophrenia (N=363) Non-affective nonschizophrenic psychoses (N=590) Non-cases (N= ) N (%) N (%) N (%) Birth weight (g), mean (S.D.) 3502 (554) 3472 (551) 3511 (523) Ponderal Index (kg/m 3 ), mean (S.D.) 27. 0(2. 5) 27. 1(2. 5) 27. 1(2. 4) Birth length (cm), mean (S.D.) 50. 5(2. 3) 50. 4(2. 5) 50. 5(2. 3) Gestational age (weeks) f36 20 (5. 5) 29 (4. 9) (4. 0) (82. 9) 481 (81. 5) (83. 1) >41 42 (11. 6) 80 (13. 6) (12. 9) Season of birth Winter 95 (26. 2) 148 (25. 1) (23. 5) Spring 107 (29. 5) 179 (30. 3) (28. 6) Summer 84 (23. 1) 147 (24. 9) (24. 9) Autumn 77 (21. 2) 116 (19. 7) (22. 9) Municipality of place of birth 1 Main cities 67 (18. 5) 96 (16. 3) (13. 5) 2 Suburbs to the municipalities above 55 (15. 2) 99 (16. 8) (14. 4) 3 Large cities 110 (30. 3) 162 (27. 5) (28. 0) 4 Cities of middle size 43 (11. 9) 79 (13. 4) (16. 0) 5 Industrial municipalities 38 (10. 5) 64 (10. 9) (9. 7) 6 Other larger municipalities 16 (4. 4) 45 (7. 6) (7. 4) 7 Municipalities in rural areas 12 (3. 3) 13 (2. 2) (4. 1) 8 Sparsely-populated areas 9 (2. 5) 18 (3. 1) (2. 4) 9 Other smaller municipalities 13 (3. 6) 14 (2. 4) (4. 4) Age of mother at birth, years f20 29 (8. 0) 46 (7. 8) (6. 1) (27. 0) 172 (29. 2) (29. 8) (34. 7) 206 (34. 9) (37. 9) (22. 1) 125 (21. 2) (19. 8) >35 30 (8. 3) 41 (7. 0) (6. 4) APGAR at 1 min f6 17 (4. 7) 28 (4. 8) (3. 8) o7 346 (95. 3) 562 (95. 2) (96. 2) Parity (46. 8) 288 (48. 8) (43. 8) (35. 0) 211 (35. 8) (37. 7) o3 66 (18. 2) 91 (15. 4) (18. 5) Caesarean section Yes 20 (5. 5) 44 (7. 5) (8. 6) No 343 (94. 5) 546 (92. 5) (91. 4) Sex Male 221 (60. 9) 294 (49. 8) (51. 3) Female 142 (39. 1) 296 (50. 2) (48. 7) Mother s education <9 years 65 (17. 9) 84 (14. 2) (13. 0) 9 10 years 65 (17. 9) 107 (18. 1) (20. 3) High/comp 125 (34. 4) 240 (40. 7) (40. 5) Higher 108 (29. 8) 159 (27. 0) (26. 3) Those with missing data tended to be of lower birth weight (101 g lighter); shorter at birth (0. 4 cm shorter), were more likely to be born before 36 weeks gestation (7. 69% v %); were more likely to have been born in the winter (December February: % v %); to be male (52. 02% v %); and to have a lower level of maternal education (<9 years, % v %). All these differences were statistically significant because of the large sample size. There was no evidence of interaction between sex and place of birth with respect to their effects on schizophrenia (P=0. 55) or

5 Association between urban birth and schizophrenia 727 Table 2. Age and sex adjusted hazard ratios of schizophrenia and non-affective non-schizophrenic psychosis for all nine categories of municipality of birth place Hazard ratios (95% CI) Schizophrenia (N=363) Non-affective, non-schizophrenic psychoses (N=590) Municipality Cases (N ) Person-years Age, sex-adjusted Cases (N ) Person-years Age, sex-adjusted 1 Main cities (0. 90, 2. 95) (1. 24, 3. 82) 2 Suburbs to the (0. 66, 2. 22) (1. 17, 3. 60) municipalities above 3 Large cities (0. 71, 2. 24) (1. 01, 3. 00) 4 Cities of middle size (0. 47, 1. 62) (0. 85, 2. 64) 5 Industrial (0. 67, 2. 37) (1. 12, 3. 55) municipalities 6 Other larger (0. 34, 1. 48) (1. 02, 3. 40) municipalities 7 Municipalities in rural (0. 46, 2. 19) (0. 47, 2. 13) areas 8 Sparsely-populated (0. 54, 2. 93) (1. 15, 4. 66) areas 9 Other smaller municipalities non-schizophrenic, non-affective psychosis (P= 0. 54). Our analyses were therefore based on the dataset combining male and female subjects. Characteristics of subjects with schizophrenia and other non-affective psychoses Table 1 shows the distribution of the variables used in the analysis for cases of schizophrenia, non-affective non-schizophrenic psychoses and for our reference population of non-cases (controls). Of the non-cases, 10. 9% were born in rural areas (categories 7, 8, 9). The figure was slightly lower for schizophrenia (9. 4%), and lower again (7. 7%) for non-affective, nonschizophrenic psychoses. Of our control population 13. 5% were born in the main cities compared to 18. 5% in the schizophrenia group and 16. 3% of the cases of other non-affective psychoses. Those with schizophrenia were more often born in the winter months and had a larger proportion of mothers of low educational status. Age and sex adjusted models for association between psychotic disorders and urbanicity Table 2 shows age and sex adjusted hazard ratios for schizophrenia and non-affective psychoses for all nine categories of municipality. Subjects born in main cities had the highest risk of both schizophrenia and non-affective nonschizophrenic psychosis. There were fewer cases in categories 7, 8 and 9, and the hazard ratios for sparsely populated areas were high compared with the other two rural municipalities. Urban patterning of obstetric complications and maternal education Table 3 shows the associations between our potentially confounding factors and the aggregated 3-level measure of urbanization. Babies born in main cities and suburbs were more likely to have lower birth weight, shorter birth length, lower APGAR scores and increased risk of birth by Caesarean section. Mother s education was clearly patterned with better educated mothers tending to live in urban areas. Multivariable models for associations between psychotic disorders and urbanicity In Table 4 we report analyses based upon the three-level measure of the degree of urbanization of place of birth. There are associations between place of birth and schizophrenia and other non-affective non-schizophrenic psychoses, but the effect size is larger for other non-affective psychoses. Controlling first for obstetric complications, and subsequently for maternal education had little effect on the strength of these associations. We found no evidence that the effects of place of birth on risk schizophrenia differed with respect to season of birth (P interaction =0. 61). However, for other

6 728 G. Harrison and others Table 3. Associations between urbanicity and potentially confounding variables Main cities and suburbs (N= ) Large cities and industry (N= ) Rural areas (N=75 891) P* Birth weight (g), mean (S.D.) 3490 (520) 3514 (525) 3548 (522) < Ponderal Index (kg/m 3 ), mean (S.D.) (2. 44) (2. 40) (2. 43) < Birth length (cm), mean (S.D.) 50. 4(2. 3) 50. 5(2. 3) 50. 7(2. 2) < Gestational age (weeks), mean (S.D.) (1. 75) (1. 76) (1. 73) Age of mother at birth (years), mean (S.D.) 27. 6(4. 7) 26. 9(4. 9) 26. 9(5. 1) < APGAR at 1 min, f (4. 7%) (3. 5%) 2592 (3. 4%) < Parity, o (15. 3%) (19. 1%) (23. 4%) < Caesarean section, Yes (9. 0%) (8. 6%) 6212 (8. 2%) < Mother s education, <9 years (9. 0%) (14. 1%) (17. 2%) < * ANOVA for continuous variables, chi-square tests for categorical variables. Table 4. Age and sex adjusted, and fully adjusted hazard ratios for schizophrenia and non-affective non-schizophrenic psychosis in relation to urbanicity of place of birth Hazard ratios (95% CI) Schizophrenia (N=363) Non-affective, non-schizophrenic psychoses (N=590) Urbanicity of birth place Age, sex-adjusted Age, sex and birth related exposures adjusted* Fully adjusted# Age, sex-adjusted Age, sex and birth related exposures adjusted* Fully adjusted# Main cities and (0. 91, 1. 96) (0. 90, 1. 93) (0. 89, 1. 93) (1. 18, 2. 26) (1. 13, 2. 17) (1. 13, 2. 18) suburbs Large cities and (0. 71, 1. 49) (0. 72, 1. 49) (0. 72, 1. 49) (0. 98, 1. 82) (0. 97, 1. 80) (0. 97, 1. 80) industry Rural areas P-trend P-difference * Controlling for age, birth weight, birth length, gestational age, season of birth, age of mother, APGAR score at 1 min, maternal parity, Caesarean section and sex. # As above but controlling also for maternal education. non-affective psychoses we found some evidence for such an interaction (P interaction =0. 046): the association with urban birth was stronger in those born in winter (HR for birth in main cities, ( )) compared with summer (HR for birth in main cities, ( )). Schizophrenia compared to other non-affective psychoses To investigate the effect of case definition, we repeated the main analysis for place of birth, using the more restricted definition for schizophrenia based described above. This eliminated those cases where the diagnosis of schizophrenia was changed to non-affective psychosis, or some other category, at later admissions. There was a reduction in the magnitude of the effect size for these cases with hazard ratio for main cities ( ) and large cities (0. 70, compared with our reference category of rural areas). DISCUSSION Main findings These data confirm previous studies reporting an association between urban place of birth and risk of schizophrenia. We show however that the association is not specific to narrowly defined schizophrenia and, indeed, may be stronger for other non-affective, non-schizophrenic psychoses. There is also an interaction with season of birth in non-schizophrenic nonaffective psychoses, the greatest risk being in winter born. Our analyses show significant associations between urban birthplace and indicators of

7 Association between urban birth and schizophrenia 729 foetal under-nutrition, such as birth weight and length. In addition, high APGAR score, maternal parity and incidence of caesarean sections also varied in relation to place of birth. Although previous studies have reported associations between these indicators and increased lifetime risk of adult onset schizophrenia (e.g. Wahlbeck et al. 2001) our data show that the risk factors associated with urbanization operate independently. Further, the association with urbanization appears not to be mediated through childhood socio-economic position despite reports of an association between higher parental social class and schizophrenia (Jones et al. 1994; Makikyro et al. 1997). These findings apply to non-affective psychoses and to schizophrenia, whether based upon broad or narrowly defined cases. Study strengths and limitations This study has a number of strengths, including the prospective cohort design and the independent nature of the assessments carried out for obstetric complications, place of birth and diagnosis. There are several limitations however. First, our period of follow-up was short and our findings apply only to cases of early age of onset. We may therefore have overestimated, or underestimated, the effects of urbanization on the incidence of schizophrenia across the life course. These findings require replication in samples containing individuals in the entire age of risk. Secondly, case detection was based on hospital admitted cases where diagnoses were recorded in an administrative database. Urbanization may be associated with factors connected with the risk of admission, rather than risk of disease, and our observed association could be accounted for by differences in the availability of psychiatric services in urban and rural areas. Mortensen et al. (1999) discount this possibility on the basis of small distances in Denmark, but we cannot eliminate this is an explanatory factor for our findings. Thirdly, it is also possible that clinicians in rural areas may be more conservative with regard to the diagnosis of schizophrenia than those in urban areas. This point would have some validity if our effect was confined to schizophrenia but in our study urban birthplace was associated with the broad range of non-affective psychoses. It is possible that there could be systematic differences in clinicians diagnostic habits within the board group of non-affective psychoses, but we think it is unlikely that there would be significant differences at the boundaries of psychosis itself. Fourthly, missing data might have introduced selection bias. We investigated this further by assessing the effect of urban place of birth using the complete dataset adjusted only for age and sex. Hazard ratios were only marginally different in the full dataset (HR 1. 26, CI for schizophrenia and HR 1. 73, CI for non-affective psychoses for main cities/suburbs compared to rural areas). Therefore, restricting the analyses to subjects with complete data probably did not have bias our main results. Controlling for paternal education in the more limited dataset (due to missing values) had no effect on the strength of the associations, so we think it highly unlikely that confounding due to socio-economic status lies behind the urban effect. Finally, we do not have data for family history of schizophrenia, although Mortensen et al. (1999) found that family history of schizophrenia did not explain or affect the urban rural differences observed in their study. Findings compared with other studies of urban rural differences The strength of the association between urban place of birth and non-affective psychoses in this Swedish population is similar to that reported by Lewis et al. (1992) for schizophrenia among male Swedish conscripts. For narrowly-defined schizophrenia, the hazard ratios were somewhat smaller in our study. If our findings are considered with Lewis et al. (1992), Torrey et al. (1997) and Marcelis et al. (1998) the range of values lies between and The findings of Haukka et al. (2001), in Finland, also lie at the lower end of this range. Although Haukka et al. did not report a point estimate for the relative risk for urban versus rural birth, their Poisson regression model showed an increased incidence of about 25% (their Fig. 1) for the cohort ascertained in Furthermore, they found that two previous birth cohorts ( and ) showed only marginal increases for urban born, and the oldest cohort ( ) had a slightly higher incidence among rural born, leading the authors to conclude that urban birth has emerged as a risk

8 730 G. Harrison and others factor for schizophrenia in Finland only in cases born after The range of values from these different European studies suggests that the magnitude of the risk associated with place of birth is smaller in most countries compared with that reported in Denmark. This may be explained by secular trends in urbanization and by differences in definitions of urban and rural. For example, Haukka et al. (2001) classified municipalities as urban if at least 95% of its inhabitants lived in settlements where the distance between neighbouring houses was <200 m and the population in these settlements exceeded 200. This is a very different urban environment from downtown New York or even the metropolitan areas of Copenhagen and Gothenburg. It is likely that the risk attributable to urban birth varies significantly in different sociocultural settings. This appears to be the case for other risk factors: Pedersen & Mortensen (2001a) showed that the risk associated with season of birth varied markedly between subpopulations, even within one country (Denmark). Factors associated with urban place of birth may behave in similar fashion. We should therefore be cautious in generalizing the population attributable risk (PAR) for urban place of birth from estimates determined in studies carried out in one country. Our data suggest that urban birthplace is a general risk factor for broadly defined clinical psychotic syndromes similar to the finding of Eaton et al. (2000). There is no indication in our analysis that associations with schizophrenia increase with greater diagnostic precision; indeed the trend was in the opposite direction. The small reduction in the point estimate of the hazard ratio is not of great consequence, but these data do not suggest that associations are likely to increase with greater diagnostic precision. In a community survey of psychotic symptoms, Van Os et al. (2001) showed that level of urbanicity was associated not only with psychotic disorders defined in terms of DSM- III-R but also, independently, with any rating of delusion and/or hallucination, and with any rating of psychosis-like symptoms. While their study was based upon residence of adults at the time of the survey rather than at birth, the progressive association between level of urbanization and the prevalence of the broad range of clinical psychotic syndromes, as well as prevalence of lesser psychotic states, is consistent with our data. Our findings cast new light upon the possible interaction effects between season and place of birth. Machon et al. (1983), O Callaghan et al. (1991) and Takei et al. (1995) found that place of residence associations were stronger among winter born cases but this was not replicated by Mortensen et al. (1999) or by Pedersen & Mortensen (2001 a). We found the interaction only in non-schizophrenic non-affective psychoses, which may partly account for conflicting findings in studies based upon schizophrenia alone. This finding requires replication, but seasonal differences may reflect seasonally varying exposures such as infection. These would be more likely to affect those born in cities because person to person transmission is more likely in densely populated areas. In conclusion, we find that the association between urban place of birth and schizophrenia is unlikely to be due to rural urban differences in obstetric complications or childhood socioeconomic position. Studies are required to further open up the black box of urban-birth risk, focusing on risk factors operating in childhood and through early adolescence. These have been reviewed by Eaton et al. (2000) and should include nutritional factors, infectious agents and psychosocial risk factors resulting from social adversity. Studies should include all nonaffective psychoses, however, and investigate associations with different models of case definition over the life course of the illness. REFERENCES Eaton, W. W., Mortensen, P. B. & Frydenberg, M. (2000). Obstetric factors, urbanization and psychosis. Schizophrenia Research 43, Gudmundsson, S., Bjorgvinsdottir, L., Molin, J., Gunnarsson, G. & Marsal, K. (1997). Socio-economic status and perinatal outcome according to residence area in the city of Malmo. Acta Obstetrica et Gynecologica Scandinavica 76, Haukka, J., Sivisaari, J., Varilo, T. & Lonnqvist, J. (2001). Regional variation in the incidence of schizophrenia in Finland: a study of birth cohort born from 1950 to Psychological Medicine 31, Jonas, O., Roder, D. & Chan, A. (1992). The association of low socio-economic status in metropolitan Adelaide with maternal demographic and obstetric characteristics and pregnancy outcome. European Journal of Epidemiology 5, Jones, P., Rodgers, B., Murray, R. & Marmot, M. (1994). Child development risk factors for adult schizophrenia in the British 1946 birth cohort. Lancet 344, Lewis, G., David, A., Andre asson, S. & Allebeck, P. (1992). Schizophrenia and city life. Lancet 340,

9 Association between urban birth and schizophrenia 731 Machon, R., Mednick, S. & Schulsinger, F. (1983). The interaction of seasonality, place of birth, genetic risk and subsequent schizophrenia in a high risk sample. British Journal of Psychiatry 143, Makikyro, T., Isohanni, M., Moring, J., Oja, H., Hakko, H., Jones, P. & Rantakallio, P. (1997). Is a child s risk of early onset schizophrenia increased in the highest social class. Schizophrenia Research 23, Marcelis, M., Navarro-Mateu, F., Murray, R., Selten, J.-P. & Van Os, J. (1998). Urbanization and psychosis: a study of birth cohorts in The Netherlands. Psychological Medicine 28, Mortensen, P. B. (1999). Effects of family history and place and season of birth on the risk of schizophrenia. New England Journal of Medicine 340, Mortensen, P. B., Pedersen, C. B., Westergaard, T., Wohlfarht, J., Ewald, E., Mors, O., Andersen, P. K. & Melbye, M. (1999). Effects of family history and place and season of birth on the risk of schizophrenia. New England Journal of Medicine 340, O Callaghan, E., Gibson, T., Colohan, H. A., Walshe, D., Buckley, P. & Larkin, C. (1991). Season of birth in schizophrenia. Evidence for confinement of an excess of winter births to patients without a family history of mental disorder. British Journal of Psychiatry 158, Pedersen, C. B. & Mortensen, P. B. (2001a). Family history, place and season of birth as risk factors for schizophrenia in Denmark: a replication and reanalysis. British Journal of Psychiatry 179, Pedersen, C. B. & Mortensen, P. B. (2001b). Evidence of a dose response relationship between urbanicity at upbringing and schizophrenia risk. Archives of General Psychiatry 58, StataCorp (2001). State Statistical Software, Release 7.0. Stata Corporation: College Station, TX. Torrey, E. F., Bowler, A. E. & Clark, K. (1997). Urban birth and residence as risk factors for psychoses: an analysis of 1880 data. Schizophrenia Research 20, Takei, N., Sham, P. C., O Callaghan, E., Glover, G. & Murray, R. M. (1995). Early risk factors in schizophrenia: place and season of birth. European Psychiatry 10, Van Os, J., Hanssen, M., Bijl, R. V. & Vollebergh, W. (2001). Prevalence of psychotic disorder and community level of psychotic symptoms. Archives of General Psychiatry 58, Wahlbeck, K., Forsen, T., Osmond, C., Barker, D. & Eriksson, J. (2001). Association of Schizophrenia with low maternal body mass index, small size at birth, and thinness during childhood. Archives of General Psychiatry 58, Wilcox, M. A., Smith, S. J., Johnson, I. R., Maynard, P. V. & Chilvers, C. E. (1995). The effect of social deprivation on birth weight, excluding physiological and pathological effects. British Journal of Obstetrics and Gynaecology 102, World Health Organization (1978). Mental Disorders: Glossary and Guide to Their Classification in Accordance with the Ninth Revision of the International Classification of Diseases. WHO: Geneva. World Health Organization (1993). The ICD-10 Classification of Mental and Behavioural Disorders, Diagnostic Criteria for Research. WHO: Geneva.