Rapid Responses to:
|
|
Rapid Responses: Rapid Responses published:
-
![[Read Rapid Response]](/icons/misc/sectionDOWN.gif)
Selection bias may have led to results
- Glyn Lewis, David Gunnell, Glynn Harrison (8 January 2002)
|
|
|||
|
Glyn Lewis, Professor of Psychiatric Epidemiology Division of Psychiatry, University of Bristol, Cotham House, Cotham Hill, Bristol BS6 6JL, David Gunnell, Glynn Harrison
Send response to journal:
|
Mulvany et al1 address an important question concerning the relationship between socioeconomic status and schizophrenia. Their main observation was of a reduced rate of schizophrenia in those with parents of a lower occupational status OR 0.59 (95% CI 0.4 to 0.85; P=0.003). We cannot understand why the abstract therefore concludes that “social class of origin does not seem to be an important risk factor for schizophrenia”. However, we are concerned that the association they report between high social class and an increased risk of schizophrenia resulted from selection bias. Selection bias can occur in case-control studies when the cases and controls are not from the same population. In essence, the controls, if they developed the disease, should be at risk of being selected as cases in the study. In Mulvany et al1 the cases were individuals born anywhere in Ireland who were admitted for schizophrenia when residing in a small area of Dublin. The controls were matched for area of birth in Ireland, but their whereabouts as adults were not known. Internal migration within Ireland and emigration from Ireland could therefore lead to selection bias2. In particular, those of a higher social class are more internally mobile, at least in the UK, 3 and this would explain the over- representation of people of higher social class in the cases compared to the controls. Likewise, emigration of those of lower socioeconomic status would also lead to a comparative increase of low social class subjects in the controls. Only 58% (352/610) of the original sample of eligible cases were included in the case-control study and this might also have introduced further selection bias. The issue of selection bias can be circumvented to ensure that both cases and controls are from the same population. In this example, it would have been necessary to check that the controls, when adult, were living in the area of Dublin from which the cases had been identified. This approach has been adopted by Harrison et al4 and Dalman et al5 in studying early factors that might affect schizophrenia incidence. Harrison et al, they found an association between low social class and the incidence of schizophrenia. Most medical conditions are commoner in people from deprived backgrounds. The results of Mulvany et al are not sufficiently robust to conclude that social class of origin is not an important risk factor for schizophrenia. Studying the association between socioeconomic status and schizophrenia using robust methods could give important clues concerning the aetiology of this important condition. Reference List 1. Mulvany F, O'Callaghan E, Takei N, Byrne M, Fearon P, Larkin C. Effect of social class at birth on risk and presentation of schizophrenia: case-control study. BMJ 2001;323:1398-401. 2. Jones ME,.Swerdlow AJ. Bias caused by migration in case-control studies of prenatal risk factors for childhood and adult diseases. American Journal of Epidemiology 1996;143:823-31. 3. Champion T. Population review: (3) Migration to, from and within the United Kingdom. Population Trends 1996;83:5-16. 4. Harrison G, Gunnell D, Glazebrook C, Page K, Kwiecinski R. Association between schizophrenia and social inequality at birth: case- control study. B.J.Psych. 2001;179:346-50. 5. Dalman C, Thomas HV, David A, Gentz J, Lewis G, Allebeck P. Signs of asphyxia at birth and risk of schizophrenia. B.J.Psych. 2001;179:403-8. |
|||