On Ethnicity, Socioeconomic Status and Schizophrenia
In their recent article, Bresnahan and colleagues have sought to investigate whether family socioeconomic status (SES) at-birth of the subject mediates the association between race/ethnicity and schizophrenia spectrum disorders (SSD) in a U.S. cohort. This line of research has a noble lineage (Harrison et al., 1988; King et al., 1994; van Os et al., 1996), previously predominantly restricted to countries of Western Europe, where the association between Black and Minority Ethnic
(BME) status and schizophrenia is particularly well-established (Fearon et al., 2006; Selten et al., 2001; Veling et al., 2006). Thus, the finding by Bresnahan et al. that the treated incidence of schizophrenia in the African-American group is between two- and threefold that of the white group in the U.S. both complements and extends findings from the U.K., Netherlands, Sweden, and Denmark (Cantor-Graae et al., 2005).
On the one hand, these findings, conducted in a different setting using a cohort study design, add further weight to the evidence suggesting factors related to race, ethnicity, or immigration are etiologically relevant to the onset of psychoses (Bradford-Hill, 1965). On the other hand, and as Bresnahan and colleagues allude to in their introduction, there are substantial differences in the immigration histories of the United States and Western Europe, both in terms of duration and pattern. Thus, in comparison to the U.S. model of immigration, the European model has a relatively young origin, with substantial immigration from former colonial countries following World War II. In the U.S., the immigration history of the African-American population has much older, deeper antecedents. The findings of Bresnahan et al. suggest that raised rates of psychoses in ethnic minority groups persist independently of the length of immigration history, something which should be of considerable interest to healthcare planners in Western Europe as we see second- and third-generation immigrants enter the main age period of risk for schizophrenia.
Bresnahan and colleagues have attempted to investigate whether family SES mediates the association between race/ethnicity and psychotic disorder. This is an important issue which has received relatively little attention to date, despite calls to do so (Cooper, 2005). After considering the mediating effects of four measures of family SES separately, using a stratified analysis, Bresnahan and colleagues find that the excess risk of SSD in the African-American group remains approximately double that of the rate in the white group, while for schizophrenia alone, this figure is closer to a threefold increase in risk. For both outcomes, the excess risk becomes non-significantly raised by a factor closer to two when all four factors are simultaneously entered into the model, though the study may be underpowered to detect significant effects given its relatively small sample size. Despite the study’s small sample size, there is some evidence that SES at birth does not wholly mediate the raised rate of schizophrenia in the African-American group.
The authors raise the possibility that other factors at both the individual and societal levels may be important in explaining this excess. In this respect, discrimination may be an important candidate socioenvironmental risk factor (Selten et al., 2005). Recently, Veling et al. (Veling et al., 2006) have found evidence that groups which face greater levels of discrimination have significantly higher rates of psychoses. This complements findings with regard to the ethnic density hypothesis (Boydell et al., 2001), while a recent study by our group has found that the rate of schizophrenia is lower in neighborhoods where ethnic groups live in more cohesive (and perhaps supportive) residential patterns (Kirkbride et al., 2007, in press). It also supports findings from a recent meta-analysis which showed that the risk of psychoses in immigrant groups increased with darker skin colors (Cantor-Graae et al., 2005), although biological explanations for this variation may also exist (see, for example, the relationship between sunlight, vitamin D, and psychoses) (McGrath et al., 2003).
While the study of Bresnahan et al. allows us to put a handle on the potential mediating effect of SES, a number of key limitations need to be overcome in order to elicit the specific risk factors underpinning the ethnicity effect. Primary in this concern is the need to design larger studies which can obtain precise estimates of risk in different ethnic groups, having considered the confounding or mediating effects of several variables. Bresnahan et al. included four confounders in their analyses, including paternal age and maternal body-mass index during pregnancy, but these were not considered simultaneously, most probably due to the small sample size. Although costly, larger studies will be required if we are to develop models which are sufficiently complex to include a number of covariates, potentially operating at multiple levels (March et al., 2006). Developing models which integrate risk factors across individual, familial, and societal levels will be vital to understanding the incidence of psychoses across different ethnic groups. Understanding person- (or gene-)environment interactions will also be important in this respect. Thirdly, prospectively collected, longitudinal study designs will be necessary to elicit when timing to ubiquitous socioenvironmental exposures is most detrimental in the onset of psychoses. Bresnahan and colleagues are to be commended on the prospective case finding design of their study, though as they acknowledge, it will be necessary to take longitudinal measures of SES to determine the exact pathways through which family SES may mediate the association between race and schizophrenia. Finally, it will be vital to acknowledge and investigate heterogeneity in exposures, mediators, confounders, and outcomes if we are to develop models with real predictive value. Although both white and African-American individuals were identified on every stratum of SES included in Bresnahan and colleagues’ study, their cohort was restricted to individuals with health insurance who were in “the middle of the income distribution in the region, and not the extremes.” Given that only very low SES may be causally relevant to the risk of psychoses (Harrison et al., 2001), and that SES and race/ethnicity are often highly correlated, it will be vital to incorporate this heterogeneity in future models to clarify the respective roles of ethnicity and socioeconomic status.
Raised rates of psychoses in Black and Minority Ethnic (BME) groups have now been observed on both sides of the Atlantic, confirming that this phenomenon is not restricted to Western Europe. The work of Bresnahan et al. suggests this cannot be entirely mediated by family SES and supports the likelihood that other socioenvironmental factors, such as discrimination, may be etiologically relevant to the raised rates in BME groups. Furthermore, it remains to be seen whether the factors which explain these excesses in the U.S. and Europe arise from similar models, given historical differences in immigration patterns.
View all comments by James Kirkbride