Social Risk or Genetic Liability for Psychosis? A Study of Children Born in Sweden and Reared by Adoptive Parents

previous studies have consistently shown that schizophrenia is more frequent in socioeconomically disadvantaged groups and deprived areas (1–4). This is partly a consequence of the disorder leading to a drift down the social class scale. However, the majority of recent studies have found increased risks for psychosis to be associated with childhood exposure to socioeconomic disadvantage, which indicates that this may also have an etiological role in the illness (5–9).

The genetic component in the etiology of schizophrenia is well established (7). Since this is also related to socioeconomic disadvantages, it is hard to disentangle possible causal mechanisms. The genetic liability may be a confounder to the presumed causal association between social disadvantages during childhood and the later risk of psychoses, or the genetic liability may be confounded by being reared by parents with a psychotic illness. In particular, there may be an interaction effect leading to further increased risk if exposed to both. An adoption design investigation of children born in Sweden and reared in adoptive families is an ideal setting where the effect of environmental factors is separated from genetic components. The availability of information about the biological parents' mental illness as well as the rearing families' socioeconomic position and mental illness offers a unique opportunity to study how socioeconomically disadvantaged conditions interact with genetic liability for psychosis. With this aim, we followed 13,116 children born in Sweden and reared by adoptive parents living in Sweden, thus enabling separation of biological family history of psychosis from social exposure. In addition, we investigated the total Swedish-born population to attain a larger study population.

Method

Study Population

Swedish-born intracountry adoptees.

The present linkage study is based on data from the national registers of Statistics Sweden and the Swedish National Board of Health and Welfare. All children born in Sweden from 1955 to 1984 with known adoptive parents and a known biological mother were identified in the Multigeneration Register via each individual's unique personal identification number. In addition to biological mothers, we were also able to include biological fathers, although a large proportion of fathers in a population of intracountry adoptees are unidentified. Linkage to the Cause of Death Register and the Register of Total Population revealed those who were still alive and residents in Sweden at the age of 16 years (N=16,388). The rearing parents were linked to the Swedish Population and Housing Census, which were performed at 5-year intervals from 1960 to 1990. To keep the two types of exposure (indicators of genetic liability and socioeconomic position) as disentangled as possible, we excluded children who lived with a biological parent at any 5-year-point when they were 1–15 years old, via linkage to the censuses and the Register of Total Population in 1995 (N=1,337). The Swedish adoptees, the rearing parents, and the biological parents were followed in the National Patient Register from 1973 to 2006 regarding inpatient care episodes for nonaffective psychoses, including schizophrenia (as defined in the exposure and outcome measures). Another 1,464 children were excluded because their adoptive parents had inpatient episodes for any psychiatric diagnosis (Table 1) during the adoptee's childhood or young adulthood (before 30 years of age), and another 206 children were excluded because they were adopted by a biological grandparent or sibling. To exclude the effect of growing up in institutions, only children living in family households during two consecutive censuses (at ages 1–10 years old) were included (children who did not: N=261). Another four persons lacked data on the date variables or were diagnosed before the age of 16 years. Thus, a total of 13,116 persons were included in the analyses.

Swedish-born nonadoptees.

As a comparison and to attain more statistical power, a population of nonadoptees born in Sweden from 1955 to 1984 was identified in the same registers. Inclusion criteria were at least one identifiable biological parent and still alive and a resident in Sweden at age 16 years. In total, 2.9 million persons were included in these analyses.

Exposure and Outcome Measures

Data on the socioeconomic position of the rearing family were obtained by linkage to the Swedish Population and Housing Census when the children were 1–5 years old (1960, 1965, 1970, 1975, 1980, or 1985). Available variables were single-parent household, parental employment (paternal employment in two-parent households and maternal employment in single-parent households [available from 1965]), and housing (coded as apartment or own house).

An indicator of genetic liability for psychosis was obtained by linkage to the National Patient Register (1973–2006) and defined as a biological mother or father hospitalized for nonaffective psychosis or affective psychosis (Table 1), since there is growing evidence that schizophrenia and affective psychosis partly share a genetic background (11, 12). In the adoptive population, the rearing environment was separated from both the genetic and environmental exposure of the biological parent. For the nonadoptees, the indicator consisting of biological parental history of psychosis was a mixture of genetic liability and everything else that the family shares in the environment; thus the indicator in this setting is referred to as familial history of psychosis as opposed to the purer measurement of genetic liability in the adopted children.

The dichotomized outcome variable, nonaffective psychosis (schizophrenia and schizotypal and delusional disorders), was defined as being hospitalized at least once for a nonaffective psychotic episode, and data were obtained through linkage to the National Patient Register (1973–2006) (main and contributory diagnosis is defined in Table 1). The outcome was specific to nonaffective psychosis because the pattern of environmental risk factors differs for schizophrenia and affective psychosis/bipolar disorder. For example, obstetric complications, migration, and urbanicity have all been associated with schizophrenia (13–15), while there is no clear evidence of association between these factors and affective psychoses/bipolar disorder (15–17).

Statistical Methods

Sex- and age-adjusted hazard ratios with 95% confidence intervals (CIs) were estimated using Cox proportional hazards models of time in the study, with nonaffective psychosis as the outcome variable. Time in the study was calculated from the starting date, which was January 1973 (or at age 16 years), until the first hospital admission recorded in the National Patient Register, or date of death recorded in the National Cause of Death Register, or date of emigration, or December 2006, whichever came first. Synergy indexes with 95% CIs were calculated according to the following formula, assuming an additive model (18, 19), with “HR” indicating hazard ratio: (HR11–1)/([HR10–1]+[HR01–1]). A synergy index of 1 indicated no biological interaction effect (i.e., the risk was the same for those exposed to both risk factors [HR11] as the sum of risks for those exposed only to the disadvantaged socioeconomic status in childhood risk [HR10] and those exposed only to the genetic liability, alternatively familial history of psychosis, risk [HR01]). If an interaction effect was present, the risk for those exposed to both risk factors was expected to exceed the sum of risk 1 and risk 2. Proportional differences were tested with chi-square statistics. The SAS software package, Version 9.1 (SAS Institute, Inc., Cary, N.C.), was used in all statistical analyses.

Results

Population of Swedish-Born Intracountry Adoptees

Among the 13,116 adoptees in the study population, 230 developed a nonaffective psychosis during 1973–2006, 47.8% were female adoptees, and 39.6% had a diagnosis of schizophrenia. The number of adoptions per year (using the inclusion criteria of the present study) decreased during this time period, from 869 persons in 1955 to 35 persons in 1984. The proportion of unidentified biological fathers was 41%. Analyses regarding possible differences between those with an identified biological father and those with an unidentified biological father were performed. A total of 5.2% of the adoptees with unidentified biological fathers had a biological mother with inpatient care for psychosis, relative to 5.7% among adoptees with identified biological fathers, a nonsignificant difference. The sex distribution was the same among adoptees with identified (men: 52%) and unidentified biological fathers (men: 53%). There was an association between the adoptee's birth year (5-year groups) and identity of the biological father. It was more frequent to have unidentified fathers in the earliest birth cohorts (χ²=180, df=5, p<0.0001), and it was also more frequent to have an unidentified father with a teenage biological mother (χ²=47, df=3, p<0.0001). There was no association between adoptees' own psychotic illness and the identity (known/unknown) of their fathers (hazard ratio=0.9, 95% CI=0.7–1.1).

Incidence Rates of Nonaffective Psychosis

Incidence rates for nonaffective psychosis were, regardless of genetic liability for psychosis (biological parental in patient care for psychosis), higher among adoptees reared in households with less than advantaged socioeconomic positions, relative to those from households with a more advantaged position (Table 2). The incidence rate of psychosis for adoptees reared in households with parental unemployment in early childhood was 1.7/1,000 person-years at risk, compared with 0.6/1,000 person-years for those reared in households with parental employment. The same pattern was observed for adoptees reared in single-parent households compared with two-parent households. A smaller difference was found among those reared in families living in apartments relative to their own house.

The incidence rate of nonaffective psychosis was, regardless of socioeconomic position, higher among adoptees with genetic liability for psychosis relative to those without (2.6./1,000 person-years and 0.5/1,000 person-years, respectively). The difference in rates would have been larger had we studied cases of schizophrenia with and without genetic liability for schizophrenia rather than psychosis, since there is a stronger association between schizophrenia and parental schizophrenia relative to psychosis and parental psychosis (11, 20).

Risk Analyses

Separate analyses were performed for the three different socioeconomic variables. In each analysis, the adoptees were grouped into whether they were exposed only to the socioeconomic indicator, only to the indicator of genetic liability, or to both the socioeconomic and genetic liability indicators and compared to the unexposed persons. Each analysis was adjusted for sex, age, and the other two exposures. Adoptees reared in families with parental unemployment (hazard ratio=2.0, 95% CI=1.0–4.2), in single-parent households (hazard ratio=1.2, 95% CI=0.6–2.6), or living in apartments (hazard ratio=1.3, 95% CI=1.0–1.8) and were without the presence of genetic liability had increased risk for psychosis (Table 3). The risk for nonaffective psychosis among persons with genetic liability for psychosis alone (without exposure to a less advantaged socioeconomic position in childhood) was increased (hazard ratio=4.7, 95% CI=3.1–7.2). Thus, there was some support that both genetic and socioeconomic indicators were independently associated with an increased risk of psychosis.

Finally, the hazard ratio for nonaffective psychosis associated with living in a family with parental unemployment in early childhood in addition to exposure to the indicator of genetic liability for psychosis was 15.0 (95% CI=5.4–42.3) (Table 3). To test for an interaction effect (i.e., an effect larger than the sum of being exposed to both social and genetic indicators [rearing family parental unemployment and biological parental inpatient care for psychosis]), a synergy index was estimated. The synergy index of 3.2 (>1) indicates an interaction effect, which is just statistically significant. Adoptees living in single-parent households in early childhood with genetic liability for psychosis had an increased risk for psychosis, with a hazard ratio of 10.3, relative to those without genetic liability who were reared in two-parent households. The synergy index of 2.6 indicates an interaction effect, although not statistically significant. The hazard ratio for psychosis among adoptees with genetic liability who were also living in apartments in early childhood was estimated to be 5.7. The synergy index of 1.2 was not statistically significant.

Population of Swedish-Born Nonadoptees

Among the 2.9 million Swedish-born persons, 24,768 had an inpatient episode for nonaffective psychosis, 57% were male nonadoptees, and 41% had a diagnosis of schizophrenia.

Parental unemployment was more frequent in the general population of Swedish-born persons (5.6%) than among the intracountry adoptees (2.0%), as was single-parent household (9.9% and 3.2%, respectively). A total of 2.9% of the Swedish-born persons had familial history of psychosis. The incidence rate for nonaffective psychosis was 0.7/1,000 person-years at risk for individuals reared in families with parental unemployment relative to 0.4/1,000 person-years for those reared in families with parental employment. Similar result patterns were found for the other two socioeconomic indicators.

Risk Analyses

Nonadoptees reared in families with parental unemployment (hazard ratio=1.5, 95% CI=1.4–1.6), in a single-parent household (hazard ratio=1.4, 95% CI=1.3–1.4), or living in apartments (hazard ratio=1.3, 95% CI=1.2–1.3) and were without the presence of familial history of psychosis had increased risks for psychosis. The risk for non-affective psychosis among persons with familial history of psychosis only (without exposure to a less advantaged socioeconomic position in childhood) was increased (hazard ratio=4.2, 95% CI=3.9–4.5).

Synergy indexes calculated to test for an interaction effect between indicator of familial history of psychosis and unemployment (synergy index=1.18, 95% CI=1.03–1.36), single-parent household (synergy index=1.22, 95% CI=1.08–1.38), and housing (synergy index=1.09, 95% CI=0.98–1.20) were all above 1.

Discussion

The study population of 13,116 adoptees born in Sweden and reared by adoptive parents in Sweden enabled us to study a situation where indicator of genetic liability for psychosis and exposure to indicators of socioeconomic disadvantage in childhood were separated. Both indicator of genetic liability and indicators of socioeconomic position (two out of three) were independently associated with an increased risk for psychosis. Adoptees reared in families with a disadvantaged socioeconomic position had an increased risk for psychosis, regardless of genetic liability. The pattern was the same in the total Swedish population (statistical significance for all indicators). This is in accordance with recent studies linking socioeconomic disadvantage at birth and in childhood to increased risk of psychosis. Concerning the question of gene and environment interaction, the risk for nonaffective psychosis was further increased for adoptees with genetic liability (biological parental inpatient care for psychosis) who also were reared in disadvantaged socioeconomic situations in childhood. This risk, when being exposed to both types of indicators, was larger than the sum of the two individual risks, suggesting an interaction effect. Thus, support was found for an interaction effect between the pure indicator of genetic liability and the indicators of socioeconomic position in childhood, although not always fully statistically significant. However, analyses in the larger study population of Swedish-born nonadoptees support these results, although the effect sizes are smaller because of the more imprecise indicator familial history of psychosis. Interaction effects of socioeconomic disadvantage and genetic liability have not been studied before, but our results are in agreement with former studies of other environment factors such as urbanicity and parental rearing patterns (21–24) and interaction effects with biological family history of psychosis. The results suggest that influencing the socioeconomic situation in childhood may have a positive effect on the risk of developing psychosis for children both with and without genetic liability for psychosis.

Unemployment among parents means exclusion from the workforce and, to a large extent, other daily social contacts. It could also mean exclusion from feelings of contributing to the society, feelings of being needed, useful, etc. The lack of steady income (or control over the income) limits the family's possibilities of housing accommodation, physical and social activities, and holidays and adds to the stress of unforeseen necessary costs. Single-parent household also has its economical consequences, and the social situation could be experienced as more stressful with feelings of lone responsibility, lack of everyday support, and exclusion from “the norm.” Housing situation is another indicator of socioeconomic position, with house owners usually having a better socioeconomic position than others. All these different conditions affect the children in the family. The socioeconomic indicators could be seen as markers of different degrees of social exclusion. Social exclusion has, along with other kinds of discrimination (skin color, age, gender, disability, and size of ethnic minorities), been associated with mental illness (25–28). Perceived disadvantage has also recently been associated with psychosis (29). A disadvantaged socioeconomic situation could lead to social exclusion, isolation, alienation, feelings of social defeat, and a stressful life situation. In a stressful life situation, there may also be fewer opportunities for social support, and these situations may be of importance in the development of psychosis in people with a biological vulnerability. Possible biological mechanisms, among several, could be the potential harmful effects by increased levels of the stress hormone cortisol (30) or that stressful situations may influence the genetic expression (i.e., epigenetics) (31). The hypothesis that chronic experience of social defeat leads to disturbance of the mesolimbic dopamine system in the brain, which increases the risk for schizophrenia, has also been put forward (32).

Strengths and Limitations

To our knowledge, there are as yet no studies where the role of social factors in childhood and genetic liability for psychosis and the interaction between the two have been considered in the development of psychosis. One strength of the adoption design is that we were able to separate socioeconomic position during childhood (in the rearing family) and genetic liability for psychosis (biological parental psychotic illness). Thus, we can study the effect of socioeconomic position in childhood and genetic liability independently as well as potential interaction effects of being exposed to both. Several methods were used to separate the two types of exposure as much as possible. Children who lived with a biological parent at any census during 1–15 years of age were excluded as were children with rearing parents with psychiatric admissions and those adopted by a grandparent or sibling. The analyses in this study are based on the assumption that the characteristics of the environment and the child before adoption were random in relation to the socioeconomic characteristics of the adoptive home. To remove any effects of growing up in institutions, children who were not living in family households at age 1–5 years and 6–10 years were excluded. Age at adoption was unknown, but the majority (97.1%) of the children lived with their adoptive parents at their first census at age 1–5 years. Thus, most children were adopted at least before 5 years of age. Other studies have shown that the great majority of the adoptees were adopted very early, most of them in infancy (33, 34)

In addition, we investigated the total Swedish-born population. The two approaches have their advantages and disadvantages. The adoptive population has small numbers but a more precise indicator of genetic liability. The nonadoptive population has more power, at the expense of intertwined indicators of genetic liability and socioeconomic position. However, the use of the two populations is a strength of the study.

The potential impact of adoption (unwanted child, nonoptimal pregnancy, risk of poor antenatal care, increased occurrence of psychiatric illness, etc.) could be a problem if adopted children are compared with others. We controlled for this because analyses were made only within the population of adopted children (i.e., they were all exposed to the circumstances of adoption). However, it is possible that adopted children relative to the general population are more sensitive to environmental adversities as a result of the possible impact of adoption.

The children were born between 1955 and 1984 and therefore had parents of various ages during the time of national coverage of psychiatric inpatient care in the National Patient Register (1973–2006). We were not able to take psychosis among grandparents into account for the same reason. In addition, a considerable number of biological fathers were unidentified, although not unexpected in a population of intracountry adopted children born between 1955 and 1984. One inclusion criterion for this study was that the biological mother should be identified, and most were (98%–100%), but 41% of the biological fathers were unknown in the adoptive population. An unidentified father was more common among adoptees born in the earliest years, and it was also more frequent to have a teenage mother among those with an unidentified father. There were no differences between those with identified biological fathers relative to those with unidentified biological fathers regarding the proportion of maternal psychosis, the sex of the adoptee, and the adoptee's own psychotic illness. However, the proportion of adoptees with genetic liability for psychosis is most likely underestimated, which in turn would attenuate our results rather than overestimate them. The choice of social indicators is equally important (35). Socioeconomic position is a multidimensional construct consisting of several different factors (36, 37), hence the importance of using more than one indicator of socioeconomic position. In this study, we were able to use three. However, the indicators of socioeconomic position in childhood as well as indicator of genetic liability are crude. Consequently, effects may be underestimated for both types of indicators as well for interaction effects.

We relied upon register-based clinical diagnoses of nonaffective psychoses. The quality of these can be questioned, but several studies have shown that the validity in Sweden is appropriate for these types of epidemiological studies (38–40).

Conclusions

In conclusion, our results indicate that adoptive and nonadoptive children reared in families with a less advantaged socioeconomic position have an increased risk for nonaffective psychosis, regardless of biological parental history of psychosis. Furthermore, there is some support for an interaction effect in children with genetic liability for psychosis who also are exposed to a less advantaged socioeconomic position during childhood. Thus, influencing the social situation in childhood may have beneficial effects on the occurrence of psychosis. This may be even more important among children with genetic liability. Further studies with a gene-environment design are needed to clarify the mechanisms that explain the interaction of genetic liability and social factors for the development of nonaffective psychosis so that effective preventive measures can be developed for vulnerable children.