Schizophrenia and urban deprivation: When the facts change, I change my mind. What do you do?

photo-1430760814266-9c81759e5e55

In my head, I had the association between increased urbanicity and risk of schizophrenia (Vassos et al, 2012) boxed off as a purely environmental risk factor. Being born in the capital city compared to being born in rural areas increases the risk of developing most mental disorders (Vassos et al, 2016).

Then along came this paper (Sariaslan et al, 2016), which followed-up an earlier study by the same group (Sariaslan et al, 2014) linking multiple data sources in Sweden (which keeps frightening amounts of data about their citizens in a variety of reliable Registries) to show that increased population density as well as deprivation increased the risk for a person being diagnosed with schizophrenia. This effect disappeared once effects of unobserved familial risk factors were accounted for. This suggested that urbanicity and deprivation effects on the risk of developing schizophrenia are due to more than the environmental effects of living in an urban and/or deprived environment.

Has previous research overemphasized the relative importance of environmental influences in the social drift of schizophrenia patients?

Has previous research over-emphasized the relative importance of environmental influences in the social drift of schizophrenia patients?

Methods

The authors used two sources of data and ran appropriate analyses for them to check if their experimental hypothesis were likely to be true in both datasets to improve confidence in their results.

The first dataset was constructed from cross-referencing a Total Population Register to identify people born between 1951-74, the Multi Generation Register to identify full and half-siblings, National Patient Register to identify cases of schizophrenia as being hospitalised twice with this diagnosis (a reliable technique but identifying a relatively severely ill population) and the Small Area Marketing Statistics to define residence between 1982 and 2009 and other databases to identify Neighbourhood characteristics such as urban density and deprivation. Final sample was 759,536 sibling pairs (including half siblings) out of 2,628,631 live births in the period.

The second dataset was from the “TCHAD (Twin Study of Child and Adolescent Development); a population-based longitudinal twin study including all twins born in Sweden between May 1985 and December 1986 who were alive and Swedish residents at the end of 1994 (n = 2960) …The final sample included 1355 twin pairs, distributed over 533 monozygotic twin pairs (genetically identical) and 822 dizygotic twin pairs (sharing 50% genes in common).” Psychotic experiences were identified from the Child Behaviour Checklist and broader definitions of deprivation used. Blood samples were also collected from twins and genotyped to create a “polygenic risk score”. “One twin from each pair was randomly selected for inclusion to yield a data set of 6,796 unrelated individuals, avoiding potential biases from analyses of genetically correlated subjects.” The polygenic risk score was based on a “discovery SNP (single nucleotide polymorphisms) set (that) was generated from the results of the genome-wide association study of schizophrenia from the Psychiatric Genomics Consortium”.

Assessments and Results

For the first dataset, the first analysis was a comparison of odds ratios of sibling pairs with one member with schizophrenia vs sibling pairs without a member diagnosed with schizophrenia living in deprived areas, using the binary logistic regression model. This found odds ratios for full siblings to be higher than half-siblings, but the confidence intervals overlapped. Sibling pairs affected by schizophrenia were twice as likely to live in deprived areas (24%) compared to sibling pairs not affected by schizophrenia (12%).

For the second dataset, a similar logistic regression model analysis was performed but using the presence of psychotic experiences for differentiating monozygotic and dizygotic twin pairs. There was a big difference in odds ratios between monozygotic twin pairs living in deprivation with psychotic experiences (OR=7.03) compared to dizygotic twin pairs living in deprivation with psychotic experiences (OR=1.23), but the 95% confidence intervals overlapped.

A univariate analysis of the “ACE” type was performed with the first dataset using evidence-based assumptions. Additive genetic factors accounted for 0.65 of the variance for deprivation and 0.72 for schizophrenia, unique environmental factors accounted for 0.32 of variance for deprivation and 0.28 for schizophrenia (shared environmental factors contributed little to variance).

In the second dataset of twins, for deprivation there was a much more even contribution of additive genetic (0.41), shared environmental (0.26) and unique environmental factors (0.32) to variance. For psychotic experiences, 0.90 of the variance was due to additive genetic factors with little contribution from shared or unique environmental factors.

A bivariate analysis in the first sample found a moderate correlation between deprivation and schizophrenia (r=0.22). A similar moderate correlation (r=0.20) was found in the second sample.

The polygenic risk score for the second sample was found to be significant in predicting a relationship between schizophrenia and living in deprivation and all the SNPs included contributed to 0.15 of the variance in neighbourhood deprivation.

Analysis was repeated varying the assumptions made and no significant differences found in the results.

This study suggests that genetic liability for schizophrenia also predicts subsequent residence in socioeconomically deprived neighbourhoods.

This study suggests that genetic liability for schizophrenia also predicts subsequent residence in socioeconomically deprived neighbourhoods.

Strengths

  • Use of large databases of comprehensive information collected nationally for first dataset
  • Use of second dataset to confirm results from first dataset, collected from large twin study
  • Multiple analysis in different datasets all triangulating to give same result increase confidence in conclusions, including the use of molecular genetics based on an existing dataset
  • Similar results in both severe form of schizophrenia plus a broader psychosis phenotype in twin studies indicates results applicable to both schizophrenia and a broader psychosis phenotype
  • Repeating analysis varying assumptions used (including other non-affective psychosis diagnosis from the Patient Registry) showed similar results.

Weaknesses

  • Data not collected prospectively for the purposes of this analysis
  • Sweden more socially mobile and less ethnically diverse than UK, so not totally identical populations when extrapolating results to the UK
  • The ACE type of analysis involves assumptions such as “no assortative mating, equal exposure across sibling types and a constant shared environmental correlation for each sibling type”. Violations of these assumptions may lead to an increase or decrease in the contribution of genetic factors but is unlikely to have had a major effect and is inevitable in this type of modelling.

Overall I was impressed by the rigour of the study and felt that the strengths outweighed the weaknesses.

Conclusions

I experienced cognitive resistance to the conclusions of this study as I had previously settled on the increased risk of schizophrenia due to urbanicity and deprivation being environmental in nature. However, this study seems to provide strong evidence that this effect of deprivation increasing the risk of schizophrenia is largely caused by genetics increasing the risk of psychosis and schizophrenia, as well as genetics increasing the risk of deprivation. The authors previous study (Sariaslan et al, 2014) showed that familial effects were more important than the duration of time spent experiencing deprivation as an adolescent.

So what now? Replication studies would obviously be helpful. People who think that the urbanicity and deprivation effects increasing risk of schizophrenia/ psychosis is an environmental effect need to conduct studies to demonstrate that this view is correct, taking into account genetic factors. It is insufficient to merely provide theoretical reasons as to why this study is wrong or to highlight any methodological problems. All research is imperfect and theories may seem correct but fail under empirical challenge. The gauntlet is thrown down. I have had to change my mind. Will further research change it back again?

Please note that as part of writing this blog I discussed methodological questions about the paper with the first author of the paper.

Family-based study designs will be an important component of future research in this field.

Family-based study designs will be an important component of future research in this field.

Links

Primary paper

Sariaslan A, Fazel S, D’Onofrio BM, Långström N, Larsson H, Bergen SE, Kuja-Halkola R, Lichtenstein P. (2016) Schizophrenia and subsequent neighborhood deprivation: revisiting the social drift hypothesis using population, twin and molecular genetic dataTranslational Psychiatry (2016) 6, e796; doi:10.1038/tp.2016.62 Published online 3 May 2016

Other references

Sariaslan A, Larsson H, D’Onofrio B, Långström N, Fazel S, Lichtenstein P. (2016) Does population density and neighborhood deprivation predict schizophrenia? A nationwide Swedish family-based study of 2.4 million individuals. Schizophr Bull. 2015 Mar;41(2):494-502. doi: 10.1093/schbul/sbu105. Epub 2014 Jul 22.

Vassos E, Pedersen CB, Murray RM, Collier DA, Lewis CM. (2012) Meta-analysis of the association of urbanicity with schizophrenia. Schizophr Bull. 2012 Nov;38(6):1118-23. doi: 10.1093/schbul/sbs096. Epub 2012 Sep 26.

Vassos E, Agerbo E, Mors O, Pedersen CB. (2016) Urban-rural differences in incidence rates of psychiatric disorders in Denmark. Br J Psychiatry. 2016 May;208(5):435-40. doi: 10.1192/bjp.bp.114.161091. Epub 2015 Dec 17. [PubMed abstract]

Share on Facebook Tweet this on Twitter Share on LinkedIn Share on Google+