Heterogeneity An issue in design & analysis for population-based neuroscience

1. Heterogeneity An issue in design & analysis for population-based neuroscience James B Kirkbride Ph.D. Sir Henry Wellcome Research Fellow University of Cambridge Part II Neuroscience Workshop

2. “Heterogeneity” “Population-based”

3. “Heterogeneity”

8. What if everyone had smoked? What would we conclude about the relationship between smoking and lung cancer?

9. 0 649 n.s 0 649

10. What if everyone had smoked? What would we conclude about the relationship between smoking and lung cancer? Other factors: alcohol, social class, genetics… Importance of heterogeneity!!!

11. Heterogeneity An issue in design & analysis for population-based neuroscience James B Kirkbride Ph.D. Sir Henry Wellcome Research Fellow University of Cambridge Part II Neuroscience Workshop

12. Population based neuroscience… What are psychotic disorders?

13. Population based neuroscience… • Debilitating set of mental disorders, characterised by: • Positive symptoms • Seeing or hearing things which aren’t there - hallucinations • Erroneous beliefs held contrary to normal view of subject - delusions • Negative symptoms (flat affect, anhedonia) • Affective symptoms – changes in mood (bipolar disorder, psychotic depression) • Chronic, debilitating socially & economically

14. Population based neuroscience… Positive Affective Negative

15. Population based neuroscience… How common are psychotic disorders?

16. Population based neuroscience… • Prevalence of psychotic symptoms: • Median: 5.3% (IQR: 1.9-14.4%) Van Os et al. 2009 • Prevalence of psychotic disorders: • Lifetime prevalence: ~1% Sahaet al. 2005; Goldneret al. 2002; Peralaet al. 2007; Robins & Regier 1990; Kessler et al. 1994 • Annual prevalence: ~0.4% Kirkbride et al, In Press; Sahaet al. 2005

17. Population based neuroscience… • Incidence of psychotic disorders: • All psychotic disorders: 31.7 per 100k/pa Kirkbride et al.PLoS One, 2012 • Schizophrenia : 15.2 per 100k/pa McGrath et al. 2004; Kirkbride et al.PLoS One, 2012 • Affective psychoses: 12.4 per 100k/pa Kirkbride et al.PLoS One, 2012 • Figures are for guidance only – preclude important heterogeneity along key socio-demographic dimensions

18. No mind is an island… [Background] Kirkbride et al,2006; Hafneret al 1993

19. Neighbourhoods & psychotic disorder

20. Population based neuroscience… Is there an association?

22. Schizophrenia

23. Population based neuroscience… • Highest rates of schizophrenia in inner city tracts • These tracts tended to be more socially disorganized, socially isolated • No comparable association with bipolar disorder • Also observed elsewhere: Mannheim (Germany), Ireland, London, Bristol, Nottingham

24. No mind is an island…

26. Population based neuroscience… “Don’t people with psychosis end up moving into inner-city neighbourhoods?”

27. Population based neuroscience… “Aren’t these results explained by reverse causation (social drift)?”

28. Population based neuroscience… • Prospectively collected national data on 1.75m Danish people • Urban birthplace & upbringing associated with increased risk of schizophrenia in later life • Cannot be explained by social drift • Dose-response relationship

29. Population based neuroscience… • Dose-response relationship: Data from Mortensen et al. (1999). Adjusted for age-sex interaction, parental age, family history and season of birth

30. Heterogeneity An issue in design & analysis for population-based neuroscience James B Kirkbride Ph.D. Sir Henry Wellcome Research Fellow University of Cambridge Part II Neuroscience Workshop

31. “Population-based”

32. Population based neuroscience… • Population-based studies: • Subjects are collected from the entire population “at-risk” • Often random selection • Representative sample • Design of study adheres to the principles of epidemiology: • Minimise chance, bias & confounding • Design may often be case-control or “cohort” based • Analysis appropriate to design

33. Population based neuroscience… • Case-control (ideal) • All cases in a given population of interest identified • Cases invited to participate in study (sub-sample) • Cases need to be representative of the total case group • Controls selected to be representative of population from which cases come from • i.e. had the case not become unwell they could have been a control • Control selection random ideally

34. Population based neuroscience… • Cohort (ideal) • Representative sample of disease free people identified • Cohort followed up over time with assessment • Wait to see who develops disease (or not) • Look for differences in assessment between the two groups • Cohort: prospective • Sample size for psychosis?

35. Population based neuroscience… • Population-based neuroscience: • Both approaches can be used for understanding differences in brain using neuroimaging studies • Principles of epidemiology central to design & analysis of subjects • Observational epidemiology may also inform the research question for neuroscience

36. What mechanisms might increase psychosis risk due to exposure to certain environmental factors?

37. Population based neuroscience… How can population heterogeneity inform clinical neuroscience? (and vice versa)

38. Population based neuroscience… • Urban living and migrant status appear to increase psychosis risk • May operate through social stress • i.e. greater social pressures lead to dopamine dysregulation triggering psychotic symptoms • If this is true, then you might expect to see neural differences in brains of those living in rural & urban areas (for example)

39. No mind is an island… [Mechanisms] • 32 health German volunteers - students • Given a stress-test (arithmetic based) to evaluate stress, while under fMRI • Asked about current & past upbringing • Second sample for validation (n=23)

40. Amygdala signals –ve affect & environmental threat

41. Perigenual anterior cingulate cortex – region controlling stress, negative affect & amygdala

42. No mind is an island… [Mechanisms] • Suggests greater sensitivity to stress in city dwellers • Amygdala signals –ve affect & environmental threat – important in psychosis symptoms • Confirms & informs epidemiology • Heterogeneity of exposure critical to detecting this

43. No mind is an island… [Mechanisms] • Next steps • Cases vs. controls? • Migrants vs. non-migrants? • Larger samples? • Interplay with genetic factors? • Measuring (variation in) specific exposures (beyond urbanicity)

44. Population based neuroscience… • Heterogeneity provides us with a window on detecting associations between exposure & disorder • Mental disorders are no exception • To detect risk factors for mental disorders, heterogeneity in the risk factor (whether genetic or environmental) is important • This is also true for neuroscientific research • To attribute differences in brain activity to a proposed factor relies on variation in the factor • fMRI studies need to be designed to take this heterogeneity into account

45. Population based neuroscience… • Peter Jones • Jeremy Coid • ÆSOP study group • ELFEP study group Acknowledgements • Wellcome Trust, UK • Gratitude & thanks to Amber Christian Osterhout, creator of the Gaining Insight campaign, www.gaining-insight.com for allowing me to useher images in this presentation

46. Population based neuroscience… jbk25@cam.ac.uk “Disease Onset" by Amber Christian Osterhout, creator of the Gaining Insight campaign, www.gaining-insight.com