The impact of cannabis use and polygenic risk scores on symptom dimension profiles at psychosis onset

Student thesis: Doctoral ThesisDoctor of Philosophy


Epidemiological and biological evidence show no boundaries between diagnostic categories of non-affective and affective psychoses, thus challenging the current nosological model developed from Kraepelin’s paradigm. My thesis aimed to address this limitation by 1) examining the transdiagnostic dimensional structure of i) psychotic symptoms in first episode psychosis (FEP) patients and ii) psychotic experiences in the general population; 2) investigating the relationship between these dimensions and a set of external factors, such as the use of cannabis and genetic common variant liability for psychotic disorders. Overall, I expected that differences in symptom profiles at FEP reflected gradients of neurodevelopmental impairment in psychosis.

This thesis uses data from a multisite incidence and case-control study, which I worked on, conducted across six countries [i.e. the ‘European Network of National Schizophrenia Networks studying Gene-Environment Interactions’ (EU-GEI) study]. To examine the latent structure of psychopathology, I analysed ratings of psychotic symptoms and experiences using multidimensional item response modelling in Mplus, and I estimated different theory-based models of psychosis, including unidimensional, multidimensional, bifactor, and hierarchical solutions. To examine the common variant liability to psychosis, I examined the population structure in the EU-GEI sample and computed ancestry-specific schizophrenia (SZ), bipolar disorder 7 (BP), and combined schizophrenia-bipolar disorder (SZ-BP) Polygenic Risk Scores (PRSs) in PRSice. To examine the relationship between the latent structure of psychopathology and demographic and context determinants, detailed patterns of cannabis consumption, and PRSs, I used multiple linear regression models fitted in STATA14.

The associations among ratings of both psychotic symptoms in FEP patients and psychotic experiences in population-based controls were best represented by a bifactor model, composed of one general psychosis factor and multiple specific dimensions. In FEP patients, the examination of general and specific dimensions with external factors showed that 1) higher scores on the negative symptom dimension were associated with being a male, and having never used cannabis; 2) higher scores on the positive symptom dimension were associated with exposure to socioenvironmental risk factors in psychosis, such as being part of an ethnic minority, and having had exposure to cannabis in a dose-response fashion, with those having used high potency varieties on a daily basis having the highest score; 3) both higher scores on the positive and negative symptom dimensions were associated with a higher SZ-PRS. In population-based controls, the examination of general and specific dimensions with external factors showed that 1) higher scores on the positive psychotic experience dimension were associated with current use of cannabis but not with the extent of lifetime exposure to cannabis; 2) higher scores on the general and all the specific psychotic experience dimensions were associated with a high SZ-PRS.

My thesis shows that symptom dimensions are useful psychosis phenotypes, that are validated by psychometric data and socioenvironmental and genetic factors. Specifically, the bifactor model of psychopathology holds across diagnostic categories of non-affective and affective psychosis at FEP and in the general population. Furthermore, my findings indicate that use of cannabis is associated with more positive and less negative symptoms at FEP, consistently with the hypothesis that cannabis users who develop psychosis have less early neurodevelopmental impairment than their non-user counterparts. Overall, these findings indicate that it is appropriate to conduct research using enhanced phenotypes, and they have translational relevance, they are important for developing secondary prevention strategies in psychosis. Currently, symptom dimensions at FEP could be used for formulating clinical impressions regardless of diagnostic categories according to a developmental-symptom approach, and for guiding tailored treatments
Date of Award1 Sept 2020
Original languageEnglish
Awarding Institution
  • King's College London
SupervisorMarta Di Forti (Supervisor), Robin Murray (Supervisor), Cathryn Lewis (Supervisor) & Ulrich Reininghaus (Supervisor)

Cite this