Interviewer-rated positive dimension

In contrast to the lack of association for the current interviewer-rated positive psychotic experience dimension, positive schizotypal experiences, as assessed by interview, have been associated with higher schizophrenia PRS derived from an earlier GWAS^{Reference Van Os, Van der Steen, Islam, Gülöksüz, Rutten and Simons26} (n = 523). The reasons for variation in results are not clear, but the schizotypy study had a broader age range (inclusion criteria 16–50 years, mean age 31.1 years (s.d. 10.7)) and, in addition to psychotic experiences, included, for example, magical ideation and derealisation.

In ALSPAC, there was no association between positive psychotic experiences from interviews at 12 or 18 years and an earlier schizophrenia PRS,^{Reference Jones, Stergiakouli, Tansey, Hubbard, Heron and Cannon13} but there was an association for self-rated positive psychotic experiences at 16 years.^{Reference Jones, Heron, Hammerton, Stochl, Jones and Cannon14} In addition to differences in assessment method (self-rated versus interview) and time frame, there were some differences in participants between the 16- and 24-year assessments, but we did not have the 16-year PLIKS questionnaire data to investigate further. However, consistent with the current study, there was no association with schizophrenia PRS in a combined analysis of self-rated positive psychotic experiences in ALSPAC at 16 years and two adolescent twin samples.^{Reference Pain, Dudbridge, Cardno, Freeman, Lu and Lundstrom23} This is because participants without psychotic experiences often had a schizophrenia PRS similar to those with high psychotic experience scores.

There is an association between the presence of self-rated positive psychotic experiences and schizophrenia PRS in a larger sample of middle-/older-aged adults, but of small effect size (odds ratio 1.09 [95% CI 1.06–1.12]) and non-specific, showing similar associations with other PRSs.^{Reference Legge, Jones, Kendall, Pardiñas, Menzies and Bracher-Smith24}

In clinical samples, most studies of probands with schizophrenia show a lack of association between the number/severity of positive symptoms and schizophrenia PRS,^{Reference Fanous, Zhou, Aggen, Bergen, Amdur and Duan4,Reference Legge, Cardno, Allardyce, Dennison, Hubbard and Pardiñas6,Reference Taylor, de Vries, van Loo and Kendler16,Reference Ahangari, Bustamante, Kirkpatrick, Nguyen, Verrelli and Fanous17} but there is an association in bipolar spectrum disorders, especially for mood-incongruent psychotic symptoms.^{Reference Allardyce, Leonenko, Hamshere, Pardiñas, Forty and Knott19,Reference Allardyce, Cardno, Gordon-Smith, Jones, Di Florio and Walters20} This may be because of greater variation in positive symptoms and schizophrenia liability within bipolar than schizophrenia samples.^{Reference Owen, Legge, Rees, Walters and O’Donovan15}

Some studies of broad psychosis spectrum samples also show associations between the number/severity of positive symptoms and schizophrenia PRS,^{Reference Quattrone, Reininghaus, Richards, Tripoli, Ferraro and Quattrone5,Reference Smigielski, Papiol, Theodoridou, Heekeren, Gerstenberg and Wotruba18} but not consistently.^{Reference Jonas, Lencz, Li, Malhotra, Perlman and Fochtmann21,Reference Sengupta, MacDonald, Fathalli, Yim, Lepage and Iyer22} Considerable methodological heterogeneity makes comparison of these studies difficult.

In the current study, positive psychotic experiences were associated with lower mental well-being, depressive and anxiety disorders and lower childhood IQ, consistent with other studies.^{Reference Siebald, Khandaker, Zammit, Lewis and Jones34,Reference Kelleher and Cannon36,Reference Cardno, Selzam, Freeman and Ronald37} Positive psychotic experiences have also been associated with other risk factors, including childhood adversity and cannabis use.^{Reference Kelleher and Cannon36,Reference Cardno, Selzam, Freeman and Ronald37}

There were no associations of global hallucination or delusion experiences, paranoid experiences or psychotic disorder with schizophrenia PRS. First-rank delusion experiences were nominally associated, but important caveats are the very small number of participants with these experiences (n = 5 (0.2%) with PRS), a wide confidence interval and the risk of a false positive result from multiple testing in this additional secondary analysis.

The concept of first-rank symptoms was devised with a view to identifying relatively specific symptoms of schizophrenia, but these have been de-emphasised in ICD-11³⁸ and DSM-5-TR,³⁹ for reasons including that they are also found in other conditions such as mania. However, as noted above, there is evidence that, within bipolar spectrum disorders, mood-incongruent psychotic symptoms, which include first-rank symptoms, are associated with higher schizophrenia PRS.^{Reference Allardyce, Leonenko, Hamshere, Pardiñas, Forty and Knott19}

In view of this, the exploratory and highly provisional association between first-rank delusions and schizophrenia PRS is interesting, but requires independent confirmation.

Interviewer-rated negative dimension

In contrast to the current findings, parent-rated negative symptoms were associated with higher scores on an earlier schizophrenia PRS in the combined analysis of ALSPAC assessments at 16 years and two adolescent twin samples.^{Reference Pain, Dudbridge, Cardno, Freeman, Lu and Lundstrom23}

Associations have also been found between negative/disorganised schizotypal features, as assessed by interview, and schizophrenia PRS,^{Reference Ahangari, Bustamante, Kirkpatrick, Nguyen, Verrelli and Fanous17,Reference Van Os, Van der Steen, Islam, Gülöksüz, Rutten and Simons26} but it is unclear whether it was the negative, disorganised or both phenomena that contributed to the association.

In clinical samples of schizophrenia and other psychotic disorders, negative symptom dimensions have frequently been associated with genetic loading for schizophrenia, including PRS.^{Reference Taylor, de Vries, van Loo and Kendler16} Again there is a caveat that the negative dimensions may include disorganised symptoms. Some studies that attempted to disaggregate these symptoms,^{Reference Fanous, Zhou, Aggen, Bergen, Amdur and Duan4} or examined the independence of their associations,^{Reference Legge, Cardno, Allardyce, Dennison, Hubbard and Pardiñas6} found that the association was predominantly with disorganised symptoms.

A further complication in clinical samples is that negative symptoms could be markers of primary developmental aetiological processes and/or secondary to having a psychotic disorder – for example, to the effects of positive symptoms, medication or social isolation.

In a study of adolescent twins at 16 years, i.e. prior to the main age at onset range of psychotic disorders, parent-rated negative symptoms (presumed to be predominantly primary symptoms) were associated with male gender, lower parental socioeconomic status, cumulative pre/perinatal complications, lower cognitive functioning in childhood and lower educational attainment.^{Reference Cardno, Selzam, Freeman and Ronald37} The association with male gender is also often found in clinical samples of psychotic disorders, along with indices of relatively severe chronic illness^{Reference Cardno, Allardyce, Bakker, Toulopoulou, Kravariti and Picchioni8} and cognitive impairment.^{Reference de Dominguez, Viechtbauer, Simons, Van Os and Krabbendam40}

In the current study there was an association with lower social rapport, which could result in social isolation, and a nominal association with lower childhood IQ; otherwise, the pattern of associations was not strikingly similar to those in adolescent population or in clinical studies. It is unclear, therefore, how much the one-off interviewer observations at 24 years are able to identify the same phenotypes as in the other studies.

Interviewer-rated disorganised dimension

Regarding the interviewer-rated disorganised dimension, there has been little comparable research in general population samples. As mentioned above, associations have been found between interview-based negative/disorganised schizotypal phenotypes and higher schizophrenia PRS,^{Reference Ahangari, Bustamante, Kirkpatrick, Nguyen, Verrelli and Fanous17,Reference Van Os, Van der Steen, Islam, Gülöksüz, Rutten and Simons26} but it is unclear which phenomena contributed to the association.

In clinical schizophrenia and psychosis samples, there is relatively consistent evidence that the disorganised symptom dimension has substantial heritability^{Reference Rietkerk, Boks, Sommer, Liddle, Ophoff and Kahn9,Reference Rijsdijk, Gottesman and Cardno10} and is associated with higher schizophrenia PRS.^{Reference Fanous, Zhou, Aggen, Bergen, Amdur and Duan4,Reference Legge, Cardno, Allardyce, Dennison, Hubbard and Pardiñas6,Reference Ahangari, Bustamante, Kirkpatrick, Nguyen, Verrelli and Fanous17}

In population samples there are investigations of self-rated cognitive disorganisation, which shows association with schizophrenia PRS.^{Reference Pain, Dudbridge, Cardno, Freeman, Lu and Lundstrom23} However, because there is little correlation between self- and observer-rated disorganised symptoms in clinical samples,^{Reference Cardno, el din Nasser, du Toit, Rehman, McIlrae and Variend41} the relevance of the self-rated cognitive disorganisation studies to the current interviewer-rated disorganised dimension is unclear.

In clinical samples, the disorganised symptom dimension is associated with lower cognitive functioning,^{Reference de Dominguez, Viechtbauer, Simons, Van Os and Krabbendam40} including lower premorbid IQ,^{Reference Cardno, Allardyce, Bakker, Toulopoulou, Kravariti and Picchioni8} which is consistent with the nominal association with lower childhood IQ in the current study. In addition, the apparently odd speech and behaviour of people with disorganised symptoms is consistent with the association with lower social rapport in the current study.

Nevertheless, the lack of association with schizophrenia PRS in the current study raises the question of how similar or different are the aetiological processes underlying interviewer-observed disorganised phenomena at a single time point in the general population compared with clinically observed disorganised symptoms, which are often clearest during acute psychotic episodes.

Self-rated negative dimension

In contrast to the results for the current self-rated negative dimension based on the CAPE questionnaire at 24 years, there was an association between the self-rated negative dimension based on CAPE at 16 years in ALSPAC and an earlier schizophrenia PRS.^{Reference Jones, Stergiakouli, Tansey, Hubbard, Heron and Cannon13,Reference Jones, Heron, Hammerton, Stochl, Jones and Cannon14} As with the positive dimension, we did not have the CAPE data at 16 years to further investigate this difference.

In other general population studies of the CAPE self-rated negative dimension, results have also been inconsistent, in that both association^{Reference Quattrone, Reininghaus, Richards, Tripoli, Ferraro and Quattrone5} and absence of association^{Reference Van Os, Van der Steen, Islam, Gülöksüz, Rutten and Simons26} with schizophrenia PRS have been found.

In contrast to the lack of association with schizophrenia PRS in the current study, there were substantive associations with depression, anxiety, neuroticism and autism PRSs, although not of sufficient effect size to be clinically actionable. Association with neuroticism PRS was also found at 16 years in ALSPAC.^{Reference Jones, Heron, Hammerton, Stochl, Jones and Cannon14} Self-rated negative schizotypal experiences have also been associated with depression and neuroticism PRSs in another general population sample.^{Reference Docherty, Shabalin, Adkins, Mann, Krueger and Bacanu25}

Consistent with the PRS results, higher scores on the self-rated negative dimension were associated with the presence of depressive and generalised anxiety disorders, as well as with lower mental well-being and social rapport, at 24 years.

There was no correlation between interviewer- and self-rated negative dimensions in the current study, which suggests that they may be assessing phenotypes with notably different aetiologies. Low correlations between self- and observer-rated negative symptoms have also been found in some clinical samples.^{Reference Cardno, el din Nasser, du Toit, Rehman, McIlrae and Variend41}

In summary, self-rated negative symptoms were not associated with schizophrenia PRS in the current study, and show inconsistent associations in other studies. They may be influenced by genetic risk variants for other disorders, but confirmation in further independent samples would be valuable.

Strengths and limitations

To our knowledge, this is the first birth cohort study to assess distinct positive, negative and disorganised psychotic experience dimensions, including the use of interviewer observations for negative and disorganised symptoms, and additionally using self-rated negative symptoms, in large sample sizes of 2500+ at 24 years.

Despite assessment around the peak age at onset of schizophrenia,^{Reference Tandon, Nasrallah, Akbarian, Carpenter, DeLisi and Gaebel7} clearly present interview-based psychotic experiences remained infrequent. This limited statistical power and exploration of the dimension construct, especially regarding the gradient of scores including co-occurring experiences.

Results could have differed if we had used other definitions of psychotic experience dimensions. We explored the use of factor scores, but were concerned about the stability of factors given the low frequency of some experiences, which also precluded splitting the sample for exploratory and confirmatory factor analysis. Furthermore, the use of simple phenotypes based on whether or not experiences were clearly present simplified interpretation and comparison with similar definitions in clinical samples.^{Reference Cardno, Allardyce, Bakker, Toulopoulou, Kravariti and Picchioni8} These simple phenotypes also show similar levels of familial aggregation to phenotypes based on factor scores in clinical samples.^{Reference Cardno, Jones, Murphy, Sanders, Asherson and Owen2,Reference Cardno, Sham, Murray and McGuffin3}

The results could also have differed if we had used other approaches to generating PRSs. However, lack of association of positive psychotic experiences and negative symptoms with schizophrenia PRS was also found in a larger sample of adolescent twins using other PRS approaches.^{Reference Cardno, Selzam, Freeman and Ronald37,Reference Maxwell, Ronald, Cardno, Breen, Rimfeld and Vassos42} Additionally, the approach in the current study facilitated comparison with other studies, most of which also used the GWAS discovery sample P-value threshold approach.

The sample size was substantial, but combination with other samples in meta-analyses may be required to detect associations of small effect.

There were considerable missing data, due to participant attrition between inception and 24 years. Individuals were more likely to have missing data at 24 years if they were male, had greater socioeconomic disadvantage or had had more severe psychotic experiences at 18 years.^{Reference Sullivan, Kounali, Cannon, David, Fletcher and Holmans12} Observed and imputed estimates are very similar for the frequency of any definite positive psychotic experience and psychotic disorder at 12–24 years,^{Reference Sullivan, Kounali, Cannon, David, Fletcher and Holmans12} but biased estimates are still possible, including because of data not missing at random.

There is also an association between non-participation and higher schizophrenia PRS in ALSPAC, leading to reduced power for analyses of schizophrenia and genetically related psychiatric phenotypes, and potentially biased estimates, even with imputation due to data not missing at random.^{Reference Martin, Tilling, Hubbard, Stergiakouli, Thapar and Davey Smith43}

Inverse probability weighting may also be used to adjust for missing data, but issues include the assumption that, for example (if being male is a predictor of missing data), male non-participants would have similar associations to male participants.^{Reference Seaman and White44}

The genetic analysis was confined to associations with PRS and did not include rarer schizophrenia risk variants.^{Reference Owen, Legge, Rees, Walters and O’Donovan15}

The sample was predominantly of White European ancestry, and exclusively so for the PRS analyses, which minimised the risk of confounding by population stratification. However, further studies in more ethnically diverse samples would be valuable, which is a major focus of the Psychiatric Genomics Consortium (https://pgc.unc.edu).

The results of this study suggest that positive, negative and disorganised experience/symptom dimensions, as defined here and assessed in this general population cohort of young adults, are not strongly associated with polygenic liability to schizophrenia, although true associations hidden by methodological constraints are not excluded, e.g. due to small effect sizes, missing data, low phenotype base rates or misclassification, or PRS limitations.

Positive psychotic experiences may be (modestly and non-specifically) influenced by commonly occurring risk alleles for schizophrenia, with PRS associations seen more from middle age onwards, as people with high schizophrenia liability but no lifetime psychotic experiences become fewer over time. Future research in ALSPAC and other samples as participants age may clarify whether this is the case.

In clinical samples, schizophrenia PRS is most consistently associated with disorganised and/or negative symptoms. The interviewer-rated negative and disorganised symptoms in the current study probably have notably different aetiologies. Also, the self-rated negative symptoms may be better indicators of social withdrawal/low motivation due to polygenic liability to depression, anxiety, neuroticism or autism than schizophrenia.

The highly provisional secondary finding, that first-rank delusions might be indicators of higher polygenic liability to schizophrenia, is of interest, but needs confirmation in other samples where information on first-rank symptoms is available before substantive conclusions can be drawn.