Discussion

In this study, we used the iPSYCH population-based case–cohort sample to examine how PGS for major psychiatric disorders are associated with age- and gender-specific risk of diagnosis across five disorders. We highlight three key findings: (a) differential risk according to age and type of psychiatric disorder; (b) the strong association between the PGS for MDD and the risk of all psychiatric disorders investigated; and (c) the gender differences that shape the strength and timing of risk associated with PGS.

First, we found that the association between disorder-specific PGS and diagnosis was strongest during adolescence and early adulthood, particularly during the late teenage years for depression and in the early twenties for schizophrenia and bipolar disorder, with decreasing estimates up to age 35 years. This pattern suggests that genetic liability may play a stronger role in early-onset presentations, whereas later diagnoses may be increasingly shaped by environmental exposures. ^{Reference Kwong, López-López, Hammerton, Manley, Timpson and Leckie24} Although associations at the youngest ages (11–13 years) were weaker, this likely reflects limited sample sizes, the rarity of childhood-onset mood disorders and the possibility that such cases are driven more by trauma or other non-genetic risk factors. ^{Reference Saveanu and Nemeroff25} Taken together, our findings align with the hypothesis that child-onset depression may have a partly distinct aetiology from adolescent- or adult-onset forms, an idea warranting further exploration. For childhood disorders, the association between genetic risk and autism diagnosis declined with age in males, whereas in females peak genetic risk occurred at later ages, perhaps reflecting a delay in diagnosis attributable to societal factors. ^{Reference Hull and Mandy26} In contrast, ADHD-PGS showed a stronger association with adult-onset diagnosis in both males and females. Evidence from studies conducted using the iPSYCH case–cohort sample has consistently shown a higher genetic burden in adult cases of ADHD ^{Reference Rajagopal, Duan, Vilar-Ribó, Grove, Zayats and Ramos-Quiroga27} , whereas studies in other cohorts report stronger genetic associations in childhood cases. ^{Reference Niarchou, Sealock, Straub, Sanchez‐Roige, Sutcliffe and Davis28} Given the scarcity of data on adult cases of childhood-onset disorders, further research is needed. Moreover, evolving diagnostic criteria for these disorders add complexity to drawing definitive conclusions. ^{Reference Sanders, Thomas, Glasziou and Doust29}

Second, we examined cross-disorder associations and found that the PGS for MDD was consistently associated with multiple psychiatric diagnoses. Interestingly, the PGS for MDD was associated not only with MDD but also with diagnoses of schizophrenia, autism and ADHD, and in some subgroups (e.g. females with schizophrenia and individuals with late-diagnosed autism), we observed stronger associations than those with the disorder-specific PGS. The overlap in predicting risk observed among the PGS for SCZ, BIP and MDD is not unexpected given their shared genetic architecture and high rates of misdiagnosis or symptom overlap. ^{Reference Doherty and Owen30–Reference Tzur Bitan, Grossman Giron, Alon, Mendlovic, Bloch and Segev33} However, we found little evidence of cross-associations between autism and ADHD, despite prior reports of shared genetic variants, ^{Reference Mattheisen, Grove, Als, Martin, Voloudakis and Meier34} suggesting that the common liability may manifest more indirectly through mood disorder liability.

Lastly, we observed important gender differences in the associations between PGS and diagnoses. For schizophrenia, the association with the SCZ-PGS was substantially higher in males and, as mentioned above, we observed that the MDD-PGS in females was more strongly associated with schizophrenia than the SCZ-PGS itself. These results are consistent with prior findings that the genetic architecture of schizophrenia may differ by gender, ^{Reference Mitjans, Papiol, Fatjó-Vilas, González-Peñas, Acosta-Díez and Zafrilla-López35} with female cases often showing more affective symptom profiles and possibly sharing liability with depressive disorders. ^{Reference Sommer, Tiihonen, van Mourik, Tanskanen and Taipale36} Although current schizophrenia GWAS indicate a genetic correlation of approximately 1 between male and female cases, suggesting no significant gender-specific differences in common variant architecture, ^{Reference Trubetskoy, Pardiñas, Qi, Panagiotaropoulou, Awasthi and Bigdeli17} existing PGS may nonetheless preferentially capture male-predominant risk profiles and thereby underestimate genetic liability in females. Similar gender-dependent variability in PGS performance has been described in population-based analyses, ^{Reference Mostafavi, Harpak, Agarwal, Conley, Pritchard and Przeworski37} where prediction accuracy differed between males and females even within the same ancestry group, potentially reflecting gender imbalances in the GWAS discovery sample sizes. This variation may reflect multiple factors, including environmental variance, indirect genetic effects and assortative mating. For childhood disorders, our results also support gender-specific dynamics. In autism, genetic risk was most strongly associated with early diagnosis in males but shifted later and remained stronger in females, consistent with the hypothesised female protective effect ^{Reference Wigdor, Weiner, Grove, Fu, Thompson and Carey10} and/or gender-biased diagnostic criteria that might delay female diagnoses. ^{Reference Hull and Mandy26} Similarly, in ADHD, stronger PGS associations were observed at older ages, particularly among females, who may face diagnostic delays due to subtler symptom presentation and comorbidities such as anxiety or depression. ^{Reference Taylor, Lichtenstein, Larsson, Anckarsäter, Greven and Ronald38,Reference Martin, Taylor, Rydell, Riglin, Eyre and Lu39} Our findings highlight the importance of accounting for gender-specific presentation and diagnostic practices in both research and clinical contexts, as differences in diagnostic rates between males and females may also contribute to the observed variability in PGS performance.

In summary, our findings indicate that psychiatric disorders with younger age at onset have greater polygenic liabilities, with gender-specific patterns in schizophrenia and childhood-onset disorders. Our study underscores the complexity of disentangling genetic liability from age at diagnosis in psychiatric disorders while highlighting the limitations of current GWAS approaches that assume independence between these traits. Future research should carefully consider gender-dependent heterogeneity in both genetic risk and age at diagnosis to improve the interpretation of polygenic liability across the life course.