Family-level Brain Pattern Similarities and Their Clinical Significance in Young Offspring of Individuals with Psychotic Disorders

Abstract

Objective

Familial high-risk offspring inherently provide an unparalleled opportunity for individual-level risk prediction in psychotic disorders: a personalized forecast of their potential brain pathology should they develop psychosis, i.e. their own parents’ brain pathology. However, whether brain patterns in offspring resembling those of their affected parents have clinical significance remains unclear. This study aims to quantify these parent-offspring brain similarities in psychotic disorders and explore their clinical significance.

Methods

Using 3T imaging data from the Bipolar Schizophrenia Network on Intermediate Phenotypes (BSNIP)-1 Study, cortical thickness similarities between offspring (ages 15-35, N=73) and their own affected parents with psychosis were quantified using a novel Familial Vulnerability Index (FVI), calculated using region-wise z-scores. FVIs were compared to the ENIGMA-derived cortical Regional Vulnerability Index (RVI) in offspring. FVIs and RVIs were examined in offspring with and without psychiatric symptoms, and relations to global and cognitive functioning were assessed.

Results

Offspring had higher FVIs than RVIs (q<0.001, d=0.78). Offspring with a psychotic disorder (q=0.001, d=2.84) or any other psychiatric diagnosis (q=0.02, d=1.40) had higher FVIs than offspring with no psychiatric diagnosis, while RVIs did not differ between these groups. FVIs, not RVIs, were negatively correlated with global functioning (r =−0.31, q=0.047). Neither FVIs nor RVIs were correlated with cognition.

Conclusion

This proof-of-concept study demonstrates the clinical significance of FVI in young offspring of individuals with psychotic disorders. Future research on the clinical utility of family-level similarities across modalities holds untapped potential to advance risk and prognostic prediction in families affected by serious mental illness.

INTRODUCTION

Psychotic disorders cause significant disability worldwide, leading to suffering for individuals and impacting healthcare systems (1,2). Given the substantial personal and public health burden associated with psychosis, along with limited treatment options, there is a pressing need for specific and accurate risk prediction approaches. Identifying novel risk indicators with high predictive clinical utility is critical for early intervention and potentially prevention in psychotic disorders (3).

A key obstacle toward identifying specific risk indicators for psychotic disorders is the biological heterogeneity observed among both those affected and individuals considered at risk (4). Previous research illustrates significant interindividual variability in brain biomarkers, including neuroimaging markers, among individuals with psychotic disorders (5-7). This biological diversity is further compounded by variability in the sources of risk, including genetic, neurodevelopmental, and environmental influences, which may converge on shared downstream clinical phenotypes, consistent with frameworks such as the watershed model (8-10). This obstacle stimulated efforts to leverage interindividual variability to develop individual-level risk predictors (11,12). One approach involves the use of normative models, which quantify deviations from brain measures observed in the general population (13,14). Another approach quantifies the similarity of an individual’s brain measures to the average changes observed in psychiatric disorders, based on large meta-analyses performed by the Enhancing Neuroimaging Genetics Meta Analyses (ENIGMA) Consortium, known as “Regional Vulnerability Indices” (RVIs)(15-17). A similar approach is Neuroscore, which combines normalized brain measures weighted by ENIGMA effect sizes into a single summary score to achieve an individual-level risk marker (18). Although instrumental in bringing individual-level perspectives, these studies draw from group-level, meta-analytic findings and do not fully account for biological heterogeneity in the context of psychopathology. As a result, achieving robust individual-level prediction remains a significant challenge.

However, one risk group offers a unique opportunity to tackle the heterogeneity problem that impedes individual-level risk prediction: Young individuals whose family members experience psychotic disorders (familial high-risk, FHR). This group represents a well-defined risk population, exhibiting not only significantly higher rates of psychotic disorders (~12%) (19,20), but also a nearly 50% likelihood of developing a psychiatric disorder by age 20 (19). At the population level, the FHR approach has a sensitivity of approximately 6.5% for identifying future psychotic disorders, comparable to that of the clinical high-risk framework (21). Importantly, unlike other risk approaches, FHR youth inherently comes with a personalized forecast of their potential future brain pathology should they develop psychosis based on their familial risk. This forecast is the brain changes documented in their own affected family members. Decades of research has established the familial aggregation of brain changes seen in psychotic disorders (22-24). First-degree relatives frequently exhibit biomarker patterns similar to their affected relatives, with increased similarity if the relative also has clinical symptoms (24). Thus, biomarker similarities between an at-risk youth and their affected family member may offer a family-specific signal for individualized risk prediction, bypassing the noise associated with broader group-level heterogeneity. This study aims to answer a straightforward yet novel question: Do biomarker similarities between an affected individual and their at-risk young family member have clinical significance as individual-level risk predictors?

To address this critical question, we quantified the family-level similarities in cortical thickness measures between young offspring (ages 15-35, N=73) and their own affected parents with psychotic disorders within the Bipolar and Schizophrenia Network for Intermediate Phenotypes (BSNIP)-1 dataset, leading to a novel Familial Vulnerability Index (FVI). Separately, we quantified the similarities between offspring’s cortical thickness measures and average thickness alterations seen in unrelated adults with schizophrenia using the established RVIs derived from meta-analytic findings by ENIGMA (15). We chose cortical thickness as the initial focus for examining parent–offspring brain similarity, given its well-established association with psychosis risk in both familial and clinical high-risk populations (25,26). We compared cortical FVIs and RVIs in offspring with and without psychiatric diagnoses as our primary aim. As a secondary aim, we examined associations of FVIs and RVIs with general, social, and cognitive functioning. We hypothesized that offspring have higher FVIs than RVIs, suggesting higher brain similarity to own parents’ brain patterns compared to average changes seen in psychotic disorders. We predicted that offspring with psychiatric disorders, particularly those with psychotic disorders, would show higher FVIs, reflecting greater similarity to the affected parent’s brain structure, compared to those without psychiatric diagnoses, and that this differentiation would be more pronounced for FVI than for RVI. Finally, we predicted a stronger correlation between general and cognitive functioning and FVIs than for RVIs.

METHODS

Study Sample and Clinical Assessments

90 young offspring (ages 15-35) and their parents with psychotic disorders (82 probands) with available imaging data were identified in the BSNIP-1 cohort and included in the analysis, along with 455 healthy individuals. After offspring and parents with missing data were excluded, FVIs were calculated for 79 offspring-parent dyads, and RVIs were computed for the same offspring. Data from six offspring were subsequently removed due to outlier FVI or RVI values using the 1.5 interquartile range (IQR) rule. 73 offspring and their 68 affected parents (18 probands with schizophrenia, 19 probands with schizoaffective disorder, and 31 probands with psychotic bipolar disorder) were included in the final analysis. 3 probands had multiple offspring (8 total) in the final sample. Among the 73 offspring included in the final analysis, 19 had no psychiatric diagnosis, 44 had at least one non-psychotic psychiatric disorder, and 10 had a psychotic disorder. Among the 6 excluded offspring, 2 had no psychiatric diagnosis, 2 had non-psychotic psychiatric disorders, and 2 had psychotic disorders. Healthy individuals had no history of psychotic disorder, bipolar disorder, or recurrent major depressive disorder, and no family history of schizophrenia-bipolar spectrum disorders in first- or second-degree relatives.

BSNIP-1 procedures, detailed elsewhere (27), were approved by Institutional Review Boards, and all participants provided written informed consent. Psychiatric diagnoses were assessed using the Structured Clinical Interview for DSM- IV Axis I disorders (28). Global and social functioning were assessed using the Global Assessment of Functioning (GAF) (28) and Birchwood Social Functioning Scales (SFS) (29). Cognition was assessed using the Brief Assessment of Cognition in Schizophrenia (BACS) (30). See Table 1 for demographic and clinical characteristics.

Table 1.

Demographic and clinical characteristics of young offspring (ages 15-35), their affected parents (probands) with psychotic disorders, and healthy control participants within the BSNIP-1 cohort included in the final analysis

	Offspring (N=73)	Affected parents (Probands) (N=68)	Healthy controls (N=455)
Age (Mean, years)	21.6	47.3	36.6
Age (SD)	5.67	8.29	12.6
Sex (n)
Female	37	56	247
Male	36	12	206
Race and Ethnicity (n)
Caucasian	33	30	279
African American	34	34	132
Other	6	4	44
Parent diagnoses (n)
Schizophrenia	-	18	-
Schizoaffective Disorder	-	19	-
Psychotic Bipolar Disorder	-	31	-
Offspring diagnoses (n)
No psychiatric disorder	19	-	-
Non-psychotic psychiatric disorder	44	-	-
Major Depressive Disorder	17
Bipolar Disorder (without psychotic features) and Other Mood Disorders	10
Other psychiatric disordersa	17
Psychotic disorder	10	-	-
Daily Chlorpromazine Equivalentb (Mean)	447.9	411.4
Daily Chlorpromazine Equivalent (SD) b	638.4	335.7
Duration of Illness (Mean) c	9.11	26.7
Duration of Illness (SD) c	3.98	10.08
GAF score (Mean)	71.7	53.1	85.7
GAF score (SD)	13.8	11.8	7.36
SFS total score (Mean)	141.96	123.77	154.7
SFS total score (SD)	21.03	25.90	17.7
BACS composite z-score (Mean)	−0.607	−1.24	−0.02
BACS composite z-score (SD)	1.06	1.28	1.17

GAF: Global Assessment of Functioning; SFS: Birchwood Social Functioning Scale; BACS: Brief Assessment of Cognition in Schizophrenia; SD: Standard deviation.

^aOther psychiatric disorders include Substance Use Disorders, Anxiety Disorders, Trauma and Stressor-Related Disorders, Obsessive Compulsive Disorder, Eating Disorders, Attention Deficit and Hyperactivity Disorder.

^bBased on participants with available antipsychotic use data (42 probands and 7 offspring).

^cYears since the first symptoms of psychotic disorder, presented only for parents and offspring with psychotic disorders.

Structural Magnetic Resonance Imaging

T1-weighted 3T images were acquired at six BSNIP sites using three different scanner platforms (GE, Siemens, Philips) with similar sequences based on the Alzheimer’s Disease Neuroimaging Initiative protocol (http://adni.loni.usc.edu). Structural data was processed using FreeSurfer 6.0. Cortical thickness for 34 bilateral Desikan-Killiany (DK) atlas regions were quantified following segmentation and visual inspection for motion artifacts and proper segmentation (31).

Familial and Regional Vulnerability Indices

Statistical analyses were performed using IBM SPSS 29 and R version 4.3.3. Bilateral cortical thickness measures from the 34 DK Atlas regions were averaged across hemispheres, yielding 34 mean cortical thickness values. Covariates, including age, sex, race, site, and intracranial volume (ICV) were regressed out from these measures and unstandardized residual measures were saved, in line with the RVI calculation method described by Kochunov et al. (15,16). In offspring and parents, z-score normalization was applied to these measures using the mean (M) and standard deviation (SD) derived from healthy controls in the BSNIP-1 cohort. These normalized values were then used to compute FVI and RVI, reflecting similarity in structural patterns, defined here as the spatial profile of region-wise cortical thickness across the cortex.

For each offspring, a cortical thickness-based Familial Vulnerability Index (FVI) was calculated as a Pearson correlation coefficient between the 34 region-wise z-values in the offspring and the corresponding z-values in their own parent.

Separately, for each offspring, a Schizophrenia Spectrum Disorder - Regional Vulnerability Index (RVI_ENIGMA) was calculated as a Pearson correlation coefficient between 34 region-wise z-values in the offspring and the corresponding effect sizes for cortical thickness differences between schizophrenia and control groups from the ENIGMA meta-analysis, as defined by Kochunov et al. (32-34).

Given that the BSNIP cohort includes not only probands (affected parents) with schizophrenia, but also those with schizoaffective disorder and psychotic bipolar disorder, we calculated a separate RVI (RVI_BSNIP) for each offspring using the corresponding effect sizes for cortical thickness differences between probands and healthy controls in the B-SNIP dataset (31) instead of ENIGMA effect sizes to more accurately represent the average pathology across these three diagnoses.

Statistical analysis

To compare FVI, RVI_ENIGMA, and RVI_BSNIP values in the offspring, a linear mixed-effects model was fitted using the lme4 and lmerTest packages in R, with Index (FVI, RVI_ENIGMA, RVI_BSNIP) as a fixed effect and family code and participant ID as random intercepts, with participant ID included to account for multiple observations (multiple indices) per participant. Each index was then compared separately across offspring diagnostic groups using linear mixed-effects models, with diagnostic group as a fixed effect and family code as a random intercept, in two separate analyses: (1) offspring with any psychiatric disorder (including both psychotic and non-psychotic disorders) versus those with no psychiatric diagnosis, and (2) offspring with no psychiatric diagnosis, non-psychotic psychiatric disorders, and psychotic disorders. Estimated marginal means, pairwise differences, and standardized mean differences (Cohen’s d) were computed. For pairwise comparisons, p values and 95% confidence intervals (CI) were adjusted for multiple comparisons using Benjamini–Hochberg false discovery rate (FDR). FDR-corrected p (q) and CI values were reported.

First, the correlations between FVI, RVI_ENIGMA, and RVI_BSNIP were assessed using linear mixed-effects models, with family code as a random factor. Then FVI’s and RVI_ENIGMA’s associations with global, social, and cognitive functioning measures (GAF, SFS total scores, and BACS composite score) were examined using linear mixed-effects models, with the index as the fixed-effect predictor, the clinical or cognitive measure as the dependent variable, and family code as a random intercept. Correlation coefficients (r) were derived from the t-statistics for the predictor. FDR-corrected p (q) values were reported (corrected for 6 comparisons, 2 indices x 3 scores). Given the strong correlation between RVI_ENIGMA and RVI_BSNIP, RVI_BSNIP was not included in the FDR correction. However, RVI_BSNIP’s correlations with functioning measures are provided in the Supplement (Table S7).

Sensitivity Analyses

To assess the robustness of our findings, these analyses were repeated (1) in all 79 offspring, including the offspring with outlier FVI or RVI values and (2) in a subsample including only one randomly selected offspring per family (N=69). Additionally, FVIs were recalculated without regressing out the covariates (total intracranial volume, age, sex, race, or site) to mitigate potential bias associated with adjusting for covariates using linear regression (35). The results of these analyses are presented in the Supplement.

Exploratory analyses examining the effect of parental diagnoses and the discriminative performance of FVI and RVI_ENIGMA were also conducted and are reported in Supplement.

RESULTS

FVI and RVI differences in offspring

Offspring had significantly higher FVIs than RVI_ENIGMA (β = 0.123, 95% CI [0.060,0.186], t = 4.72, q < 0.001, d = 0.78) and RVI_BSNIP (β = 0.122, 95% CI [0.059,0.19], t = 4.68, q < 0.001, d = 0.77), while there were no differences between RVI_ENIGMA and RVI_BSNIP (q = 0.97) (Figure 1, Table S1).

Figure 1.

Boxplots with individual datapoints demonstrating Familial Vulnerability Index (FVI) and ENIGMA- and BSNIP-derived Regional Vulnerability Index (RVI_ENIGMA and RVI_BSNIP) values in offspring. q = FDR-corrected p value, d = Cohen’s d effect size.

Comparisons of FVIs/RVIs across diagnostic groups among offspring

Offspring with psychiatric disorders had significantly higher FVIs than those without psychiatric diagnoses (β = 0.158, 95% CI [0.057,0.26], t = 3.11, q = 0.008, d = 1.7), while RVI_ENIGMA and RVI_BSNIP did not differ between the two groups (q’s = 0.784) (Figure S1, Table S2). Offspring with psychotic disorders had higher FVIs than those with non-psychotic psychiatric disorders (β = 0.135, 95% CI [−0.025,0.296], t = 2.08, q = 0.041, d = 1.45) and those with no psychiatric disorders (β = 0.266, 95% CI [0.089,0.443], t = 3.71, q = 0.001, d = 2.84) (Figure 2, Table S3). Offspring with non-psychotic psychiatric disorders also had higher FVIs than those with no psychiatric disorders (β = 0.131, 95% CI [0.005,0.257], t = 2.55, q = 0.020, d = 1.40) (Table S3). RVI_ENIGMA and RVI_BSNIP did not differ between the three groups (Figure 2, Tables S4-S5).

Figure 2.

Boxplots of FVI (A), RVI_ENIGMA (B), and RVI_BSNIP (C) across the three diagnostic groups (No psychiatric disorders vs. non-psychotic psychiatric disorders vs. psychotic disorders) with individual datapoints. q = FDR-corrected p value, d = Cohen’s d effect size.

Associations between FVIs and RVIs

FVIs were not correlated with either RVI_ENIGMA (r =−0.062, q = 0.81) or RVI_BSNIP (r = −0.031, q = 0.81). There was a strong correlation between RVI_ENIGMA and RVI_BSNIP (r = 0.80, 95% CI [0.72, 0.86], q <0.001) (Figure 3).

Figure 3.

Scatterplots with regression lines and shaded 95% confidence intervals, displaying the correlations between FVI and RVI_ENIGMA (A), FVI and RVI_BSNIP (B), RVI_ENIGMA and RVI_BSNIP (C). r = correlation coefficient, q = FDR-corrected p value.

Associations between index values and global, social, and cognitive functioning

Higher FVI scores were associated with lower GAF scores (r = −0.31, 95% CI [−0.49, −0.09], q = 0.047). There was no correlation between RVI_ENIGMA and GAF scores (r = −0.02, q = 0.933) (Figure 4). Neither FVI nor RVI_ENIGMA scores were correlated with SFS total or BACS composite scores (see Supplement for full statistical results) (Table S6).

Figure 4.

Scatterplots with regression lines and shaded 95% confidence intervals, displaying the correlation between Global Assessment of Functioning (GAF) scores and FVI (A), RVI_ENIGMA (B), RVI_BSNIP (C). r = correlation coefficient, q = FDR-corrected p value.

In a post-hoc analysis, when offspring with psychotic disorders (n=10) were excluded, higher FVI scores were still correlated with lower GAF scores, though at a trend-level in this smaller sample (r = −0.23, p = 0.069).

Sex differences and relationship with age

In exploratory analyses, FVIs or RVIs (RVI_ENIGMA or RVI_BSNIP) were not different between male and female offspring, or between offspring with affected mothers vs. fathers. There was no significant difference in FVI between offspring who were sex-concordant (n=35) versus sex-discordant (n=38) with their parent.

None of the indices were significantly associated with age. Because age and sex were regressed out during FVI computation, we repeated these analyses using FVIs calculated without covariate adjustment (including age and sex; see Supplement). Results were consistent; FVI was not associated with age and did not differ by sex.

Sensitivity analyses

In sensitivity analyses, offspring with psychiatric disorders had higher FVIs than those without, both when all 79 offspring (including outliers) were included, when analyses were repeated using only one randomly selected offspring per family, and when FVIs were calculated without covariates. Additionally, the stepwise pattern of FVIs persisted across offspring with psychotic disorders, non-psychotic psychiatric disorders, and no psychiatric disorders. Please see the Supplementary Material for full results.

DISCUSSION

In this study, we quantified the family-level similarities in cortical thickness measures between young offspring (ages 15-35) and their own affected parents with psychotic disorders, leading to a novel Familial Vulnerability Index. We compared this novel individual-level index to the established Regional Vulnerability Index derived from ENIGMA meta-analytic findings. Our results demonstrated that young offspring showed greater structural pattern similarities to their own affected parents than to patterns found in schizophrenia-control differences in the ENIGMA meta-analysis or those observed in the overall BSNIP-1 psychotic disorders cohort, as evidenced by higher FVIs compared to RVIs in the offspring. Offspring with a psychotic disorder or any psychiatric disorder had higher FVIs than offspring with no psychiatric diagnosis, following a stepwise pattern, while RVIs did not differ between these groups, suggesting that FVIs may index disease risk more faithfully than RVIs. FVIs were negatively correlated with global functioning scores, indicating that greater brain similarity to the affected parent was linked to lower functioning. On the other hand, RVIs were not associated with GAF scores, highlighting the greater clinical significance of FVIs in offspring. Together, these results suggest that FVI is a novel, distinct, and promising metric for identifying brain patterns that are associated with psychosis risk at an individual level among youth at familial risk for psychosis.

Our findings suggesting greater brain pattern similarity to the affected parent align with evidence from the emerging field of “Intergenerational Neuroimaging”. Familial aggregation of various neuroimaging metrics has been established in large twin/sibling studies (23,36,37), parent-child studies(38-43), and extended pedigree studies (44-46) in healthy families and families affected by psychiatric illnesses, including psychotic disorders. Parent-child studies have measured the structural and functional brain similarity between a given sample of parents and children, and some investigated their relationship with a given phenotype (47). However, these studies investigated a given region of interest at a time by measuring the correlation between a group of parents and their children (39,40,47,48), not focusing on individual-level similarities.

Some recent studies have taken a different approach by directly measuring the concordance of several imaging metrics between parent-offspring pairs, thus obtaining a correlation for each dyad as an index of similarity (43,49-57). For example, Takagi et al. showed that it was possible to identify parent-child dyads based on functional connectivity and gray matter volume similarities (50). Recently, Matsudaira et al. demonstrated that correlations in brain structural patterns (cortical thickness, surface area, and local gyrification index) between parent-offspring dyads were significantly stronger than those between unrelated individuals (58), aligning with our findings.

Despite its promising application in psychiatric research for individual-level risk prediction, only a handful of studies have applied this approach in psychiatric populations. This limited research (59,60), focused on depression, investigated the intergenerational transmission of brain structure and function in depressed mothers and offspring. Minami et al. demonstrated correlations in brain structure in default mode and central executive networks in depressed mother and never-depressed daughter dyads, but not in mother-son, father-daughter, or mother-son dyads (53). Colich et al. investigated the correlation between mothers’ and daughters’ putamen response to the anticipation of loss, and its relation to maternal depression history (60). Although there is extensive literature showing structural and functional brain changes in offspring and siblings of individuals with psychotic disorders (23,31,61-64), individual-level concordance of brain measures between affected and unaffected family members has not been explored previously.

Our results demonstrate that FVIs are significantly associated with global functioning, but not with cognition. Consistent with this, intergenerational studies in healthy families have reported associations between parent–offspring brain similarities and behavioral phenotypes, which appear to strengthen in the context of impairments in those phenotypes. Takagi et al. demonstrated that children with higher developmental problem scores, based on Childhood Behavior Checklist scores, showed more brain similarities (based on gray matter volume and functional connectivity) to their parents than other children (50). Dimanova et al. showed that mother-child similarity of gray matter volume in neocortical parts of the corticolimbic circuitry (anterior cingulate, medial orbitofrontal areas) was correlated with the degree of similarity in mental wellbeing (54). Although studies on brain similarity in relationship to cognition are even more scarce (43,57), Constant-Varlet et al. demonstrated that brain activity in the insula and precentral gyrus during a mental arithmetic task was correlated in mothers and their 8-year-old children, with stronger correlations in dyads with mothers who have lower math skills than higher skills (57). These results suggest that examining FVIs based on specific regions and using different imaging modalities could provide novel markers of risk for distinct phenotypes across psychiatric disorders. The lack of association between FVI and cognition was contrary to our hypothesis. This analysis was limited to cortical FVI and the BACS composite score, both of which are summary measures, limiting the ability to capture more specific brain–behavior relationships. Future studies examining domain-specific cognitive measures in relation to regional or multimodal FVIs may provide greater insight into this relationship.

In our sample, cortical RVIs did not differ significantly between offspring with and without psychiatric illness, nor were they correlated with general or cognitive functioning. While Kochunov et al. observed significant cortical RVI differences between healthy controls and early-course schizophrenia (15), our sample showed no differences between offspring with psychosis and those without psychiatric disorders. This analysis, however, was limited by small sample size and the fact that both groups comprised offspring, who may share endophenotypic structural brain changes. In line with our findings, Kochunov et al. demonstrated that cortical RVIs were not associated with cognition in individuals with schizophrenia (although multimodal RVIs were). Although research on RVIs is limited in risk groups, there were no differences in RVIs between children with and without a family history of psychosis, despite a correlation between RVIs and family history loadings (10). Similarly, Karcher et al. showed that, in comparison to healthy controls, only children with persistent and distressing psychotic-like experiences (PLEs) had higher RVIs (16), while children with persistent nondistressing, transient distressing, and transient nondistressing PLEs exhibited no significant differences in RVIs compared to controls. Together, these suggest that RVIs may have higher specificity, but low sensitivity when applied broadly as a marker of psychosis risk. Our results indicate cortical FVIs have more clinical significance than cortical RVIs in familial high-risk groups and hold greater promise as potential risk indicators. Comparisons of multi-modal FVIs and RVIs might provide further insights into their potentially distinct clinical applications.

This study is a proof-of-concept study that demonstrates the clinical significance of parent-offspring brain similarities as potential individual-level risk indicators in young individuals from families affected by psychotic disorders. Although this report is limited to cortical thickness similarities in families with psychotic disorders, this approach has the potential to grow in depth and scope, paving the way for a new line of research within psychiatry. We introduce a preliminary framework we refer to as Family-Informed Precision Psychiatry, emphasizing the potential of within-family neurobiological patterns for personalized prediction. Future directions include developing family-level similarity indices using other structural measures (e.g. surface area, local gyrification index), other modalities (e.g. diffusion MRI, functional MRI, magnetic resonance spectroscopy, EEG, cognition), and investigating region-specific similarities as potential risk indicators. This line of research offers clinical translational potential not only for psychotic disorders, but also for other heritable neuropsychiatric conditions, such as mood, developmental, and neurocognitive disorders. The clinical significance of reduced FVIs in healthy parent-offspring dyads, reflecting increased dissimilarity, also warrants investigation as a potential indicator of emerging illness. Further, potential clinical applications beyond risk prediction warrant exploration, including the mechanistic implications of these similarities. Identifying the regions most affected in the parent and those that contribute most strongly to parent–offspring similarity may help define individual-level biological targets for prevention and treatment, as well as inform the development of novel markers of prognosis and treatment response in familial cases. Negative FVIs, as well as regions showing increased parent–offspring dissimilarity, warrant further investigation to determine whether they reflect potential compensatory processes. Future work should also integrate genetic measures and key exposomal factors (e.g., substance use, metabolic comorbidities, and stress) to understand their respective contributions to parent–offspring brain similarity and refine the biological interpretation of FVI.

To advance clinical translation, longitudinal studies are needed to evaluate the long-term predictive utility of FVIs in youth from families affected by psychotic disorders, and to determine whether it adds predictive value beyond clinical symptoms and established risk factors. A family-specific risk marker that improves prediction beyond existing risk factors could help prioritize scarce resources, such as individual and family-based psychotherapies, for FHR youth not yet exhibiting symptoms of serious mental illness, and guide preventive interventions targeting modifiable risk factors. Furthermore, identifying family-specific regions that are most strongly affected, and those that contribute most to parent–offspring similarity, may provide new biological targets for the development of targeted preventive interventions. Examining longitudinal associations between FVIs and clinical symptom trajectories will clarify whether FVIs can serve as a dynamic biomarker of risk, with potential utility for tracking risk progression and evaluating the effects of interventions. Overall, FVIs hold promising clinical potential for improving individualized risk stratification and guiding early identification and preventive interventions. Robust longitudinal studies will be essential to realize this clinical potential.

The limitations of this study include the modest sample size of parent-offspring dyads, the analyses being limited to cortical thickness measures, and potential sampling bias, as families who were able to participate in the B-SNIP study may differ systematically from those who were not. Future studies in larger, independent samples are needed to assess the replicability of these findings. In addition, we were not able to systematically examine how parent-offspring brain similarities differed among multiple siblings in affected families, given the limited number of families with multiple offspring in our dataset. The developmental heterogeneity among offspring and parent-offspring dyads is also important to consider. To address this, we regressed out age, sex, and intracranial volume — key developmental factors influencing brain measures — and standardized imaging data using z-normalization. While we could not fully account for developmental differences, our results indicate that parent-offspring brain similarities hold clinical relevance despite these differences. Additionally, following the approach used to calculate RVIs, we quantified FVIs using Pearson’s r, which assumes linear relationships and equal weighting across regions. This approach may not fully capture more complex, non-linear patterns of similarity between parents and offspring, and may overlook region-specific contributions shaped by multiple interacting risk factors, including shared genetic liability and environmental exposures within families. Furthermore, Pearson correlation may disproportionately weight regions with lower variance while underrepresenting those with greater variability. Future work is needed to refine FVI estimation using non-linear, regionally informed, and data-driven approaches that better capture the multidimensional nature of familial risk. Another limitation of this study is the challenge of distinguishing familial similarity from illness-related similarity. While we assessed the clinical relevance of overall brain similarity between affected parents and offspring, we cannot separate familial risk traits from psychopathology-related traits. This study provides convergent evidence and will facilitate future research using multi-generational designs to differentiate familial brain similarity from illness-related similarity.

In conclusion, we demonstrated the clinical significance of individual-level cortical thickness similarities between young offspring - affected parent dyads in families impacted by psychotic disorders in this proof-of-concept study. Familial Vulnerability Indices hold clinical translational potential as truly personal and biological risk indicators. Furthermore, this approach will open avenues to assess the clinical utility of family-level similarity indices derived from multiple modalities in a broad spectrum of heritable neuropsychiatric disorders. This new framework, Family-Informed Precision Psychiatry, holds immense untapped potential for psychiatric research to advance risk prediction and clinical care in families affected by serious mental illness.

Supplementary Material

Supplement Results, Figures S1-S5, Tables S1-S7