Brain structural trajectories in youth at familial risk for schizophrenia or bipolar disorder according to development of psychosis spectrum symptoms

Background: The evaluation of child and adolescent offspring of patients with schizophrenia (SzO) or bipolar disorder (BpO) may help understand changes taking place in the brain in individuals at heightened risk for disease during a key developmental period. Methods: One hundred twenty-eight individuals (33 SzO and 46 BpO, considered jointly as ‘Familial High Risk’ (FHR), and 49 controls) aged 6 – 17 years underwent clinical, cognitive and neuroimaging assessment at baseline, 2- and 4-year follow-up. Twenty FHR participants (11 SzO and 9 BpO) developed psychotic spectrum symptoms during follow-up, while 59 FHR participants did not. Magnetic resonance imaging was performed on a 3Tesla scanner; cortical surface reconstruction was applied to measure cortical thickness, surface area and grey matter volume. Results: FHR participants who developed psychotic spectrum symptoms over time showed greater time-related mean cortical thinning than those who did not and than controls. By subgroups, this effect was present in both BpO and SzO in the occipital cortex. At baseline, FHR participants who developed psychotic spectrum symptoms over time had smaller total surface area and grey matter volume than those who did not and than controls. Over time, all FHR participants showed less longitudinal decrease in surface area than controls. In those who developed psychotic spectrum symptoms over time, this effect was driven by BpO, while in those who did not, this was due to SzO, who also showed less grey matter volume reduction. Conclusion: The emergence of psychotic spectrum symptoms in FHR was indexed by smaller cross-sectional surface area and progressive cortical thinning. Relative preservation of surface area over time may signal different processes according to familial risk. These ﬁndings lay the foundation for future studies aimed at stratiﬁcation of FHR youth. the case of grey matter volume and surface area) were also included as ﬁxed effect, and subject ID and family membership were added as random variables. The combined results of the two hemispheres are presented throughout the article. Analyses of lobar measures were based on the same models, where false discovery rate (FDR) correction was used to adjust main group and group by time effects. Post hoc pairwise analyses of all contrasts showing signiﬁcant main effects were also subjected to multiple comparison correction. The same analyses were also run for subcortical structures including the hippocampus, caudate, putamen and pallidum. Sensitivity analyses were performed using the same method- ology as above, by assessing the effects of general cognitive capacity, other psychiatric conditions or cannabis use on the models showing signiﬁcant effects. In addition, analyses were repeated in youth with no previous exposure to psychotropic medications, and by dividing FHR offspring according to whether they had experienced psychotic spectrum symptoms at any point in time (including baseline).


Introduction
Over the last three decades, researchers have attempted to characterize changes in the brain taking place prior to onset of psychosis. To this end, researchers have sought to identify samples exhibiting either subthreshold symptoms or a combination of familial and functional criteria (Miller et al., 2003) and have followed them over time, through progression to health or disease. However, despite evidence from animal models (Gomes, Rinc on-Cort es, & Grace, 2016) and birth cohort studies (Cannon et al., 2002) suggesting that changes in the brain and behaviour associated with psychosis begin early in the course of neurodevelopment, the majority of studies in individuals considered at increased for psychosis have been performed in adults. The field has only recently begun to look towards childhood and adolescence as a critical period for understanding development of major adult mental health outcomes. Studies in adolescents at varying degrees of risk for psychosis, identified either following an 'ultrahigh risk' approach (Cannon et al., 2015) or through community-based samples (Jalbrzikowski et al., 2019), have associated psychotic spectrum symptoms with global and regional grey matter volume and cortical thickness reductions. However, all above studies have relied on either cross-sectional designs or limited follow-up periods, which curtails the ability to capture differences in trajectories. This is especially relevant given the dynamic maturational changes that characterize this developmental phase (Horga, Kaur, & Peterson, 2014) and that the preclinical stage for psychosis may be longer during youth (Schultze-Lutter et al., 2015).
Another unanswered question is whether there is such a thing as a common neural signature to 'psychosis'. Comparative studies between patients with schizophrenia or bipolar disorder (Rimol et al., 2012) have revealed shared differences in their brain structural characteristics, such as thinner frontal cortices, but also specificities, such as widespread surface area reduction, which has been identified in patients with schizophrenia only. Brain changes associated with disease are hereditary ( Van Haren et al., 2012), and the different measures which contribute to brain volumecortical thickness and surface areaare thought to be driven by different genetic influences (Winkler et al., 2012). While there is considerable genetic overlap in the common variants conferring risk for schizophrenia and bipolar disorder (Purcell et al., 2009), evidence of distinct genetic loci, different polygenic risk scores and greater enrichment of copy number variants (Georgieva et al., 2014) in schizophrenia confirm that a certain degree of specificity remains between conditions.
Longitudinal studies of young offspring of patients with schizophrenia or bipolar disorder have the potential to overcome limitations of previous studies, and are able to shed light on the common premorbid trajectories leading to psychosis, while also highlighting areas of specificity between disorders. In our cohort study, which has recruited offspring aged 6-17 years, so far we have reported reduced global and regionalfrontal, parietal and occipitalgrey matter volume and surface area, which were specific to offspring of patients with schizophrenia at crosssection . Over two years, we found greater longitudinal decrease in cortical thickness in the frontal lobe in offspring of patients with bipolar disorder relative to controls. In addition, for older offspring only, we observed greater longitudinal decrease in cortical thickness in offspring of patients with bipolar disorder relative to offspring of patients with schizophrenia, and less decrease in surface area in offspring of patients with schizophrenia relative to controls (Sugranyes, Sol e-Padull es, et al., 2017).
Here, we set out to collect data at three time points over a 4-year period, aiming to assess longitudinal changes in measures of brain structurecortical thickness, surface area and grey matter volumewhich would identify children and adolescents at familial high risk (FHR) for bipolar disorder or schizophrenia who experienced psychotic spectrum symptoms over time. This approach aimed at investigating changes taking place in the brain during a period of heightened risk for psychosis, and to examine specificities of these changes between groups at FHR for schizophrenia versus bipolar disorder.
Based on existing literature and in the light of our own findings so far, we hypothesized that smaller surface area and grey matter volume at baseline, and progressive cortical thinning, would characterize FHR who developed psychotic spectrum symptoms over time. When analysing offspring subtypes, we speculated that the deficits in grey matter volume and surface area observed in offspring of patients with schizophrenia (SzO) at baseline would normalize in those who remained well, while progressive changes in cortical thickness would be especially notable in offspring of patients with bipolar disorder (BpO). As secondary analyses, we set out to examine how these changes were distributed across the brain at a lobar level.

Methods
This is a naturalistic, longitudinal study including assessments at 0, 2 and 4 years. The recruitment and evaluation of the sample have been described in detail elsewhere . Recruitment was performed systematically through parents: patients with a diagnosis of schizophrenia or bipolar disorder from adult psychiatry units with offspring aged 6-17 years were identified and invited to participate in the study. The exclusion criteria for proband parents were intellectual disability and drug or medically induced psychosis or mania. Exclusion criteria for offspring included intellectual disability, head injury with loss of consciousness or severe neurological conditions. Community control parents were recruited through advertisements posted in the same geographic area as the patients. The exclusion criteria were the same as for probands, in addition to personal or first-degree family history of schizophrenia or bipolar spectrum disorders. All 6-to 17-year-old offspring of community control parents were invited to participate in the study; exclusion criteria were the same as those for FHR offspring. To decrease selection bias, control parents who stated motivation to participate due to concerns about school performance or emotional or behavioural problems in their offspring were excluded.

Clinical and cognitive assessment
Participants underwent a comprehensive clinical and cognitive assessment at each visit performed by clinically experienced psychiatrists and psychologists. Clinical diagnoses of the index parents were based on the Spanish version of the Structured Clinical Interview for DSM-IV-TR Axis I Disorders (First, Spitzer, Gibbon, & Williams, 2002). This interview was also used for clinical assessment of coparents and offspring aged 18 or above. Participants younger than 18 years were assessed by child and adolescent psychiatrists or psychologists using the Spanish version of the Kiddie Schedule for Affective Disorders and Schizophrenia for School-Age Children-Present and Lifetime Version (Kaufman et al., 1997). Accumulated prevalence of axis I disorders was calculated at each time point. A measurement of general cognitive capacity was estimated using the Vocabulary, Similarities, Block Design, and Matrix Reasoning subtests of the Wechsler Intelligence Scale for Children (Wechsler, 2003) or the Wechsler Adult Intelligence Scale (Wechsler, 2001) (for participants over 16 years). Pubertal development was assessed using a self-report pictorial questionnaire according to Tanner criteria. (Tanner & Whitehouse, 1976). Cannabis use was measured dichotomously using the Teen-Addiction Severity Index administered by an interviewer to the participant individually (Kaminer, Bukstein, & Tarter, 1991).
Participants were assessed at each visit with the Scale of Prodromal Symptoms Scale (SOPS) within the Structured Interview for Prodromal Symptoms (Miller et al., 2003). Reliability for the SOPS has been undertaken by the team members performing clinical assessments (Kappa statistic for both SOPS total score and subscales >0.8). Scores in the SOPS were used to divide the sample between those who had experienced psychotic spectrum symptoms at at least one longitudinal assessment and those who had not. A definition of 'psychotic spectrum symptoms' was defined as any SOPS positive or negative rating equal or over 3, and included both individuals experiencing an attenuated psychosis syndrome (American Psychiatric Association, 2013) and those meeting criteria for a psychotic disorder (schizophrenia spectrum and other psychotic disorders (295.XX, 298.X, 297.1), or bipolar (296.4, 296.5) or depressive disorders (296.2, 296.3) with psychotic features.
The protocol was approved by the local ethics review board. All participants provided written informed consent or assent after receiving a complete description of the study at each time point. For participants under age 18 years, parents gave informed consent and participants provided assent.

Image acquisition and processing
A high-resolution T1-weighted 3-dimensional magnetization prepared rapid sequence was obtained on a 3-T Siemens Magnetom Trio Tim at the Center for Image Diagnosis at the Hospital Clinic of Barcelona, using the following parameters: 240 sagittal slices, 2,300-ms repetition time, 3.01-ms echo time, 1-mm slice thickness, 900-ms inversion time, 394 _ 240 field of view, 256 _ 256 matrix size and 9_ flip angle. To rule out possible underlying pathology, an axial T2 structural image was acquired and subsequently reviewed by a neuroradiologist blinded to group classification. All scans were visually checked for inaccuracies by two trained raters. Editing procedures detailed on the FreeSurferWiki (http://freesurfer.net/fswiki/Edits), consisting of adding control points and/or modification of the brain mask, were required in 5% of the sample, with no difference in rates of corrections between groups (chi-square .92, p = .82).
After reorientation along the anterior and posterior commissural line, measurements of cortical thickness and surface area were computed using the standard FreeSurfer 5.3.0 pipeline (http://www.surfer.nmr.mgh.harvard.edu) (Fischl et al., 2002). The FreeSurfer automated procedure includes motion correction, nonuniform intensity normalization, registration to Montreal Neurological Institute stereotactic space, skull strip, white matter segmentation, definition of pial surface and parcellation of the brain into 34 cortical regions per hemisphere using the Desikan-Killiany brain atlas (Desikan et al., 2006).
Measures of mean hemisphere cortical thickness and total hemisphere surface area and grey matter volume were computed and divided into frontal, temporal, parietal and occipital parcellations; they were then extracted and subjected to between group comparisons.

Statistical analysis
Comparisons were performed between FHR divided according to whether they had experienced developed psychotic spectrum symptoms at at least one visit over the follow-up period (FHR-P and FHR-NP), and controls. For secondary analyses, the FHR group was further subdivided into offspring subgroups according to whether they had developed psychotic spectrum symptoms over time (SzO-P and BpO-P) or not (SzO-NP and BpO-NP).
Statistical analysis: demographic and clinical measures. Statistical analyses of demographic and clinical variables were performed using SPSS 22.0. For demographic variables, chi-square statistics with Yates correction and Fisher's exact test were used to compare percentages of discrete variables. Univariate analyses of variance and Mann-Whitney U or Kruskal-Wallis tests were used to compare continuous variables. Pairwise contrasts were carried out when the 3-group comparison was statistically significant. Clinical variables were assessed with linear mixed models, where family was included as random variable so as to account for relatedness between participants.
Statistical analysis: neuroimaging measures. For global measures of cortical thickness, surface area and grey matter volume, linear mixed-effects models were used for (a) cross-sectional analysis, where baseline measures of brain structure were included as outcome variables and 'group' as predictor, and (b) for longitudinal analyses, where measures of brain structure at all available time points were included as outcome variable, and 'group', 'time' (computed as time elapsed from baseline scan to each follow-up assessment) and 'group by time interaction' were also included as predictors. For each model, age at baseline and gender were included as covariates in a stepwise procedure, when showing effects at a p < .20 threshold. Hemisphere (right/left) and total intracranial volume (only in the case of grey matter volume and surface area) were also included as fixed effect, and subject ID and family membership were added as random variables. The combined results of the two hemispheres are presented throughout the article. Analyses of lobar measures were based on the same models, where false discovery rate (FDR) correction was used to adjust main group and group by time effects. Post hoc pairwise analyses of all contrasts showing significant main effects were also subjected to multiple comparison correction. The same analyses were also run for subcortical structures including the hippocampus, caudate, putamen and pallidum.
Sensitivity analyses were performed using the same methodology as above, by assessing the effects of general cognitive capacity, other psychiatric conditions or cannabis use on the models showing significant effects. In addition, analyses were repeated in youth with no previous exposure to psychotropic medications, and by dividing FHR offspring according to whether they had experienced psychotic spectrum symptoms at any point in time (including baseline).

Demographic and clinical measures
One hundred twenty-eight participants underwent baseline assessment. Table S1 provides information on relatedness between participants, and Figure S1 illustrates a flow chart illustrating retention rates over the three time points and availability of MRI scans at each assessment. Although there were no demographic, diagnostic or cognitive differences, participants who failed to attend all assessments had higher rates of psychotic spectrum symptoms at baseline than those who did not (see Appendix S1).
Reasons for lack of scan included wearing dental braces, claustrophobia or refusal to undergo scanning. Among participants with at least one valid MRI scan during the study period (N = 33 SzO, N = 46 BpO and N = 49 controls), eleven (33.3%) SzO, nine (19.6%) BpO and five (10.2%) controls experienced psychotic spectrum symptoms at at least one longitudinal assessment. Among these individuals, N = 4, all within the group of SzO, had experienced clinical psychosis, while the remaining participants had experienced an attenuated psychosis syndrome. Given the small number of controls who developed psychotic spectrum symptoms over time, these were excluded from the imaging analyses. Table 1, Tables S2 and S3, and Figure S2 illustrate demographic and clinical information of the sample included in neuroimaging analyses. There were no age or sex differences between FHR-P, FHR-NP or controls. FHR-P had lower general cognitive capacity © 2020 Association for Child and Adolescent Mental Health Longitudinal imaging in high-risk youth 3   3   1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59 than controls at baseline and at 4 years ( Figure S2). FHR-P, and FHR-NP at 2 years, had a greater history of use of psychotropic medications than controls, and FHR-P only had greater use of cannabis than controls across all time points. Both FHR-P and FHR-NP had higher rates of attention deficit hyperactivity disorders (ADHD) than controls at baseline and 2 years, while (nonbipolar) mood disorders where more prevalent at 2-year follow-up in FHR-P only ( Figure S2).

Neuroimaging measures
Cortical thickness. We observed no effect of group on mean hemisphere cortical thickness at baseline. There was an overall effect of time on mean hemisphere cortical thickness (F = 603.3, p < .001, Beta = À.0273, CI: [À.0316 to À.0230]). Examination of longitudinal effects revealed a group by time effect (F = 3.18, p = .042) consisting of greater timerelated loss of cortical thickness in FHR-P than controls (Beta = À.0064, t = À2.51, p = .012, CI: [À.0115 to À.0014]) and than FHR-NP (Beta = À.0051, t = À2.02, p = .044, CI:[À.0101 to À.0001]) (Figure 1). This was observed in the occipital lobe for FHR-P relative to controls and FHR-NP and in the frontal lobe for FHR-P vs controls at near significance (Table S4). By subgroups, BpO-P showed greater decrease in cortical thickness over time relative to controls, to BpO-NP and to both SzO subgroups (Table S5, Figure 2). At a lobar level, these effects in cortical thickness were observed in the temporal and occipital cortices, and frontal lobe at near significance (Table S5).

Surface area.
There was an effect of group at baseline (F = 5.59, p = .005) whereby FHR-P showed smaller surface area than controls (Beta = À3122.5, t = À3.32, p = .001, CI: [À4984.23 to À1260-70]) and than FHR-NP (Beta = À2000.1, t = À2.49, p = .015, CI: [À3598.9 to À401.41]). At a lobar level, FHR-P showed a near-significant smaller surface area than controls in the frontal, temporal and parietal lobes, and than FHR-NP in the temporal lobes (Table S4). Details on these effects divided by subgroups of FHR are depicted in Figure 2 and Table S6, which illustrate how both SzO-P and SzO-NP showed smaller total surface area than controls, while BpO-P exhibited smaller surface area than both controls and BpO-NP. These effects were observed in the temporal and parietal lobes.
There was an effect of time on surface area (F = 57.66, p < .001, Beta = À153.07, CI: [À289.47 to À16.68]), and examination of group by time effects (F = 4.16, p = .016) revealed less time-related decrease in surface area in both FHR-P (Beta = 208.1, t = 2.56, p = .011, CI: [48.36 to 367.74]) and FHR-NP (Beta = 129.5, t = 2.18 p = .029, CI: [12.97 to 246.01]) than controls (Figure 1). By offspring subgroups, group by time effects for total surface area were observed in SzO-NP relative to controls, SzO-P and BpO-NP, and in BpO-P relative to controls, BpO-NP and SzO-P (see Table S5 and Figure 2). At a lobar level, effects described for SzO-NP were observed in the parietal lobe and for BpO-P in the frontal lobe (Table S5).
There was an effect of time on grey matter volume (F = 559.6, p < .001, Beta = À3993.9, CI: [À4690.7 to À3297.1]) but no group by time effects for grey matter volume between FHR groups. When subdividing according to offspring subgroup, there was a significant group by time interaction for total grey matter volume, whereby SzO-NP showed less timerelated loss of total grey matter volume than controls (Table S5). No group or group by time effects were observed in the hippocampus, caudate, putamen or pallidum (see Appendix S2).

Sensitivity analyses
Results from sensitivity analyses are depicted in Appendix S3. There was no significant effect of potential confounders on any of the above findings, with the following exceptions: in participants with no previous exposure to psychotropic medication, the main effects of FHR group on surface area showed the same direction and similar beta values than with the whole sample (retaining trend-level significance for baseline measures and statistical significance for group by time effects). For mean total cortical thickness, while the direction of the effects was the same as for the whole group, group by time effects became nonsignificant. Nevertheless, there continued to be a group by time effect for cortical thickness of the occipital lobe and for the frontal lobeat trend levelwithin the medication na€ ıve subgroup. While there were no effects of co-occurring ADHD or mood disorders on any of the measures, we did observe that FHR without a history of mood disorders experienced less surface area decrease over time than controls, and FHR with a resolved mood disorder showed less grey matter volume decrease over time than those with a persistent mood disorder and than controls.

Discussion
This study extends previous neuroimaging reports in our familial risk cohort, by identifying longitudinal differences in brain structure according to development of psychotic spectrum symptoms over 4-year follow-up. This is possible as the sample grows older, with children entering adolescence and early adulthood, approaching the peak age of onset of psychosis.
In our previous, 2-year follow-up study (Sugranyes, Sol e-Padull es, et al., 2017), the only observed changes in cortical thickness consisted of greater cortical thinning in BpO than SzO in older youth. Here, we have continued to observe no differences in cortical thickness at baseline. However, over time, FHR-P experienced greater mean hemisphere and occipital (and near-significant frontal) cortical thinning relative to FHR-NP and to controls. When assessing offspring subgroups, similar effects were found in BpO, in whom there was also a nearsignificant effect in the temporal lobe, while in SzO, these effects were observed in the occipital cortex only. Our observations provide evidence that cortical thinning associated with illness observed in adult patients (Rimol et al., 2012) and in young adults with 'ultrahigh' risk criteria who convert to psychosis (Cannon et al., 2015) could begin progressively during childhood and adolescence, and possibly exhibit early areas of overlap between SzO and BpO in the occipital cortex. Although 6 cortical thinning has been observed in patients with bipolar disorder regardless of the presence of psychotic symptoms (Hibar et al., 2018), in the current sample greater cortical thinning characterized BpO-P, and was independent of mood diagnosis. To our knowledge, this is the first study so far to assess structural imaging measures in youth at FHR for bipolar disorder experiencing psychotic spectrum symptoms. Cortical thinning has been associated with synaptic pruning, associated glial and vascular changes and/or cell shrinkage (Morrison & Hof, 1997); changes which have been suggested to take place proximally to illness, which could explain the lack of findings in our sample at cross-section, and the progressive differences as individuals grow older, potentially nearer to illness onset. The only area showing greater cortical thinning in SzO-P, the occipital cortex, is among the earliest developing cortical regions (Tamnes et al., 2017). More widespread cortical thinning, extending to anterior regions, may emerge as SzO grow into late adolescence, potentially denoting different timing of cortical maturation between SzO versus BpO.
At baseline, smaller total surface area was specific to FHR-P. When assessing these results by subgroups, we observed that this finding was present in all SzO versus controls, regardless of whether they went on to develop psychotic spectrum symptoms, while in BpO, this was only present in BpO-P. Therefore, our results suggest that for SzO, smaller baseline surface area may be a trait characteristic, which extends our findings at 2 years (Sugranyes, Sol e-Padull es, et al., 2017), while for BpO, this may index cases experiencing psychotic spectrum  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59 symptoms during follow-up. Comparative studies in adult patients have pointed to specificity of surface area reduction for schizophrenia (Rimol et al., 2012), while a recent meta-analysis has suggested that surface area decrease in bipolar disorder may only be present in patients with psychotic symptoms (Hibar et al., 2018), which coincides with our observations. Our findings also concur with recent data from a community-based study in adolescents which also described cross-sectional frontal and parietal surface area reduction in youth showing psychoticbut not bipolarspectrum symptoms (Jalbrzikowski et al., 2019), which mirrors the lack of effect of mood diagnoses on our findings at cross-section. Only follow-up of the current sample into adulthood will resolve questions concerning prediction of specific diagnostic outcomes in these youth.
Previous reports on changes in surface area in individuals at increased risk for psychosis either for clinical or familial reasons are mixed: another adolescent sample of SzO (Prasad et al., 2010) has described cross-sectional reduction similar to ours, while studies in individuals with ultrahigh risk for psychosis documenting data on surface area are few and provide contradicting results (Takayanagi et al., 2017). Smaller cortical surfaces have also been found to characterize other conditions traditionally considered to be of 'neurodevelopmental' origin such as ADHD (Ambrosino, de Zeeuw, Wierenga, van Dijk, & Durston, 2017); indeed, we have recently reported a gradient of severity in childhood developmental difficulties (delayed or disrupted acquisition of language, motor, reading/ writing skills and elimination disorders), as well as increased rates of ADHD and lower general cognitive capacity, between SzO, BpO and controls . While 7 these factors appear to aggregate in these individuals, it is noteworthy that neither ADHD nor lower general cognitive capacity significantly changed the current imaging findings and is thus likely to result from a combination of geneenvironmental effects impacting on both cortical surface and behavioural and cognitive measures.
Our findings concerning longitudinal change in surface area were apparently contradictory, as both FHR-P and FHR-NP showed less time-related decrease in total surface area. When assessing offspring subgroups, we observed that the effects in FHR-NP were found in SzO in total surface area and specifically in the parietal lobe. This suggests that our findings at 2 years (Sugranyes, Sol e-Padull es, et al., 2017) of less longitudinal decrease in surface area in SzO than in controls were likely driven by those who would not develop psychotic spectrum symptoms at follow-up. These group by time differences in surface area in SzO-NP were coupled with less longitudinal grey matter volume decrease. A different sample of healthy first-degree child and adolescent relatives of patients with schizophrenia also experienced less time-related decrease in grey matter volume which normalized as the sample matured, which the authors interpreted as compensatory for early deficits, although these authors did not provide data for surface area (Mattai et al., 2011). Therefore, less time-related decrease in surface area in SzO-NP during adolescence may signal the effect of influences conferring protection against disease. In contrast, we observed relative preservation of surface area in BpO-P. In this case, this change was not coupled with changes in grey matter volume, and was located in the frontal lobe, coinciding with one of the regions of greater cortical thinning associated with the development of psychotic spectrum symptoms. Indeed, a number of reports have related cortical thickness reduction to surface area increase, in association with developmental changes in sulcal metrics (Aleman-Gomez, Janssen, & Schnack, 2013). In addition to mechanistic explanations, there is discussion with regard the genetic link between surface area and cortical thickness, which may be stronger earlier in life than during maturity (Schmitt et al., 2019), and could vary according to the different genetic background between schizophrenia and bipolar disorders. A further issue potentially playing a role in longitudinal change in cortical surface and grey matter volume concerns mood disorders. Inclusion of mood disorders in the models dividing FHR according to psychotic spectrum symptoms at follow-up did not significantly change the results. However, when subdividing FHR according to history of mood disorders, we observed that those who had no history of a mood diagnosis experienced less time-related decrease in surface area than controls and that those with a mood diagnosis which had resolved over follow-up showed less grey matter volume loss than those who continued to meet criteria for a mood disorder at followup and than controls. Taken together, our observations help understand previous mixed findings concerning cortical surface area in samples of individuals at increased risk for psychosis, and stress the need for longitudinal studies taking into account different morphometric measures, familial background and trajectories of mood symptoms, in order to fully understand the brain changes leading to disease.
A number of methodological considerations need to be taken into account when interpreting the current findings. First, we acknowledge the small sample size of the individual offspring groups, which limited further subdivisions of the sample (for instance according to pubertal stage or according to co-occurring diagnoses). Schizophrenia patients, particularly, have low fertility rates, which is a challenge in all offspring studies, especially when performing systematic recruitment. Second, we have combined participants experiencing attenuated psychotic symptoms with those manifesting clinical psychotic symptoms, in an attempt to capture the effects of psychosis dimensionally, although this needs to be taken into account when comparing our findings with others following a different approach. Despite strong inter-rater reliability in the SOPS, inter-rater reliability for the K-SADS or SCID-I interview was not determined. Nevertheless, it is worth noting that interviews were performed by a small number of experienced child and adolescent psychiatrists and psychologists, and supervision was provided between team members during a transition period between assessments. Furthermore, all clinical information was collected by experienced clinicians through face-to-face assessments of the offspring and their parents, which grants greater reliability to our data in comparison with other similar studies based on self-reports. A shortcoming of our design is that we have not been able to assess bipolar prodromal symptoms in our sample with a measure designed for this purpose; however, we have presented subanalyses examining the effect of co-occurring mood disorders, which have also been assessed in the same context with standardized measures. Further, we cannot entirely rule out attrition bias whereby individuals more severely affected could have been less likely to attend follow-up visits; although there were no diagnostic or demographic differences, those who did not attend all assessments had higher rates of psychotic spectrum symptoms at baseline, which may have reduced power to detect differences. The strength of some of the associations decreased when limiting analyses to the medication na€ ıve samplewhile we cannot fully exclude that psychotropic medications may have an effect on brain structure, it is likely that those on medication are in fact those either at greatest risk for the disease or those with manifest symptoms, and therefore that the decrease in power may explain most of the reduction of effects. Furthermore, a large portion of cases with medication were stimulants, which have in fact been suggested to have a normalizing effect on brain structural deficits (Nakao, Radua, Rubia, & Mataix-Cols, 2011). Finally, for secondary analyses examining crosssectional and longitudinal effects of group for each cortical lobe, while corrected for multiple comparisons, our models did not take into account potential correlations between the different lobes.

Conclusions
Our sample provides the opportunity to examine the imprint of familial risk for disease in individuals who carry a combination of both genetic liability and risk markers of disease. The current study provides novel evidence consisting of early cross-sectional surface area and volumetric decrease, and progressive cortical thinning, in FHR who go on to develop psychotic spectrums at follow-up. Future studies in FHR samples should extend these findings, examining their potential to help stratify FHR individuals according to risk of progression to disease, aiding tailored intervention.