Obesity as a Risk Factor for Accelerated Brain Ageing in First-Episode Psychosis—A Longitudinal Study

Abstract

Background

Obesity is highly prevalent in schizophrenia, with implications for psychiatric prognosis, possibly through links between obesity and brain structure. In this longitudinal study in first episode of psychosis (FEP), we used machine learning and structural magnetic resonance imaging (MRI) to study the impact of psychotic illness and obesity on brain ageing/neuroprogression shortly after illness onset.

Methods

We acquired 2 prospective MRI scans on average 1.61 years apart in 183 FEP and 155 control individuals. We used a machine learning model trained on an independent sample of 504 controls to estimate the individual brain ages of study participants and calculated BrainAGE by subtracting chronological from the estimated brain age.

Results

Individuals with FEP had a higher initial BrainAGE than controls (3.39 ± 6.36 vs 1.72 ± 5.56 years; β = 1.68, t(336) = 2.59, P = .01), but similar annual rates of brain ageing over time (1.28 ± 2.40 vs 1.07±1.74 estimated years/actual year; t(333) = 0.93, P = .18). Across both cohorts, greater baseline body mass index (BMI) predicted faster brain ageing (β = 0.08, t(333) = 2.59, P = .01). For each additional BMI point, the brain aged by an additional month per year. Worsening of functioning over time (Global Assessment of Functioning; β = −0.04, t(164) = −2.48, P = .01) and increases especially in negative symptoms on the Positive and Negative Syndrome Scale (β = 0.11, t(175) = 3.11, P = .002) were associated with faster brain ageing in FEP.

Conclusions

Brain alterations in psychosis are manifest already during the first episode and over time get worse in those with worsening clinical outcomes or higher baseline BMI. As baseline BMI predicted faster brain ageing, obesity may represent a modifiable risk factor in FEP that is linked with psychiatric outcomes via effects on brain structure.

Introduction

Psychotic disorders are among the most disabling conditions, which also present with high rates of medical comorbidities and premature mortality.^1,2 Overweight and obesity are particularly prevalent in individuals with schizophrenia,^3–5 already early in the course of illness,^6–8 including participants with first episode of psychosis (FEP).⁴ Obesity contributes not only to poor somatic health and shorter life expectancy,^9,10 but also to adverse psychiatric outcomes.^11,12 It is one of the strongest contributors to functional deterioration in psychosis.¹¹ The pathoplastic effects of obesity on mental health indices may be related to the negative effects of obesity on the brain.

The brain is now recognized as one of the end organs for obesity-related damage.¹³ Replicated cross-sectional findings from thousands of individuals have demonstrated associations between measures of obesity and brain structure, including basal ganglia, hippocampus, thalamus, frontal, temporal, and cerebellar regions.^13–16 In addition, brains of obese/overweight individuals appear older than their chronological age.^17,18 Similar neurostructural alterations are frequently reported already early in the course of psychotic disorders,^19–21 but their origins remain unknown. Considering the multiple mutual links between obesity, FEP, and brain structure, it is possible that obesity contributes to brain alterations in psychosis. While the effects of obesity on the brain are of major interest in medicine, they remain markedly under-researched in psychiatry.

We have previously demonstrated that obesity was associated with advanced brain age and lower cerebellar volume in FEP.^18,22 Yet, the temporal direction of the association remains unclear. Obesity could be contributing to brain changes or vice versa. Longitudinal design can help address the direction of association and the issue of reverse causality. However, there are no longitudinal studies investigating links between obesity and brain atrophy in schizophrenia/FEP. Such studies are necessary to test whether obesity is a risk factor for brain alterations in FEP. Identification of clinical risk factors for brain alterations in psychotic disorders is the first step towards their management. Specifically, a longitudinal study in the early stages of psychotic illness, when the brain and metabolic markers may be most dynamic and additional confounds related to chronicity/long-term use of medications/social factors are relatively limited, offers a particularly strong research design. Focusing on changes that develop early in the course of illness is also clinically relevant for early intervention and prevention of long-term negative outcomes.

Previous brain imaging studies have traditionally focused on individual brain regions. Recently, normative models and cumulative measures of brain structure have become available through access to large databases of brain scans and advances in neuroimaging analyses involving machine learning. One such approach is to estimate the biological age of the brain from structural magnetic resonance imaging (MRI).²³ The difference between estimated brain age and chronological age, so called Brain Age Gap Estimate (BrainAGE), captures diffuse, multivariate neurostructural alterations into a single number, thus simplifying the analyses and aiding in their interpretation.²⁴ The BrainAGE metric is sensitive to the presence of schizophrenia as well as obesity¹⁸ and provides a unique opportunity to better understand longitudinal brain changes in major psychiatric disorders.

In this longitudinal study in participants with FEP and controls, we used machine learning and structural MRI to study brain ageing/neuroprogression around the time of illness onset and 1–2 years later. Our a priori hypothesis was that both the diagnosis of FEP and baseline body mass index (BMI) would each be associated with accelerated brain ageing over time. In addition, we explored the effects of other clinical, biochemical variables and medications on annual rates of brain ageing.

Methods

Patient Recruitment

This was a part of the Early-Stage Schizophrenia Outcome (ESO) project,^18,25 an ongoing prospective study of individuals with FEP, conducted at the National Institute of Mental Health, Czech Republic (NIMHCZ). Cross-sectional, baseline findings from this study were published previously,^18,22,25 but this is the first analysis of the longitudinal data. Individuals with FEP were recruited during their first hospitalization through the ESO Patient Enrolment Network, which involves 10 Czech inpatient psychiatric facilities. Subsequently, participants were centrally assessed by psychiatrists at the NIMHCZ. The largest contributing facility was Psychiatric Hospital Bohnice (1200 beds), which serves the Prague and part of Central Bohemia regions—catchment area of >1.5 million subjects.

Consistent with the literature and our previous studies, FEP was defined as the first hospitalization for psychosis.^26,27 Individuals with FEP met the following inclusion criteria: (1) were undergoing their first psychiatric hospitalization; (2) had the ICD-10 diagnosis of schizophrenia (F20), acute and transient psychotic disorders (F23), or schizoaffective disorders (F25) based on Mini-International Neuropsychiatric Interview²⁸; (3) had fewer than 24 months of untreated psychosis; and (4) were at least 18 years old. Patients with psychotic mood disorders were excluded from the study.

We focused on participants at the early stages of illness, to minimize the effects of illness and medications on brain structure. Thus, individuals who were hospitalized before meeting the duration criteria for schizophrenia are a particularly interesting group. These participants were included in the study and received the working diagnosis of acute and transient psychotic disorders, congruent with the Diagnostic and Statistical Manual of Mental Disorders, fourth edition (DSM-IV) brief psychotic disorder. This approach is in keeping with other studies of FEP.^5,29

Healthy controls, at least 18 years old, were recruited via advertisement, using the following exclusion criteria: (1) lifetime history of any psychiatric disorders and (2) psychotic disorders in first- or second-degree relatives. Additional exclusion criteria for both groups included history of neurological or cerebrovascular disorders and any MRI contraindications.

The baseline MRI scanning (Visit 1) was performed during the first hospitalization and study participants were invited for the second scan and assessment (Visit 2) ~1 year later. At each visit, we rated the symptoms, illness severity, and functioning using the Positive and Negative Syndrome Scale (PANSS),³⁰ Clinical Global Impression (CGI), and the Global Assessment of Functioning (GAF)³¹ and also collected fasting blood samples for biochemical analyses (glucose, high-density lipoprotein [HDL]-cholesterol, low-density lipoprotein [LDL]-cholesterol, triglycerides, and high-sensitivity C-reactive protein [hsCRP]). Duration of illness and treatment variables were determined via interview together with all available collateral data from medical records, treatment providers, and family members. We expressed antipsychotic doses in chlorpromazine equivalents and calculated the cumulative antipsychotic exposure as dosage times duration of treatment in months. We measured BMI using the formula: BMI = weight (kg)/height (meters).² All diagnostic assessments, ratings of symptoms and medications, were performed by psychiatrists. Biochemical analyses were done in a single clinical laboratory using standard clinical methods.

The study was carried out according to the latest version of the Declaration of Helsinki. All study individuals were informed about the study procedures and signed informed consent, which was approved by the local Research Ethics Board.

MRI Acquisition

Structural MRI data were collected at 2 sites, the NIMHCZ, N = 137, and Institute of Clinical Experimental Medicine in Prague (IKEM), N = 221. We acquired T1-weighted 3D MPRAGE scans (TE = 4.63 ms, TR = 2300 ms, bandwidth 130 Hz/pixel, FOV = 256 × 256 mm, matrix 256 × 256, voxel size 1 × 1 × 1 mm³) on 3T Siemens Trio MRI scanner (IKEM) or 3T Siemens Prisma MAGNETOM (NIMHCZ) MRI equipped with standard head coil.

Brain Age Estimation

We estimated the brain age of each participant using a machine learning method which was developed by us (K.F.), extensively validated,^23,32 shown to be sensitive to metabolic or psychiatric disorders,^18,25,33,34 and robust to inter-scanner differences.^23,32 Briefly, this included (1) standard voxel-based morphometry preprocessing of structural MRI data, (2) feature reduction via smoothing and principal component analysis, and (3) age estimation using relevance vector regression (RVR). This RVR model was trained using an independent sample of 504 healthy individuals from the IXI database (http://www.brain-development.org). For more details, see Franke et al^23,33,34 and supplementary material.

Our outcome measure was the BrainAGE, ie, the difference between the estimated brain age and the chronological age. We removed the association between age and BrainAGE using previously documented procedure.³⁵ Specifically, for the cross-sectional analyses, where BrainAGE was the dependent variable, we used age as a covariate. For the longitudinal analyses, we regressed out the fixed effect of age from BrainAGE using a linear mixed model, which also included a random grouping factor of subject ID (one ID per subject) to capture the longitudinal, within-subject design. Model residuals provided BrainAGE_cleaned, which was not associated with age. The BrainAGE – BrainAGE_cleaned represents the age-related error, which we subtracted from each individual’s estimated brain age, Formula 1, in order to obtain estimated brain age, which was not biased by age.

$${\displaystyle \begin{array}{l}{\displaystyle \mathrm{E}\mathrm{s}\mathrm{t}\mathrm{i}\mathrm{m}\mathrm{a}\mathrm{t}\mathrm{e}\mathrm{d}\mathrm{B}\mathrm{r}\mathrm{a}\mathrm{i}\mathrm{n}\mathrm{A}\mathrm{g}{\mathrm{e}}_{\mathrm{c}\mathrm{l}\mathrm{e}\mathrm{a}\mathrm{n}\mathrm{e}\mathrm{d}}=\mathrm{E}\mathrm{s}\mathrm{t}\mathrm{i}\mathrm{m}\mathrm{a}\mathrm{t}\mathrm{e}\mathrm{d}\mathrm{B}\mathrm{r}\mathrm{a}\mathrm{i}\mathrm{n}\mathrm{A}\mathrm{g}\mathrm{e}-(BrainAGE-BrainAG{E}_{\mathrm{c}\mathrm{l}\mathrm{e}\mathrm{a}\mathrm{n}\mathrm{e}\mathrm{d}})}\end{array}}$$ (1)

The annual rate of brain ageing (R) was calculated by using Formula 2:

$${\displaystyle \begin{array}{l}{\displaystyle R=\frac{\mathrm{E}\mathrm{s}\mathrm{t}\mathrm{i}\mathrm{m}\mathrm{a}\mathrm{t}\mathrm{e}\mathrm{d}\mathrm{B}\mathrm{r}\mathrm{a}\mathrm{i}\mathrm{n}\mathrm{A}\mathrm{g}{\mathrm{e}}_{\mathrm{c}\mathrm{l}\mathrm{e}\mathrm{a}\mathrm{n}\mathrm{e}\mathrm{d}\mathrm{\_}\mathrm{V}\mathrm{i}\mathrm{s}\mathrm{i}\mathrm{t}2}-\mathrm{E}\mathrm{s}\mathrm{t}\mathrm{i}\mathrm{m}\mathrm{a}\mathrm{t}\mathrm{e}\mathrm{d}\mathrm{B}\mathrm{r}\mathrm{a}\mathrm{i}\mathrm{n}\mathrm{A}\mathrm{g}{\mathrm{e}}_{\mathrm{c}\mathrm{l}\mathrm{e}\mathrm{a}\mathrm{n}\mathrm{e}\mathrm{d}\mathrm{\_}\mathrm{V}\mathrm{i}\mathrm{s}\mathrm{i}\mathrm{t}1}}{\mathrm{\Delta }T}}\end{array}}$$ (2)

where ΔT is time elapsed in years between the 2 visits. This provided an annual estimate of brain ageing relative to real time, with 1.0 indicating a change of 1 year in estimated brain age over a 1-year period.

Statistical Analyses

All statistical analyses were performed using R version 3.6.2, with mixed modeling using the package mgcv.³⁶ To compare clinical and demographic variables, we used t test, or chi-square test, as appropriate. We first investigated the association between BrainAGE and clinical or demographic variables, using linear mixed modeling, controlling for a random effect of data collection site, and a fixed effect of participant age. We also used this same method to test the association between follow-up BrainAGE and follow-up clinical or demographic variables to verify that associations were consistent and mirrored longitudinal changes. In longitudinal analyses, we tested the association between group (controls vs FEP) and/or BMI as predictor(s) and the annual rate of brain ageing as dependent variable, while controlling for a random effect of data collection site. To verify that baseline variables were significantly associated with annual rate of brain ageing beyond their association with baseline brain measures, we also additionally controlled for baseline BrainAGE. All of these longitudinal analyses also controlled for age, by using the Estimated Brain Age_cleaned, as described above.

In each group separately, we tested whether the annual rate of brain ageing was greater than the expected rate of 1.0 (ie, 1 year of brain ageing per chronological year) and whether the changes in estimated brain age were greater than the interval between the 2 visits, using linear mixed model to control for a random effect of data collection site.

We also explored the association between baseline clinical variables or their changes over time, with annual rate of brain ageing. Lastly, we tested whether the annual rate of brain ageing was best predicted by BMI, biochemical measures, or some combination of the 2. All of these analyses were performed using linear mixed modeling and controlling for a random effect of data collection site. We calculated variance inflation factor (VIF) to check for multicollinearity, which was acceptable in all models and Q-Q plots to ensure normally distributed residuals.

Results

We recruited 183 participants with FEP and 155 controls (table 1; supplementary table S1). Individuals with FEP had a higher baseline BrainAGE (3.39 ± 6.36) than controls (1.72 ± 5.56, β = 1.68, t(336) = 2.59, P = .010; figure 1). Among individuals with FEP, lower baseline functioning (GAF, β = −0.06, t(164) = −2.06, P = .041) and higher PANSS negative scores (β = 0.20, t(175) = 2.85, P = .005) were associated with higher baseline BrainAGE (supplementary table S2).

Table 1.

Demographic and Clinical Characteristics of Participants at the Time of Their First Visit

	Controls (N = 155)	FEP (N = 183)	Group Difference
Sex: male, N (%)	68 (43.8%)	102 (55.7%)	χ 2 = 4.26, P = .039
Age, M (SD)	28.99 (7.41)	29.29 (7.66)	t(330) = 0.36, P = .714a
BMI, M (SD)	23.32 (3.37)	23.81 (4.16)	t(332) = 1.20, P = .230
BMI group, N (%)b			χ 2 = 5.86, P = .053
Normal weight	117 (76.5%)	119 (65.4%)
Overweight	31 (20.3%)	49 (26.9%)
Obese	5 (3.3%)	14 (7.7%)
NIMHCZ/IKEM, N (%)	56 (36.13)/99 (63.87)	69 (37.70)/114 (62.30)	χ 2 = 0.03, P = .853
Glucose (mmol/L), M (SD)c	4.15 (0.52)	4.15 (0.51)	t(162) = 0.10, P = .918
Cholesterol (mmol/L), M (SD)c	4.53 (0.89)	4.53 (0.92)	t(247) = 0.05, P = .960
HDL (mmol/L), M (SD)c	1.54 (0.39)	1.30 (0.35)	t(235) = 4.98, P < .001
LDL (mmol/L), M (SD)c	2.51 (0.74)	2.57 (0.74)	t(243) = 0.68, P = .494
TGC (mmol/L), M (SD)c	1.08 (0.48)	1.44 (0.99)	t(197) = 3.72, P < .001
hsCRP (mmol/L), M (SD)c	1.65 (2.58)	2.61 (4.48)	t(217) = 2.14, P = .034
Medical comorbidities, N (%)	n/a	Hypertension: 10 (5.46%)	n/a
		Type 2 diabetes: 2 (1.09%)
		Dyslipidemia: 0 (0)
		Hypothyroidism: 8 (4.37%)
Substance abuse/dependence, N (%)	n/a	57 (31.14)	n/a
Diagnosis, N (%)	n/a		n/a
Schizophrenia		101 (55.2%)
Acute and transient psychotic disorders		79 (43.2%)
Schizoaffective disorder		3 (1.6%)
Duration of illness (months)	n/a	7.90 (10.94)	n/a
Duration of treatment (months)	n/a	3.90 (6.49)	n/a
Duration untreated (months)	n/a	3.91 (9.48)	n/a
CHLPZ equivalent (mg)d	n/a	365 (225)	n/a
Age of onset	n/a	28.46 (7.48)	n/a
PANSS Positive Scale, M (SD)	n/a	11.96 (4.03)	n/a
PANSS Negative Scale, M (SD)	n/a	15.21 (5.82)	n/a
PANSS General Psychopathology Scale, M (SD)	n/a	29.71 (8.15)	n/a
GAF, M (SD)	n/a	69.21 (15.14)	n/a
CGI-S, M (SD)	n/a	3.07 (1.35)	n/a

Note: BMI, body mass index; CGI-S, Clinical Global Impression—Severity scale; CHLPZ, chlorpromazine; GAF, Global Assessment of Functioning; HDL, high-density lipoprotein; hsCRP, high-sensitivity C-reactive protein; IKEM, Institute of Clinical Experimental Medicine in Prague; LDL, low-density lipoprotein; n/a, not applicable; NIMHCZ, National Institute of Mental Health, Czech Republic; PANSS, Positive and Negative Syndrome Scale; TGC, triglyceride.

^aWe used the Welch two-sample t-test (unequal variance assumed), which relies on a Welch–Satterthwaite degrees of freedom adjustment.

^bBMI was missing in 3 individuals.

^cGlucose was available in 53% of participants, with remaining biochemical measures (cholesterol, HDL, LDL, TGC, and hsCRP) in 73%–75% of participants.

^dAll FEP individuals received atypical antipsychotics, while 7 people were treated with a combination of atypical and first-generation antipsychotics.

Fig. 1.

Significant group difference in baseline BrainAGE between individuals with first episode of psychosis and controls (A), and significant predictors of a higher annual rate of brain ageing, including higher baseline BMI (B) in the whole sample, worsening global functioning (C), and increasing negative symptoms (D) between visits among individuals with FEP. BMI, body mass index; FEP, first-episode psychosis; GAF, Global Assessment of Functioning; PANSSN, Positive and Negative Syndrome Scale Negative Scale.

During the 1.61 ± 1.22 years interval between the scans, the annual rate of brain ageing in individuals with FEP was 1.28 ± 2.40 years per 1 calendar year, which was not significantly greater than the expected rate of 1.0 (β = 0.28, t(218) = 1.59, P = .057). Among controls, the 1.07 ± 1.74 annual rate of brain ageing was also comparable to the expected rate of 1.0 (β = 0.07, t(186) = 0.51, P = .305). Similarly, the change in estimated brain age between the first and second visit was not significantly different from the interval between the scans in either group (FEP: β = 0.36, t(182) = 1.72, P = .090; controls: β = 0.20, t(154) = 1.12, P = .270). In addition, the annual rate of brain ageing did not significantly differ between the 2 groups (β = 0.21, t(333) = 0.93, P = .176).

In the whole sample, baseline BMI predicted the annual rate of brain ageing (β = 0.08, t(333) = 2.59, P = .010; figure 1). Specifically, for every 1-point increase in BMI, the annual rate of brain ageing increased by approximately an additional month. Even when we controlled for baseline BrainAGE, BMI remained associated with rate of change in brain age (β = 0.079, t(332) = 2.69, P = .007). In a model containing both BMI and diagnosis, only BMI (β = 0.08, t(332) = 2.54, P = .012) was significantly associated with the annual rate of brain ageing. When jointly modeling the association between baseline BMI, glucose, LDL-cholesterol, HDL-cholesterol, triglycerides, hsCRP, and annual rate of brain ageing, only BMI remained significant (β = 0.11, t(138) = 1.98, P = .049) (supplementary table S3).

Among individuals with FEP, none of the baseline clinical characteristics, including diagnosis, type and severity of symptoms, global functioning, or history of substance abuse, predicted the annual rate of brain ageing (supplementary table S2). Worsening of functioning between the visits (GAF; β = −0.04, t(164) = −2.48, P = .014; figure 1) and increases in negative (PANSS N; β = 0.11, t(175) = 3.11, P = .002; figure 1) and general psychopathology (PANSS G; β = 0.06, t(175) = 2.58, P = .011) were all associated with greater annual rate of brain ageing. Associations between follow-up variables and follow-up BrainAGE were consistent with the longitudinal analyses (supplementary table S4).

Antipsychotic Medications, BMI, Symptoms, and Brain Structure

The average duration of treatment at baseline was 3.90 ± 6.49 months and all participants received atypical antipsychotics. The baseline cumulative antipsychotic exposure was associated with higher BMI (β = 0.01, t(168) = 4.23, P < .001), but not with baseline BrainAGE (β = 0.003, t(168) = 1.37, P = 0.174), or with annual rate of brain ageing (β = 0.002, t(169) = 1.11, P = .269). At follow-up, cumulative antipsychotic exposure was associated with both higher BMI (β = 0.003, t(142) = 3.19, P = .002) and higher BrainAGE (β = 0.003, t(145) = 2.73, P = .007; supplementary table S4). When modeled jointly, only antipsychotic exposure (β = 0.003, t(141) = 2.14, P = .033), but not higher BMI (β = 0.17, t(141) = 1.74, P = .084), was associated with higher BrainAGE at follow-up, with no interaction between the 2. The VIF for BMI (1.09), antipsychotic dose (1.06), and duration of treatment (1.14) did not suggest multicollinearity among these measures.

The antipsychotic dose at the first visit was associated with PANSS N (β = 7.28, t(170) = 2.46, P = .014), PANSS P (β = 18.01, t(170) = 4.41, P < .001), and PANSS G (β = 7.06, t(166) = 3.25, P = .001). This pattern changed at follow-up, when PANSS N was associated with the cumulative exposure (β = 11.97, t(144) = 2.14, P = .034), while PANSS P and G were not. When modeled jointly and controlling for BMI and age, only PANSS N, but not the cumulative medication exposure, was a significant predictor of BrainAGE at follow-up (β = 0.25, t(139) = 2.90, P = .004).

Discussion

Relative to controls, individuals with FEP, and especially those with more negative symptoms and lower functioning, showed advanced BrainAGE already during their first episode. The diagnosis of FEP was not associated with accelerated brain ageing over time. The annual rate of brain ageing in FEP was not significantly different from expected rate of change or from the annual rate of brain ageing in controls. At the same time, those FEP individuals who showed greater worsening of clinical outcomes, ie, increasing rates of negative and global symptoms and worsening in functioning between the visits, also demonstrated faster brain ageing. Importantly, across both cohorts, higher baseline BMI predicted acceleration of brain ageing in the next 1–2 years.

This is the first study to prospectively investigate the association between BMI and BrainAGE. In the whole sample, one additional point in BMI was associated with approximately an additional month of brain ageing each calendar year. This is in keeping with other prospective studies indicating that changes in BMI precede changes in brain volumes.^37–45 This is further supported by a mendelian randomization study, which suggested a causal relationship between obesity and whole brain or gray matter volumes⁴⁶ and by studies showing that treatment of obesity slowed brain atrophy.^39,41,47,48 In addition, our findings extend the previous literature by suggesting that the effects of obesity are sufficiently diffuse to affect a composite measure of brain structure, such as BrainAGE.

Some longitudinal studies reported progressive regional brain atrophy in FEP,^19,49 while others found no such changes.^50,51 We did not find accelerated rate of brain ageing in FEP. Interestingly, the only previous study which also estimated brain age⁵² showed a very similar extent of brain ageing over time in schizophrenia, ie, 1.36 years/year, as compared to 1.28 years/year in our study. In addition to that study, we also showed that the annual rate of ageing was comparable between FEP and controls. Together with the already higher BrainAGE at baseline, this suggests that brain alterations in FEP may precede the illness onset, as also demonstrated by others.^25,52–58 It is also possible that progressive brain changes occur only in some individuals with FEP. Indeed, while as a group FEP did not show progressive brain age changes, those with worsening functioning and increasing negative symptoms did demonstrate acceleration of brain ageing over time. This is remarkably consistent with 2 recent studies that also documented association between changes in neurostructural measures and worsening negative symptoms^59,60 or global functioning.⁵⁹ Interestingly, the baseline symptoms did not predict these longitudinal changes, making it more likely that the progressing brain alterations led to the worsening of clinical symptoms. Our findings support the hypothesis that there is heterogeneity in progressive brain changes in FEP, which may in part explain the heterogeneity in clinical outcomes and the inconsistent findings in previous studies.

We were particularly interested to test whether the obesitogenic effects of antipsychotic medications contributed to our findings and to the associations between antipsychotics and brain structure. Indeed, greater cumulative antipsychotic exposure was associated with higher BMI. When modeled jointly, only antipsychotic exposure, but not BMI, was associated with greater BrainAGE at the second visit. At least in the first years after initiation of treatment, the effect of antipsychotic medication on brain structure was not mediated by BMI. This is in keeping with other studies^61,62 and may also be related to the fact that metabolic alterations in FEP may pre-date AP treatment⁶³ or that most of the antipsychotic-related metabolic changes may happen in the first months of treatment,⁶⁴ even prior to the first scan. We also cannot rule out that following a longer exposure, metabolic side effects could become more relevant for the association between antipsychotics and brain alterations.

Interestingly, we found association between cumulative antipsychotic exposure and greater BrainAGE only at Visit 2, but not at Visit 1. This pattern is congruent with 2 explanations. First, medications contributed to neurostructural alterations, but the duration of medication exposure at the first scan, ie, 3.90 ± 6.49 months, was too short for these effects to occur. Alternatively, medication dosage at Visit 1 primarily reflects symptoms, that are not associated with brain alterations, but over time the cumulative medication exposure becomes a more reliable proxy for symptoms that are associated with brain changes. Indeed, medications were most strongly associated with positive symptoms at Visit 1, but with negative symptoms at Visit 2. It is not surprising that higher doses of medications during hospitalization are primarily used to target positive symptoms, but following hospitalizations, medication exposure predominantly tracks negative symptoms. Importantly, it was negative, not positive symptoms, which were associated with brain age, as in another study,⁵⁶ and with acceleration of brain ageing over time. When analyzed jointly, only the negative symptoms, but not the cumulative exposure to medications, remained associated with BrainAGE at follow-up. This is broadly similar to findings of another study, which also demonstrated a greater effect of symptoms than medications on longitudinal brain changes in FEP.⁶⁵ All in all, our findings are most congruent with the explanation that individuals with more brain alterations showed more negative symptoms and required more medications over time.

We can only speculate about the mechanisms underlying the effect of obesity on brain structure. Interestingly, BMI was a stronger predictor of accelerated brain ageing than any individual metabolic measures, including glucose, cholesterol, triglycerides, and hsCRP. This is in keeping with other studies^66,67 and may be related to the relatively young age (29.29 years), absence of other pathology, and mostly normal levels of these biochemical measures. For the same reasons, contribution of vascular factors is also perhaps less likely. Obesity could also affect brain structure through adipokines, oxidative stress, and insulin resistance.^13,68–73 Additionally, we may be detecting effects of lifestyle factors, including high-fat diets, lack of exercise, alcohol abuse, smoking, or psychosocial factors including poverty and disparities in health care/monitoring.^74,75 While we do not precisely understand which of the obesity-associated factors contributed to brain alterations, the fact that an easy-to-measure index, such as BMI, predicted accelerated brain ageing is highly relevant for research and clinical practice.

While we know that neurostructural alterations are frequent in FEP and have clinical implications,^59,60,65 we do not understand their origins and have no means of alleviating/preventing them. Our findings suggest that obesity contributes to brain alterations in FEP. This emphasizes the need to improve weight monitoring and management in FEP and to better integrate psychiatric and medical services. Such efforts may be important not only for prevention of adverse cardiovascular outcomes and increased mortality,^2,72 but also to maintain brain health. In addition, identification of obesity as a risk factor for brain alterations in FEP could lead to testing of novel treatment methods, which would target the brain substrate rather than just the symptoms. Indeed, obesity-related structural brain abnormalities might be preventable or even reversible with lifestyle/surgical/medication interventions focused on weight management.^39,41,47,48 Last but not least, these efforts may help address some of the clinical outcomes, which are frequent, but currently difficult to treat, such as negative symptoms and worse functioning.

With regards to limitations, this study was not designed to test the effects of medications and was not randomized. However, the longitudinal design allowed us to better clarify the temporal direction of associations. Biochemical measures were missing in some individuals and we did not study other psychiatric disorders. Waist circumference or waist-hip-ratio may show more extensive associations with gray matter, but BMI is much easier to acquire, correlates with other obesity-related alterations, and is by far the most frequently used measure.^13,16 We excluded participants with large vessel disease but cannot rule out microangiopathy. Multiple factors such as chronic stress, lifestyle variables, including diet, exercise, and substance use/abuse, and socioeconomic factors impact clinical characteristics and potentially brain structure. Yet, they are difficult to quantify and track over time, especially in clinical populations. Thus, it is beneficial to use a composite measure, such as BMI, which captures a large part of variance related to these additional variables.

With 341 individuals, this is one of the largest prospective studies in FEP. Longitudinal studies continue to be rare but will be necessary to address many key questions which cannot be resolved by cross-sectional comparisons, such as issues of neuroprogression or reverse causality. All diagnostic and clinical assessments were performed by psychiatrists, which could have increased precision. Indeed, many of our findings had excellent face validity. It is noteworthy that the brain imaging measures corresponded with system-level clinical variables including negative symptoms and worse functioning not only cross-sectionally but also longitudinally. The additional strengths of this study include the focus on early stages of illness, use of an increasingly popular, cumulative measure of brain structure, and conservative application of machine learning, where we completely separated training from testing. The results were robust, replicated many of the previously reported findings, while providing some highly original and new information.

In conclusion, brain alterations in psychosis are manifest already during the first episode and over time become more pronounced in those with worsening clinical outcomes or higher baseline BMI. Importantly, BMI was the only baseline variable that predicted acceleration of brain ageing in the next 1–2 years. These findings support a causal role of obesity in brain alterations. Here, we have a risk factor for progressive accumulation of brain changes, which is easy to measure, modifiable, and frequently abnormal already in FEP. Consequently, improving weight monitoring and management may slow neuroprogression in individuals with FEP. As accelerated brain ageing tracked worsening of clinical outcomes, this could also help alleviate accumulation of negative symptoms and improve functioning over time. Future prospective studies should investigate the impact of weight management strategies on brain health and clinical outcomes in FEP and attempt to better understand the factors underlying the neuroprogressive effects of obesity.

Funding

This study was supported by funding from the Canadian Institutes of Health Research (142255), Ministry of Health of the Czech Republic (16-32791A, NU20-04-00393), and Brain & Behavior Research Foundation Young and Independent Investigator Awards to T.H. The sponsors of the study had no role in the design or conduct of this study; in the collection, management, analysis, and interpretation of the data; or in the preparation, review, or approval of the manuscript.