Accelerated aging of functional brain networks supporting cognitive function in psychotic disorders

Abstract

Background:

Across networks, connectivity within the fronto-parietal (FPN) and cingulo-opercular (CON) networks exhibits reductions earliest during healthy aging, contributing to cognitive impairment. Individuals with psychotic disorders demonstrate evidence of accelerated aging across multiple biological systems. By leveraging a large sample of psychosis patients from early to chronic illness-stages, this study seeks to determine whether CON and FPN exhibit evidence of accelerated aging in psychotic disorders, confirm associations between network efficiency and cognition, and determine whether reduced network efficiency is observed in early-stage psychosis.

Methods:

Resting-state fMRI and cognitive data were obtained on 240 psychotic disorder and 178 healthy control (HC) participants. Global efficiency, a measure of functional integration, was calculated for the CON, FPN, subcortical, and visual networks. Associations with age and cognition were assessed and compared between groups.

Results:

Consistent with accelerated aging, significant group by age interactions reflected significantly stronger relationships between efficiency and age in psychosis than HC for both CON (psychosis: r=−.37; HC: r=−.16) and FPN (psychosis: r=−.31; HC: r=−.05). Accelerated aging was not observed in subcortical or visual networks, suggesting specificity for cognitive networks that decline earliest in healthy aging. Replicating prior findings, efficiency of both CON and FPN correlated with cognitive function across all participants (r’s>.11, p’s<.031). Furthermore, chronic (p=.004), but not early psychosis patients (p=.553) exhibited significantly lower FPN efficiency compared to HC.

Conclusions:

Functional integration of higher-order cognitive networks is intact in early psychosis but exhibits evidence of accelerated aging, suggesting the potential for intervention targeting cognition within the early psychosis period.

Introduction

Psychotic disorders typically manifest in late adolescence/early adulthood as the end result of an abnormal neurodevelopmental process (1) likely shaped by a combination of genetic risk (2), pre- and perinatal factors (3,4), and environmental stress (5). While neurodevelopmental in nature, psychotic disorders are also characterized by accelerated aging, indicating more rapid aging processes following illness onset (6). Markers of accelerated aging in psychosis span biological systems and include shorter telomere length (7), increased inflammatory markers (8), and functional deterioration in the context of cognitive decline (9). Psychotic disorder patients also demonstrate accelerated brain aging as demonstrated by sharper decline in gray matter volume (11) and white matter integrity (10,11), and increased brain age gap (12,13) than healthy individuals and patients with depression.

Another marker of brain aging is resting-state functional connectivity, which estimates the degree of inter-connectivity between brain regions. Resting-state functional connectivity declines throughout healthy aging, particularly for connections within networks (14). Age-related reductions in functional connectivity occurs earliest in higher-order cognitive networks, such as the fronto-parietal network (FPN) and cingulo-opercular network (CON), which support domain-general cognitive ability (15–18). Reduced FPN and CON connectivity occurs in the context of stable visual network connectivity (15,17), suggesting that cognitive networks are most vulnerable to age-related decline in the healthy population (19).

Abnormal resting-state functional connectivity has been widely reported in psychotic disorders (20–22). While average functional connectivity is the most commonly used metric of network connectivity, efficiency, which reflects integration of information transfer within a given network (23), is a potentially more sensitive measure of network alterations and vulnerability to aging abnormalities (24–26). Older adults with vascular dementia, for instance, exhibit lower global efficiency, which is associated with greater cognitive impairment (27). Stroke patients with lesions in nodes critical for highly efficient network connectivity exhibit greater cognitive and functional impairment compared to patients with lesions in areas less critical for network integration (28). Network efficiency is therefore sensitive to alterations in neurodegenerative disorders and localized brain lesions.

The only prior study of accelerated aging of functional networks in schizophrenia by Sheffield et al. (29) found that functional network efficiency of the CON and FPN declines at an accelerated rate in schizophrenia. However, limitations of this study, including modest sample sizes (especially for a study on age effects), focus on individuals in the chronic stage of schizophrenia, and use of regressed task-effects to estimate “resting-state” functional connectivity, which yields different estimates of functional connectivity than pure resting-state fMRI (30) and drives heterogeneous findings in clinical populations (31), raise concerns about reproducibility and generalization. Moreover, growing evidence indicates that accelerated aging processes observed in schizophrenia are also present in bipolar disorder, including reduced telomere length (32), altered immune functioning (33), and increased oxidative stress (34). However, whether accelerated aging of functional networks extends to psychotic bipolar disorder has not been investigated.

The current investigation utilized a large sample of individuals with schizophrenia and bipolar disorder, including a significant number of individuals in the early stage of psychosis, to address the limitations and knowledge gaps described above. The primary goal of this study was to determine if accelerated aging of FPN and CON, identified in an earlier, smaller study of largely chronic patients, replicates in a substantially larger independent dataset, and, if so, confirm the specificity of differential aging effects compared to other networks. Secondary goals included: 1) determine if bipolar disorder is associated with similar patterns of accelerated aging as schizophrenia; 2) replicate prior associations between network efficiency and cognition; and 3) examine the effect of illness stage (i.e. chronic, early stage) on CON and FPN efficiency.

Methods & Materials

Participants

506 individuals that participated in one of three MRI studies (CT00762866; 1R01MH070560; 1R01MH102266) were included in this investigation. This included 197 healthy controls (HC) and 319 individuals with a schizophrenia spectrum disorder (schizophrenia, schizoaffective disorder, schizophreniform) or bipolar disorder type I with psychotic features (i.e. psychotic bipolar disorder). Psychotic disorder participants were recruited from the Psychotic Disorders Program in the Department of Psychiatry and Behavioral Sciences at Vanderbilt University Medical Center. Healthy subjects were recruited from Nashville and the surrounding area via advertisements and word of mouth. The studies were approved by the Vanderbilt Institutional Review Board and all study participants provided written informed consent.

Diagnosis was confirmed using the Structured Clinical Interview of the DSM-IV (SCID) (35). Exclusion criteria were similar across the three studies including age less than 16 or older than 65 (age 55, 1R01MH102266); estimated pre-morbid IQ less than 70 based on the Wechsler Test of Adult Reading (WTAR) (36); history of significant head trauma, medical illness or central nervous system disorder; pregnancy or lactation; substance abuse within the last one month for patients (three months, 1R01MH102266) or lifetime history of substance abuse/dependence in HC; first degree relative with a psychotic illness (HC only); and MRI contra-indicators.

Of the 506 individuals included in the study, 48 were excluded for neuroimaging processing failures (e.g. segmentation/normalization errors due to poor quality structural scans), 38 resting-state scans (32 psychosis patients, 6 HC) did not meet our QA threshold described below, and 2 subjects were identified as multivariate outliers (age and network efficiency were tested using Mahalanobis Distance). Thus, the final sample included 240 psychosis patients and 178 HC (Table 1).

Table 1:

Demographics and Participant Characteristics

Demographics Mean(SD)	Healthy Controls N=178	Psychosis N=240	Omnibus Statistic
Age	28.99 (10.45)	28.71 (10.73)	t(416)=.267, p=.790
Gender (M/F)	109/69	155/85	X2=.492, p=.483
Race (C/AA/O)	124/42/12	174/52/13	X2=2.55, p=.467
Parental Education	14.46 (2.24)	14.73 (2.65)	t(372)=1.02, p=.308
Personal Education	15.15 (2.22)	13.52 (2.16)	t(391)=7.26, p<.001
WTAR	111.28 (11.20)	102.33 (15.14)	t(415)=6.63, p<.001
Cognitive Factor Score	.535 (.56)	−.392 (.94)	t(413)=11.71, p<.001
Duration of Illness (Months)	-	87.88 (111.96)
PANSS Positive	-	16.29 (8.01)	-
PANSS Negative	-	14.03 (6.48)	-
PANSS General	-	29.64 (8.22)	-
PANSS Total	-	59.64 (18.56)	-
Montgomery-Asberg Depression Rating Scale	-	8.83 (8.52)	-
Young Mania Rating Scale	-	4.67 (7.39)	-
Chlorpromazine Equivalent (CPZ)	-	390.10 (237.07)
Schizophrenia/Psychotic Bipolar Disorder	-	177/63	-
Early/Chronic Psychosis	-	110/130	-

M=Male, F=Female; C=Caucasian, AA = African-American; O = Other Race; WTAR: Wechsler Test of Adult Reading; PANSS: Positive and Negative Syndrome Scale

Cognitive Testing

All participants completed the Screen for Cognitive Impairment in Psychiatry (SCIP) (37), a brief (10-15 minute) measure that assesses immediate verbal memory, working memory, verbal fluency, delayed verbal memory, and processing speed (see (21) for a description of the SCIP). General cognitive ability was estimated as the first factor of an exploratory principal axis factor analysis including SCIP sub-scores across all final subjects and explained 57% of the variance in performance across tasks. The first factor was the only factor with an eigenvalue >1.

Neuroimaging Data Acquisition and Preprocessing

Neuroimaging data were collected on one of two identical 3T Philips Intera Achieva scanners located at the Vanderbilt Institute for Imaging Sciences. Scanning parameters, number of participants scanned under each protocol, and imaging preprocessing steps are described in detail in the Supplement. Briefly, a 7 min. (eyes closed) or 10 min. (eyes open, fixation) echo-planar imaging (EPI) resting-state fMRI scan and T1-weighted anatomical (1mm isotropic resolution) were collected on each subject. FMRI data preprocessing was performed using the SPM12 package (http://www.fil.ion.ucl.ac.uk/spm/software/spm12/) and included correction for head motion using rigid body motion correction, spatial coregistration to T1-weighted anatomical images, spatial normalization to MNI-space, and spatial smoothing (6mm FWHM kernel).

Functional Connectivity

Following preprocessing, fMRI data were band-pass filtered (0.01-0.10 Hz) and BOLD time-series data were extracted from 264 regions-of-interest (ROIs) (6mm spheres corresponding to the Power atlas (38)). Following removal of nuisance signals related to head motion, white matter, and CSF, and the global grey matter signal regression, the ROI-to-ROI correlation matrix was calculated. A priori Power network assignments for each ROI were used to define the CON, FPN, subcortical, and visual network (29) (Supplement).

Multiple steps were taken to minimize the effects of head motion on functional connectivity estimates (Supplement). Briefly, 12 nuisance regressors, anatomical CompCor (44), and signal-to-noise ratio (SNR) calculations were applied. Subjects with median temporal SNR below the 5^th percentile or median intra-frame voxel displacement above the 95^th percentile of the entire sample were excluded from further analysis.

Global Efficiency

Network efficiency was calculated using the Brain Connectivity Toolbox (23) in MatLab (Mathworks). A priori network ROIs were extracted from thresholded whole brain graphs (5-10% strongest positive connections, see Supplement), resulting in a CON, FPN, subcortical, and visual sub-graph for each subject, at each threshold. The global efficiency algorithm was then applied to network-specific sub-graphs. Global efficiency yields a single metric estimating the potential for information transfer within a network and reflects network integration (23). Efficiency metrics calculated at each threshold (5%-10%) were averaged to produce a single estimate for each network, used in data analysis.

Statistical Analysis

Similar to previous work (29), presence of accelerated network aging was tested using linear regression analysis. Age, group (psychosis/HC), gender, race, study, and group by age interaction terms were entered as predictors of network efficiency. Significant interactions reflecting accelerated aging were further explored by examining within-group bivariate correlations and comparing correlations between groups using Fisher r-to-z transformation and subsequent one-tailed independent samples t-test of z-transformed coefficients. Relationships between age and efficiency were followed-up controlling for chlorpromazine (CPZ) equivalent and duration of illness to help determine whether medication use and illness chronicity were driving accelerated aging effects. The accelerated aging hypothesis was tested on efficiency of the visual and subcortical networks to determine specificity of our CON/FPN results.

The association between network efficiency and overall cognition were examined using linear regression with cognition (i.e. SCIP factor score) entered as the dependent variable and diagnostic group, gender, race, protocol, network metric, and metric by group interaction entered as predictors, allowing for analysis of behavioral consequences related to changes in efficiency.

Group differences in efficiency for illness stage (early psychosis/chronic psychosis/HC) were analyzed through repeated-measures ANOVA and follow-up univariate ANOVA controlling for age, gender, race, and protocol. These analyses were included to determine whether network efficiency is already reduced at illness onset compared to HC, to aid in our interpretation of the accelerated aging results (e.g. (39)). Follow-up analyses comparing just early and chronic psychosis, controlling for age, were also explored, as group differences that remain after controlling for age would suggest a substantial impact of illness wear-and-tear on the accelerated aging findings. Subcortical efficiency was included in these analyses to test for further replication of previously published findings (40). Early psychosis was defined as study participation within two years following onset of psychosis.

Results

Accelerated Aging of Functional Networks

CON efficiency was significantly negatively associated with age across all subjects (t₄₁₆=−5.61, p<.001). A significant group by age interaction (t₄₁₆=−2.15, p=.034) (Figure 1) was due to a stronger (z=2.34, p=.019) association between age and CON efficiency in psychosis (r=−.37, p<.001) than HCs (r=−.16, p=.036).

Figure 1: Differential association between age and diagnosis for CON and FPN efficiency.

A significant age × diagnosis interaction was observed when predicting CON global efficiency (t₄₁₆=−2.15, p=.034) and FPN global efficiency (t₄₁₆=−2.33, p=.02) but not subcortical network (t₄₁₆=−.270, p=.787) or visual network efficiency (t₄₁₆=−.761, p=.447), controlling for age, group, race, gender, and protocol. Shaded 95% confidence intervals.

Age was also negatively associated with FPN efficiency (t₄₁₆=−3.627, p<.001) with a significant group by age interaction (t₄₁₆=−2.33, p=.02) also due to a stronger (z=2.66, p=.008) negative association between age and FPN efficiency in psychosis (r=−.31, p<.001) than HC (r=−.05, p=.489).

To determine the impact of disorder chronicity on our findings, we examined the effect of CPZ-equivalent and duration of illness on associations between efficiency and age. Age remained significantly associated with efficiency when controlling for CPZ-equivalent (CON: t₁₉₃=−6.75, p<.001; FPN: t₁₉₃=−3.80, p<.001). CPZ-equivalent was not significantly associated with efficiency (CON: t₁₉₃=.33, p=.739; FPN: t₁₉₃=−.55, p=.580). Furthermore, both CON (t₂₂₇=−.4.15, p<.001) and FPN (t₂₂₇=−2.39, p=.018) were significantly associated with age after controlling for duration of illness.

To assess specificity of accelerated aging to higher-order cognitive networks, we tested for accelerated aging in two additional networks: the subcortical network, which may be influenced by antipsychotic drug use, and the visual network, which has shown preserved connectivity in healthy aging (16). Although age was significantly negatively associated with subcortical efficiency (t₄₁₆=−3.68, p<.001) no significant group interaction was observed (t₄₁₆=−.270, p=.787). In the psychosis group alone, the relationship between age and subcortical efficiency (t₁₉₃=−2.52, p=.012) was attenuated when controlling for CPZ-equivalent (t₁₉₃=−1.74, p=.083) and CPZ-equivalent was significantly associated with efficiency ((t₁₉₃=−2.69, p=.008). Visual network efficiency was not associated with age across subjects (t₄₁₆=−1.11, p=.267) with no significant group interaction (t₄₁₆=−.761, p=.447). Accelerated aging effects of the CON and FPN in psychosis remained when controlling for subcortical and visual network efficiency (CON: t₄₁₆=−2.15, p=.032; FPN: t₄₁₆=−2.30, p=.022).

We repeated the analysis of CON and FPN considering psychotic disorder diagnosis to determine if schizophrenia and psychotic bipolar disorder patients demonstrate similar patterns of accelerated aging (results discussed in detail in the Supplemental Material). Briefly, in each group, age was similarly associated with efficiency of CON (bipolar: r=−.41, schizophrenia: r=−.36) and FPN (bipolar: r=−.23, schizophrenia: r=−.33).

Network Efficiency and Cognitive Ability

CON efficiency was significantly associated with cognition across all subjects (t₄₁₃=2.17, p=.03) in the context of a significant group interaction (t₄₁₃=2.39, p=.018), driven by a positive association between CON efficiency and cognition in psychosis participants (r=.16, p=.012) but not HC (r=−.06, p=.420) (Figure 2).

Figure 2: General cognitive ability is associated with CON and FPN efficiency.

General cognitive ability was significantly positively associated with CON (t₄₁₃=2.17, p=.03) and FPN (t₄₁₃=3.26, p=.001) global efficiency across all subjects, when controlling for age, group, gender, race, and protocol. A significant group × efficiency interaction was observed in the CON, due to a significant positive association in the psychosis group (r=.16, p=.012) not observed in HC (r=−.06, p=.420). Shaded 95% confidence intervals.

FPN efficiency was also associated with cognition across all subjects (t₄₁₃=3.26, p=.001) with no significant group interaction (t₄₁₃=−.22, p=.830); the correlations were similar in psychosis (r=.12, p=.066) and HC (r=.23, p=.002).

Subcortical network efficiency was not associated with cognition across all subjects (t₄₁₃=1.58, p=.114) and there was no group interaction (t₄₁₃=1.18, p=.239; psychosis: r=.18, p=.005; HC: r=.10, p=.185).

We further examined the group by age interaction when predicting higher-order cognition, to better elucidate our findings. This interaction was not significant (t₄₁₃=−1.04, p=.301) as age was similarly related to cognition in both psychosis (r=−.27, p<.001) and HC (r=−.30, p<.001).

Finally, to test specificity of FPN and CON efficiency with current cognitive ability, relationships with pre-morbid, crystallized knowledge from the WTAR were assessed. Analyses revealed non-significant associations between CON/FPN efficiency and crystallized knowledge and continued relationships with current cognitive ability when controlling for crystallized knowledge (Supplement).

Group and Illness Stage Effects on Network Efficiency

While regression analysis revealed significantly different slopes between age and efficiency in each group (an estimate of rate of decline in our cross-sectional sample), we aimed to determine whether there were: a) group differences in overall network efficiency; and b) whether accelerated aging could be further understood through group differences based on illness-stage, to aid in inference about trajectory. In repeated-measures ANOVA with network (FPN/CON/subcortical) as a within-subjects factor and group (HC/psychosis) as a between-subjects factor, psychosis participants demonstrated significant overall reduction in network efficiency compared to HC (F_1,411=11.52, p=.001) (Figure 3a), a significant main effect of network (F_2,822=37.52, p<.001), and a significant group by network interaction (F_2,822=3.54, p=.029). Follow-up univariate ANOVA revealed that FPN (F_1,411=5.06, p=.025) and subcortical (F_1,411=12.76, p<.001) efficiency, but not CON (F_1,411=.16, p=.688) efficiency, were significantly reduced in psychosis patients.

Figure 3: Group differences in network efficiency.

Group differences in network efficiency were tested in a repeated-measure ANOVA with three networks as within-subject factors. A significant main effect of group was found for both HC vs. Psychosis (F_1,411=11.52, p=.001) and HC vs. Early Psychosis vs. Chronic Psychosis (F_2,410=7.89, p<.001). Follow-up Univariate ANOVAs demonstrated that this omnibus group difference was driven by reduced global efficiency in the FPN and subcortical network, primarily in the chronic psychosis patients.

Regarding illness stage, there was a significant overall group difference in network efficiency (F_1,410=7.89, p<.001) (Figure 3b), a significant main effect of network (F_2,820=43.75, p<.001) and a significant group by network interaction (F_2,820=4.08, p=.003). Univariate ANOVAs revealed no main effect of group for CON global efficiency (F_2,410=.44, p=.643) and no differences between HC and both early-stage (p=.398) and chronic patients (p=.888). FPN efficiency, however, showed a significant main effect of group (F_2,410=4.27, p=.015) with reduced FPN efficiency in chronic psychosis (p=.004) but similar FPN efficiency in early psychosis (p=.553) compared to HC. Subcortical efficiency also significantly differed between groups (F_2,410=10.66, p<.001), with chronic patients (p<.001) but not early psychosis patients (p=.317) demonstrating reduced subcortical efficiency compared to HC. After controlling for age, early and chronic patients did not significantly differ on CON (F_2,237=2.68, p=.103), FPN (F_2,237=1.12, p=.291), or subcortical (F_2,237=3.58, p=.06) efficiency.

Finally, while connectivity of the whole brain was lower in psychosis (t₄₁₅=1.88, p=.06), group differences in efficiency were largely unchanged when including whole brain connectivity as a covariate, suggesting graph results are minimally influenced by overall connectivity differences (41)(Supplement).

Discussion

In a large sample of psychotic disorder patients and HCs, we confirm evidence of accelerated functional network aging in psychosis. Accelerated aging was observed in the FPN and CON, two networks that show the earliest evidence of connectivity decline in prior studies of healthy aging (15,18). Furthermore, prior findings of stable visual network connectivity were confirmed in both HCs and psychosis patients, strengthening the evidence that differential associations between network efficiency and age in the CON and FPN are due to an accelerated aging process, as opposed to a general reduction in network efficiency. When controlling for CPZ-equivalence, relationships between efficiency and age remained for CON and FPN, but were attenuated in the subcortical network, arguing against a general medication effect on network topology. Accelerated aging of FPN and CON efficiency has real-world consequences, as lower efficiency of these networks was associated with worse general cognition. Finally, FPN/CON network efficiency were intact in early psychosis, but FPN efficiency was significantly reduced in the chronic state, providing additional validation for efficiency reductions following illness onset. This implies that intervention even after the first episode of psychosis may provide significant benefits.

Typically, functional network connectivity declines with age, occurring earliest within higher-order cognitive networks (15,18). In older versus younger adults, FPN and CON efficiency, but not visual or somatosensory network efficiency, is reduced (17) in the context of reduced whole brain modularity, suggesting that lower within-network efficiency results in a less segregated whole brain system. Older adults’ functional networks are also more susceptible to perturbation, with a larger decline in network efficiency following computational “lesioning” (i.e. node removal) than is seen in younger adults (42). Comparing cognitive with emotion processing networks, only cognitive networks exhibit reduced functional connectivity in older adults, which is associated with poorer cognitive performance (43). Furthermore, reduced effective connectivity between the anterior insula and dorsal anterior cingulate cortex, two hub nodes of the CON, is reduced in older adults and associated with worse cognition (44). In the current study, CON efficiency was also significantly associated with age in our healthy cohort, however the magnitude of this relationship was significantly less than in patients with psychosis. This suggests that a normative pattern of decline in network integrity occurs in psychotic disorders, but the rate of decline is more rapid than expected in healthy aging.

Accelerated decline of resting-state network integrity in psychotic disorders occurs in the context of abnormal resting-state functional connectivity throughout the life course. Ultra-high risk patients demonstrate significantly reduced anti-correlation between default mode network and regions comprising the FPN and CON (45–47), and there is reduced within and between-network connectivity of the FPN and CON in chronic states (48)(49,50). Furthermore, reduced connectivity and efficiency of the CON has occurs across the psychosis spectrum, including in psychosis-spectrum youths (51) and individuals in the general population who endorse psychotic-like experiences (52). In our data, CON efficiency was not reduced in either chronic or early psychosis, even in the context of accelerated aging, suggesting that CON efficiency is intact, and possibly even slightly greater in early psychosis than HC, although the difference was non-significant. Given the presence of connectivity alterations in high-risk and psychosis-spectrum youths, we were surprised to find intact CON and FPN efficiency in our early psychosis patients; however there is growing evidence that dysconnectivity of higher-order cognitive networks is not present in early psychosis. For instance functional connectivity within the FPN and CON is normal in medication-naive first episode patients (53,54), global efficiency abnormalities were not present at baseline or 12 months later in an early psychosis sample (55), and analysis of network topology during a cognitive control task has revealed intact global efficiency in first episode psychosis versus healthy controls (56). Findings of intact network efficiency in early psychosis, and non-significant differences in network efficiency for early and chronic patients when controlling for age, aid our understanding of network trajectories (39); together, they suggest that network efficiency is intact in early psychosis and the network alterations more reliably observed in chronic states may be manifestations of an accelerated aging process.

Accelerated aging in psychotic disorders is a growing literature, based on mounting evidence that processes observed in healthy aging are occurring earlier and/or at a more rapid rate in patients with psychosis (57). A recent review reported fairly consistent evidence of increased inflammation, oxidative stress, prolactin, and insulin resistance in schizophrenia (58). While the influence of environmental factors on accelerated aging have yet to be fully explored, abnormal glucose tolerance and metabolic profiles have been observed in antipsychotic-naive schizophrenia patients after controlling for a multitude of environmental confounds (59,60). Furthermore, gray matter volume and white matter integrity demonstrate evidence of accelerated aging in multiple studies, and are unrelated to antipsychotic drug use (10,11) or smoking (61). Our finding that CON/FPN efficiency were unrelated to CPZ-equivalence, but subcortical efficiency was, fits well with previous literature demonstrating that antipsychotics exert greatest effect in subcortical regions (62) and chronic administration of antipsychotics decreases striatal neuron firing (63). Specificity of accelerated aging in CON and FPN, but not subcortical network, therefore suggests that medication use is not driving accelerated aging of higher-order networks. Although still an open question, degradation of gray matter volume and white matter pathways may contribute to alterations in functional connectivity observed in the current study; a hypothesis recently supported by accelerated decline of white matter tracks associated with higher-order cognitive functioning in schizophrenia, not seen in tracks supporting sensorimotor information (61).

Finally, accelerated aging of FPN and CON efficiency has important implications for cognitive functioning in a population already plagued with significant cognitive impairment. Activation in the FPN and CON supports domain-general cognitive ability, with regions in the FPN modulating activity for adaptive-task control (e.g. error cues) and regions in the CON maintaining activation for stable-task control (64). Reduced efficiency of these networks has now been associated with poorer general cognition in four independent datasets (40, 52, 65) providing strong evidence that the functional topology of these networks support cognitive ability. It is notable, however, that a similar relationship between age and cognitive ability was observed in both groups, indicating that more rapid decline in FPN/CON efficiency in psychosis is not leading to a significantly steeper decline in cognition. Although speculative, this discrepancy may be due to our use of a relatively brief cognitive screen (SCIP), such that differential age effects may have been detected using a more comprehensive cognitive battery. Lastly, we show specific associations between FPN/CON efficiency and fluid, as opposed to crystallized, cognitive ability. This finding mirrors data in older adults, who typically show stable crystallized knowledge, such as word reading, and declining higher-order cognitive abilities, such as processing speed and verbal learning (68). Given the recently shown impact of accelerated aging on functional outcome (9), accelerated decline of information integration within high-order cognitive networks implies an accelerated decline of functional outcome, which may in turn contribute to the increased mortality rates recently reported in schizophrenia (69).

The present study has several limitations. First, the data presented are cross-sectional and therefore cannot speak directly to changes in network efficiency over each individual’s lifetime. Review of accelerated aging studies has noted this as a major limitation of most accelerated aging reports (58); however given the relatively gradual changes in markers of aging, longitudinal studies over decades may be necessary to appropriately address this issue. Another limitation is the potential influence of confounding factors, for instance antipsychotic drug use and metabolic disease. While not directly addressable here, the lack of accelerated aging in the subcortical network, continued accelerated aging in the CON/FPN when CPZ-equivalent was included as a covariate, and the normative relationship between age and visual network efficiency in psychosis, argues against a strong influence of antipsychotics, metabolic disease or smoking on our results, as cardiovascular risk factors should influence connectivity whole brain-wide, not within specific networks. We also had several limitations for data collection, for instance we cannot rule out that some subjects fell asleep during their resting-state scans, and also we used slightly different scanning protocols across studies, which we addressed by including protocol as a covariate. Finally, the age range for our sample was <65, truncating our ability to detect influences of age on efficiency in older adults.

Conclusions

Here, in a cross-sectional study, we present compelling evidence that functional integration of the fronto-parietal and cingulo-opercular networks, which support general cognitive ability, are subject to differential age effects in individuals with a primary psychotic disorder. These findings may explain network abnormalities observed in chronic patients that are often not observed in first episode psychosis. What is hopeful about these findings is that efficiency of these critical networks may be responsive to intervention during the early stages of a psychotic disorder. In fact, moderate aerobic exercise and cognitive training have both been found to increase functional connectivity within the FPN and CON in older adults (70, 71), and exercise increases some areas of cognition in patients with schizophrenia (72). If connectivity and cognition can be increased or protected in early stages of psychosis, functional outcome and ultimately mortality have the potential for improvement even following the diagnosis of a psychotic disorder.