Studies directly testing the accelerated ageing hypothesis using different methodology

Nine studies using non-BrainAGE neuroimaging in early psychosis largely supported disrupted age-related brain trajectories. In a 5-year study, van Haren et al ^{Reference Pol, Schnack, Cahn, Brans and Carati77} found accelerated grey matter loss and ventricular expansion before the age of 45 years, suggestive of early neurodegeneration. Pham et al ^{Reference Pham, Sasabayashi, Takahashi, Takayanagi, Kubota and Furuichi78} reported progressive frontotemporal gyrification loss in individuals with schizophrenia but not in those with schizotypal disorder. Dukart et al ^{Reference Dukart, Smieskova, Harrisberger, Lenz, Schmidt and Walter79} found grey matter decline in both first-episode patients and those in at-risk states. Lin et al ^{Reference Lin, Li, Zhou, Deng, Ma and Wang80} observed steeper cortical thinning with age in drug-naïve patients. Sheffield et al ^{Reference Sheffield, Rogers, Blackford, Heckers and Woodward81} and Kuo et al ^{Reference Kuo, Lee, Hung, Liu, Lee and Chung82} demonstrated functional declines in key cognitive networks (Sheffield et al observed an accelerated decline in the functional integration and efficiency of the frontoparietal and cingulo-operatic networks, whereas Kuo et al used structural covariance networks to demonstrate associations of schizophrenia with advanced ageing across multiple systems, notably within the default mode network and salience network). Shivakumar et al ^{Reference Shivakumar, Kalmady, Rajasekaran, Chhabra, Anekal and Narayanaswamy83} reported telomere shortening associated with hippocampal volume loss in drug-free patients. Chiapponi et al ^{Reference Chiapponi, Piras, Fagioli, Girardi, Caltagirone and Spalletta84} identified accelerated microstructural degeneration in the hippocampus, whereas Thormodsen et al ^{Reference Thormodsen, Rimol, Tamnes, Juuhl-Langseth, Holmén and Emblem85} noted absent normative cortical thinning in early-onset schizophrenia. An exception was the study of Bose et al, ^{Reference Bose, Mackinnon, Mehta, Turkheimer, Howes and Selvaraj86} which found increased white but reduced grey matter decline.

Fifteen 15 studies provided mixed support for accelerated ageing in patients with chronic schizophrenia. Man et al ^{Reference Man, Ding, Chai, An, Liu and Qin87} found a small but significant BAG, and Rowland et al ^{Reference Rowland, Krause, Wijtenburg, McMahon, Chiappelli and Nugent88} reported steeper declines in medial frontal γ-aminobutyric acid. Longitudinal cortical thinning with downregulation of specific genes was observed by González-Peñas et al, ^{Reference González-Peñas, Alloza, Brouwer, Díaz-Caneja, Costas and González-Lois89} whereas De Picker et al ^{Reference Picker, Ottoy, Verhaeghe, Deleye, wyffels and Fransen90} reported age-related microglial activation. Kochunov et al ^{Reference Kochunov, Chiappelli, Wright, Rowland, Patel and Wijtenburg91,Reference Kochunov, Ganjgahi, Winkler, Kelly, Shukla and Du92} described accelerated white matter decline independent of perfusion, and Sheffield et al ^{Reference Sheffield, Repovs, Harms, Carter, Gold and Macdonald93} found worsening network inefficiency. Additional studies showed selective age-related changes in grey or white matter: Wright et al ^{Reference Wright, Kochunov, Chiappelli, McMahon, Muellerklein and Wijtenburg94} confirmed white matter ageing, and Kochunov et al ^{Reference Kochunov, Glahn, Rowland, Olvera, Winkler and Yang95} linked this to genetic risk. Only Voineskos et al ^{Reference Voineskos, Lobaugh, Bouix, Rajji, Miranda and Kennedy96} reported no evidence of accelerated frontotemporal tract decline.

Indirect imaging evidence relating to accelerated ageing

Although not specifically designed to test the accelerated ageing hypothesis, 12 imaging studies identified through PROSPERO offered relevant indirect evidence. A meta-analysis by Hua and Mathalon ^{Reference Hua and Mathalon97} reported widespread morphometric abnormalities consistent with advanced brain age in patients with early psychosis. Longitudinal work by Pruessner et al ^{Reference Pruessner, Lepage, Collins, Pruessner, Joober and Malla98,Reference Pruessner, Bechard-Evans, Pira, Joober, Collins and Pruessner99} showed hippocampal volume loss linked to hypothalamic–pituitary–adrenal axis dysfunction, whereas Hua et al ^{Reference Hua, Cummings, Roach, Fryer, Loewy and Stuart100} found progressive disruption in structural connectome integrity. Other studies of early psychosis, such as that of Skudlarski et al, ^{Reference Skudlarski, Schretlen, Thaker, Stevens, Keshavan and Sweeney101} identified white matter abnormalities, but findings from the studies of Schuster et al, ^{Reference Schuster, Schuller, Paulos, Namer, Pull and Danion102} Mamah et al, ^{Reference Mamah, Chen, Shimony and Harms103} Feng et al ^{Reference Feng, Palaniyappan, Robbins, Cao, Fang and Luo52} and Seitz et al ^{Reference Seitz, Rathi, Lyall, Pasternak, del Re and Niznikiewicz104} were less conclusive. Koutsouleris et al ^{Reference Koutsouleris, Davatzikos, Borgwardt, Gaser, Bottlender and Frodl67} observed atrophy compatible with accelerated ageing, although this was confounded by illness duration and medication.

Oxidative stress, inflammation and metabolism

Evidence from meta-analytic and early-course data was robust, suggesting systemic physiological dysregulation. Jorgensen ^{Reference Jorgensen, Baago, Rygner, Jorgensen, Andersen and Kessing4} meta-analysed oxidative-stress-induced nucleic acid damage (82 studies, comprising 10 151 patients and 10 532 controls), describing elevated levels in patients with psychosis (medium to large effect) and suggesting involvement in the pathophysiology of age-related medical conditions. Notably, these markers appeared to be intrinsic rather than solely treatment driven as in FEP treatment-naïve samples. Garcia-Rizo described abnormal glycaemic homeostasis ^{Reference Garcia-Rizo, Fernandez-Egea, Bernardo and Kirkpatrick118} and higher blood cell counts, ^{Reference Garcia-Rizo, Casanovas, Fernandez-Egea, Oliveira, Meseguer and Cabrera119} whereas Fernandez-Egea described reduced biologically active testosterone. ^{Reference Fernandez-Egea, García-Rizo, Miller, Parellada, Justicia and Bernardo120} These findings in treatment-naïve participants indicate that metabolic and hormonal dysregulation occurs before the onset of psychosis, representing a pre-existing risk that may contribute to earlier morbidity and subsequent mortality.

In addition to these core findings, the literature converged on the following four distinct mechanistic domains:

1. (a) Mitochondrial function: mitochondrial signatures suggest widespread impairment in cellular energy homeostasis, with Shivakumar et al describing reduced DNA copy number variations, ^{Reference Shivakumar, Rajasekaran, Subbanna, Kalmady, Venugopal and Agrawal121} whereas Garcia de la Cruz observed significantly elevated levels of circulating cell-free mitochondrial DNA. ^{Reference Juarez-Rojop, Tovilla-Zarate, Nicolini and Genis-Mendoza122}

2. (b) Oxidative stress: alongside nucleic acid damage, specific markers of ageing and oxidation are altered. Lotan et al described a reduction in physiological age-dependent iron accumulation and decreased ferritin, implying an imbalance that theoretically implies oxidative damage; ^{Reference Lotan, Luza, Opazo, Ayton, Lane and Mancuso123} Jørgensen et al described higher urinary excretion of systemic DNA and RNA oxidation markers independent of treatment; ^{Reference Jorgensen, Broedbaek, Fink-Jensen, Knorr, Greisen Soendergaard and Henriksen124} and Zhang described decreased levels of vitamin B12 and folate-dependent enzyme methionine synthase forms. ^{Reference Zhang, Hodgson, Trivedi, Abdolmaleky, Fournier and Cuenod125} In addition, Hagen described a higher marginal accumulation rate of advanced glycation end products over a 2-year follow-up in a treated FEP sample. ^{Reference Hagen, Sutterland, Edrisy, Tan and de Haan126} However, discrepancies arose; Rebouças found no changes in glutathione peroxidase or protein carbonyl content, ^{Reference Sartori, Librenza-Garcia, Rabelo-da-Ponte, Massuda and Czepielewski127} and Busch found no difference in hair cortisol concentrations in forensic autopsies. ^{Reference Busch, Wang, Lynnerup, Jacobsen, Jørgensen and Linnet128}

3. (c) Multi-omics, inflammation and metabolism: broader profiling revealed disruption in inflammatory and metabolic pathways. Campeau et al performed a multi-omics analysis, finding altered proteomic and metabolomic signatures that overlapped with frailty and metabolic ageing. ^{Reference Campeau, Mills, Stevens, Rossitto, Meehan and Dorrestein129} Postmortem studies aligned, with Yu et al describing an ‘older’ lipidome state in the prefrontal cortex ^{Reference Yu, He, Zubkov, Huang, Kurochkin and Yang130} and Weissleder et al describing decreased IGF1 mRNA levels involved in metabolic regulation. ^{Reference Weissleder, Webster, Barry and Shannon Weickert131} Other studies described higher levels of inflammatory markers (CrP, chemokines, TNF-α, IL-6, insulin–glucose homeostasis parameters) associated with psychopathology, chronicity and ageing endophenotypes, including increased eotaxin levels ^{Reference Sirivichayakul, Kanchanatawan, Thika, Carvalho and Maes48,Reference Hong, Lee, Martin, Soontornniyomkij, Soontornniyomkij and Achim132–Reference Squassina, Manchia, Pisanu, Ardau, Arzedi and Bocchetta137} accompanied by increased allostatic load. ^{Reference Nugent, Chiappelli, Rowland and Hong138} Allostatic load (Table 1), which is defined by several physiological measures that reflect the cumulative physiological effect of stress, increases from roughly ages 20 to 60 years, after which it is usually relatively constant. ^{Reference Crimmins, Johnston, Hayward and Seeman139}

4. (d) Vascular endothelium and gut microbiota: Nguyen et al, in separate studies, analysed the gut microbiota, describing differences in beta-diversity associated with microbial metabolic pathways linked to atherosclerosis and immunity ^{Reference Nguyen, Kosciolek, Daly, Vázquez-Baeza, Swafford and Knight140} and in levels of vascular endothelial biomarkers, which are key players in the intersection of inflammatory response and vascular risk. ^{Reference Nguyen, Dev, Chen, Liou, Martin and Irwin141}

The confluence of increased inflammatory levels, oxidative stress biomarkers and metabolic load suggests a complex interplay underlying the ageing process in schizophrenia; however, interpretation requires some caution. The evidence gathered is cross-sectional and vulnerable to confounders. Markers could have been influenced by lifestyle habits or environmental factors, such as childhood trauma, that promote worse insulin–glucose homeostasis.

Telomere length

Four meta-analyses described a significant decrease in telomere length. ^{Reference Ayora, Fraguas, Abregú-Crespo, Recio, Blasco and Moises3,Reference Rao, Kota, Li, Yao, Tang and Mao142–Reference Russo, Prinzi, Proietti, Lamonaca, Frustaci and Boccia144} Ayora et al (22 studies, 4145 patients and 4184 controls) described a small-to-medium effect (standardised mean difference (SMD) = −0.388)], ^{Reference Ayora, Fraguas, Abregú-Crespo, Recio, Blasco and Moises3} and Russo et al ^{Reference Russo, Prinzi, Proietti, Lamonaca, Frustaci and Boccia144} (18 studies, 1827 patients and 1741 controls) reported a medium effect size (SMD = −0.52), whereas Rao et al ^{Reference Rao, Kota, Li, Yao, Tang and Mao142} and Polho et al ^{Reference Polho, De-Paula, Cardillo, dos Santos and Kerr143} performed similar main analyses (11 studies, 1243 patients and 1274 controls) and found a medium effect size (SMD = −0.48). Another meta-analysis involving a wide array of mental diagnoses including schizophrenia (six studies, 772 patients and 763 controls) described a smaller effect size for leukocyte telomere length shortening (Hedge’s g = −0.23). ^{Reference Darrow, Verhoeven, Révész, Lindqvist, Penninx and Delucchi133}

Several individual studies supported a reduction in telomere length: Fernandez-Egea et al reported this finding in an antipsychotic-naïve FEP sample, ^{Reference Fernandez-Egea, Bernardo, Heaphy, Griffith, Parellada and Esmatjes145} Zakharova et al in individuals with schizophrenia specifically with suicidal risk, ^{Reference Zakharova, Bravve, Mamedova, Kaydan, Ershova and Martynov146} Squassina et al in patients with treatment-resistant schizophrenia ^{Reference Squassina, Manchia, Pisanu, Ardau, Arzedi and Bocchetta137} and Russo et al in patients below 50 years of age. ^{Reference Russo, Prinzi, Proietti, Lamonaca, Frustaci and Boccia144} Van Mierlo et al, in a study of post-mortem brain, described a significant decrease in telomere length in the superior temporal gyrus, ^{Reference van Mierlo, Wichers, He, Sneeboer, Radstake and Kahn147} suggesting that cell senescence was affected in white but not grey matter. Czepielewski et al, ^{Reference Czepielewski, Massuda, Panizzutti, Grun, Barbé-Tuana and Teixeira26,Reference Czepielewski, Massuda, Panizzutti, da Rosa, de Lucena and Macêdo148} Rao et al ^{Reference Rao, Kota, Li, Yao, Tang and Mao142} and Galletly et al. ^{Reference Galletly, Dhillon, Liu, Balzan, Hahn and Fenech149} described shorter telomere lengths in a male sample.

Other authors reported no differences or an increase in telomere length. In individuals with antipsychotic-naïve FEP, Shivakumar et al (studying a mixed sample of antipsychotic-naïve and antipsychotic-free individuals) described no differences; however, males displayed a non-significant reduction in telomere length compared with female and control male participants, with worse psychopathology associated with reduced telomere length. ^{Reference Shivakumar, Kalmady, Rajasekaran, Chhabra, Anekal and Narayanaswamy83} Li et al described no differences in telomere length; ^{Reference Li, Hu, Zong, He, Wang and Dai150} however after 8 weeks, treatment responders showed a trend towards longer telomere length compared with non-responders, and Talarico et al described longer telomere lengths. ^{Reference Talarico, Xavier, Ota, Spindola, Maurya and Tempaku151}

Çevik et al described a non-significant reduction in telomere length, whereas childhood adversities were associated with reduced telomere length. ^{Reference Çevik, Mançe-Çalışır, Atbaşoğlu, Saka, Alptekin and Üçok152} Riley et al described no differences in telomere length, but significant gender differences arose in the relationship of early trauma with telomere length (in male patients). ^{Reference Riley, Perrin, Vaez-Azizi, Ruby, Goetz and Dracxler153} Monroy-Jaramillo et al described no differences in telomere length; however, when participants were stratified by metabolic risk due to antipsychotics, patients receiving high-metabolic-risk treatment presented reduced telomere length. ^{Reference Monroy-Jaramillo, Rodríguez-Agudelo, Aviña-Cervantes, Roberts, Velligan and Walss-Bass154} Wolkowitz et al reported no diagnostic differences; however, they analysed the effects of biological gender and described prematurely shortened telomere length in women with schizophrenia, with a differential gender effect in the pattern of decline pattern. ^{Reference Wolkowitz, Jeste, Martin, Lin, Daly and Reuter155} Pisanu et al reported mixed results but found that specific pleiotropic variants may contribute to the cellular ageing processes. ^{Reference Pisanu, Congiu, Meloni, Paribello, Patrinos and Severino156}

Overall, the current literature offers moderate support for telomere shortening as a marker of accelerated ageing in schizophrenia. All four meta-analyses consistently showed significant small-to-medium size effects (ranging from −0.39 to −0.52); however, individual studies showed some inconsistencies and confounders, particularly those due to differential aspects of biological gender, influence of antipsychotic treatment, stressful environmental events and psychopathology (suicidal behaviour), which seemed to promote telomere shortening.

Epigenetics

The epigenetic landscape in schizophrenia is complex, with results heavily dependent on the methodology and tissue type used. A recent systematic review and meta-analysis of eight studies (N = 6490) by Chrusciel ^{Reference Chrusciel, Orso, de Mattos, Fries, Kristensen and Grassi-Oliveira5} found no significant difference in biological age between patients and controls when all epigenetic clocks globally were analysed globally. However, this global null result masked significant heterogeneity. When participants were stratified by specific clock type, a small but significant ‘accelerated ageing’ effect emerged for the PhenoAge and Hannum clocks (SMD = 0.29). Unpacking this heterogeneity requires distinguishing between illness stages and the specific biological dimensions captured by different clocks.

With respect to markers related to the trajectory, evidence suggests a dynamic shift in epigenetic status over the course of the illness. In contrast to the accelerated ageing hypothesis, early-stage and at-risk cohorts often displayed an unexpected ‘epigenetic deceleration’ or youthfulness. Naïve FEP samples, ^{Reference Talarico, Xavier, Ota, Spindola, Maurya and Tempaku151,Reference Li, Zong, Li, He, Tang and Hu157} youth at familial risk and small-sample cohorts ^{Reference Jeremian, Malinowski, Chaudhary, Srivastava, Qian and Zai158} have shown decelerated ageing according to Horvath’s clock. Epigenetic deceleration (or ‘youthfulness’) refers to a biological clock that seems younger than chronological age owing to a relative maturational delay in epigenetic changes, rather than a reversal of DNA methylation errors.

However, this pattern seemed to reverse with illness chronicity and treatment. Longitudinal data indicated that although patients may appear ‘decelerated’ at baseline, an accelerated pattern emerged after 8–10 weeks of antipsychotic treatment. ^{Reference Talarico, Xavier, Ota, Spindola, Maurya and Tempaku151,Reference Li, Zong, Li, He, Tang and Hu157} This shift was consistent with findings in chronic populations, in which older adults – particularly females at high genetic risk – exhibited accelerated PhenoAge and pace-of-ageing (DunedinPACE). ^{Reference Caspi, Shireby, Mill, Moffitt, Sugden and Hannon159} Higgins-Chen et al ^{Reference Higgins-Chen, Boks, Vinkers, Kahn and Levine160} suggested that this reversal may be driven by slowing of ‘mitotic clocks’ (linked to cell division and immunity), while ‘mortality clocks’ accelerate owing to accumulated damage. An exception to this chronic acceleration pattern was found by Okazaki et al, ^{Reference Okazaki, Numata, Otsuka, Horai, Kinoshita and Sora161} who reported decreased epigenetic age (Hannum) in patients undergoing long-term hospital stays, although replication results were mixed.

Regarding the role of lifestyle confounders, the ‘acceleration’ observed in chronic stages seemed to be strongly driven by secondary lifestyle factors rather than intrinsic disease mechanisms. The ‘GrimAge’ clock, a strong predictor of mortality, showed accelerations of up to 5 years in schizophrenia, but this effect was largely attenuated after controlling for smoking and age-associated proteins. ^{Reference Higgins-Chen, Boks, Vinkers, Kahn and Levine160} Teeuw et al ^{Reference Teeuw, Ori, Brouwer, de Zwarte, Schnack and Hulshoff Pol73} found that accelerations in PhenoAge often did not survive correction for smoking, whereas Rosoff et al ^{Reference Rosoff, Hamandi, Bell, Mavromatis, Park and Jung162} suggested that smoking, rather than a direct genetic link to longevity, explained the apparent ageing effects. Despite these confounders, sensitivity analyses still linked Hannum’s clock to mortality: ^{Reference Kowalec, Hannon, Mansell, Burrage, Ori and Ophoff163} Muntané et al ^{Reference Bosch, Martorell, Navarro and Vilella164} described a polygenic overlap between schizophrenia and shorter lifespan, whereas Segura et al ^{Reference Segura, Prohens, Mezquida, Amoretti, Bioque and Ribeiro165} found accelerated patterns using DNAmTL in patients experiencing clinical relapse. Previous findings have suggested that distinct clocks capture different signals: some measure intrinsic biological ageing (often normal or decelerated in patients with schizophrenia), whereas others capture cumulative environmental stress (accelerated).

With respect to divergence in brain tissue, peripheral blood findings did not necessarily mirror brain ageing. Direct examination of post-mortem brain tissue generally failed to support the accelerated ageing hypothesis. Large analyses of blood and brain tissue by McKinney et al ^{Reference McKinney, Lin, Ding, Lewis and Sweet166} and Wu et al ^{Reference Wu, Ye, Wang and Zhao167} reported no differences or delayed acceleration. In direct brain analyses, Voisey et al ^{Reference Voisey, Lawford, Morris, Wockner, Noble and Young168} and McKinney et al ^{Reference McKinney, Lin, Ding, Lewis and Sweet169} described no significant methylation differences. Liu et al ^{Reference Liu, Qi, Cheng, Meng, Yang and Pan170} described decelerated epigenetic ageing in the cerebellum of aged patients However, although methylation clocks remained neutral, transcriptomic studies revealed functional shifts. Arakelyan et al ^{Reference Arakelyan, Avagyan, Kurnosov, Mkrtchyan, Mkrtchyan and Zakharyan116} found that the gene expression profiles of patients aged 50–65 and 65–80 years closely resembled those of healthy controls more than 85 years old. Other studies have identified disrupted age-related gene expression networks ^{Reference Kim, Jo, Webster and Lee171,Reference Sabunciyan172} and histone modifications ^{Reference Tang, Dean and Thomas173} that imply an ‘older’ functional state. Conversely, Torkamani et al ^{Reference Torkamani, Dean, Schork and Thomas117} described age-related dysregulation in developmental gene networks in an opposite direction to that associated with ageing.

Overall, current epigenetic evidence does not support a uniform ‘accelerated ageing’ phenotype. Instead, the data point to a decoupled process, with intrinsic deceleration or delay in early development and brain tissue, overlaid by a secondary lifestyle-driven acceleration in peripheral markers that accumulates with chronicity and treatment exposure.