Clinical Sciences and Clinical Research

Abstract

Background: Post-intensive care syndrome (PICS) encompasses cognitive, physical, and psychiatric impairments that persist after critical illness. The construct has organized survivorship care, but it is defined by what survivors experience rather than by the biology that produces it. This symptom-level definition may help explain why broad, undifferentiated interventions have yielded inconsistent results.

Objective: We argue for disaggregating PICS into biologically defined subtypes and outline a testable framework for mechanism-matched neurorecovery, using glucagon-like peptide-1 (GLP-1) receptor agonists as a worked example.

Hypothesis: We propose that PICS is best understood not as one condition but as a family of overlapping neuroimmunometabolic processes—microglial activation, blood-brain barrier disruption, insulin resistance, endothelial injury, gut dysbiosis, and unremitting catabolism.

Framework: Drawing on the logic that resolved acute respiratory distress syndrome into subphenotypes, we set out eight candidate mechanistic axes (putative endotype dimensions to be tested empirically) and map the documented actions of GLP-1 receptor agonists onto each. The mapping is deliberately two-sided: it marks axes where these agents are mechanistically promising and the catabolic-sarcopenic axis, where, by promoting weight loss, they may instead cause harm unless paired with anabolic co-therapy.

Translational implications: Operationalizing the framework will require validated biomarkers, deeply phenotyped survivor cohorts, unsupervised methods to test whether the proposed clusters are real rather than imposed, and endotype-stratified trials.

Conclusion: For PICS, GLP-1 receptor agonism is investigational and unapproved; endotype-guided precision neurorecovery remains to be empirically validated. The framework's value is heuristic—it supplies a falsifiable plan rather than a treatment—and its central claim is that moving PICS from a symptom-defined syndrome toward a biologically stratified one may be a prerequisite for testing rational neurorecovery strategies.

Introduction 

Improved survival after critical illness has shifted the clinical challenge from acute mortality to long-term recovery. As survival from sepsis, acute respiratory failure, and major trauma has improved, a large and growing population now carries the durable aftermath of intensive care: delayed executive function, fragmented memory, affective dysregulation, fatigue, and sarcopenia. Collectively termed post-intensive care syndrome (PICS), these sequelae affect a substantial share of intensive care unit (ICU) survivors within the first post-discharge year, with comparable burdens after sepsis, critical COVID-19, and major burns [1–5]. The origins are not interchangeable: the sterile inflammatory storm of major trauma or burns and the protracted immune provocation of sepsis or COVID-19 plausibly imprint different recovery trajectories—a distinction we return to below. 
The construct has been clinically valuable, drawing attention to a neglected phase of illness and prompting survivorship pathways. Its breadth, however, is also a limitation. PICS is defined phenomenologically and stratified, when stratified at all, by severity rather than biology. The interventions produced by this framing—structured follow-up, early mobilization, nutritional optimization, transfusion strategies, trauma-focused psychotherapy—have been applied with little reference to mechanism, and their results have been correspondingly inconsistent [6–9]; pharmacological attempts to modify acute brain injury have also disappointed [10]. We contend that the problem is not that PICS is untreatable, but that it is heterogeneous, and that treating a heterogeneous condition as a single entity may dilute any mechanism-specific therapy until it becomes statistically invisible. 
One idea organizes what follows—the PICS Endotype Hypothesis: that PICS comprises biologically distinct but overlapping neuroimmunometabolic processes, and that recognizing them may be a prerequisite for testing rational therapy. We examine GLP-1 receptor agonists as a candidate platform for precision neurorecovery because their pharmacology converges on the mechanisms we propose. We state at the outset, and repeat where it matters, that these agents are not an established treatment for PICS; the case rests on biological plausibility and testable predictions, not on current practice. Figure 1 presents the conceptual model.
A terminological caution. In the critical care literature, “PICS” has two unrelated meanings: post-intensive care syndrome and persistent inflammation, immunosuppression, and catabolism syndrome. Here, PICS always denotes the former; for the latter, we write PIICS [11]. Box 1 provides key definitions for non-specialist readers.

Methods

Study design: This work is a hypothesis and conceptual framework based on a SANRA-informed critical narrative review. Its review component is structured and purposive rather than systematic: its aim is to integrate heterogeneous evidence into a coherent, falsifiable framework, not to estimate a pooled effect. Consistent with this design, and because the design does not require them, we did not perform a PRISMA or PRISMA-ScR protocol, meta-analysis, GRADE rating, or formal risk-of-bias appraisal. Reporting was guided by SANRA (Scale for the Assessment of Narrative Review Articles). The search was structured and purposive, guided by mechanistic relevance, and intended to assemble the evidence most useful for building and stress-testing a hypothesis.

Information sources and searches: We used PubMed/MEDLINE as the primary database, complemented by Scopus, Web of Science, Google Scholar, and author-assembled bibliographic files. The literature spans 2010–2026, with priority to 2021–2026 to reflect the rapid evolution of GLP-1 neuroscience and post-intensive care research; the last comprehensive search was on 31 May 2026. Principal terms, combined with Boolean operators, included post-intensive care syndrome, ICU survivors, sepsis-associated encephalopathy, delirium, neuroinflammation, immunometabolism, GLP-1 receptor agonists (semaglutide, liraglutide, exenatide, tirzepatide), blood-brain barrier, microglia, mitochondrial dysfunction, insulin resistance, sarcopenia, gut-brain axis, endothelial dysfunction, and persistent inflammation–immunosuppression–catabolism syndrome.

Search transparency and two-tier strategy: To allow approximate reproduction of the corpus without converting this into a systematic review, we used a deliberately asymmetrical two-tier strategy. The survivorship core underwent broad, structured purposive mapping; the GLP-1 intersection was a bounded, plausibility-driven sample of the far larger incretin-neuroscience literature. Records were exported to a reference manager, deduplicated by digital object identifier and then by title, and screened by a single reviewer (initials withheld in the blinded version), with uncertain inclusions resolved with the senior authors. A representative PubMed/MEDLINE string and full term blocks are provided in the Supplementary Material. We make this asymmetry explicit so that the GLP-1 mapping is not mistaken for an exhaustive or systematically unbiased survey, and we note single-reviewer screening as a recognized limitation of this design.

Selection criteria: We included narrative and systematic reviews, meta-analyses, observational studies, clinical trials, and translational or experimental studies relevant to the PICS–neuroimmunometabolism–GLP-1 axis, and excluded studies without a clear mechanistic or conceptual link, duplicates, and sources whose bibliographic details could not be verified. For the GLP-1 tier, a record was eligible only if it reported a central-nervous-system-relevant endpoint—cognition or behavior, neuroinflammation or glial activation, blood-brain barrier or neurovascular measures, neurotrophic signaling, or direct target engagement—rather than a purely peripheral metabolic outcome.

Figure1: Conceptual model of the PICS Endotype Hypothesis.

Heterogeneous acute insults of critical illness, both disease-driven (sepsis, shock) and iatrogenic (mechanical ventilation, sedation, antibiotic and nutritional disruption), converge on a shared neuroimmunometabolic engine that is amplified by the gut and funneled through a disrupted blood-brain barrier. Beyond this barrier, biology diverges into eight partially overlapping candidate endotype axes (A–H), each defined by a dominant mechanism, a candidate biomarker signature, and a hypothesized, differential relationship to GLP-1 receptor agonism. Index etiology modifies which axis predominates, and arrow weight encodes the strength of supporting evidence (established, indirect/translational, or hypothesized). GLP-1 receptor agonism is annotated as investigational.

Handling of conflicting evidence.

Where evidence conflicted, we prioritized it in the order human randomized > human observational > animal > in vitro, and weighted demonstrated target engagement heavily, since a mechanism that cannot be shown to be engaged in vivo cannot be intercepted. We deliberately sought discordant and null findings—preclinical neuroprotection that has not replicated, mixed or null psychiatric outcomes, and neurodegenerative trials with absent or modest cognitive signals—because these countercurrents are integral to a falsifiable framework (see Section 9).

Levels of mechanistic support.

Throughout the text and tables, we label the evidence underlying each claim using a single mechanistic-support scheme, distinct from formal clinical certainty grading (e.g., GRADE): direct human PICS/ ICU evidence; indirect human evidence; adjacent-disease evidence; experimental evidence (animal or cell models); translational evidence; inferential/mechanistic reasoning; and requires verification. These labels describe the source and proximity of the evidence to the post-ICU context, not the certainty of an estimated treatment effect, and they keep the strength of each claim separable from the strength of the overall hypothesis. Because this article is hypothesis-generating, the evidence map should be interpreted as a structured plausibility framework rather than as an exhaustive or unbiased estimate of intervention effects.

PICS as a Biologically Heterogeneous Syn-drome

The case for stratification rests on three observations. First, PICS is phenotypically heterogeneous to a degree the unitary label conceals: cognitive, psychiatric, and physical trajectories diverge and are frequently uncoupled, and risk factors and one-year outcomes differ between the physical and cognitive components, implying distinct underlying processes [12–14]. Acute delirium, itself pathophysiologically heterogeneous, is an early marker of this divergence and a recognized antecedent of durable cognitive injury [15]; the neurocognitive burden borne by sepsis survivors has drawn particular attention [16]. Second, PICS overlaps substantially with post-sepsis syndrome, long COVID, and the neuropsychiatric sequelae of critical illness without clear biological boundaries, suggesting that clinically drawn lines do not correspond to mechanism [1-17]. Third, the field lacks validated, mechanism-anchored biomarkers and accepted molecular stratification, so interventions are necessarily general. Making the dissociation auditable The claim that PICS domains dissociate should be inspectable rather than asserted. Trajectory-based sub-phenotyping of sepsis survivors has identified functional-recovery classes that diverge over the first year and are not reducible to baseline severity [18]; secondary latent-class analysis of a large mixed-ICU cohort separated survivors into subtypes with different one-year outcomes after accounting for acute illness [19]; and a genomic structural-equation analysis points to partially distinct heritable architecture across PICS-associated traits [20]. These studies vary in instruments, follow-up windows, and adjustment for confounders, and we note this heterogeneity rather than present the dissociation as uniform. Competing, non-biological explanations for inconsistent post-ICU trials. Biological heterogeneity is not the only reason broad post-ICU interventions have produced inconsistent results, and presenting it as such would overstate the case. Variable intervention dose and fidelity, differences in timing relative to discharge, outcome-instrument variability and practice effects, and survivorship and attrition bias can each blur a real average effect [21–23]. The endotype hypothesis earns its place only if it predicts above and beyond these: if biomarker-defined clusters add incremental predictive value after adjustment for rehabilitation dose and socioeconomic variables, and if cluster membership is stable across centers. We state this as a condition the hypothesis must meet, not as a rhetorical concession. Critical care has confronted a comparable problem before. The recognition that acute respiratory distress syndrome (ARDS) encompasses hyper- and hypo-inflammatory subphenotypes, with different mortality and divergent treatment responses, has transformed how the syndrome is studied; comparable host-response phenotyping has been extended to sepsis survivorship and cardiogenic shock [24-25]. Sepsis-associated encephalopathy is now being parsed into mechanism-based subphenotypes, a direct precedent for the present argument [26]. Where a clinically uniform syndrome is underwritten by measurable, heterogeneous biology, stratification may be a prerequisite for progress. PICS satisfies both conditions but has not been subjected to the same discipline. ARDS as a methodological template, not a promise. We invoke the ARDS subphenotypes as a template for how clusters are discovered and validated, not as a forecast that PICS will yield equally clean groups; PICS has more heterogeneous baselines, environments, and outcome instruments, and the analogy should be read at the level of method rather than expected tidiness [19,25]. Framed this way, the hypothesis makes an explicit, falsifiable prediction: after adjustment for severity, frailty, and ICU exposures, multimodal biomarker-derived clusters will (i) replicate across independent cohorts, (ii) predict longitudinal cognitive, psychiatric, and physical trajectories, and (iii) show differential GLP-1 target engagement or treatment responses. Conversely, if the candidate clusters collapse into a severity or frailty gradient or fail to replicate externally, the hypothesis is refuted.

Neuroimmunometabolic Mechanisms after Critical illness

The biology linking an acute insult to chronic disability can be read as a sequence that converges on the central nervous system and then diverges. The index illness floods the circulation with pathogen- and damage-associated molecular patterns and a cytokine surge led by interleukin-6 (IL-6) and tumor necrosis factor-α (TNF-α); superimposed are the iatrogenic exposures of intensive care—mechanical ventilation, which can itself injure the brain; sedation and analgesia; immobilization; and antibiotic and nutritional disruption of the gut [27–29]. The intestine is an early amplifier: critical illness collapses microbial diversity and depletes short-chain fatty acid (SCFA)-producing taxa; loss of butyrate weakens the epithelial barrier, and a permeable gut translocates lipopolysaccharide into the circulation, where endotoxemia sustains inflammation [30–32]. Tryptophan metabolism is diverted along the kynurenine pathway [33]. Sustained peripheral inflammation acquires a metabolic character, reprogramming immune-cell bioenergetics and driving skeletal muscle into a catabolic, energy-depleted state signaled by rising myostatin, while the adipose-brain axis adds inflammatory tone [34–36]. The blood-brain barrier is pivotal: inflammatory mediators and microbial products breach endothelial tight junctions, and sepsis-associated encephalopathy is an acute clinical expression of this crossing; amyloid-β- and caspase-1-associated signatures suggest the breach lays molecular groundwork for chronic change [37–39]. Experimentally, sepsis aggravates Alzheimer-type pathology over time, a persistent neuroinflammatory state can be demonstrated in sepsis-surviving animals long after recovery [40-41], and dedicated rodent PICS models are beginning to expose candidate protective pathways [42] (experimental evidence). Beyond the barrier, microglia adopt a primed phenotype, astrocytes become reactive, and the NLRP3 inflammasome propagates a self-amplifying neuroinflammatory state [43,44]. Host factors—cognitive and mitochondrial reserve, baseline microbiome, age, sex, and genetic susceptibility—influence which process predominates [14]. The nature of the index insult is a further determinant of trajectory and bears on the external validity of any single-axis prediction. A massive sterile insult (major surgical trauma or extensive burns) drives a damage-associated molecular pattern surge and emergency myelopoiesis that tilt toward the catabolic and mixed (PIICS) axes, whereas prolonged sepsis and the protracted antigenic and endothelial provocation of severe COVID-19 favor sustained microglial and vascular involvement [4-17]. The same host may therefore reach a different dominant axis depending on how the critical illness began; index etiology should be treated as a stratifying covariate rather than averaged away. Temporal structure: which mechanisms are acute and which persist Because the therapeutic-window argument depends on persistence, the acute drivers of delirium and encephalopathy should be separated from the processes proposed to persist after discharge. In the acute ICU phase (days), a cytokine surge, acute barrier opening (sepsis-associated encephalopathy), and sedation- and ventilation-related insults predominate [45-46]. In the early post-discharge weeks, microglial priming, reactive astrogliosis, NLRP3-driven self-amplification, persistent endotoxemia, and endothelial and neurovascular dysfunction are hypothesized to drive the process forward, although much direct evidence at this stage currently comes from animal models [47-48]. In the late months, chronic low-grade neuroinflammation, aggravated neurodegenerative-type pathology, and a persistent catabolic or PIICS trajectory are candidate substrates for long-term impairment [40-49]. Human verification of persistence requires longitudinal measurement of the same markers from discharge across the first year; Figure 2 encodes this temporal axis alongside the distinction between established and hypothesized links.

Proposed Candidate Endotype Axes

We propose eight candidate mechanistic axes (Table 1), framed as putative endotype dimensions rather than validated, discrete classes. Real survivors are mixtures; the claim is only that one axis often predominates enough to shape trajectory and treatment response and is, in principle, identifiable. We do not claim eight discrete bins already exist: the more likely reality is a smaller number of latent classes superimposed on continuous mechanistic dimensions that co-occur, and some axes (for example, neuroglial, endothelial, and mitochondrial) may empirically merge. We therefore treat the eight as candidate axes to be discovered, agnostic about whether they reflect latent classes or a dimensional representation, with validation requiring unsupervised clustering or latent-class analysis and external replication. “Dominance” is operational: a survivor is assigned to an axis only when its posterior probability exceeds a pre-specified threshold; multi-label membership is permitted; and where no single axis dominates, we default to a “top-two axes” description. For each axis we state the core mechanism, probable phenotype, candidate biomarkers, plausible relationship to GLP-1 receptor agonism, and level of mechanistic support.

Figure 2: Neuroimmunometabolic mechanisms linking critical illness to cognitive, psychiatric, and physical impairment.

A mechanistic schematic tracing the sequence from index insult through the intestinal amplifier (dysbiosis, SCFA depletion, endotoxemia), systemic immunometabolic reprogramming and catabolic muscle-brain signaling, to blood-brain barrier breakdown, microglial priming, reactive astrogliosis, NLRP3 activation, and mitochondrial failure, culminating in cognitive, psychiatric, and physical impairment. Solid arrows denote established links (human or robust preclinical); dashed arrows denote hypothesized or mainly animal-model links. The schematic is organized into three temporal phases (ICU acute, early post-discharge, late months) to indicate which mechanisms are proposed to persist after discharge.

Hyperinflammatory-neuroglial axis

Unresolved central inflammation (primed microglia, reactive astrocytes, NLRP3 activation, elevated IL-6 and TNF-α) produces cognitive impairment that tracks inflammatory intensity [43-44]. The plausibility of an IL-6-anchored subgroup is reinforced by observations that IL-6 inhibition during critical illness is associated with altered mental-health outcomes in survivors [50] (indirect human evidence). GLP-1 receptor agonists suppress microglial activation, inflammasome assembly, and central Toll-like receptor-driven inflammation [51–53] (experimental evidence). Mechanistic support level: experimental + indirect human.

Metabolic, insulin-resistant axis:

Persistent insulin resistance and disturbed brain glucose handling may contribute to cognitive slowing via impaired central insulin signaling. GLP-1 receptor agonists act directly on this axis, and preclinical work links them to improved diabetes-associated cognition via metabolic and microbiome-mediated routes [54-55] (experimental/translational evidence). Mechanistic support level: experimental for the metabolic mechanism; indirect for PICS.

Catabolic-sarcopenic axis:

Bioenergetic failure, catabolism, and sarcopenia define a phenotype in which the physical domain dominates and recovers most reluctantly, marked by elevated myostatin and low metabolic reserve [34,-35] (direct human biomarker evidence). Here GLP-1 receptor agonism is double-edged: it may improve bioenergetics but cannot reverse muscle loss and may, through weight reduction, aggravate sarcopenia [56]. This is a design constraint, not merely an opportunity: in this axis, GLP-1 should be conceived only as one arm of an obligate combination paired with anabolic and anti-catabolic strategies (supervised high-intensity resistance training, nutritional optimization, targeted protein and amino-acid provision), so that any incretin-driven weight loss is preferentially adipose rather than lean mass.

Mechanistic support level: direct human (biology); safety concern requiring mandatory anabolic co-therapy.

Operationalizing lean-mass protection and stopping rules:

To make the obligate co-therapy principle actionable, lean-mass monitoring should precede any numeric threshold: body composition tracked with dual-energy X-ray absorptiometry where feasible, or opportunistic computed-tomography muscle analysis or bedside ultrasound where it is not, and function assessed with grip dynamometry and timed physical tests [21,34]. Pre-specified stopping triggers would include unacceptable lean-mass decline, persistent failure to meet protein and calorie targets, recurrent dehydration or acute kidney injury, and severe gastrointestinal intolerance. Describing the structure of these triggers, rather than committing prematurely to cut-offs, is what we mean by a safety-sentinel arm.

Vascular-endothelial axis

Endothelial injury and neurovascular dysfunction are proposed to sustain barrier leakage and impair cerebral perfusion. Human data link endothelial and neurologic injury markers during critical illness (including S100B and E-selec-tin) to long-term cognitive impairment and disability [57], a PICS review identifies endothelial dysfunction as a contrib-utor to acute brain injury in survivors [2], experimental sep-sis impairs neurovascular coupling before cognitive decline [48], and endothelial markers persist beyond resuscitation, with glycocalyx degradation and angiopoietin-2 as candidate readouts [58–60]. Liraglutide reduces vascular damage in mixed metabolic-neurodegenerative models [61] (experimental evidence). This axis is mechanistically plausible and translationally important but not yet validated as a discrete PICS subtype. Mechanistic support level: indirect human + experimental.

Gut-brain dysbiosis axis

In a dysbiotic gut with a failed barrier, the engine remains peripheral, with neurobehavioral symptoms proportional to endotoxemia and SCFA depletion [30-32]. Because the microbiome is the driver, GLP-1 monotherapy is likely necessary but insufficient, and durable benefit may require microbiome-directed co-therapy [54-55]. Mechanistic support level: experimental (strong preclinical microbiome-cognition link; human causality unproven).

What “microbiome-directed” should and should not mean in recent survivors

To prevent open-ended interpretation, we specify a stepped set ordered by feasibility and risk: dietary fiber and protein modulation first; then defined prebiotics, probiotics, or synbiotics; then postbiotics or SCFA provision; and only as a last, cautious option, fecal microbiota transplantation. In a recently critically ill, often immunologically perturbed population, recent antibiotic exposure, bacteremia risk, and immunosuppression must guide any escalation [23]. The lower-risk steps are appropriate for hypothesis testing; higher-risk options would require strong justification and dedicated safety oversight.

Neuropsychiatric stress axis:

Neuroendocrine stress and affective dysregulation predominate, manifesting as depression, anxiety, and post-traumatic stress [12]. GLP-1 receptor agonists show signals on mood, anxiety, and neurotrophic (BDNF) pathways across psychiatric and neurodegenerative contexts, although psychiatric safety, including vigilance for suicidality, must be monitored [62–65]. Mechanistic support level: indirect human (psychiatric); safety signals require attention.

Mitochondrial-bioenergetic axis:

Mitochondrial dysfunction and oxidative stress produce fatigue and cognitive slowing distinct from inflammation. GLP- 1 receptor agonists improve cellular bioenergetics and redox balance in experimental models [56,66]. Mechanistic sup-port level: experimental; indirect for PICS.

Mixed (PIICS) axis

This axis maps onto the well-characterized PIICS state of chronic critical illness: self-perpetuating inflammation, innate-immune reprogramming with emergency myelopoiesis, expansion of myeloid-derived suppressor cells, and protein catabolism [67–71]. The language of endotyping is already established in this literature [72]; PIICS is predictable from early SOFA, albumin, and lymphocyte data [73]; it occurs after trauma as well as sepsis [74]; and PIICS during ICU admission is directly associated with post-discharge physical dysfunction within the PICS spectrum, though not, at one month, with cognitive or mental components [49] (direct hu-man evidence). Its relationship with GLP-1 receptor agonism is the least defined of the eight. Mechanistic support level: direct human (physical dysfunction); inferential for cognitive/ neuropsychiatric components. Summary of the two least-supported axes. Among the proposed axes, the mixed PIICS axis has the strongest biological and clinical support, particularly in chronic critical illness and post-sepsis recovery. By contrast, the vascular-endothelial axis rests on converging but indirect evidence and should be presented as mechanistically plausible and translationally important, but not yet validated as a discrete PICS subtype.

GLP-1 receptor agonism beyond glycemic con-trol

Therapeutic interest in GLP-1 receptor agonists derives from pharmacology that has long outgrown glucose control [75]. These receptors are expressed centrally and peripherally, including in glia [43]. Central agonism dampens microglial activation and reactive astrogliosis, inhibits the NLRP3 inflammasome, and attenuates Toll-like-receptor-driven inflammation [44-52-76-77]; it improves mitochondrial bioenergetics and redox balance, supports neurotrophic signaling, modulates autophagy, and engages amyloid, tau, and α-synuclein pathways in neurodegenerative models [64-66-78-79-103] (experimental evidence). Effects on the vasculature, microbiome, and gut-brain axis broaden its reach [54-55-61], though an umbrella review of non-cardiometabolic outcomes under-scores the unevenness of the supporting evidence [80-102]. The appetite- and weight-lowering actions are mechanistically separable from related pathways such as GDF15 signal-ing—a distinction that matters when weighing anorexigenic risk in cachectic survivors [81] (experimental evidence). Clinically, the most mature neurological data come not from critical-illness survivorship but from neurodegenerative and psychiatric disease. Phase 3 programs have tested semaglutide in early Alzheimer’s disease, and trials in amyloid-positive adults are ongoing [82–84]; reviews report cognitive and affective signals across agents alongside real uncertainty [85-89], and tirzepatide and other dual incretin agonists widen the space further [90-91]. What matters for the present argument is the correspondence itself: the mechanisms these agents have been reported to engage in experimental and adjacent-disease settings track closely with those we propose as the substrate of the PICS axes (Table 2). This overlap is the foundation of the hypothesis and, in the same breath, the reason it demands testing rather than assumption.

Distinguishing target engagement from clinical benefit

The pivotal EVOKE and EVOKE+ phase 3 programs of oral semaglutide in early Alzheimer's disease [82-83] were designed specifically to test whether biomarker-level target engagement translates into clinically meaningful benefit. Until their primary outcome data are reported in full and independently verified, no clinical efficacy in neurodegeneration- let alone in PICS-should be assumed; reported biomarker signals do not establish clinical benefit. This caution reinforces the need to distinguish mechanistic target engagement from clinically meaningful neurorecovery and tempers any expectation that a positive PICS effect can be inferred from neurodegenerative-disease rationale. It is integral to, rather than at odds with, a framework framed as a hypothesis to be disproven as readily as confirmed.

Agent-level specificity and central target engagement:

Treating these effects as a uniform class property risks obscuring differences that may determine feasibility in survivors. Liraglutide, semaglutide, and exenatide differ in brain penetrance and half-life; oral and subcutaneous formulations differ in exposure; dual incretin agonists broaden the receptor profile; and biased agonism may dissociate metabolic from neural effects [56-90-91]. Central target engagement at clinically used doses is implied more often than demonstrated, and candidate pharmacodynamic readouts for PICS cohorts- cerebrospinal fluid drug levels, downstream signaling measures, neuroinflammation imaging-remain to be established [92-93]. The hypothesis is therefore most defensible for agents with the best central-exposure data; whether it is truly class-wide is itself an empirical question [94–96].

Biomarkers for endotype-guided stratification

Operationalizing the hypothesis requires endotypes that are measurable with existing assays, integrated with an analytic layer that does not yet exist (Table 3). The inflammatory tier comprises IL-6, C-reactive protein, TNF-α, and IL-1β. The neuroinflammatory and neuronal-injury tiers comprise GFAP, neurofilament light chain, and S100B, with phosphorylated tau and amyloid-β indices extending toward the neurodegenerative axis [39-57-97]. Metabolic markers index the insulin-resistant axis; catabolic markers (myostatin, albumin, prealbumin, grip strength, muscle mass) index sarcopenia [34-35]. Endothelial markers (E-selectin, angiopoietin-2, syndecan-1, thrombomodulin, von Willebrand factor) index the vascular axis [57-58-60], and PIICS-relevant markers include SOFA, day-1 albumin, and lymphocyte counts [73]. Mi-tochondrial/oxidative-stress and gut-brain markers (SCFAs, metagenomics, permeability) complete the biological panel, added to neuropsychological testing, patient-reported outcomes, and continuous digital biomarkers [98]. No single analyte assigns an endotype; the endotype is defined by a joint pattern of analytes. A pivotal, still-unproven element is computational: unsupervised latent-class and clustering methods—the family that revealed the ARDS subphenotypes—applied to multimodal survivor data to test whether the proposed axes emerge without being imposed, followed by supervised models returning a probabilistic assignment and predicted treatment response [24]. Success is not any clustering, which is trivial, but a few stable clusters whose membership predicts trajectory.

Causal nodes versus epiphenomena: the pivotal caveat

A deeper caveat conditions the entire enterprise. Identifying a marker that correlates with an endotype is not the same as identifying a driver that GLP-1 can intercept. Analytes such as IL-6, S100B, and neurofilament light chain may be causal nodes whose interruption changes trajectory, or epiphenomena—passive readouts of injury that rise and fall with damage without participating in its propagation. A therapy aimed at a causal node can modify disease; the same therapy aimed at a bystander marker will move the number without moving the patient. Establishing causal interceptability is therefore a primary objective, demanding Mendelian-randomization, mediation, and target-engagement designs in which a candidate marker is shown to lie on the pathway between mechanism and outcome and to shift in step with clinical benefit under GLP-1 exposure. Until then, the proposed panels are hypotheses about which biology is actionable, not validated targets.

A feasible, tiered biomarker plan

Multi-site cohorts are constrained by assay availability, pre-analytics, batch and center effects, and informative dropout, so the panel is best organized into tiers: a core tier of widely available assays (inflammatory and metabolic markers, neurofilament light chain and GFAP, grip strength, basic body composition) feasible in most cohorts, and a modular extended tier (metagenomics, permeability testing, advanced imaging). A common sampling template—early ICU predictors for the PIICS axis, then discharge and 1, 3, 6, and 12 months—would support discovery and longitudinal modeling. [99-100]. Quality control, harmonization, and explicit handling of missingness from death or rehospitalization (inverse-probability weighting, joint modeling, sensitivity analyses) are prerequisites for trusting that any cluster reflects biology rather than artifact [23]. Pivotal nodes warrant concrete readouts: barrier integrity probed with permeability imaging (e.g., dynamic contrast-enhanced MRI) or the cerebrospinal-fluid-to-serum albumin ratio in selected cohorts; glial activation tracked with soluble TREM2 and GFAP, with translocator-protein PET reserved for exploratory substudies [46-57]. External validation across health systems, stratified by index etiology, is the decisive test of whether any endo-type is real or merely local [19-101].

Mechanistic overlap between PICS biology and GLP-1 receptor agonist effects

The preceding axes can be read against the documented pharmacology of GLP-1 receptor agonists, and this corre-spondence is the central evidentiary rationale for the review (Table 2). Each major node of PICS biology has a reported pharmacological counterpart: microglial and NLRP3-driven neuroinflammation against suppression of microglial acti-vation and inflammasome assembly; reactive astrogliosis against attenuated astrocyte reactivity; systemic Toll-like-re-ceptor-driven inflammation against central anti-inflammatory signaling; mitochondrial and oxidative injury against improved bioenergetics; insulin resistance against direct metabolic ac-tion; dysbiosis against microbiome remodeling; neurotrophic deficit against BDNF support; amyloid, tau, and synaptic in-jury against target engagement in neurodegenerative mod-els; and endothelial injury against reduced vascular damage. Critically, this convergence is drawn overwhelmingly from ex-perimental and adjacent-disease evidence rather than from PICS itself; it is a reason to test these agents in stratified survivors—not evidence that they work in PICS.

Discussion

Central contribution and mechanistic rationale If PICS is a set of endotype axes and GLP-1 receptor agonists act on the mechanisms that define them, the question is not whether these agents work in PICS but in whom, when, and toward what end. Timing matters: crossing the blood-brain barrier may mark a point beyond which the central injury machinery becomes self-sustaining, so barrier-stabilizing and microglia-repolarizing interventions are plausibly more effective early, and the neurodegenerative trajectory may have a finite window [37-40]. Endotype selection matters: the inflammatory and neurodegenerative axes offer the strongest mechanistic matches; the gut-brain axis likely requires combination therapy; and the catabolic-sarcopenic axis carries a real risk of harm through weight loss, so uniform application would average benefit against harm and obscure both [56]. Finally, prevention, recovery, and disease modification are distinct aims with distinct designs, and conflating them has complicated the wider GLP-1 neurology literature. Precision here requires both restraint and ambition. The framework delineates where GLP-1 receptor agonism may be beneficial, where combination therapy is warranted, and where it should be avoided, positioning these agents within a comprehensive recovery strategy—nutrition, resistance exercise, sleep, rehabilitation—rather than as a sole intervention [30]. These are not established clinical practices; the rationale is mechanistic and indirect, outlining a research agenda rather than a prescription.

Discordant evidence and uncertainty:

A framework built to be falsified must foreground the evidence that could refute it. Several countercurrents bear directly on the hypothesis. First, GLP-1 and cognition: neurodegenerative and psychiatric trials report mixed or null cognitive signals across agents, and the EVOKE/EVOKE+ programs in early Alzheimer’s disease are testing whether biomarker-level engagement yields clinical benefit—an open question that, on current evidence, illustrates that target engagement need not translate into clinical benefit. Second, central penetration and engagement at clinically tolerated doses remain unproven and may be inconsistent across an already disrupted, heterogeneous blood-brain barrier. Third, the bulk of supporting data is extrapolated from animal models and neurodegenerative disease rather than from PICS, and such extrapolation has repeatedly failed in critical care. Two possibilities would specifically undermine the precision claim. A generic neuroprotective effect, if present, could prove uniform across axes, refuting endotype-specific precision while leaving a blunt, non-stratified benefit; conversely, candidate biomarkers may be epiphenomena rather than causal targets, so that intercepting them changes the number without changing the patient. There is also the possibility of no clinically meaningful benefit despite encouraging mechanistic signals. Finally, in some survivors these agents may be positively hazardous—the severely catabolic or cachectic patient, the gastroparetic or high-aspiration-risk survivor, the profoundly dysautonomic patient, and those with active gastrointestinal intolerance or unstable nutritional status. We regard each of these as a way the hypothesis can fail, and we have designed the program (Section 9.6) so that it can fail informatively.

Translational implications: biomarkers, deep cohorts, and stratified trials.

As a conceptual model and explicitly not a clinical recommendation, we sketch an endotype-stratified adaptive platform trial. The population would be adult ICU survivors with persistent PICS-domain impairment at a defined post-discharge interval, phenotyped and allocated to endotype strata before randomization. Exclusions would include established GLP-1 contraindications, severe malnutrition, and high aspiration risk. Within strata, inflammatory and neurodegenerative arms would compare early GLP-1 receptor agonism with placebo (cognitive trajectory primary; earlier-versus-later initiation secondary); the gut-brain arm would compare GLP-1 monotherapy with GLP-1 plus microbiome-directed therapy (SCFA recovery as a pharmacodynamic mediator); and the catabolic-sarcopenic arm would function as a safety sentinel, testing titrated GLP-1 with obligatory anabolic and nutritional support against rehabilitation alone (lean body mass and physical function co-primary). The adaptive architecture would refine strata and response signatures as evidence accrues and would be built to refute the precision claim as readily as to support it. Figure 3 summarizes the workflow. To keep this sketch tentative yet judgeable, any realization should pre-specify a minimum set of parameters: persistent impairment defined with named instruments and minimal clinically important differences across domains; “early” versus “late” initiation anchored to a defined post-discharge interval; the adaptive feature stated explicitly (e.g., response-adaptive randomization, dropping a non-responding arm, or pre-planned sample-size re-estimation); and endotype uncertainty handled probabilistically rather than by hard stratification. Core-outcome-set work in post-ICU nutrition and metabolism offers a starting point for instrument selection [23].

These are parameters to be pre-specified, not a finished protocol. Risks, safety, and ethical considerations:

The pharmacology that makes GLP-1 receptor agonists attractive also makes them hazardous in this fragile population. Gastrointestinal effects are common, and delayed gastric emptying raises aspiration concerns relevant to perioperative and readmitted survivors [80]; the same delay produces drug-drug interactions in the heavily co-medicated survivor. The most consequential hazard is to nutritional status and muscle: weight loss is therapeutic in obesity but dangerous in the catabolic, sarcopenic, or cachectic survivor, in whom it could deepen frailty [35-56]. For this reason, in the catabolic-sarcopenic axis GLP-1 receptor agonism should not be pursued as monotherapy outside carefully monitored research protocols, and would be difficult to justify without obligatory pairing with anabolic and anti-catabolic strategies that actively protect lean mass. Additional risks warrant explicit attention: dehydration and acute kidney injury in the setting of volume depletion; acute pancreatitis and gallbladder disease (cholelithiasis and biliary disease); accelerated diabetic retinopathy when glycemia falls rapidly; and the rodent thyroid C-cell signal of uncertain human relevance. Older, frail, and renally impaired survivors require particular caution, and polypharmacy is the rule after critical illness. Psychiatric safety, including vigilance for suicidality, must be monitored even as aggregate evidence remains under study [65].

Figure 3: Precision GLP-1 neurorecovery framework.

A translational workflow: multimodal biomarker acquisition feeds an unsupervised computational classifier that assigns a probabilistic endotype; assignment then routes survivors to mechanism-matched, endotype-stratified investigational arms (for example, early GLP-1 in inflammatory/neurodegenerative axes; GLP-1 plus microbiome-directed therapy in the gut-brain axis; and, in the catabolic-sarcopenic axis, GLP-1 only with obligatory anabolic safeguards, functioning as a safety sentinel) within an adaptive platform trial. The schematic foregrounds falsifiability—including an explicit gate at which the hypothesis can be refuted before any patient is exposed—and labels GLP-1 receptor agonism as investigational rather than established for PICS. Independent of endotype, and consistent with regulatory labeling, a personal or family history of medullary thyroid carcinoma or multiple endocrine neoplasia type 2 should be treated as a contraindication to GLP-1 receptor agonism—one of the few absolute exclusions the stratified logic does not override. Beyond individual safety lie equity and access, cost-effectiveness, and the risk of medicalizing survivorship. Given the cost and supply constraints of these agents and the real risk of premature off-label use, responsible development requires stewardship measures—registries, shared protocols, and data sharing—and explicit attention to cost and disparities so that a precision framework does not become a vector of inequity. The same endotype logic should guide non-pharmacological recovery (sleep optimization, graded rehabilitation, nutritional periodization, microbiome-aware care), with GLP-1 receptor agonism, if it has any place at all, positioned within that broader strategy rather than above it [21-23]. Shared decision-making and rigorous oversight are constitutive of any responsible use of these agents in critical-illness survivors, not optional add-ons.

Limitations

This was a SANRA-informed purposive narrative synthesis, not a systematic review; relevant studies may have been missed, single-reviewer screening was used, and no quantitative or formal bias appraisal was conducted. The central claim is integrative and therefore inferential; although each component mechanism is individually supported, their assembly into discrete GLP-1-responsive endotypes is a hypothesis, not a finding. The evidence base is uneven—weighted toward neurodegenerative and psychiatric models and animal and adjacent-disease data, with no direct GLP-1 trial in PICS—and the vascular-endothelial axis in particular rests on indirect support.

A further limitation concerns the agent rather than the argument: persistent neuroinflammation of established disease may be self-sustaining and no longer incretin-responsive once the therapeutic window has closed. These limitations do not undermine the framework but define the conditions under which it must be tested. We therefore frame GLP-1 receptor agonism as a falsifiable investigational hypothesis to be disproved as readily as confirmed, not as a therapy in search of justification.

Future research agenda:

We envisage a phased program. Phase 1 establishes deep-phenotyping cohorts with serial multimodal sampling from discharge across the first year [98-104]. Phase 2 covers biomarker discovery and validation against trajectories. Phase 3 uses endotype-informed observational studies to test whether cluster membership forecasts outcomes [49]. Phase 4 conducts small proof-of-concept GLP-1 studies in the best-matched axes with pharmacodynamic endpoints. Phase 5 conducts randomized trials in biomarker-enriched populations. Phase 6 integrates findings with rehabilitation and implementation science. Each phase can fail informatively: if survivor data do not resolve into reproducible, outcome-predictive clusters, the hypothesis is refuted in Phase 1 before any patient is exposed to an intervention.

Conclusion

PICS likely represents not a single condition but a hetero-geneous set of biological recovery trajectories. The eight neuroimmunometabolic axes proposed here are testable hy-potheses, not validated classes, and may collapse into fewer dimensions or prove continuous rather than discrete. GLP-1 receptor agonism is best understood as a heuristic platform for testing mechanism in selected survivors—not an estab-lished treatment for PICS. Validation will require validated biomarkers, longitudinal deeply phenotyped cohorts, demon-strated central target engagement, and endotype-stratified trials, and any future use must be designed to protect against harm, especially in catabolic and sarcopenic survivors. The framework is built so that either answer, affirmative or nega-tive, advances the care of critically ill survivors.

Tables

The eight axes below are putative endotype dimensions to be tested empirically, not validated discrete classes; they may merge, coexist, or prove continuous. “Mechanistic support level” denotes the source and proximity of supporting evidence, not formal clinical-certainty grading.

Ang-2, angiopoietin-2; BDNF, brain-derived neurotrophic factor; CRP, C-reactive protein; HPA, hypothalamic-pituitary-adrenal; PIICS, persistent inflammation-immunosuppression-catabolism syndrome; PROs, patient-reported outcomes; PTSD, post-traumatic stress disorder; SCFAs, short-chain fatty acids; SOFA, Sequential Organ Failure Assessment.

Table 1: Proposed PICS neuroimmunometabolic candidate endotype axes: clinical features, biomarkers, and GLP-1 relevance.

This mapping is a hypothesis, not therapeutic proof: the GLP-1 actions below are drawn largely from experimental and non-PICS evidence, and no entry denotes demonstrated efficacy in PICS.

Evidence labels follow the mechanistic-support scheme defined in the Methods. Parenthetical qualifiers denote evidence setting, not a separate tier: “metabolic setting” marks indirect human evidence from metabolic disease; “review-level” marks a synthesis whose primary evidence is experimental or observational; “non-PICS” marks evidence from outside the post-intensive-care context. Translation and extrapolation risk are qualitative judgments reflecting the distance between cited evidence and the post-ICU context. BDNF, brain-derived neurotrophic factor; TLR, Toll-like receptor.

Table 2: Mechanistic overlap between PICS biology and GLP-1 receptor agonist effects.

CRP, C-reactive protein; CSF, cerebrospinal fluid; GFAP, glial fibrillary acidic protein; HbA1c, glycated hemoglobin; IL, interleukin; PET, positron-emission tomography; PIICS, persistent inflammation-immunosuppression-catabolism syndrome; SCFAs, short-chain fatty acids; SOFA, Sequential Organ Failure Assessment; TNF-α, tumor necrosis factor-α. Endotype letters are defined in Table 1.

Table 3: Candidate biomarker panel for endotype-guided PICS research.

Boxes

Box 1.

Key definitions for non-specialist readers:

PICS (post-intensive care syndrome): new or worsening cognitive, psychiatric, and physical impairment that persists after critical illness.

PIICS (persistent inflammation, immunosuppression, and catabolism syndrome): a chronic-critical-illness state of simultaneous ongoing inflammation, immune suppression, and muscle wasting.

Endotype: a patient subgroup defined by a distinct biological mechanism rather than by shared outward features.

Phenotype: an observable clinical profile, which may arise from more than one underlying mechanism.

Neuroimmunometabolism: the interacting neural, immune, and metabolic signaling that links systemic illness to the brain.

GLP-1 receptor agonists: drugs that activate the glucagon-like peptide-1 receptor (e.g., semaglutide, liraglutide), with metabolic and candidate neural effects.

Target engagement: direct evidence that a drug reaches and acts on its intended molecular target in living tissue.

Latent class: a statistically inferred subgroup, not directly observed, derived from patterns across multiple measurements.

Precision neurorecovery: matching recovery-oriented interventions to a survivor’s dominant biological mechanism rather than applying them uniformly.

Box 4. Clinical implications

No current indication: GLP-1 receptor agonists should not be used to treat PICS outside trials. If validated, stratification would target therapy and identify survivors to spare. Routine ICU practices affecting the gut may be modifiable contributors. Any future use demands shared decision-making and nutritional safeguards.

Declarations

Acknowledgements None.

Funding

This research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors.

Data Availability Statement

This is a narrative review; no new datasets were generated or analyzed. All findings discussed derive from the publications listed in the reference list. A representative search string and full term blocks are provided in the Supplementary Material.

Ethics Statement:

Not applicable. This work did not involve human partici-pants, identifiable human data, or animal experimentation, and therefore did not require ethics committee approval or informed consent.

Use of Artificial Intelligence

A generative AI assistant was used to support English-language editing and the drafting of schematic figure concepts. The tool is not an author and does not meet ICMJE authorship criteria; it did not generate primary data, perform analyses, or draw scientific conclusions. All AI-assisted content, including the figures, was reviewed, verified, and revised by the authors, who take full responsibility for the integrity, accuracy, and originality of the work.

Conflicts of Interest: The authors declare no conflicts of interest regarding the publication of this paper.

References

1. Fleischmann-Struzek, Carolin, Franka EA Joost, Mathias W. Pletz, Björn Weiß, Nicolas Paul, E. Wesley Ely, Konrad Re-inhart, and Norman Rose. "How are Long-Covid, Post-Sep-sis-Syndrome and Post-Intensive-Care-Syndrome related? A conceptional approach based on the current research litera-ture." Critical Care 28, no. 1 (2024): 283.
   View at Publisher | View at Google Scholar
2. Gupta, Lovish, Maazen Naduthra Subair, Jaskaran Munjal, Bh-upinder Singh, Vasu Bansal, Vasu Gupta, and Rohit Jain. "Be-yond survival: understanding post-intensive care syndrome." Acute and critical care 39, no. 2 (2024): 226-233.
   View at Publisher | View at Google Scholar
3. Zhang, Yu, Xiao-yu Yin, Meng-qing Zhang, Jian-jun Yang, Mu-huo Ji, and Jin-chun Shen. "Incidence and risk factors of post-intensive care syndrome (PICS) in sepsis survivors: a prospective, observational study." BMC anesthesiology 26, no. 1 (2026): 42.
   View at Publisher | View at Google Scholar
4. Berger, Mette M., Kristine WAC Koekkoek, Lara Montoro, and Olivier Pantet. "What does “PICS” mean in major burns? Per-sistent critical illness or post-intensive care syndrome." Burns 52, no. 5 (2026): 107991.
   View at Publisher | View at Google Scholar
5. Rousseau, Anne-Françoise, Pauline Minguet, Camille Colson, Isabelle Kellens, Sourour Chaabane, Pierre Delanaye, Eti-enne Cavalier, J. Geoffrey Chase, Bernard Lambermont, and Benoit Misset. "Post-intensive care syndrome after a critical COVID-19: cohort study from a Belgian follow-up clinic." An-nals of intensive care 11, no. 1 (2021): 118.
   View at Publisher | View at Google Scholar
6. Oshima, Taku, and Junji Hatakeyama. "Nutritional therapy for the prevention of post-intensive care syndrome." Journal of in-tensive care 12, no. 1 (2024): 29.
   View at Publisher | View at Google Scholar
7. Walsh, Timothy S., Lydia Emerson, Jo Singleton, Rachel Locherty, David Hope, Stephanie Cholbi, Annabel Giddings et al. "Anaemia management with red blood cell transfusion to improve post-intensive care disability: Protocol for the ABC post-ICU randomised controlled trial." Journal of the Intensive Care Society 27, no. 1 (2026): 98-106.
   View at Publisher | View at Google Scholar
8. Gawlytta, Romina, Miriam Kesselmeier, Andre Scherag, Hel-en Niemeyer, Maria Böttche, Christine Knaevelsrud, and Jen-ny Rosendahl. "Internet-based cognitive-behavioural writing therapy for reducing post-traumatic stress after severe sep-sis in patients and their spouses (REPAIR): results of a ran-domised-controlled trial." BMJ open 12, no. 3 (2022): e050305.
   View at Publisher | View at Google Scholar
9. Rousseau, Anne-Françoise, Gabriel Thierry, Bernard Lamber-mont, Vincent Bonhomme, and Joana Berger-Estilita. "Preha-bilitation to mitigate postintensive care syndrome in surgical patients: The rationale for a peri-critical illness pathway involv-ing anaesthesiologists and intensive care physicians." Europe-an Journal of Anaesthesiology| EJA 42, no. 5 (2025): 419-429.
   View at Publisher | View at Google Scholar
10. Mart, Matthew F., Leanne M. Boehm, Amy L. Kiehl, Michelle N. Gong, Atul Malhotra, Robert L. Owens, Babar A. Khan et al. "Long-term outcomes after treatment of delirium during critical illness with antipsychotics (MIND-USA): a randomised, place-bo-controlled, phase 3 trial." The Lancet Respiratory Medicine 12, no. 8 (2024): 599-607.
    View at Publisher | View at Google Scholar
11. Chadda, Karan R., and Zudin Puthucheary. "Persistent in-flammation, immunosuppression, and catabolism syndrome (PICS): a review of definitions, potential therapies, and re-search priorities." British Journal of Anaesthesia 132, no. 3 (2024): 507-518.
    View at Publisher | View at Google Scholar
12. Schmidt, Konrad FR, Jochen S. Gensichen, Maya Schroevers, Martina Kaufmann, Friederike Mueller, Gustav Schelling, Sa-bine Gehrke-Beck et al. "Trajectories of post-traumatic stress in sepsis survivors two years after ICU discharge: a secondary analysis of a randomized controlled trial." Critical Care 28, no. 1 (2024): 35.
    View at Publisher | View at Google Scholar
13. Miyazaki, Michiteru, Hirokuni Tagaya, Kazumasa Miida, Hide-nori Kariya, Yuichi Kataoka, and Masayasu Arai. "Association between risk factors and 1-year mortality of physical and/or cog-nitive components of post-intensive care syndrome in patients with sepsis." Medicine 105, no. 16 (2026): e48425-e48425.
    View at Publisher | View at Google Scholar
14. Costas-Carrera, Ana, Marta Mónica Sánchez-Rodríguez, Silvia Cañizares, Antonio Ojeda, Inés Martín-Villalba, Mireia Primé-Tous, Manuel Arturo Rodríguez-Rey et al. "Neuropsychological functioning in post-ICU patients after severe COVID-19 infec-tion: the role of cognitive reserve." Brain, Behavior, & Immuni-ty-Health 21 (2022): 100425.
    View at Publisher | View at Google Scholar
15. Stollings, Joanna L., Katarzyna Kotfis, Gerald Chanques, Brenda T. Pun, Pratik P. Pandharipande, and E. Wesley Ely. "Delirium in critical illness: clinical manifestations, outcomes, and management." Intensive care medicine 47, no. 10 (2021): 1089-1103.
    View at Publisher | View at Google Scholar
16. Salahuddin, Zainab. "Confronting the Neurocognitive Toll of Sepsis Survivors." Juntendo Medical Journal 71, no. 6 (2025): 411-413.
    View at Publisher | View at Google Scholar
17. Rousseau, Anne-Françoise, Camille Colson, Pauline Minguet, Isabelle Kellens, Maxine Collard, Camille Vancraybex, Julien Guiot, Bernard Lambermont, and Benoit Misset. "Character-istics of mid-term post-intensive care syndrome in patients at-tending a follow-up clinic: A prospective comparison between COVID-19 and non-COVID-19 survivors." Critical care explora-tions 5, no. 1 (2023): e0850.
    View at Publisher | View at Google Scholar
18. Puthucheary, Zudin A., Jochen S. Gensichen, Aylin S. Caki-roglu, Richard Cashmore, Lara Edbrooke, Christoph Heintze, Konrad Neumann, Tobias Wollersheim, Linda Denehy, and Konrad FR Schmidt. "Implications for post critical illness trial design: sub-phenotyping trajectories of functional recovery among sepsis survivors." Critical Care 24, no. 1 (2020): 577.
    View at Publisher | View at Google Scholar
19. Soussi, Sabri, Divya Sharma, Peter Jüni, Gerald Lebovic, Laurent Brochard, John C. Marshall, Patrick R. Lawler et al. "Identifying clinical subtypes in sepsis-survivors with different one-year outcomes: a secondary latent class analysis of the FROG-ICU cohort." Critical Care 26, no. 1 (2022): 114.
    View at Publisher | View at Google Scholar
20. Lv, Quankun, Guoxin Wu, Zihao Huang, Jiaxian Huo, Xinming Huang, Bin Yang, Yi Ye et al. "A genomic structural equation modelling study elucidates shared genetic architecture of poly-genic traits associated with post-intensive care syndrome." Scientific Reports (2026).
    View at Publisher | View at Google Scholar
21. Nakanishi, Nobuto. "Intensive care unit-acquired muscle at-rophy and weakness in critical illness: a review of long-term recovery strategies." Acute and critical care 40, no. 3 (2025): 361-372.
    View at Publisher | View at Google Scholar
22. Paulus, Michelle Carmen, Imre Willemijn Kehinde Kouw, Yente Florine Niké Boelens, Anoek Jacqueline Hubertine Hermans, Bert Strookappe, and Arthur Raymond Hubert van Zanten. "Feasibility challenges in protein supplementation research: Insights from the convalescence of functional outcomes after intensive care unit stay in a Randomised Controlled Trial." Clin-ical Nutrition 46 (2025): 119-130.
    View at Publisher | View at Google Scholar
23. Blaser, Annika Reintam, Thomas Davies, and Kaspar F. Bach-mann. "Core outcome sets and trials of nutrition and metabo-lism interventions." Current Opinion in Clinical Nutrition & Met-abolic Care 29, no. 2 (2026): 159-168.
    View at Publisher | View at Google Scholar
24. Slim, Marleen A., Rombout BE van Amstel, Lieuwe DJ Bos, Olaf L. Cremer, W. Joost Wiersinga, Tom van der Poll, and Lon-neke A. van Vught. "Inflammatory subphenotypes previously identified in ARDS are associated with mortality at intensive care unit discharge: a secondary analysis of a prospective ob-servational study." Critical Care 28, no. 1 (2024): 151.
    View at Publisher | View at Google Scholar
25. Soussi, Sabri, Mojtaba Ahmadiankalati, Jacob C. Jentzer, John C. Marshall, Patrick R. Lawler, Margaret Herridge, Alexandre Mebazaa et al. "Clinical phenotypes of cardiogenic shock sur-vivors: insights into late host responses and long-term out-comes." ESC Heart Failure 11, no. 2 (2024): 1242-1248.
    View at Publisher | View at Google Scholar
26. Song, Qin, and Bo Song. "Decoding sepsis-associated enceph-alopathy: From blood–brain barrier injury to mechanism-based subphenotypes." Brain Research Bulletin (2025): 111660.
    View at Publisher | View at Google Scholar
27. Van Hollebeke, Marine, Zhonghua Shi, and Martin Dres. "Ven-tilator-induced brain injury: another iatrogenic complication of mechanical ventilation." Current Opinion in Critical Care (2026): 10-1097.
    View at Publisher | View at Google Scholar
28. Turner, Ashley D., Travis Sullivan, Kurt Drury, Trevor A. Hall, Cydni N. Williams, Kristin P. Guilliams, Sarah Murphy, and Iqbal O’Meara AM. "Cognitive dysfunction after analgesia and sedation: out of the operating room and into the pediatric inten-sive care unit." Frontiers in Behavioral Neuroscience (2021).
    View at Publisher | View at Google Scholar
29. Shimoyama, Yuichiro, Noriko Kadono, and Osamu Umegaki. "Association of the National Database Code “General Anesthe-sia” and Mechanical Ventilation with New-Onset Neuro-Psychi-atric Medication Use in Sepsis Survivors Aged≥ 90 Years: A Nationwide Retrospective Cohort Study." International Journal of General Medicine (2026): 609863.
    View at Publisher | View at Google Scholar
30. Moțățăianu, Anca, Georgiana Șerban, and Sebastian Andone. "The role of short-chain fatty acids in microbiota–gut–brain cross-talk with a focus on amyotrophic lateral sclerosis: a sys-tematic review." International journal of molecular sciences 24, no. 20 (2023): 15094.
    View at Publisher | View at Google Scholar
31. Zhang, Qiulei, Chang Lu, Weixuan Fan, Jingxiao Zhang, and Yongjie Yin. "Application background and mechanism of short-chain fatty acids in sepsis-associated encephalopathy." Fron-tiers in Cellular and Infection Microbiology 13 (2023): 1137161.
    View at Publisher | View at Google Scholar
32. Zhao, Lianwu, Yan Huang, Zhiwen Ye, Wei Chen, Nuobei Zhang, Zhili Wen, and Chenglong Ge. "Short-chain fatty acids attenuate sepsis-induced gut dysbiosis and hippocampal neu-roinflammation via NLRP6 inflammasome activation in mice." International Journal of Surgery 112, no. 1 (2026): 570-582.
    View at Publisher | View at Google Scholar
33. Kanova, Marcela, and Pavel Kohout. "Tryptophan: a unique role in the critically ill." International journal of molecular sci-ences 22, no. 21 (2021): 11714.
    View at Publisher | View at Google Scholar
34. Shirahata, Ayaki, Nobuto Nakanishi, Yuko Ono, Shigeaki Inoue, and Joji Kotani. "Utility of Plasma Myostatin as a Predictive Biomarker for Post Intensive Care Syndrome in Patients With Sepsis." Acute Medicine & Surgery 12, no. 1 (2025): e70098.
    View at Publisher | View at Google Scholar
35. Mellen, Rodrigo Haber, Otavio Simoes Girotto, Eduarda Boni Marques, Lucas Fornari Laurindo, Paulo Cesar Grippa, Clau-demir Gregorio Mendes, Lorena Natalino Haber Garcia et al. "Insights into pathogenesis, nutritional and drug approach in sarcopenia: a systematic review." Biomedicines 11, no. 1 (2023): 136.
    View at Publisher | View at Google Scholar
36. Mateus-Gomes, Saudina, Amnah Al-Sayyar, Baptiste Lobey, Agnes Nadjar, Rejane Rua, and Rasheed Ahmad. "Metabolic inflammation at the adipose-brain axis." Frontiers in Physiology 17 (2026): 1795058.
    View at Publisher | View at Google Scholar
37. Gao, Qingzeng, and Marina Sorrentino Hernandes. "Sepsis-as-sociated encephalopathy and blood-brain barrier dysfunction." Inflammation 44, no. 6 (2021): 2143-2150.
    View at Publisher | View at Google Scholar
38. Li, Jin, Qi Jia, Lin Yang, You Wu, Yuliang Peng, Lixia Du, Zong-ping Fang, and Xijing Zhang. "Sepsis-associated encephalop-athy: mechanisms, diagnosis, and treatments update." Inter-national Journal of Biological Sciences 21, no. 7 (2025): 3214.
    View at Publisher | View at Google Scholar
39. Tuckey, Amanda N., Arcole Brandon, Yasaman Eslaamizaad, Waqar Siddiqui, Talha Nawaz, Christopher Clarke, Erica Sutherland et al. "Amyloid-β and caspase-1 are indicators of sepsis and organ injury." ERJ open research 10, no. 1 (2024).
    View at Publisher | View at Google Scholar
40. Vo, Quan, Dina C. Nacionales, Karen N. McFarland, Carmelina Gorski, Evan L. Barrios, Gwoncheol Park, Lyle L. Moldawer et al. "Temporal impact of sepsis on Alzheimer's disease pa-thology and neuroinflammation." Progress in neurobiology 250 (2025): 102775.
    View at Publisher | View at Google Scholar
41. Alves, Vinícius Santos, Joyce Pereira da Silva, Fabiana Cris-tina Rodrigues, Suzana Maria Bernardino Araújo, André Luiz Gouvêa, Raíssa Leite-Aguiar, Stephanie Alexia Cristina Silva Santos et al. "P2X7 receptor contributes to long-term neuroin-flammation and cognitive impairment in sepsis-surviving mice." Frontiers in Pharmacology 14 (2023): 1179723.
    View at Publisher | View at Google Scholar
42. Imai, Yumiko, Yasuha Kinugasa, Ryota Nukiwa, Mara Anais Llamas Covarrubias, Khin Kyaemon Thwin, Kumiko Yonezaki, Takashi Shimizu et al. "Protective Role of Apelin in a Mouse Model of Post–Intensive Care Syndrome." American Journal of Respiratory Cell and Molecular Biology 74, no. 2 (2026): 241- 256.
    View at Publisher | View at Google Scholar
43. Cui, Qi N., Lauren M. Stein, Samantha M. Fortin, and Matthew R. Hayes. "The role of glia in the physiology and pharmacology of glucagon-like peptide-1: implications for obesity, diabetes, neurodegeneration and glaucoma." British Journal of Pharma-cology 179, no. 4 (2022): 715-726.
    View at Publisher | View at Google Scholar
44. Wang, Lei, Jiangwei Ding, Changliang Zhu, Baorui Guo, Wu Yang, Wenxin He, Xinxiao Li et al. "Semaglutide attenuates sei-zure severity and ameliorates cognitive dysfunction by block-ing the NLR family pyrin domain containing 3 inflammasome in pentylenetetrazole-kindled mice." International journal of mo-lecular medicine 48, no. 6 (2021): 219.
    View at Publisher | View at Google Scholar
45. Fan, Tracey H., Lavienraj Premraj, Jacob Roberts, Melissa Lydston, Chiara Robba, David Hager, Jose I. Suarez, Denise Battaglini, and Sung-Min Cho. "In-hospital neurologic compli-cations, neuromonitoring, and long-term neurologic outcomes in patients with sepsis: A systematic review and meta-analy-sis." Critical care medicine 52, no. 3 (2024): 452-463.
    View at Publisher | View at Google Scholar
46. Tang, Fajuan, ShanShan Wu, Zhuan Zou, Xihong Li, and Lina Qiao. "The core role of central nervous system in sepsis-related organ damage." Frontiers in Immunology 16 (2025): 1694003.
    View at Publisher | View at Google Scholar
47. Dal-Pizzol, Felipe, Monique Michels, Mazeraud Aurélien, Fer-nando Augusto Bozza, Cristiane Ritter, and Sharshar Tarek. "What animal models can tell us about long-term psychiatric symptoms in sepsis survivors: a systematic review." Neu-rotherapeutics 18, no. 2 (2021): 1393-1413.
    View at Publisher | View at Google Scholar
48. Csipo, Tamas, Benjamin R. Cassidy, Priya Balasubramanian, Douglas A. Drevets, Zoltan I. Ungvari, and Andriy Yabluchans-kiy. "Endothelial dysfunction and impaired neurovascular cou-pling responses precede cognitive impairment in a mouse model of geriatric sepsis." Frontiers in Aging Neuroscience 13 (2021): 644733.
    View at Publisher | View at Google Scholar
49. Suganuma, Shinya, Masafumi Idei, Hidehiko Nakano, Yasuaki Koyama, Hideki Hashimoto, Nobuyuki Yokoyama, Shunsuke Takaki, and Kensuke Nakamura. "Impact of persistent inflam-mation, immunosuppression, and catabolism syndrome during intensive care admission on each post-intensive care syn-drome component in a PICS clinic." Journal of Clinical Medi-cine 12, no. 16 (2023): 5427.
    View at Publisher | View at Google Scholar
50. Van Gelder, Thomas G., Arief Lalmohamed, Toine CG Egberts, Arjen JC Slooter, Marieke Zegers, Mark van den Boogaard, and Irene J. van Diem-Zaal. "The association between interleu-kin-6 inhibitors as treatment for COVID-19 and mental health problems in ICU survivors." Journal of Critical Care 94 (2026): 155558.
    View at Publisher | View at Google Scholar
51. Kopp, Katherine O., Elliot J. Glotfelty, Yazhou Li, and Nigel H. Greig. "Glucagon-like peptide-1 (GLP-1) receptor agonists and neuroinflammation: Implications for neurodegenerative disease treatment." Pharmacological research 186 (2022): 106550.
    View at Publisher | View at Google Scholar
52. Wong, Chi Kin, Brent A. McLean, Laurie L. Baggio, Jacque-line A. Koehler, Rola Hammoud, Nikolaj Rittig, Julian M. Yabut, Randy J. Seeley, Theodore J. Brown, and Daniel J. Drucker. "Central glucagon-like peptide 1 receptor activation inhibits Toll-like receptor agonist-induced inflammation." Cell metabo-lism 36, no. 1 (2024): 130-143.
    View at Publisher | View at Google Scholar
53. Shafiek, Mena Z., Hala F. Zaki, Ahmed F. Mohamed, and Weam W. Ibrahim. "Novel trajectories towards possible effects of semaglutide for amelioration of reserpine-induced fibromyal-gia in rats: contribution of cAMP/PKA/p-CREB and M1/M2 mi-croglia polarization." Journal of Neuroimmune Pharmacology 20, no. 1 (2025): 43.
    View at Publisher | View at Google Scholar
54. Feng, Jing, Zhenjie Teng, Yu Yang, Jingzhen Liu, and Shuchun Chen. "Effects of semaglutide on gut microbiota, cognitive function and inflammation in obese mice." PeerJ 12 (2024): e17891.
    View at Publisher | View at Google Scholar
55. Qi, Liqin, Huimin Kang, Feihui Zeng, Menglan Zhan, Cuihua Huang, Qintao Huang, Lijing Lin et al. "Gut microbiota medi-ates semaglutide attenuation of diabetes-associated cognitive decline." Neurotherapeutics (2025): e00615.
    View at Publisher | View at Google Scholar
56. Wong, Sabrina, Gia Han Le, Christine E. Dri, Kayla M. Teopiz, and Roger S. McIntyre. "Evaluating biased agonism of gluca-gon-like peptide-1 (GLP-1) receptors to improve cellular bioen-ergetics: a systematic review." Diabetes, Obesity and Metabo-lism 27, no. 11 (2025): 6105-6115.
    View at Publisher | View at Google Scholar
57. Hughes, Christopher G., Mayur B. Patel, Nathan E. Brummel, Jennifer L. Thompson, J. Brennan McNeil, Pratik P. Pandhari-pande, James C. Jackson et al. "Relationships between mark-ers of neurologic and endothelial injury during critical illness and long-term cognitive impairment and disability." Intensive care medicine 44, no. 3 (2018): 345-355.
    View at Publisher | View at Google Scholar
58. Zeineddin, Ahmad, Feng Wu, Wei Chao, Lin Zou, Roumen Ves-selinov, Amanda M. Chipman, Jing Fei Dong, Huang Huang, Shibani Pati, and Rosemary A. Kozar. "Biomarkers of endo-thelial cell dysfunction persist beyond resuscitation in patients with hemorrhagic shock." Journal of Trauma and Acute Care Surgery 93, no. 5 (2022): 572-578.
    View at Publisher | View at Google Scholar
59. Patterson, Eric K., Gediminas Cepinskas, and Douglas D. Fra-ser. "Endothelial glycocalyx degradation in critical illness and injury." Frontiers in medicine 9 (2022): 898592.
    View at Publisher | View at Google Scholar
60. Yu, Wen-Kuang, J. Brennan McNeil, Nancy E. Wickersham, Ciara M. Shaver, Julie A. Bastarache, and Lorraine B. Ware. "Angiopoietin-2 outperforms other endothelial biomarkers as-sociated with severe acute kidney injury in patients with severe sepsis and respiratory failure." Critical care 25, no. 1 (2021): 48.
    View at Publisher | View at Google Scholar
61. Carranza-Naval, Maria Jose, Angel Del Marco, Carmen Hi-erro-Bujalance, Pilar Alves-Martinez, Carmen Infante-Garcia, Maria Vargas-Soria, Marta Herrera et al. "Liraglutide reduces vascular damage, neuronal loss, and cognitive impairment in a mixed murine model of Alzheimer’s disease and type 2 diabe-tes." Frontiers in aging neuroscience 13 (2021): 741923.
    View at Publisher | View at Google Scholar
62. Pierret, Aureliane CS, Yuya Mizuno, Pippa Saunders, Esha-ya Lim, Riccardo De Giorgi, Oliver D. Howes, Robert A. Mc-Cutcheon et al. "Glucagon-like peptide 1 receptor agonists and mental health: a systematic review and meta-analysis." JAMA psychiatry 82, no. 7 (2025): 643-653.
    View at Publisher | View at Google Scholar
63. Yi, Yu Tian, Yang Jing Zheng, Stavroula I. Bargiota, Christine E. Dri, Kayla M. Teopiz, Sabrina Wong, Gia Han Le et al. "The effect of GLP-1 receptor agonists on anxiety: A systematic re-view." Psychoneuroendocrinology (2026): 107844.
    View at Publisher | View at Google Scholar
64. Xu, Tianyi, Yang Jing Zheng, Sabrina Wong, Christine E. Dri, Xin Tong Zhou, Kayla M. Teopiz, Gia Han Le, and Roger S. McIntyre. "Molecular Evidence Associating GLP-1 Receptor Agonists and Brain-Derived Neurotrophic Factors in Neurode-generative and Psychiatric Disorders: A Systematic Review." Asian Journal of Psychiatry (2026): 104870.
    View at Publisher | View at Google Scholar
65. Ebrahimi, Pouya, Juan Carlos Batlle, Aryan Ayati, M. Haisum Maqsood, Clarine Long, Constantine Tarabanis, Natalie Mc-gowan et al. "Suicide and self-harm events with GLP-1 recep-tor agonists in adults with diabetes or obesity: a systematic review and meta-analysis." JAMA psychiatry 82, no. 9 (2025): 888-895.
    View at Publisher | View at Google Scholar
66. Ghosh, Puja, Rosaria Anna Fontanella, Lucia Scisciola, Ada Pesapane, Fatemeh Taktaz, Martina Franzese, Armando Puocci et al. "Targeting redox imbalance in neurodegeneration: characterizing the role of GLP-1 receptor agonists." Theranos-tics 13, no. 14 (2023): 4872.
    View at Publisher | View at Google Scholar
67. Mira, Juan C., Lori F. Gentile, Brittany J. Mathias, Philip A. Efron, Scott C. Brakenridge, Alicia M. Mohr, Frederick A. Moore, and Lyle L. Moldawer. "Sepsis pathophysiology, chron-ic critical illness, and persistent inflammation-immunosuppres-sion and catabolism syndrome." Critical care medicine 45, no. 2 (2017): 253-262.
    View at Publisher | View at Google Scholar
68. Gentile, Lori F., Alex G. Cuenca, Philip A. Efron, Darwin Ang, Azra Bihorac, Bruce A. McKinley, Lyle L. Moldawer, and Fred-erick A. Moore. "Persistent inflammation and immunosuppres-sion: a common syndrome and new horizon for surgical inten-sive care." Journal of Trauma and Acute Care Surgery 72, no. 6 (2012): 1491-1501.
    View at Publisher | View at Google Scholar
69. Hawkins, Russell B., Steven L. Raymond, Julie A. Stortz, Hi-royuki Horiguchi, Scott C. Brakenridge, Anna Gardner, Philip A. Efron et al. "Chronic critical illness and the persistent in-flammation, immunosuppression, and catabolism syndrome." Frontiers in immunology 9 (2018): 1511.
    View at Publisher | View at Google Scholar
70. Horiguchi, Hiroyuki, Tyler J. Loftus, Russell B. Hawkins, Ste-ven L. Raymond, Julie A. Stortz, McKenzie K. Hollen, Brett P. Weiss et al. "Innate immunity in the persistent inflammation, immunosuppression, and catabolism syndrome and its impli-
    View at Publisher | View at Google Scholar
71. Efron, Philip A., Scott C. Brakenridge, Alicia M. Mohr, Evan L. Barrios, Valerie E. Polcz, Stephen Anton, Tezcan Oz-razgat-Baslanti et al. "The persistent inflammation, immuno-suppression, and catabolism syndrome 10 years later." Journal of Trauma and Acute Care Surgery 95, no. 5 (2023): 790-799.
    View at Publisher | View at Google Scholar
72. Fenner, Brittany P., D. B. Darden, Lauren S. Kelly, Jaimar Rincon, Scott C. Brakenridge, Shawn D. Larson, Frederick A. Moore, Philip A. Efron, and Lyle L. Moldawer. "Immunological endotyping of chronic critical illness after severe sepsis." Fron-tiers in Medicine 7 (2021): 616694.
    View at Publisher | View at Google Scholar
73. Zhong, Ming, Tingting Pan, Na-Na Sun, Ruo-Ming Tan, Wen Xu, Yu-Zhen Qiu, Jia-Lin Liu, Er-Zhen Chen, and Hong-Ping Qu. "Early prediction for persistent inflammation-immunosup-pression catabolism syndrome in surgical sepsis patients." International Journal of General Medicine (2021): 5441-5448.
    View at Publisher | View at Google Scholar
74. Hesselink, Lillian, Ruben J. Hoepelman, Roy Spijkerman, Mark CH de Groot, Karlijn JP van Wessem, Leo Koenderman, Luke PH Leenen, and Falco Hietbrink. "Persistent inflammation, im-munosuppression and catabolism syndrome (PICS) after poly-trauma: a rare syndrome with major consequences." Journal of Clinical Medicine 9, no. 1 (2020): 191.
    View at Publisher | View at Google Scholar
75. Laurindo, Lucas Fornari, Sandra Maria Barbalho, Elen Land-graf Guiguer, Maricelma da Silva Soares de Souza, Gabriela Achete de Souza, Thiago Marques Fidalgo, Adriano Cressoni Araujo et al. "GLP-1a: going beyond traditional use." Interna-tional journal of molecular sciences 23, no. 2 (2022): 739.
    View at Publisher | View at Google Scholar
76. An, Ji-Ren, Jun-Tong Liu, Xiao-Meng Gao, Qing-Feng Wang, Gui-Yan Sun, Jia-Nan Su, Chi Zhang, Jia-Xiang Yu, Yu-Feng Yang, and Yan Shi. "Effects of liraglutide on astrocyte polariza-tion and neuroinflammation in db/db mice: focus on iron over-load and oxidative stress." Frontiers in cellular neuroscience 17 (2023): 1136070.
    View at Publisher | View at Google Scholar
77. Iwasa, Kensuke, Shinji Yamamoto, Kota Yamashina, Nan Yagishita-Kyo, Kei Maruyama, Takeo Awaji, Yoshinori Takei, Akira Hirasawa, and Keisuke Yoshikawa. "A peripheral lip-id sensor GPR120 remotely contributes to suppression of PGD2-microglia-provoked neuroinflammation and neurode-generation in the mouse hippocampus." Journal of neuroin-flammation 18, no. 1 (2021): 304.
    View at Publisher | View at Google Scholar
78. Au, Hezekiah CT, Pak Ho Lam, Poh Khuen Lim, and Roger S. McIntyre. "Role of glucagon-like peptide-1 on amyloid, tau, and α-synuclein: target engagement and rationale for the develop-ment in neurodegenerative disorders." Neuroscience & Biobe-havioral Reviews 173 (2025): 106159.
    View at Publisher | View at Google Scholar
79. Kornelius, Edy, Sing-Hua Tsou, Ching-Chi Chang, Ying-Jui Ho, Sheng-Chieh Lin, Wei-Liang Chen, Chien-Ning Huang, and Chih-Li Lin. "Liraglutide attenuates glucolipotoxicity-induced RSC96 Schwann cells’ inflammation and dysfunction." Biomol-ecules 12, no. 10 (2022): 1338.
    View at Publisher | View at Google Scholar
80. Yang, Kaijie, Changyuan Liu, Qiqiang Guo, and Yongze Li. "GLP-1 Receptor Agonists and Noncardiometabolic Outcomes: An Umbrella Review of Meta-Analyses." JAMA Network Open 9, no. 3 (2026): e264722.
    View at Publisher | View at Google Scholar
81. Ghidewon, M., H. S. Wald, A. D. McKnight, B. C. De Jonghe, D. M. Breen, A. L. Alhadeff, T. Borner, and H. J. Grill. "Growth differentiation factor 15 (GDF15) and semaglutide inhibit food intake and body weight through largely distinct, additive mech-anisms." Diabetes, Obesity and Metabolism 24, no. 6 (2022): 1010-1020.
    View at Publisher | View at Google Scholar
82. Cummings, Jeffrey L., Alireza Atri, Howard H. Feldman, Oskar Hansson, Mary Sano, Filip K. Knop, Peter Johannsen, Teresa León, and Philip Scheltens. "evoke and evoke+: design of two large-scale, double-blind, placebo-controlled, phase 3 stud-ies evaluating efficacy, safety, and tolerability of semaglutide in early-stage symptomatic Alzheimer’s disease." Alzheimer's research & therapy 17, no. 1 (2025): 14.
    View at Publisher | View at Google Scholar
83. Scheltens, Philip, Alireza Atri, Howard H. Feldman, Henrik Zetterberg, Mary Sano, Peter Johannsen, Teresa León Colom-bo et al. "Baseline characteristics from evoke and evoke+: Two phase 3 randomized placebo-controlled trials of semaglutide in participants with early-stage symptomatic Alzheimer's dis-ease." Alzheimer's & Dementia: Translational Research & Clin-ical Interventions 12, no. 1 (2026): e70200.
    View at Publisher | View at Google Scholar
84. Koychev, Ivan, Amanda I. Adler, Paul Edison, Brian Tom, Jo-anne E. Milton, Joe Butchart, Adam Hampshire et al. "Protocol for a double-blind placebo-controlled randomised controlled trial assessing the impact of oral semaglutide in amyloid posi-tivity (ISAP) in community dwelling UK adults." BMJ open 14, no. 6 (2024): e081401.
    View at Publisher | View at Google Scholar
85. De Giorgi, Riccardo, Ana Ghenciulescu, Courtney Yotter, Max-ime Taquet, and Ivan Koychev. "Glucagon-like peptide-1 re-ceptor agonists for major neurocognitive disorders." Journal of Neurology, Neurosurgery & Psychiatry (2025): jnnp-2024.
    View at Publisher | View at Google Scholar
86. Zhou, Zhi Dong, Lingxiao Yi, Karolina Popławska-Domasze-wicz, Kallol Ray Chaudhuri, Joseph Jankovic, and Eng King Tan. "Glucagon-like peptide-1 receptor agonists in neurode-generative diseases: Promises and challenges." Pharmaco-logical Research 216 (2025): 107770.
    View at Publisher | View at Google Scholar
87. Lee, Serene, Liyang Yin, Kayla M. Teopiz, Sabrina Wong, Gia Han Le, Naomi Xiao, Stephen Stahl et al. "Effects of gluca-gon-like peptide-1 receptor agonists on psychiatric disorders: a systematic review." Therapeutic Advances in Psychopharma-cology 15 (2025): 20451253251396304.
    View at Publisher | View at Google Scholar
88. Bi, Zheng, Zhiyu Jiao, Jinju Li, and Zhaohui Fang. "Gluca-gon-like peptide-1 receptor agonists and risk of depression: a systematic review and meta-analysis." BMC psychiatry 26, no. 1 (2026): 379.
    View at Publisher | View at Google Scholar
89. Kumar, Visesh, T. Y. Sudha, Debanjan Bhattacharjee, Azfar Mateen, and Sumit Kumar Mahato. "Glucagon-like Peptide-1 Receptor Agonists as a Dual-target Strategy for Depression and Metabolic Health: A Narrative Review." Annals of African Medicine (2026).
    View at Publisher | View at Google Scholar
90. Alshehri, Ghadah H., Hayder M. Al-Kuraishy, Ali I. Al-Gareeb, Mohamed N. Fawzy, Huda Jaber Waheed, Marios Papada-kis, Athanasios Alexiou, and Gaber El-Saber Batiha. "A novel therapeutic prospect: a dual-acting tirzepatide for Alzheimer's disease." European journal of pharmacology (2025): 177979.
    View at Publisher | View at Google Scholar
91. Alhowail, Ahmad, Mohammed F. Aldawsari, and Maha Aldu-bayan. "Comparative analysis of pioglitazone and tirzepatide on body weight, glucose levels, neuroinflammation, and oxi-dative stress in diabetic rats." Drug Design, Development and Therapy (2025): 6605-6618.
    View at Publisher | View at Google Scholar
92. Ognard, Julien, Sevda Alipour Khabir, Sherief Ghozy, Gerard El Hajj, Kevin M. Kallmes, John J. Chen, Ramanathan Kadirvel, David F. Kallmes, and Waleed Brinjikji. "Use of glucagon-like peptide-1 receptor agonists in idiopathic intracranial hyperten-sion: a systematic review." The Journal of Headache and Pain 26, no. 1 (2025): 202.
    View at Publisher | View at Google Scholar
93. Wong, Sabrina, Gia Han Le, Kayla Teopiz, and Roger S. McIn-tyre. "Investigating the Functional Connectivity Between Cen-tral Glucagon-like Peptide-1 (GLP-1) and Glutamatergic Sig-naling: A Systematic Review." CNS spectrums: 1-40.
    View at Publisher | View at Google Scholar
94. Ballum, Hana, Christine Dri, Sonya Liao, Minyi Yu, Gia Han Le, Sabrina Wong, Kayla M. Teopiz et al. "The effect of gluca-gon-like peptide 1 (GLP-1) receptor agonists on cognition: A systematic review of systematic reviews and meta-analyses." Journal of Affective Disorders (2026): 121310.
    View at Publisher | View at Google Scholar
95. Choudhury, Ishraq, John Headly Ward, Sahana Mahesh, Uazman Alam, Shazli Azmi, and Matthew Anson. "Effect of Glucagon-Like-Peptide-1 Receptor Agonists (GLP-1 RA) on Neuropsychiatric Outcomes: A Systematic Review and Me-ta-Analysis." Clinical therapeutics (2026).
    View at Publisher | View at Google Scholar
96. Stefanou, Maria-Ioanna, Evangelos Panagiotopoulos, An-astasios Tentolouris, Aikaterini Theodorou, Georgia Papa-giannopoulou, Athanasia Athanasaki, Panagiota-Eleni Tsa-louchidou et al. "Efficacy and safety of glucagon-like peptide-1 receptor agonists in Parkinson’s disease: a systematic review and meta-analysis of randomized placebo-controlled clinical trials." Therapeutic Advances in Neurological Disorders 19 (2026): 17562864251408269.
    View at Publisher | View at Google Scholar
97. Smith, Camryn J., Dasia Hodge, Fiona E. Harrison, and Shawniqua Williams Roberson. "The pathophysiology and bio-markers of delirium." In Seminars in neurology, vol. 44, no. 06, pp. 720-731. Thieme Medical Publishers, Inc., 2024.
    View at Publisher | View at Google Scholar
98. Nelson, David W., Tobias Granberg, Pia Andersen, Elias Jokhadar, Jessica Kåhlin, Anna Granström, Helena Hallinder et al. "The Karolinska NeuroCOVID study protocol: Neurocog-nitive impairment, biomarkers and advanced imaging in critical care survivors." Acta Anaesthesiologica Scandinavica 66, no. 6 (2022): 759-766.
    View at Publisher | View at Google Scholar
99. Haastrup, Adeyeye I., Justin T. Roberts, Sheetal Gandotra, Emily M. Hartsell, Grant T. Daly, Viktor M. Pastukh, Lina D. Pur-cell et al. "Metabolomic biomarkers predict long-term physical function in survivors of acute respiratory failure." Respiratory Research 27, no. 1 (2026): 4.
    View at Publisher | View at Google Scholar
100. Mottale, Raphael, Claire Dupuis, Sylwia Szklarzewska, and Jean-Charles Preiser. "The metabolic response to stress in critical illness: updated review on the pathophysiological mech-anisms, consequences, and therapeutic implications." Annals of Intensive Care 15, no. 1 (2025): 174.
     View at Publisher | View at Google Scholar
101. Docherty, C., C. Page, J. Wilson, P. Ross, K. Garrity, T. Quasim, M. Shaw, and J. McPeake. "Association between inflammation and post-intensive care syndrome: a systematic review." An-aesthesia 79, no. 7 (2024): 748-758.
     View at Publisher | View at Google Scholar
102. Kong, Fanjing, Yanru Zhao, Weiming Zhang, Xinyi Wang, Tianyu Wu, Ziyi Zhou, Ying Xu, Lina Xia, and Tao Sun. "Com-prehensive evaluation of GLP-1 receptor agonists: an umbrella review of clinical outcomes across multiple diseases." Nature Communications (2026).
     View at Publisher | View at Google Scholar
103. Sioufi, Maria-Christina, Isabela Heroiu, Sabrina Wong, Gia Han Le, Christine E. Dri, Yang Jing Zheng, Taeho Greg Rhee et al. "The effect of GLP-1 receptor agonists on autophagy: insights gathered from research evaluating neurodegenerative disorders with these agents." Acta Neuropsychiatrica (2026): 1-36.
     View at Publisher | View at Google Scholar
104. Liu, Keibun, Toru Kotani, Kensuke Nakamura, Takai Chihiro, Yasunari Morita, Kenzo Ishii, Kenji Fujizuka et al. "Effects of evidence-based ICU care on long-term outcomes of patients with sepsis or septic shock (ILOSS): protocol for a multicentre prospective observational cohort study in Japan." BMJ open 12, no. 3 (2022): e054478.
     View at Publisher | View at Google Scholar