Hypothesis-Generating Brief: Brain age MRI

Hypothesis-Generating Brief: Brain age MRI

Abstract

Evidence-honesty note: 63/65 retained sources are coded as null or no extracted directional signal; this corpus is non-supportive for clinical efficacy claims and hypothesis-generating only. Source-bundle reconciliation note: Directional coding is conservative claim-level coding from extracted claim records, not a statement that the source texts contain no directional findings; source-level positive, negative, or unclear findings should be interpreted through the coded outcome class, directness, and claim-count fields. 64/65 retained sources are indirect, review-level, adjacent, or mechanistic and are used only to bound interpretation. The conclusion therefore does not support broad causal, clinical, or policy claims.

This paper synthesizes evidence on Brain age MRI across 65 accepted source papers and 1135 high-confidence extracted claims.

The evidence profile contains 1 direct clinical source, 64 adjacent clinical sources, and no sources classified primarily as mechanistic or model-system evidence, with 66 cross-study disagreements across the evidence base.

Positive study-level signals are summarized in the cardiometabolic outcome class, null signals in the contextual adjacent evidence, safety and comorbidity, cardiometabolic outcome classes, and negative signals in no dominant outcome class. The paper therefore interprets the corpus as a tiered evidence profile rather than as a single pooled effect.

The conclusion is that Brain age MRI remains a bounded geroscience case: the retained clinical and adjacent evidence profile defines the scope for targeted testing, while mixed and null findings limit any unqualified anti-aging claim.

Methods

Risk-of-bias appraisal summary: The public appraisal artifact reports 65 source-level rating row(s) using ROBINS-I, RoB-2, SYRCLE; overall ratings are some concerns=65. These ratings summarize preliminary source-level appraisal and do not upgrade indirect or adjacent evidence into direct clinical proof.

Review type and protocol

This manuscript is reported as a Evidence brief. A deterministic protocol governed source retrieval, screening, extraction, and synthesis; the protocol was frozen before manuscript rendering. The full audit trail is in the supplementary methods_pack.json and the timestamped submission directory synthesis-brain_age_mri-v06-DAILY-2026-06-21T15-19-07Z-R2.

Information sources

Sources were retrieved across PubMed, Europe PMC, OpenAlex, Semantic Scholar, Crossref, DOAJ, OpenAIRE, PMC OAI, bioRxiv, medRxiv, arXiv, and ClinicalTrials.gov. Retrieval window: 2026-06-21.

Search strategy

The following topic-anchored queries were executed against the information sources listed above:

• brain age MRI AND aging AND human
• brain age MRI AND older adults
• brain age MRI AND randomized controlled trial
• brain age AND aging AND human
• brain age AND older adults
• brain age AND randomized controlled trial
• MRI brain age AND aging AND human
• MRI brain age AND older adults
• MRI brain age AND randomized controlled trial
• neuroimaging aging AND aging AND human

Eligibility criteria

• Sources whose primary content addresses brain age mri.
• Sources with extractable quantitative or qualitative findings.
• Peer-reviewed primary research, systematic reviews, or meta-analyses; preprints accepted only when source-traceable.
• Sources with verifiable bibliographic identifiers (DOI / PMID / canonical handle).

Selection of sources of evidence

The synthesis did not begin from an unfiltered database export. It began from a pre-curated receipt-candidate set generated by the retrieval and claim-binding pipeline. Of 467 records in the receipt-candidate union, 181 were classified as source candidates and 65 were admitted as traceable synthesis sources. Mixed partial-or-none and partial-only rows are separate claim-binding audit buckets, not additive exclusion totals. No additional records were excluded after final source admission.

source admission funnel

Admission bucket	n
Receipt candidate union	467
Classified source candidates	181
No extractable claims	41
None-only claim binding	18
Mixed partial-or-none claim-binding candidates	182
Partial-only claim-binding candidates	32
Strict high-confidence sources	13
Admitted final sources	65

Exclusion reasons

• No records were excluded at the gates instrumented for this run: the eligibility criteria above were applied during retrieval and claim-binding but produced no post-screening exclusions with recorded counts for this corpus.

Data items

The following fields were extracted from each included source: study design, population / cohort, intervention or exposure, comparator, outcome class, effect direction, effect size, confidence interval or credible interval, p-value, sample size, follow-up duration, risk-of-bias rating. Under the calibration rule, source verification in the public bundle is limited to reference-level metadata; exact statistics and effect directions are drawn from these structured extraction artifacts (the synthesis manifest, risk-of-bias sidecar when populated, and claim registry) rather than from re-parsed full text.

Risk-of-bias appraisal

Risk-of-bias framework assignment follows study design (RoB-2 for RCTs, ROBINS-I for non-randomised studies, AMSTAR-2 for systematic reviews / meta-analyses). Public appraisal claims are limited to populated risk_of_bias.json rows; when no populated ratings are present, interpretation remains bounded by source tier and directness rather than formal RoB certification.

Synthesis approach

Evidence-tension synthesis: claims grouped by outcome class (cardiometabolic, cognitive, contextual adjacent evidence, frailty, immune and inflammation, muscle function, safety and comorbidity); within-class agreement, disagreement, and directness gaps surfaced explicitly. Quantitative pooling applied only where ≥3 sources reported a comparable endpoint with extractable effect estimates.

AI-use disclosure

Source retrieval, claim extraction, evidence routing, and prose drafting were assisted by large language models under a deterministic audit-trail protocol. Every manuscript claim is traceable to a source record in the supplementary manifest.json. Final eligibility and interpretation decisions are author-verified.

Accountability

Accountability is established through reproducible artifacts: a deterministic protocol (methods_pack.json), a complete claim and citation registry, extracted numeric trace, deterministic gates (full_paper.journal_surface.json, pre_submit_gate.json, artifact_consistency.json), and a versioned correction path documented in the run's submission record. Certification under the researka_agent_certified model verifies that the manuscript is machine-verifiable, internally consistent, provenance-traced, and format-checked against these artifacts; it does not adjudicate domain correctness, corpus fit, or novelty, which remain subject to expert and reader review.

Evidence Landscape

Directional coding note: Null or no extracted directional signal means no coded positive, negative, or mixed effect was extracted for that specific outcome class; it is not an absence-of-support finding. Positive, negative, mixed, unclear, and null are outcome-specific codes, so a bounded rationale can be supported by adjacent or different outcome evidence while another outcome remains null or unclear. Contextual claims contain bibliographic background, mechanism, methods, exposure definitions, or population context rather than effect-direction evidence. When an outcome-class summary uses no extracted directional signal, it should state the source proportion, such as X/Y sources, to avoid ambiguity.

RCT-count reconciliation: Reviewer feedback indicates that at least one included source aggregates more than one randomized trial, so this manuscript treats any prior single-RCT wording as a source-coding count, not as a claim that the underlying trial evidence contains only one RCT.

Substantive evidence synthesis: The manifest includes 65 retained sources, 1 direct-source row(s), and directional coding across null=63, positive=1, unclear=1. Representative source-level signals are: Levakov 2023: outcome=Cardiometabolic; direction=positive; directness=indirect; tier=B2; claims=56; Yilmaz 2025: outcome=Contextual Adjacent Evidence; direction=unclear; directness=indirect; tier=B2; claims=29; Huang 2025: outcome=Immune and Inflammation; direction=null; directness=indirect; tier=B2; claims=60; Ran 2022: outcome=Contextual Adjacent Evidence; direction=null; directness=indirect; tier=B2; claims=58; Tanner 2025: outcome=Safety and Comorbidity; direction=null; directness=indirect; tier=B2; claims=50; Selitser 2025: outcome=Contextual Adjacent Evidence; direction=null; directness=review; tier=B2; claims=43; Narula 2026: outcome=Contextual Adjacent Evidence; direction=null; directness=indirect; tier=B2; claims=43; Bao 2022: outcome=Contextual Adjacent Evidence; direction=null; directness=indirect; tier=B2; claims=43. These signals inform the bounded conclusion by separating effect direction from evidence tier/directness; indirect, review-level, mechanistic, or contextual evidence remains hypothesis-generating.

Key Findings

Key findings from source synthesis: First, the strongest positive or favorable signals are treated as narrow source-level signals, not broad clinical proof (Levakov 2023: outcome=Cardiometabolic; direction=positive; directness=indirect; tier=B2; claims=56; Yilmaz 2025: outcome=Contextual Adjacent Evidence; direction=unclear; directness=indirect; tier=B2; claims=29; Huang 2025: outcome=Immune and Inflammation; direction=null; directness=indirect; tier=B2; claims=60). Second, negative, mixed, unclear, or no-directional-signal rows are given equal interpretive weight (Ran 2022: outcome=Contextual Adjacent Evidence; direction=null; directness=indirect; tier=B2; claims=58; Tanner 2025: outcome=Safety and Comorbidity; direction=null; directness=indirect; tier=B2; claims=50; Selitser 2025: outcome=Contextual Adjacent Evidence; direction=null; directness=review; tier=B2; claims=43). Third, the bounded conclusion follows from the balance of source direction, outcome class, evidence tier, and directness rather than from source count alone.

Results

Evidence domain	Corpus slice	Strongest signal	Directness	Main limitation
Contextual Adjacent Evidence	n=51; claims=842	no extracted directional signal in 50/51 sources	1 direct; 47 indirect; 3 review	limited corpus depth in this outcome class
Safety and Comorbidity	n=5; claims=109	no extracted directional signal in 5/5 sources	5 indirect	limited corpus depth in this outcome class
Cardiometabolic	n=3; claims=86	no extracted directional signal in 2/3 sources	3 indirect	limited corpus depth in this outcome class
Frailty	n=2; claims=15	no extracted directional signal in 2/2 sources	2 indirect	limited corpus depth in this outcome class
Muscle Function	n=2; claims=2	no extracted directional signal in 2/2 sources	2 indirect	limited corpus depth in this outcome class
Cognitive	n=1; claims=21	no extracted directional signal in 1/1 sources	1 indirect	single-source slice; hypothesis-generating
Immune and Inflammation	n=1; claims=60	no extracted directional signal in 1/1 sources	1 indirect	single-source slice; hypothesis-generating

Outcome-class note: Contextual Adjacent Evidence denotes background, boundary-condition, or adjacent-outcome sources. It is not pooled with direct outcome evidence; these sources bound scope, safety, methods, and translation rather than serving as equal-weight support for the main efficacy claim.

This evidence brief reports outcome packets as a map of retained evidence rather than as a full journal Results narrative or pooled effect estimate.

Contextual Adjacent Evidence Outcomes

51 included sources were assigned to this outcome class. Directional coding: null=50, unclear=1. Directness coding: direct=1, indirect=47, review=3.

Safety Comorbidity Outcomes

5 included sources were assigned to this outcome class. Directional coding: null=5. Directness coding: indirect=5.

Cardiometabolic Outcomes

3 included sources were assigned to this outcome class. Directional coding: null=2, positive=1. Directness coding: indirect=3.

Frailty Outcomes

2 included sources were assigned to this outcome class. Directional coding: null=2. Directness coding: indirect=2.

Muscle Function Outcomes

2 included sources were assigned to this outcome class. Directional coding: null=2. Directness coding: indirect=2.

Cognitive Outcomes

1 included source were assigned to this outcome class. Directional coding: null=1. Directness coding: indirect=1.

Immune Inflammation Outcomes

1 included source were assigned to this outcome class. Directional coding: null=1. Directness coding: indirect=1.

Limitations

Verification note: Reference-only or no-abstract records are treated as verification-limited context, not as equal-weight support for the main claim.

The curated corpus on brain-age MRI is overwhelmingly observational, with a single randomized trial (Haudry 2025, an RCT with a mechanistic/biomarker endpoint) supplying direct interventional evidence in older adults; no long-term mortality or hard-outcome RCTs in non-diabetic or non-meditation populations are present, so causal claims about anti-aging benefit cannot be sustained. The cardiometabolic and immune-inflammation outcome classes are represented only by cohort designs (Levakov 2023, Motaghi 2025, Huang 2025, Mouches 2022, Derboghossian 2024, Selitser 2025, Tavakoli 2025), and even within those cohorts effect directions diverge — Levakov 2023 reports a positive weight-loss effect after 18 months of lifestyle intervention while Mouches 2022 and Derboghossian 2024 report null associations between cardiovascular risk factors and brain-age gap, leaving the cardiometabolic signal unresolved. The absence of replication-grade interventional evidence means the headline synthesis is constrained to biomarker associations rather than clinical benefit, and the headline-level null-vs-positive tension in cardiometabolic outcomes is not adjudicable from this corpus alone.

Several outcome claims rest on a single source and therefore cannot be internally replicated within the corpus. The Tai-Chi/balance-exercise MRI analysis (Narula 2026) and the unilateral exercise-in-schizophrenia brain-age-gap finding (Yilmaz 2025, n=134) similarly stand alone, so their directional signals — including the null and unclear direction codes — cannot be triangulated, and the synthesis cannot promote any of them to a robust claim without external replication.

The enrolled populations are narrow on demographic and clinical axes, restricting external validity.

Hard clinical endpoints are not measured in the included evidence. Falls, hospitalization, disability, and mortality are similarly absent; the only survival-related signal is Casanova 2024's elastic-net Cox model against all-cause mortality using SOMAscan proteins. As a result, the brain-age-MRI case is built entirely on surrogate associations, which carry the well-documented risk that biomarker movement does not translate into clinical benefit, and the corpus cannot adjudicate whether observed brain-age gap reductions (e. For example, Yilmaz 2025 in schizophrenia, Levakov 2023 with weight loss, Haudry 2025 with meditation) would yield fewer events if scaled.

Where the corpus might appear to support a clinically actionable claim, the underlying evidence is mechanistic rather than clinical. The cross-study disagreements the synthesis surfaces — most prominently mechanism vs clinical cross-domain pairs in which a direct contextual-other trial must be kept separate from indirect cardiometabolic, frailty, immune, safety, cognitive, and muscle-function cohorts, and indirectness gap pairs separating Haudry 2025 from the broader contextual other literature — are a direct consequence of this mechanism-to-clinic gap, and the corpus provides no longitudinal data linking an MRI-derived brain-age change to a subsequent clinical event within the same cohort.

Conclusion

For Brain age MRI, the final interpretation is deliberately tiered: the retained clinical and adjacent evidence profile defines a bounded geroscience rationale, but the corpus does not support treating mechanistic target engagement, intermediate biomarkers, and patient-relevant outcomes as interchangeable evidence. The closing claim should therefore be read as a map of what the retained studies can support, not as a clinical recommendation or a general anti-aging endorsement. Positive signals identify hypotheses and candidate contexts; null, mixed, or adverse signals identify the boundaries that future work must test directly. The evidence hierarchy remains load-bearing here: direct interventional hard-endpoint records carry more interpretive weight than adjacent clinical evidence, and both carry more translational weight than mechanistic or model systems. A stronger future conclusion would require larger direct human samples, prespecified endpoints, longer follow-up, comparable intervention characterization, transparent safety capture, and a consistent direction of effect across clinically proximate outcomes. Until that evidence exists, the paper's conclusion is that the topic is worth structured follow-up only within the boundaries defined by the included source set. That boundary is not a weakness in the paper; it is the main claim that keeps the synthesis reusable. Readers should carry forward the evidence classes separately: favorable mechanistic or surrogate findings can motivate experiments, indirect human findings can prioritize populations and endpoints, and direct clinical findings define the current ceiling for applied interpretation. The current corpus is non-supportive for clinical efficacy or general health-intervention claims; it supports only hypothesis generation and structured follow-up within the limits of indirect evidence. Any downstream use should preserve that tiered reading rather than compressing the corpus into a simple yes/no verdict for clinical practice or public messaging.

What This Synthesis Adds

This synthesis maps 65 included sources on Brain Age MRI across 7 outcome classes and 66 cross-study disagreements. It separates endpoint-specific evidence from broad geroprotection claims so that favorable biomarker signals are not treated as proof of durable healthspan benefit.

Across 65 curated reference papers, the evidence base for Brain shows a context-dependent profile. Positive signals appear in: cardiometabolic. Null findings dominate: contextual other, safety comorbidity. The synthesis surfaces cross-study disagreements across outcome classes — see Cross-Domain Synthesis. The Brain anti-aging case as currently constituted is incomplete: mechanistic plausibility coexists with mixed or sparse human-RCT evidence, and the boundary conditions remain to be established.

The strongest unresolved contrast is the null vs positive between Levakov 2023 and Derboghossian 2024 on cardiometabolic (severity 4/5), which defines the boundary condition future studies must test rather than smooth over.

This synthesis adds a design-level evidence-weighting layer and an explicit cross-study disagreement map, keeping boundary conditions visible instead of averaging them away in narrative summary.

Boundary-Condition Matrix

Evidence domain	Direct sources	Indirect / mechanism sources	Direction profile	Interpretation boundary
cardiometabolic	0	3	null, positive	conflict-resolution gap
cognitive	0	1	null	direct interventional hard-endpoint gap
frailty	0	2	null	direct interventional hard-endpoint gap
muscle function	0	2	null	direct interventional hard-endpoint gap
immune and inflammation	0	1	null	direct interventional hard-endpoint gap
safety and comorbidity	0	5	null	direct interventional hard-endpoint gap
contextual adjacent evidence	1	50	null, unclear	replication gap

Evidence-Gap Priority

Priority	Gap	Rationale
P1	cardiometabolic: conflict-resolution gap	0 direct and 3 indirect sources; direction profile: null, positive
P2	cognitive: direct interventional hard-endpoint gap	0 direct and 1 indirect source; direction profile: null
P3	frailty: direct interventional hard-endpoint gap	0 direct and 2 indirect sources; direction profile: null
P4	muscle function: direct interventional hard-endpoint gap	0 direct and 2 indirect sources; direction profile: null
P5	immune and inflammation: direct interventional hard-endpoint gap	0 direct and 1 indirect source; direction profile: null

Next-Study Design Recommendation

The next high-yield study for Brain Age MRI should target the cardiometabolic evidence gap, pre-register the primary endpoint, separate clinical from mechanistic endpoints, preserve safety and adherence capture, and include an analysis plan that can falsify the current boundary-condition claim rather than only confirming a favorable direction. Minimum useful design: at least 200 participants per arm, a priority population of adults or older adults with baseline risk in the target outcome domain, and follow-up lasting at least 24 weeks; shorter or smaller studies should be treated as hypothesis-generating.

Evidence Snapshot

The manuscript foregrounds the load-bearing evidence; the full evidence tables remain in the supplement.

Load-Bearing Included Studies

• Haudry 2025; tier=A1; directness=direct; endpoint=contextual adjacent evidence; direction=null; representative statistic=P = 0.14.
• Huang 2025; tier=B2; directness=indirect; endpoint=immune inflammation; direction=null.
• Ran 2022; tier=B2; directness=indirect; endpoint=contextual adjacent evidence; direction=null; representative statistic=P = 0.139.
• Levakov 2023; tier=B2; directness=indirect; endpoint=cardiometabolic; direction=positive; representative statistic=P < 0.001.
• Tanner 2025; tier=B2; directness=indirect; endpoint=safety comorbidity; direction=null; representative statistic=P = 0.061.
• Bao 2022; tier=B2; directness=indirect; endpoint=contextual adjacent evidence; direction=null.
• Narula 2026; tier=B2; directness=indirect; endpoint=contextual adjacent evidence; direction=null; representative statistic=P > 0.05.
• Selitser 2025; tier=B2; directness=review; endpoint=contextual adjacent evidence; direction=null.
• Lu 2024; tier=B2; directness=indirect; endpoint=contextual adjacent evidence; direction=null; representative statistic=P = 0.077.
• Liew 2023; tier=B2; directness=indirect; endpoint=contextual adjacent evidence; direction=null; representative statistic=P = 0.386.

Source Classification Map

Each retained source is mapped to its public evidence role so the evidence landscape can be checked without opening the supplement.

• Impact of meditation on brain age derived from multimodal neuroimaging in experts and older adults from a randomized trial: outcome=contextual adjacent evidence; directness=direct; tier=A1; direction=null; claims=17.
• Association of peripheral immune markers with brain age and dementia risk estimated using deep learning methods: outcome=immune inflammation; directness=indirect; tier=B2; direction=null; claims=60.
• Brain age vector: A measure of brain aging with enhanced neurodegenerative disorder specificity: outcome=contextual adjacent evidence; directness=indirect; tier=B2; direction=null; claims=58.
• The effect of weight loss following 18 months of lifestyle intervention on brain age assessed with resting-state functional connectivity: outcome=cardiometabolic; directness=indirect; tier=B2; direction=positive; claims=56.
• More than chronic pain: behavioural and psychosocial protective factors predict lower brain age in adults with/at risk of knee osteoarthritis over two years: outcome=safety comorbidity; directness=indirect; tier=B2; direction=null; claims=50.
• Prediction of brain age using quantitative parameters of synthetic magnetic resonance imaging: outcome=contextual adjacent evidence; directness=indirect; tier=B2; direction=null; claims=43.
• The impact of balance exercise on brain age and brain morphometry: insights from MRI analysis: outcome=contextual adjacent evidence; directness=indirect; tier=B2; direction=null; claims=43.
• Cardiometabolic risk factors and brain age: a meta-analysis to quantify brain structural differences related to diabetes, hypertension, and obesity: outcome=contextual adjacent evidence; directness=review; tier=B2; direction=null; claims=43.
• Predictive values of pre-treatment brain age models to rTMS effects in neurocognitive disorder with depression: Secondary analysis of a randomised sham-controlled clinical trial: outcome=contextual adjacent evidence; directness=indirect; tier=B2; direction=null; claims=41.
• Association of Brain Age, Lesion Volume, and Functional Outcome in Patients With Stroke: outcome=contextual adjacent evidence; directness=indirect; tier=B2; direction=null; claims=40.
• Brain age revisited: Investigating the state vs. trait hypotheses of EEG-derived brain-age dynamics with deep learning: outcome=contextual adjacent evidence; directness=indirect; tier=B2; direction=null; claims=34.
• Machine Learning–Based Sleep Electroencephalographic Brain Age Index and Dementia Risk: outcome=contextual adjacent evidence; directness=indirect; tier=B2; direction=null; claims=33.
• MRI-informed machine learning-driven brain age models for classifying mild cognitive impairment converters: outcome=contextual adjacent evidence; directness=indirect; tier=B2; direction=null; claims=32.
• Multimodal brain age estimates relate to Alzheimer disease biomarkers and cognition in early stages: a cross-sectional observational study: outcome=contextual adjacent evidence; directness=indirect; tier=B2; direction=null; claims=32.
• Increased MRI-based Brain Age in chronic migraine patients: outcome=safety comorbidity; directness=indirect; tier=B2; direction=null; claims=31.
• Novel Volumetric and Surface-Based Magnetic Resonance Indices of the Aging Brain – Does Male and Female Brain Age in the Same Way?: outcome=contextual adjacent evidence; directness=indirect; tier=B2; direction=null; claims=30.
• Brain age gap reduction following exercise mirrors clinical improvements in schizophrenia spectrum disorders: outcome=contextual adjacent evidence; directness=indirect; tier=B2; direction=unclear; claims=29.
• Brain age in genetic and idiopathic Parkinson's disease: outcome=contextual adjacent evidence; directness=indirect; tier=B2; direction=null; claims=27.
• Genome-wide analysis of brain age identifies 59 associated loci and unveils relationships with mental and physical health: outcome=contextual adjacent evidence; directness=indirect; tier=B2; direction=null; claims=25.
• A deep learning model for brain age prediction using minimally preprocessed T1w images as input: outcome=contextual adjacent evidence; directness=indirect; tier=B2; direction=null; claims=24.
• Associations between contralesional neuroplasticity and motor impairment through deep learning-derived MRI regional brain age in chronic stroke (ENIGMA): a multicohort, retrospective, observational study: outcome=safety comorbidity; directness=indirect; tier=B2; direction=null; claims=24.
• Developmental Brain Age Estimation From MRI Data: A Systematic Review of Deep Learning Approaches and Open Datasets: outcome=contextual adjacent evidence; directness=review; tier=B2; direction=null; claims=24.
• Association between low‐frequency oscillations in blood pressure variability and brain age derived from neuroimaging: outcome=contextual adjacent evidence; directness=indirect; tier=B2; direction=null; claims=22.
• Brain age gap, dementia risk factors and cognition in middle age: outcome=cognitive; directness=indirect; tier=B2; direction=null; claims=21.
• The value of arterial spin labelling perfusion MRI in brain age prediction: outcome=contextual adjacent evidence; directness=indirect; tier=B2; direction=null; claims=19.
• Predicting brain age for veterans with traumatic brain injuries and healthy controls: an exploratory analysis: outcome=contextual adjacent evidence; directness=indirect; tier=B2; direction=null; claims=16.
• ASSOCIATIONS BETWEEN CARDIORESPIRATORY FITNESS, BRAIN AGE, AND NEURODEGENERATION AMONG OLDER ADULTS: outcome=cardiometabolic; directness=indirect; tier=B2; direction=null; claims=16.
• Quantitative assessment of neurodevelopmental maturation: a comprehensive systematic literature review of artificial intelligence-based brain age prediction in pediatric populations: outcome=contextual adjacent evidence; directness=review; tier=B2; direction=null; claims=16.
• Toward MR protocol-agnostic, unbiased brain age predicted from clinical-grade MRIs: outcome=contextual adjacent evidence; directness=indirect; tier=B2; direction=null; claims=16.
• Decoding MRI-informed brain age using mutual information: outcome=contextual adjacent evidence; directness=indirect; tier=B2; direction=null; claims=15.
• Longitudinal accelerated brain age in mild cognitive impairment and Alzheimer’s disease: outcome=contextual adjacent evidence; directness=indirect; tier=B2; direction=null; claims=14.
• An exploratory causal analysis of the relationships between the brain age gap and cardiovascular risk factors: outcome=cardiometabolic; directness=indirect; tier=B2; direction=null; claims=14.
• Meditation Linked to Enhanced MRI Signal Intensity in the Pineal Gland and Reduced Predicted Brain Age: outcome=contextual adjacent evidence; directness=indirect; tier=B2; direction=null; claims=13.
• Increased Brain Age Among Psychiatrically Healthy Adults Exposed to Childhood Trauma: outcome=contextual adjacent evidence; directness=indirect; tier=B2; direction=null; claims=11.
• Lifespan brain age prediction based on multiple EEG oscillatory features and sparse group lasso: outcome=contextual adjacent evidence; directness=indirect; tier=B2; direction=null; claims=11.
• MRI-based whole-brain elastography and volumetric measurements to predict brain age: outcome=contextual adjacent evidence; directness=indirect; tier=B2; direction=null; claims=10.
• Investigating the Association of Frailty Score and Diabetes with Relative Brain Age: Insights from the UK Biobank: outcome=frailty; directness=indirect; tier=B2; direction=null; claims=9.
• Sleep Patterns in Midlife and Brain Age: outcome=contextual adjacent evidence; directness=indirect; tier=B2; direction=null; claims=8.
• Plasma‐based Brain Age as a Biomarker for Cognitive Health and Risk of Brain‐Related Diseases: outcome=contextual adjacent evidence; directness=indirect; tier=B2; direction=null; claims=8.
• Brain age gap estimation using attention-based ResNet method for Alzheimer’s disease detection: outcome=contextual adjacent evidence; directness=indirect; tier=B2; direction=null; claims=7.

Classification Criteria

• Outcome class is assigned from the source's bound endpoint, population, and claim text; adjacent/background sources are separated from clinical outcome slices.
• Directness is coded as direct only when a source tests the topic against a clinically proximate outcome in the relevant population; a qualifying direct source would be a human interventional or hard-endpoint study of the topic itself. Indirect human, review-level, and mechanistic sources are weighted separately.
• Directional signal is counted within the assigned outcome class only. A no extracted directional signal cell means the retained sources in that outcome slice did not yield a coded positive, negative, or mixed direction for that slice; it is not a claim that the source reports no associations anywhere else.
• Evidence tier follows the deterministic tier/directness taxonomy used in the source builder; the prose writer cannot move a source between classes after sources are frozen.

Load-Bearing Tensions

• Severity 4 null vs positive: Levakov 2023 vs Derboghossian 2024; Levakov 2023 (positive on cardiometabolic) vs Derboghossian 2024 (null on cardiometabolic) — partial conflict
• Severity 4 null vs positive: Levakov 2023 vs Mouches 2022; Levakov 2023 (positive on cardiometabolic) vs Mouches 2022 (null on cardiometabolic) — partial conflict
• Severity 3 indirectness gap: Dijsselhof 2023 vs Haudry 2025; Haudry 2025 (direct, A1) vs Dijsselhof 2023 (indirect) on contextual other — direct vs indirect must be kept separate
• Severity 3 indirectness gap: Liew 2023 vs Haudry 2025; Haudry 2025 (direct, A1) vs Liew 2023 (indirect) on contextual other — direct vs indirect must be kept separate
• Severity 3 indirectness gap: Jonemo 2023 vs Haudry 2025; Haudry 2025 (direct, A1) vs Jonemo 2023 (indirect) on contextual other — direct vs indirect must be kept separate
• Severity 3 indirectness gap: Valdes-Hernandez 2023 vs Haudry 2025; Haudry 2025 (direct, A1) vs Valdes-Hernandez 2023 (indirect) on contextual other — direct vs indirect must be kept separate
• Severity 3 indirectness gap: Kim 2023 vs Haudry 2025; Haudry 2025 (direct, A1) vs Kim 2023 (indirect) on contextual other — direct vs indirect must be kept separate
• Severity 3 indirectness gap: Dartora 2024 vs Haudry 2025; Haudry 2025 (direct, A1) vs Dartora 2024 (indirect) on contextual other — direct vs indirect must be kept separate

Additional corpus sources informed the synthesis without anchoring a foregrounded quantitative claim and are catalogued for completeness: Gemein 2024, Sun 2026, Lu 2024b, Millar 2023, Navarro-Gonzalez 2023, Podgorski 2021, Teipel 2024, Jawinski 2025, Ull 2025, Park 2026, Heffernan 2025, Stefaniak 2024, Dragendorf 2024, Coetzee 2025, Li 2024, Ly 2024, Plini 2025, Hendrikse 2025, Hu 2025, Claros-Olivares 2024, Cavailles 2025, Wang 2025, Hanson 2024, Aghaei 2024, Pang 2024, Ahmadi 2025, Dunk 2025, Kim 2025, Pallapothu 2025, Meysami 2025, Roman 2025, Meysami 2026, Wang 2021, Kou 2024, Toraih 2025, Yu 2025, Yu 2025b, Rajabli 2025, Satpathi 2025, Rajabli 2026, Dorfel 2024, Aithal 2025, Raji 2025, Raji 2026.

References

• Huang 2025. Association of peripheral immune markers with brain age and dementia risk estimated using deep learning methods. International Journal of Surgery (London, England), 2025. DOI: 10.1097/JS9.0000000000002746. PMID: 40561180.
• Ran 2022. Brain age vector: A measure of brain aging with enhanced neurodegenerative disorder specificity. Human Brain Mapping, 2022. DOI: 10.1002/hbm.26066. PMID: 36094058.
• Levakov 2023. The effect of weight loss following 18 months of lifestyle intervention on brain age assessed with resting-state functional connectivity. eLife, 2023. DOI: 10.7554/eLife.83604. PMID: 37022140.
• Tanner 2025. More than chronic pain: behavioural and psychosocial protective factors predict lower brain age in adults with/at risk of knee osteoarthritis over two years. Brain Communications, 2025. DOI: 10.1093/braincomms/fcaf344. PMID: 41020178.
• Selitser 2025. Cardiometabolic risk factors and brain age: a meta-analysis to quantify brain structural differences related to diabetes, hypertension, and obesity. Journal of Psychiatry & Neuroscience: JPN, 2025. DOI: 10.1503/jpn.240105. PMID: 40068862.
• Narula 2026. The impact of balance exercise on brain age and brain morphometry: insights from MRI analysis. Aging Clinical and Experimental Research, 2026. DOI: 10.1007/s40520-026-03322-6. PMID: 41566095.
• Bao 2022. Prediction of brain age using quantitative parameters of synthetic magnetic resonance imaging. Frontiers in Aging Neuroscience, 2022. DOI: 10.3389/fnagi.2022.963668. PMID: 36457759.
• Lu 2024. Predictive values of pre-treatment brain age models to rTMS effects in neurocognitive disorder with depression: Secondary analysis of a randomised sham-controlled clinical trial. Dialogues in Clinical Neuroscience, 2024. DOI: 10.1080/19585969.2024.2373075. PMID: 38963341.
• Liew 2023. Association of Brain Age, Lesion Volume, and Functional Outcome in Patients With Stroke. Neurology, 2023. DOI: 10.1212/WNL.0000000000207219. PMID: 37015818.
• Gemein 2024. Brain age revisited: Investigating the state vs. trait hypotheses of EEG-derived brain-age dynamics with deep learning. Imaging Neuroscience, 2024. DOI: 10.1162/imag_a_00210. PMID: 40800431.
• Sun 2026. Machine Learning–Based Sleep Electroencephalographic Brain Age Index and Dementia Risk. JAMA Network Open, 2026. DOI: 10.1001/jamanetworkopen.2026.1521. PMID: 41854616.
• Lu 2024b. MRI-informed machine learning-driven brain age models for classifying mild cognitive impairment converters. Journal of Central Nervous System Disease, 2024. DOI: 10.1177/11795735241266556. PMID: 39049837.
• Millar 2023. Multimodal brain age estimates relate to Alzheimer disease biomarkers and cognition in early stages: a cross-sectional observational study. eLife, 2023. DOI: 10.7554/eLife.81869. PMID: 36607335.
• Navarro-Gonzalez 2023. Increased MRI-based Brain Age in chronic migraine patients. The Journal of Headache and Pain, 2023. DOI: 10.1186/s10194-023-01670-6. PMID: 37798720.
• Podgorski 2021. Novel Volumetric and Surface-Based Magnetic Resonance Indices of the Aging Brain – Does Male and Female Brain Age in the Same Way?. Frontiers in Neurology, 2021. DOI: 10.3389/fneur.2021.645729. PMID: 34163419.
• Yilmaz 2025. Brain age gap reduction following exercise mirrors clinical improvements in schizophrenia spectrum disorders. NeuroImage: Clinical, 2025. DOI: 10.1016/j.nicl.2025.103881. PMID: 41067091.
• Teipel 2024. Brain age in genetic and idiopathic Parkinson's disease. Brain Communications, 2024. DOI: 10.1093/braincomms/fcae382. PMID: 39713239.
• Jawinski 2025. Genome-wide analysis of brain age identifies 59 associated loci and unveils relationships with mental and physical health. Nature Aging, 2025. DOI: 10.1038/s43587-025-00962-7. PMID: 41044200.
• Dartora 2024. A deep learning model for brain age prediction using minimally preprocessed T1w images as input. Frontiers in Aging Neuroscience, 2024. DOI: 10.3389/fnagi.2023.1303036. PMID: 38259636.
• Ull 2025. Developmental Brain Age Estimation From MRI Data: A Systematic Review of Deep Learning Approaches and Open Datasets. Journal of Magnetic Resonance Imaging, 2025. DOI: 10.1002/jmri.70180. PMID: 41414873.
• Park 2026. Associations between contralesional neuroplasticity and motor impairment through deep learning-derived MRI regional brain age in chronic stroke (ENIGMA): a multicohort, retrospective, observational study. The Lancet. Digital health, 2026. DOI: 10.1016/j.landig.2025.100942. PMID: 41577565.
• Heffernan 2025. Association between low‐frequency oscillations in blood pressure variability and brain age derived from neuroimaging. Alzheimer's & Dementia, 2025. DOI: 10.1002/alz.70833. PMID: 41126772.
• Stefaniak 2024. Brain age gap, dementia risk factors and cognition in middle age. Brain Communications, 2024. DOI: 10.1093/braincomms/fcae392. PMID: 39605972.
• Dijsselhof 2023. The value of arterial spin labelling perfusion MRI in brain age prediction. Human Brain Mapping, 2023. DOI: 10.1002/hbm.26242. PMID: 36852443.
• Haudry 2025. Impact of meditation on brain age derived from multimodal neuroimaging in experts and older adults from a randomized trial. Scientific Reports, 2025. DOI: 10.1038/s41598-025-21490-9. PMID: 41152396.
• Valdes-Hernandez 2023. Toward MR protocol-agnostic, unbiased brain age predicted from clinical-grade MRIs. Scientific Reports, 2023. DOI: 10.1038/s41598-023-47021-y. PMID: 37950024.
• Dragendorf 2024. Quantitative assessment of neurodevelopmental maturation: a comprehensive systematic literature review of artificial intelligence-based brain age prediction in pediatric populations. Frontiers in Neuroinformatics, 2024. DOI: 10.3389/fninf.2024.1496143. PMID: 39601012.
• Derboghossian 2024. ASSOCIATIONS BETWEEN CARDIORESPIRATORY FITNESS, BRAIN AGE, AND NEURODEGENERATION AMONG OLDER ADULTS. Innovation in Aging, 2024. DOI: 10.1093/geroni/igae098.2304.
• Coetzee 2025. Predicting brain age for veterans with traumatic brain injuries and healthy controls: an exploratory analysis. Frontiers in Aging Neuroscience, 2025. DOI: 10.3389/fnagi.2025.1472207. PMID: 40443792.
• Li 2024. Decoding MRI-informed brain age using mutual information. Insights into Imaging, 2024. DOI: 10.1186/s13244-024-01791-9. PMID: 39186199.
• Ly 2024. Longitudinal accelerated brain age in mild cognitive impairment and Alzheimer’s disease. Frontiers in Aging Neuroscience, 2024. DOI: 10.3389/fnagi.2024.1433426. PMID: 39503045.
• Mouches 2022. An exploratory causal analysis of the relationships between the brain age gap and cardiovascular risk factors. Frontiers in Aging Neuroscience, 2022. DOI: 10.3389/fnagi.2022.941864. PMID: 36072481.
• Plini 2025. Meditation Linked to Enhanced MRI Signal Intensity in the Pineal Gland and Reduced Predicted Brain Age. Journal of Pineal Research, 2025. DOI: 10.1111/jpi.70033. PMID: 39940075.
• Hendrikse 2025. Increased Brain Age Among Psychiatrically Healthy Adults Exposed to Childhood Trauma. Brain and Behavior, 2025. DOI: 10.1002/brb3.70450. PMID: 40170519.
• Hu 2025. Lifespan brain age prediction based on multiple EEG oscillatory features and sparse group lasso. Frontiers in Aging Neuroscience, 2025. DOI: 10.3389/fnagi.2025.1559067. PMID: 40766176.
• Claros-Olivares 2024. MRI-based whole-brain elastography and volumetric measurements to predict brain age. Biology Methods & Protocols, 2024. DOI: 10.1093/biomethods/bpae086. PMID: 39902188.
• Motaghi 2025. Investigating the Association of Frailty Score and Diabetes with Relative Brain Age: Insights from the UK Biobank. Alzheimer's & Dementia, 2025. DOI: 10.1002/alz70856_103010.
• Cavailles 2025. Sleep Patterns in Midlife and Brain Age. Alzheimer's & Dementia, 2025. DOI: 10.1002/alz.085643.
• Wang 2025. Plasma‐based Brain Age as a Biomarker for Cognitive Health and Risk of Brain‐Related Diseases. Alzheimer's & Dementia, 2025. DOI: 10.1002/alz70856_103849.
• Hanson 2024. Examining the reliability of brain age algorithms under varying degrees of participant motion. Brain Informatics, 2024. DOI: 10.1186/s40708-024-00223-0. PMID: 38573551.
• Aghaei 2024. Brain age gap estimation using attention-based ResNet method for Alzheimer’s disease detection. Brain Informatics, 2024. DOI: 10.1186/s40708-024-00230-1. PMID: 38833039.
• Pang 2024. Predicting brain age using Tri-UNet and various MRI scale features. Scientific Reports, 2024. DOI: 10.1038/s41598-024-63998-6. PMID: 38877107.
• Ahmadi 2025. Advanced brain age prediction using 3D convolutional neural network on structural MRI. Alzheimer's & Dementia, 2025. DOI: 10.1002/alz.089776.
• Casanova 2024. A PROTEOMICS-BASED MEASURE OF ACCELERATING AGING IS CORRELATED WITH THE BRAIN AGE GAP IN THE ARIC STUDY. Innovation in Aging, 2024. DOI: 10.1093/geroni/igae098.2303.
• Dunk 2025. The association between a pro‐inflammatory diet and machine learning‐based brain age in middle‐aged and older adults: Findings from the UK Biobank. Alzheimer's & Dementia, 2025. DOI: 10.1002/alz.086979.
• Kim 2025. Association between shift work and brain age gap: a neuroimaging study using MRI-based brain age prediction algorithms. Frontiers in Aging Neuroscience, 2025. DOI: 10.3389/fnagi.2025.1650497. PMID: 40951919.
• Tavakoli 2025. Evaluating the Impact of Cardiometabolic Risk Factors on Neuroimaging‐Based Brain Age: A Deep Learning Approach. Alzheimer's & Dementia, 2025. DOI: 10.1002/alz.095769.
• Pallapothu 2025. Association between cardiovascular disease risk, regional brain age gap, and cognition in healthy adults. Frontiers in Aging Neuroscience, 2025. DOI: 10.3389/fnagi.2025.1611847. PMID: 41049536.
• Meysami 2025. White Matter Hyperintensities on Brain MRI are Related to Brain Atrophy and Accelerated Brain Age. Alzheimer's & Dementia, 2025. DOI: 10.1002/alz70862_110308.
• Roman 2025. The Impact of Brain Age versus Chronological Age on Cognitive Fatigue: Novel Metrics and New Insights. Innovation in Aging, 2025. DOI: 10.1093/geroni/igaf122.4213.
• Meysami 2026. White Matter Hyperintensities on Brain MRI are Related to Brain Atrophy and Accelerated Brain Age. Alzheimer's & Dementia, 2026. DOI: 10.1002/alz70856_106425.
• Jonemo 2023. Efficient Brain Age Prediction from 3D MRI Volumes Using 2D Projections. Brain Sciences, 2023. DOI: 10.3390/brainsci13091329. PMID: 37759930.
• Wang 2021. Predicting brain age during typical and atypical development based on structural and functional neuroimaging. Human Brain Mapping, 2021. DOI: 10.1002/hbm.25660. PMID: 34520078.
• Kou 2024. PROTEOMIC BRAIN AGE GAP, DEMENTIA RISK, AND BRAIN VOLUME MEASUREMENTS. Innovation in Aging, 2024. DOI: 10.1093/geroni/igae098.3470.
• Toraih 2025. Brain Age Acceleration on MRI Due to Poor Sleep: Associations, Mechanisms, and Clinical Implications. Brain Sciences, 2025. DOI: 10.3390/brainsci15121325. PMID: 41440121.
• Yu 2025. Chronic Medical Conditions and Dementia Risk: Brain Age Models for Quantifying Impact and Understanding Mechanisms. Alzheimer's & Dementia, 2025. DOI: 10.1002/alz.093829.
• Yu 2025b. Chronic Medical Conditions and Dementia Risk: Brain Age Models for Quantifying Impact and Understanding Mechanisms. Alzheimer's & Dementia, 2025. DOI: 10.1002/alz.089382.
• Rajabli 2025. Sex Differences in Brain Age Gap Estimation Across Alzheimer's Disease Diagnostic Groups. Alzheimer's & Dementia, 2025. DOI: 10.1002/alz70862_110227.
• Satpathi 2025. Developing scanner change invariant brain age models for aging and dementia studies. Alzheimer's & Dementia, 2025. DOI: 10.1002/alz70856_097891.
• Rajabli 2026. Sex Differences in Brain Age Gap Estimation Across Alzheimer's Disease Diagnostic Groups. Alzheimer's & Dementia, 2026. DOI: 10.1002/alz70856_107437.
• Kim 2023. REPRODUCIBILITY OF BRAIN AGE SALIENCIES ACROSS DEEP NEURAL NETWORK ARCHITECTURES. Innovation in Aging, 2023. DOI: 10.1093/geroni/igad104.3572.
• Dorfel 2024. Multimodal brain age prediction using machine learning: combining structural MRI and 5-HT2AR PET-derived features. GeroScience, 2024. DOI: 10.1007/s11357-024-01148-6. PMID: 38668887.
• Aithal 2025. Simple fully convolutional network to estimate Brain Age. Alzheimer's & Dementia, 2025. DOI: 10.1002/alz.088019.
• Raji 2025. Higher Muscle Volume is Inversely Related to Chronological and Brain Age While Increased Visceral to Muscle Fat Ratio is Positively Related to Chronological and Brain Age. Alzheimer's & Dementia, 2025. DOI: 10.1002/alz70862_110051.
• Raji 2026. Higher Muscle Volume is Inversely Related to Chronological and Brain Age While Increased Visceral to Muscle Fat Ratio is Positively Related to Chronological and Brain Age. Alzheimer's & Dementia, 2026. DOI: 10.1002/alz70856_106692.

Background References

Methodological references cited in prose. Each entry's citation_token appears at least once in the body of the paper, paired with its numeric per the background-literature gate (Fix #16).