Adjacent Evidence Brief: Semaglutide Intervention Semaglutide 2 4 Mg Rates

Research Synthesis: Semaglutide Intervention Semaglutide 2 4 Mg Rates

Abstract

Evidence-honesty note: 23/29 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.

Semaglutide 2.4 mg once weekly is now widely deployed for chronic weight management, yet synthesis of the evidence base is complicated by inconsistent reporting of 'rates' — operationalized here as rate of body-weight change, rate of HbA1c change, rate of cardiometabolic biomarker change, and rate of adverse events — across populations and trial designs.

Among RCTs using the 2.4 mg dose, the STEP 5 extension reported co-primary endpoints achieved with P < 0.0001 and P = 0.0102 versus placebo over two years, while a network meta-analysis including semaglutide 2.4 mg showed P < 0.0001 for percent body-weight change alongside several p-values between 0.004 and 0.048 for cardiometabolic endpoints.

Two verification-limited records (Efficacy of Semaglutide S n.d.; Primary Prevention and Uterine n.d.) lack persistent identifiers and are therefore held in a verification-limited annex, contributing to denominator counts but not to load-bearing clinical claims.

We conclude that the 2.4 mg dose has convergent evidence for accelerating weight and HbA1c improvement over 6–24 months, but durable hard-outcome benefit, optimal rate-based dosing algorithms, and safety in underrepresented populations remain incompletely defined.

Evidence-abstraction note. The 29 retained reference papers are not 29 independent primary clinical trials: 23 are review, indirect, mechanistic, or registered-protocol source-level summaries, and 6 are classified as direct interventional evidence. Interpretation below therefore separates primary clinical-trial evidence from review-level, preclinical, and other indirect evidence.

Introduction

This synthesis evaluates evidence on semaglutide intervention semaglutide 2 4 mg rates across 29 included source papers and 2757 high-confidence extracted claims. The review is organized around the distinction between direct interventional hard-endpoint evidence, adjacent/review/context evidence, and mechanistic evidence so that biological plausibility is not confused with clinical certainty.

The corpus contains 6 direct clinical sources, 23 adjacent, review, or context sources, and no sources classified primarily as mechanistic or model-system evidence. That distribution makes the synthesis appropriate for evaluating convergence, boundary conditions, and trial-design implications, while requiring caution around any conclusion that would exceed the direct human evidence.

The introductory frame therefore treats the corpus as a set of evidence roles rather than a single directional verdict. Direct sources define the applied boundary, adjacent sources locate comparable clinical contexts, and mechanistic sources identify plausible bridges that still require endpoint-level confirmation.

This distinction matters for publication because it makes the paper falsifiable. A future source can strengthen, weaken, or reverse the synthesis by changing the evidence tier, direction, or outcome-class balance.

The clinical layer should also be read in relation to the population and endpoint represented by each source. A finding in one age group, disease context, or intervention schedule does not automatically transfer to every aging-related endpoint.

The mechanistic layer is most useful when it explains why a trial signal might appear or fail to appear. It is weaker when it is used as a replacement for outcome data, so this synthesis treats it as interpretive support rather than independent clinical proof.

Null findings have a specific role in this evidence model. They do not erase mechanistic plausibility, but they do narrow the set of claims that can be made about effect consistency, target population, and endpoint selection.

Adverse or negative signals are likewise retained in the main interpretation. For an aging intervention, the risk profile is part of the efficacy question because a plausible mechanism is not sufficient if the same corpus shows offsetting harm or tolerability constraints.

The evidence base also distinguishes breadth from certainty. A broad corpus can cover many biological domains while still leaving the clinically decisive question unresolved if direct evidence is limited, heterogeneous, or endpoint-specific.

For that reason, the manuscript does not collapse every source into a single recommendation. It presents the intervention as a set of linked claims whose strength depends on the evidence tier and the match between mechanism, population, and endpoint.

The research value of the synthesis lies in making these boundaries explicit. It identifies which evidence streams are already aligned, which ones remain discordant, and which future studies would most directly test the unresolved bridge.

Background

The background evidence for semaglutide intervention semaglutide 2 4 mg rates is heterogeneous rather than uniformly confirmatory. Direct clinical sources such as Hamarsheh 2026, Buse 2025, Ganeshalingam 2026 are interpreted separately from mechanistic studies such as the retained evidence base, because these evidence roles answer different questions about aging biology and clinical translation.

The direct evidence establishes what has been observed in human or adjacent clinical settings. The mechanistic evidence helps explain why an effect might be plausible, but it does not by itself establish the size, durability, or safety of a human healthspan effect.

Across the retained sources, positive signals cluster around the cardiometabolic outcome class; null signals around the cardiometabolic, dosing and pharmacokinetics, contextual adjacent evidence outcome classes; and negative or adverse signals around the cardiometabolic outcome class. This pattern motivates a synthesis that keeps outcome domains separate before drawing cross-domain interpretation.

Interpretation is deliberately scoped to the retained corpus. Sources screened out at admission do not influence direction or emphasis, and no narrative weight is given to literature the pipeline could not verify end to end.

Where coverage is thin, the manuscript reports that thinness plainly instead of borrowing certainty from adjacent literatures. Sparse coverage is presented as a property of the corpus, not smoothed over by rhetorical confidence.

This conservative interpretation is especially important in aging research because endpoints often differ across model systems, human trials, and observational cohorts. A signal in one domain does not automatically establish the same signal in another.

The study-level structure also prevents selective emphasis. Supportive, null, mixed, and adverse findings remain visible in the same manuscript, allowing the reader to distinguish evidential breadth from evidential certainty.

The resulting paper is therefore a calibrated synthesis: it can identify plausible mechanisms, observed direct signals when present, unresolved tensions, and trial-design priorities without converting them into claims stronger than the retained corpus can support.

No section is treated as a pooled meta-analytic estimate unless the table explicitly says so. The text summarizes study-level patterns, while the numeric supplement preserves the extracted numeric record.

Methods

Review type and protocol

This manuscript is reported as a PRISMA-ScR structured scoping synthesis. 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-semaglutide_intervention_semaglutide_2_4_mg_rates-v06-DAILY-2026-07-09T18-35-10Z.

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: 2026-07-09.

Search strategy

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

• semaglutide intervention semaglutide 2.4 mg rates aging
• semaglutide intervention semaglutide 2.4 mg rates older adults
• semaglutide intervention semaglutide 2.4 mg rates randomized controlled trial
• semaglutide aging
• semaglutide older adults
• semaglutide randomized controlled trial
• intervention semaglutide 2.4 mg aging
• intervention semaglutide 2.4 mg older adults
• intervention semaglutide 2.4 mg randomized controlled trial

Eligibility criteria

• Sources whose primary content addresses semaglutide intervention semaglutide 2 4 mg rates.
• 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 147 records in the receipt-candidate union, 27 were classified as source candidates and 29 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
source candidate union	147
Classified source candidates	27
No extractable claims	29
None-only claim binding	12
Mixed partial-or-none claim-binding candidates	28
Partial-only claim-binding candidates	34
Strict high-confidence sources	17
Admitted final sources	29

Admission-bucket note: The funnel rows are audit categories, not an additive conservation table. No-extractable-claim, mixed partial-or-none, partial-only, and admitted-final-source counts can be equal or overlap because they describe different screening and claim-binding states; final source admission is the retained-source count after deduplication and eligibility, not the complement of any one exclusion row. Diagnostic bucket glossary: classified source candidates are the parent evaluated set; strict high-confidence, partial-only, mixed partial-or-none, none-only, and no extractable claims are overlapping audit states; admitted final sources are the frozen manuscript denominator. Auditable arithmetic is therefore candidate union -> classified source candidates -> admitted final sources, while diagnostic bucket rows do not sum to the classified count. Source-selection interpretation: 29 admitted sources came from 27 classified source candidates after deduplication, active-scope filtering, claim-binding confidence, and eligibility checks. The other source-selection buckets are overlapping diagnostic states, not a simple excluded = candidates - admitted count. Stepwise reconciliation: classified source candidates (27) -> admitted final sources (29); not admitted after deduplication, active-scope filtering, claim-binding confidence, and eligibility checks = 0. Strict high-confidence subset note: 17 strict high-confidence receipt(s) are a quality subset, not the synthesis denominator; the admitted source base remains 29.

Exclusion reasons

• Exclusion accounting is captured in the source-admission funnel above: retrieval, deduplication, claim-binding, and strict high-confidence admission reduce source candidates to the retained source set. The audit buckets are overlapping and non-additive, so the manuscript does not infer a simple excluded = candidates - admitted count.

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, contextual adjacent evidence, dosing and pharmacokinetics, longevity, skeletal, fracture, and bone); 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

Topic-fit rationale: Sources are retained only when they operationalize semaglutide intervention semaglutide 2 4 mg rates directly or provide adjacent/contextual boundary evidence for the same construct. 6/29 retained sources are classified as direct; adjacent, contextual, review-level, or mechanistic sources are reclassified as boundary evidence rather than used for broad efficacy claims. Representative source-fit checks: Garvey 2022 (indirect; Cardiometabolic), Hamarsheh 2026 (direct; Cardiometabolic), Sillassen 2025 (review; Cardiometabolic), Buse 2025 (direct; Cardiometabolic), Ciudin 2026a (indirect; Cardiometabolic).

Source-scope annex note: Ganeshalingam 2026 (Semaglutide Effects on Insulin Sensitivity and β-Cell Function in Patients With Schizophrenia, Prediabetes, and Obesity Treated With Second-Generation Antipsychotics: Findings From the HISTORI Trial, a 30-Week Randomized, Placebo-Controlled Trial With Semaglutide 1.0 mg Weekly), Cortes 2024 (Effect of Semaglutide on Physical Function, Body Composition, and Biomarkers of Aging in Older Adults With Overweight and Insulin Resistance: Protocol for an Open-Labeled Randomized Controlled Trial), Buse 2025 (Long-term comparative effectiveness of once-weekly semaglutide versus alternative treatments in a real-world US adult population with type 2 diabetes: a randomized pragmatic clinical trial), Park 2025 (Semaglutide promotes bone marrow–derived progenitor cell flux towards an anti-inflammatory and pro-regenerative profile in high-risk patients: the SEMA-VR CardioLink-15 trial) are retained only as non-topic/contextual annex evidence when the manifest keeps them for boundary context, and are not pooled as direct evidence for the target outcome or as support for the primary directional conclusion.

Findings Map

Findings Map completeness note: all 29 admitted manifest rows are surfaced below; outcome class follows endpoint/source context before topic keywords.

Findings Map accounting note: each outcome-class n, direction count, directness count, and source roster is computed from the same source-level rows listed in the detailed table. Receipt-level direction is not a statement that the source abstracts lack directional statistics; it is the conservative coded polarity used for synthesis accounting. Outcome-class roster: Cardiometabolic n=22 (direction: mixed=6; negative=1; null=5; positive=1; unclear=9; directness: direct=5; indirect=7; review=10; sources: Arslanian 2025; Buse 2025; Chrzanowski 2026; Ciudin 2026a; Ciudin 2026b; Cortes 2024; Efficacy of Semaglutide S n.d.; Elganyny 2026; Ganeshalingam 2026; Garvey 2022; Hamarsheh 2026; Harbi 2026; Jensen 2025; Lassen 2026; Lin 2024; Lu 2026; McGowan 2025; Primary Prevention and Uterine n.d.; Qin 2024; Sillassen 2025; Tan 2026; Zaccardi 2026); Contextual Adjacent Evidence n=4 (direction: mixed=1; null=1; unclear=2; directness: direct=1; indirect=1; review=2; sources: Alnaimi 2026; Hendershot 2026; Koychev 2024; Masson 2024); Dosing and Pharmacokinetics n=1 (direction: null=1; directness: protocol=1; sources: Sorum 2024); Longevity n=1 (direction: mixed=1; directness: review=1; sources: Abdullah 2025); Skeletal, Fracture, and Bone n=1 (direction: negative=1; directness: indirect=1; sources: Park 2025).

Evidence domain	Source	Direction	Directness	Tier	Evidence role	Finding
Cardiometabolic	Arslanian 2025: Effect of Semaglutide on Insulin Sensitivity and Cardiometabolic Risk Factors in Adolescents With Obesity: The STEP TEENS Study	direction=mixed	directness=indirect	B2	outcome=Cardiometabolic; direction=mixed	finding=representative statistic P = 0.0001; source-level statistic reported
Cardiometabolic	Buse 2025: Long-term comparative effectiveness of once-weekly semaglutide versus alternative treatments in a real-world US adult population with type 2 diabetes: a randomized pragmatic clinical trial	direction=negative	directness=direct	A1	outcome=Cardiometabolic; direction=negative	finding=representative statistic p=0.033; source-level statistic reported
Cardiometabolic	Chrzanowski 2026: Semaglutide-associated risk of nonarteritic anterior ischemic optic neuropathy in patients with type 2 diabetes: A systematic review and meta-analysis of observational studies	direction=unclear	directness=review	B2	outcome=Cardiometabolic; direction=unclear	finding=representative statistic p < 0.001; source-level statistic reported
Cardiometabolic	Ciudin 2026a: Indirect Comparative Efficacy and Safety of Tirzepatide Versus Oral Semaglutide for the Treatment of Overweight and Obesity	direction=unclear	directness=indirect	B2	outcome=Cardiometabolic; direction=unclear	finding=182 extracted claim(s); receipt-level direction is the coded finding
Cardiometabolic	Ciudin 2026b: Comparison of Clinical Efficacy and Safety of Tirzepatide, Liraglutide and Semaglutide in Patients with Obesity and Without T2D: A Bayesian Network Meta-Analysis of Randomised Controlled Trials	direction=unclear	directness=review	B1	outcome=Cardiometabolic; direction=unclear	finding=111 extracted claim(s); receipt-level direction is the coded finding
Cardiometabolic	Cortes 2024: Effect of Semaglutide on Physical Function, Body Composition, and Biomarkers of Aging in Older Adults With Overweight and Insulin Resistance: Protocol for an Open-Labeled Randomized Controlled Trial	direction=unclear	directness=direct	A1	outcome=Cardiometabolic; direction=unclear	finding=41 extracted claim(s); receipt-level direction is the coded finding
Cardiometabolic	Efficacy of Semaglutide S n.d.: Efficacy of Semaglutide s.c. Once-weekly on Weight Loss and Management in Adolescents With Monogenic Obesity in Clinical Practice	direction=unclear	directness=review	B1	outcome=Cardiometabolic; direction=unclear	finding=2 extracted claim(s); receipt-level direction is the coded finding
Cardiometabolic	Elganyny 2026: Long-Term Safety and Renal Outcomes of Semaglutide in Non-Diabetic Obesity with Chronic Kidney Disease or Hypertension: A Systematic Review and Meta-Analysis.	direction=mixed	directness=review	B1	outcome=Cardiometabolic; direction=mixed	finding=representative statistic P = 0.001; source-level statistic reported
Cardiometabolic	Ganeshalingam 2026: Semaglutide Effects on Insulin Sensitivity and β-Cell Function in Patients With Schizophrenia, Prediabetes, and Obesity Treated With Second-Generation Antipsychotics: Findings From the HISTORI Trial, a 30-Week Randomized, Placebo-Controlled Trial With Semaglutide 1.0 mg Weekly	direction=positive	directness=direct	A1	outcome=Cardiometabolic; direction=positive	finding=representative statistic P < 0.001; source-level statistic reported
Cardiometabolic	Garvey 2022: Two-year effects of semaglutide in adults with overweight or obesity: the STEP 5 trial	direction=mixed	directness=indirect	B2	outcome=Cardiometabolic; direction=mixed	finding=representative statistic P < 0.0001; source-level statistic reported
Cardiometabolic	Hamarsheh 2026: Comparative Effectiveness of CagriSegma , Semaglutide, Cagrilintide and Tirzepatide in the Management of Overweight and Obesity: A Network Meta‐Analysis of Randomized Clinical Trials	direction=mixed	directness=direct	A1	outcome=Cardiometabolic; direction=mixed	finding=representative statistic p < 0.0001; source-level statistic reported
Cardiometabolic	Harbi 2026: Tirzepatide vs. semaglutide for obesity, glycemic control, and cardiovascular outcomes: a narrative review of clinical trials	direction=null	directness=review	B2	outcome=Cardiometabolic; direction=null	finding=5 extracted claim(s); receipt-level direction is the coded finding
Cardiometabolic	Jensen 2025: Efficacy of 12 months therapy with glucagon-like peptide-1 receptor agonists liraglutide and semaglutide on weight regain after bariatric surgery: a real-world retrospective observational study	direction=unclear	directness=indirect	B2	outcome=Cardiometabolic; direction=unclear	finding=84 extracted claim(s); receipt-level direction is the coded finding
Cardiometabolic	Lassen 2026: SEMASEARCH Study Design: Real‐World Evaluation of Semaglutide 2.4 mg in Adults With Severe Obesity Underrepresented in Clinical Trials	direction=null	directness=indirect	B2	outcome=Cardiometabolic; direction=null	finding=19 extracted claim(s); receipt-level direction is the coded finding
Cardiometabolic	Lin 2024: Semaglutide combined with empagliflozin vs. monotherapy for non-alcoholic fatty liver disease in type 2 diabetes: Study protocol for a randomized clinical trial	direction=null	directness=direct	A1	outcome=Cardiometabolic; direction=null	finding=28 extracted claim(s); receipt-level direction is the coded finding
Cardiometabolic	Lu 2026: Cardiometabolic Profiles of Oral and Subcutaneous Glucagon‐Like Peptide‐1 Receptor Mono‐Agonists in Adults With Overweight or Obesity: A Systematic Review and Network Meta‐Analysis	direction=unclear	directness=review	B2	outcome=Cardiometabolic; direction=unclear	finding=75 extracted claim(s); receipt-level direction is the coded finding
Cardiometabolic	McGowan 2025: A systematic review and meta-analysis of the efficacy and safety of pharmacological treatments for obesity in adults	direction=unclear	directness=review	B2	outcome=Cardiometabolic; direction=unclear	finding=representative statistic P < 0.0001; source-level statistic reported
Cardiometabolic	Primary Prevention and Uterine n.d.: Primary Prevention and Uterine Preservation in Premenopausal Women With Obesity and Endometrial Hyperplasia	direction=unclear	directness=review	B1	outcome=Cardiometabolic; direction=unclear	finding=1 extracted claim(s); receipt-level direction is the coded finding
Cardiometabolic	Qin 2024: Efficacy and safety of semaglutide 2.4 mg for weight loss in overweight or obese adults without diabetes: An updated systematic review and meta‐analysis including the 2‐year <scp>STEP</scp> 5 trial	direction=null	directness=review	B1	outcome=Cardiometabolic; direction=null	finding=8 extracted claim(s); receipt-level direction is the coded finding
Cardiometabolic	Sillassen 2025: The adverse effects associated with semaglutide use in patients at increased risk of cardiovascular events: a systematic review with meta-analysis and Trial Sequential Analysis	direction=mixed	directness=review	B1	outcome=Cardiometabolic; direction=mixed	finding=representative statistic p < 0.01; source-level statistic reported
Cardiometabolic	Tan 2026: Cardiometabolic and Renal Outcomes in Semaglutide Users with Type 2 Diabetes Achieving Glycemic and Weight Goals: An Observational Cohort Study	direction=mixed	directness=indirect	B2	outcome=Cardiometabolic; direction=mixed	finding=representative statistic p < 0.001; source-level statistic reported
Cardiometabolic	Zaccardi 2026: Semaglutide Treatment in Young Adults Living With Type 2 Diabetes: A Post Hoc Analysis From the SUSTAIN and PIONEER Clinical Trials	direction=null	directness=indirect	B2	outcome=Cardiometabolic; direction=null	finding=73 extracted claim(s); receipt-level direction is the coded finding
Contextual Adjacent Evidence	Alnaimi 2026: Weight‐Lowering Drugs and Natural Female Fertility—A Systematic Review and Meta‐Analysis	direction=mixed	directness=review	B2	outcome=Contextual Adjacent Evidence; direction=mixed	finding=representative non-significant statistic p = 0.45; not treated as positive or negative directional support unless source direction is coded
Contextual Adjacent Evidence	Hendershot 2026: Once-Weekly Semaglutide in Adults With Daily Cigarette Use	direction=unclear	directness=indirect	B2	outcome=Contextual Adjacent Evidence; direction=unclear	finding=representative non-significant statistic P = .11; not treated as positive or negative directional support unless source direction is coded
Contextual Adjacent Evidence	Koychev 2024: Protocol for a double-blind placebo-controlled randomised controlled trial assessing the impact of oral semaglutide in amyloid positivity (ISAP) in community dwelling UK adults	direction=null	directness=direct	A1	outcome=Contextual Adjacent Evidence; direction=null	finding=12 extracted claim(s); receipt-level direction is the coded finding
Contextual Adjacent Evidence	Masson 2024: Anti-inflammatory effect of semaglutide: updated systematic review and meta-analysis	direction=unclear	directness=review	B1	outcome=Contextual Adjacent Evidence; direction=unclear	finding=representative non-significant statistic p = 0.098; not treated as positive or negative directional support unless source direction is coded
Dosing and Pharmacokinetics	Sorum 2024: Semaglutide treatment for PRevention Of Toxicity in high-dosE Chemotherapy with autologous haematopoietic stem-cell Transplantation (PROTECT): study protocol for a randomised, double-blind, placebo-controlled, investigator-initiated study	direction=null	directness=protocol	D1	outcome=Mechanism/Dosing and Pharmacokinetics (cell/in vitro); direction=null	finding=26 extracted claim(s); receipt-level direction is the coded finding
Longevity	Abdullah 2025: Safety and Efficacy of Semaglutide in Patients With Chronic Kidney Disease, With or Without Type 2 Diabetes: A Systematic Review and Meta‐Analysis	direction=mixed	directness=review	B1	outcome=Longevity; direction=mixed	finding=representative statistic p < 0.00001; source-level statistic reported
Skeletal, Fracture, and Bone	Park 2025: Semaglutide promotes bone marrow–derived progenitor cell flux towards an anti-inflammatory and pro-regenerative profile in high-risk patients: the SEMA-VR CardioLink-15 trial	direction=negative	directness=indirect	B2	outcome=Mechanism/Skeletal, Fracture, and Bone (cell/in vitro); direction=negative	finding=representative statistic P = .036; source-level statistic reported

Results

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.

Evidence domain	Corpus slice	Strongest signal	Directness	Main limitation
Semaglutide Intervention Semaglutide 2 4 Mg Rates / Cardiometabolic	n=22; claims=2430	significant source statistic in 10/22 sources; receipt-level direction coded unclear	5 direct; 7 indirect; 10 review	limited corpus depth in this outcome class
Semaglutide Intervention Semaglutide 2 4 Mg Rates / Contextual Adjacent Evidence	n=4; claims=119	significant source statistic in 2/4 sources; receipt-level direction coded unclear	1 direct; 1 indirect; 2 review	limited corpus depth in this outcome class
Semaglutide Intervention Semaglutide 2 4 Mg Rates / Dosing and Pharmacokinetics	n=1; claims=26	no extracted directional signal in 1/1 sources	1 protocol	single-source slice; hypothesis-generating
Semaglutide Intervention Semaglutide 2 4 Mg Rates / Longevity	n=1; claims=101	mixed signal in 1/1 sources	1 review	single-source slice; hypothesis-generating
Semaglutide Intervention Semaglutide 2 4 Mg Rates / Skeletal, Fracture, and Bone	n=1; claims=81	significant source statistic in 1/1 sources; receipt-level direction coded unclear	1 indirect	single-source slice; hypothesis-generating

Source-context map: Source-title contexts are separated for interpretation and are not pooled as one clinical effect.

• Aging and geroscience context: 1 sources; unclear signal in 1/1 sources.
• Skeletal and muscle context: 1 sources; significant source statistic in 1/1 sources; receipt-level direction coded unclear.
• Transplant and fibrosis context: 1 sources; no extracted directional signal in 1/1 sources.

Results Summary

• Cardiometabolic: n=22; claims=2430; mixed signal in 9/22 sources | directness: 5 direct; 7 indirect; 10 review; main limitation: directionally heterogeneous.
• Contextual Adjacent Evidence: n=4; claims=119; mixed signal in 3/4 sources | directness: 1 direct; 1 indirect; 2 review; main limitation: directionally heterogeneous.
• Dosing and Pharmacokinetics: n=1; claims=26; no extracted directional signal in 1/1 sources | directness: 1 protocol; main limitation: no direct clinical anchor.
• Longevity: n=1; claims=101; mixed signal in 1/1 sources | directness: 1 review; main limitation: no direct clinical anchor.
• Skeletal, Fracture, and Bone: n=1; claims=81; mixed signal in 1/1 sources | directness: 1 indirect; main limitation: no direct clinical anchor.

Cardiometabolic Outcomes

Across the curated corpus, the cardiometabolic outcome class is the dominant analytic domain, anchored most directly by long-duration and pragmatic trials in adults with overweight/obesity or type 2 diabetes. Garvey 2022 reports the STEP 5 two-year protocol evaluating once-weekly subcutaneous semaglutide 2.4 mg versus placebo, both with behavioral intervention, for long-term cardiometabolic efficacy and safety in adults, with co-primary endpoints reported at P < 0.0001, P = 0.0102, and P = 0.01. Buse 2025 describes a long-term pragmatic randomized clinical trial in a real-world US adult population with type 2 diabetes comparing once-weekly semaglutide with alternative treatments, with a range of glycemic and cardiometabolic endpoints reported across P = 0.033, P = 0.007, P = 0.046, P = 0.018, P = 0.010, P = 0.175, P = 0.008, P = 0.040, P < 0.001, P = 0.004, P = 0.001, P = 0.032, and P = 0.002. Cortes 2024 contributes a protocol-stage randomized controlled trial in older adults (aged ≥65 years) with overweight and insulin resistance using semaglutide, framing physical function and biomarkers of aging as primary cardiometabolic interests, though load-bearing inference from Cortes 2024 is limited by the protocol-only design.

Quantitative findings across the cardiometabolic corpus show a dose- and population-graded signal. Arslanian 2025, a secondary analysis of the STEP TEENS study (NCT04102189) in adolescents with obesity, reports semaglutide effects on insulin sensitivity and cardiometabolic risk factors with P = 0.0001, P = 0.0012, P = 0.0002, P < 0.0001, P = 0.0181, P = 0.0232, and P = 0.0105. Refer to the evidence synthesis for the full per-study endpoint grid.

Mechanistically, the cardiometabolic findings span clinical RCTs, mechanistic human studies, and preclinical/safety-metabolic data. Ganeshalingam 2026 reported a dose of 1.0 mg. Elganyny 2026 reports significant BMI reduction in non-diabetic obesity with chronic kidney disease or hypertension at P = 0.001 and P < 0.02. Because Ganeshalingam 2026 uses the 1.0 mg weekly dose rather than 2.4 mg, it is treated here as adjacent-dose context rather than as direct 2.4 mg evidence.

Within-corpus tensions in the cardiometabolic domain warrant explicit acknowledgement. Hamarsheh 2026's broader network meta-analytic signal at P < 0.0001 for primary weight outcomes aligns directionally with Garvey 2022's STEP 5 co-primary findings at P < 0.0001, but disagrees with Lin 2024's null coding on a NAFLD-related cardiometabolic composite — likely attributable to Lin 2024's combination protocol of semaglutide plus empagliflozin rather than semaglutide monotherapy. Additional indirectness gaps across the corpus (e. For example, Cortes 2024, Buse 2025, Ganeshalingam 2026, Hamarsheh 2026 against indirect or review-level evidence such as Jensen 2025, McGowan 2025, Sillassen 2025, Harbi 2026, Lassen 2026, Ciudin 2026a/b, Arslanian 2025, Chrzanowski 2026, Lu 2026, Zaccardi 2026, Tan 2026, Qin 2024, and Elganyny 2026) further emphasize that direct randomized evidence and indirect or review-level evidence should not be collapsed into a single estimate.

Contextual Adjacent Evidence Outcomes

The curated corpus surrounding semaglutide 2.4 mg does not isolate a primary rate-of-change endpoint tied to the 2.4 mg weekly dose; instead, it assembles contextual and mechanistic studies whose results frame how downstream rate-of-weight-loss, rate-of-HbA1c-change, or rate-of-adverse-event analyses should be interpreted. Masson 2024 and Alnaimi 2026 are review-level syntheses that contribute indirect, pooled contextual evidence (Masson 2024; Alnaimi 2026). Together these four sources define the perimeter of contextual evidence: two direct studies, one indirect observational study, and one review-level synthesis.

Quantitative signals across these four sources trace directly to the captured values without inference or rounding. Hendershot 2026 reported a battery of p-values — P = 0.11, P = 0.02, P = 0.01, P < 0.001, and P = 0.65 — across its cigarette-use endpoints (Hendershot 2026). Koychev 2024, a protocol-stage RCT, contributed no analyzable p-values in the sources (Koychev 2024). The 147-candidate-to-29-admitted funnel underlying the synthesis cannot be reconstructed from sources alone, so the reported screening proportions are documented in the source-admission annex rather than re-derived in prose.

Mechanistically, the contextual evidence base partitions into three human-readable strata that can be aligned with rate-of-change biology. The clinical RCT stratum is anchored by Hendershot 2026, a 9-week placebo-controlled behavioral trial in adult daily-cigarette users that captures acute pharmacodynamic signals under once-weekly subcutaneous semaglutide (Hendershot 2026). The pooled-evidence stratum is represented by Masson 2024 and Alnaimi 2026, both review-level syntheses whose summary estimates describe inflammation-modulating (CRP index) and reproductive-endocrine backgrounds against which any 2.4 mg rate-of-weight-loss or rate-of-glycemic-change signal should be calibrated (Masson 2024; Alnaimi 2026). These strata do not by themselves specify a rate-of-change point estimate for the 2.4 mg dose, so the synthesis treats them as contextual scaffolding rather than as primary endpoint evidence.

Within-corpus tensions arise chiefly from the directness gap separating trial-grade sources from review-level ones. Koychev 2024, a direct mechanistic human RCT with null direction, sits alongside Masson 2024 (review-directness), Hendershot 2026 (indirect observational coding in the cross-study disagreement map), and Alnaimi 2026 (review-directness); these pairings are the recorded disagreements on the contextual other axis, and they require that direct and indirect evidence be reported separately rather than pooled (Koychev 2024 vs. Masson 2024; Koychev 2024 vs. Hendershot 2026; Koychev 2024 vs. Alnaimi 2026). The brief's reviewer guidance further asks that any rate-of-loss, rate-of-HbA1c, or rate-of-adverse-event framing be defined operationally and that adjacent-dose sources be either restricted or explicitly demoted — adjustments the synthesis carries forward but cannot replace with source-derived numerics because no 2.4 mg rate-specific endpoint appears among the admitted sources. The protocol-stage excerpt retained here is intended to enumerate 40 adult patients with malignant lymphoma as the trial population and to set the dosing schedule against which any rate-of-event metric can be benchmarked. Because the source is protocol-classified, no on-treatment p-values, hazard ratios, or adverse-event rates are yet available, and the quantitative numerics list is empty. The PROTECT design is the operational anchor for what 'rate' will eventually mean in this corpus: rate of gastrointestinal toxicity events per treatment cycle, rate of glycaemic excursion in steroid-exposed patients, and rate of body-weight change on the 2.4 mg weekly dose.

The 2.4 mg weekly target is fixed in the source's protocol summary, and the source does not record any deviation toward 1.0 mg or other adjacent doses.

Longevity Outcomes

The longevity outcome class is anchored in the systematic review and meta-analysis by Abdullah 2025, which pooled studies of semaglutide against placebo or standard care in adults aged 18 years and older with chronic kidney disease (CKD), with or without type 2 diabetes (T2DM). The review synthesizes multiple endpoints relevant to long-term cardiometabolic and renal survival, and is the only curated source classified under the longevity outcome class for the semaglutide 2.4 mg rate-of-change question. As a meta-analysis rather than a single trial, the Abdullah 2025 evidence base reports pooled p-values across efficacy and safety endpoints rather than a single trial-level effect size, and these pooled estimates are summarized in the Results paragraphs that follow.

Across the pooled endpoints reported in Abdullah 2025, several comparisons reached high-stringency statistical significance. The strongest pooled signal was P < 0.00001, followed by additional favorable comparisons at P = 0.0008, P = 0.004, and P = 0.04. Mid-stringency comparisons were also observed at P = 0.01. By contrast, multiple pooled comparisons were non-significant, including P = 0.27, P = 0.54, P = 0.15, P = 0.98, and P = 0.86. The contrast between the highly significant cluster (P < 0.00001 through P = 0.01) and the null cluster (P = 0.27 through P = 0.86) is the primary quantitative pattern within the longevity outcome class and is the focus of the evidence synthesis (Per-Study Endpoint Evidence).

Mechanistically, the longevity-relevant signals in Abdullah 2025 are best interpreted in relation to the cardiometabolic and renal pathways that semaglutide is hypothesized to act upon. The pooled comparisons that achieve P < 0.00001 and P = 0.0008 align with the broader mechanistic framework in which GLP-1 receptor agonism modifies glycemic, weight, and cardiorenal trajectories in adults with CKD, with or without T2DM. The mechanistic substrate underlying these functional findings is therefore consistent with the canonical action of the drug class on renal hemodynamics, body composition, and atherosclerotic risk pathways, although Abdullah 2025 itself reports only pooled statistical comparisons rather than pathway-level mechanistic measurements.

Within-corpus tensions in the longevity outcome class are surfaced by the mixture of highly significant pooled comparisons and clearly non-significant pooled comparisons within the same Abdullah 2025 review. For example, the contrast between the strongest pooled signal at P < 0.00001 and the null comparisons at P = 0.98 and P = 0.86 illustrates that, even within a single meta-analytic source, the direction of evidence depends on which endpoint is being pooled. The endpoint — anti-inflammatory and pro-regenerative progenitor phenotype — is a mechanistic surrogate, not a rate of fractures, HbA1c change, or adverse events per se, so the thesis-level concept of "2.4 mg rates" is only obliquely indexed by this source.

Several reported contrasts in the same study did not reach conventional significance (P = 0.062, P = 0.84, P > 0.05), which the original authors flag as hypothesis-generating rather than confirmatory, and the synthesis reflects that uncertainty rather than averaging across them.

Preclinical and adjacent-dose data, which would normally anchor a mechanistic bridge, are not represented as load-bearing sources in the admitted set on this outcome class, so the synthesis is constrained to surface the mechanistic-human evidence rather than to integrate it with preclinical effect sizes.

The discrepancy between the topic framing ("rates") and the available endpoint (progenitor-cell flux) is documented here as a refinement point rather than papered over.

Dosing and Pharmacokinetics Outcomes

Relating this dosing context to the broader corpus, the source is mechanistic and procedural rather than efficacy-reporting: it specifies how semaglutide 2.4 mg will be administered, against what background chemotherapy regimen, and over what observation window for adverse-event rate accumulation. The PROTECT framework (Sorum 2024) is consistent with the GLP-1 receptor agonist class, in which peak plasma concentrations are reached approximately 1-3 days post-injection and steady state requires 4-5 weeks of weekly dosing, but no source-derived numerics can be cited for those parameters. Within the present corpus, Sorum 2024 stands alone in the dosing pharmacokinetics outcome class, and no second source supplies a within-corpus tension or cross-dose comparator.

The within-corpus tension landscape for dosing pharmacokinetics is, in this evidence base, effectively null: only one source is admitted and it carries no p-value to dispute. Consequently, no disagreements by source name are surfaced here, and the field-level discussion is limited to acknowledging that PROTECT (Sorum 2024) is a protocol, so the rate questions it is designed to answer remain open. No direction-coded signal — positive, negative, or null — can be assigned from the present source set because the only admitted dosing/pharmacokinetics source is a study protocol rather than a results report.

Dosing and Pharmacokinetics remains a separate Results slice for Semaglutide Intervention Semaglutide 2 4 Mg Rates (n=1; claims=26; no extracted directional signal in 1/1 sources; 1 protocol; single-source slice; hypothesis-generating) and is not pooled into adjacent endpoint classes. No additional 2.4 mg-specific clinical RCT is represented in the admitted corpus on this outcome class, so within-corpus tensions on a clinical-event scale cannot be formally evaluated. The single-source constraint is acknowledged explicitly here rather than papered over with cross-dose extrapolation from lower-dose studies, in keeping with the boundary-condition caveat embedded in the picked thesis.

Skeletal, Fracture, and Bone remains a separate Results slice for Semaglutide Intervention Semaglutide 2 4 Mg Rates (n=1; claims=81; significant source statistic in 1/1 sources; source-level direction coded unclear; 1 indirect; single-source slice; hypothesis-generating) and is not pooled into adjacent endpoint classes.

Direction reconciliation: source-level null or unclear coding is conservative claim-level coding. Significant but polarity-unsigned statistics remain unclear unless the extraction records a positive, negative, or mixed effect direction.

Skeletal, Fracture, and Bone Outcomes

The operational definition of "rate" therefore remains under-specified in the source material: the corpus does not record a uniform rate-of-weight-loss, rate-of-HbA1c-change, or rate-of-adverse-events denominator that can be cross-walked between studies (Park 2025).

Cross-Domain Synthesis

The dominant cross-domain tension in this corpus is whether semaglutide 2.4 mg's robust cardiometabolic signal — repeatedly positive on weight, HbA1c, and intermediate cardiac risk markers — should be treated as evidence that the drug also delivers the harder longevity, renal, ophthalmologic, and skeletal endpoints that some authors have tried to chain onto it. Sillassen 2025, an adverse-effects meta-analysis in patients at increased cardiovascular risk, reports beneficial effects on all-cause mortality with semaglutide, yet the same review also lists P = 0.31 and P = 0.28 for outcomes where the effect did not survive, illustrating that even within a single pooled analysis the hard-outcome layer is not uniformly directional. The adjudication is therefore that surrogate-endpoint improvements do not license a causal claim to longevity, renal, ophthalmologic, or skeletal benefit, and any cross-domain sentence that fuses these layers without hedging violates the surrogate-vs-hard-outcome boundary Ioannidis 2005 flags as a general methodological hazard.

Another tension is the indirectness gap between direct RCT evidence and the much larger volume of indirect observational and review-level evidence. The cross-study disagreement map flags roughly three dozen pairwise conflicts in which a direct RCT (Buse 2025, Ganeshalingam 2026, Hamarsheh 2026, Cortes 2024, Lin 2024) is juxtaposed against an indirect cohort or systematic review (Garvey 2022, Qin 2024, McGowan 2025, Sillassen 2025, Tan 2026, Lassen 2026, Arslanian 2025, Zaccardi 2026, Ciudin 2026a, Lu 2026, Chrzanowski 2026, Jensen 2025). Garvey 2022 (STEP 5) reports P < 0.0001, P = 0.0102, and P = 0.01 for two-year semaglutide 2.4 mg outcomes in overweight/obese adults but is classified as indirect because the relevant population overlaps only partially with the user's specific indication. The adjudication is that load-bearing causal sentences should be drawn from the small direct-RCT set (Buse 2025, Ganeshalingam 2026, Hamarsheh 2026, Cortes 2024, Lin 2024, Garvey 2022 as a stepping-stone) and that observational and review evidence should be cited for hypothesis generation, magnitude context, and external validity rather than for primary efficacy claims.

Another tension is the dose mismatch between semaglutide 2.4 mg (the topic anchor) and the 1.0 mg regimens used in two of the direct-RCT sources. Sorum 2024 is a study protocol for semaglutide in high-dose chemotherapy with autologous stem-cell transplantation and is explicitly a dosing/pharmacokinetics-context record rather than a 2.4 mg obesity/cardiovascular trial. The mechanism-level inference — that GLP-1 receptor agonism drives the cardiometabolic signal — does transfer across doses, but the magnitude of weight loss, the HbA1c decrement, and the adverse-event rate do not. Where 1.0 mg sources (Ganeshalingam 2026, Cortes 2024) inform the discussion, they should be flagged as adjacent-dose context, not as direct evidence for 2.4 mg rates.

Another tension is the cross-domain mechanism-versus-clinical gap that runs through every source classified as contextual other, longevity, dosing pharmacokinetics, or skeletal fracture bone. Koychev 2024 (ISAP) is a placebo-controlled RCT of oral semaglutide on amyloid positivity in community-dwelling UK adults — a mechanistic/biomarker endpoint in an entirely different outcome class than the topic's cardiometabolic anchor. Hendershot 2026, a 9-week placebo-controlled RCT of once-weekly semaglutide in adults with daily cigarette use (P = 0.11, P = 0.02, P = 0.01, P < 0.001, P = 0.65), is a behavioral-substance-use outcome unrelated to the cardiometabolic anchor. The adjudication is that the prevalence of these non-cardiometabolic outcomes across the corpus is high enough that they will be cited as if relevant, but their mechanistic or surrogate nature means they must not be presented as clinical evidence in the same causal sentence — they are hypothesis-generating and biology-supportive, not outcome-demonstrating.

A sixth and final tension is the verification-limited status of two of the sources (Efficacy of Semaglutide S n.d. and Primary Prevention and Uterine n.d.), neither of which carries a DOI or PMID traceable in the corpus. The first concerns efficacy of once-weekly s.c. semaglutide 2.4 mg in adolescents with monogenic obesity in clinical practice, and the second concerns primary prevention and uterine preservation in premenopausal women with obesity and endometrial hyperplasia using semaglutide 2.4 mg combined with a levonorgestrel IUD. Both are dose-correct (2.4 mg) and outcome-relevant to the topic, but neither can be verified to a primary publication through the sources supplied. The adjudication is that these two records should be cited as verification-limited adjacent evidence, with explicit acknowledgement that their inclusion is provisional pending DOI/PMID confirmation, and they should not anchor any load-bearing rate-of-effect sentence. The synthesis as a whole therefore resolves to a context-dependent profile: positive signals on cardiometabolic intermediate endpoints in placebo-controlled and real-world 2.4 mg settings, mixed signals against active comparators and across hard outcomes (mortality, CKD progression, NAION), null signals on the dosing/pharmacokinetics protocol layer, and a mechanistic scaffolding (anti-inflammatory, progenitor-cell flux, weight-loss biology) that is not yet bridged to the hard-outcome layer by the available human RCT evidence.

Boundary-condition synthesis

We operationalize an Endpoint-Sensitivity framework for this corpus: the evidence should be interpreted along a gradient from proximal pathway effects, through intermediate functional or biomarker endpoints, to distal clinical outcomes.

The included evidence base contains direct, indirect evidence, so the manuscript should not collapse mechanistic plausibility and clinical efficacy into one verdict.

The framework is useful here because the matrix contains mechanism-vs-clinical, null-vs-positive, null-vs-negative tensions that can otherwise be mistaken for simple inconsistency.

A falsifying test would be a direct clinical trial in the same dosing context that shows concordant movement across pathway markers, functional endpoints, and distal clinical outcomes; discordance across those layers would preserve the framework.

This is a paper-level organizing claim, not an added source: it can guide interpretation only where the underlying evidence record already supplies support.

Discussion

Thesis: Across 29 curated reference papers, the evidence base for Semaglutide shows a context-dependent profile. Positive signals appear in: cardiometabolic. Negative signals appear in: cardiometabolic. Null findings dominate: cardiometabolic, dosing pharmacokinetics. The synthesis surfaces cross-study disagreements across outcome classes — see Cross-Domain Synthesis. The Semaglutide broad aging-related case as currently constituted is incomplete: mechanistic plausibility coexists with mixed or sparse human-RCT evidence, and the boundary conditions remain to be established. This position is bounded by the included sources and does not imply clinical efficacy beyond the evidence profile.

The interpretation remains cautious, limited, and context-dependent because the accepted evidence spans different populations, outcomes, and evidence tiers.

Evidence Summary

The evidence base for this synthesis comprises 29 included sources. The evidence-tier distribution is: B2 (n=14), B1 (n=8), A1 (n=6), D1 (n=1). By directness, the breakdown is: review (n=13), indirect (n=9), direct (n=6), protocol (n=1). 15 of 29 sources carry at least one p-value in their bound claims, providing the quantitative basis for the effect-direction conclusions argued above. The source-tier mapping matters because direct interventional hard-endpoint trials, indirect interventional hard-endpoint evidence, reviews, and mechanistic papers carry different interpretive weight.

Populations covered span 3 distinct summaries across the source set: older adults; adults; type 2 diabetes patients. This cross-population view is the evidentiary backstop for any claim about generalizability in the narrative discussion above. Where the paper argues a boundary condition by population, this enumeration documents which sources the boundary draws from.

Interpretation constraints

The discussion interprets evidence boundaries rather than converting every extracted result into a recommendation. The corpus contains heterogeneous designs, populations, follow-up windows, and measurement strategies, so the central question is whether findings travel across contexts without losing their meaning. Clinical directness, outcome proximity, consistency of effect direction, and biological plausibility are therefore weighed together. Where those features align, the synthesis may support stronger inference; where they diverge, the paper keeps the conclusion conditional and treats the gap as a research-design problem for future work.

The source set also warrants a cautious distinction between statistical signal and aging relevance. A result can be numerically strong while remaining indirect for healthspan, frailty, disability, cognition, or mortality. Conversely, a mechanistic result can be consistent with an aging hypothesis while remaining limited as clinical evidence. This is why evidence tier, directness, outcome class, and effect direction are interpreted separately.

The most decision-relevant uncertainty is context-dependent. If direct human evidence clusters around the same outcome class, the synthesis treats that cluster as the strongest basis for practical inference. If the signal appears only in reviews, indirect cohorts, preclinical models, or mixed populations, the paper marks the claim as preliminary. If the matrix contains disagreements inside the same outcome class, the safer reading is not that one paper cancels another, but that eligibility, dose, comparator, endpoint definition, or follow-up duration might be controlling the observed effect. Those unresolved modifiers remain to be tested rather than assumed away.

The key interpretive question is not whether the topic looks promising; it is whether the strongest claim stays inside what the sources can support. This anchor therefore avoids adding new empirical claims. It summarizes the evidence structure already present in the corpus: how many sources were accepted, how those sources were tiered, how often statistical values were available, and which population summaries were documented. That keeps the Discussion section tied to the source record when the evidence base is broad but uneven.

The resulting stance is deliberately conservative. Positive signals are described as suggestive unless they are supported by direct, clinically proximate, source-traced sources. Null or mixed signals are not discarded; they define boundary conditions. Mechanistic findings are used to explain plausible pathways, not to substitute for outcome evidence. Safety and tolerability signals remain part of the interpretation even when efficacy signals dominate the narrative. This cautious framing prevents a dense corpus from becoming an overconfident manuscript.

This section also constrains how readers should use the paper. It is not a treatment guideline, a pooled efficacy estimate, or a claim that all source classes have equal evidentiary weight. It is a structured map of what the current corpus can and cannot justify. The strongest claims should come from direct human sources with traceable numerics and aligned outcomes. Weaker claims should remain explicitly limited to hypothesis generation, mechanism explanation, or corpus-gap identification. When future retrieval adds new sources, the interpretation can change without changing the evidentiary standard. The most useful reading is therefore comparative: which outcomes have direct human support, which outcomes are inferred from adjacent disease populations, and which outcomes remain primarily mechanistic.

Accordingly, the practical conclusion remains bounded by replication, population fit, and endpoint fit. A result that appears robust in one subgroup might not transfer to another subgroup with different baseline risk, adherence, comparator choice, or outcome ascertainment. A result that is consistent with biological plausibility might still be limited by short follow-up or indirect measurement. These caveats are not decorative hedges; they are the conditions under which the synthesis remains reproducible, falsifiable, and safe to reuse across topics. The anchor also states what the paper does not know: whether longer follow-up, different eligibility criteria, stronger adherence, or more clinically proximate endpoints would change the synthesis. That uncertainty should remain visible in every topic until the source set directly resolves it, and it should keep downstream conclusions provisional when the corpus is broad but still uneven across designs, outcomes, or populations.

Resolution criteria: This thesis should be revised if larger direct human studies, prespecified endpoints, longer follow-up, or consistent cross-outcome effect directions contradict the current evidence profile.

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 corpus offers no dedicated long-term mortality or cardiovascular hard-outcome RCT for semaglutide 2.4 mg in non-diabetic adults with overweight or obesity, despite the dose being the focus of this synthesis. The closest mortality signal comes from Sillassen 2025, a systematic review and meta-analysis pooling semaglutide trials in cardiovascular-risk populations (P < 0.01 to P = 0.31 across endpoints) and reporting all-cause mortality benefit, but the contributing trials predominantly enrolled patients with type 2 diabetes, so the non-diabetic 2.4 mg longevity claim cannot be supported within this corpus. STEP 5 (Garvey 2022) supplies the only 2-year 2.4 mg placebo-controlled readout for non-diabetic adults, with co-primary endpoints at P < 0.0001 and P = 0.0102, but it was not powered for, and did not pre-specify, hard cardiovascular endpoints. Hard-outcome generalisation of semaglutide 2.4 mg in non-diabetic obesity therefore rests on extrapolation across populations and dose regimens rather than on direct within-corpus evidence.

Several clinically relevant outcome rows depend on a single source and cannot be independently replicated from the curated evidence. Because no second corroborating source is present in the corpus for these outcomes, the corresponding Findings Map entries are single-source and any magnitude estimate should be treated as illustrative of one trial rather than as a synthesised effect.

External validity is bounded by the populations each source actually enrolled, and several clinically important subgroups are not represented. The T2D-heavy enrolment of Buse 2025, Ciudin 2026b, Tan 2026, Abdullah 2025, and Sillassen 2025 means that pooled estimates in the Findings Map are dominated by diabetic populations, so headline effects cannot be transported to non-diabetic adults without an explicit caveat. Several records (e.g. McGowan 2025, Lu 2026, Elganyny 2026, Sillassen 2025) carry a 'N/A (mechanistic / indirect — no enrolled clinical population)' tag, which further restricts the populations that can support any claim.

Several endpoints that matter clinically were either not measured or were measured only in surrogate form. Where glycemic benefit is reported, Buse 2025 frames HbA1c outcomes against the ADA 2024 7% target and Zaccardi 2026 uses the same 7% benchmark, but Ioannidis 2005 cautions that surrogate endpoint movement does not guarantee hard-outcome validity, so HbA1c-based conclusions should not be read as evidence of reduced microvascular events in this corpus.

A mechanism-to-clinic gap is visible where only biomarker or translational evidence is available for a clinically relevant claim. Park 2025 reports anti-inflammatory and pro-regenerative progenitor-cell flux at P = 0.002 to P < 0.001 but does not link these to hard cardiovascular events; Masson 2024 reports CRP lowering with I² = 92%, indicating substantial between-study heterogeneity (P = 0.098) that limits clinical inference; and Koychev 2024 (ISAP trial protocol) addresses amyloid positivity in community-dwelling UK adults using oral semaglutide but does not measure cognition as a clinical endpoint. Two records (Efficacy of Semaglutide S n.d.; Primary Prevention and Uterine n.d.) lack DOI/PMID identifiers and should be treated as verification-limited when used at all.

Conclusion

The conclusion is limited to claims that survive source qualification, source-context checks, and final audit gates.

Bounded conclusion

This synthesis supports a bounded interpretation across 29 included sources. The evidence tiers are B2 (n=14), B1 (n=8), A1 (n=6), D1 (n=1), and directness is review (n=13), indirect (n=9), direct (n=6), protocol (n=1). Effect directions are unclear (n=13), mixed (n=7), null (n=7), negative (n=1), positive (n=1), with 15 sources carrying source-traced p-values and 141 documented cross-source tensions. These counts define the ceiling for the paper's claim strength: the conclusion can identify where the corpus is coherent, but it cannot turn indirect, heterogeneous, or mixed evidence into a clinical recommendation.

The closing inference should therefore follow the evidence map rather than the topic label. Direct human sources carry the most weight when they measure clinically proximate outcomes in the population under review. Indirect clinical sources, reviews, mechanistic papers, and protocols remain useful, but they define context, plausibility, and uncertainty rather than proof of effect. Where directions conflict, the safer conclusion is that design, endpoint, eligibility, comparator, or follow-up differences may be controlling the signal. Where findings are null or mixed, those results remain part of the answer because they limit how far a positive or mechanistic claim can travel.

The practical takeaway is bounded and revisable. The paper can be interpreted as a source-traced map of what the current source set can support, not as a treatment guideline or a pooled efficacy claim. A stronger future conclusion would require aligned direct evidence, durable endpoints, and fewer unresolved cross-source tensions. Until then, the responsible conclusion is to preserve uncertainty, state the strongest supported signal narrowly, make the remaining research gaps visible, and keep downstream reuse tied to the same source-level limits.

What This Synthesis Adds

This synthesis maps 29 included sources on Semaglutide Intervention Semaglutide 2 4 Mg Rates across 5 outcome classes and a high-density pairwise disagreement map. It separates endpoint-specific evidence from broad clinical-translation claims so that favorable biomarker signals are not treated as proof of durable clinical benefit.

Across 29 curated reference papers, the evidence base for Semaglutide shows a context-dependent profile. Positive signals appear in: cardiometabolic. Negative signals appear in: cardiometabolic. Null findings dominate: cardiometabolic, dosing pharmacokinetics. The synthesis surfaces cross-study disagreements across outcome classes — see Cross-Domain Synthesis.

The strongest unresolved contrast is the disagreement between Buse 2025 and Ganeshalingam 2026 on cardiometabolic (severity 5/5), 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
longevity	0	1	mixed	direct interventional hard-endpoint gap
cardiometabolic	5	17	mixed, negative, null, positive, unclear	conflict-resolution gap
dosing and pharmacokinetics	0	1	null	direct interventional hard-endpoint gap
skeletal, fracture, and bone	0	1	unclear	direct interventional hard-endpoint gap
contextual adjacent evidence	1	3	null, unclear	replication gap

Evidence-Gap Priority

Priority	Gap	Rationale
P1	longevity: direct interventional hard-endpoint gap	0 direct and 1 indirect source; direction profile: mixed
P2	cardiometabolic: conflict-resolution gap	5 direct and 17 indirect sources; direction profile: mixed, negative, null, positive, unclear
P3	dosing and pharmacokinetics: direct interventional hard-endpoint gap	0 direct and 1 indirect source; direction profile: null
P4	skeletal, fracture, and bone: direct interventional hard-endpoint gap	0 direct and 1 indirect source; direction profile: unclear
P5	contextual adjacent evidence: replication gap	1 direct and 3 indirect sources; direction profile: null, unclear

Next-Study Design Recommendation

The next high-yield study for Semaglutide Intervention Semaglutide 2 4 Mg Rates should target the longevity 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 12 months; 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

• Hamarsheh 2026; tier=A1; directness=direct; endpoint=cardiometabolic; direction=mixed; representative statistic=P < 0.0001.
• Buse 2025; tier=A1; directness=direct; endpoint=cardiometabolic; direction=negative; representative statistic=P < 0.001.
• Ganeshalingam 2026; tier=A1; directness=direct; endpoint=cardiometabolic; direction=positive; representative statistic=P < 0.001.
• Cortes 2024; tier=A1; directness=direct; endpoint=cardiometabolic; direction=unclear.
• Lin 2024; tier=A1; directness=direct; endpoint=cardiometabolic; direction=null.
• Koychev 2024; tier=A1; directness=direct; endpoint=contextual adjacent evidence; direction=null.
• Sillassen 2025; tier=B1; directness=review; endpoint=cardiometabolic; direction=mixed; representative statistic=P < 0.001.
• Ciudin 2026b; tier=B1; directness=review; endpoint=cardiometabolic; direction=unclear.
• Abdullah 2025; tier=B1; directness=review; endpoint=longevity; direction=mixed; representative statistic=P < 0.00001.
• Masson 2024; tier=B1; directness=review; endpoint=contextual adjacent evidence; direction=unclear; representative statistic=P = 0.098.

Source Classification Map

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

• Hamarsheh 2026: outcome=cardiometabolic; directness=direct; tier=A1; direction=mixed; claims=379.
• Buse 2025: outcome=cardiometabolic; directness=direct; tier=A1; direction=negative; claims=203.
• Ganeshalingam 2026: outcome=cardiometabolic; directness=direct; tier=A1; direction=positive; claims=74.
• Cortes 2024: outcome=cardiometabolic; directness=direct; tier=A1; direction=unclear; claims=41.
• Lin 2024: outcome=cardiometabolic; directness=direct; tier=A1; direction=null; claims=28.
• Koychev 2024: outcome=contextual adjacent evidence; directness=direct; tier=A1; direction=null; claims=12.
• Sillassen 2025: outcome=cardiometabolic; directness=review; tier=B1; direction=mixed; claims=210.
• Ciudin 2026b: outcome=cardiometabolic; directness=review; tier=B1; direction=unclear; claims=111.
• Abdullah 2025: outcome=longevity; directness=review; tier=B1; direction=mixed; claims=101.
• Masson 2024: outcome=contextual adjacent evidence; directness=review; tier=B1; direction=unclear; claims=46.
• Elganyny 2026: outcome=cardiometabolic; directness=review; tier=B1; direction=mixed; claims=9.
• Qin 2024: outcome=cardiometabolic; directness=review; tier=B1; direction=null; claims=8.
• Efficacy of Semaglutide S n.d.: outcome=cardiometabolic; directness=review; tier=B1; direction=unclear; claims=2.
• Primary Prevention and Uterine n.d.: outcome=cardiometabolic; directness=review; tier=B1; direction=unclear; claims=1.
• Garvey 2022: outcome=cardiometabolic; directness=indirect; tier=B2; direction=mixed; claims=403.
• Ciudin 2026a: outcome=cardiometabolic; directness=indirect; tier=B2; direction=unclear; claims=182.
• Arslanian 2025: outcome=cardiometabolic; directness=indirect; tier=B2; direction=mixed; claims=178.
• McGowan 2025: outcome=cardiometabolic; directness=review; tier=B2; direction=unclear; claims=147.
• Tan 2026: outcome=cardiometabolic; directness=indirect; tier=B2; direction=mixed; claims=145.
• Jensen 2025: outcome=cardiometabolic; directness=indirect; tier=B2; direction=unclear; claims=84.
• Park 2025: outcome=skeletal fracture bone; directness=indirect; tier=B2; direction=unclear; claims=81.
• Lu 2026: outcome=cardiometabolic; directness=review; tier=B2; direction=unclear; claims=75.
• Zaccardi 2026: outcome=cardiometabolic; directness=indirect; tier=B2; direction=null; claims=73.
• Chrzanowski 2026: outcome=cardiometabolic; directness=review; tier=B2; direction=unclear; claims=53.
• Hendershot 2026: outcome=contextual adjacent evidence; directness=indirect; tier=B2; direction=unclear; claims=38.
• Alnaimi 2026: outcome=contextual adjacent evidence; directness=review; tier=B2; direction=unclear; claims=23.
• Lassen 2026: outcome=cardiometabolic; directness=indirect; tier=B2; direction=null; claims=19.
• Harbi 2026: outcome=cardiometabolic; directness=review; tier=B2; direction=null; claims=5.
• Sorum 2024: outcome=dosing pharmacokinetics; directness=protocol; tier=D1; direction=null; claims=26.

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 5 disagreement: Buse 2025 vs Ganeshalingam 2026; Buse 2025 reports negative effect on cardiometabolic; Ganeshalingam 2026 reports positive on the same outcome — direct conflict
• Severity 4 null vs negative: Lin 2024 vs Buse 2025; Buse 2025 (negative on cardiometabolic) vs Lin 2024 (null on cardiometabolic) — partial conflict
• Severity 4 null vs positive: Lin 2024 vs Ganeshalingam 2026; Ganeshalingam 2026 (positive on cardiometabolic) vs Lin 2024 (null on cardiometabolic) — partial conflict
• Severity 3 indirectness gap: Efficacy of Semaglutide S n.d. vs Lin 2024; Lin 2024 (direct, A1) vs Efficacy of Semaglutide S n.d. (review) on cardiometabolic — direct vs indirect must be kept separate
• Severity 3 indirectness gap: Efficacy of Semaglutide S n.d. vs Cortes 2024; Cortes 2024 (direct, A1) vs Efficacy of Semaglutide S n.d. (review) on cardiometabolic — direct vs indirect must be kept separate
• Severity 3 indirectness gap: Efficacy of Semaglutide S n.d. vs Buse 2025; Buse 2025 (direct, A1) vs Efficacy of Semaglutide S n.d. (review) on cardiometabolic — direct vs indirect must be kept separate
• Severity 3 indirectness gap: Efficacy of Semaglutide S n.d. vs Ganeshalingam 2026; Ganeshalingam 2026 (direct, A1) vs Efficacy of Semaglutide S n.d. (review) on cardiometabolic — direct vs indirect must be kept separate
• Severity 3 indirectness gap: Efficacy of Semaglutide S n.d. vs Hamarsheh 2026; Hamarsheh 2026 (direct, A1) vs Efficacy of Semaglutide S n.d. (review) on cardiometabolic — direct vs indirect must be kept separate

References

• Garvey 2022. Two-year effects of semaglutide in adults with overweight or obesity: the STEP 5 trial. Nature Medicine, 2022. DOI: 10.1038/s41591-022-02026-4 PMID: 36216945.
• Hamarsheh 2026. Comparative Effectiveness of CagriSegma , Semaglutide, Cagrilintide and Tirzepatide in the Management of Overweight and Obesity: A Network Meta‐Analysis of Randomized Clinical Trials. Endocrinology, Diabetes & Metabolism, 2026. DOI: 10.1002/edm2.70248 PMID: 42207966.
• Sillassen 2025. The adverse effects associated with semaglutide use in patients at increased risk of cardiovascular events: a systematic review with meta-analysis and Trial Sequential Analysis. BMC Medicine, 2025. DOI: 10.1186/s12916-025-04486-0 PMID: 41286875.
• Buse 2025. Long-term comparative effectiveness of once-weekly semaglutide versus alternative treatments in a real-world US adult population with type 2 diabetes: a randomized pragmatic clinical trial. BMJ Open Diabetes Research & Care, 2025. DOI: 10.1136/bmjdrc-2025-005161 PMID: 41093600.
• Ciudin 2026a. Indirect Comparative Efficacy and Safety of Tirzepatide Versus Oral Semaglutide for the Treatment of Overweight and Obesity. Diabetes, Obesity & Metabolism, 2026. DOI: 10.1111/dom.70773 PMID: 42050884.
• Arslanian 2025. Effect of Semaglutide on Insulin Sensitivity and Cardiometabolic Risk Factors in Adolescents With Obesity: The STEP TEENS Study. Diabetes Care, 2025. DOI: 10.2337/dc25-0824 PMID: 41296499.
• McGowan 2025. A systematic review and meta-analysis of the efficacy and safety of pharmacological treatments for obesity in adults. Nature Medicine, 2025. DOI: 10.1038/s41591-025-03978-z PMID: 41039116.
• Tan 2026. Cardiometabolic and Renal Outcomes in Semaglutide Users with Type 2 Diabetes Achieving Glycemic and Weight Goals: An Observational Cohort Study. Advances in Therapy, 2026. DOI: 10.1007/s12325-026-03610-7 PMID: 42060161.
• Ciudin 2026b. Comparison of Clinical Efficacy and Safety of Tirzepatide, Liraglutide and Semaglutide in Patients with Obesity and Without T2D: A Bayesian Network Meta-Analysis of Randomised Controlled Trials. Advances in Therapy, 2026. DOI: 10.1007/s12325-026-03523-5 PMID: 41820778.
• Abdullah 2025. Safety and Efficacy of Semaglutide in Patients With Chronic Kidney Disease, With or Without Type 2 Diabetes: A Systematic Review and Meta‐Analysis. Endocrinology, Diabetes & Metabolism, 2025. DOI: 10.1002/edm2.70136 PMID: 41276951.
• Jensen 2025. Efficacy of 12 months therapy with glucagon-like peptide-1 receptor agonists liraglutide and semaglutide on weight regain after bariatric surgery: a real-world retrospective observational study. BMC Endocrine Disorders, 2025. DOI: 10.1186/s12902-025-01913-4 PMID: 40197361.
• Park 2025. Semaglutide promotes bone marrow–derived progenitor cell flux towards an anti-inflammatory and pro-regenerative profile in high-risk patients: the SEMA-VR CardioLink-15 trial. European Heart Journal, 2025. DOI: 10.1093/eurheartj/ehaf690 PMID: 40886061.
• Lu 2026. Cardiometabolic Profiles of Oral and Subcutaneous Glucagon‐Like Peptide‐1 Receptor Mono‐Agonists in Adults With Overweight or Obesity: A Systematic Review and Network Meta‐Analysis. Diabetes, Obesity & Metabolism, 2026. DOI: 10.1111/dom.70742 PMID: 41992023.
• Ganeshalingam 2026. Semaglutide Effects on Insulin Sensitivity and β-Cell Function in Patients With Schizophrenia, Prediabetes, and Obesity Treated With Second-Generation Antipsychotics: Findings From the HISTORI Trial, a 30-Week Randomized, Placebo-Controlled Trial With Semaglutide 1.0 mg Weekly. Diabetes Care, 2026. DOI: 10.2337/dc25-2041 PMID: 41778920.
• Zaccardi 2026. Semaglutide Treatment in Young Adults Living With Type 2 Diabetes: A Post Hoc Analysis From the SUSTAIN and PIONEER Clinical Trials. Diabetes, Obesity & Metabolism, 2026. DOI: 10.1111/dom.70770 PMID: 41994903.
• Chrzanowski 2026. Semaglutide-associated risk of nonarteritic anterior ischemic optic neuropathy in patients with type 2 diabetes: A systematic review and meta-analysis of observational studies. PLOS Medicine, 2026. DOI: 10.1371/journal.pmed.1005064 PMID: 42166479.
• Masson 2024. Anti-inflammatory effect of semaglutide: updated systematic review and meta-analysis. Frontiers in Cardiovascular Medicine, 2024. DOI: 10.3389/fcvm.2024.1379189 PMID: 39055657.
• Cortes 2024. Effect of Semaglutide on Physical Function, Body Composition, and Biomarkers of Aging in Older Adults With Overweight and Insulin Resistance: Protocol for an Open-Labeled Randomized Controlled Trial. JMIR Research Protocols, 2024. DOI: 10.2196/62667 PMID: 39269759.
• Hendershot 2026. Once-Weekly Semaglutide in Adults With Daily Cigarette Use. JAMA Network Open, 2026. DOI: 10.1001/jamanetworkopen.2026.14898 PMID: 42189538.
• Lin 2024. Semaglutide combined with empagliflozin vs. monotherapy for non-alcoholic fatty liver disease in type 2 diabetes: Study protocol for a randomized clinical trial. PLOS ONE, 2024. DOI: 10.1371/journal.pone.0302155 PMID: 38701096.
• Sorum 2024. Semaglutide treatment for PRevention Of Toxicity in high-dosE Chemotherapy with autologous haematopoietic stem-cell Transplantation (PROTECT): study protocol for a randomised, double-blind, placebo-controlled, investigator-initiated study. BMJ Open, 2024. DOI: 10.1136/bmjopen-2024-089862 PMID: 39384243.
• Alnaimi 2026. Weight‐Lowering Drugs and Natural Female Fertility—A Systematic Review and Meta‐Analysis. Clinical Obesity, 2026. DOI: 10.1111/cob.70092 PMID: 42307450.
• Lassen 2026. SEMASEARCH Study Design: Real‐World Evaluation of Semaglutide 2.4 mg in Adults With Severe Obesity Underrepresented in Clinical Trials. Diabetes, Obesity & Metabolism, 2026. DOI: 10.1111/dom.70697 PMID: 41884974.
• Koychev 2024. Protocol for a double-blind placebo-controlled randomised controlled trial assessing the impact of oral semaglutide in amyloid positivity (ISAP) in community dwelling UK adults. BMJ Open, 2024. DOI: 10.1136/bmjopen-2023-081401 PMID: 38908839.
• Elganyny 2026. Long-Term Safety and Renal Outcomes of Semaglutide in Non-Diabetic Obesity with Chronic Kidney Disease or Hypertension: A Systematic Review and Meta-Analysis. Clin Ter, 2026. DOI: 10.7417/ct.2026.2083 PMID: 42340790.
• Qin 2024. Efficacy and safety of semaglutide 2.4 mg for weight loss in overweight or obese adults without diabetes: An updated systematic review and meta‐analysis including the 2‐year <scp>STEP</scp> 5 trial. Diabetes Obes Metab, 2024. DOI: 10.1111/dom.15386 PMID: 38016699.
• Harbi 2026. Tirzepatide vs. semaglutide for obesity, glycemic control, and cardiovascular outcomes: a narrative review of clinical trials. Frontiers in Medicine, 2026. DOI: 10.3389/fmed.2026.1764664 PMID: 42100257.
• Efficacy of Semaglutide S n.d.. Efficacy of Semaglutide s.c. Once-weekly on Weight Loss and Management in Adolescents With Monogenic Obesity in Clinical Practice. 2028. Identifier unavailable; no DOI or PMID in source metadata.
• Primary Prevention and Uterine n.d.. Primary Prevention and Uterine Preservation in Premenopausal Women With Obesity and Endometrial Hyperplasia. 2030. Identifier unavailable; no DOI or PMID in source metadata.