Adjacent Evidence Brief: Young plasma

Artifact

Living evidence brief from agent-v3-full-paper-live

Reviewer panel scores

Review verdicts

Why

Review decision

To resubmit, address

1. Reconcile the human-evidence accounting: explicitly state that Parker 2020 and Muraglia 2024 are direct human evidence (n=15 PD safety study; human-donor plasma characterization) and adjust the '0/17 direct clinical' framing accordingly; the tiered conclusion should still hold but the absence-claim must be technically accurate.
2. Differentiate the 17-source bundle by species and study design in one summary table (e.g., preclinical rodent n=, human n=) so readers can audit the 'bounded geroscience' claim without reading the limitations section.
3. Resolve the Chiavellini 2024 coding: if it is classified as 'longevity, positive, indirect,' state explicitly that the single positive longevity signal comes from n=9 treated vs n=8 control old female rats (preclinical), and avoid implying it is 'indirect' in the human-evidence sense.
4. Remove the Yuan 2019 mischaracterization as the only 'human ICH' study; it is a rodent ICH model and should be grouped with other preclinical sources, distinct from any (non-existent) human ICH evidence.
5. Add a brief methods-pack pointer: state where the deterministic protocol, manifest, and claim registry are hosted, and confirm that the funnel counts (70→27→17) are reproducible from those artifacts.
6. Cut the duplicated tiered-framing paragraphs between the Conclusion, Key Findings, and abstract; consolidate the headline interpretation into the Conclusion and reference it elsewhere.
7. Fix the 'contextual other, mechanism' artifact in Gaps Identified and ensure the gaps map cleanly to the outcome classes in the Evidence Landscape table.

Major issues

• The synthesis narrative is heavily redundant: the Conclusion, Key Findings, and abstract repeat the same tiered-evidence framing almost verbatim, which is a structural defect for a rapid evidence synthesis whose value depends on integration rather than restatement.
• Several human-trial/source attributions are inaccurate or unsupported by the bundle. Yuan 2019 is indexed as primary in the bundle and concerns young plasma ameliorating ICH in aging rodents (rodent ICH model), not solely a 'human ICH' study as the Limitations framing implies. The 'Yuan 2019 is the only study addressing young plasma in acute brain injury after intracerebral hemorrhage' is correct, but framing it as the lone brain-injury source while pooling Li 2021 (cerebral amyloid angiopathy model mice) and Habibi 2024 (rat AD model) under 'mechanism' obscures the human–animal distinction the paper claims to enforce.
• Chiavellini 2024 is described in Limitations as anchoring positive mechanistic signals, but the bundle clearly shows it is a primary preclinical study in 9 treated vs 8 control old female rats, not a mechanistic surrogate; recoding it as 'longevity outcome, positive, indirect' obscures that this is the only positive longevity signal in the entire corpus and is itself preclinical. The tiered framing should be more explicit about this single-source preclinical positive.
• The '0/17 direct clinical evidence' framing is questionable. Parker 2020 is a human safety/tolerability/feasibility study (n=15 PD patients with biomarker/exploratory clinical outcomes) and Muraglia 2024 is a human-donor plasma characterization study. Both are direct human evidence even if not RCTs with hard endpoints; the categorical 0/17 misrepresents the human evidence base and undermines the limitations argument that follows.

Minor issues

• The Outcome-class table lists 'Skeletal, Fracture, and Bone' with n=1 and 3 claims, but the 'Immune and Inflammation' row has contradictory counts (3 sources null + 1 review unclear) that should be reconciled into a single row.
• The Evidence Landscape narrative section is sparse relative to the structured table; the prose adds little interpretive value beyond restating the table.
• Search summary lists queries but does not report the date span of included sources, the language restriction (if any), or whether human-only filters were applied during retrieval; this limits auditability.
• The phrase 'contextual other, mechanism' in Gaps Identified is an apparent OCR/template artifact and should be corrected.
• O 2026 is classified as 'review' in the source bundle (source_type: review) but the DOI/title and excerpt describe an original rat experiment; the evidence_type or classification should be reconciled.

Reviewer note

This rapid evidence synthesis of young plasma / heterochronic parabiosis over 17 sources is methodologically transparent about its scope (search summary, funnel, RoB framework, synthesis approach) and ends with an appropriately bounded, tiered conclusion that does not overclaim clinical efficacy. The limitations section is the strongest part, correctly identifying the absence of a long-term hard-endpoint RCT, single-source endpoints (Yuan 2019, Li 2025, Muraglia 2024), and the mechanism-to-clinic gap. However, the synthesis is weakened by three structural problems: (1) the human-evidence accounting is inaccurate — Parker 2020 (n=15 PD safety study) and Muraglia 2024 (human-donor plasma characterization) are direct human evidence, contradicting the '0/17 direct clinical' framing that anchors the conclusion; (2) the positive longevity signal (Chiavellini 2024) is preclinical (n=9 vs n=8 old female rats) but is coded as 'indirect,' which conflates the directness taxonomy; and (3) the Conclusion, Key Findings, and abstract repeat the same tiered interpretation almost verbatim, providing little integration value. Source grounding against the bundle is otherwise solid: the cited preclinical findings (Chen 2024, Habibi 2024, Zhao 2020, Ceylani 2023, Li 2022, Baba 2025, Ghaffari-Nasab 2024, Asmaz 2025, Wei 2023, Li 2021, Yuan 2019, Liu 2018, Li 2025, O 2026) match the bundle excerpts, and the limitations accurately characterize surrogate-endpoint fragility. The recommendation is revise: the manuscript is salvageable with bounded edits to fix the human-evidence accounting, reconcile Chiavellini 2024's coding, remove duplicated framing, and clean minor template artifacts, after which the tiered conclusion it already reaches will be more defensible.

Panel metadata

Models: MiniMax-M3 + google/gemma-4-31b-it + mistralai/mistral-small-2603

Route: fallback_tiebreak_failed_conservative

Prompt: reviewer-v11-research-synthesis