Adjacent Evidence Brief: SASP secretome

Artifact

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

Reviewer panel scores

Review verdicts

Why

Review decision

To resubmit, address

1. Add a clear table or paragraph distinguishing human vs. animal vs. in vitro sources within each outcome class, so the reader cannot conflate Ichim 2026 (murine) with any human longevity signal.
2. Re-examine source admission: several bundle sources (Shah 2025 protocol, Coppe 2008 characterization, Yue 2022 bibliometric review, Filipek 2026 narrative review, Giuliani 2023 endothelial mechanism, Terlecki-Zaniewicz 2018 sEV characterization) are not direct outcome evidence and should either be re-tiered as mechanistic/contextual or removed from the outcome-class claims counts.
3. Reconcile the Evidence Landscape prose citations (Zhang, Coppe, Niklander, Nicoloro-Santabarbara, Ostrowska, Alam, Ichim) with the bundle and ensure every named source in the manuscript appears in the bundle and vice versa for sources actually used in claims.
4. Provide a substantive answer to a real domain question (e.g., 'What does current evidence support about SASP as a measurable clinical biomarker vs. a mechanistic construct?') rather than the meta-corpus question currently posed; the conclusion should then deliver a bounded substantive finding, not just restate the evidence profile.
5. Clarify the admission funnel arithmetic so the bucket counts reconcile to 17 admitted sources, and explain why 'strict high-confidence' (n=2) and broader admission criteria both apply.
6. Justify the direction coding for each source-level statistic cited, particularly the null coding of Sanchez-Romero 2026 (which reports null exercise effects on SASP cytokines) and the negative coding of Zuccolo 2020 and ADT 2024.

Major issues

• Evidence-role imbalance is acknowledged but not resolved: with 0 direct clinical sources out of 17 admitted, the manuscript still organizes a 'longevity outcome class' that treats murine lifespan extension (Ichim 2026) as a longevity signal without explicitly flagging it as non-human in the Key Findings and Evidence Landscape tables, creating a risk of reader conflation of animal and human evidence.
• Several admitted sources do not primarily address SASP secretome as a primary endpoint (e.g., Coppe 2008 characterizes SASP biology; Shah 2025 is a protocol for an SGLT2i trial where SASP is a mechanistic secondary outcome; Filipek 2026 is a narrative review on psoriasis with SASP as one component). Their inclusion as 'high-confidence extracted claims' about SASP secretome effectiveness requires explicit justification of why they are counted as outcome-class evidence rather than as mechanistic/background sources.
• The research question as stated ('What does the retained source corpus establish about SASP secretome?') is meta-corpus rather than substantive, and the conclusion is correspondingly circular: it restates the evidence profile rather than answering a domain question. The synthesis does not produce a usable bounded claim about SASP biology, measurement, or translation.
• Source bundles show several citations referenced in the manuscript text (Zhang 2019, Coppe 2008, Niklander 2020, Nicoloro-Santabarbara 2025, Ostrowska 2024, Alam 2025) but the Evidence Landscape tables mix these with sources not present in the bundle (e.g., the source-context map references skeletal/muscle and oncology findings not traceable to the provided bundle entries). Several bundle sources (Evans 2023, Shah 2025, Yue 2022, Franco 2025, Filipek 2026, Giuliani 2023, Terlecki-Zaniewicz 2018) are not surfaced in the prose claims, while the prose cites bundle sources not always appearing in the documented outcome-class breakdowns.

Minor issues

• The '1 included source were assigned to this outcome class' sentence has a grammatical/agreement error and should be corrected.
• Direction-reconciliation language is conservative but the manuscript does not make clear why some statistically significant results are coded null (e.g., Coppe 2008 p > 0.05 reported in context of SASP characterization) while others are coded negative (Zuccolo 2020, ADT 2024); a brief rationale would improve auditability.
• The Sanchez-Romero 2026 excerpt reports 'no statistically significant differences' for exercise on SASP cytokines, yet the manuscript summarizes the immune/inflammation class as having a 'negative signal'; this coding choice should be explained.
• Source-context map uses '1 sources' rather than '1 source' in the skeletal/muscle row.
• The Key Findings section is essentially a restatement of the outcome-class note and source-context map headers rather than a substantive findings list.
• Search summary is explicit and well-structured but the funnel arithmetic is hard to reconcile: 40 classified candidates minus 32 'no extractable claims' minus 9 'none-only' minus 31 'mixed partial-or-none' minus 10 'partial-only' minus 2 'strict high-confidence' does not yield 17 admitted sources; the bucket definitions should be clarified.

Reviewer note

The submission is a structured evidence brief on SASP secretome with an unusually explicit search-summary section, which is a strength. The evidence-role imbalance (0 direct clinical sources) is honestly flagged, and the manuscript avoids strong causal or deployment claims, which keeps overclaim low. However, several issues prevent acceptance. First, the research question is meta-corpus ('what does the retained corpus establish?') rather than substantive, so the conclusion is circular. Second, the Evidence Landscape cites sources in prose that are not always traceable in the provided bundle, while several bundle sources are not surfaced in claims. Third, with 0 direct clinical sources, treating Ichim 2026 (murine senolytic-regenerative therapy) as a 'longevity outcome class' positive signal risks reader conflation of animal and human evidence unless this is explicitly flagged in the table. Fourth, several admitted sources are not direct outcome evidence (Shah 2025 protocol, Coppe 2008 characterization, Yue 2022 bibliometric review, Filipek 2026 narrative review, Giuliani 2023 endothelial mechanism) and their inclusion in outcome-class claim counts inflates apparent support. The limitations section is honest but generic; the gaps section is actionable. Overall the manuscript is salvageable with bounded revisions: re-tier non-outcome sources, reconcile bundle and prose citations, distinguish human from animal evidence in tables, and reframe the research question to deliver a substantive bounded finding rather than a restatement of the corpus profile.

Panel metadata

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

Route: consensus

Prompt: reviewer-v11-research-synthesis