Review papers with raw data transparency

Overview

This narrative review (Communications Medicine, DOI 10.1038/s43856-025-01165-2) brings together evidence from multiple omic layers to explain how preeclampsia (PE) may (1) reflect maternal cardiovascular susceptibility, (2) drive persistent maternal vascular and metabolic changes postpartum, and (3) program offspring cardiovascular trajectories via placental dysfunction and molecular signalling Multi-omic PE CVD Review

What the paper does well

• Comprehensive cross-omic synthesis: integrates findings across genomics, epigenomics, transcriptomics, proteomics, metabolomics and microbiomics and maps them to clinical phenotypes (EO-PE vs LO-PE, SGA, preterm) and CVD endpoints Multiome integration claim
• Clear focus on biologically plausible axes: angiogenic imbalance (sFlt-1/PlGF/FLT1), RAAS/AT1-autoantibodies, senescence and metabolic dysregulation are linked mechanistically to both acute PE and later CVD risk Angiogenic imbalance
• Actionable biomarker discussion: evaluates cfDNA/cfRNA fragmentomics, cfDNA methylation, PAPPA2, FLT1, sFlt-1/PlGF ratio and proteomic subclasses with explicit links to prediction performance (eg cfDNA methylation combined risk score predicting 72% EO-PE at 80% specificity) cfDNA methylation performance

Major limitations and blindspots

1. Narrative review design limits causal inference — the paper pools association studies, animal models and small cohort omics datasets; it cannot by itself establish causality linking molecular patterns in pregnancy to later CVD events in mothers or offspring Review limitations
2. Heterogeneity in PE definitions and populations — EO-PE and LO-PE show different biology; meta-synthesis risks conflating placental-driven and maternal-driven disease mechanisms without uniform case definitions across cited studies EO-vs-LO heterogeneity
3. Longitudinal evidence sparse — the paper repeatedly notes the lack of large, long-term, multi-omic cohorts linking in‑pregnancy molecular signatures to hard cardiovascular endpoints years-to-decades later; this is the critical translational gap Need for longitudinal data
4. Potential publication and selection biases — omic studies with positive biomarker findings are preferentially published; small sample sizes and single‑centre proteomic/metabolomic studies can overstate effect sizes and subtyping robustness Proteome sample size limits

Detailed evidence mapping (claims and support)

Practical implications and next steps recommended by reviewer

• Prioritize prospective, longitudinal multi-omics cohorts that sample mothers in preconception, each trimester, peripartum and repeatedly postpartum with standard clinical cardiovascular phenotyping and adjudicated CVD outcomes to test predictive and causal models (this is the critical gap highlighted in the paper) Need for longitudinal follow up
• Harmonize definitions and biospecimen protocols (PE subtype criteria, cfDNA assay standards, methylation pipelines) to reduce heterogeneity and make multi-cohort meta-analysis feasible.
• Use causal inference frameworks (Mendelian randomization, longitudinal mediation analysis, genetic colocalization) combined with multi-omic QTL mapping to prioritize biomarkers with causal evidence rather than associative signals alone Genetic architecture and PRS
• Prioritize external validation of cfDNA methylation and fragmentomic predictors in diverse populations before clinical translation.

Confidence and where the conclusion could change

Confidence in the review conclusions is moderate: the paper accurately synthesizes current evidence and identifies plausible mechanisms linking PE to later CVD risk, but its overall conclusions depend heavily on observational and often small omics studies; high-quality longitudinal data or robust causal genetic evidence that contradicts placental/anti-angiogenic mediation would change the interpretation markedly Review limitations and uncertainty

Useful data visualizations

Quick actionable checklist for researchers

1. Assemble prospective multi-ethnic pregnancy cohorts with standardized biospecimens and cardiovascular follow-up.
2. Pre-register analytic pipelines for cfDNA fragmentomics and methylation to reduce HARKing and p-hacking risk.
3. Use genetics (colocalization, MR) to triage biomarkers for downstream mechanistic work.
4. Share processed multi-omics data in public repositories with clear metadata to enable reproducibility.

Interactive next steps

To evolve this critique into reproducible analyses (summary statistics meta-analysis, cfDNA classifier validation, WGCNA reanalysis) you can run the BGPT AI Biology agent below to iteratively fetch datasets, harmonize, and compute robust cross-study assessments.

Author review buttons

Open bespoke Author reviews