Methodology

The ASCVD Risk Factor Atlas is a de novo, evidence-linked catalog of every risk factor reported in the biomedical literature for atherosclerotic cardiovascular disease — coronary heart disease, ischemic stroke, peripheral artery disease, and aortic/large-artery atherosclerosis. It was built ground-up by mining primary sources rather than transcribing an existing guideline list, so it surfaces both established factors and uncommon or emerging ones.

1. Literature harvest — 6,441 distinct articles from PubMed and OpenAlex across broad, per-category, and study-type queries.
2. Candidate extraction — named-entity extraction over abstracts surfaced every mentioned risk factor / exposure / biomarker, each linked to its source article.
3. Trials & genetics — 5,715 ClinicalTrials.gov studies and 6,227 GWAS Catalog associations (trait- and gene-anchored) were harvested for linking.
4. Canonicalization — synonyms were merged into 997 distinct canonical factors, each assigned a category, type, direction, and modifiability.
5. Evidence linking & grading — references, trials, and genetic associations were attached to each factor and combined into an evidence-strength grade and score.

Each factor receives a grade from its evidence profile:

• A Strong & causal — multiple meta-analyses, or genetic/Mendelian-randomization support plus trial evidence, with high literature volume.
• B Moderate — consistent evidence with a meta-analysis or genetic association.
• C Emerging — several primary studies, limited meta-analytic/causal confirmation.
• D Preliminary — few reports or single studies.

The score (0–100) weights literature volume, meta-analyses, linked trials, genetic associations, and a Mendelian-randomization bonus. This is a pragmatic ranking to gauge how much evidence a factor carries — not a formal GRADE assessment.

• Extraction is automated; category/direction assignments are best-effort and may contain errors.
• Trial and gene→factor links are keyword/curation based and can miss or over-include.
• Reference counts reflect this harvest, not the entire literature — they index attention, not truth.
• Some bidirectional conditions (e.g. heart failure) appear as factors where the literature reports them as such.