Methods: how items are chosen, summarised and scored

Selection

Candidates come from a standing PubMed survey: everything carrying CETP in the title, every paper naming one of the five agents, recent work mentioning CETP, the major journals, the review and trial slice, the genetics literature and the cognition and longevity angle. A paper is included when CETP or a CETP inhibitor is a primary exposure, intervention or outcome and the abstract reports a result that can be quoted. A paper that mentions CETP inhibitors in a list of drugs in development is not included.

CETP biology outside cardiovascular medicine is included where CETP or an inhibitor is the intervention, and is scored low and labelled as preclinical.

Summarising

Every summary is written from the published abstract. Figures appear exactly as published: no rounding, no unit conversion, no numbers imported from memory or from elsewhere in the paper. Where the abstract reports a primary and secondary endpoint, the primary one leads. Each summary ends with a clause of judgement: what the paper adds, or the limitation that bounds it.

Negative and terminated trials are written with the same care as positive ones. This class has a graveyard, and reporting it accurately is most of the value of the site.

How new studies arrive

Since 20 August 2026 the PubMed survey runs nightly and is automated. A language model running on our own hardware reads each new abstract, decides whether the paper belongs here against a written editorial brief, and drafts the headline and the summary. Every draft then passes an automated gate that rejects it unless the tags are from this site's vocabulary, the headline carries the journal and year, the summary is the right length, and every number in it appears in the abstract. Spelling is checked against English dictionaries plus a medical word list. A paper is scheduled days ahead, in plain sight, and one is published each morning.

The automation only ever adds. It never edits or removes an existing item, and it never marks one landmark: the canon stays a human judgement. Where a summary is wrong, the abstract and the DOI remain the source of truth, and Corrections below applies exactly as before.

Significance

Each item carries a score from 1 to 10. 10 means the paper changed the field's direction (ILLUMINATE, dal-OUTCOMES, ACCELERATE, REVEAL, the Mendelian randomisation work). 9 is a phase 3 result or a practice-changing document; 8 a practice-relevant trial or large cohort; 7 a solid confirmation or mechanism; 6 useful and incremental; 5 an orientation piece or small cohort; 4 a small, single-centre, in vitro or animal study. Below 3 the item is not included. The score is an editorial judgement, not a metric.

Dates and links

Items are dated by the earlier of the journal issue date and the electronic publication date, so the library sorts by when the science appeared rather than when it was entered here. Each study links to its DOI, with the PubMed record as a second link. Nothing is presented as open access when it is not: the summary is written at abstract level and the source link is where the full text lives.

Corrections

If something here is wrong, the source wins. Corrections are made to the item and its date of last update changes accordingly.