Low-dose vaginal estrogen for the genitourinary syndrome of menopause.

The signal was indexed against menopause through the Direct Research arm, which reads peer-reviewed literature for compound–condition pairs and extracts a verbatim claim for each finding. For this pair the arm surfaced a single source: a 2023 review of menopausal symptom management.

“Low-dose vaginal estrogen is associated with subjective improvement in GSM symptom severity by approximately 60% to 80%”PubMed · PMID 36749328

Local estrogen restores estrogen-dependent vaginal and urogenital epithelial tissue, relieving genitourinary menopausal symptoms. That mechanism is well established clinically, which is exactly why the modest score below is worth walking through carefully.

The headline is the Direct arm's reading. Each of the five dimensions below carries the engine's own 0–2 score and the rationale it wrote for that score. Nothing here is curated by hand; it is the live row.

Rigor and specificity are both at the ceiling: the source is a review (the highest design tier the rubric recognises) and it names the intervention and the condition explicitly. What holds the score back is corroboration and consistency. The engine held corroboration at 1 because a single review is one synthesis; the trials pooled inside it do not count as independent sources. It scored consistency 1 because, with only one source on file, agreement across independent studies cannot be assessed and defaults to neutral.

This is the engine refusing to inflate. The underlying therapy is guideline-backed and well replicated in the wider literature, but the model only credits what it has ingested and verified verbatim. A single point of consistency is the entire distance between this Moderate signal and a Strong one. The companion walkthrough, aromatase inhibitors for endometriosis, shows the mirror image: the same corroboration score of 1, but a systematic review whose pooled studies agree pushes consistency to 2 and the pair to Strong.

A fair test of any evidence engine is to ask what its ingested sources miss. The broader clinical record for vaginal estrogen in postmenopausal women, including its recurrent-UTI prevention role, extends well past the one review on file. The references below sit outside the ingested corpus today. They are kept out of the score by design, so the score stays a measure of what the engine has actually read. Ingesting these would legitimately raise the corroboration and consistency dimensions from section 02.

The earlier version of this page called this signal the cleanest Strong case in the database. The current engine disagrees with its own predecessor, and the disagreement is the feature. It scores corroboration and consistency on verbatim-verified claims, so the model lands this pair at Moderate. That is an honest reflection of a corpus that holds one review, while the wider field has produced dozens of trials and guidelines.

The clinical follow-on question, why a guideline-backed therapy remains under-prescribed in routine postmenopausal care, sits outside the engine's scope. The engine records the evidence it has read and makes the gaps legible. Closing those gaps, in the corpus and in the clinic, is the work the platform is built to inform.