WhelWomen's Health Evidence Lab
Home/Full index (preview)/WHEL-C-085
WHEL-C-085 · Moderate evidence · 7/10

metformin for PCOS

In meta-analyses informing the 2023 PCOS guidelines, metformin may improve waist-hip ratio and hirsutism versus inositol but is inferior to combined oral contraceptives on androgen markers (FAI, SHBG, testosterone) in women with PCOS.

Origin · FDA ApprovedPathway · 505(b)(2) · existing active ingredient, new indicationEvidence arm · Direct researchContradiction present
How to read thisThe summary above and the proposed mechanism are generated by the model from the sources it ingested, and are written as the model’s reasoning rather than established fact. Any figure quoted from MATRIX is a model-derived association score, not a clinical measurement. How far the published record backs this pair is carried by the score’s own rigor dimension and traced to verbatim sources at the foot of the page.

Hypothesized mechanism

Metformin's insulin-sensitizing effects may reduce hyperandrogenism and central adiposity in PCOS, though it is less effective than COCP at lowering androgen indices.

This is the model’s proposed mechanism from the sources on file, not a demonstrated causal pathway. How well the published record supports it is reflected in the rigor and plausibility dimensions of the score, and traced to the verbatim sources at the foot of the page.

How the score was reached, for this pair

The composite score is the sum of five dimensions, each scored 0 to 2 by the model from the evidence on file. Below is the sub-score this specific pair received on each, with what that dimension measures. It scored 7 of 10 overall, a moderate reading, from a direct rated moderate in strength.

The model’s overall reasoning for this pair is the summary at the top of the page, and the mechanism it proposed is in the section above.

Direct research arm · anchors the headline7.0 / 10 · Moderate

Scored for women. Evidence generated in women (female population, ~100% female). (band F1, ×1.00).

Corroboration

Evidence comes from two systematic reviews/meta-analyses (one on inositol comparisons, one on COCP comparisons), each a single synthesis rather than independent replication of the same comparison. No single large low-bias RCT is described, so this stays at 1.

1 / 2

Rigor

Both sources are systematic reviews and meta-analyses informing the 2023 international PCOS guidelines, which is the highest design tier. Claims 6-8 report pooled effect estimates with confidence intervals.

2 / 2

Specificity

Both metformin and PCOS are named directly throughout (e.g., 'Metformin may improve waist-hip ratio and hirsutism' in PCOS, and 'Metformin and Combined Oral Contraceptive Pills in the Management of Polycystic Ovary Syndrome'). The drug and condition are unambiguous.

2 / 2

Plausibility

Metformin's insulin-sensitizing action plausibly underlies improvements in waist-hip ratio and androgen-related outcomes, but no explicit mechanistic statement is given in the claims. Mechanism is plausible but only implied, not evidenced here.

1 / 2

Consistency

Results are mixed and comparator-dependent: metformin appears better than inositol for waist-hip ratio and hirsutism but inferior to COCP on FAI, SHBG, and testosterone. The directions differ across comparisons rather than agreeing, so consistency is not strong.

1 / 2
Community arm5.0 / 10 · Emerging

Scored for women. Evidence generated in women (female population). (band F1, ×1.00).

Corroboration

Independence (patient accounts): 0 distinct account(s) across 3 thread(s) (author handles unavailable — counting distinct posts).

1 / 2

Rigor

All three claims are vague subjective testimonials with no dose, timing, duration, or specific symptom changes reported. Phrases like 'changed my life' and 'worked a little too well' lack any concrete clinical detail.

0 / 2

Specificity

The drug (metformin) and condition (PCOS) are both named and linked positively, e.g. 'Metformin made me realize how much PCOS actually affected me.' However, the specific outcome/symptom that improved is never stated, weakening the drug-to-outcome link.

1 / 2

Plausibility

Metformin improving PCOS symptoms is highly consistent with its known pharmacology as an insulin sensitizer addressing the insulin resistance central to PCOS. The reported positive effects fit established mechanism.

2 / 2

Consistency

All three accounts agree directionally, uniformly reporting strong positive effects ('changed my life', 'worked well', noticeable impact). There are no contradicting or denying reports, though no dose/timing details exist to assess coherence.

1 / 2
How the scoring rubric works, in general →

Independent reading, reported beside the score

One outside model cross-reference is reported alongside the composite score. It is recorded separately and is not combined into the score.

MATRIX cross-reference

Every Cure’smachine-learned treatment-probability model, drawn from a biomedical knowledge graph across roughly 1,800 drugs and 22,000 diseases. It provides a model-based estimate of how plausible a drug-disease link is given the structure of biomedical knowledge, reported alongside the substrate’s own evidence.

For this pair. MATRIX maps this drug and disease in its graph but returned no treat-score for the pair, which can mean the predicted link fell below the model's publication threshold.

Scored over MATRIX’s own entities, confirming the same drug and disease: CHEBI:6801 (drug) and MONDO:0008487 (disease). Validate against the source: Every Cure’s MATRIX dataset ↗.

More on the MATRIX cross-reference and its provenance →

Layers not covered for this pair

Sex-specific pharmacokineticsNone on file

Not covered for this pair. This layer holds documented sex-specific pharmacokinetics for a limited set of drugs, and this compound is not among them yet. A blank here means the drug is not covered by the layer, not that no sex difference exists.

More on the sex-specific pharmacokinetics layer and its sources →
Cycle-phase dependenceNone on file

Not covered for this pair. The cycle-phase layer is seeded for the strongest-evidence cases so far (PMDD), and this pair is not among them yet. A blank here means the pair is not covered by the layer, not that the effect was found to be phase-independent.

More on the cycle-phase layer and its sources →

Source evidence · what the pipeline ingested

These are the sources the pipeline ingested to detect and score this signal, the published literature the model actually read, each tagged by study type. Where the model combined findings the claim is marked as a synthesis (S), and where the literature disagrees the contradiction is shown (!).

Every source below belongs to this signal’s evidence arm, Direct research. Whel reads each drug-condition pair through four such arms, each held to its own inclusion bar; a signal is surfaced through one of them.

These are the verbatim sources the pipeline surfaced and read; they may not be the full published record for a pair, and the score reflects the strength and agreement of the evidence rather than its volume. The strength of these source types is what the rigor dimension of the score reads off. MATRIX, sex-specific pharmacokinetics, and cycle phase are separate layers the pipeline does not ingest, external cross-references reported beside the score, and they link to their own sources in their sections above.

The primary sources and pipelines this evidence is drawn from →
← Back to the full index