Polycystic ovary syndrome (PCOS) is a prevalent, insulin-resistance-linked reproductive/metabolic condition — it affects 5-20% of women of childbearing age, insulin resistance is present in ~75% of lean and ~95% of overweight women with PCOS, and 60-70% of women with PCOS are overweight (Greff et al., 2023). It enters this wiki on the prevention-and-improvement-by-lifestyle-lever terms (like MASLD / metabolic syndrome / T2D): what an exposure does to a patient-important outcome for the stratum, not the fertility-clinic management of a diagnosed case. The lever with gold SR/MA evidence held so far is inositol, a nutraceutical insulin sensitiser positioned as the folk alternative to metformin (the standard metabolic drug). This page is the nucleus; diet / weight-loss / exercise levers are named gaps below.
Verdict — the surrogates move; the reproductive endpoint that matters does not
The single held gold source (Greff 2023, SR+MA of 26 RCTs / 1691 women; 24 in the quantitative pool) concludes «Inositol is an effective and safe treatment in PCOS» and recommends it be added to PCOS guidelines (Greff et al., 2023). Read against the surrogate-vs-patient-important axis, that headline is a surrogate-driven overclaim: the statistically significant effects sit almost entirely on biochemical and anthropometric surrogates (androgens, SHBG, BMI, fasting glucose, AUC-insulin) plus a very-low-certainty primary outcome (cycle normalization); the one patient-important reproductive outcome the MA pooled — clinical pregnancy — was null vs both placebo and metformin, and live birth and miscarriage were never pooled [INFERRED, from the outcome-by-outcome table below]. The decision-relevant finding is narrower and different: inositol is a plausible metformin substitute for the metabolic surrogates with far fewer GI side effects, on evidence that does not yet reach the outcome a woman seeking fertility actually cares about -> Surrogate Outcomes.
Inositol vs placebo — outcomes ranked by patient-importance
Effect sizes are pooled mean difference (MD) or risk ratio (RR) with 95% CI; GRADE certainty is the MA’s own per-outcome rating (Table 2). Sig = CI excludes the null.
| Outcome (tier) | n studies (pts) | Effect (95% CI) | GRADE | Sig? |
|---|---|---|---|---|
| Pregnancy rate, +adjuvant Rx (patient-important) | 4 (308) | RR 1.24 (0.85; 1.81) | very low | no |
| Pregnancy rate, no other Rx (patient-important) | 1 (42) | RR 3.3 (0.4; 27.13) | low | no |
| Cycle normalization (intermediate, PRIMARY outcome) | 2 (118) | RR 1.79 (1.13; 2.85) | very low | yes |
| BMI (surrogate) | 8 (419) | MD -0.45 kg/m2 (-0.89; -0.02) | high | yes |
| Total testosterone (surrogate) | 6 (284) | MD -20.39 ng/dl (-40.12; -0.66) | moderate | yes |
| Free testosterone (surrogate) | 4 (152) | MD -0.41 ng/dl (-0.69; -0.13) | moderate | yes |
| SHBG (surrogate) | 4 (152) | MD +32.06 nmol/l (1.27; 62.85) | moderate | yes |
| Androstenedione (surrogate) | 6 (198) | MD -0.69 ng/ml (-1.16; -0.22) | moderate | yes |
| Fasting glucose (surrogate) | 5 (266) | MD -3.14 mg/dl (-5.75; -0.54) | moderate | yes |
| AUC-insulin (surrogate) | 4 (132) | MD -2081 uU/ml/min (-2745; -1417) | high | yes |
| HOMA-IR (surrogate) | 3 (92) | MD -0.23 (-1.19; 0.72) | low | no |
| Fasting insulin (surrogate) | 6 (286) | MD -2.78 uU/ml (-7.13; 1.57) | low | no |
| AUC-glucose (surrogate) | 4 (132) | MD -763 (-1925; 398) | moderate | no |
| DHEAS (surrogate) | 4 (152) | MD -92.54 (-206.31; 21.22) | low | no |
| Ferriman-Gallwey / hirsutism (surrogate) | 1 (43) | MD -1.23 (-5.37; 2.92) | low | no |
Two internal tells the headline glosses. (1) The direct insulin-resistance markers did not move. The MA states: «In the case of fasting plasma insulin, HOMA-IR, and AUC-glucose, inositols compared to placebo showed no favorable effect» (Greff et al., 2023). So the significant glycemic signal rests on fasting glucose and AUC-insulin — not on HOMA-IR or fasting insulin, the canonical IR surrogates — which weakens the mechanistic insulin sensitiser story on its own terms (the AUC-insulin subgroup benefit is myoinositol-specific and does not reverse the HOMA-IR null). (2) The primary outcome is very-low certainty. Cycle normalization — the registered primary outcome (down-switched from ovulation/cycle length for lack of data) — is significant but GRADEd very low on only 2 studies / 118 women, so it carries little decision weight -> Rating Certainty of Evidence.
The patient-important gap — pregnancy null, live birth never pooled
Pregnancy was the only patient-important reproductive endpoint pooled, and it is null and untrustworthy: «The overall pregnancy outcome was heterogenous regarding its definition, carrying a significant risk of bias. Only one study reported on the pregnancy rate for the inositol placebo comparison without additional therapy and found no difference (RR = 3.3 CI: 0.4; 27.13)» (Greff et al., 2023). The pooled pregnancy estimates lean on trials where inositol was followed by adjuvant fertility therapy (letrozole, rFSH+HCG), so they do not isolate an inositol effect, and the MA itself concedes «The generalization of pregnancy rate results was problematic since only one study analyzed women who wanted pregnancy» (Greff et al., 2023).
G-gap (patient-important, load-bearing): live birth and miscarriage were never pooled — 0
occurrences across both chunks (searched: live birth / miscarriage, chunks 01-02, corpus zero over
a non-empty denominator). This is a named gap, not a null: the MA is silent on the two
outcomes that decide whether inositol helps a woman conceive and carry, so no fertility claim is
supported at any certainty. The MA’s own next-step ask is «Further well-designed RCTs are needed to
assess the beneficial effect on pregnancy rate» (Greff et al., 2023). -> — a fertility-endpoint SR (live birth /
clinical pregnancy) would move this cell.
Inositol vs metformin — the substitution/Layer-1 comparator
Metformin is PCOS’s standard metabolic drug but induces GI side effects (nausea, diarrhea, vomiting, flatulence) (Greff et al., 2023). Because a mature standard drug is the realistic alternative, the decision-relevant question is not inositol-vs-nothing but inositol-vs-metformin -> Layer 1 - Ranking Interventions for a Stratum. Head-to-head (Table 3):
| Outcome | n studies (pts) | Effect (95% CI) | GRADE | Read |
|---|---|---|---|---|
| HOMA-IR | 6 (419) | MD -0.18 (-0.41; 0.06) | high | non-inferior |
| BMI | 9 (593) | MD -0.11 kg/m2 (-0.25; 0.04) | high | non-inferior |
| Fasting insulin | 6 (427) | MD -0.37 (-1.52; 0.78) | high | non-inferior |
| Total testosterone | 4 (320) | MD 0.2 (-5.72; 6.12) | moderate | non-inferior |
| SHBG | 3 (220) | MD +2.78 nmol/l (0.02; 5.54) | moderate | slight inositol edge |
| Ferriman-Gallwey / hirsutism | 3 (220) | MD 0.6 (0.24; 0.96) | high | metformin better |
| Cycle normalization | 6 (424) | RR 1.42 (0.8; 2.53) | very low | non-inferior (NS) |
| Pregnancy rate | 5 (383) | RR 1.22 (0.84; 1.78) | very low | non-inferior (NS) |
| Side effects | 4 studies | RR 0.16 (0.09; 0.28) | — | inositol far fewer |
The one decision-changing head-to-head finding is tolerability: «four studies comparing inositol to metformin showed a lower rate of side effects in the inositol group (7 vs. 53%, RR = 0.16, CI: 0.09; 0.28)» (Greff et al., 2023). On metabolic surrogates inositol is broadly non-inferior to metformin (high certainty for HOMA-IR / BMI / fasting insulin), but metformin is better on hirsutism (FG-score) and both are null on the reproductive endpoint. So the net-of-alternative case is: for a woman who cannot tolerate metformin’s GI effects, inositol delivers similar metabolic-surrogate movement with a much lower side-effect burden — a genuine substitution finding, bounded to surrogates. The choice stays the person’s (layer 3); this only sizes the marginal rock.
Myo-inositol vs D-chiro-inositol — isomer and a dose caution
The two clinically-studied isomers behave differently, so inositol is not one exposure -> Is the Food Category Doing Any Work. Myo-inositol carried the most-supported and best-GRADEd benefits — BMI -0.71 kg/m2 (-1.00; -0.43, high), fasting glucose -4.03 mg/dl (-6.59; -1.47), AUC-insulin -2034 (high), cycle normalization 1.76 (1.06; 2.92) — while its pooled pregnancy rate was null (3 studies, RR 0.92, 0.53; 1.61, very low) (Greff et al., 2023). D-chiro-inositol (DCI) drove the largest androgen/SHBG shifts (total testosterone -41.71, free testosterone -0.58, SHBG +55.45, DHEAS -168.48; several high/moderate) but on very few, small trials. Trial doses were typically myo-inositol ~2000-4000 mg/day or DCI ~600-1200 mg/day over 7-24 weeks (Greff et al., 2023).
DCI carries a dose/duration caution (mechanism-flagged, route-c flavor): «DCI is an aromatase inhibitor and promotes glycogen synthase, which inhibits the conversion of androgens to estrogens, resulting in the accumulation of androgens and lack of estrogens. That is why long-term or high-dose DCI administration will worsen the symptoms of PCOS» (Greff et al., 2023). The MA reports no adverse androgen effect at 6-8-week DCI dosing, but this is a mechanistic/short-term claim, not a long-term safety finding — the pooled evidence cannot confirm or refute the long-term-DCI-harm direction, so it stays a directional caution, not a settled harm.
Layer-1 placement and named gaps
Where inositol sits in the PCOS lever hierarchy: it is a surrogate-moving, well-tolerated metformin-alternative whose patient-important reproductive benefit is unestablished — a modest, adjunct-grade rock, dominated for metabolic risk by the same established levers that govern insulin resistance generally -> Insulin Resistance Surrogates and Cardiovascular Risk. It is the PCOS sibling of Cinnamon and Glycemic Control: a supplement with real but surrogate-only signal and no held patient-important-outcome evidence.
Named gaps (G) driving the next PCOS source choices:
- Weight loss / diet as first-line — the standard first metabolic lever for overweight
PCOS is weight loss, but no gold SR on diet/weight-loss -> PCOS incidence/remission or ovulation is
held. — directional-only,
confidence: low. - Low-GI diet vs weight-loss-does-the-work in PCOS — separating the diet composition from the energy-deficit effect..
- Exercise -> PCOS metabolic/reproductive outcomes — unheld..
- Fertility endpoint (live birth / clinical pregnancy) for inositol — the reproductive cell above..
Provenance, quality, and conflicts
Greff 2023 is a gold SR+MA (26 RCTs, RoB 2, GRADE, PROSPERO-registered CRD42021283275); most outcomes were GRADEd moderate, the reproductive and cycle-normalization outcomes very low (Greff et al., 2023). Its own stated limits: small number of small, heterogeneous trials; differing follow-up and doses; heterogeneous pregnancy definition; missing AUC forest plots; moderate-to-high RoB in some domains (Greff et al., 2023). Funding was public/academic (Hungarian NKFIH-FK129206, Semmelweis University, Hungarian Hypertension Society; funder had no role), and the authors declare no competing interests (Greff et al., 2023) — so the meta-analysis is not manufacturer-linked, though the underlying inositol RCTs’ funding was not appraised here.
Synthesis
This is a
provisional-C nucleus — a single gold source opens the PCOS cluster and the decision question; the banked cross-source A/F moves are OWED when a second PCOS lever lands. Within this one
source, the move beyond its own summary is a surrogate-vs-patient-important re-sorting: the MA
presents one flat beneficial on all aspects verdict; sorted by outcome tier it becomes a
surrogate-strong, patient-important-null result with a very-low-certainty primary outcome and a
live-birth/miscarriage hole the MA never names as such. That re-read uses the MA’s own GRADE ratings
and tables against its headline — an intra-source critical reading, not a cross-source tension. The
genuinely decision-changing residue is the metformin-substitution-on-tolerability finding (7% vs
53% side effects, non-inferior surrogates), not the fertility framing the title invites.
[E-independent] is not claimed — this is a single source; the sorting and the gap are the value,
and both are reproducible-checkable against the MA’s own tables.