Hyperandrogenism in PCOS and the putative postmenopausal deficit: same axis, different dysfunctions

Cutoff date: 2026-09-07
Stage: REPORT_EN
Project: L8-6 — Adrenal androgens and DHEA
Status: complete scientific report, version 1
Next stage: BILINGUAL_QA

1. Executive scientific abstract

The comparison “hyperandrogenism in PCOS versus postmenopausal androgen deficiency” contains a useful intuition—androgen signaling matters across the life course—but its scalar form does not withstand the evidence. Polycystic ovary syndrome (PCOS) is not simply “too much androgen,” postmenopause is not simply “too little androgen,” and neither DHEA-S nor total testosterone represents androgen receptor (AR) activity on its own. Steroid families age at different rates: DHEA-S, DHEA, androstenedione, and classical testosterone usually decline, whereas 11-ketotestosterone (11KT) may be preserved. Some women with PCOS retain hyperandrogenism after menopause, although cohorts of very old survivors show convergence in testosterone and FAI. Life stage changes the distribution of inputs; it does not assign a woman’s molecular state.

This synthesis proposes replacing the mirror with a prereceptor partition topology. In adipose tissue, AKR1C3 can convert 11-ketoandrostenedione (11KA4) into 11KT and favor AR signaling. HSD11B1 can reduce 11KT to 11OHT, attenuating that branch, while converting cortisone into cortisol and favoring glucocorticoid receptor (GR) signaling. At the same time, aromatization, estrogen receptor (ER) signaling, inflammation, SHBG, liver function, insulin, and cellular composition alter the effective dose and functional direction. The same net function can represent three different states: chemistry without a material consequence, one active branch, or two AR/GR contributions that oppose or interact. A serum correlation or isolated HSD11B1 inhibition cannot resolve that distinction.

Human evidence supports with moderate confidence that the 11-oxygenated profile differs from the classical profile and that PCOS is associated with metabolic and vascular risk in some contexts. It does not establish that 11KT causes adipose insulin resistance, that postmenopausal androgen deficiency is universal, or that normalizing a concentration increases health or longevity. The closest human counterexample is small but important: in 20 women with PCOS, seven days of DHEA or 11KA4 changed the steroidome without a statistically significant detected change in clamp M-value. This is not an equivalence test, but it elevates biochemical conversion without a measurable material function to the top-ranked hypothesis.

The adversarial review removes two shortcuts. First, the published genetic signal linking bioavailable testosterone to type 2 diabetes is sensitive to SHBG and does not demonstrate a testosterone effect independent of transport. Second, late-life null findings for morbidity and mortality in PCOS come from small, selected cohorts; they prove neither protection nor equivalence, just as excess events in younger cohorts do not establish androgen mediation. The chain to longevity remains open: steroid mixture → receptor → organ-specific function → event → years free of disease/disability → survival.

The minimum experiment does not initially compare PCOS with postmenopause and does not start by inhibiting an enzyme. It first qualifies metrology and function; it then independently varies 11KT and cortisol in a 2×2 factorial in female human adipocytes, with fixed E2/TNF, measured target engagement, and a frozen primary endpoint: change in fractional suppression of net NEFA efflux at an insulin concentration selected in an independent pilot. Equivalence requires a CI90% within ±0.10 absolute units, not p>0.05. Only if both branches are material are AR and GR loss/rescue tested, followed by AKR1C3/HSD11B1 flux with tracers and bypass. This design can kill the dual topology if one branch is equivalent and can close a pharmacological opportunity early if chemistry changes without function.

Central conclusion: the “same axis” exists as a network of supply, transport, conversion, and receptors; the “same dysfunction in opposite directions” has not been demonstrated. The most parsimonious explanation today is that much of the systemic contrast reflects hormonal composition, antecedent metabolism, and parallel organs, with narrower and more contextual local functions. The most informative hypothesis—two separable AR/GR contributions—remains H0 and must be earned experimentally.

2. Scientific question and relevance

Question: what mechanisms connect PCOS hyperandrogenism and the putative postmenopausal androgen deficit with women’s health and longevity, and what new hypotheses can be falsified?

The question decomposes into four estimands that must not be merged:

1. Life-course composition: whether classical and 11-oxygenated androgens change concordantly within the same woman around the final menstrual period.
2. Local causality: whether 11KT and cortisol, at measurable free exposure, alter the same adipose function through AR and GR.
3. Prereceptor partition: whether AKR1C3/HSD11B1 modify that function in the direction predicted by their products, beyond chemical target engagement.
4. Systemic translation: whether that function subsequently mediates type 2 diabetes (T2D), a vascular outcome, or another defined endpoint, and only thereafter years free of disease or survival.

The relevance to longevity derives from possible effects on lipid distribution and mobilization, glucose uptake, liver function, muscle, vasculature, brain, bone, and reproduction. These organs may respond in different directions. Therefore, there is no universal valence in which “more androgen is better” or “less androgen is better,” and no defensible longevity composite can be derived from fertility, BMD, a clamp, or mortality in isolation.

3. Scope, population, and life stage

The human contrast includes adult women with documented PCOS during reproductive age, women undergoing the menopausal transition, and women after natural menopause, with and without a history of PCOS. Molecular exposure is measured; it is not assigned by diagnosis or stage.

Operational populations:

• Biochemically characterized hyperandrogenic PCOS, approximately 18–40 years, distinguishing clinical features, classical androgens, and 11-oxygenated androgens.
• Reproductive controls recruited using criteria independent of the metabolic endpoint.
• Menopausal transition with adjudicated final menstrual period (FMP) and repeated within-woman measures.
• Early natural postmenopause, approximately 1–10 years after FMP, with later postmenopause analyzed separately.
• Documented, documented-absent, or indeterminate PCOS history; PCOS must not be reconstructed solely from recalled cycles or current testosterone.

Surgical/induced menopause, premature ovarian insufficiency, pregnancy/lactation, secreting tumors, congenital adrenal hyperplasia, and advanced hepatic or renal disease belong in separate strata or benchmarks. Hormone therapy, contraceptives, glucocorticoids, antiandrogens, aromatase inhibitors, and metabolic treatments are recorded by compound, route, and timing; discontinuation is not requested. The research uses no current Lua data, PHI, private genomes, or user samples.

The prioritized proximal tissue is non-tumoral human abdominal SAT. SAT does not represent VAT, liver, muscle, breast, bone, ovary, or brain. A negative result in SAT lowers the priority of the adipose branch; it does not establish nullity across the entire axis. Mexico/LATAM is a future transportability population, not a magnitude extrapolated from European or US cohorts.

4. Background knowledge and mechanism map

4.1 Sources and C19 families

The ovary and adrenal contribute classical C19 precursors; the adrenal is an important source of DHEA/DHEA-S and 11-oxygenated steroids. Sulfation of DHEA by SULT2A1 depends on PAPS generated by PAPSS2; STS can release DHEA. Low DHEA-S can therefore represent lower supply, lower sulfation, greater desulfation/use, or greater clearance. The human PAPSS2 case is a causal counterexample: very low DHEA-S coexisted with androgenization, so DHEA-S cannot be treated as AR dose.

In a cross-sectional comparison of 200 women, classical androgens were lower in the group aged ≥60 years, whereas 11KT was not detectably lower and 11OHT was higher. This design separates age and stage and does not demonstrate an individual trajectory, but it invalidates the premise that all families follow a single decline. SWAN, in turn, modeled a transitional increase in DHEA-S in many women superimposed on a general age-related trend. “Adrenopause” is therefore not a simple synchronized decline either.

In reproductive-age PCOS, 11-oxygenated steroids represent a large fraction of the measured molar pool and several are elevated, but that pool includes precursors and does not equal the percentage of AR activity. A clamp study found associations in different directions: 11KT/11OHT were directly associated with insulin sensitivity, whereas free testosterone was inversely associated. This cross-sectional association does not justify calling 11KT protective; it does refute use of a single androgen sum with a fixed metabolic sign.

4.2 Transport, liver, and reverse causality

SHBG and albumin modify availability; liver and kidney modify binding-protein synthesis and clearance. In HepG2 models and mice carrying a human SHBG transgene, monosaccharide-induced lipogenesis reduced HNF4A and SHBG transcription. This is a preclinical competing mechanism: energetic/hepatic state may alter metabolic function and the available hormonal fraction simultaneously. It does not show that insulin universally suppresses SHBG in women.

The genetics study by Ruth et al. links bioavailable testosterone with PCOS and T2D, but the T2D signal did not persist with the testosterone-specific instrument independent of SHBG. The correct inference is neither “SHBG causes everything” nor “testosterone is irrelevant,” but that transport and exposure were not cleanly separated for T2D.

4.3 Prereceptor conversion

AKR1C3 reduces A4 to testosterone and 11KA4 to 11KT. In recombinant human protein and SGBS adipocytes, catalytic efficiency for 11-keto substrates was approximately four times that for classical substrates, and 11KT formation responded to AKR1C3 inhibition. This is DI evidence of capacity, not a quantification of contribution in intact female SAT.

HSD11B1 combines two logics:

• 11KT → 11OHT, which may reduce relative AR signaling;
• cortisone → cortisol, which may increase GR signaling.

Schiffer et al. observed that HSD11B1 inhibition reduced GR activation and increased AR activation in reporters, and increased 11KT in adipose explants; the 2025 editorial correction must be used for the affected figure. The clinical subset was small and predominantly male and did not measure the adipose function of interest. The topology is supported; functional cancellation is not.

Moreover, 11KT is not an AR-pure intervention by definition: aromatase-containing systems can generate 11-oxygenated estrogens and ER signaling. The corresponding study did not detect a material contribution to the circulating estrogen pool, but that finding does not exclude tissue microenvironments. Any attribution to AR requires product measurement and receptor-specific loss/rescue.

4.4 Receptors, cells, and context

AR, GR, and ER share some cofactors and regulatory motifs. A generic ARE/GRE reporter does not separate receptors. Receptor loss may also alter differentiation, ligand metabolism, or cellular state. Target engagement must precede function and be measured with orthogonal readouts: translocation/occupancy, a specific transcriptional panel, and quantified loss-rescue.

In adipocytes derived from SAT of healthy women, chronic testosterone reduced maximal insulin-stimulated glucose uptake, and AR antagonists attenuated the effect. Nominal concentrations and pharmacology prevent direct transport to 11KT, PCOS, or postmenopause, but they refute a universal hormonal null. In another human system, TNF shifted the antilipolytic curve and dexamethasone modified the effect; GR has no fixed direction outside inflammatory context.

4.5 Adipose function

“Lipolysis” is not one measurement. The following must remain separate:

• unstimulated basal release;
• capacity under β-adrenergic stimulation;
• suppression by insulin at a prespecified concentration;
• inhibitable amplitude, residual, slope, and curve position only if identifiable;
• glycerol, NEFA, re-esterification, and glucose uptake.

Jönsson et al. showed in female human adipocytes that the insulin response can be biphasic and that glycerol and NEFA diverge because of re-esterification. Lee and Fried measured insulin at 0, 30, 120, and 600 pM against 8-bromo-cAMP-stimulated lipolysis; the zero point is not unstimulated basal release. Less glycerol may mean lower mobilization, greater re-esterification, or lower capacity—not automatically better sensitivity.

4.6 From tissue to health and longevity

The candidate chain is:

ovarian/adrenal source + liver/SHBG/clearance
                         ↓
             free mixture reaching tissue
                         ↓
  AKR1C3: 11KA4→11KT→AR ───────────┐
  HSD11B1: 11KT→11OHT              ├─→ adipose response to insulin
  HSD11B1: cortisone→cortisol→GR ──┘       ↑
                       ER + TNF + cellular identity
                                         ↓
                systemic lipid and glucose balance
                                         ↓
                    T2D or specific vascular event
                                         ↓
          years free of disease/disability and survival

The chemical links are better supported than the functional links; SAT→T2D and T2D/CVD→survival mediated by this mixture are unproven. Ovary, brain, muscle, liver, and bone may act in parallel and generate trade-offs.

5. Evidence method

The scientific charter, retrieved lifetime memory, and all prior project artifacts were read. The synthesis accumulated a directed map of primary sources and sought adversarial evidence. The update on September 7, 2026 again located in primary sources the Dokras et al. article published online in August 2026 and the March 2026 Schiffer et al. work still as preprint v1; a negative search is not interpreted as proof that no other version exists worldwide.

The following categories were used:

• DH: human observational, genetic, or interventional evidence.
• DX: ex vivo human tissue.
• DI: in vitro, recombinant protein, cell line, or cultured human cell.
• DA: animal.
• CP: published computation or Lua Labs simulation with known truth.
• I: new inference still requiring a test.

For decisive sources, design, population/life stage, sample, exposure, outcome, sample size, direction, and limitations were extracted. Discrepancies were preserved: glycerol numbers whose prose contradicts the figure, an adjusted estimate with a point estimate outside its CI, the effect of SHBG in genetics, and the Schiffer correction. Conference abstracts and preprints are labeled and do not receive the same weight as peer-reviewed articles.

The search was directed rather than a comprehensive systematic review. Heterogeneous studies were not combined in a new meta-analysis. Bars were not digitized to invent numbers, and CIs were not reconstructed from p-values. Causality requires perturbation, temporality, or a valid causal design; association and chemical capacity are insufficient.

6. Evidence map

No human, animal, or computational evidence was identified that closes the complete chain from androgen composition to female longevity. This absence is a gap in the directed corpus, not proof of worldwide nonexistence.

7. Contradictory evidence and null findings

7.1 Persistent hyperandrogenism versus late convergence

Markopoulos finds higher androgens in postmenopausal women with PCOS; Forslund finds convergence of T/FAI in survivors at approximately 81 years. These findings are not incompatible if excess attenuates with age, if 11KT remains unmeasured, if treatments and selection change the cohort, or if heterogeneity exists. The smallest discriminator is repeated within-woman measurement of free T, DHEA-S, and free 11KT—not comparison of studies at different ages.

7.2 11KT and metabolic function

11KT is elevated in some PCOS cohorts, but Tosi observed a direct association with sensitivity while free T was inverse; McDonnell changed steroid branches without a statistically significant detected clamp change. Ranked explanations are: (1) conversion without a material function or an exposure window that is too short; (2) confounding by adiposity/SHBG/organ; (3) different functions of different ligands; and (4) insufficient power. The downstream factorial and formal equivalence distinguish the first three; another cross-sectional study would not.

7.3 Elevated AKR1C3 versus comparable protein

O’Reilly demonstrates contextual conversion and adipose changes; Dumesic does not find higher AKR1C3 protein in normal-weight PCOS. RNA/protein, activity, substrate, adiposity, and cellular state may be discordant. The discriminator is traced flux per viable mass at equal substrate with loss-rescue, not another expression correlation.

7.4 Favorable versus null DHEA findings

Villareal observed aggregate improvements; Nair did not observe broad female functional benefits despite restoring DHEA-S. The studies differ in duration, population, endpoint, sex-specific estimation, and precision. This neither validates replacement nor eliminates a local effect; it requires comparable female estimates and prevents use of hormonal increase as a functional surrogate.

7.5 Younger vascular risk versus late events/mortality

Berni finds a moderate excess of a vascular composite in a predominantly young population. Schmidt/Forslund do not detect higher MI, stroke, or mortality in small older cohorts. Possible explanations include survivor selection, outcome definition, obesity, treatment, power, phenotype change, and competing risks. They must not be informally averaged, and a late null must not be converted into protection.

7.6 Error and editorial quarantine

The graphical direction of glycerol in O’Reilly 2017 contradicts the printed group values; only the figure/author direction is retained. In Kugelman 2025, aOR 2.92 appears with CI95% 1.06–2.82, which is impossible because the point estimate lies outside the interval; the adjusted magnitude is not used pending correction. Schiffer 2024 is interpreted with its 2025 correction. Preserving these errors prevents fabricated precision.

8. Multiscale mechanistic synthesis

Molecular

The relevant information is not an isolated concentration but the vector DHEA-S, DHEA, A4, T, DHT, 11OHA4, 11KA4, 11KT, 11OHT, cortisone, cortisol, E1, and E2, their free fractions, and temporal histories. PAPSS2/SULT2A1 and STS control storage/delivery; AKR1C3 controls activation of classical and 11-keto branches; HSD11B1 simultaneously redistributes androgens and glucocorticoids; CYP19A1/HSD17B modify ER signaling.

Cellular

Intracellular dose depends on transport, binding, adsorption, and metabolism. AR/GR/ER may alter genes controlling storage, mobilization, uptake, and cellular state, while sharing motifs and cofactors. A transcript or reporter does not identify which receptor caused a function. Post-maturation loss, resistant rescue, and ligand remeasurement are required to avoid mistaking signaling for dedifferentiation.

Tissue

SAT contains adipocytes, progenitors, endothelium, immune cells, and stroma. TNF, adipocyte size, and surgical source alter response. A homogenate AKR1C3:HSD11B1 ratio combines abundance, activity, cofactors, and composition. The net NEFA endpoint integrates lipolysis, re-esterification, and export; glycerol and glucose uptake decompose branches and do not replace the primary endpoint.

Systemic

Insulin and adiposity may modify AKR1C3 and SHBG, while steroids may modify adipose tissue, liver, muscle, and brain. The system contains feedback and reverse causality. Automatically adjusting for BMI/insulin/SHBG may block mediation; not adjusting may leave common cause. The DAG and timing must be fixed for each estimand.

Life course

FMP strongly changes E2/FSH, but does not synchronize all C19 branches. PCOS history shifts distributions without guaranteeing lifelong excess. Aging, treatments, renal/hepatic function, and survival select which women enter late-life cohorts. “Postmenopause” is a stage, not a state of low AR activity.

Longevity

Long-term health can be attributed only after demonstrating persistent function and mediation of an outcome. T2D, CVD, cancer, bone, cognition, physical function, and mortality are distinct domains with potentially opposite signs. No post hoc composite is created to rescue a hormone with discordant effects.

9. Computational layer

Conventional simulation was used, not BioNeMo, new genetics, or transcriptomics. The objective was to determine whether the design could distinguish scale, amplitude, curve position, and AR/GR cancellation. The public anchor was the Lee and Fried architecture [0,30,120,600] pM; biological data were not digitized. Hill curves were simulated:

L(c)=R+A/[1+(c/K)^h]

with known truth, additive/proportional error, donor-level heterogeneity, and synthetic n values. Decisive results:

• D4 [0,30,120,600] was unstable for K with free h; in the central scenario K=12 pM, n=12, and CV=15%, 56.03% of individual fits reached a boundary under a scale change and only 39.5% detected K×2.
• D7 [0,3,10,30,60,120,600] improved range coverage, but an isolated reduction in A induced ΔlogK=0.1264 bias, exceeding the 0.1 operational criterion.
• Fixing h=1 to gain precision is invalid: when the truth was h=0.5, D7 biased ΔlogK by −0.2325 and detected K×2 only 14.5% of the time.
• With D7/free h, detection of K×2 ranged from 100% at CV=5% to 17% at CV=30%; n=12 does not establish equivalence.
• A coupled AR/GR perturbation classified both true nullity and exact cancellation as “null” in ~95% of 10,000 simulations.
• The downstream 2×2 recognized exact cancellation in 95.01% and generated 0.06% false cancellation calls under nullity, but confused partial cancellation with interaction when a simple net rule was applied.
• Treating 70% receptor loss as complete underestimated the component by 30%; cross-ligand changes generated illustrative bias up to −50%.

Consequence: D7 is only a qualification grid. The confirmatory study uses a functional contrast at a prespecified c* unless pilot data identify K/h without boundary estimates. The 2×2 factorial is necessary, but interaction, a net-effect margin, actual receptor loss, and intracellular ligands remain mandatory. The simulation changes the design; it does not raise any biological hypothesis to H2.

10. Primary hypothesis

L8-6-AR-H2L v1 — conversion/engagement without a material functional effect

Role: primary by ranking.
Status: strengthened_as_rival, not validated.
Maturity: H1 for chemistry–function discordance; H0 for causal equivalence in SAT.
Confidence: higher than the report’s alternatives by parsimony, not to be interpreted as a probability.

Falsifiable statement: within the free range observed or justifiably attainable in female human adipocytes, 11KT and cortisol will produce measurable AR/GR engagement, but their main effects and interaction on fractional NEFA suppression will remain within the material margin ±0.10, with fixed E2/TNF and cellular state.

Lineage: L8-6-HG-H2 v1 split by adversarial review; E30/E31; L8-6-COMP-C1/C3; E04, E13, corrected E15; L8-6-AR-C2; L8-6-ED-C2/C3.

Mechanism: local conversion is a capacity or correlate whose magnitude is insufficient to move the endpoint; liver/SHBG, adiposity, antecedent metabolism, and parallel organs explain much of the association. A molecular signal or change in a secondary branch may exist without materially altering the primary endpoint.

Predictions:

1. LC–MS/MS and occupancy/transcription will demonstrate exposure and engagement for both ligands.
2. The CI90% values for δA (11KT), δG (cortisol), and δAG will lie within [-0.10,+0.10].
3. AR or GR loss/rescue may move proximal readouts without a concordant change in the primary endpoint.
4. AKR1C3/HSD11B1 will change products, but downstream bypass will not uncover a hidden material function.
5. In humans, a mixture model will not improve external error/calibration after prior function, adiposity, liver/SHBG, and treatment.

Evidence for: E30 changed steroids without a statistically significant detected clamp change; E04 did not detect greater group insulin resistance after menopause; Nair restored DHEA-S without broad functional benefit; E15 loses the T2D signal when T is separated from SHBG. Evidence against: Corbould demonstrates AR capacity to alter uptake; E21 from the testosterone/E2 program and animal models preserve effects in other contexts and organs.

Kill criteria: a reproducible material effect of direct ligand with valid free exposure and receptor-specific loss and rescue; tracer+bypass linking precursor→product→receptor→function; or a hormonal trajectory that precedes and mediates an out-of-sample human outcome. p<0.05, a transcript, or a cross-sectional correlation does not kill it.

Discriminating experiment: ED1-DOWNSTREAM, followed only if positive by ED2-RECEPTOR and ED2-FLUX.

11. Competing hypothesis

L8-6-AR-H1 v1 — dual AR/GR contribution to the same function

Role: mechanistic competitor.
Status: weakened_reformulated.
Maturity: H0.
Confidence: low; the broad equifinality version was superseded because it could not be killed.

Falsifiable statement: within a prespecified free-exposure window and a single E2/TNF context, 11KT→AR and cortisol→GR will each produce a receptor-specific effect greater than ΔF=0.10 on the same functional endpoint, and HSD11B1 perturbation will move that endpoint in the direction prospectively predicted by measured downstream fluxes.

Lineage: superseded L8-6-HG-H1 v1; L8-6-MECH-H5 v1; E26/E28; L8-6-COMP-C2; L8-6-AR-C1/C7; L8-6-ED-C2/C4.

Mechanism: AKR1C3 generates 11KT and AR signaling; HSD11B1 consumes 11KT toward 11OHT and generates cortisol from cortisone. Within a fixed context, both branches contribute materially to the same output through additivity, opposition, or interaction. The enzyme matters functionally only if product changes prospectively predict the direction without refitting.

Predictions:

1. δA and δG will lie outside equivalence and replicate their signs in another batch.
2. AR loss will selectively eliminate the 11KT effect and rescue will restore it; likewise for GR/cortisol.
3. Product and engagement will change before function.
4. With traced 11KA4/cortisone, HSD11B1/AKR1C3 will move measured fluxes and the endpoint will follow the prediction frozen from the downstream factorial.
5. Equalizing intracellular 11KT and cortisol will eliminate the enzyme-attributable fraction.

Evidence for: AKR1C3/HSD11B1 biochemistry; AR/GR divergence in reporters and explants; glucocorticoid modification of antilipolysis in a TNF context. Evidence against: the links come from different systems; E30 favors chemistry without function; no dual demonstration exists in the same female SAT.

Kill criteria: one branch remains within the margin with adequate precision; loss/rescue is not selective; chemistry/engagement changes but the primary endpoint does not; the HSD11B1 direction contradicts products; the required exposure lacks a human bridge; or the effect depends on toxicity, one tool, or one batch. A single branch is not “partial support”: it kills this dual version and requires a new ID.

Discriminating experiment: the same ED1, followed by ED2 only if both branches are material.

Secondary tests that are not merged with the competing hypothesis

• L8-6-AR-H4 v1: free 11KT changes less than free T/DHEA-S around FMP. H1 cross-sectional/H0 longitudinal; tested with ED3-FMP.
• L8-6-AR-H2S v1: common cause/parallel organs explain the systemic association. H0–H1 observational; tested through bidirectional temporality and incident T2D in ED4-SYSTEM.
• L8-6-AR-H3 v1: a shared U-shaped AR-activity curve. H0, parked; reactivated only if real tissue extremes and direct ligand function are established.

12. Translational hypothesis

L8-6-AR-HT1 v1 — experimental topology qualification

Role: methodological translation, not a clinical biomarker.
Status: weakened_parked.
Maturity: H0.
Requirement: HUMAN_QA_REQUIRED before any human extension.

Falsifiable statement: only after causality and metrology are validated will chemical and receptor variables measured before the endpoint distinguish, in donor- and laboratory-level holdouts, known perturbations of low input, AKR1C3 loss, HSD11B1 partition, and AR/GR loss better than DHEA-S, total testosterone, or diagnosis alone.

Lineage: L8-6-HG-HT1 v1; L8-6-AR-H2L/H1/H4; L8-2-K6/K7; E11/E18/E28/E31; L8-6-COMP-C2; L8-6-ED-C7.

Mechanism/signatures: low input reduces precursor/products; AKR1C3 loss preserves 11KA4 but reduces 11KT and permits bypass; HSD11B1 partition changes 11KT/11OHT and cortisol/cortisone; receptor loss preserves product but eliminates function and is rescued only by a resistant receptor.

Predictions: frozen confusion matrix and thresholds; pre-endpoint variables only; preserved calibration in another batch/laboratory; gain over absolute analytes/context; labels defined by an independent intervention, not by the outcome.

Evidence for: the perturbations have distinguishable chemical signatures, and the computation shows why a coupled trajectory is insufficient. Evidence against: no integrated assay, natural-state reference truth, or external replication exists; the classifier may learn dose/batch.

Kill criteria: failure of identity, recovery, ICC/CV, or stability; absence of causal function; separation explained by batch/dose; no independent label; no improvement in external calibration; or death of the corresponding biological mechanism.

Discriminating experiment: ED6-METROLOGY, a blinded benchmark after causality, with training and holdout separated by donor and laboratory. Until then it is not called a biomarker, test, or clinical opportunity.

13. Integrated falsifiable predictions and kill criteria

14. Discriminating experiment

ED0-Q — qualification

Non-tumoral female abdominal SAT obtained for an independent indication. Fresh mature adipocytes for function and differentiated female ASC/SVF for post-maturation perturbation. Donor is the unit; wells are averaged. Aliquots are randomized, plates balanced, and the operator blinded.

A 12-donor run-in is used for feasibility/variance, not hypothesis testing. Test D7 [0,3,10,30,60,120,600] pM under isoproterenol with a separate basal condition. Choose c* as the lowest concentration producing median NEFA suppression of 40–60%, a descending slope, and an antilipolytic direction in ≥75% of donors. If none exists, repeat qualification with a new range; do not select c* inside the confirmatory study.

Gates: LC–MS/MS bias/CV ≤15%, recovery 80–120%, LLOQ below the low level; viability ≥85%; PLIN1/ADIPOQ markers and lipid content without >20% change; isoproterenol raises NEFA/glycerol ≥2× basal; technical CV ≤15%; MDC95≤0.10; CRISPRi with ≥70% protein loss, two concordant guides, and 80–120% rescue. If Q0 fails, the system is non-identifiable; no biological hypothesis gains support.

ED1-DOWNSTREAM — minimum test

A new cohort of at least 30 analyzable donors, recalculated using run-in covariance. Paired factorial low/high 11KT × low/high cortisol, with fixed E2 and TNF. Levels reproduce a traceable human intracellular range; if no bridge exists, they are labeled experimental.

Temporal order:

1. 0–1 h: free/intracellular ligand and AR/GR translocation.
2. 2–6 h: receptor-specific occupancy and frozen transcriptional panel.
3. 24 h: primary F_NEFA = 1 − NEFA(c*)/NEFA(0) under isoproterenol; glycerol/re-esterification and uptake as decomposition/orthogonal readouts.

Mixed model with donor intercept and 11KT, cortisol, and interaction effects. AR-H2L requires CI90% values for δA, δG, and δAG within ±0.10 plus valid engagement. n=30 is a conditional minimum: if paired SD exceeds 0.14, MDC95>0.10, or exact simulation requires >48 analyzable donors, improve the assay before scaling.

ED2-RECEPTOR and ED2-FLUX

Only for material branches. Post-maturation inducible CRISPRi of AR/NR3C1 with two guides, resistant rescue, and ligand remeasurement. Minimum 12 donors and a second batch, recalculated for precision. Then, only if both branches are causal: separate tracers for 11KA4 and cortisone, reduced AKR1C3/HSD11B1, genetic rescue, and 11KT/cortisol bypass. Measure cell+medium+conjugate balance and require product→occupancy→function ordering.

AKR1C3 requires separation of catalysis from possible AR stabilization suggested in SGBS: catalytically competent rescue versus a catalytic mutant with comparable abundance/localization. HSD11B1 is first studied by separate substrate; its null is cancellation only if both branches have already been demonstrated.

ED3-FMP — independent life-course composition

Archived natural-menopause cohort with four visits: two between −3 and −0.5 years and two between +0.5 and +3 years of FMP. Measure 11KT, T, DHEA-S, A4, 11OHA4/11KA4, SHBG/albumin, E1/E2, liver/kidney function, and treatments. Free 11KT by equilibrium dialysis or a validated method. Estimands Δ11, ΔT, ΔD, C_T=Δ11−ΔT, and C_D=Δ11−ΔD; external replication mandatory. This does not establish engagement.

ED4-SYSTEM — systemic causality

Only after a local signal or replicated longitudinal composition. Women aged 40–65 years with serial measurements and incident T2D as the primary outcome; death as a competing risk. Compare mixture(t)→function(t+1) with function/metabolism(t)→mixture(t+1), using a frozen DAG and total/direct estimands. Event-driven cohort, at least 200 events for a reduced model as a conservative gate, plus external replication. CVD, cancer, bone, and mortality require separate studies.

Stop rules

• Do not replace the primary endpoint with glycerol, uptake, or pAKT after a null.
• Do not call p>0.05 equivalence or call a null net effect cancellation.
• Do not count wells or nuclei as donors.
• Do not continue if effective exposure lacks a human bridge or if toxicity/drift occurs.
• Do not open the U-shaped curve or classifier until extremes and causality are demonstrated.
• A positive ex vivo finding does not establish T2D, CVD, or longevity.

15. Biomarkers and stratification

No biomarker has been validated in this project. DHEA-S, total T, FAI, 11KT, enzyme ratios, AKR1C3/HSD11B1 expression, and stage are analytes, proxies, or experimental candidates, not biomarkers of function or longevity.

Future stratification must separate:

• STRAW+10, age, and years since FMP;
• documented/absent/indeterminate PCOS;
• free T and 11KT, DHEA-S, and full mixture;
• BMI, adipocyte size, and SAT/VAT;
• liver/renal function, SHBG/albumin, insulin/glycemia;
• hormonal, metabolic, and glucocorticoid treatment;
• surgical source, ischemia, batch, and center;
• ancestry/self-report, only after mechanism and metrology.

A topology candidate would advance only after demonstrating ICC/CV, stability, blinded classification, and external gain over analytes/context, without using the endpoint as both label and predictor.

16. Individual variability

Heterogeneity is not incidental noise. Women at the same stage may differ in adrenal/ovarian source, sulfation, STS activity, SHBG, 11KT, aromatization, HSD11B1, inflammation, fat distribution, and therapeutic exposure. A PCOS history may attenuate with age or persist through different branches. The same serum value may generate a different intracellular dose; the same dose may produce a different response because of receptor, cofactor, and cellular state.

The design must estimate within-donor effects before comparing groups, retain raw curves, and report distributions rather than means alone. Prespecified modifiers do not rescue a null primary result. Without common support for age/stage, within-woman follow-up is prioritized; regression adjustment does not create comparability.

17. Pharma relevance and maturity

The present Pharma opportunity is mechanistic de-risking, not compound selection.

BAY1128688 is not proposed: AKRENDO1 was stopped for hepatotoxicity. This does not kill AKR1C3 as a target, but it separates target validation from molecule validation and raises the safety/chemistry bar. An HSD11B1 inhibitor is not selected simply because it lowers cortisol either: it could increase 11KT/AR and change the net sign.

Kill criteria for the Pharma opportunity: local functional equivalence; unattainable exposure; absence of enzyme/receptor dependence; benefit in one branch offset by greater risk in another; lack of mediation to an outcome; or inability to separate target engagement from toxicity. Overall status: H1 links, H0 chains; not H5. HUMAN_QA_REQUIRED before partnering or a human protocol.

18. Limitations

1. The review was directed and not systematic; literature may have been missed, and a recent search does not prove worldwide absence.
2. Much of the mechanistic evidence comes from recombinant protein, cell lines, selected explants, or animal models; it is not equivalent to longitudinal female physiology.
3. Postmenopausal PCOS studies are small, heterogeneous, and vulnerable to retrospective classification, treatment, and survival.
4. Nanba separates age groups without an individual trajectory; p=.3 does not establish 11KT equivalence.
5. E30 is a conference abstract, n=10 per assignment and seven days; it strengthens a competitor but does not establish nullity.
6. E31 uses predominantly male livers and does not measure SAT or function.
7. The simulation uses synthetic truth and operational criteria; it provides neither biological magnitudes nor clinical power.
8. ΔF=0.10, QC thresholds, and gates are design decisions, not agreed clinical thresholds.
9. CRISPRi/rescue may alter differentiation and ligands; post-maturation induction and controls reduce but do not eliminate the problem.
10. The NEFA endpoint represents net efflux and may not mediate systemic sensitivity or T2D.
11. No validated surrogate exists for intracellular 11KT exposure or AR engagement in cohorts.
12. Transportability to VAT, liver, muscle, brain, ovary, bone, and Mexico/LATAM populations remains open.
13. Available clinical outcomes do not demonstrate androgen mediation of healthspan or survival.

19. Conclusions

PCOS excess and the putative postmenopausal deficit do not currently form a demonstrated symmetry. They share the network, not necessarily the lesion. The network includes source, sulfation, transport, local conversion, AR/GR/ER, inflammation, and organ function. The strongest evidence changes the variable: active composition and tissue context are more informative than isolated DHEA-S or total T.

The best-positioned competitor is conversion/engagement without a material effect on the prioritized adipose function, accompanied systemically by common cause and parallel organs. The dual AR/GR hypothesis is more novel and informative, but it begins at H0 and one equivalent branch kills it. The “excess–deficit” U remains parked until the extremes are shown to exist in tissue and to harm the same function in the same context.

The smallest decisive test is a downstream 11KT×cortisol factorial with equivalence, interaction, and quantified receptor losses. Enzymes, life-stage groups, and clinical translation come afterward. This sequence makes a negative result into knowledge: it can close the adipose branch and prevent attribution to a hormone or target of what actually belongs to transport, context, or chemistry without function.

Female longevity cannot be inferred from a hormone. It requires a causal mechanism, persistent function, an organ-specific outcome, mediation, competing risks, and years free of disease/disability. No distal link in that chain has been validated for the proposed topology.

20. References

1. Crawford S, Santoro N, Laughlin GA, et al. Circulating dehydroepiandrosterone sulfate concentrations during the menopausal transition. J Clin Endocrinol Metab. 2009. DOI: 10.1210/jc.2009-0386. PMID: 19470626.
2. Nanba AT, Rege J, Ren J, et al. 11-Oxygenated C19 steroids do not decline with age in women. J Clin Endocrinol Metab. 2019. DOI: 10.1210/jc.2018-02527. PMID: 30753518.
3. Markopoulos MC, Rizos D, Valsamakis G, et al. Hyperandrogenism in women with polycystic ovary syndrome persists after menopause. J Clin Endocrinol Metab. 2011;96:623–631. DOI: 10.1210/jc.2010-0130. PMID: 21177795.
4. Markopoulos MC, Valsamakis G, Kouskouni E, et al. Study of carbohydrate metabolism indices and adipocytokine profile and their relationship with androgens in polycystic ovary syndrome after menopause. Eur J Endocrinol. 2013;168:83–90. DOI: 10.1530/EJE-12-0550. PMID: 23076844.
5. O’Reilly MW, Kempegowda P, Jenkinson C, et al. 11-Oxygenated C19 steroids are the predominant androgens in polycystic ovary syndrome. J Clin Endocrinol Metab. 2017. DOI: 10.1210/jc.2016-3285. PMID: 27901631.
6. Tosi F, Villani M, Garofalo S, et al. Clinical value of serum 11-oxygenated androgen measurements in women with polycystic ovary syndrome. J Clin Endocrinol Metab. 2022. DOI: 10.1210/clinem/dgab920. PMID: 34951635.
7. Dokras A, et al. 11-Oxygenated androgen profiles in women with polycystic ovary syndrome: phenotype and intervention analyses. J Clin Endocrinol Metab. Published online 2026-08-13. DOI: 10.1210/clinem/dgag329. PMID: 42592844.
8. O’Reilly MW, House PJ, Tomlinson JW. AKR1C3-mediated adipose androgen generation and metabolic dysfunction in women with PCOS. J Clin Endocrinol Metab. 2017. DOI: 10.1210/jc.2017-00947. PMID: 28645211.
9. Zang H, Carlström K, Arner P, et al. Effects of testosterone and estrogen treatment on lipolysis signaling pathways in subcutaneous adipose tissue of postmenopausal women. Fertil Steril. 2007. DOI: 10.1016/j.fertnstert.2006.11.088. PMID: 17408628.
10. Dumesic DA, et al. The subcutaneous adipose microenvironment in normal-weight women with polycystic ovary syndrome. J Endocr Soc. 2024. DOI: 10.1210/jendso/bvae162.
11. Paulukinas RD, Mesaros C, Penning TM. Conversion of classical and 11-oxygenated androgens by human AKR1C3 and HSD11B1 in adipocyte models. Endocrinology. 2022. DOI: 10.1210/endocr/bqac068. PMCID: PMC9162389.
12. Schiffer L, et al. HSD11B1 regulates the balance between 11-ketotestosterone/androgen receptor and cortisol/glucocorticoid receptor signalling. FASEB J. 2024. DOI: 10.1096/fj.202302131R. PMID: 38551804. 2025 correction: 10.1096/fj.202503904, PMID: 41144895.
13. Schiffer L, et al. Inhibition of the androgen-activating enzyme AKR1C3 selectively decreases systemic and intra-adipose 11-oxygenated androgens in women. bioRxiv, preprint v1. 2026. DOI: 10.64898/2026.03.27.714735.
14. McDonnell E, et al. Metabolic and steroid responses to DHEA or 11-ketoandrostenedione in women with PCOS. Endocrine Abstracts. 2024;104:OC8. Primary abstract. DOI: 10.1530/endoabs.104.oc8.
15. McDonnell E, et al. Human hepatic metabolism of 11-oxygenated androgens after 11-ketoandrostenedione exposure. Eur J Endocrinol. 2025. DOI: 10.1093/ejendo/lvaf166. PMID: 40795393.
16. Corbould A. Chronic testosterone treatment induces insulin resistance in subcutaneous adipocytes of women. 2007. PMID: 17332526.
17. Lee MJ, Fried SK. Glucocorticoids antagonize TNF-stimulated lipolysis and insulin resistance in human adipocytes. Am J Physiol Endocrinol Metab. 2012. DOI: 10.1152/ajpendo.00228.2012. PMID: 22949029.
18. Lee MJ, Fried SK, et al. Depot-specific regulation of cortisone-to-cortisol conversion in human adipose tissue. Obesity. 2008. DOI: 10.1038/oby.2008.207. PMID: 18388900.
19. Jönsson C, et al. Bimodal regulation of lipolysis by insulin in human adipocytes and divergence of glycerol and fatty-acid release. Biochem J. 2019. DOI: 10.1042/BCJ20190594. PMID: 31519735.
20. Paulukinas RD, et al. Catalytic and non-catalytic roles of AKR1C3 in androgen receptor stability and adipocyte lipid metabolism. Endocrinology. 2023. DOI: 10.1210/endocr/bqad033. PMID: 36799021.
21. Selva DM, Hogeveen KN, Innis SM, Hammond GL. Monosaccharide-induced lipogenesis regulates the human hepatic sex hormone-binding globulin gene. J Clin Invest. 2007. DOI: 10.1172/JCI32249. PMID: 17992261.
22. Rege J, Nakamura Y, Satoh F, et al. 11-Oxygenated estrogens are a novel class of human estrogens but do not contribute materially to the circulating estrogen pool. Endocrinology. 2021. DOI: 10.1210/endocr/bqaa231. PMCID: PMC7814299.
23. Noordam C, Dhir V, McNelis JC, et al. Inactivating PAPSS2 mutations in a patient with premature pubarche. N Engl J Med. 2009. DOI: 10.1056/NEJMoa0810489. PMID: 19474428.
24. Nair KS, Rizza RA, O’Brien P, et al. DHEA in elderly women and DHEA or testosterone in elderly men. N Engl J Med. 2006. DOI: 10.1056/NEJMoa054629. PMID: 17050889.
25. Villareal DT, Holloszy JO. Effect of DHEA on abdominal fat and insulin action in elderly women and men: a randomized controlled trial. JAMA. 2004;292:2243–2248. DOI: 10.1001/jama.292.18.2243. PMID: 15536111.
26. Ruth KS, Day FR, Tyrrell J, et al. Using human genetics to understand the disease impacts of testosterone in men and women. Nat Med. 2020. DOI: 10.1038/s41591-020-0751-5. PMCID: PMC7025895.
27. Schmidt J, Landin-Wilhelmsen K, Brännström M, Dahlgren E. Cardiovascular disease and risk factors in PCOS women of postmenopausal age: a 21-year controlled follow-up study. J Clin Endocrinol Metab. 2011. DOI: 10.1210/jc.2011-1677. PMID: 21956415.
28. Berni TR, Morgan CL, Rees DA. Women with polycystic ovary syndrome have an increased risk of major cardiovascular events: a population study. J Clin Endocrinol Metab. 2021. DOI: 10.1210/clinem/dgab392. PMCID: PMC8372630.
29. Forslund M, et al. Androgen and gonadotropin profiles in women with PCOS followed to a mean age of 81 years. J Clin Endocrinol Metab. 2021. DOI: 10.1210/clinem/dgaa840. PMID: 33205205.
30. Forslund M, et al. Morbidity and mortality in women with PCOS followed to a mean age of 81 years. Eur J Obstet Gynecol Reprod Biol. 2022. DOI: 10.1016/j.ejogrb.2022.02.020. PMID: 35220175.
31. Kugelman N, et al. Polycystic ovarian morphology and long-term health outcomes among women with PCOS. JAMA Network Open. 2025. DOI: 10.1001/jamanetworkopen.2025.40818. PMID: 41171273.
32. Caldwell ASL, Middleton LJ, Jimenez M, et al. Characterization of reproductive, metabolic, and endocrine features of polycystic ovary syndrome in female mice with tissue-specific deletion of the androgen receptor. Proc Natl Acad Sci USA. 2017. DOI: 10.1073/pnas.1616467114. PMID: 28320971.
33. Millán-de-Meer M, Luque-Ramírez M, Nattero-Chávez L, Escobar-Morreale HF. PCOS during the menopausal transition and after menopause: a systematic review and meta-analysis. Hum Reprod Update. 2023. PMID: 37353908.

Scientific delta of REPORT_EN

This report preserves the factual manifest and operationalizes the same model change as its Spanish twin: it replaces the scalar PCOS–postmenopause androgen mirror with a falsifiable AR/GR/ER composition-and-partition topology, elevates conversion without material function to the top-ranked hypothesis, and defines a downstream equivalence factorial that can distinguish nullity, a single branch, and dual contribution before enzymes, life-stage groups, or Pharma relevance are evaluated. The consequence for longevity is restrictive but productive: only an organ-specific and longitudinal causal chain can open that translation; current data do not close it.