Allopregnanolone as a neurosteroid: GABA-A modulation, anxiety, and insomnia in perimenopause

Stage: REPORT_EN
Scientific cutoff: 2026-08-03
Core population: women undergoing spontaneous ovarian aging in STRAW+10 −2/−1
Classification: progesterone and neurosteroid metabolomics · neuroendocrinology · sleep
Global maturity: H0 for the integrated chain; H1 for molecular capacity, preclinical plasticity, and several systemic competitors
Use: scientific research report. It does not diagnose, prescribe, or recommend treatments.

1. Executive scientific abstract

Allopregnanolone (Allo; 3α,5α-THP) is a neuroactive progesterone metabolite capable of directly modulating GABA-A receptors. That molecular capacity is well demonstrated in recombinant human receptors. It does not follow, however, that a low blood Allo concentration causes anxiety or insomnia during the menopause transition. The available human evidence does not contain a longitudinal chain in the same women that measures stereospecific Allo → GABA-A function → objective sleep/anxiety at sufficient resolution, much less a causal link to healthspan or longevity.

The main conclusion of this project is that the linear model low Allo → less GABA → anxiety/insomnia should be abandoned. The scientifically defensible model has two gates. The first is arousal initiation: vasomotor symptoms, circadian phase, temperature, respiration, the HPA axis, pain, substances, and homeostatic sleep pressure contribute to whether an awakening or anxiety episode occurs. The second is event-conditioned recovery: recent cerebral Allo history and the adaptive state of GABA-A could modify how long a network takes to recover inhibition, return to stable sleep, or extinguish arousal after an event. Human evidence supports the first gate better than the second; the second is an H0 mechanistic hypothesis, not a clinical finding.

Adversarial review narrowed the hypothesis further. A difference observed while Allo remains present may simply be acute allosteric occupancy. To demonstrate post-ligand memory, two conditions must receive the same blocks of free Allo in a different order, end at the same concentration, undergo chemically verified washout in medium and cells, and only then receive an identical excitatory perturbation. The difference must depend on a persistent functional GABA-A modification, preserve stable E_GABA, survive control for cumulative activity, and replicate with loss and rescue of a preselected mediator. The priority causal unit therefore becomes exposure order × washout × post-event recovery, not mean concentration × symptom.

The best-supported explanation today is a stratified systemic cause: vasomotor and non-vasomotor events, pre-event state, age, circadian rhythm, HPA activity, ovulation, estradiol, and progesterone explain in parallel much of the variation in sleep, anxiety, and peripheral Allo. This explanation also must not be absolutized: objective hot flashes cluster near some awakenings, but do not uniformly explain objective transitions or the entire insomnia phenotype.

The relation to health and longevity is distal. In SWAN, persistent insomnia trajectories were associated with incident cardiovascular disease, but those studies measured neither Allo nor GABA-A and did not demonstrate mediation. Persistent sleep and functional recovery may be future healthspan intermediates; they are not longevity surrogates. There is currently no evidence that Allo, GABA-A, or an intervention on this pathway prolongs life, reduces mortality, or increases disease-free years.

Scientific delta of REPORT_EN: integrating metrology, pharmacology, receptor plasticity, sleep physiology, and adversarial critique reorganizes the program around three estimands that must not be conflated: acute occupancy, post-washout GABA-A memory, and human recovery conditioned on event type. This separation transforms a nonspecific hormonal association into an experimental sequence capable of producing four interpretable outcomes: specific memory, acute occupancy without memory, network homeostasis, or material equivalence.

2. Scientific question and relevance

Primary question

In women at STRAW+10 −2/−1, does a specific temporal history of Allo first modify GABA-A function and then recovery after an arousal event, beyond ovulatory architecture, E2/P4, vasomotor symptoms, HPA activity, circadian factors, respiration, medications, and prior function?

Mechanistic question

After removing the ligand and its neuroactive metabolites, does the order of Allo blocks leave a persistent GABA-A state that changes recovery in a human network, or is any effect better explained by contemporaneous occupancy, E_GABA, cumulative activity, maturation, or system drift?

Health and longevity question

If the proximal chain survives, do repeated episodes of slow recovery contribute to a persistent trajectory of fragmentation and reduced daytime function? Only after demonstrating that persistence would it be legitimate to evaluate organ-specific healthspan domains. Mortality, lifespan, and longevity remain outside inference.

The relevance is twofold. First, the menopause transition combines ovulatory irregularity with vasomotor, circadian, and affective changes; a hormonal mean may lose the causal temporal dimension. Second, an incorrect “deficiency” narrative can conflate substrate, conversion, compartment, receptor, and symptom, directing pharmacologic research toward sustained exposure increases without showing that this is the correct biological intervention.

3. Scope, population, and life stage

The core population is women approximately 40–55 years old undergoing spontaneous ovarian aging, prospectively classified as STRAW+10 −2 or −1. Reproductive stage and ovulatory state are measured separately; chronological age does not replace stage. Scientific comparators are late reproductive stage −3b, early postmenopause +1a, and, most importantly, contrasted windows within the same woman.

The primary human causal unit is within-woman change, not the average difference between women. Sleep and anxiety outcomes remain separate. The following are distinguished:

• awakening hazard;
• post-event recovery burden;
• accumulated nocturnal continuity;
• post-event state anxiety;
• daytime functional recovery.

Ovulatory and anovulatory cycles, early and late transition, events with and without physiologic VMS, insomnia and sleep without complaint, low and high baseline anxiety, and spontaneous exposure versus exogenous hormones or GABAergic drugs are not pooled.

Surgical menopause, primary ovarian insufficiency, pregnancy, postpartum, PMDD, PTSD, epilepsy, and neurodegenerative disease are not substitutes for perimenopause. They may inform a molecular node or plasticity when labeled as indirect evidence. Hysterectomy without oophorectomy requires its own rules and does not automatically enter the core cohort.

4. Background knowledge and mechanism map

4.1 Metabolic pathway

The canonical pathway is:

progesterone --SRD5A1/5α-reductase--> 5α-DHP --AKR1C2/3α-HSD + NADPH--> allopregnanolone.

Resected human brain tissue and recombinant human enzymes support 5α-reductase/3α-HSD activity and the capacity of AKR1C2 to form Allo. Net flux nevertheless depends on substrate, cofactors, back-oxidation, epimer formation, 5β-reduction, conjugation, and clearance. AKR1C1/3, SDR enzymes, and other pathways can produce or consume related steroids. An isolated Allo concentration or the Allo/P4 ratio does not locate the metabolic bottleneck.

4.2 Compartment

This report distinguishes L_p(t), measured peripheral Allo, from L_c(t), effective free Allo in a cerebral microenvironment. L_c(t) integrates blood entry, local synthesis, neuron–glia metabolism, membrane binding, back-oxidation, and clearance. The small post-mortem study of five luteal and five postmenopausal women showed regional heterogeneity but did not calibrate blood against brain within women. No adequate longitudinal plasma–CSF series exists in STRAW −2/−1.

4.3 Receptor and electrochemistry

GABA-A is a pentameric channel. Mutagenesis and electrophysiology in recombinant human receptors demonstrated separable transmembrane neurosteroid sites for potentiation of GABA responses and direct activation. This proves molecular capacity, not clinical sign.

The net effect depends on:

• pentameric composition and assembly;
• synaptic and extrasynaptic receptors;
• phosphorylation, insertion, internalization, and desensitization;
• ambient GABA;
• the reversal potential E_GABA and Cl⁻ homeostasis through KCC2/NKCC1;
• circuit, cell type, maturation, and network state;
• free concentration, duration, order, and residual ligand presence.

Allo is not selective for δ receptors; in mouse granule neurons, δ and γ2 populations contributed to tonic inhibition, whereas phasic inhibition depended mainly on γ2. A change in GABRD or GABRA4 therefore does not define the pathway by itself, and surface abundance of one subunit does not demonstrate assembly or function.

4.4 Plasticity and memory

Steroid history can modify GABA-A on several timescales:

• milliseconds–seconds: modulation and desensitization;
• minutes–hours: phosphorylation and trafficking;
• hours–days: programs such as EGR3→GABRA4 and subunit remodeling.

Puberty, pharmacologic withdrawal, and the peripartum period demonstrate causal plasticity in animals, but they are extreme transitions in regions distinct from human sleep networks. One preclinical study showed that brief Allo exposure followed by washout could leave a sustained increase in inhibition dependent on PKC and β3; this makes post-ligand memory testable, but does not demonstrate that it occurs in perimenopause.

4.5 Two-gate model

The integrated causal map is:

ovulatory/P4 irregularity + adrenal source + local metabolism

→ peripheral and cerebral Allo history

→ acute occupancy or persistent GABA-A adaptation, modulated by E_GABA

× vasomotor/non-vasomotor arousal pulse

→ event initiation and/or recovery burden

→ accumulated fragmentation and post-event anxiety

→ possible functional persistence

→ organ-specific healthspan, not yet attributed.

Initiation and recovery are different estimands. A hot flash may precede an awakening without Allo initiating it; Allo could, in theory, modify recovery. The event and pre-event state may instead explain both, or sleep/anxiety may later alter HPA activity, ovulation, and Allo.

5. Evidence method

This report extends, corrects, and synthesizes the SCOPING, EVIDENCE_MAP, EVIDENCE_VERIFICATION, MECHANISTIC_SYNTHESIS, COMPUTE_DECISION, HYPOTHESIS_GENERATION, ADVERSARIAL_REVIEW, and EXPERIMENT_DESIGN artifacts. It does not restart the research.

Primary studies, full texts, PubMed/PMC, and official registries were prioritized. For each decisive source, design, population or system, species, sex/stage, size, exposure, comparator, outcome, magnitude when available, and limitations were extracted. Reviews were used to locate original studies, not as the only evidence for causal claims.

Evidence labels are:

• direct human: measures the claimed node in the relevant population or window;
• proximal human: measures a neighboring construct, such as MRS-GABA, P4, or sleep, without measuring Allo and functional GABA-A together;
• causal animal: permits perturbation with restricted transportability;
• in vitro/ex vivo: tests molecular or cellular capacity in a defined system;
• computational: prioritizes cell/component, not function;
• inferred: concatenates nodes not observed in one source.

Formal verification closed on 2026-08-02. A bounded primary check on 2026-08-03 found no new publication combining repeated stereospecific Allo, GABA-A function, objective sleep, and anxiety in STRAW −2/−1. NCT06238700 remains without results and studies pregnenolone rather than isolated Allo; NCT05329779 remains terminated for low enrollment with two participants and no interpretable biological estimate.

Inference rules: association is not causation; p>0.05 is not equivalence; MRS-GABA is not receptor or current; P4 is not Allo; blood is not brain; one night of sleep is not healthspan; CVD is not longevity.

6. Evidence map

Certainty judgment by link

• P4 → 5α-DHP → Allo: high biochemical capacity; stage/cell flux unknown.
• Allo → GABA-A: high in-vitro molecular certainty; low certainty for the clinical sign in perimenopause.
• Allo history → GABA-A memory: moderate preclinical plausibility; absent human evidence.
• perimenopause → regional GABA: low–moderate and associational.
• Allo → anxiety: very low in humans.
• Allo → objective insomnia: not established.
• VMS/systemic arousal → sleep: moderate, heterogeneous, and insufficient for the entire phenotype.
• persistent sleep disturbance → CVD: moderate observational evidence.
• Allo/GABA-A → healthspan/longevity: not established.

7. Contradictory evidence and null findings

1. Mean level versus dynamics. Some studies show lower peripheral Allo after menopause; another found no follicular–postmenopausal difference. In the closest weekly perimenopausal series, Allo did not predict group-level symptoms. Method, matrix, adrenal source, clearance, substrate, stage, and frequency may explain the discrepancy.

2. Apparent heterogeneity versus analytic selection. The D02 “8 of 10 sensitive” result selected the larger of two correlations for each woman, used |r|≥0.30, had no holdout, and relied on 6–8 pairs. It is not confirmatory evidence of subgroups.

3. Biphasic curve versus pseudoreplication. D04 suggested worse mood in an intermediate range, but grouped 215 observations from 37 women into post hoc bins and did not adjust dependence, dose, period, or multiplicity. It only exploratorily refutes a naive monotonic narrative; it does not estimate a causal curve.

4. PAM capacity versus paradoxical sign. Allo can potentiate GABA-A and also reduce a specific tonic current through desensitization in a defined α4β2δ and Cl⁻ context. This is not a chemical contradiction: receptors, electrochemistry, and timescales differ.

5. Regional GABA versus global deficit. sgACC changed longitudinally; mPFC was borderline and age-dependent; DLPFC was null. A single MRS result cannot represent all inhibitory function.

6. P4 and sleep versus mediation by Allo. Oral P4 improved continuity in very small postmenopausal samples under some conditions; intranasal P4 promoted sleep without zolpidem's EEG signature. P4 acts through multiple metabolites, PR, thermoregulation, and other pathways.

7. VMS near awakening versus vasomotor sufficiency. Many VMS cluster temporally around awakenings, but physiologic VMS do not uniformly predict transitions or all metrics. Perception/recall and a non-vasomotor insomnia phenotype are separable constructs.

8. Human causal trial failed for feasibility. NCT05329779 ended with two participants. This is not a biological null; it does show that the decisive human test was not executed.

9. Sleep–CVD versus neurosteroid mediation. The longitudinal association exists, but may reflect confounding, bidirectionality, apnea, depression, prior health, or autonomic/metabolic mechanisms that do not include Allo.

8. Multiscale mechanism synthesis

Molecular

Allo is a potent GABA-A neurosteroid ligand, but its effective concentration cannot be inferred from P4 or blood. The balance among synthesis, reversible conversion, 5α/5β products, epimers, conjugation, and clearance determines availability. Stereospecific metrology is part of the mechanism, not a technical detail.

Membrane and receptor

Acute occupancy and post-ligand memory are different mechanisms. The first requires ligand presence; the second requires a persistent phosphorylation, trafficking, or assembly state after ligand removal. PKC, EGR3, α4, δ, γ2, and β3 are candidates, not an established signature. A memory hypothesis survives only if a preselected mediator changes before function, shows a functional complex, is lost with perturbation, and returns with rescue.

Cell and network

More GABA-A conductance does not automatically equal more inhibition. E_GABA, KCC2/NKCC1, ambient GABA, baseline excitability, maturation, excitatory/inhibitory ratio, and astrocytes can change sign or recovery. Cumulative activity during Allo sequences can leave network homeostasis even without neurosteroid-specific memory; equivalence of activity or an independent clamp is therefore required.

Human circuit

MRS signals suggest regionality, not receptor. SICI/TMS may approximate GABA-A intracortical inhibition, but it is motor, variable, and does not automatically represent sleep networks. The human functional readout must pass its own repeatability gate; substituting MRS after failed SICI would change the construct.

Nocturnal phenotype

The event must be classified before inferring mechanism: physiologic VMS, VMS perception, apnea/respiration, pain, movement, or unclassified arousal. Recovery burden, not merely the number of awakenings, is the outcome aligned with the hypothesis. Post-event anxiety and sleep remain separate.

System and life course

Persistent nocturnal burden could activate HPA/sympathetic signaling, affect metabolism and inflammation, and contribute to CVD; it may also be a consequence of prior health. In the reverse direction, repeated fragmentation may modify HPA activity and then ovulation, P4, and Allo. Only series with frozen physiologic lags can distinguish these directions. Even if the proximal chain replicates, extension first ends at functional recovery and organ-specific domains, not global longevity.

9. Computational layer

Decision: skip. The donor-aware analysis inherited from L7-2 already examined 317,097 cells/nuclei from 24 datasets and 48 female dataset×donor pairs aged 36–55 years. It prioritized SRD5A1, downstream candidates, and the neuronal GABRA4/GABRD/GABRG2 context, together with a neuron-versus-astrocyte-support comparison.

That analysis resolved the only decision a static atlas could change: which model and components to test first. It did not measure protein, assembly, flux, Allo, chloride, waveform, receptor surface, sleep, or anxiety. Repeating an atlas or using Geneformer or sequence/structure models would not distinguish post-washout memory, acute occupancy, network homeostasis, or systemic sufficiency. Current uncertainty is experimental, not annotational.

10. Primary hypothesis

L7-3-HG-H1 v2 — post-washout GABA-A memory of exposure order

Falsifiable statement: In a qualified female human neural preparation, two sequences composed of the same free-Allo blocks will leave different recovery burden after stereospecifically verified washout and an identical perturbation; the difference will require a persistent functional GABA-A modification rather than E_GABA, residual ligand/metabolites, cumulative activity, or baseline drift.

Mechanism: Allo order changes GABA-A phosphorylation, trafficking, or assembly on a slower timescale than occupancy; that state persists after ligand removal and modifies the response to a later perturbation.

Predictions: donor-level order effect; metabolites below the functional residual gate; E_GABA, baseline, and activity in equivalence; assembly/function change before the pulse; mediator loss removes the contrast and rescue restores it; independent replication without retuning pulse, window, or model.

Evidence for: direct molecular capacity; steroid-history plasticity in animals; PKC/β3-dependent post-Allo persistence in a preclinical model; executability of components in an atlas.

Evidence against: no human perimenopausal demonstration; group-null D02; exploratory D04; extreme preclinical windows; blood and MRS do not measure the mechanism; the signal may be occupancy or homeostasis.

Status: weakened_narrowed_testable.
Maturity: integrated H0; molecular/preclinical parents H1.
Confidence: 0.16.

11. Competing hypothesis

L7-3-HG-H2 v2 — stratified systemic sufficiency

Falsifiable statement: In humans, the type and intensity of a thermovasomotor or non-vasomotor event and the pre-event state will explain initiation and recovery; after measuring them faithfully, peripheral Allo history and GABA-A function will remain within incremental equivalence.

Mechanism: ovulatory irregularity, age, thermoregulation, circadian factors, HPA activity, respiration, and sleep pressure change arousal and peripheral steroids in parallel. Perception and anxiety amplify some events. Allo provides no material independent mediation.

Predictions: M_sys will explain recovery within VMS and non-VMS events; adding Allo history and SICI will not improve error/calibration outside woman and cycle; symptoms will persist with stable Allo while triggers persist; sleep/HPA lags may precede endocrine change.

Evidence for: VMS/endocrine suppression, non-vasomotor insomnia, age and sleep architecture, E2/ovulation variability, the group-null D02, and absence of a human chain.

Evidence against: Allo can modulate GABA-A and leave plasticity; objective VMS do not explain all sleep; null blood does not prove null brain.

Nested mechanistic rivals that D0 must distinguish:

• L7-3-AR-H1O: acute occupancy without memory; orders equivalent after washout, effect only with Allo present.
• L7-3-ED-HN1: network homeostasis; post-washout effect follows cumulative activity/baseline and disappears when they are matched.
• L7-3-HG-H3 v2: reverse causality; nocturnal burden precedes HPA activity and later ovulation/P4/Allo.

Status: strengthened_but_split.
Maturity: H1 for systemic components; H0 for complete sufficiency.
Confidence: 0.48.

12. Translational hypothesis

L7-3-HG-HT1 v2 — conditional event-locked signature

Falsifiable statement: Only after Q0c→D0-C→Q0h/QF/Q1→D1, one Allo order/history parameter and one functional GABA-A readout, frozen before validation, will improve external error and calibration for post-event recovery beyond stage, ovulation, P4/E2, VMS, HPA activity, circadian factors, respiration, event severity, and prior function.

Proposed mechanism: an exposure dynamic may capture ligand–receptor mismatch lost by an isolated sample; a functional readout may localize a proximal node. The combination makes sense only if the cellular mechanism and human transport have already been demonstrated.

Predictions: reproducible stereospecific metrology; a parameter more stable/informative than isolated concentration or Allo/P4; a functional readout with adequate ICC/MDC; leave-one-woman/cycle-out gain; external calibration; signal preferentially for recovery, not necessarily awakening count.

Evidence against: there is no standardized assay, central proxy, functional measure validated for this use, or external cohort. A positive D0 does not itself enable translation.

Status: parked_weakened; HUMAN_QA_REQUIRED.
Maturity: H0.
Confidence: 0.03.
Note: this is not a validated biomarker or clinical instrument.

13. Falsifiable predictions and kill criteria

Cross-cutting death criteria: a molecular signal without function does not count; significance without materiality does not count; p>0.05 does not demonstrate equivalence; an effect from one clone/batch/pulse does not replicate; technical failure is not presented as universal biological nullity.

14. Discriminating experiment

14.1 Smallest decisive sequence

Q0c → Gate A → D0-S → D0-C.

This is the smallest sequence capable of changing the hypothesis. Human study D1 is conditional on a positive D0-C.

14.2 Q0c: cellular metrology

Stereospecific LC-MS/MS in medium, conditioned medium, free fraction, and cellular extract. It must resolve Allo, isoallopregnanolone, pregnanolone, P4, 5α-DHP, and 5β-DHP using isotopic standards, matrix curves, blanks, accuracy, precision, recovery, carryover, and stability.

L/H levels are defined functionally in the model, not as human brain concentrations. Common washout T_wash must leave each analyte below LLOQ and with a functional upper bound incapable of explaining more than one quarter of the minimum material effect ΔM according to the acute curve.

14.3 Gate A: model

A human cortical network derived from female-donor iPSCs with a fixed excitatory/inhibitory ratio and astrocyte support. Two differentiations per donor; well/electrode/cell are technical replicates.

It must blindly demonstrate hyperpolarizing GABA or a defined shunt by gramicidin, stable E_GABA, reproducible GABA-A and MEA responses, maturation, identity, viability, controlled temperature/pH/osmolarity, and baseline without drift. The MEA electrical pulse is calibrated once to be reversible and non-toxic.

14.4 D0-S: screening and mediator selection

Six donors, two differentiations, and ≥4 technical wells per condition/batch. Primary conditions:

• O_late = H | H | L | L | H | H;
• O_early = H | L | L | H | H | H.

Both contain four H blocks, two L blocks, one H→L transition, and one L→H transition; they share start, peak, AUC, time at each level, and terminal concentration, and differ in recency of the low/reentry window.

Arms: two orders→washout→pulse; vehicle; acute Allo during pulse; orders without pulse; vehicle without pulse. A stable arm is secondary and is not called overmatched.

Primary outcome:

RB_MEA = ∫_0^60 max[0, log(BR_t) − log(BR_baseline)] dt.

BR is network burst rate in predefined windows. Hierarchical secondary outcomes are time to equivalence, tonic/phasic currents, E_GABA, excitability, assembly/surface/phosphorylation, and viability.

One mediator advances to confirmation if it changes before the pulse, exceeds MDC, demonstrates assembly/function, permits non-toxic loss-rescue, and repeats direction in ≥4/6 donors.

14.5 D0-C: confirmation

New donors, preregistered protocol, two operators/series, and a temporally independent batch. It includes mediator perturbation and rescue, controls without Allo history, and an accumulated-activity-matched branch if D0-S does not demonstrate equivalence.

Donor-level mixed model:

RB ~ order × perturbation + pulse_real + baseline + cumulative_activity + (1 + order | donor) + (1 | donor:batch).

Variance from six D0-S donors informs blinded re-estimation. The confirmatory maximum is 24 donors; if more are required or MDC95 ≥ ΔM, this is a technical no-go. Equivalence is tested with TOST and ±ΔM, not absence of significance.

Interpretation:

• post-washout effect + assembly + rescue + replication → supports HG-H1 memory;
• equivalent orders + effect with ligand present → supports AR-H1O occupancy;
• effect abolished when activity/baseline are matched → supports ED-HN1 homeostasis;
• CI90% within ±ΔM with valid controls → informative equivalence and memory deprioritization;
• interval crossing zero and margins → indeterminate, not support.

14.6 Human D1, only if D0-C is positive

Follow-up for 12 weeks or two cycles/windows in women at STRAW −2/−1, without selecting only insomnia/VMS. Frequent stereospecific sampling, multi-time cortisol, continuous actigraphy/light/temperature, at least four home PSG nights, physiologic VMS and autonomic intensity, respiration, and one prespecified SICI readout at three points.

Primary human outcome: minutes in wake/N1 during 30 minutes after a PSG awakening, separating VMS and non-VMS events. Nested M_sys, M_allo, and M_gaba models; validation leaves out woman and cycle. Pilot of 20 women for feasibility/variance components; initial confirmatory set of 48 complete women, maximum 80 by simulation. Lags and responder classes are not searched post hoc.

15. Measurement candidates and stratification

No validated biomarker currently exists. Candidates are experimental tools subject to gates:

• Repeated stereospecific Allo: candidate peripheral exposure; not brain.
• P4, 5α/5β-DHP, ISO, and pregnanolone: metabolic context; no ratio identifies flux by itself.
• SICI/TMS: indirect cortical GABA-A function; requires ICC/MDC and does not automatically represent sleep circuits.
• MRS-GABA: regional pool, not a substitute for SICI or receptor.
• VMS by conductance/temperature + perception: physiologic event and subjective construct kept separate.
• PSG/actigraphy: architecture/events and continuity; actigraphy does not replace PSG staging.
• HPA activity, light, circadian phase, respiration, and temperature: causes/modifiers, not decorative adjustments.

Prespecified strata are STRAW −2/−1, ovulatory/anovulatory/indeterminate, insomnia yes/no, baseline anxiety low/high, VMS/non-VMS event, and perception yes/no. An “Allo responder” category could only be defined in one cohort and tested in another.

16. Individual variability

Variability must not be invoked to rescue every null. It must be localized and prespecified.

1. Endocrine input: ovulation frequency, P4 amplitude/duration, adrenal source, and STRAW transition.
2. Metabolism/compartment: enzymes, cofactors, transport, local synthesis, clearance, matrix, and sampling time.
3. Receptor: composition/assembly, trafficking, prior history, circuit, and plasticity.
4. Electrochemistry: KCC2/NKCC1, E_GABA, inflammation, maturation, and ambient GABA.
5. Event: physiologic VMS, perception, respiration, pain, circadian phase, and sleep pressure.
6. Baseline vulnerability: prior anxiety, insomnia, depression, medication, alcohol/caffeine, and apnea.
7. Temporal course: the same exposure may have different effects during occupancy, washout, and recovery.

The correct test of heterogeneity is out-of-sample replication of a defined interaction, not individual correlations selected after analysis.

17. Pharma relevance and maturity

The current opportunity is precompetitive target/assay discovery, not clinical readiness.

Post-ligand memory branch

If D0-C demonstrates memory with assembly/rescue, it could justify investigating trafficking or assembly modulators that prevent or reverse an adverse state without continuous occupancy. Before a screen, the goal must be specified as preventing, accelerating, or reversing memory. PKC/EGR3 are pleiotropic and cannot be advanced by correlation.

Acute occupancy branch

If only present Allo changes recovery, the opportunity concerns PAMs and exposure pharmacokinetics, not memory. Sedation, cognition, tolerance, falls, and circuit specificity must be assessed; an acute effect does not demonstrate perimenopausal utility.

Systemic branch

If M_sys explains recovery and Allo/GABA remain in equivalence, the correct outcome is to deprioritize the neurosteroid target for that phenotype, not to search post hoc for a subgroup.

Pharma gates

• Go to exploratory target validation: D0-C with loss/rescue and replication.
• Go to limited screening: mechanism defined as memory, reversal, or occupancy, not a mixture.
• Go to partnering assessment: only after temporal D1, metrology/compartment, reliability, external replication, pharmacology, and human QA.
• No-go: mean serum Allo, Allo/P4, response to P4, or MRS-GABA as a companion biomarker.
• No-go: assume sustained Allo elevation is beneficial.

HG-H1, ED-HN1, and translation are H0; molecular occupancy is H1, but its perimenopausal relevance is H0. There is no target product profile, validated responder population, dose, cerebral exposure, efficacy, or safety for this mechanism. H5 is not warranted; any partnering advance requires HUMAN_QA_REQUIRED.

18. Limitations

1. No human study measures the complete chain in STRAW −2/−1.
2. Direct Allo data are small, cross-sectional, peripheral, or metrologically insufficient.
3. D02 does not validate heterogeneity; D04 does not validate a causal biphasic curve.
4. Plasma/serum do not calibrate cerebral exposure; the compartment bridge may be ethically difficult.
5. MRS, SICI, current, assembly, and behavior are different constructs.
6. Plasticity comes from puberty, pharmacologic withdrawal, peripartum, and animal regions; transportability is low.
7. Female iPSC does not preserve brain age, STRAW stage, ovulatory history, or sleep architecture.
8. The MEA pulse tests network recovery, not hot flash, awakening, or anxiety.
9. The D0 endpoint lacks published donor-level variance; the design uses pilot and blinded re-estimation.
10. The activity clamp can alter GABA-A and requires its own validation.
11. SICI is motor and indirect; it may not represent nocturnal networks.
12. D1 is intensive and may fail for feasibility even if biology exists; that failure does not prove universal cerebral nullity.
13. Objective hot flashes do not explain all insomnia, and their perception must not substitute for physiology.
14. Sleep–CVD is observational; it does not prove Allo, GABA-A, global healthspan, or longevity.
15. The search through the cutoff does not prove absolute nonexistence of unindexed studies or future results.

19. Conclusions

Allopregnanolone has real and potent GABA-A biology, but the perimenopausal explanation cannot be reduced to a mean deficiency. Ovarian substrate is irregular, blood is not brain, GABA-A is plastic and chloride-dependent, and sleep has vasomotor and non-vasomotor generators. Current human evidence favors a stratified systemic cause and leaves Allo–GABA-A memory as a novel low-maturity hypothesis.

The most important conceptual advance is to separate occupancy, memory, and recovery. D0-C can distinguish them using permuted orders, functional washout, donor as the unit, equivalence, controlled activity/chloride, and confirmed loss-rescue. Only replicated memory enables the event-locked human study; only human temporality and external replication enable a translational hypothesis.

The link to women's health is plausible only through persistence of sleep disturbance and functional recovery. The sleep–CVD association justifies studying a future bridge, not attributing it to Allo. There is no evidence of an effect on longevity and no basis to prescribe, clinically stratify, or promote sustained Allo elevation.

20. References

1. Steckelbroeck S, et al. Characterization of the 5α-reductase-3α-hydroxysteroid dehydrogenase complex in human brain tissue. J Clin Endocrinol Metab. 2001. DOI 10.1210/jcem.86.3.7325. PMID 11238528.
2. Trauger JW, et al. Human AKR1C steroid metabolism and allopregnanolone formation. Biochemistry. 2002. DOI 10.1021/bi026109w. PMID 12416991.
3. Hosie AM, Wilkins ME, da Silva HMA, Smart TG. Endogenous neurosteroids regulate GABA-A receptors through two discrete transmembrane sites. Nature. 2006;444:486–489. DOI 10.1038/nature05324. PMID 17108970.
4. Shen H, et al. A critical role for α4βδ GABA-A receptors in shaping the response to neurosteroids at puberty. Nat Neurosci. 2007;10:469–477. DOI 10.1038/nn1868. PMID 17351635.
5. Maguire J, Mody I. GABA-A receptor plasticity during pregnancy: relevance to postpartum depression. Neuron. 2008;59:207–213. DOI 10.1016/j.neuron.2008.06.019. PMID 18667149.
6. Hsu FC, Smith SS. Withdrawal from repeated progesterone administration increases neuronal excitability: role of GABA-A receptor α4 subunit. J Neurophysiol. 2003;89:186–198. DOI 10.1152/jn.00195.2002. PMID 12522171.
7. Gangisetty O, Reddy DS. Neurosteroid withdrawal regulates GABA-A receptor α4-subunit expression through Egr3. Neuroscience. 2010. PMID 20670676.
8. Abramian AM, et al. Neurosteroids promote phosphorylation and membrane insertion of extrasynaptic GABA-A receptors. Proc Natl Acad Sci USA. 2014;111:7132–7137. DOI 10.1073/pnas.1403285111. PMID 24778259.
9. Endogenous and synthetic neuroactive steroids evoke sustained increases in GABAergic inhibition through a protein kinase C-dependent mechanism. 2016. PMID 27743930.
10. Lu X, Zorumski CF, Mennerick S. Lack of neurosteroid selectivity at δ versus γ2-containing GABA-A receptors in dentate granule neurons. Front Mol Neurosci. 2020;13:6. DOI 10.3389/fnmol.2020.00006. PMID 32038169.
11. Słopień R, et al. Correlation between allopregnanolone levels and depressive symptoms during late menopausal transition and early postmenopause. Gynecol Endocrinol. 2018;34:144–147. DOI 10.1080/09513590.2017.1371129. PMID 28857628.
12. Sander BJ. Examining the Relation Between Weekly Sex Hormones and Mood in Perimenopausal Females. MSc thesis. University of Regina; 2023. Non-peer-reviewed evidence.
13. Pacchierotti C, et al. Plasma allopregnanolone levels in women attending a menopause clinic: association with panic and climacteric symptoms. Maturitas. 2004;48:147–154. DOI 10.1016/j.maturitas.2003.08.003. PMID 15172089.
14. Andréen L, et al. Allopregnanolone concentration and mood during progesterone treatment in postmenopausal women. Psychoneuroendocrinology. 2006. DOI 10.1016/j.psyneuen.2005.06.002.
15. Genazzani AR, et al. Circulating levels of allopregnanolone in humans: sex differences and gonadal/adrenal control. J Clin Endocrinol Metab. 1998;83. DOI 10.1210/jcem.83.6.4905. PMID 9626145.
16. Kimball A, et al. The allopregnanolone to progesterone ratio across the menstrual cycle and in menopause. Psychoneuroendocrinology. 2019;112:104512. DOI 10.1016/j.psyneuen.2019.104512. PMID 31780185.
17. Bixo M, et al. Progesterone, 5α-pregnane-3,20-dione and 3α-hydroxy-5α-pregnan-20-one in specific regions of the human female brain. Brain Res. 1997. DOI 10.1016/S0006-8993(97)00455-1. PMID 9295207.
18. Wang Y, et al. Menopause-related GABA changes in subgenual anterior cingulate cortex and affective symptoms. Front Neurosci. 2019;13:785. DOI 10.3389/fnins.2019.00785. PMID 31481863.
19. Tran KH, et al. Medial prefrontal GABA during the menopause transition. Psychoneuroendocrinology. 2022; DOI 10.1016/j.psyneuen.2022.105977. PMID 36146906.
20. Tran KH, et al. Dorsolateral prefrontal GABA in perimenopause. Psychoneuroendocrinology. 2022; DOI 10.1016/j.psyneuen.2022.105967. PMID 36386999.
21. Joffe H, et al. Effects of gonadal steroid suppression on sleep and vasomotor symptoms in healthy premenopausal women. Menopause. 2013. DOI 10.1097/GME.0b013e31828292d1. PMID 23481119.
22. de Zambotti M, et al. Nocturnal hot flashes: relationship to objective awakenings and sleep-stage transitions. J Clin Sleep Med. 2016;12. DOI 10.5664/jcsm.5936. PMID 26951410.
23. Thurston RC, Santoro N, Matthews KA. Are vasomotor symptoms associated with sleep characteristics among symptomatic midlife women? Menopause. 2012;19:742–748. DOI 10.1097/gme.0b013e3182422973. PMID 22415568.
24. Baker FC, et al. Insomnia in women approaching menopause: beyond perception. Psychoneuroendocrinology. 2015;60:96–104. DOI 10.1016/j.psyneuen.2015.06.005. PMID 26142241.
25. Lampio L, et al. Sleep during menopausal transition: a 6-year follow-up. J Sleep Res. 2017. DOI 10.1111/jsr.12525. PMID 28379985.
26. Caufriez A, et al. Progesterone prevents sleep disturbances and modulates GH, TSH and melatonin secretion in postmenopausal women. J Clin Endocrinol Metab. 2011;96:E614–E623. DOI 10.1210/jc.2010-2558. PMID 21289261.
27. Schüssler P, et al. Progesterone reduces wakefulness in sleep EEG and has no effect on cognition in healthy postmenopausal women. Psychoneuroendocrinology. 2008. DOI 10.1016/j.psyneuen.2008.05.013. PMID 18676087.
28. Schüssler P, et al. Intranasal progesterone and sleep in postmenopausal women. Psychoneuroendocrinology. 2018. DOI 10.1016/j.psyneuen.2018.04.001. PMID 29649764.
29. Thurston RC, et al. Midlife sleep trajectories and incident cardiovascular disease in women. Circulation. 2024. DOI 10.1161/CIRCULATIONAHA.123.066491. PMID 38284249.
30. Thurston RC, et al. Sleep characteristics and carotid atherosclerosis among midlife women. Sleep. 2017. DOI 10.1093/sleep/zsw052. PMID 28364498.
31. ClinicalTrials.gov. Study on Allopregnanolone and Depression in Perimenopausal Women. NCT05329779. Terminated for low enrollment; two participants, no interpretable result.
32. ClinicalTrials.gov. Study on Allopregnanolone and Depression in Women Across the Menopause Transition. NCT06238700. Pregnenolone vs placebo; no results at cutoff.

Traceable claims from this report

• L7-3-REPORT-EN-C01: the priority causal unit is exposure order × washout × post-event recovery; mean concentration and concurrent symptoms do not identify GABA-A memory.
• L7-3-REPORT-EN-C02: acute occupancy, post-ligand memory, network homeostasis, and material equivalence are distinct experimental outcomes and must be tested in one shared design.
• L7-3-REPORT-EN-C03: systemic causation is the best-supported human explanation, but it must be stratified into physiologic VMS, perception, and non-vasomotor arousal.
• L7-3-REPORT-EN-C04: current Pharma relevance is conditional target/assay discovery; no biomarker, clinical readiness, or partnering rationale exists.
• L7-3-REPORT-EN-C05: the health bridge ends at the persistent sleep–CVD association; mediation by Allo/GABA-A and effects on global healthspan or longevity have not been demonstrated.