Estradiol pharmacology

Reference synthesis · inline C1–C5 confidence tags · cross‑linked to evidence and reference.

1Molecule and metabolism

17β‑estradiol (C₁₈H₂₄O₂, MW 272.4) is a four‑ring steroid (A, B, C, D) with two metabolically reactive hydroxyls. The C3 hydroxyl is a phenol on the aromatic A ring (pK_a ~10.7, so essentially fully protonated and uncharged at physiologic pH; log P ~4.0, high passive permeability).C1 The C17 hydroxyl is a 17β‑configured aliphatic secondary alcohol on the saturated D ring — chemistry that admits oxidation to the C17 ketone (E1) by removing the carbinol hydrogen.

Enzymes that act on E2

Directional pumps

The HSD17B1/HSD17B2 pair and the SULT1E1/STS pair each form a tissue‑directional pump. Liver and gut express the “off” arms (HSD17B2 with NAD⁺; SULT1E1) and inactivate E2 on absorption. Ovary, breast, brain, endometrium, and bone express the “on” arms (HSD17B1 with NADPH; STS) and reactivate E1S and E1 to E2. The cytoplasmic cofactor pools enforce this directionality — fed‑state hepatocyte [NAD⁺]/[NADH] ~700 (Williamson 1967) drives oxidation; cytoplasmic [NADPH]/[NADP⁺] ~100 in target tissues drives reduction.C1 Plasma E1/E2 ~5 at oral steady state is the empirical readout of this hepatocyte equilibrium (the thermodynamic calculation predicting an intracellular ratio of ~11 has correct arithmetic but fragile cofactor inputs, so the observed plasma ratio is the load‑bearing anchor).C3

Mass balance for oral 1 mg E2

Of an oral 1 mg dose, ~98% is absorbed across the enterocyte (high passive permeability). Gut‑wall conjugation by UGT1A10 removes an inferred ~30–50% of absorbed E2 before portal entry.C3 Of the residual that reaches the liver, ~90% is extracted on first pass (HSD17B2 oxidation to E1, SULT1E1 sulfation to E1S, UGT glucuronidation with biliary export via MRP2). Only ~40 µg (~4% of dose) exits the liver as free E2 — the canonical “F_oral ≈ 5%” figure.C1 The v4 PK model partitions the remaining ~96% into ~30% E1, ~32% E1S, ~29% glucuronides, and ~5% catechols/E3/methylated metabolites; these fractions are model calibration outputs, not measured human radiolabel data (no portal‑vein cannulation study has been done on E2 the way Back & Rogers 1982 did for EE).C3 Kuhl 2005's textbook fractions (~15% E1, ~25% E1S, ~50% glucuronides) are looser anchors of the same shape.C2

The qualitative load‑bearing point: only a few percent of an oral E2 dose becomes free plasma E2; most becomes circulating E1 and E1S that target tissues reactivate over the following 12–24 h. The “5% bioavailability” figure understates total estrogenic activity over time, and overstates the dose actually presented to peripheral receptors.

Catechol pathway and COMT polymorphism

The catechol arm of E2 metabolism breaks into three downstream species that matter clinically: 2‑OH‑E2 (CYP1A1/1A2/3A4, ~30% of hepatic E2 clearance in the v4 module), 4‑OH‑E2 (CYP1B1, ~8%), and the methylated terminal products 2‑MeO‑E2 / 4‑MeO‑E2 made by COMT. (These hepatic‑flux fractions are higher than the ~5% “catechols / E3 / methylated” bucket cited in the mass balance above — the discrepancy is real and reflects that the older bucket counted only what is excreted as catechol or methoxy conjugates after sulfation/glucuronidation, while the catechol module tracks the fraction of E2 that passes through a catechol intermediate on its way to elsewhere.) The chemistry that makes 4‑OH‑E2 a candidate carcinogenicity driver is the same that makes 2‑MeO‑E2 anti‑proliferative: catechols can redox‑cycle to semiquinone and quinone intermediates that adduct DNA, but the methylated catechols are stable and have anti‑angiogenic / anti‑tubulin activity.C2 COMT prefers C2 hydroxyls over C4, so the methylation pump empties the 2‑OH‑E2 pool faster than the 4‑OH‑E2 pool — producing the steady‑state ratio of [4‑OH‑E2]/[2‑OH‑E2] ≈ 0.5 in the v4 catechol module after calibration to Bradlow 1995.C3

The COMT Val158Met polymorphism shifts this balance. The Val158 enzyme (~50% allele frequency in European populations) has ~3–4× the activity of the Met158 enzyme; Met/Met homozygotes (~25%) accumulate ~1.7× as much 2‑OH‑E2 and produce ~0.45× as much 2‑MeO‑E2 at the same parent‑E2 input (deterministic consequence of the v4 module's COMT methylation fractions Val 0.55 / Met 0.25).C3 Implication: the protective methylated‑catechol arm is genotype‑dependent in a way the standard PK summary tables do not capture. Pregnancy term levels (parent E2 ~20 ng/mL) push the model's 2‑OH‑E2 to ~1.4 ng/mL (Val) and ~2.3 ng/mL (Met) — ~290× the cycling baseline within the same genotype — consistent with the known elevated catechol burden in late gestation.C3

2Pharmacokinetics by route

Route determines two distinct quantities: how much parent drug reaches systemic circulation (F), and how much the hepatocyte sees on first pass (the portal‑vein concentration during absorption). These can diverge dramatically. Oral and sublingual deliver the dose through portal vein; transdermal, IM ester depot, and SC depot bypass it.

The E1:E2 ratio is a route signature. Premenopausal cycling produces ~1:1 (the ovary makes both, peripheral conversion adds a little). Transdermal and injectable routes give ~1:1 because they bypass the portal vein. Oral routes give 5:1 or higher because hepatic HSD17B2 runs the NAD⁺‑driven equilibrium during first‑pass.C1

Oral E2

Free‑E2 plasma bioavailability is ~5% for the parent drug; the apparent plasma half‑life is 13–20 h, much longer than the free‑E2 IV half‑life of 1–2 h.C1 The discrepancy is the E1S reservoir slowly back‑feeding the free pool via STS reactivation (§4). Steady‑state plasma E1S on 1 mg/d is 2,000–4,000 pg/mL (~50–100× the E2 level).C1 E2‑valerate is hydrolyzed to E2 + valeric acid by gut‑wall and plasma esterases within ~30 min, then behaves identically to oral E2 (MW correction: E2V is ~76% E2 by mass).

A third PK feature specific to oral routes is enterohepatic recirculation (EHR). The qualitative loop is well‑established: hepatocyte‑made E2 / E1 glucuronides are excreted into bile via the canalicular transporter MRP2, intestinal bacterial β‑glucuronidases (primarily in the distal small intestine and colon) hydrolyze a fraction back to free aglycone, and the deconjugated free E2 / E1 are reabsorbed across the enterocyte. The Adlercreutz & Martin 1980 review tabulates the cumulative biliary fractions of an administered dose: ~65% E2, ~48% E1, ~23% E3, with the rank-ordering reflecting the increasing polarity of E3 (which is preferentially renal-excreted).C2 Quantitatively, the loop's OUTPUT (reabsorbed fraction) is much less well-characterized than the input: Adlercreutz & Martin 1980 review the loop but explicitly say “the quantitative contribution of liver, kidneys, and intestine to estrogen metabolism cannot be accurately assessed at present,” and no subsequent human radiolabel study has cleanly partitioned biliary excretion vs reabsorption fractions for oral E2.C3 The qualitative signature is the broad 4–8 h plateau (instead of a sharp single peak) on oral PK curves and the 4–6 h secondary shoulder visible on the Goldzieher 1990 EE single‑dose curves below.C2 The popular “antibiotics raise contraceptive failure via EHR disruption” framing is mechanistically plausible but not supported by controlled human PK data: Adlercreutz 1984 found no significant plasma E2 / EE reduction during ampicillin co‑administration despite measurable fecal‑estrogen‑metabolite shifts; population‑level OC failure studies show a robust signal only for the rifamycins (which act via hepatic CYP3A induction, not via EHR).C3 The v4 model now has an explicit EHR extension (v4_ehr_ext.py, MODEL-PLAN #5 closed) that adds a bile compartment, 6 h gamma‑shaped transit, intestinal β‑glucuronidase deconjugation fraction (default 0.20), and enterocyte reabsorption fraction (default 0.70) layered on top of the baseline v4 oral‑E2 trajectory. The model's central‑parameter prediction: the EHR loop contributes only ~0.1% to plasma E2 AUC and ~0.1% to C_max at steady state on oral 1 mg q24h. Sweeping the deconjugation fraction from 0.05 (broad‑spectrum‑antibiotic flora) to 0.30 (healthy gut) moves AUC by <0.3% — consistent with Adlercreutz 1984's controlled human PK finding that ampicillin co‑administration produces no significant plasma E2 / EE reduction. The folk‑PK 20–30% antibiotic AUC drop story is not supported by either the model or the controlled data; the loop is real and produces the visible plateau on the time course, but its quantitative plasma contribution is heavily attenuated by the same first‑pass that limits oral E2 bioavailability in the first place (each loop traversal sees only ~4% F_oral on its way back to systemic plasma).C3

Sublingual E2

Some fraction of a sublingual dose absorbs across the buccal/sublingual mucosa directly to systemic circulation as free E2, bypassing portal first‑pass; the rest is swallowed and follows the oral route. C_max is ~3–4× oral at the same doseC1 and the curve is distinctly biphasic: a sharp ~1.7 h early decline reflecting free‑E2's true IV‑like clearance, then a slower late tail driven by the smaller E1S reservoir from the swallowed fraction.C2 The absolute direct‑sublingual bioavailability has never been cleanly measured in humans. The v4 model uses ~5% (calibrated against Doll 2022's AUC ratio of 1.8× oral); older textbooks cite 20–25%; the only primary anchor for absolute F is the Kuhnz 1993 marmoset study at ~10%.C2

Sublingual is not reliably liver‑sparing. Cirrincione 2021 (LC‑MS/MS) finds sublingual produces plasma E1 levels closer to oral than to transdermal — meaning the swallowed fraction and the slowly redistributing direct fraction both hit hepatocytes.C1 Bar 2024 (ECE conference abstract, EP592) reports sublingual E2 produces meaningful free protein S depression, similar in magnitude to oral.C4

Measured

Aggregated mean curves from Burnier 1981, Casper & Yen 1981, Fiet 1982, Kuhnz 1993, Price 1997, Wiegratz 2001, Wren 2003, Pickar 2015 — single‑dose 0.25–2 mg micronized estradiol. Sublingual gives a sharp 1–2 h peak followed by fast biphasic decay; oral gives a broad ~8 h peak with smoother decline. Inter‑study C_max spread for 1 mg SL is ~100–400+ pg/mL — a factor of 4. From transfemscience.org sublingual review. Model-vs-data digitization caveat: a pixel-extraction overlay of v4 predictions on the digitized curves (model/figures/v4/plot_overlay_transfemscience.png) shows v4 under-predicts the SL 1 h peak by ~5× (model ~158 pg/mL vs digitized ~770 pg/mL) and over-predicts the oral 12 h AUC by ~70% (model 694 vs digitized 416 pg·h/mL). v4's structural shape (broad oral peak vs sharp SL peak) is correct; the magnitudes need additional calibration.C3

Transdermal

Transdermal patch and gel deliver E2 directly to systemic circulation through skin, bypassing both gut and liver first‑pass. A 50 µg/d patch produces ~50–80 pg/mL steady state with E1:E2 ~1:1 and minimal SHBG induction. Hepatic exposure equals systemic exposure, so the receptor sees free E2 at sub‑K_d levels at typical doses — well below the SHBG EC50.C2

Injectable (IM, SC)

IM ester depots (E2‑valerate t½ ~2 d release, E2‑cypionate t½ ~6.5 d, E2‑undecylate t½ ~22 d) hydrolyze in the depot and slowly release free E2 into systemic circulation. Single‑dose IM E2V 5 mg peaks at ~667 pg/mL around day 2; IM EC 5 mg at ~338 pg/mL around day 4 (peak values commonly understated as ~200/~100 pg/mL in earlier drafts of this synthesis — those are C_avg, not C_peak). SC E2 aqueous and depot routes are increasingly used in trans HRT for self‑injection convenience; PK matches the IM equivalent ester.

Oral esters (E2V, EC, EUn)

Oral E2‑valerate is essentially oral E2 from minute 30 onward; oral EC and EUn are less commonly used orally (the longer aliphatic chains slow ester hydrolysis but most clinical use is parenteral depot). The European HRT and COC market (Climen, Qlaira) uses E2V as the oral E2 source by convention; the US market uses unesterified E2.

3Hepatic ER‑α and SHBG

Hepatic ER‑α (ESR1) is the molecular mediator of route‑dependent estrogen effect on the liver. When activated, it transcriptionally upregulates coagulation factors II, VII, VIII, IX, X, XI, fibrinogen, SHBG, CBG, TBG, angiotensinogen, IGFBPs, HDL, and triglycerides; and downregulates protein S, antithrombin, and IGF‑1. The net hemostatic shift is procoagulant, and it tracks the concentrations the hepatocyte sees — not systemic plasma E2.

The two curves

The single most important framing correction from round‑2 fact‑checking: receptor occupancy and SHBG induction are governed by two different dose‑response curves at very different concentrations. Earlier drafts conflated them under one Hill curve at K_d = 0.25 nM (~68 pg/mL free), concluding “routes below ~68 pg/mL avoid hepatic effects.” This is too crisp.

• ER‑α occupancy tracks free hepatocyte E2 with K_d ~0.25 nM (Vickers et al. 1989, rat hepatocyte).C2 Recombinant human ER‑α sits at K_d 0.05–0.5 nM, so the cross‑species applicability is defensible.
• SHBG mRNA induction in HepG2 cells responds to E2 in the 0.5–2.5 µM range (Kalme 1999), three to four orders of magnitude above the receptor K_d.C1 The mechanism: transcriptional activation requires sustained occupancy and downstream ESR1‑cofactor recruitment, not just transient binding.
• The v4 PK model uses ~1150 pg/mL (~5.5 nM) free hepatic E2 as the SHBG EC50 — a calibration target chosen so the model's predicted oral‑E2 dose‑response (1 mg → ~+91%, 2 mg → ~+130%, 4 mg → ~+166% SHBG)C3 lands inside the measured +67–171% band Ropponen et al. 2005 reported for oral E2 in humans (the 1 mg point is the model's, not the study's — that trial used oral E2 valerate 2–4 mg/d and published no per‑dose breakdown).C1 This is not a measured EC50, and it should NOT be cited to Selva & Hammond 2009 — that paper is on HNF‑4α/thyroid‑hormone regulation of SHBG, not E2 dose‑response. Earlier drafts contained this misattribution.C1

The two timescales

The “two curves” above separate occupancy from induction along the concentration axis. They are also separated along the time axis, and the time gap is what makes the protein readout sluggish. A single oral dose floods the hepatocyte for only minutes during first‑pass, but the protein it ultimately changes — SHBG, clotting factors — turns over across days. The signal therefore passes through a ladder of increasing time constants:

• Ligand–receptor association can be rapid — the on‑rate is near diffusion‑limited and a high‑concentration portal pulse drives the complex toward its equilibrium occupancy quickly — but the estradiol–ER complex has a slow off‑rate: dissociation half‑lives of ~90 min (intact rat uterine cells) to ~2 h (37 °C) are reported for classic ER systems. The Hill occupancy curve below is therefore an equilibrium approximation, not a claim that binding equilibrates in milliseconds.C3
• Receptor occupancy — the fraction of ER‑α bound at any instant — therefore responds on a minutes–hours scale: a high first‑pass E2 spike raises occupancy quickly, but the slow off‑rate means the bound‑receptor signal is smoothed (it integrates over the pulse) rather than tracking the portal concentration second‑by‑second.C3
• Transcriptional output (ESR1‑cofactor recruitment → RNA Pol II → mRNA accumulation) builds over hours; it integrates the occupancy signal rather than mirroring its instantaneous value.C3
• Protein output is the slowest step. Secreted SHBG accumulates in plasma against a multi‑day clearance (terminal t½ ~4 days in primate tracer studies, ~7 days commonly cited clinically), so plasma SHBG is still changing after about two weeks and approaches a new steady state over ~3–4 half‑lives (~12–16 days for the ~4‑day primate terminal phase; ~21–28 days for the ~7‑day human postpartum decline) — well after the dose that drove it has cleared.C2

The hepatic exposure rule

Plasma protein levels track the concentrations the hepatocyte sees during absorption, not the systemic E2 level. Oral E2 at 1 mg/d produces ~50 pg/mL systemic plasma E2, while the v4 model estimates a transient portal/hepatocyte free‑E2 exposure around ~5000 pg/mL during absorption; this portal figure is model‑derived, not directly measured (no human portal‑vein E2 study exists).C3 Pregnancy is the cleanest natural contrast for the “route, not dose” claim: term pregnancy plasma E2 reaches roughly hundreds‑fold above cycling baseline (tens of thousands of pg/mL), yet pregnancy VTE risk is only about 3–10× baseline (CDC summary ~5× overall; trimester/postpartum estimates vary).C1 The route/exposure explanation — that pregnancy delivers estrogen placenta‑to‑systemic rather than as a portal‑vein bolus, so the liver sees a sustained rather than spiked signal — is a mechanistic inference, not a directly measured causal decomposition.C3

Plasma protein binding and the free fraction

SHBG enters the story twice. The previous subsection covered SHBG as a hepatic transcriptional output (the protein the liver makes more of under oral‑style first‑pass exposure). It is also the dominant carrier for E2 in plasma, and the two roles couple: raising SHBG output lowers the free fraction of every steroid SHBG carries, which feeds back on how much hormone reaches peripheral receptors. This subsection treats the carrier role.

Circulating E2 partitions across three pools. In a non‑pregnant adult only ~2–3% is unbound (“free”); ~38% is bound to SHBG and ~60% to albumin.C1 The free‑hormone hypothesis holds that only the unbound steroid crosses cell membranes to reach intracellular ER, so it is free — not total — E2 that sets receptor occupancy. A weaker variant, the bioavailable‑hormone model, counts free + albumin‑bound together, on the grounds that the albumin complex dissociates fast enough during capillary transit to surrender hormone to tissue; the SHBG‑bound pool is treated as sequestered. Which model is “correct” is tissue‑ and transit‑time‑dependent and still debated; both agree the SHBG‑bound fraction is the least available.C3

Albumin's affinity for E2 is three to four orders of magnitude below SHBG's, but its plasma concentration is ~1000× higher (Hammond 2016), so the two carriers end up holding comparable shares of E2.C2 Because neither protein is more than slightly occupied by the small steroid pool, the free fraction has a simple closed form:

Published SHBG–E2 K_d values span ~1.5–30 nM across assays. Structural/equilibrium work reports ~10–30 nM (Avvakumov 2010), whereas the operative constant embedded in clinical free‑hormone calculators (Vermeulen‑type, ~1.5 nM) is several‑fold tighter — partly because SHBG is a homodimer presenting two binding sites, and partly assay convention. The page uses ~20 nM as the molecular constant and a tighter ~2–5 nM operative value for free‑fraction arithmetic. The gap is not a temperature effect: SHBG–steroid affinity actually decreases (weakens) with rising temperature, so it is slightly lower at 37°C than in cold (4°C) assays for all three steroids (PMID 3702439) — the opposite direction from the structural-vs-operative gap.C2

Why this matters for trans HRT: SHBG suppresses free T more than free E2

SHBG does not bind all steroids equally. Its affinity ranks DHT (~1 nM) > testosterone (~3–5 nM) > estradiol (~20 nM) — androgens bind several‑fold tighter than E2.C1 In the free‑fraction equation, a tighter K_d means the [SHBG]/Kd term grows faster as SHBG rises, so the higher‑affinity ligand loses a larger proportion of its free fraction. Oral E2 raises SHBG ~67–171% (dose‑dependent; Ropponen et al. 2005);C1 that same SHBG rise suppresses free testosterone and free DHT proportionally more than it suppresses free E2 — a feature, not a bug, in transfeminine HRT, where the goal is to lower free androgen while maintaining free estrogen. It is also the mechanism behind the antiandrogenic reputation of oral (vs transdermal) estrogen and of EE‑containing combined oral contraceptives.

SHBG is unusually dynamic for a plasma carrier. Across healthy adults it spans ~20–100 nM (men lower, women higher); pregnancy — the largest sustained physiological estrogen exposure — raises maternal SHBG roughly 5–10× over the non‑pregnant level (Anderson 1976: ~5×, with later assays reporting up to ~10× at term).C2 The same study clocks the postpartum decline at a ~7‑day half‑life — an independent human cross‑check on the multi‑day SHBG turnover that sets the ~2–4‑week lag before plasma SHBG reaches a new steady state after any estrogen dose change.C2 Because [SHBG] enters the free‑fraction denominator directly, this several‑fold dynamic range is what makes route‑ and dose‑driven SHBG shifts clinically meaningful for free‑hormone exposure.

The clinical consequence of the slow protein turnover is that SHBG — and the free fraction it sets — lags any dosing change by weeks. Starting, stopping, or switching the route of oral estrogen does not move SHBG to its new level on the next dose; the plasma pool fills (or drains) along a first‑order curve parameterized by the SHBG clearance half‑life, ~4 days (primate tracer) to ~7 days (postpartum human); the exponential time constant is τ = t½/ln 2 (~5.7–10.2 days), so it is somewhat longer than the half‑life itself.C2 A laboratory SHBG (or a free‑testosterone) drawn one week into a regimen change is therefore reading a value still in transit, not the steady state.

The clinical signature of this hepatic effect is its route dependence: the same molecule raises SHBG sharply by mouth and barely at all through skin, because only the oral route delivers the first‑pass portal concentration spike that drives hepatic ER well into the upper-arm region of the SHBG induction curve. Ropponen et al. 2005 measured exactly this contrast in a within‑subject crossover, and it is the single cleanest demonstration that route, not systemic E2 level, sets the hepatic readout.

4The E1S reservoir

Estrone‑3‑sulfate (E1S) is the most abundant circulating estrogen species in every hormonal state. It is metabolically inert at ER itself (sulfate group blocks C3 binding) but is reactivated to E1 in target tissues by STS, then reduced to E2 by HSD17B1. Its plasma half‑life of 10–12 h (Ruder 1972) makes it the rate‑limiting step for the apparent elimination of oral E2.C1

Postmenopausal endogenous E1 production is ~40 µg/day (Longcope 1986; Grodin 1973), not ~80 µg/day as earlier drafts had it — that higher number was likely a premenopausal early‑luteal value misapplied to postmenopause.C1

The reservoir explains the long apparent oral half‑life. Free E2 itself clears with IV t½ of 1–2 h. The 13–20 h apparent half‑life of oral E2 is the rate at which the E1S reservoir bleeds back into the free pool: ~320 µg‑equivalent of a 1 mg oral dose enters systemic E1S, then reactivates over 12–24 h via STS in target tissues.C2 Sublingual loads a smaller reservoir (only the swallowed fraction enters portal first‑pass), giving a shorter, sharper plasma curve. Transdermal builds essentially no oral‑style reservoir — the steady‑state E1S elevation comes from peripheral E2→E1→E1S conversion at standard cycling rates.

5VTE risk by route

The clinical end‑point of route‑dependent hepatic exposure is venous thromboembolism (VTE). Oral estrogen elevates risk roughly 4× baseline; transdermal does not.C1

The ESTHER 4.2× figure is from the original case‑control study (Canonico 2007, Circulation, PMID 17309934); pooled meta‑analyses dilute the signal because they include lower‑dose oral regimens and heterogeneous transdermal protocols. The 4.2 vs 0.9 contrast is the load‑bearing population‑level number behind modern guideline preference for transdermal over oral.C1

The cascade of attenuation (a heuristic)

Pregnancy plasma E2 is ~250× cycling baseline; pregnancy VTE risk is ~3–10×. The intuitive question is how the 250× plasma signal is buffered down to a 3–10× clinical signal. Earlier drafts presented a stepwise multiplicative chain (÷3 SHBG buffering × ÷30 SHBG induction saturation × ÷2 TFPI/protein S compensation × ×3 Virchow's triad) terminating at the empirical risk. The round‑2 fact‑check found multiple invented multipliers in this chain, and it is now framed strictly as a heuristic:

• SHBG buffering of free E2: earlier draft said ÷3; pregnancy SHBG rise drops the free fraction only to ~60% of nonpregnant baseline, so the actual buffering is closer to ÷1.7.C2
• SHBG induction saturation: the dominant attenuator (the dose‑response curve plateaus near pregnancy levels), but the specific ÷30 was modeled, not measured.
• TFPI ↑ / protein S ↑: wrong direction. In pregnancy, free protein S falls to ~38% of nonpregnant (Comp 1986), and TFPI is lower than nonpregnant (~7.1 vs 8.2 ng/mL).C1C1 There is no net anticoagulant compensation; this arm is also procoagulant.
• ×3 Virchow's triad: invented multiplier. Virchow's triad is qualitative; the empirical 3–10× comes from cohort studies (Heit 2005, Pomp 2008, Sultan 2013), not a derivable mechanistic chain.C2

The qualitative claim — plasma E2 ≠ free E2 ≠ hepatic effect ≠ net coagulation ≠ VTE incidence — is fully supported. The specific multipliers are not. Read the chain as a teaching device for the structure of attenuation, not as a sourced calculation.

6Ethinyl estradiol

Ethinyl estradiol (EE) is 17α‑ethinyl‑17β‑estradiol. The single 17α‑ethinyl substitution makes C17 a quaternary carbon — bearing the 17β‑OH, the 17α‑ethinyl (−C≡CH), and the two ring carbons C13/C16 — with no carbinol hydrogen for HSD17B2 to abstract.C1 The 17β‑OH is still there. An earlier draft of this synthesis incorrectly stated that EE has “no C17‑OH” — the round‑2 codex fact‑check caught this. The chemistry is right (HSD17B2 cannot oxidize EE), the prior wording was wrong.C1

The mechanistic consequences:

• EE persists as parent drug. Without HSD17B2 oxidation, the dominant inactivation arm for E2 (E2 → E1 → E1S in the liver) is gone for EE. Clearance proceeds only via sulfation (SULT1E1 forms EE‑3‑sulfate; reversible via STS), glucuronidation (UGT1A1/1A3/1A9, plus UGT2B7 at C17), and CYP3A4/2C9 hydroxylation. All slower than the HSD17B2 arm for E2.
• Gut wall does ~44% of EE first‑pass conjugation. Back & Rogers 1982 cannulated the human portal vein and measured this directly — the single load‑bearing primary measurement for “gut wall is a major first‑pass site,” made specifically on EE.C1 Back 1990 review notes the gut‑wall product is principally sulfate (likely SULT1E1‑dominated, despite SULT1A1 being more abundant, because SULT1E1 wins at physiological substrate concentrations on K_m — the earlier draft had this backwards).C2
• Hepatic extraction of free EE is ~25% on first pass (Back & Rogers 1982), not ~20% as earlier drafts had it.C1 Compare with ~90% extraction for E2 — EE escapes most of the hepatic first‑pass that destroys oral E2.
• Single‑dose oral EE C_max is ~70 pg/mL for 30 µg (extrapolated linearly from Brody/Turkes/Goldzieher 1989 70 µg data, mean C_max ~160 pg/mL).C1 The frequently cited ~95 pg/mL is the steady‑state C_max (Yasmin label; 1.5–2× accumulation factor); earlier drafts mislabeled this as single‑dose.
• EE half‑life is 18–24 h at steady state.C1
• EE does not meaningfully bind SHBG (relative affinity ~0.18 vs E2 ~9, i.e. ~50× weaker).C1 When EE drives SHBG up via hepatic ER, the rising SHBG does not trap EE itself but crashes the free fraction of endogenous E2 — the mechanism by which combined OCs maintain biological activity while suppressing ovarian E2 to near‑castrate.
• Per‑weight hepatic potency vs oral E2 is large but endpoint‑specific: 75–1000× on the various hepatic proteins per Kuhl 2005's weight‑basis table.C1 The “~100× per mole” shorthand is a reasonable average over the SHBG endpoint specifically; it is NOT a universal constant across hepatic outputs. At the peripheral ER, EE is only ~1.5× as potent as E2 per mole.

Measured

Brody, Turkes & Goldzieher 1989, Contraception 40:269 (reproduced by Goldzieher 1990). Single‑dose 2‑tablet (= 70 µg EE) administration to n=24 women. Mean C_max ~160 pg/mL at 1 h; highest‑responder peak ~300 pg/mL — 5× the lowest responder. The 4–6 h plateau is enterohepatic recirculation. The 5× spread at n=24 with the same drug and dose is the inter‑individual variability the model curves do not show. Model-vs-data digitization caveat: v4 overlay (model/figures/v4/plot_overlay_goldzieher.png) shows v4 under-predicts the EE 1 h peak by ~2× (model 78 pg/mL vs digitized 166 pg/mL) but matches the 12–48 h terminal slope (model −0.059/h vs digitized −0.050/h). v4's EE absorption rate is too slow; this is a known model limitation that the §6 EE single‑dose validation hasn't yet been retuned for.C3

SHBG response to combined oral contraceptives (Stegeman 2013)

Source: Stegeman 2013, J Thromb Haemost.C1 Androgenic progestins (LNG) blunt SHBG rise; antiandrogenic ones (drospirenone, cyproterone) amplify it. Diane‑35 reaches roughly pregnancy‑level SHBG. The estrogen choice (EE vs E2‑valerate) matters at least as much as the progestin choice — Qlaira (E2V + dienogest) sits at +50–80% SHBG and ~4–6/10,000 woman‑years VTE, vs Yasmin's +200–300% SHBG and ~10–12/10,000 woman‑years.

7Prodrug strategy: EC508, E2MATE

The unsolved formulation problem: design an oral E2 prodrug that delivers physiological systemic E2 levels without hepatic ER‑α saturation. The chemical approach attempted twice: install a sulfamoyl group (−SO₂NH₂) that tightly ligates the active‑site Zn²⁺ of carbonic anhydrase II (CAII) inside red blood cells. RBC sequestration carries the drug past the liver in the cellular compartment rather than as free drug in plasma; plasma esterases then slowly release E2 in systemic circulation.

E2MATE

E2MATE (estradiol‑3‑sulfamate) installed the sulfamoyl group on the C3 phenolic oxygen. This protected C3 from SULT/UGT and bound CAII, but the C3‑O‑SO₂NH₂ bond is also recognized by STS. Instead of being cleanly cleaved to release E2, the sulfamate covalently modifies STS's active‑site formylglycine residue via mechanism‑based sulfamoyl transfer.C1 The FGly‑N‑sulfate adduct is stable; STS is permanently inactivated; both the prodrug's own activation and the body's normal E1S reactivation cycle are abolished. This is a mechanism‑based irreversible inhibitor, not a transition‑state mimic — earlier drafts of this synthesis described it as a TSI, which is the wrong mechanistic class (round‑2 mechanism fact‑check finding 1).C1 E2MATE has been repurposed as an STS inhibitor for endometriosis.

EC508

EC508 = estradiol‑17β‑(1‑(4‑(aminosulfonyl)benzoyl)‑L‑proline). The redesign moves the sulfamoyl group off the C3 oxygen (where STS attacks) and onto a separate L‑proline‑based handle attached at C17 via an esterase‑cleavable ester linker.C2 Estradiol is intact; the sulfamoyl head ligates CAII inside RBCs; plasma esterases slowly cleave the C17 ester bond to release free E2 plus a prolyl‑aryl‑sulfonamide fragment.

Development status: EC508 was developed by Evestra Pharmaceuticals. As of 2021, the transfemscience.org EC508 article notes “No Further Development.” It never progressed to human clinical trials; small‑pharmaceutical funding constraints are the most plausible explanation. The chemistry works in principle and in animals.C2

8Inter‑individual variability

Real PK variability is much larger than central‑estimate tables suggest. The Goldzieher 1990 reproduction of Brody/Turkes/Goldzieher 1989 shows a 5× spread in single‑dose oral EE C_max across n=24 women receiving identical 70 µg doses.C1 Stanczyk 2013 reviews comparable spreads across oral and transdermal E2.C1 Drivers of variability:

• UGT2B7 and SULT1E1 polymorphisms — affect C17 and C3 conjugation rates; can move oral F by 2–3×.
• UGT1A1*28 (Gilbert's syndrome variant) — reduces hepatic glucuronidation; modestly raises E2 AUC.
• Body composition — adipose tissue expresses aromatase; obese individuals have higher peripheral testosterone→E2 conversion. BMI affects E1S apparent V_d.
• Age — UGT activity decreases ~30% in older adults; gut motility and microbiome change.
• Ethnicity — CYP1A2 and CYP3A4 inducibility varies between populations; affects catechol pathway contribution and EE clearance.
• Gut microbiome composition — controls bacterial β‑glucuronidase activity, hence enterohepatic recirculation magnitude.

Bimodal oral EE bioavailability has been reported in some studies, with a high‑F subpopulation distinct from a low‑F majority.C2 Clinical implication: literature central‑estimate doses are starting points, not endpoints. Real PK has CV ~30–60% on most parameters and follow‑up bloodwork is essential for any HRT regimen targeting a specific E2 range.

What the v4 model says about variability

The deterministic curves elsewhere in this document collapse a population to its central tendency, which is exactly the visual lie the field has been making for decades. To put a number on the envelope, the v4 model was run as a 200‑sample Monte Carlo (12 scenarios × 200 ODE integrations) with nine of its parameters sampled from log‑normals centred on the deterministic values: MCR of E2 / E1 / E1S / EE (CV 28–45%, from Longcope 1968 and Ruder 1972 dispersion), the SULT1E1 and HSD17B2 throughput constants (CV 40–50%, from the SULT/HSD expression‑polymorphism literature), SHBG baseline (CV 30%, Vermeulen 1969 / WHO), and the oral bioavailability of E2 and EE (CV 30–60%, the dominant source of inter‑individual variability).C2

For oral E2, the model produces a p95/p05 steady‑state spread of ~3.8× across the 1 mg, 2 mg, and 4 mg dose levels — matching the Goldzieher 1989 / Stanczyk 2013 reported responder range of 3–5× without explicit tuning (the per-parameter CVs in the Monte Carlo wrapper are drawn from primary literature, not fit to the responder-range target). For oral EE 30 µg in COC, the spread is ~4.7×. For transdermal, IM ester, and SC depots, the spread is tighter at ~2.6× — but this is because the only variance these routes inherit in the model is whole‑body MCR (CV ~28%, log‑normal); the gut F_oral variability that dominates oral PK is bypassed by construction, not modeled away. So the comparison reads “oral spread = MCR variability × F_oral variability, multiplied; non‑oral spread = MCR variability alone.”C2 The takeaway holds clinically — oral routes do amplify population PK variability in real life — but the ~2.6× non‑oral number should be read as the model's MCR‑only floor rather than an independent prediction. A real patient lands somewhere inside the light envelope below, not on the median line.

9Steady‑state vs single‑dose

Single‑dose PK and steady‑state PK differ in three ways relevant to HRT.

Accumulation. For a drug with apparent t½ longer than the dosing interval, plasma concentrations accumulate to steady‑state values of (single‑dose) × (accumulation factor). For oral E2 (apparent t½ ~15 h, q24h dosing) the accumulation factor is ~1.4×. For EE (t½ ~20 h, q24h) it is 1.5–2× (Yasmin label: single‑dose 30 µg C_max ~70 pg/mL → steady‑state ~95 pg/mL).C1

E1S reservoir loading. The reservoir takes ~3–5 days (3–5× E1S t½ of 10–12 h) to reach steady state. Until then, plasma E2 trough is lower than it will be at chronic dosing — new oral HRT patients commonly under‑estimate their eventual steady‑state level from day‑1 measurements.

SHBG transcriptional time constants. SHBG protein turnover sets a slow time constant of ~4 days; chronic dose changes don't show in plasma SHBG for ~2 weeks. Clotting factor changes are faster (~3–7 days) but still substantially slower than the PK time constants. Single‑dose hepatic effects are essentially undetectable; only chronic dosing produces the protein‑synthesis shifts that drive VTE risk.

10Trans HRT specifics

Practical estrogen choices in transfeminine HRT, with the synthesized pharmacology applied to clinical decisions.

(The “Common clinical range” column in the table above reflects ranges in routine clinical use; Endocrine Society 2017 lists oral E2 2–6 mg/d and patch 0.025–0.2 mg/d, but the SC E2 1–6 mg q4–7d and IM EV 4–10 mg q5–7d ranges are common practice patterns not directly tabulated in the guideline. Not dosing advice.)

Target serum E2: guidelines usually frame the goal as roughly 100–200 pg/mL (or simply E2 <200 pg/mL to avoid the supraphysiologic range), with serum testosterone monitored separately toward the female range (~30–100 ng/dL).C2 Within-population PK spread is ~3–5× at a fixed dose (§8), so a single reading is a feedback-loop check, not a measurement of where a particular patient actually sits at steady state — and SHBG / free-T draw at <2 weeks after any dose change reads a value still in transit (§3). The ~100 pg/mL lower bound is commonly cited as roughly the threshold for adequate gonadotropin suppression in monotherapy, but a 2025 systematic review found the evidence does not establish that any specific E2 range guarantees feminization sufficiency — so the range is a pragmatic monitoring target, not a proven feminization optimum.C2 The feminization process itself unfolds over years, not lab draws: Endocrine Society 2017 notes early changes over 3–12 months with breast development generally maximal around 2 years; UCSF guidance describes breast development as usually Tanner 2–3 with maximal feminization taking 2–5 years; clinical endpoints (not a single E2 draw) are what carry the assessment.

Anti‑androgen co‑medication. Spironolactone (in the U.S.) and cyproterone acetate (where available, not US‑marketed) are the two main agents in routine use. Bicalutamide is an off‑label androgen‑receptor blocker used by some clinicians but carries a small but documented risk of liver injury (including rare fulminant hepatitis per LiverTox) and is generally framed as “other approaches” rather than first‑line.C2 GnRH analogs (leuprolide, triptorelin) provide potent gonadotropin suppression and are typically used when other strategies are inadequate, but are commonly limited by cost, insurance/access, and injection/implant logistics — not by efficacy. Estrogen monotherapy (no co‑med anti‑androgen) can achieve gonadotropin–testosterone suppression by negative feedback on the HPG axis, particularly with injectable routes that reach steady‑state E2 levels above ~200 pg/mL; an injectable‑estradiol cohort (PMC11649620 / JCEM 2025) reported >80% testosterone <50 ng/dL among weekly injectors not on GnRHa, but monotherapy was only a minority (~23%) of patients without gonadectomy and median monotherapy E2 was ~250 pg/mL. Oral monotherapy by contrast is not reliably suppressive (Leinung 2018: oral E2 alone “not infrequently unable” to suppress T).C2 A 2025 RCT (Angus et al., JCEM, PMID 39288020) found no breast-growth difference between spironolactone and cyproterone acetate at 6 months despite stronger testosterone suppression with CPA — weakening the case that more T-suppression maps directly to more feminization.C2

Recent (2024-2025) trans-cohort data add three substantive updates to the older HRT literature this section relies on. (a) Slack et al. 2024 (PMID 39233010) reports a 2,126-patient transfeminine retrospective cohort with VTE history 0.8% and estrogen use not independently associated with VTE after adjustment for age/comorbidity — the cis-postmenopausal ESTHER 4× transfer to trans HRT may overstate the absolute risk in younger, healthier trans cohorts. (b) Stangl, Wiepjes et al. 2025 (PMID 40377124) is the first prospective GAHT binding-protein time-course: oral E2 + CPA raises SHBG +81% at 12 months; oral E2 + GnRHa raises SHBG +242%; transdermal raises SHBG only with GnRHa — concrete GAHT-specific numbers (rather than borrowed cis-postmenopausal anchors). (c) Misakian et al. 2025 (PMID 39797602) 562-patient multicenter injectable cohort found median in-target weekly dose was 4 mg E2-equivalent; IM vs SC and EV vs EC differences were not statistically significant — the §10 injectable dose table's q5d vs q14d distinctions matter less than the literature historically implied.C2

EE is contraindicated in modern trans HRT (Endocrine Society 2017, WPATH SOC8). CEE (Premarin) is similarly out of favor for the same hepatic‑disproportion reason.

Route selection in patients with VTE risk factors (factor V Leiden, prior DVT, smoking, obesity, age >45–50) commonly defaults to transdermal or IM‑ester depot rather than oral, with WPATH SOC8 specifically flagging older age and elevated VTE risk as transdermal‑preferred indications.C1 The 4× OR contrast between oral and transdermal is the clinical reason this is the conservative default. Guidelines do not pin a hard BMI cutoff (BMI >30 is one common risk‑factor heuristic but is not a formal route‑selection threshold).

Putting it together: route choice as a three‑way trade‑off

The pharmacology synthesized above collapses into a small number of trade‑offs that actually drive route choice. None of them is unique to trans HRT, but the relative weights are.

1. Hepatic exposure. Routes that pass through portal vein (oral, sublingual, oral esters) expose the hepatocyte to a transient bolus that is much higher than the steady‑state systemic level. The hepatocyte ER‑α sees that bolus and induces SHBG, clotting factor II/VII/IX/X, fibrinogen, and angiotensinogen on a multi‑day time constant (§3). Transdermal, IM‑ester depot, and SC depot avoid this bolus by construction. Empirically (§5) this maps to the ESTHER ~4× VTE OR for oral over transdermal; mechanistically it maps to the hepatic‑exposure rule.C1 — If VTE risk is elevated, this consideration dominates.
2. PK stability. Within‑interval fluctuation differs by ≥an order of magnitude across routes. Transdermal patches deliver near‑constant levels for 3–7 d. The v4 model's directly‑computed peak:trough ratios for the depot routes are IM‑EC q14d 2.11× and SC‑EUn monthly 2.58×;C2 shorter‑interval IM‑EV q5d gives a tighter swing than IM‑EC q14d because the dosing interval is short relative to the depot release time (model does not yet compute the exact ratio — qualitative claim). Oral q24h fluctuates ~2–4× over a dose interval; sublingual q12h fluctuates the most, with a sharp ~1 h peak followed by fast biphasic decay — the transfemscience cross‑study mean curves show peak:trough >5× over a q12h interval.C2 If patient symptoms cluster around mood / breast tenderness / energy fluctuations tied to dosing, smoother routes (patch > IM ester > SC depot > oral > SL) help.
3. Practical convenience and patient‑driven titration. Oral and sublingual are titratable in fine increments and need only swallowing; SC self‑injection is fast and uses a thin needle; IM requires a longer needle and more discipline. Patches need a flat, dry skin site and lose dose on swimming/heat. Gel transfers to family members in casual contact. If adherence is the rate‑limiting input, the safest molecular choice is the route the patient will actually use consistently.

Reading the trade‑off in practice. Most decisions reduce to picking which of the three axes to optimize for and accepting non‑ideal performance on the other two:

• VTE‑dominant (factor V Leiden carrier, prior DVT, age >45–50, smoker, elevated BMI or other documented VTE risk factor): the pharmacology favors transdermal patch or IM E2‑valerate over oral; the trade‑off is higher cost (patch) or injection logistics (IM).
• Monotherapy‑dominant (avoiding spironolactone side effects, no GnRH access): a route that drives SHBG up while delivering high systemic E2 helps suppress free testosterone via the protein‑binding shift. High‑dose oral E2 (4–6 mg/d) does both. High‑dose IM achieves the same gonadotropin suppression via hypothalamic‑pituitary negative feedback (HPG axis) rather than via SHBG induction. The hepatic‑SHBG arm of oral is doing useful work here, but the same arm is the VTE risk.
• Stability‑dominant (mood lability tied to peaks/troughs, sensitive breast tissue, perimenopausal‑style symptoms): the smoothness ranking is patch > IM ester > SC depot > oral > SL. q3.5d patch changes get the closest to the cycling‑follicular smoothness the body originally evolved with.
• Adherence‑dominant (early in HRT, traveling lifestyle, needle aversion): the pharmacology's preferred route is moot if the patient can't take it consistently. Oral or SL is the cost‑of‑consistency choice; re‑evaluating route as life circumstances allow more‑clinically‑optimal options is the long arc.

What target serum E2 does and doesn't tell you. The 100–200 pg/mL window is a feedback‑loop monitoring target (will this dose suppress LH/FSH? Are we in the supraphysiologic range that raises VTE risk further?), not a feminization target.C2 Two patients with identical 150 pg/mL serum E2 can have very different breast development, mood, and body‑composition trajectories — depending on the time integral of free‑E2 over months (route- and adherence-dependent), androgen receptor signaling (residual testosterone level), and individual receptor / target‑tissue genetics. A single‑timepoint serum E2 is the cheapest‑to‑measure proxy for “the system has the substrate it needs,” not for “the patient is achieving their feminization goals.” The latter is set by clinical follow‑up, not laboratory chemistry.

What this synthesis can't tell you. The biggest gaps in the field are: (a) no head‑to‑head RCT comparing routes for trans HRT feminization endpoints (most data are observational or from cis postmenopausal HRT trials); (b) no validated free‑E2 assay in routine clinical use, so all monitoring is on total E2 with the free fraction inferred; (c) no longitudinal trans-population PK data of the granularity that would let the v4 model (or any model) be calibrated against a real trans HRT cohort instead of borrowed cis-postmenopausal anchors; and (d) the patient‑experience endpoints (mood, energy, body composition, dysphoria) are not measured in the studies that report serum E2 levels, so the inferential chain “serum E2 → feminization” is more model than data.C2

11Confidence index

The C1–C5 inline claim tags throughout this page each link to the specific entry in evidence.html that supports (or doesn't support) the claim. (The tier badges in the legend table below are unlinked reference labels, not claims; hover any of them to see the tier definition.) A non‑exhaustive map of the load‑bearing claims by tier:

The C1–C5 system is intentionally generous — every numerical or mechanistic claim above carries a tag. The pattern: basic enzyme biochemistry and population‑level PK numbers are stable C1; model‑calibrated quantities and inferred mechanisms degrade to C2–C3; preliminary clinical signals and unmeasured chemistry are C4–C5.

The same provenance discipline applies to the figures. Every plot on this page carries a corner badge stating what kind of object it is, so a reader never has to guess whether a curve is data or inference: Measured = points plotted directly from a cited primary or literature source; Modeled = a curve computed from a model (Hill saturation, free‑fraction equation, the v4 PK model) whose shape follows from assumptions, not measured points; Schematic = an illustrative structure or mechanism diagram carrying no quantitative claim. Hover any badge for the specifics of that figure. (The badge labels are an editorial provenance taxonomy, not a primary-source datum.)C3

What round‑2 fact‑checks moved (since the earlier narrative)

Listed compactly; full discussion in evidence.html §11.

• SHBG EC50 1150 pg/mL (was 1500 in earlier drafts; re-tuned when the operative free-fraction Kd was anchored to PMC7839820 binding split) — re‑framed as v4 calibration target, not measured. Selva & Hammond 2009 misattribution removed; Kalme 1999 is the right HepG2 anchor (at 0.5–2.5 µM).C1
• EE has a 17β‑OH — earlier “no C17‑OH” wording corrected. C17 is quaternary; HSD17B2 fails because there is no carbinol H to abstract, not because the alcohol is missing.C1
• E2MATE inhibits STS by covalent modification of formylglycine, not by transition‑state mimicry.C1
• Cascade buffering ÷3 → ÷1.7; TFPI in pregnancy decreases (procoagulant), not increases; ×3 Virchow's triad is invented.C1C2
• EE single‑dose C_max ~70 pg/mL for 30 µg; earlier ~95 was steady‑state mislabeled as single‑dose.C1
• EE hepatic extraction ~25% (Back & Rogers 1982), not ~20%.C1
• Sublingual direct F is uncertain; v4 uses 5%, older textbooks 20–25%, marmoset 10%, human absolute not cleanly measured.C2
• Gut SULT1E1 dominates over SULT1A1 at physiological [E2] (K_m argument, not abundance); earlier draft had this backwards.C2
• SULT1E1 lowest K_m “for estrogens”, not “for any substrate.”C2
• AUC values in narrative time‑course curves are model integrals, not published anchors. Doll 2022 only directly reports AUC(0–8) and a 1.8× SL/oral ratio.
• E2 mass balance fractions are v4 calibration outputs, not radiolabel data.C3

12See also

• evidence.html — per‑claim primary‑source quotes, figures, and confidence tags (64 entries)
• reference.html — glossary, enzyme list, PK parameter definitions, papers (~110 anchors)
• synthesis‑narrative.html — the long‑form narrative version (156 KB), preserved as archive with full v4 model figures and complete pre‑round‑2 fact‑check discussion
• Interactive v4 PK viewer — pick a scenario, scrub time, watch concentrations and fluxes update
• transfemscience.org — literature‑reviewed PK aggregations for trans HRT
• Wikipedia: Pharmacokinetics of estradiol; Wikipedia: Ethinylestradiol — well‑referenced indices to primary literature
• Aly's hormones — community PK references for IM (paper‑aly‑im) and SL (paper‑aly‑sl) routes