Per-claim primary-source quotes, figures, and confidence tags backing the synthesis narrative.
This page is the cited-evidence companion to synthesis.html. For each load-bearing claim in the synthesis, the corresponding entry below quotes the primary source verbatim (with citation and URL), tags it with a confidence tier C1–C5, and notes contradictions or caveats. The synthesis links forward here via anchor IDs; you can also browse this page directly. Where the synthesis says “Vickers 1989 reports a Kd of 0.25 nM,” the corresponding entry below shows the actual sentence from the paper, plus what the fact-check passes added or corrected. Numerical fact-checking was run in three rounds (Claude + codex round 1, then Claude PK + Claude mechanism + codex round 2); the round-2 corrections are surfaced explicitly in §11.
The synthesis “Molecule and metabolism” section (“The enzymes that act on those positions”) leans on a few enzymological claims that are well-established in the primary literature. The most load-bearing are HSD17B2 oxidative / HSD17B1 reductive directionality, SULT1E1's low Km for estrogens, and the UGT regiospecificity at C3 vs C17.
← synthesis: Molecule and metabolism (enzyme table)
Claim: 17β-HSD type 2 dominates in adult liver and gut, prefers NAD⁺, catalyzes oxidative inactivation of E2 to E1 (and testosterone to androstenedione). This is the molecular basis of the “liver quietens, target tissues reawaken” model.
“HSD17B2 prefers NAD⁺ and catalyzes oxidative inactivation of E2 to E1 (and also testosterone → androstenedione). It is highly expressed in liver, gut, endometrium, and placenta.”
— Compiled from Wu et al. 1993; Mindnich/Adamski reviews; UniProt P37059. As summarized inresearch/fact-check.md A2, citing the primary expression-cloning paper PMID 8099587.
C1 Confirmed by primary expression cloning + multiple subsequent reviews.
← synthesis: Molecule and metabolism
Claim: HSD17B1 is the reductive isoform, NADPH-dependent, highly expressed in placental syncytiotrophoblast and ovarian granulosa cells.
“HSD17B1 strongly prefers NADPH and catalyzes E1 → E2 reduction. Highly expressed in placental syncytiotrophoblast and ovarian granulosa cells.”
— Peltoketo et al. 1999; Lin et al. 2006. Summary viaresearch/fact-check.md A3. Side note: HSD17B1 has modest reverse activity for E2→E1, but in vivo NADPH/NADP⁺ ratio strongly favors reduction (PMC2736091).
C1 Confirmed.
← synthesis: Molecule and metabolism
Claim: SULT1E1 has a Km for E2 of ~5–20 nM, the lowest of any human SULT for estrogen substrates. Above ~100 nM substrate inhibition kicks in (the “bell-shaped” SULT1E1 curve).
“Among the SULT isoforms, SULT1E1 has the lowest Km values for estrone (E1), estradiol (E2), and catecholestrogen sulfation.”
— Estrogen Sulfotransferase review, PMC8001535. Km value 5–20 nM compiled from Falany 1997; see also PMID 12922923 and the SULT1E1 mechanism paper doi:10.1021/bi902190r.C1 Confirmed for “lowest Km for estrogens”.
Round-2 mechanism fact-check finding 8: an earlier synthesis draft said SULT1E1 had the lowest Km of any human SULT for any substrate. That stronger universal claim is plausible (most other SULTs have μM Km for their substrates) but isn't stated as such in primary sources. Soften to “lowest among human SULTs for estrogen substrates.”
← synthesis: Molecule and metabolism
Claim: The standard human UGT regiospecificity for estradiol: phenolic-C3 glucuronidation by UGT1A1, 1A3, 1A8, 1A10; secondary-alcohol-C17 glucuronidation by UGT2B7. UGT1A10 has the highest E2-3-G activity.
“This is the standard regiospecificity reported by Lépine et al. 2004 (JCEM) and reviewed in Itäaho et al. 2008 (Drug Metab Dispos). UGT1A10 has highest E2-3-G activity; UGT2B7 dominates E2-17-G.”
— Summarized inresearch/fact-check.md A5, citing Lépine 2004 JCEM (“Specificity and regioselectivity of the conjugation of estradiol… by human UDP-glucuronosyltransferases”) and Itäaho 2008 Drug Metab Dispos.
C1 Confirmed by direct enzyme-panel measurements.
← synthesis: Molecule and metabolism (mass balance)
Claim: UGT1A10 in gut wall has ~10× the activity of hepatic UGT1A1 on estradiol — the molecular basis of large gut-wall first-pass conjugation. The correct primary citation is Basu et al. 2004, not Strassburg 1998 (which only characterized expression patterns).
“UGT1A10 in cells supported 10-fold higher glucuronidation of 17β-estradiol than UGT1A1.”
— Basu NK et al. 2004, J Biol Chem 279:28320, PMID 15117964. (As re-attributed in round-2 mechanism fact-check finding 6; earlier synthesis drafts misattributed this number to Strassburg 1998.)C1 Direct quote from primary kinetic measurement.
UGT1A8 and UGT1A10 are the gut-specific isoforms; UGT1A10 is predominantly intestinal (Tukey & Strassburg 2000; Strassburg 1998). The earlier draft conflated Strassburg's discovery of expression patterns with Basu's quantitative kinetic measurement.
← synthesis: Molecule and metabolism (STS row)
Claim: Steroid sulfatase (STS) reverses sulfation in many tissues including target/breast tissue; this is the basis of the SULT1E1↔STS sulfation/desulfation pump and of the STS-inhibitor therapeutic concept.
“Steroid sulfatase (STS) hydrolyzes E1S → E1 and DHEAS → DHEA. Expressed widely; the breast/endometrial expression is the basis for the STS-inhibitor therapeutic concept.”
— Reed et al. 2005, Endocr Rev. Summary viaresearch/fact-check.md A7.
C1 Standard textbook biochemistry.
← synthesis: Molecule and metabolism (HSD17B2 directionality)
Claim: Cytosolic free [NAD⁺]/[NADH] in fed hepatocyte is ~700, strongly driving the HSD17B2 equilibrium toward E2 oxidation (E2 → E1). This is the thermodynamic underpinning of liver-as-inactivator.
“Williamson, Lund, and Krebs calculated the mean free cytoplasmic [NAD+]/[NADH] ratio in rat liver as 725 in fed rats, 528 in starved rats, and 208 in alloxan-diabetic rats. This is still repeatedly cited as a canonical estimate for free cytosolic redox state, although it is rat liver and inferred from metabolite indicator systems, not a direct human hepatocyte measurement.”
— Williamson DH, Lund P, Krebs HA. 1967. Biochem J 103:514. PMID 4291787. Summary via codex fact-check #2.C1 Confirmed for the canonical fed-state value.
Caveat: this is rat liver, inferred from the lactate/pyruvate ratio (LDH equilibrium), not a direct human measurement. The 200–1000 range covers fed/fasted/diabetic states. The synthesis's downstream derivation (Keq × NAD⁺/NADH → cytoplasmic E1/E2 ≈ 11) is C3 inferred — see below.
← synthesis: Molecule and metabolism (footnote)
Claim: Hepatocyte interior equilibrates to ~5–10:1 E1:E2 under HSD17B2, with the high end (~11) being a thermodynamic estimate the original synthesis built. Empirically, oral E2 produces plasma E1/E2 of ~3–5.
“If E°' E1/E2 is near −0.26 V and free cytosolic [NAD+]/[NADH] is about 700, the Nernst calculation gives an E1/E2 ratio in the rough single-digit to tens range; using −0.260 V gives about 6, while −0.275 V gives about 20. So ~11 is a plausible midpoint, not a measured cellular ratio, and it depends strongly on the unverified formal potential and pH/proton convention.”
— codex fact-check #3, citing Williamson 1967 (NAD ratio) and noting that no primary electrochemical measurement of E°'(E1/E2) was located.C3 Inference plausible; arithmetic correct given inputs; the inputs themselves (NAD ratio in human hepatocyte, E°' for E1/E2 couple) are not robustly measured.
The empirical anchor is the observed plasma E1/E2 ratio of 5–7 in patients on oral E2 (Kuhl 2005) — this is what the synthesis should lean on. The thermodynamic chain is qualitative support, not a load-bearing quantitative prediction.
Synthesis §3 (“The oral path”) and §4 (routes table) make several PK claims grounded in Kuhl 2005, Stanczyk 2013, and Doll 2022. The most load-bearing: ~5% bioavailability, 13–20 h apparent half-life from the E1S reservoir, Cmax ~35 pg/mL at 8 h for 1 mg, plasma E1/E2 ratio ~5:1.
← synthesis: Pharmacokinetics by route
Claim: Oral E2 systemic bioavailability for free unconjugated E2 plasma AUC is approximately 5%. This 5% figure refers to free-E2 AUC; total estrogen exposure (E2 + E1 + E1S + glucuronides) is much higher (~30%), and the hepatic effect is effectively saturated regardless.
“The absolute bioavailability of oral micronized estradiol is approximately 5%, with a possible range of 0.1% to 12%.”
— Wikipedia, Pharmacokinetics of estradiol, “Absorption and bioavailability / Oral administration.” Primary anchor: Kuhl H. 2005, Climacteric 8 Suppl 1:3–63. PMID 16112947.C1 Confirmed; nuance below.
The fact-check noted that “5%” alone is sloppy without specifying which compartment. For free-E2 plasma AUC, F ≈ 2–5%. For total estrogen (E2+E1+E1S+glucuronides), F ≈ 20–40%. For hepatic estrogenic effect (SHBG induction etc.), the relevant exposure is the high portal concentration during absorption — neither plasma F nor total-estrogen F describes it well. 1 mg oral PO ≈ ~100 μg/d transdermal in hepatic equivalence (Kuhl 2005; transfemscience “Approximate Comparable Dosages of Estradiol”).
← synthesis: Pharmacokinetics by route
Claim: Standard plasma E2 protein binding distribution in non-pregnant women, follicular phase: ~2–3% free, ~38% reversibly bound to SHBG, ~60% reversibly bound to albumin.
“Approximately 38% of estradiol is reversibly bound to SHBG and 60% is reversibly bound to albumin in women under normal physiological circumstances, with 2 to 3% of total estradiol circulating free or unbound at any given time.”
— Wikipedia, Pharmacokinetics of estradiol, “First-pass effect and differences from other routes.” Anderson 1974 / Dunn 1981 classics.C1 Confirmed.
Round-2 finding: this means “free E2 = 50 pg/mL in cycling women” (in older drafts) is a labeling error — the 50 pg/mL is total serum E2; free is ~1 pg/mL at 2% free fraction.
← synthesis: Pharmacokinetics by route
Claim: Apparent half-life of oral E2 is 13–20 h, but this reflects E1S reservoir decay, not free-E2 metabolism. IV-bolus free-E2 half-life is only 1–2 h.
“Oral: 13–20 hours” (elimination half-life, oral).
— Wikipedia, Pharmacokinetics of estradiol, PK data table. IV t½ 1–2 h: Düsterberg & Nieuweboer 1985, Hormone Res 21:145, PMID 2987096.C1 Confirmed; mechanistically driven by E1S reservoir.
← synthesis: Pharmacokinetics by route (time-course)
Claim: 1 mg single-dose oral E2 produces Cmax ~35 pg/mL at ~8 h post-dose — the modern LC-MS/MS-anchored value, used as the central curve in the synthesis time-course plot.
“sublingual E2 had a significantly higher peak serum E2 concentration of 144 pg/mL, measured using LC-MS/MS, compared with an oral E2 concentration of 35 pg/mL”
— Doll TG et al. 2022, “A pharmacokinetic comparison of sublingual vs oral estradiol in transgender women,” Endocr Pract 28:237, PMID 34781041. n=10, single 1 mg dose each route, 1-week washout.C1 Direct measurement.
← synthesis: Pharmacokinetics by route (anchors list)
Claim: Oral 1 mg/d steady state in postmenopausal women: plasma E2 ~30–50 pg/mL, E1 ~150–300 pg/mL, E1S ~2,000–4,000 pg/mL. The plasma E1/E2 ratio of ~5–7 is the hallmark of oral dosing.
“Stanczyk 2013 and Kuhl 2005 give very similar numbers: oral 1 mg micronized E2 in PMP women → mean E2 30–50 pg/mL, mean E1 150–300 pg/mL, E1S ~2,000–3,000 pg/mL. The ratio E1/E2 ≈ 5–7 is the hallmark of oral E2.”
—research/fact-check.md C2, citing Kuhl 2005 PMID 16112947 and Stanczyk FZ et al. 2013, Contraception.
C1 Confirmed via two independent review-level anchors.
← synthesis: Pharmacokinetics by route (mass balance figure)
Claim: The 1 mg oral E2 mass balance shown in the synthesis (40 μg free E2 / 300 μg E1 / 320 μg E1S / 290 μg glucuronides / 50 μg other) is the v4 model's calibrated fate estimate, not a primary radiolabel human study.
“A single administered dose of estradiol is absorbed 15% as estrone, 25% as E1S, 25% as estradiol glucuronide, and 25% as estrone glucuronide.”
— Wikipedia, Estrone sulfate (medication), mass balance summary citing primary literature.C3 The synthesis's diagram uses v4 model-calibrated percentages (30% E1, 32% E1S, 29% glucuronide); the Wikipedia/Kuhl textbook mass balance gives looser fractions (15% E1, 25% E1S, 50% glucuronides). The synthesis's caption explicitly flags this as “v4 calibration, not a directly measured radiolabel mass balance study.”
Round-2 PK fact-check finding 6: the F_oral_E1 fraction was bumped from 0.18 to 0.30 specifically to match observed plasma E1/E2 ≈ 5 at steady state. The qualitative point (most of the dose is transformed, only ~40 μg ends up as free systemic E2) holds; the exact buckets should be read as model output. No human portal-vein cannulation study has been done on E2 (the Back & Rogers 1982 study was on EE, not E2).
← synthesis: Molecule and metabolism (model parameters)
Claim: Whole-body MCR of E2 in women is ~1,400 L/day (or ~1,360 ± 40 L/d/m² BSA-normalized); MCR of E1S is much lower at ~80 L/d/m² (~140–150 L/d absolute).
“Whole-blood MCR of E2 in women: 1,360 ± 40 L/d/m² → ~2,300 L/d for a 1.7 m² woman, but the 'absolute' 1400 L/d figure used by the AI is what gets quoted in textbooks when you don't normalize, and is roughly OK. MCR of E1S: ~80 L/d/m² (Ruder et al. 1972, JCI; PMC302214) → ~140 L/d absolute.”
—research/fact-check.md B6, citing Longcope, Layne, Tait 1968, J Clin Invest 47:93, PMC297151; Ruder et al. 1972, JCI, PMC302214.
C1 Confirmed for order of magnitude and ordering.
Minor: MCR_E1 in women is closer to 2,700–3,200 L/d when de-normalized to a 1.7 m² body; the synthesis's working figure of 2,200 L/d is ~30% low. Doesn't change conclusions.
Synthesis §3 and §7 make heavy use of the Doll 2022 LC-MS/MS measurement and the transfemscience aggregated review. Three key claims: Cmax 144 pg/mL at 1 h, AUC ratio 1.8× vs oral, and biphasic decay with an inferred late-tail half-life of ~13 h.
Primary-data anchor for the synthesis's oral and sublingual E2 curves. Aggregated mean curves from Burnier 1981, Casper & Yen 1981, Fiet 1982, Kuhnz 1993, Gausau & Mahler 1993, Price 1997, Wiegratz 2001, Wren 2003, Pickar 2015; single dose 0.25–2 mg micronized E2. Reproduced from transfemscience.org sublingual E2 review. Inter-study Cmax variability for 1 mg SL spans ~100–400+ pg/mL.
← synthesis: Pharmacokinetics by route (SL curve)
Claim: Sublingual 1 mg E2 in transgender women produces Cmax ~144 pg/mL at ~1 h, with AUC(0–8 h) 1.8× that of oral 1 mg in the same subjects.
“sublingual E2 had a significantly higher peak serum E2 concentration of 144 pg/mL, measured using LC-MS/MS, compared with an oral E2 concentration of 35 pg/mL… The area under the curve (AUC) (0–8 hours) for sublingual E2, measured using LC-MS/MS, was 1.8-fold higher than the AUC (0–8 hours) for oral E2.”
— Doll TG et al. 2022, “A pharmacokinetic comparison of sublingual vs oral estradiol in transgender women,” Endocr Pract 28:237, PMID 34781041. n=10, crossover with 1-week washout.C1 Direct LC-MS/MS measurement.
Round-2 PK fact-check finding 1 (and Cirrincione 2021): older RIA-era studies sometimes reported Tmax ‘~30 min’ for sublingual — this likely reflects first detectable rise or RIA over-reading at low concentrations, not the actual peak. Doll's LC-MS/MS finds Tmax = 1 h.
← synthesis: Pharmacokinetics by route
Claim: The often-quoted “20–25% sublingual bioavailability” is not directly measured in humans; the only primary IV-comparator data is Kuhnz 1993 in marmoset monkeys at 10%. The v4 model uses 5% direct + ~75% swallowed.
“A study in marmoset monkeys found that the absolute bioavailability of sublingual estradiol was 10%; approximately twice that of conventional absolute bioavailability estimates of oral estradiol (5%).”
— transfemscience.org, “Sublingual Estradiol as an Alternative to Oral Estradiol in Transfeminine People”, citing Kuhnz et al. 1993, Arzneimittelforschung 43:966.C2 Marmoset anchor confirmed; direct human absolute F has not been cleanly measured.
Round-2 PK fact-check finding 4 and codex round-2 finding 4: the synthesis's earlier “20–25% direct” was inferred from relative AUC ratios (1.8–5× vs oral), not measured. There's an internal inconsistency in older drafts: §3/§4 said 20–25%, but the v4 model uses 5%. The transfemscience review is also explicit about the unquantified swallowed fraction.
← synthesis: Pharmacokinetics by route (SL curve shape)
Claim: The directly-absorbed fraction of sublingual E2 clears with free-E2 kinetics (t½ 1–2 h) over the first ~4 hours; the slower late tail (~13 h) reflects the E1S reservoir from the swallowed fraction. The biphasic shape is supported but the exact 13 h late-tail t½ is inferred, not directly measured.
“Estradiol levels are found to rapidly rise on the order of about five to ten times that of the peak of oral estradiol, then rapidly decline, with an elimination half-life of only a few hours.”
— transfemscience.org sublingual E2 review. Original biphasic observation: Price TM et al. 1997, Obstet Gynecol 89:340, PMID 9052581 — “estradiol levels drop steeply within 4 hours, and this is followed by a more gradual decline.”C2 Biphasic shape confirmed by Price 1997. C3 for the specific 13 h late-tail half-life (inferred from E1S reservoir kinetics; not directly measured in any SL study with sampling to 24 h).
← synthesis: Pharmacokinetics by route
Claim: The clean “sublingual bypasses first-pass” model is naive; significant swallowed fraction means the route is mixed sublingual + oral delivery.
“Because accidental swallowing of some of the estradiol seems probable, the sublingual route is, most likely, actually a combination of sublingual and oral delivery of estradiol.”
— transfemscience.org sublingual E2 review.C1 Directly stated in the canonical review.
← synthesis: Pharmacokinetics by route
Claim: Sublingual E2 produces plasma E1 levels much closer to oral than to transdermal; SHBG induction is in the oral range; Bar 2024 finds clinically significant free protein S decrease with low-dose sublingual.
“Cirrincione 2021 and Bar 2024 both show sublingual generates higher E1 and E1S than transdermal — markers of meaningful hepatic estrogenic exposure…The Bar 2024 abstract (ECE 2024; endocrine-abstracts.org ea0099ep592) reports a clinically significant decrease in free protein S under low-dose sublingual E2 in treatment-naïve trans women — consistent with meaningful hepatic estrogenic exposure.”
—research/fact-check.md C7 and D9, citing Cirrincione LR et al. 2021, “PK of sublingual vs oral estradiol in transgender women,” Endocrine Practice, PMID 34781041, and Bar et al. 2024, ECE abstract EP592.
C2 Cirrincione direct measurement is C1; Bar abstract is preliminary so C4. The composite “sublingual has meaningful hepatic effects” claim is C2.
The synthesis's earlier strong claim “sublingual ≈ oral on hepatic side” was probably over-corrected (codex round 1). Honest framing: sublingual produces intermediate-to-high hepatic effect, likely greater than transdermal, but no direct VTE outcomes data exists. The mechanism (peak-responsive protein S vs AUC-responsive SHBG) is mechanistically plausible but unproven.
Synthesis §8 (“Ethinyl estradiol: the structural outlier”) is anchored to Goldzieher/Brody 1989 single-dose curves (figure below), Klipping 2012 pop-PK, Yasmin label SS data, and Back & Rogers 1982 portal-vein extraction.
Primary-data anchor for EE single-dose kinetics. Reproduced from Goldzieher JW 1990, “Selected aspects of the pharmacokinetics and metabolism of ethinyl estrogens,” Am J Obstet Gynecol 163(1 Pt 2):318–22, PMID 2196804 (Figure 1), originally Brody, Turkes & Goldzieher 1989, Contraception 40:269–84. Single-dose 2-tablet (= 70 μg EE) administration to n=24 women. Mean Cmax ~160 pg/mL at 1 h; highest-responder peak ~300 pg/mL is 5× the lowest responder.
← synthesis: Ethinyl estradiol
Claim: Oral EE bioavailability is approximately 45% (range 38–48%), an order of magnitude higher than oral E2's ~5%. The reason is the 17α-ethinyl group blocking HSD17B2 oxidation, leaving EE only sulfation, glucuronidation, and CYP hydroxylation as clearance routes.
“Bioavailability: 38–48%”
— Wikipedia, Ethinylestradiol, clinical data table. Anchored to Stanczyk FZ, Archer DF, Bhavnani BR 2013, Contraception 87(6):706, PMID 23375353.C1 Multiple sources converge.
← synthesis: Ethinyl estradiol
Claim: EE is on the order of 100× more potent than natural estrogens for hepatic protein induction (SHBG, CBG, TBG, angiotensinogen), but only ~1.5× at peripheral ER. The disproportion is hepatic, not general.
“Orally, ethinylestradiol is on the order of 100 times as potent by weight as natural estrogens like micronized estradiol and conjugated estrogens, which is largely due to substantially greater resistance to first-pass metabolism.”
— Wikipedia, Ethinylestradiol. Primary: Kuhl H. 2005, Climacteric 8 Suppl 1:3–63, PMID 16112947.C1 for the order of magnitude. C2 for the exact “100× per mole” figure (depends on endpoint).
Codex round-2 finding 8: Kuhl's Table 3 actually gives EE SHBG potency of 50,000 vs E2 = 100 on a weight basis, i.e. ~500×, not ~100×. Other summaries give lower equivalence values. “~100× per mole” is a modeling shorthand; the per-endpoint range is 75–1000× (Kuhl 2005). The synthesis should treat this as derived shorthand, not a universal measured constant.
← synthesis: Ethinyl estradiol (EE mass balance)
Claim: In EE, C17 is quaternary (OH, ethinyl, C13, C16) with no carbinol H to abstract, so HSD17B2 cannot oxidize EE to an “E1-equivalent.” This is why EE persists in the liver. The C17-OH is still present — the ethinyl blocks oxidation by removing the carbinol H, not by removing the OH.
“The 17α-ethynyl group prevents oxidation of the C17β position of ethinylestradiol by 17β-HSD, and for this reason, ethinylestradiol is not inactivated in these tissues…”
— Wikipedia, Ethinylestradiol. Independently: “the 17a-ethinyl group prevents the oxidation of the 17b-hydroxy group” — Kuhl H. 2005 PDF.C1 Directly stated in two independent sources.
Round-2 codex finding 1: an earlier synthesis draft said EE has “no C17-OH” in the mass balance diagram — that is a hard structure error. The C17-OH is preserved (EE is 17α-ethinyl-17β-estradiol); the ethinyl blocks oxidation of that OH by removing the C17 carbinol H. PubChem confirms: “Ethinylestradiol can be glucuronidated by UGT1A1, UGT1A3, UGT1A4, UGT1A9, and UGT2B7” — if the C17-OH were absent, UGT2B7 (the C17 glucuronosyl transferase) couldn't act on it.
← synthesis: Ethinyl estradiol (EE mass balance)
Claim: The single load-bearing primary measurement for “gut wall is a major first-pass site” comes from Back & Rogers 1982, who cannulated the portal vein in humans dosed with EE. They found gut-wall conjugation fraction = 0.44; hepatic = 0.25.
“The fraction of conjugated ethinylestradiol in the gut wall was 0.44, while the conjugated fraction in the liver was only 0.25. Calculations showed that the gut wall appeared to be twice as effective as the liver in conjugating ethinyloestradiol on the first pass.”
— Back DJ et al. 1982, “The gut wall metabolism of ethinyloestradiol,” Br J Clin Pharmacol, PMID 7059434, PMC1402099. Cited extensively in subsequent Simcyp PBPK work (EG = 0.44, EH = 0.25, FG = 0.56).C1 Direct measurement; the load-bearing primary number.
Round-2 PK fact-check finding 2: an earlier synthesis draft used “~20% hepatic extraction” for EE; the correct Back & Rogers number is 25%. This shifts the EE mass-balance arithmetic slightly: 30 μg × (1 − 0.44) × (1 − 0.25) = 12.6 μg systemic free EE, F ≈ 42% (still within the 38–48% Stanczyk range). Back's 1990 review adds: “Ethinyl estradiol … is extensively metabolized, principally to a sulfate conjugate” (Back 1990, Am J Obstet Gynecol, doi:10.1016/0002-9378(90)90554-K) — so the gut-wall product is principally sulfate, not predominantly glucuronide as analogizing from E2 would suggest.
← synthesis: Pharmacokinetics by route (EE curve)
Claim: Single-dose 30 μg EE Cmax is ~50–70 pg/mL at 1.5 h; steady-state Cmax (Yasmin label) is ~95 pg/mL. The 1.5× accumulation factor reflects deep-compartment loading.
“Following a single dose, maximum serum concentrations of ethinyl estradiol of 62 ± 21 pg/mL are reached at 1.5 ± 0.5 hours (note this was for a 20 mcg formulation)… at steady state, attained from at least day 6 onwards, maximum concentrations of ethinyl estradiol were 77 ± 30 pg/mL.”
— Wikipedia Ethinylestradiol PK summary, citing Alesse label and Klipping 2012 pop-PK, PMC3632974. Linear scaling 30/20 × 62 = 93 pg/mL.C1 Confirmed.
Round-2 PK fact-check finding 1: an earlier synthesis draft had the EE single-dose curve peaking at ~95 pg/mL (which is the steady-state Yasmin value, not single dose). Corrected to ~70 pg/mL for single-dose at 30 μg.
← synthesis: Ethinyl estradiol
Claim: EE single-dose t½ is ~18 h; steady-state t½ extends to ~20–24 h due to slow release from the deep peripheral compartment.
“Single-dose t1/2 ~10–16 h; steady-state t1/2 ~ 17–24 h because of slow release from the deep peripheral compartment. The 7–36 h range in Wikipedia covers all reported study means.”
—research/EE-parameters.md compilation from Klipping 2012 (PMC3632974), Devineni 2007 (PMC4285808), Yasmin label (NDA 021098).
C1 Multi-source convergent.
← synthesis: Ethinyl estradiol
Claim: EE relative SHBG binding affinity is 0.18 vs E2's 8.7–12, a ratio of ~50×. EE essentially doesn't bind SHBG — explaining why EE in COCs maintains free-fraction biological activity while suppressing endogenous E2 to near-castrate levels.
“Wikipedia: EE relative SHBG binding affinity = 0.18 vs E2 = 8.7–12 → ratio ~50–67×. EE doesn't meaningfully bind SHBG.”
—research/fact-check-mech-round2.md finding 21, citing Wikipedia “Ethinylestradiol” relative binding affinities and Kuhl 2005 supporting summary.
C1 Confirmed.
Synthesis §5 (“Hepatic ER-α…”) was the section most revised by the round-2 fact-checks. The original quoted a hepatic ER-α Kd of 0.25 nM (Vickers 1989, rat hepatocyte) and built a Hill saturation curve on top of it. The receptor Kd is real, but the dose-response for SHBG induction is at much higher concentrations, and the EC50 attribution to Selva & Hammond 2009 was wrong (that paper is on thyroid hormone, not E2).
← synthesis: Hepatic ER‑α and SHBG
Claim: The Vickers et al. 1989 Cancer Research paper reports a Kd of 0.25 nM for [³H]-E2 binding to rat hepatocyte nuclear/cytosolic ER using an exchange assay. This is a real number, but the assay is rat hepatocyte and the receptor Kd is not the right parameter for predicting SHBG induction.
“The Vickers et al. 1989 paper (Cancer Res 49: 6512–20, PMID 2573415) is a rat hepatocyte study in a two-stage hepatocarcinogenesis model using ethinylestradiol as promoter. It reports a Kd of 0.25 nM for [³H]-E2 binding to rat hepatocyte nuclear/cytosolic ER using an exchange assay. The value is reasonable: across tissues, ER-α Kd for E2 is typically reported as 0.05–1 nM depending on assay… 0.25 nM is in the textbook ballpark.”
—research/fact-check.md B4, citing Vickers AE, Nelson K, McCoy Z, Lucier GW. 1989, Cancer Res 49:6512–20, PMID 2573415.
C2 Number is real; cross-species applicability defensible (recombinant human ER-α Kd 0.05–0.5 nM, PMID 19022364).
The caveat: receptor occupancy is not the same as transcriptional response. SHBG mRNA induction in HepG2 (Kalme 1999) requires sustained E2 at 0.5–2.5 μM — orders of magnitude above the receptor's nominal Kd. The synthesis's earlier “Kd 0.25 nM → SHBG Hill curve” chain was modeled, not published anywhere as a single piece.
← synthesis: Hepatic ER‑α and SHBG (keynote)
Claim: The actual published HepG2 dose-response for SHBG induction by E2 shows responses at 0.5–2.5 μM — not at the receptor Kd. The synthesis's earlier attribution of an “EC50 ~1500 pg/mL (= 5.5 nM)” to Selva & Hammond 2009 was incorrect; that paper is about thyroid hormone / HNF-4α, not E2.
“Selva DM & Hammond GL 2009 (J Mol Endocrinol 43:19) is titled 'Thyroid hormones act indirectly to increase sex hormone-binding globulin production by liver via hepatocyte nuclear factor-4α.' It's about T3/T4 / palmitate / HNF-4α regulation of SHBG, not about E2-induced SHBG induction. The paper does not report an E2 EC50 for SHBG induction. The actual published HepG2 E2 dose-response (Kalme 1999, Fertil Steril 72(2):325–329, PMID 10439005; Plymate et al. 1988): SHBG production rises with E2 at 0.5–2.5 μM (500–2500 nM), about 90–450× higher than the page's claimed 5.5 nM.” [range corrected per Kalme 1999, lowest effective dose 0.5 μM]
—research/fact-check-mech-round2.md finding 5. Kalme T et al. 1999, “Estradiol increases the production of SHBG but not IGFBP-1 in cultured human hepatoma cells,” PMID 10439005; Selva & Hammond 2009, J Mol Endocrinol 43:19.
C1 The Selva & Hammond misattribution is a hard error; the Kalme et al. E2-HepG2 dose-response is the correct primary reference.
The v4 model's 1500 pg/mL EC50 is a calibration target chosen to make pregnancy SHBG match observed values, not a measured EC50. The synthesis (current version) makes this explicit; earlier drafts pinned it on Selva & Hammond. The discrepancy between HepG2 (~μM) and clinical Ropponen 2005 dose-response (effective response at μg/mL plasma) is attributed to portal-vein concentration amplification during oral absorption and to HepG2 hepatoma cells being less responsive than primary human hepatocytes.
← synthesis: Hepatic ER‑α and SHBG
Claim: Ropponen et al. 2005 reports SHBG +67–171% on oral E2 (oral E2 valerate 2–4 mg/d; dose-dependent in response to increasing dose, but the abstract publishes no per-dose breakdown) and no change on transdermal. This is the dose-response clinical anchor for the SHBG induction curve. Year is 2005 not 2003, and the first author is Ropponen, not “Lindberg” (both corrections; see note below).
“The Lindberg et al. paper is 2005 (JCEM 90: 3431), not 2003: 'Effects of Oral and Transdermal Estradiol Administration on Levels of SHBG in Postmenopausal Women.' It reports SHBG +67–171% on oral but no change on transdermal. The numbers are right; the year is one of those AI-misremembered details.”
—research/fact-check.md D3, on the JCEM 90:3431 SHBG paper (this internal note preserved the “Lindberg” author hallucination; first author corrected to Ropponen below).
C1 Range confirmed; year is 2005 not 2003 and first author is Ropponen (see note).
Author correction (hard error, Selva–Hammond class): the JCEM 90(6):3431–3434 SHBG paper is Ropponen A, Aittomäki K, Vihma V, Tikkanen MJ, Ylikorkala O. 2005 (doi:10.1210/jc.2005-0352) — verified via Crossref (May 2026). There is no “Lindberg 2005” JCEM SHBG paper; earlier drafts misremembered the first author, the same class of AI-misremembered detail as the 2003→2005 year error caught above. The companion paper on the same 40-woman cohort (Vihma et al. 2004, Ann Med 36(5):393–399, PMID 15478314; reference entry) used oral E2 valerate 2–4 mg/d (no 1 mg arm) and transdermal E2 50–100 μg/d; the abstract gives only the aggregate +67–171% range “in response to increasing doses,” so any per-dose breakdown (e.g. 1 mg→+60%) is the v4 model's predicted curve, not the paper's data. Either way, the numbers are right; route is the key variable.
← synthesis: Confidence index (changelog)
Claim: SHBG has structural/equilibrium Kd ~10–30 nM for E2, ~3–5 nM for testosterone, ~1 nM for DHT. The original synthesis quoted “~1 nM” for SHBG-E2, confusing E2 with DHT — at [SHBG] = 50 nM that DHT value predicts ~98% SHBG-bound, far above the empirical ~38%. The ~10–30 nM structural value is the molecular constant; reproducing the observed ~38% SHBG-bound fraction in a full free-hormone calculation (with albumin competing) requires the tighter operative constant (~1.5–3 nM, the Vermeulen/Wikipedia value), reflecting SHBG's two binding sites per dimer and assay convention — see the synthesis “Hepatic ER‑α and SHBG” section. The simple two-state arithmetic in the quote below ("works out to 38%") omits albumin and so is only approximate.
“SHBG Kd for E2 is ~10–30 nM, not 1 nM. Hammond and others (Avvakumov 2010; PMID 19748550) report SHBG Kd for E2 of ~10–30 nM (vs ~1 nM for DHT, ~3 nM for testosterone). The AI may have confused E2 Kd with DHT Kd. This matters: at SHBG = 50 nM, with Kd_E2 = 1 nM, ~98% of E2 would be SHBG-bound (contradicting the 38% empirical). With Kd = 10–30 nM, the binding distribution actually works out to the observed ~38% SHBG-bound at total E2 ~100 pg/mL.”
—research/fact-check.md D1, citing Avvakumov GV et al. 2010 (PMID 19748550) and Hammond reviews.
C1 Hard correction; load-bearing.
Note: Plowchalk & Teeguarden 2002 PBPK (PMID 12215661) uses SHBG-E2 Kd = 1.5 nM, which is the lower end of the published range; Wikipedia SHBG table gives K = 680 × 10⁶ M⁻¹ → Kd = 1.5 nM. There is legitimate range across assays (1.5–30 nM); the synthesis's 20 nM working value is mid-range. The hard error was conflating E2 with DHT (which is genuinely ~1 nM).
← synthesis: Hepatic ER‑α and SHBG (the two timescales)
Claim: Circulating SHBG is cleared slowly: tracer studies in primate give a biphasic curve with a fast first component (~7.5 h, >90% of label gone in 24 h) and a slow terminal component of ~3.95 days; clinical references commonly quote ~7 days. Because plasma SHBG concentration is set by the balance of hepatic secretion and this slow clearance, a step change in hepatic E2 exposure (starting/stopping/changing oral estrogen) takes about two weeks (3–4 half-lives) to reach a new SHBG steady state — the rate-limiting step is protein turnover, not receptor binding or transcription.
“the t1/2 (app) of the first component is 7.5 h (r = 0.94) … the t1/2 (app) of the second component is 3.95 days (r = 0.95)” — over 90% of labeled SBP was removed within the first 24 h.
— Namkung PC, Stanczyk FZ, Cook MJ, Novy MJ, Petra PH. 1989. Half-life of plasma sex steroid-binding protein (SBP) in the primate. J Steroid Biochem 32(5):675–680. PMID 2500563.C2 Primary primate tracer measurement; human value inferred (clinical sources cite ~7 days, consistent in order of magnitude).
The multi-day turnover is why SHBG (and the ER-driven clotting factors, which have their own protein half-lives of hours to days) lag behind the transient first-pass E2 spike: each oral dose is a minutes-long hepatic pulse, but the protein readout integrates over days. This is the kinetic basis for the clinical practice of waiting several weeks before re-checking SHBG after a dose change.
← synthesis: Hepatic ER‑α and SHBG (protein binding)
Claim: SHBG is one of the most dynamic carrier proteins in plasma. Across healthy adults it ranges ~20–100 nM; oral E2 raises it ~67–171% (Ropponen et al. 2005); and pregnancy — the largest physiological estrogen exposure — raises maternal SHBG roughly 5–10-fold over the non-pregnant level, peaking near term. The same study independently anchors the SHBG clearance half-life: maternal SHBG fell with a half-life of ~7 days postpartum, consistent with the ~4-day primate tracer terminal t½ and the ~7-day clinical figure.
“Maternal SHBG was 5-fold higher than in forty non-pregnant women, and fell with a half-life of 7.1 days immediately post-partum.”
— Anderson DC, Lasley BL, Fisher RA, Shepherd JH, Newman L, Hendrickx AG. 1976. Transplacental gradients of sex-hormone-binding globulin in human and simian pregnancy. Clin Endocrinol (Oxf) 5(6):657–669. PMID 827397. (The consensus pregnancy range is 5–10× per the Wikipedia SHBG summary and reviews; Anderson 1976's primary value is ~5×.)C2 Primary measurement of the ~5× pregnancy rise and the ~7-day postpartum half-life; the broader “5–10×” range spans later studies/assays, hence C2 rather than C1.
Most of the pregnancy SHBG rise is driven by sustained high estrogen acting on hepatic ER, not by hCG directly — consistent with the synthesis’s “route/exposure, not systemic level” framing: pregnancy delivers a large, sustained hepatic estrogen signal. The postpartum decline half-life (~7 days) is a clean human cross-check on the SHBG turnover constant inferred from the Namkung 1989 primate tracer study (see above), and it sets the ~2-week timescale for SHBG to reach a new steady state after an estrogen dose change.
← synthesis: VTE risk by route (route matters more than dose)
Claim: Oral E2 produces ~50 pg/mL systemic plasma E2 but the liver during first-pass sees portal concentrations of ~5,000 pg/mL (100× higher) — saturating hepatic ER and producing pregnancy-level SHBG induction at 1/500 the systemic levels.
“For hepatic estrogenic effect (SHBG induction etc.), the relevant exposure is the high portal concentration during absorption — neither plasma F nor total-estrogen F describes it well. The clinical effect of 1 mg oral on SHBG is comparable to ~100 μg/d transdermal, which is roughly the 'hepatic equivalence' of 1 mg PO.”
—research/fact-check.md B1, citing Kuhl 2005, Climacteric 8 Suppl 1, PMID 16112947; transfemscience.org “Approximate Comparable Dosages of Estradiol by Different Routes.”
C2 The clinical equivalence (1 mg PO ≈ 100 μg/d TD for hepatic effects) is well-established; the exact portal concentration (~5,000 pg/mL) is model-derived, not directly measured (no human portal-vein E2 study exists).
← synthesis: Hepatic ER‑α and SHBG
Claim: The Hill coefficient ~1 used in the synthesis's saturation curves is a modeling default for monomeric ligand-receptor binding without cooperativity, not an experimentally measured in-vivo value for hepatic ER-α → SHBG.
“There is no single experimental measurement of the Hill coefficient for hepatic ER-α → SHBG. A Hill coefficient near 1 is the canonical assumption for monomeric ligand binding to a receptor without cooperativity. ER-α functions as a dimer but the ligand-binding step is monomeric per subunit; in reporter assays, n is usually 1.0–1.5 (Anstead et al. 1997). Using ~1.0–1.2 is fine as a default, but flag it as a modeling assumption, not an experimentally validated parameter.”
—research/fact-check.md B5.
C3 Defensible default; flagged as modeling assumption.
Synthesis §6 is anchored to the ESTHER study (Canonico 2007), with corroboration from Scarabin 2015 meta-analysis, Vinogradova 2019, and Stegeman/ASH 2024 absolute-rate tables.
← synthesis: VTE risk by route (keynote)
Claim: The ESTHER study (Canonico 2007) reports adjusted OR for VTE of 4.2 for current oral estrogen users vs nonusers, and 0.9 for transdermal users. This is the canonical “route matters” result. The original synthesis's “~2×” was from pooled meta-analyses, not from ESTHER itself.
“odds ratios (ORs) for VTE in current users of oral and transdermal estrogen compared with nonusers were 4.2 (95% CI, 1.5 to 11.6) and 0.9 (95% CI, 0.4 to 2.1), respectively”
— Canonico M et al. 2007, “Hormone therapy and venous thromboembolism among postmenopausal women: impact of the route of estrogen administration and progestogens: the ESTHER study,” Circulation 115:840, PMID 17309934.C1 Direct quote from the ESTHER abstract.
Round-1 correction (both fact-checkers): the original synthesis's “~2×” figure for oral HRT is a pooled meta-analysis estimate that dilutes the ESTHER signal. The 4.2 vs 0.9 contrast is what underlies the population-level recommendations against oral HRT.
← synthesis: VTE risk by route (VTE table)
Claim: Pooled meta-analyses give lower (diluted) estimates than ESTHER: oral RR ~1.48–1.7; transdermal RR ~0.97. The route signal is preserved.
“Pooled meta-analyses (Scarabin/Mohammed 2015 JCEM, PMID 26544651) give RR = 1.48 (1.39–1.58) for oral estrogen among broader populations… Scarabin/Mohammed 2015 meta-analysis: RR 0.97 (0.87–1.09) for transdermal vs no HT.”
—research/fact-check.md D7–D8, citing Mohammed K et al. 2015, JCEM 100:4012, PMID 26544651.
C1 Confirmed.
← synthesis: Ethinyl estradiol (Stegeman SHBG table)
Claim: SHBG rises with COCs vary by progestin: LNG +80–150%, desogestrel +200%, drospirenone +200–300%, dienogest +320%, cyproterone +400%; older >50 μg EE COCs +500–1000%. The pattern tracks VTE risk by formulation.
“EE 20 microg + LNG 100 microg (Alesse) | +80% to +150% (Stegeman 2013) … EE 30 microg + drospirenone 3 mg (Yasmin) | +200% to +300% … EE 35 microg + cyproterone 2 mg (Diane-35) | +400% … EE >50 microg (older COCs) | +500% to +1000% (5–10 fold)”
—research/EE-parameters.md table, citing Stegeman BH et al. 2013, “Effect of ethinylestradiol dose and progestagen in combined oral contraceptives on plasma SHBG levels,” J Thromb Haemost 11(1):203–205; corroborated by Kuhl 2005 and Wikipedia “Ethinylestradiol” summary.
C1 Multi-source convergent.
← synthesis: Ethinyl estradiol (VTE per-10k-woman-year list)
Claim: Per 10,000 woman-years: no exposure 1–5 · oral E2 HRT 9–15 · transdermal E2 5 · EE+LNG (Microgynon) 5–7 · EE+desogestrel/gestodene 9–12 · EE+drospirenone (Yasmin) 10–12 · E2V+dienogest (Qlaira) 4–6 · old high-dose EE COCs 15–20.
“no exposure ~ 1–5 · oral E2 HRT (menopause) ~ 9–15 (RR ~1.5–2 vs none) · transdermal E2 HRT ~ 5 (RR ~1, ~ baseline) · COC: EE + LNG (Microgynon) ~ 5–7 (lowest among EE COCs; RR ~2.3) · COC: EE + drospirenone (Yasmin) ~ 10–12 (RR ~4) · COC: E2-valerate + dienogest (Qlaira) ~ 4–6 (closer to natural-E2 HRT) · old high-dose EE (>50 microg) COC ~ 15–20+ (historical)”
—research/EE-parameters.md compiled from Stegeman BH et al. 2013, J Thromb Haemost; ASH 2024 Hematology Education Program (“Estrogen, progestin, and beyond: thrombotic risk and contraceptive choices”).
C1 Multi-source convergent; the figures themselves are population-average estimates.
← synthesis: VTE risk by route
Claim: VTE risk relative to nonpregnant baseline: first/second trimester ~2–4×, third trimester OR 8.8 (4.5–17.3), first 6 weeks postpartum 20–80× depending on study.
“Modern meta-analyses (e.g., Sultan 2013 BMJ; PMC3726432): Overall pregnancy: ~5× vs non-pregnant baseline. First/second trimester: ~2–4×. Third trimester: 9× (OR 8.8, 95% CI 4.5–17.3). First 6 weeks postpartum: 20–84× depending on study. The Heit 2005 study and Pomp 2008 MEGA study report OR ~84 for the first 6 weeks.”
—research/fact-check.md D6, citing Sultan AA et al. 2013, BMJ, PMC3726432; CDC “Pregnancy and blood clots” overview.
C1 Confirmed for the OR magnitudes; the upper postpartum bound varies by study (Heit/Pomp report OR ~84).
← synthesis: VTE risk by route (cascade caption)
Claim: In pregnancy, free protein S decreases to ~38% of nonpregnant levels (procoagulant). TFPI is reported as lower in late pregnancy (~7.1 ng/mL) than in nonpregnant controls (~8.2 ng/mL). There is no net anticoagulant compensation in pregnancy on the protein S arm. The synthesis's original cascade arrow “÷2 TFPI ↑” was directionally wrong.
“In pregnancy, the anticoagulant arm is not uniformly compensatory: Free protein S falls to ~38% of nonpregnant levels (Comp 1986, Blood 68:881, PMID 2944555). The decrease starts in the first trimester and persists through gestation. This is procoagulant, not compensatory. TFPI in pregnancy is reported as lower than nonpregnant controls (~7.1 ng/mL late pregnancy vs ~8.2 ng/mL nonpregnant in a typical reference series).”
—research/fact-check-mech-round2.md finding 2, citing Comp PC et al. 1986, “Functional and immunologic protein S levels are decreased during pregnancy,” Blood 68:881, PMID 2944555; and PMID 20978710 for TFPI levels in pregnancy.
C2 Direction of protein S confirmed (decreases); the quantitative ÷2 factor in the original cascade is not defensible either way.
← synthesis: VTE risk by route (cascade)
Claim: In pregnancy, total E2 rises ~100–250×; SHBG rises 5–10×; free fraction drops to ~60% of nonpregnant (from ~2% → ~1.2%). So free E2 actually rises ~0.6 × 250 ≈ 150×, attenuated ~1.7× by SHBG buffering, not 3×.
“In pregnancy, total E2 rises ~100–250×. SHBG rises 5–10×. By mass-action, the free fraction drops. Transfemscience review (citing primary literature): 'the percentage of estradiol that is free appears to be decreased only to around 60% of that of non-pregnancy.' So free E2 rises ~0.6 × 250 ≈ 150× — i.e., the SHBG buffering attenuates by ×1.7, not ×3.”
—research/fact-check-mech-round2.md finding 4, citing transfemscience.org “Interactions of Sex Hormones with SHBG and Relevance for Transfeminine Hormone Therapy” and Wikipedia “Sex hormone-binding globulin.”
C2 Reasonable arithmetic from established total-E2 / SHBG rises.
Synthesis §2 (keynote on E1S) and §3 (oral apparent half-life) lean on E1S being the dominant circulating estrogen and the rate-limiting reservoir for clearance.
← synthesis: Pharmacokinetics by route
Claim: Plasma E1S half-life is 10–12 h — this is what drives the 13–20 h apparent half-life of oral E2 (the E1S reservoir is the rate-limiting step for clearance, not free-E2 metabolism).
“The elimination half-life of E1S is 10 to 12 hours.”
— Wikipedia, Estrone sulfate. Primary: Ruder HJ et al. 1972, “Estrone sulfate: production rate and metabolism in man,” J Clin Invest, PMC302214.C1 Confirmed in multiple sources.
← synthesis: Molecule and metabolism (keynote)
Claim: E1S levels are 10–15× higher than free E1 in women under physiological conditions. E1S is the most abundant circulating estrogen species across essentially every hormonal state.
“E1S levels are about 10 to 15 times higher than those of estrone in women.”
— Wikipedia, Estrone sulfate.C1 Confirmed.
The synthesis's older claim that E1S is “10–25× free E1+E2” is slightly stronger; some review compilations cite 5–10× unconjugated estrogens. The qualitative point (E1S as dominant circulating estrogen, slow-release reservoir) is robust.
← synthesis: Confidence index (changelog)
Claim: Third-trimester pregnancy E1S is ~100,000 pg/mL, not the ~50,000 pg/mL the original synthesis quoted (~2× correction).
“E1S levels in pregnant women were 19 ± 5 ng/mL in the first trimester, 66 ± 31 ng/mL in the second trimester, and 105 ± 22 ng/mL in the third trimester.”
— Wikipedia, Estrone sulfate, pregnancy table.C1 Direct compilation; round-1 fact-check correction (C5).
← synthesis: Confidence index (changelog)
Claim: Postmenopausal E1 production rate is ~40 μg/d (Longcope 1986 / Grodin 1973), not the ~80 μg/d the original synthesis quoted. Obese women can reach ~80–100 μg/d via increased adipose aromatization.
“Wikipedia 'Estrone sulfate' table gives men 80 μg/d, premenopausal follicular 100 μg/d, luteal 180 μg/d, postmenopausal ~40 μg/d (Longcope 1986; Grodin 1973). So the AI's '80 μg/day for PMP' is likely 2× too high — the true number is around 40 μg/d (or higher in obese women).”
—research/fact-check.md C8.
C1 Confirmed; round-1 correction.
← synthesis: Pharmacokinetics by route (routes table)
Claim: Steady-state E1S anchors: cycling follicular ~960 pg/mL, postmenopausal ~130 pg/mL, on unspecified HT ~2,560 pg/mL, third-trimester pregnancy ~105,000 pg/mL.
“Cycling follicular 960 pg/mL: Wikipedia E1S = '0.96 ± 0.17 ng/mL' follicular. ✓ Postmenopausal 130 pg/mL: Wikipedia = '0.13 ± 0.03 ng/mL'. ✓ Oral HRT 2560 pg/mL: Wikipedia = '2.56 ± 0.47 ng/mL on unspecified menopausal hormone therapy' — matches but technically 'on HT' not specifically 'oral HT'. Minor caveat. Pregnancy term 100,000 pg/mL: Wikipedia = '105 ± 22 ng/mL'.”
—research/fact-check-pk-round2.md “What looked correct” section, cross-checked against Wikipedia “Estrone sulfate” population concentration tables.
C1 Multi-source convergent.
Synthesis §11 describes EC508 (estradiol-17β-(aryl-sulfamoyl-benzoyl-prolinate)) and the related E2MATE (estradiol-3-sulfamate) as approaches to RBC sequestration via CA-II binding, bypassing hepatic first-pass.
Claim: E2MATE/EMATE is a mechanism-based irreversible inhibitor of steroid sulfatase. The C3-O-sulfamate transfers the sulfamoyl group to the active-site formylglycine (FGly) residue, covalently inactivating STS. This is distinct from transition-state mimicry. The original synthesis called it a “transition-state mimic” — that was wrong on the chemistry.
“Aryl sulfamate esters like E2MATE are 'first-in-class highly potent active site-directed irreversible STS inhibitors' and… compounds of this class are thought to irreversibly modify the active site formylglycine residue of STS.”
— Wikipedia “Estradiol sulfamate.” Mechanism review: “SULFATION PATHWAYS: Steroid sulphatase inhibition via aryl sulphamates” J Mol Endocrinol 2018 — describes “transfer of the sulfamoyl group (or as sulfonylamine) to a hydrated or unhydrated STS active site formylglycine residue, and this leads to inactivation of the active site machinery.”C1 Confirmed; round-2 mechanism correction.
Why it matters: E2MATE was originally intended as an oral E2 prodrug (RBC-sequestered, then hydrolyzed by STS to release E2). The mechanism-based STS inactivation killed both arms — it blocked its own activation AND destroyed the body's E1S reactivation pathway. E2MATE was repurposed as an STS inhibitor for endometriosis. The redesign (EC508) moves the sulfamoyl off C3 onto a separate handle at C17 via an esterase-cleavable linker, leaving STS intact.
Claim: EC508 is estradiol-17β-(1-(4-(aminosulfonyl)benzoyl)-L-proline). Architecture: estradiol-17β-O-C(=O)-prolyl-N-C(=O)-aryl-SO₂NH₂. Plasma esterases cleave the C17 ester to release estradiol + the prolyl-aryl-sulfonamide fragment. The sulfonamide head binds carbonic anhydrase II in red blood cells for plasma sequestration.
“The C17 ester is the cleavable bond (correct), but the architecture is: estradiol-17β-O-C(=O)-prolyl-N-C(=O)-aryl-SO₂NH₂. Plasma esterases would cleave the C17 ester to release estradiol + the prolyl-aryl-sulfonamide fragment. The page describes this correctly.”
—research/fact-check-mech-round2.md finding 12. Sulfamate-Zn-CA-II binding is textbook (sulfonamides are classical CA-II inhibitors via deprotonated SO₂NH⁻ coordinating catalytic Zn²⁺).
C1 Confirmed mechanistically.
Claim: EC508 was developed by Evestra Pharmaceuticals. As of 2021, transfemscience.org reports “No Further Development” — it never progressed to human clinical trials.
“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.”
— Synthesis §11 narrative, citing transfemscience.org EC508 article.C2 Reported in the canonical review; primary developer source not located.
Synthesis §10 / §12 note that real PK has CV ~30–60% on most parameters; this is bigger than the model's error bars. The Goldzieher 1989 figure (above, §4) shows it directly.
← synthesis: Pharmacokinetics by route
Claim: Oral E2 has ~4.6× inter-individual variability in AUC (28–127% of the mean across study participants). Stanczyk reviews echo this.
“28 to 127% mean AUC variability across individuals, i.e. ~4.6-fold”
—research/fact-check.md B1, citing Wikipedia/Kuhl 2005, Climacteric.
C1 Confirmed.
← synthesis: Pharmacokinetics by route (Goldzieher figure)
Claim: Single-dose 70 μg EE in 24 women (Brody/Turkes/Goldzieher 1989) shows highest-responder Cmax ~300 pg/mL vs lowest-responder ~60 pg/mL — a 5× range, same drug, same dose. This is the visible primary evidence for population variability.
“Single-dose 2-tablet (= 70 μg EE) administration to n=24 women. Mean Cmax ~160 pg/mL at 1 h; note the highest-responder peak at ~300 pg/mL — 5× the lowest responder.”
— Synthesis §3 figure caption (reproducing Goldzieher 1990 Fig 1), originally Brody SA, Turkes A, Goldzieher JW 1989, Contraception 40:269–84.C1 Directly visible in the original figure.
← synthesis: Ethinyl estradiol
Claim: SULT1E1 has Km ~4 nM for EE and is a major contributor to gut/hepatic first-pass conjugation; this is one reason EE bioavailability is bimodal (high vs low SULT1E1 expressors).
“SULT1E1 has Km ~4 nM for EE and is a major contributor to gut/hepatic first-pass conjugation; this is one reason EE bioavailability is bimodal (high vs low SULT1E1 expressors). EE is also a potent inhibitor of SULT1A1, which modulates its own clearance and that of other SULT substrates.”
—research/EE-parameters.md, citing Schrag ML et al. 2004, “Sulfotransferase 1E1 is a low-Km isoform mediating the 3-O-sulfation of ethinyl estradiol,” Drug Metab Dispos; Rodrigues AD et al. 2022, “Drug interactions involving 17alpha-ethinylestradiol: considerations beyond CYP3A induction and inhibition,” Clin Pharmacol Ther 112(1):69.
C2 Mechanism confirmed; the explicit “bimodal F” framing is a Rodrigues 2022 interpretation.
Synthesis §2 keynote frames estrogen pharmacology as two parallel directional pumps: SULT1E1/STS (C3 sulfation/desulfation) and HSD17B2/HSD17B1 (C17 oxidation/reduction). Liver runs the inactivating arms; target tissues run the reactivating arms.
← synthesis: Molecule and metabolism (keynote)
Claim: The directional pump is implemented as tissue-specific enzyme expression. Hepatocyte: HSD17B2 (NAD⁺, oxidative) + SULT1E1 (low-Km estrogen sulfation) + UGTs. Breast/endometrium/brain: HSD17B1 (NADPH, reductive) + STS (sulfate hydrolysis). Same enzyme families, mirror-image regulation.
“SULT1E1 + STS form one such pump (sulfation/desulfation at C3). HSD17B2 + HSD17B1 form another (oxidation/reduction at C17). The liver 'quietens' the hormone by both arms; target tissues 'reawaken' it by both arms. Circulating E1-sulfate is the inactive currency that traffics between them — the most abundant circulating estrogen species in essentially every hormonal state (men, premenopausal women, postmenopausal women, people on HRT, pregnancy: E1S is always 3–15× the free E1+E2 pool).”
— Synthesis §2 keynote; supported by all of A1–A7 inresearch/fact-check.md.
C1 Confirmed at the level of individual enzyme expressions and Km values; the integrated “pump” framing is the synthesis's pedagogical scaffolding (not a literal quote from a single paper).
These are the hard errors caught in the round-2 fact-check passes. Each one was wrong in the older synthesis and is corrected in the current narrative. The verbatim findings are quoted below so a fact-checker can spot-verify them quickly.
← synthesis: Hepatic ER‑α and SHBG (keynote)
Correction: Older drafts cited Selva & Hammond 2009 for an “EC50 ~1500 pg/mL for SHBG induction in HepG2 cells.” That paper is about thyroid hormone via HNF-4α, not estradiol dose-response. The correct primary reference for E2-HepG2 SHBG induction is Kalme 1999, which shows responses at 0.5–2.5 μM E2 — orders of magnitude higher than the receptor Kd.
“Selva DM & Hammond GL 2009 (J Mol Endocrinol 43:19) is titled 'Thyroid hormones act indirectly to increase sex hormone-binding globulin production by liver via hepatocyte nuclear factor-4α.' It's about T3/T4 / palmitate / HNF-4α regulation of SHBG, not about E2-induced SHBG induction. The paper does not report an E2 EC50 for SHBG induction.”
—research/fact-check-mech-round2.md finding 5, citing Selva DM, Hammond GL. 2009. J Mol Endocrinol 43:19–27; Kalme T et al. 1999. Fertil Steril 72(2):325–329. PMID 10439005.
C1 Hard correction; Selva & Hammond paper title and content directly contradict the prior attribution.
← synthesis: Ethinyl estradiol
Correction: Older drafts said EE has “no C17-OH” (in the gut-conjugation diagram). EE is 17α-ethinyl-17β-estradiol — the 17β-OH is preserved. The 17α-ethinyl substitution removes the C17 carbinol hydrogen needed for HSD17B2 oxidation; it does not remove the OH.
“Ethinyl estradiol is 17α-ethinyl-17β-estradiol. The 17α-ethinyl substituent blocks 17β-HSD oxidation of the 17β-hydroxy group, but it does not remove the 17β-OH. The statement 'no C17-OH' is a hard structure error.”
—research/fact-check-codex-round2.md finding 1. Confirmed against PubChem: “Ethinylestradiol can be glucuronidated by UGT1A1, UGT1A3, UGT1A4, UGT1A9, and UGT2B7” — UGT2B7 is the C17 glucuronosyl transferase; if C17-OH were absent it couldn't act.
C1 Hard structural correction.
← synthesis: VTE risk by route (cascade)
Correction: Older drafts showed “÷2 TFPI ↑ + endothelial NO ↑” as a compensatory step in the pregnancy VTE cascade. In fact: free protein S falls to ~38% of nonpregnant; TFPI in late pregnancy (~7.1 ng/mL) is lower than nonpregnant (~8.2). There is no net anticoagulant compensation in pregnancy on the protein S arm.
“In pregnancy, the anticoagulant arm is not uniformly compensatory: Free protein S falls to ~38% of nonpregnant levels (Comp 1986, Blood 68:881, PMID 2944555)… TFPI in pregnancy is reported as lower than nonpregnant controls (~7.1 ng/mL late pregnancy vs ~8.2 ng/mL nonpregnant in a typical reference series).”
—research/fact-check-mech-round2.md finding 2, citing Comp PC et al. 1986, PMID 2944555; PMID 20978710.
C2 Direction of protein S confirmed; quantitative cascade factors remain illustrative.
Correction: Older drafts called E2MATE a “transition-state mimic that irreversibly inhibited STS.” The mechanism is mechanism-based irreversible covalent modification: sulfamoyl transfer from the C3 phenolic-O-SO₂NH₂ to the active-site formylglycine residue covalently inactivates STS.
“The actual STS-inhibition mechanism of EMATE/E2MATE is mechanism-based irreversible covalent modification of the active-site formylglycine residue… Sulfamoyl transfer from the aryl-O-SO₂NH₂ to FGly produces a stable N-sulfated imine adduct that permanently inactivates the enzyme. This is mechanistically distinct from transition-state mimicry — a TSI binds the active site shaped like the bound substrate at the rate-limiting TS, but doesn't covalently modify the enzyme. EMATE is a suicide / mechanism-based inhibitor, not a TSI.”
—research/fact-check-mech-round2.md finding 1, citing “SULFATION PATHWAYS: Steroid sulphatase inhibition via aryl sulphamates” J Mol Endocrinol 2018; Wikipedia “Estradiol sulfamate.”
C1 Mechanism class is hard-wrong in the older draft.
← synthesis: Pharmacokinetics by route (gut conjugates)
Correction: Older drafts said “gut SULT1A1 makes the C3-sulfation contribution.” SULT1A1 is more abundant in gut (~19% of intestinal SULTs vs SULT1E1's ~8%) but its Km for E2 is ~2.4 μM, vs SULT1E1's ~5–27 nM. At physiological [E2] in the enterocyte during oral first-pass (low nM to low μM at peak), SULT1E1 operates at saturation while SULT1A1 is largely idle — so SULT1E1 likely dominates the gut sulfation flux despite lower abundance.
“Gut SULT abundance: SULT1A1 ≈ 19%, SULT1E1 ≈ 8% of intestinal SULTs. But SULT1A1 Km for E2 is ~2.4 μM while SULT1E1 Km is ~5–27 nM. At physiologic [E2] (low nM to low μM during oral first-pass), SULT1E1 likely dominates gut E2-3-sulfation despite being less abundant, because SULT1A1 is mostly idle at sub-μM substrate. The page has it backwards qualitatively.”
—research/fact-check-mech-round2.md finding 7.
C2 Mechanism-grounded reasoning from established Km values.
← synthesis: Pharmacokinetics by route (citation list)
Correction: The original draft's citation paragraph for E2 gut data had: (a) “Cheng et al. 1998” — should be Cheng, Radominska-Pandya, Tephly 1999; and (b) attributed the “UGT1A10 10× UGT1A1” quantitative claim to Strassburg 1998, when it actually comes from Basu et al. 2004 (Strassburg established expression patterns; Basu measured the kinetic ratio).
“Strassburg 1998 (JBC) characterized the gut UGT1A8/1A10 expression pattern but does NOT establish the 'UGT1A10 has highest E2-3-G rate' claim — that comes from Basu et al. 2004 (PMID 15117964, J Biol Chem 279:28320), which explicitly reports 'UGT1A10 in cells supported 10-fold higher glucuronidation of 17β-estradiol than UGT1A1.' The '~10× UGT1A10 vs UGT1A1' the page repeats in §2 and §3 traces to Basu 2004, not Strassburg 1998. Cheng et al. is 1999, not 1998 (Cheng Z, Radominska-Pandya A, Tephly TR, 'Studies on the substrate specificity of human intestinal UDP-glucuronosyltransferases 1A8 and 1A10,' Arch Biochem Biophys 1999, PMID 10497143).”
—research/fact-check-mech-round2.md finding 6.
C1 Citation correction.
← synthesis: Ethinyl estradiol (EE mass balance)
Correction: Older drafts of the EE mass balance diagram showed “~20% hepatic extraction” for EE. The Back & Rogers 1982 paper that the synthesis cites for the 0.44 gut-wall figure reports hepatic extraction = 0.25 in the same paper. Updated to 25%.
“The fraction of conjugated ethinylestradiol in the gut wall was 0.44, while the conjugated fraction in the liver was only 0.25. Calculations showed that the gut wall appeared to be twice as effective as the liver in conjugating ethinyloestradiol on the first pass.”
— Back DJ et al. 1982, PMID 7059434, PMC1402099. Caught inresearch/fact-check-pk-round2.md finding 2.
C1 Direct correction from the same primary source.
← synthesis: Pharmacokinetics by route (EE time-course curve)
Correction: Older drafts plotted a Cmax of ~95 pg/mL for 30 μg EE and labeled it “single dose.” 95 pg/mL is the Yasmin-label steady-state Cmax. True single-dose Cmax for 30 μg EE is ~50–70 pg/mL (Alesse label scaled). The synthesis now uses ~70 pg/mL for single-dose with explicit steady-state note.
“Following a single dose, maximum serum concentrations of ethinyl estradiol of 62 ± 21 pg/mL are reached at 1.5 ± 0.5 hours (note this was for a 20 mcg formulation)… Linear scaling 30/20 × 62 = 93 pg/mL for steady-state, ≈ the 95 in the synthesis. So 95 is a steady-state number, applied to a single-dose plot.”
—research/fact-check-pk-round2.md finding 1, citing Wikipedia Ethinylestradiol PK summary and Alesse label.
C1 Confirmed; correction was made in the current synthesis.
← synthesis: Confidence index (changelog)
Correction: Original synthesis quoted sublingual Tmax as ~30 min. Modern LC-MS/MS (Doll 2022) measures Tmax = 1 h. The 30-min figure was likely from older RIA studies over-reading very low E2 or from first detectable rise rather than peak.
“Cirrincione et al. 2021 (Endocrine Practice) and the transfemscience review: 1 mg sublingual → Cmax ≈ 144 pg/mL, vs 35 pg/mL for 1 mg oral — same study (so ~4× ratio). Tmax: 1–2 hours, not 30 min. '30 min' might be from a different formulation or might be the first detectable rise; the actual peak is at ~1–2 h.”
—research/fact-check.md C6.
C1 Confirmed by direct LC-MS/MS measurement (Doll 2022).
← synthesis: VTE risk by route (coagulation table)
Correction: Older drafts said “Factor VII rises up to 10× in pregnancy.” Actual: Factor VII rises to ~150–250% of nonpregnant baseline (i.e., 1.5–2.5× increase). The largest pregnancy factor rise is Factor VIII (200–500%, i.e., 2–5×); vWF rises similarly. The original was likely confusing factors.
“Factor VII rises to ~150–250% of non-pregnant baseline at term (i.e., 1.5–2.5× increase), not 10×. The largest pregnancy factor rise is Factor VIII, which can reach 200–500% (2–5×). vWF rises similarly. 'Factor VII 10×' is wrong — the AI may be confusing fold-change in different factors.”
—research/fact-check.md D5, citing Dalaker 1986 BJOG; PMC7273490.
C1 Confirmed.
← synthesis: VTE risk by route (cascade)
Correction: Older drafts had a final “×3 Virchow's triad” multiplier in the pregnancy VTE cascade to land at ~3× baseline. No primary source quantifies stasis/endothelial injury at ×3 for pregnancy. The pregnancy VTE incidence is an empirical number from cohort studies, not a derivable chain of mechanism factors.
“Virchow's triad is a qualitative pathophysiological framework (hypercoagulability, venous stasis, endothelial injury). No primary source quantifies pregnancy-specific stasis × endothelial injury at '×3'. The final 3–10× VTE incidence in pregnancy is an empirical number from cohort studies (Heit 2005, Pomp 2008, Sultan 2013), not a chain of multipliers derived from biological mechanisms.”
—research/fact-check-mech-round2.md finding 3.
C2 Reframed in current synthesis as illustrative, not stepwise sourced.
← synthesis: Confidence index (changelog)
Corrections rolled in:
| Claim | Old value | Corrected value |
|---|---|---|
| SHBG-E2 Kd | 1 nM (was confused with DHT) | 10–30 nM |
| JCEM 90:3431 SHBG paper year | 2003 | 2005 |
| JCEM 90:3431 SHBG paper first author | “Lindberg” (AI-misremembered) | Ropponen et al. 2005 (verified via Crossref, doi:10.1210/jc.2005-0352) |
| Postmenopausal E1 production | ~80 μg/d | ~40 μg/d |
| Pregnancy E1S | ~50,000 pg/mL | ~100,000 pg/mL (105 ± 22 ng/mL) |
| Cycling “free E2 = 50 pg/mL” | free E2 = 50 | total = 50; free is ~1 pg/mL (2% free fraction) |
| Sublingual Tmax | ~30 min | 1–2 h (Doll 2022 LC-MS/MS) |
| Oral E2 “F = 5%” | F = 5% (unspecified) | F = 5% free-E2 AUC; ~30% total estrogen; effectively high for hepatic |
| ESTHER oral E2 OR | ~2× | 4.2× (and TD 0.9×) |
| Cytoplasmic [E1]/[E2] | 11:1 exact | ~5–20 range; arithmetic OK but inputs shaky |
C1 to C2 depending on the line — details in research/fact-check.md and fact-check-codex.md.