Clinical overview
The objective is to appraise, not "describe." With the physiology and the principal indications taken from the Intermediate chapter on progesterone in pregnancy, the harder task is to weigh the evidence behind a hormone that is cheap, ubiquitous, perceived as harmless and therefore reflexively prescribed. The consultant discipline is to separate the few subgroups in which progesterone is genuinely disease-modifying from the larger penumbra of unproven or negative uses, to read the pivotal trials critically (subgroup signals, intention-to-treat dilution, relative versus absolute effect), and to translate that into resource-conscious practice under the National Integrated Maternal and Perinatal Care (NIMPC) Guidelines (NDoH, 2024).
The easily-forgotten half of the objective is the gynaecological and antagonist side: progesterone as luteal-support agent, the distinction between bio-identical progesterone and the synthetic progestogens of contraception, and the clinical use of anti-progesterone action — mifepristone for medical evacuation and the cautionary tale of selective progesterone-receptor modulators (SPRMs) in fibroid therapy. The thread running through all of it is the same consultant discipline: which molecule, which route, which woman, and what does the best trial actually show?
Core knowledge
This section assumes the receptor biology and the luteoplacental physiology taught at Intermediate (progesterone in pregnancy) and builds the advanced layer — the mechanisms that justify a prescribing decision.
Mechanism worth re-stating at Final level
Progesterone acts through nuclear receptors (PR-A and PR-B isoforms) and faster membrane-associated receptors. The crucial mechanistic nuance is functional progesterone withdrawal: in humans, term and preterm labour are triggered not by a fall in serum progesterone (unlike many species) but by a local shift in the PR-A:PR-B ratio and coregulators that renders the myometrium progesterone-resistant despite high circulating levels. This is why serum progesterone levels do not guide supplementation — there is no validated replacement threshold — and why the preterm-birth hypothesis is one of reinforcing an endogenous brake, not correcting a deficiency. Vaginal micronised progesterone exploits a uterine first-pass effect, giving high myometrial concentrations at low systemic levels; that pharmacology, not a serum target, underlies the route choice.
The luteoplacental shift (corpus luteum until ~7–9 weeks, then placental syncytiotrophoblast) is assumed. Its corollary: exogenous progesterone has its strongest physiological rationale before the shift (luteal-phase support after ART, where stimulation/down-regulation impairs the corpus luteum) and a weaker one after — one reason most obstetric regimens stop around 16 weeks.
The pharmacology that decides route and molecule
"Progesterone" is a trap word because the clinical objects behind it are pharmacologically distinct, and the trial evidence is preparation-specific — you cannot read a result for one and prescribe another.
- Micronised progesterone is bio-identical (the same molecule the ovary and placenta make). Given orally it suffers heavy hepatic first-pass metabolism to 5β-pregnanolone and allopregnanolone — neurosteroids that cause the somnolence and dizziness women describe, and which make oral dosing pharmacokinetically unreliable for a uterine target. Given vaginally it bypasses the liver and concentrates in the uterus by a local "first-uterine-pass" effect, achieving high endometrial/myometrial tissue levels at modest serum concentrations. This dissociation of serum from tissue level is precisely why a serum progesterone reading is uninterpretable on vaginal therapy and why the obstetric trials (PRISM, PREGNANT, OPPTIMUM) almost all used the vaginal route.
- Dydrogesterone is a retroprogesterone — orally active, selective for the progesterone receptor, with negligible androgenic, oestrogenic or glucocorticoid activity, and it does not cross-react with the common immunoassay for progesterone. It is the molecule that makes an oral regimen pharmacologically credible (LOTUS programme below).
- 17-hydroxyprogesterone caproate (17-OHPC) is a synthetic caproate ester of a hydroxylated progesterone, formulated for slow intramuscular release — a fundamentally different molecule from micronised progesterone, which is one reason the vaginal and intramuscular evidence streams diverge and must never be pooled in your reasoning.
- The contraceptive progestins (levonorgestrel, etonogestrel, medroxyprogesterone acetate, drospirenone) are distinct synthetic molecules with their own receptor cross-reactivities (androgenic, anti-mineralocorticoid, glucocorticoid). Their safety datasets — bone-mineral density on depot medroxyprogesterone, VTE class effects — do not transfer to obstetric micronised progesterone, a confusion that surfaces constantly in counselling.
The consultant link to make explicit: route changes the evidence base, molecule changes the safety profile — these must be kept separate rather than treating "progesterone" as one thing.
Assessment — appraising the pivotal evidence
The question is not "what is progesterone for" but "how good is the evidence, and where does the headline result mislead?" Each subtype of use below has a different evidentiary footing, and the judgement call differs accordingly.
Early-pregnancy bleeding and recurrent loss: PRISM and PROMISE, read properly
PROMISE (first-trimester vaginal progesterone in unexplained recurrent miscarriage) was negative for live birth: progesterone is not a treatment for recurrent loss per se. The subtlety lives in PRISM (NEJM 2019; n = 4,153 with early-pregnancy bleeding). Overall live birth was 75% with progesterone vs 72% with placebo — relative rate 1.03 (95% CI 1.00–1.07), p = 0.08, not significant — so no benefit in unselected threatened miscarriage. But a pre-specified subgroup graded by prior losses became meaningful at three or more previous miscarriages: 72% vs 57%, RR 1.28 (95% CI 1.08–1.51) — an absolute live-birth difference of ~15 percentage points.
The discipline in appraising this is to read the whole gradient, not just the positive cell. PRISM's pre-specified subgroups by prior-loss count form a clean dose–response:
- No previous miscarriage: 74% vs 75%, RR 0.99 (95% CI 0.95–1.04) — no effect, and the point estimate is on the wrong side of unity.
- One or two previous miscarriages: 76% vs 72%, RR 1.05 (95% CI 1.00–1.12) — a small effect whose CI just touches 1.
- Three or more previous miscarriages: 72% vs 57%, RR 1.28 (95% CI 1.08–1.51) — a large effect.
- Test for interaction across the three strata: p = 0.007 — a statistically significant gradient, not noise.
A positive subgroup nested in a negative trial is normally treated with suspicion (multiplicity, post-hoc dredging, regression to the mean). PRISM survives that scrutiny for three concrete reasons: the subgroup was pre-specified; the effect is biologically gradated (monotonic across prior-loss count, with a significant interaction test — the strongest single defence against a chance subgroup); and it is externally coherent — the ESHRE recurrent-pregnancy-loss guideline (update 2022, published 2023) re-analysed pooled PRISM/PROMISE data and adopted the same bleeding-plus-prior-loss signal. Compute the absolute effect to keep the magnitude honest: at ≥3 losses the ARR is 72% − 57% = 15 percentage points, giving an NNT of 1/0.15 ≈ 7 women treated for one extra live birth — a genuinely large effect for a cheap, safe drug. In the unselected trial the ARR is ~3 points (NNT ≈ 33), which is why the overall result is unconvincing.
There is a health-economic layer too. The PRISM cost-effectiveness analysis found an incremental cost-effectiveness ratio of £3,305 per additional live birth overall, but in the ≥1-prior-miscarriage subgroup progesterone was cost-saving (the averted miscarriages and their downstream management more than paid for the drug). This is the formal basis on which a cheap intervention with a modest absolute effect becomes defensible policy — and the argument that lets NICE pragmatically lower the treatment threshold to one prior loss when the robust efficacy signal was at three. That threshold mismatch is the live debate: erring toward access is reasonable given a cheap, safe drug and a cost-saving signal, but it should be stated plainly that the strongest efficacy evidence sits at ≥3 losses and that certainty is being traded for reach.
NICE NG126 (updated August 2023) translates this into a recommendation: vaginal micronised progesterone 400 mg twice daily for a woman with a scan-confirmed intrauterine pregnancy, bleeding and a previous miscarriage, continued to 16 completed weeks if a heartbeat is seen. It supplements, not replaces, the recurrent-loss work-up (antiphospholipid screen, uterine assessment, thyroid and parental karyotype where indicated).
Preterm birth: EPPPIC as the reference appraisal
Beyond "short cervix and prior spontaneous preterm birth," the argument rests on EPPPIC (Lancet 2021), the individual-participant-data meta-analysis of 31 trials, 11,644 women, 16,185 offspring — a model of how IPD beats aggregate meta-analysis (uniform outcome definitions, proper subgroup interaction tests, the ability to test effect-modifiers like cervical length at the individual level):