Progesterone After Three Losses: Which Trial Actually Describes Her
Two well-designed randomized trials reached genuinely different bottom lines on the same drug — and which one actually describes this patient is a matter of matching her specific presentation, not picking a favorite.
Sofia P., a 35-year-old woman, works as a translator and has had three first-trimester pregnancy losses over the past four years, the most recent eight months ago, each occurring between six and nine weeks with no identified cause found on a full recurrent-loss workup — normal karyotype for both her and her partner, no antiphospholipid antibodies, no uterine anomaly on saline-infusion sonogram. She confirmed a new pregnancy three days ago via a positive urine test and quantitative hCG consistent with roughly five weeks' gestation, and came in today reporting light spotting since yesterday — not heavy, but enough to frighten her given her history.
Two major randomized trials on vaginal progesterone in recurrent pregnancy loss reached genuinely different bottom-line conclusions, and the reason isn't that one trial is simply wrong. PROMISE (Coomarasamy et al., NEJM 2015), the larger of the two, found no significant live-birth-rate benefit from progesterone in an unselected recurrent-loss population, where bleeding was not required to enter. PRISM (Coomarasamy et al., NEJM 2019) asked a different question of a different group: every one of its 4,153 participants was already bleeding in the current pregnancy. It too was negative overall — live birth 75% against 72% — but one of ten prespecified subgroups, stratified by number of prior miscarriages, separated sharply: 72% against 57% among women with three or more. Sofia clears PRISM's entry criterion on today's spotting and sits in that stratum on her three losses, which is why the two trials point opposite ways for her and not for most patients carrying either feature alone. The caveat is real and worth naming plainly rather than glossing over: PRISM's own authors flagged that finding as one of ten prespecified subgroups tested without adjustment for multiple comparisons, the kind of result that can look real without replicating. But absent a test that would tell Sofia something more specific than PRISM's own subgroup already does, this is the best evidence available for today's decision, not a lesser one dressed up as certain. Vaginal progesterone's own safety profile in early pregnancy is well established regardless of which trial turns out to matter more for her, which is part of why starting it today while the rest of the workup continues costs her little if the benefit turns out not to hold.
Urgent visit, spotting since yesterday
My default read is PROMISE — it's the larger, primary trial, and it found no significant benefit from vaginal progesterone in an unselected recurrent-loss population. Absent a specific, well-supported reason to think Sofia differs from that population, I'd want a clear justification before adding a drug on the strength of a smaller subgroup finding.
If she'd come in today with no bleeding at all, I don't think this is a close call — PROMISE's null result would be the more directly applicable finding, and I wouldn't be pushing back the way I am now.
But she has come in bleeding, which is how PRISM defined its whole population — PROMISE never required that. I'll concede PRISM missed its primary endpoint, so I'm not claiming a positive trial. What I'm claiming is narrower: within it, the prior-miscarriage stratification was prespecified, and the three-or-more group separated at 72% against 57%. Sofia is in that group, on today's presentation and on her history both.
I'd agree PROMISE's null result stands for the unselected population it studied — my point is that Sofia was never in that population. She would have been eligible for PRISM and not for PROMISE, and that is the whole distance between the two answers.
This is a genuine case of two well-conducted trials answering related but distinct questions, and the honest approach is matching her specific presentation against each trial's own entry criteria rather than treating either result as the universal answer. On that test she is inside PRISM's entry criterion and inside the stratum where its effect sat, and outside PROMISE's population altogether. That is a real reason to offer progesterone — held at the strength a prespecified subgroup of a trial that missed its primary endpoint will actually carry, which is a reasonable offer rather than a clear indication.
Agreed: vaginal progesterone started today, with Sofia told directly why — that she is eligible for the trial that found a signal and was never in the one that didn't, and that the signal itself comes from a subgroup of a trial that missed its main endpoint. All three voices converged once the disagreement was reframed from "does progesterone work" to "which trial's population does she actually match," a question her current bleeding answered directly.