Pregnancies of unknown location (PUL) require repeated visits to exclude ectopic pregnancy (EP). We externally validated the M6P (with progesterone) and M6NP (without progesterone) risk models.

In this prospective single-centre cohort (Rennes University Hospital, 2024–2026), consecutive women with a PUL and initial hCG of 25–3,510 IU/L were included. The models were applied directly, without the first progesterone step of two-step triage, at a 5% risk threshold, and were not available to clinicians. We estimated sensitivity, specificity, predictive values and AUC, and compared the models in women evaluable by both (exact McNemar and DeLong tests).

Among 456 women included, 444 were analysable for M6NP (127 EP, 28.6%) and 252 for M6P (105 EP, 41.7%). Sensitivity was 96.2% (95% CI 90.6–98.5) for M6P and 96.1% (91.1–98.3) for M6NP; negative predictive values were 94.4% and 97.1%; specificities were 46.3% and 53.0%. In the 252 women evaluable by both models, sensitivity, specificity and AUC did not differ significantly (all p>0.5).

Both models were highly sensitive but only moderately specific, and a low-risk result did not exclude EP. The absence of a significant difference does not demonstrate equivalence, and progesterone testing was not random.

M6P and M6NP identify a low-risk subgroup of PUL; interpretation must integrate clinical and ultrasound findings. Prospective evaluation of risk-adapted follow-up is warranted.