Objective: To assess prospectively the ability of two multiparameter diagnostic models and their individual components to predict the outcome of early pregnancies which could not be identified on transvaginal ultrasound scan.
Design: Prospective observational study.
Setting: Dedicated early pregnancy unit in an inner city teaching hospital.
Population: Women with a positive urine pregnancy test and clinical suspicion of early pregnancy complications.
Methods: A full medical history, clinical examination and transvaginal ultrasound scan were carried out at the initial visit. When the location of the pregnancy could not be ascertained by ultrasound, serum beta-human chorionic gonadotrophin (beta-hCG) and progesterone levels were measured. All women were managed expectantly until either a normal pregnancy was visualised on scan; the pregnancy resolved spontaneously or intervention was required due to a worsening of clinical symptoms or non-declining beta-hCG levels.
Main outcome measures: Spontaneous resolution of pregnancy (i.e. cessation of symptoms and decline in serum beta-hCG level to < 20 iu/L) without need for any active intervention.
Results: Of the 104 women recruited, 72 (69%) pregnancies resolved spontaneously. Both multiparameter diagnostic models identified resolving pregnancies with positive predictive values > or = 95%. Their performances were not significantly better compared with individual progesterone levels which achieved a positive predictive value of 97% using a cutoff level of 20 nmol/L.
Conclusion: Serum progesterone measurement alone is as accurate as more complex diagnostic models for the prediction of successful expectant management in pregnancies of unknown location.