Using a logistic model to identify women with first-trimester spontaneous abortion suitable for expectant management

Objective To design a method to identify women with first trimester spontaneous abortion suitable for expectant management.

Design A stepwise logistic regression analysis based on retrospective analysis of clinical and biochemical variables.

Setting Department of Obstetrics and Gynaecology, Sahlgrenska University Hospital, Gothenburg, Sweden.

Participants One hundred and three women with inevitable or incomplete spontaneous abortion at < 13 weeks of gestation were chosen for expectant management. Eighty-one women with complete spontaneous abortion (i.e. complete expulsion and/or resolution of pregnancy products within three days of expectant management) were compared with those (n= 22) who underwent surgical evacuation of the uterus, most commonly owing to retained products of conception after three days.

Results Employing a stepwise logistic regression procedure, five diagnostic variables possessing prognostic power were identified: serum progesterone, daily serum hCG change, serum CA125, serum alpha fetoprotein and intrauterine diameter. The logistic regression analysis was also applied to three diagnostic variables chosen for routine clinical use: serum progesterone, serum hCG and intrauterine diameter. The probability of complete spontaneous abortion within three days of expectant management in each woman could be calculated.

Conclusion We have used a logistic model to calculate the probability of complete spontaneous abortion within three days in women with first trimester miscarriages. Such information may be of clinical use in caring for women, as well as for development of management guidelines for those with miscarriages.