SAN ANTONIO – An easy-to-use risk stratification tool accurately predicted which pregnancies of unknown location were ectopic pregnancies by using a model validated by retrospective chart review.
Reeva Makhijani, MD, and her colleagues built the tool using a composite of risk factors to create a “generalized additive model,” or GAM, in combination with beta HCG levels. They presented the results during a poster session at the annual meeting of the American Society for Reproductive Medicine.
The model showed that a prior history of ectopic pregnancy (EP) (P = .0045), a history of pelvic surgery (P = .397), and a presentation of vaginal bleeding (P = .0003) all significantly increased the risk of EP.
Another statistical measure, the area under the receiver operating curve (AUC), helps estimate the likelihood of EP according to beta-HCG levels. When the initial beta-HCG was considered together with the ratio of the initial beta HCG to the presenting beta-HCG, the AUC was 0.889. For the initial beta-HCG level alone, the AUC was 0.793, while for the ratio alone, the AUC was 0.88. Higher AUC figures indicate more predictive power.
, an ob.gyn. resident physician at Brown University, Providence, R.I., and her colleagues have built a prototype of a computer application that calculates risk of EP when the significant risk factors and lab values are entered.
After reviewing the electronic medical records of 800 patients who had pregnancies of unknown location (PUL), in the final analysis Dr. Makhijani and her coauthors included 398 patients whose medical histories allowed assessment of risk factors and whose record included at least two beta-HCG values taken 36-72 hours apart. The investigators also excluded patients with molar pregnancies, ruptured EPs, or who had undergone surgery before a second beta-HCG was obtained.
Of the 398 patients, 40 (10%) were eventually found to have EP, while 168 (42%) had an intrauterine pregnancy, and 190 (48%) were diagnosed with spontaneous abortion.
The patients were about 27 years old on average, and just over half (n = 224) were parous. Vaginal bleeding was a presenting sign in 233 patients, and 284 had abdominal pain. Of those with EP, 34 of 40 had vaginal bleeding, and 25 of 40 had abdominal pain.
In addition to the three factors found to have significant association with EP, the investigators initially considered a number of other patient characteristics, including age, parity, and presentation with abdominal pain. Additional risk factors examined included history of infertility, pelvic inflammatory disease, sexually transmitted disease, intrauterine device placement, and diethylstilbestrol (DES) exposure. None of these were significantly associated with risk of EP.
“Our model can be translated into an easy-to-use risk stratification tool that can accurately predict the risk of EP,” said Dr. Makhijani and her coauthors. “This tool could potentially be used by clinicians and ob.gyn. residencies nationally as [pregnancies of unknown location] are a very common management scenario.”
Dr. Makhijani reported having no disclosures and no outside sources of funding.