Open AccessRationalizing the management of pregnancies of unknown location: Diagnostic accuracy of human chorionic gonadotropin ratio‐based decision tree compared with the risk prediction model M4
Author(s) -
Nadim Batool,
Leonardi Mathew,
Infante Fernando,
Lattouf Ihab,
Reid Shan,
Condous George
Publication year - 2020
Publication title -
acta obstetricia et gynecologica scandinavica
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.401
H-Index - 102
eISSN - 1600-0412
pISSN - 0001-6349
DOI - 10.1111/aogs.13752
Subject(s) - medicine , ectopic pregnancy , confidence interval , human chorionic gonadotropin , pregnancy , logistic regression , gynecology , obstetrics , gonadotropin , retrospective cohort study , hormone , genetics , biology
The objective was to compare the diagnostic accuracy of the decision tree analysis prediction model P1, which incorporates the human chorionic gonadotropin (hCG) ratio (hCG 48 hours/hCG 0 hour), and risk prediction model M4 in the management of women with pregnancy of unknown location (PUL). Material and methods A retrospective diagnostic accuracy study was performed on PUL data collected between August 2011 and September 2018. Women with a PUL were prospectively managed according to the P1 prediction model, which utilizes the hCG ratio and, if necessary, a day (D) 7 hCG. We compared the performance of P1 with the M4 model, a logistic regression mathematical model using initial hCG and hCG ratio, to classify PULs as low risk (failed PUL [failed] or intrauterine pregnancy) or high risk (ectopic pregnancy or persistent PUL). The reference standard was defined as the final PUL outcome. Results Transvaginal ultrasound was done in 3847 consecutive women for early pregnancy complications, 437 (11.3%) of whom were classified as PUL. Final analysis comprised 413 cases with complete data. Final PUL clinical outcomes were: 247 (59.8%) failed PUL, 94 (22.7%) intrauterine pregnancy, 49 (11.8%) ectopic pregnancy and 23 (5.5%) persistent PUL. The sensitivity of P1 and M4 in predicting high‐risk PUL were 81.9% (95% confidence interval [CI] 71.1‐90.0) and 80.6% (95% CI 69.5‐88.9), respectively. The specificities were 74.5% (95% CI 69.5‐79.1) and 75.6% (95% CI 70.7‐80.1), respectively. Conclusions P1 and M4 performed similarly with respect to diagnostic accuracy in predicting PUL outcome. P1 needs to be externally validated.