Evaluation of a New Model for Human Chorionic Gonadotropin Rise in Pregnancies of Unknown Viability

Abstract

Abstract For early pregnancy management, quantitative serum human chorionic gonadotropin (hCG) measurements and transvaginal ultrasonography are the standard clinical guides. Current guidelines use the change in hCG levels to determine likely location of the pregnancy but represent a minimal rate to predict viability. When incorrectly used, these are inaccurate. This study examined the performance of a new hCG threshold model and compared it with 3 established models. The study is a retrospective cohort study. All individuals with at least 2 consecutive quantitative hCG serum levels seen at the University of Missouri–Columbia from January 1, 2015, until March 1, 2020, were screened for study eligibility. Eligibility requirements were 3-fold: first, patients who presented with pregnancy of unknown viability; second, they had an initial hCG level between 2 and 5000 mIU/mL, maintaining the first interval of no greater than 7 days between testing; and third, the final pregnancy outcome was confirmed. Exclusion occurred for patients who had germ cell tumors, had molar pregnancies, had an intrauterine device in place, utilized hCG monitoring for a molar pregnancy or elective abortion, utilized assisted reproductive technology to achieve pregnancy, or were lost to follow-up. Results of the study came from a pool of 688 patients who met all the inclusion criteria. They contributed an average of 2.3 measurements per patient for a total of 1554 hCG measurements. There was an average of 10.1 days for follow-up. Of those patients included, 167 had viable intrauterine pregnancy (IUP) (24.3%), 463 experienced early pregnancy loss (67.3%), and 58 had ectopic pregnancies (8.4%). The rates of rise in values up to 4 days and again between 4 and 6 days were added for a total rate or rise. Because this was a retrospective study, a calculation was used when hCG levels were not available on day 4 or 6. The values were calculated to be conservative and to avoid incorrectly classifying any viable pregnancy as nonviable. Strengths of the study include its large sample size, its inclusion of all 3 pregnancy outcomes, verified pregnancy outcomes, and subsequent misclassification assessments based off of the aforementioned strength. This study confirmed 122 of 168 viable IUPs (72.6%) had live births, and the simple nature of the study allows for easy integration into clinical practice. Finally, the new model allows for greater variation in days on which a patient must present for care. Study limitations included imputed values from hCG samples not specifically collected on day 4 or 6. Despite this, application of the proposed model utilizing merely known values resulted in better performance. In addition, minimal rise for hCG rates in this data set may be a result of internal validation that may differ in larger, multi-institutional data sets. Furthermore, it must be remembered that current methods using hCG thresholds to determine pregnancy viability rely predominantly on mathematical data analysis and thereby do not reach 100% accuracy thresholds due to biologic variations. Based on this reason, hCG rises below cutoffs established should not mandate a reflexive nonviability diagnosis. The study concluded that although shared decision-making remains paramount in determining management steps, the proposed new model simultaneously reduced unnecessary interventions and optimized correct classification rates for all subgroups of pregnancy. Further prospective studies and investigations of cost-effectiveness are needed before widespread model implementation for pregnancy classification.