The HCG values at which the gestational sac is predicted to be seen based on this study are similar to those reported by Connolly et al. with the discriminatory level for visualization of the gestational sac being slightly higher (3,994 mIU/mL here compared to 3,510 mIU/mL). The highest reported HCG value with no gestational sac seen on transvaginal ultrasound is 9,083 mIU/mL for a patient with a triplet pregnancy and 4,336 mIU/mL for a singleton pregnancy [11, 12].
The HCG values for visualizing a yolk sac in this study are higher than found in the Connolly study. The HCG level for predicting visualization of a yolk sac 99% of the time of 39,454 mIU/mL is within the confidence interval found in the Connolly study. Even though the Connolly study has a larger sample size, the confidence interval reported in their study is still wide since the sample size is relatively small. One of the biggest limitations of this study is also sample size. Our sample size was data limited as we were not able to review data from before 2016.
Even in the modern era of medicine where electronic medical records are the norm, compiling data on pregnancies with known outcomes and corresponding ultrasound and HCG values in early pregnancy is a laborious task. The initial study by Kadar et al. in 1981 examined the records of 53 patients. The Connolly study had a sample size of 366 patients who presented with pain or bleeding and went on to have a viable pregnancy. In this study we examined records of patients who had a live delivery and retrospectively collected data on those with a recorded early TVUS and HCG. We also reviewed gynecology triage records of patients presenting for evaluation of early pregnancies to obtain data on early ectopic pregnancies and spontaneous abortion. This allowed for in depth clinical correlation which can help to evaluate pregnancies of unknown location.
Figure 3 includes HCG values for pregnancies with no gestational sac on initial TVUS that went on to become a spontaneous abortion. It is likely that some of these pregnancies have high HCG values with no intrauterine gestational sac because the gestational sac had already passed. These could be confused clinically with ectopic pregnancy because some of these pregnancies presented with HCG above the 99% discriminatory level of 3,994 mIU/mL. Serial HCG measurement is helpful for detecting spontaneous abortion with a high presenting HCG because the level typically decreases at a mean rate of 70–75% over 2 days [13].
As seen in Fig. 3, most of the ectopic pregnancies presented with an HCG value under 2,000 mIU/mL. We found that 90% of ectopic pregnancies had an HCG at presentation less than the value at which 99% of viable pregnancies can be detected by visualizing a gestational sac on transvaginal ultrasound (3,994 mIU/mL in this study). Few of the ectopic pregnancies included would have qualified for immediate intervention (medical or surgical treatment for ectopic pregnancy) based on HCG above the discriminatory level at the time of presentation. Table 2 shows that there is significant symptom overlap between pregnancies of unknown location regardless of eventual pregnancy outcome (SAB, ectopic, or viable pregnancy). More emphasis should be placed on repeat HCG values for early detection of ectopic pregnancy in addition to assessing clinical presentation.
The commonly accepted practice of assessing pregnancy of unknown location by repeating the HCG after two days dates back to a 1981 study by Kadar et al [14]. The 48-hour sampling interval was recommended because “after 1 day the difference between the mean percent hCG increase of intrauterine and ectopic pregnancies (20%) is less than twice the interassay variability.” The intrerassay and intraassay coefficient of variation cited in the 1981 Kadar study is “less than 15%”. Since modern day assays have coefficients of variation of 5% or less, repeating the HCG after 24 hours is appropriate in modern practice. This would allow for more rapid management and could decrease the risk of ruptured ectopic pregnancy.
This study is accompanied by the dataset used for the logistic regression as well as all of the Stata code to completely recreate the statistical analysis [9]. This allows other investigators to use our methods with their datasets or to combine datasets from multiple studies.
In conclusion, this study is in agreement with the 2013 study by Connolly et al. since the logistic regression model for this data predicts that 99% of early viable singleton pregnancies will have a visible gestational sac on transvaginal ultrasound when the HCG level reaches 3,994 mIU/mL. This limits the utility of a discriminatory level to detect ectopic pregnancies. Since only 10% of ectopic pregnancies included in this study had an HCG value above 3,994 mIU/mL, the discriminatory level concept is not very useful in detecting ectopic pregnancies in modern practice. We feel that rapid repeat HCG measurement (such as repeating after only 24 hours) is an underutilized strategy for evaluating early pregnancies and is appropriate based on modern HCG assays that are much more precise than they were in the 1980s.