Predicting pregnancy failure in ‘empty’ gestational sacs

It is an unfortunate fact that spontaneous pregnancy loss (SPL) is a common condition among humans1. Approximately half of all embryos are lost before a pregnancy is known to be present by the mother or clinician2. Among clinically known pregnancies, approximately 15% are doomed to miscarry. Loss rates are particularly high in the 5–6-week range, even in women who appear clinically normal. The risk of pregnancy loss is well known to be even higher among women with symptoms of vaginal bleeding, and among older women3. Sonographers have long recognized their potential contribution to the evaluation of early pregnancy and embryonic viability. Ian Donald, the early pioneer of obstetrical sonography, stated: ‘We are particularly interested in studying the first 12 weeks of uterine development which are even more interesting than the last 12 weeks. It is surely the most crucial period in any being's existence…’4. Another pioneer, Dr Hugh Robinson, produced much of the original work on early normal and abnormal pregnancies5-7. He not only set the stage for future studies, he defined the goals. ‘The primary objective’, said Dr Robinson, ‘is to formulate criteria for the sonar (identification) of abnormal pregnancies such that these diagnoses (can) be applied prospectively and with complete reliability in the active management of established early pregnancy failures.’ Before making reliable diagnoses of SPL we must first accurately characterize normal pregnancies. In doing so, the current authors distinguish between threshold levels and discriminatory levels of any measurement of normal development. A threshold level tells us the earliest we can expect the pregnancy to reach a certain landmark, while the discriminatory level tells us when we should also expect this landmark. For example, a living embryo may be detected by 5.5 weeks in some patients and that may be considered a threshold value. Conversely, diagnosis of a failed pregnancy should only be suggested when there is failure to detect a living embryo after a certain discriminatory level. That level has been determined to be 6 + 4 weeks by at least two studies using transvaginal ultrasound8, 9. Similarly, a gestational sac may be seen as early as 4 + 3 weeks in some patients (threshold level)10, and should always be seen by 5 + 2 weeks (discriminatory level)11. While the threshold value may vary with ultrasound frequency and resolution, the discriminatory value is more dependent on biological variation. It is clear then that in making reliable diagnoses of SPL using a single criterion, we should always consider the discriminatory values. While Dr Robinson's goal of complete reliability is usually straightforward when a discrete embryo is identified12, 13, in many doomed pregnancies the embryo either fails to develop or does not develop beyond a rudimentary stage. In these patients, trophoblastic activity may continue despite the absence of a developing embryo14. The trophoblastic cells continue to elaborate pregnancy hormone, although usually at a much-reduced rate, thus attempting to perpetuate this already lost gestation. Because early normal pregnancies also show a gestational sac but no detectable embryo during a brief but finite stage of early development (approximately 4.5–6 weeks for most normal pregnancies)15, the diagnostic dilemma of an ‘empty’ sac is a common one. Despite the frequency with which ‘empty’ gestational sacs are encountered, there is considerable variability in how such pregnancies are interpreted and managed. We must balance Dr Robinson's goal of certainty with practicality and spare patients from unnecessary follow-up examinations to enable this diagnosis. We cannot agree with the diagnostic guidelines issued by the Board of the Faculty of Clinical Radiology of the Royal College of Radiologists and the Council of the Royal College of Gynaecologists which state, in the setting of an empty sac, the diagnosis of pregnancy failure requires at least one additional examination a minimum of 7 days later, and in the case of an empty sac < 15 mm in diameter, a follow-up scan is recommended 2 weeks later16. Although all diagnosticians recognize the value of adding the dimension of time and serial observation to particularly difficult cases, this philosophy should not become so pervasive that one virtually abdicates diagnostic responsibility. Once a failed pregnancy has been reliably diagnosed, nothing is gained by additional assessment. In addition to the luxury of follow-up, other methods to assess early pregnancies include hormonal assays (human chorionic gonadotropin (hCG), estrogen, progesterone, human placental lactogen, pregnancy specific B-glycoprotein and alpha-fetoprotein) and sonography17-19. Among the hormonal studies, progesterone and hCG levels are more accurate than the others. Falling hCG levels predict pregnancy failure quite accurately. Unfortunately, serial hCG determinations are not uncommonly equivocal. There are a number of sonographic criteria for diagnosing SPL20, 21 and these have been analyzed in detail22. Among the most predictive criteria for diagnosing a failed pregnancy is a large gestational sac without a living embryo23. Gestational sac size is normally a good predictor of gestational age: the sac is first visible by 4.5–5 menstrual weeks and then grows approximately 1 mm per day during the first trimester24. For this reason, there is an excellent correlation between sac size, gestational age, and hCG levels25. As the pregnancy progresses, normal development of the embryo and embryonic structures also occur in a very predictable fashion until a living embryo is visible. A living embryo may be seen in normal pregnancies when the sac is as small as 10 mm (threshold level) and should always be seen above a critical discriminatory sac size. This discriminatory sac size has been proposed to be in the range 20–30 mm using transabdominal ultrasound, and in the range 16–20 mm with transvaginal scans22, 26-30. Because of occasional exceptions seen using a 16-mm discriminatory sac size31, we support a conservative level of 18 mm or larger as an isolated criterion, although this must be correlated with other sonographic and clinical data, and it is certainly possible to diagnose pregnancy failure for sacs that are smaller. In evaluating discriminatory sac sizes, it is worth noting that using transabdominal ultrasound, Robinson reported in 1975 that non-visualization of a living embryo in sacs larger than 2.5 cc (corresponding to 17 mm in diameter) was evidence for a failed gestation5. Another common and useful sonographic feature for predicting SPL is a disproportionately enlarged amnion compared to embryonic development32. During very early pregnancy, the crown–rump length is approximately equal to amnion sac diameter, again illustrating that normal embryological development is predictable and highly consistent. Therefore, a disproportionately small or non-visible embryo within an ‘enlarged’ amnion is a good marker for a failed pregnancy (Figure 1). This has been referred to as the ‘empty amnion’ sign33. Such enlarged amniotic cavities may be confused for an enlarged yolk sac. Larger amnion cavities tend to be found in larger gestational sacs but this sign is also valid when the amnion is small. Empty amnion. Transvaginal scan shows a gestational sac measuring 16 mm in mean diameter, without a living embryo. The yolk sac (YS) is smaller than seen with most normal pregnancies, but is normally pushed to the periphery by the expanding amnion (arrows). An embryo of approximately equal size to the amnion is expected within it at this time. The findings are diagnostic of a failed pregnancy. In addition to criteria of a disproportionately enlarged sac size or amnion compared to embryonic development, other sonographic criteria for SPL include delayed embryonic development, delayed sac growth, non-visualization of a yolk sac, and a number of morphologic criteria such as irregular sac shape, thin choriodecidual reaction, and abnormal position. Also, disproportionately low hCG levels compared to gestational sac size may indicate SPL34, 35. In most cases we have more than one sonographic criterion in addition to clinical data to indicate that a pregnancy has failed. It is with this background that the current issue of Ultrasound in Obstetrics and Gynecology contains two related papers addressing the issue of an ‘empty’ gestational sac36, 37. In agreement with the literature, both studies reported a high frequency of SPL: Elson and colleagues reported SPL among 59% of patients presenting with an ‘empty’ gestational sac and Falco and colleagues found that nearly two-thirds of such patients suffered a pregnancy loss among those presenting with symptoms of miscarriage. Also consistent with previous studies, both groups found that clinical factors associated with an increased risk of a pregnancy failure included: increased maternal age, increased menstrual age, lower hCG levels, lower progesterone levels, and vaginal bleeding. Unlike the existing literature, neither study found that an absolute sac size reliably distinguished normal from failed pregnancies. However, this is not unexpected since Falco et al. excluded sacs larger than 16 mm from their study, and Elson et al. excluded sacs 20 mm or larger. Both studies suggest a higher accuracy of interpretation when clinical and laboratory data are correlated with ultrasound findings for diagnosing failed gestations. Falco et al. found that maternal age, menstrual age, sac size compared with menstrual history and β-hCG level were significantly correlated with the subsequent occurrence of miscarriage. Using a combination of factors, they report no ongoing pregnancies in patients older than 35 years, with a serum β-hCG level < 1200 mUI/mL), a gestational age > 7 weeks, or a mean sac diameter < −1.64 standard deviations. However, caution should be exercised in this interpretation since, while it is true that the vast majority of ‘empty’ sacs associated with low hCG levels will result in SPL, an early normal pregnancy may also have low hCG levels when the sac is very small (<5 mm). This fact again emphasizes the importance of correlating hCG levels with sac size, rather than treating all empty sacs equally. One also wonders about the necessity of hCG levels, if the menstrual dates are certain, since lack of a living embryo by 7 menstrual weeks would be considered diagnostic of a failed gestation without the need for other criteria. Conversely, if the menstrual dates are uncertain, correlation with ultrasound and the menstrual dates again proves to be very helpful. For example, if the menstrual dates exceed 6 + 4 weeks, and there is no evidence of a living embryo with transvaginal scans, and the gestational sac appears abnormal sonographically, then we can be reasonably certain that the pregnancy has failed even when the sac is less than 16 mm in diameter. Conversely, if the menstrual dates are unreliable or potentially in error and the gestational sac appears normal, then we must disregard the dates and assume the pregnancy is viable until proven otherwise. Progesterone levels may be more helpful than hCG levels for evaluation of early viability. Low progesterone levels are known to be associated with a poor prognosis38, and this observation is confirmed by Elson et al. However, because low progesterone levels may occasionally be seen in normal pregnancies, progesterone levels alone may not be diagnostic. A unique observation by Elson et al. is to combine progesterone levels with sonographic evidence of an empty sac to predict a failed pregnancy. Given the extensive literature comparing sonographic findings with hCG levels, it is perhaps surprising that there are not more studies correlating ultrasound findings with progesterone levels. Incorporation of gestational sac size with maternal age is a logical improvement on the use of progesterone for diagnosing failed pregnancies. As the model of Elson and colleagues predicts, low progesterone levels may be seen with normal pregnancies, but only when the pregnancy is very early. Their model shows that the probability of SPL increases rapidly as progesterone levels fall below 30 nmol/L. According to their model, a 20-year-old woman with a sac of 5 mm carries a probability of normal outcome of 31% when the progesterone level is 30 nmol/L and 3.7% for a progesterone level of 20 nmol/L. These values drop to 4% and 0.3%, respectively, when the sac is 20 mm. In comparison, for a 40-year-old woman with a progesterone level of 30, the probability of normal outcome is only 1.7% for a 5-mm sac and 0.16% for a 20-mm sac. The two papers presented here illustrate that experienced clinicians rarely work with a single piece of information and must consider a number of variables simultaneously. No formula is a substitute for clinical skill and experience in considering all of the information to obtain a diagnostic impression. The range of examination and interpretive skills in sonography is simply too broad for that. The vast majority of failed pregnancies will have more than one sonographic and clinical criterion to add credibility to the impression of a failed pregnancy, and as clinicians we use all of this information to arrive at a final diagnosis. In the final analysis, when considering criteria useful for diagnosing SPL, we once again return to Hugh Robinson's objective by aiming for a certain diagnosis before actively managing such pregnancies. This goal is just as valid today as it ever was. Future studies can be anticipated that also attempt to simulate clinical experience with objective parameters on a single examination. As the current papers illustrate, the most useful methods will undoubtedly correlate a variety of sonographic, clinical, and laboratory results in predicting pregnancy outcome.