Abstract
BACKGROUND:
Pregnancy loss prediction based on routinely measured ultrasound characteristics is generally aimed toward distinguishing nonviability. Physicians also use ultrasound indicators for patient counseling, and in some cases to decide upon the frequency of follow-up sonograms. To improve clinical utility, allocation of cut-points should be based on clinical data for multiple sonographic characteristics, be specific to gestational week, and be determined by methods that optimize prediction.
OBJECTIVES:
To identify routinely measured features of the early first trimester ultrasound and the gestational age–specific cut-points that are most predictive of pregnancy loss.
MATERIALS AND METHODS:
This was a secondary analysis of 617 pregnant women enrolled in the Effects of Aspirin in Gestation and Reproduction (EAGeR) trial; all women had 1–2 previous pregnancy losses and no documented infertility. Each participant had a single ultrasound with a detectable fetal heartbeat between 6 weeks 0 days and 8 weeks 6 days. Cut-points for low fetal heart rate and small crown–rump length were separately defined for gestational weeks 6, 7, and 8 to optimize prediction. Identity and log-binomial regression models were used to estimate absolute and relative risks, respectively, and 95% confidence intervals between jointly categorized low fetal heart rate, small crown–rump length, and clinical pregnancy loss. Adjusted models accounted for gestational age at ultrasound in weeks. Missing data were addressed using multiple imputation.
RESULTS:
A total of 64 women experienced a clinical pregnancy loss following the first ultrasound (10.4%), 7 were lost to follow-up (1.1%), and 546 women (88.5%) had a live birth. Low fetal heart rate and small crown–rump length (≤122, 123, and 158 bpm; ≤6.0, 8.5, and 10.9 mm for gestational weeks 6, 7, and 8, respectively) were independent predictors of clinical pregnancy loss, with greatest risks observed for pregnancies having both characteristics (relative risk, 2.08; 95% confidence interval, 1.24–2.91). The combination of low fetal heart rate and small crown–rump length was linked to a 16% (95% confidence interval, 9.1–23%) adjusted absolute increase in risk of subsequent loss, from 5.0% (95% confidence interval, 1.5–8.5%) to 21% (95% confidence interval, 15–27%). Abnormal yolk sac diameter or the presence of a subchorionic hemmhorage did not improve prediction of clinical pregnancy loss.
CONCLUSION:
Identified cut-points can be used by physicians for patient counseling, and in some cases to decide upon the frequency of follow-up sonograms. The specified criteria should not be used to diagnose nonviability.
Keywords: fetal loss, miscarriage, sonogram, sonographic
Although pregnancy loss is a relatively common event that occurs in about 30% of pregnancies, the factors that predict pregnancy losses are not well understood. As prenatal ultrasound has become a standard element of routine prenatal care, many studies have focused on the ability of early ultrasound characteristics, such as fetal heart rate and crown–rump length, to predict nonviability. Studies designed with the objective of determining nonviability with certainty do not allow for patient counseling and more frequent follow-up for women at high risk for a pregnancy loss, but may not initially meet the nonviability thresholds.
Existing studies aimed to predict pregnancy loss have primarily investigated specific components of the first trimester ultrasound scan, including subchorionic hemorrhage,1–4 fetal heart rate,5–8 crown–rump length,9–11 and yolk sac diameter.12–15 Individually these factors have been associated with pregnancy loss. Although there have been studies examining relationships between multiple ultrasound parameters and risk of pregnancy loss,16–22 prediction studies have been performed on a limited basis,17,19–21 and none of these studies have identified data-driven cut-points most predictive of pregnancy loss, limiting ease of clinical implementation. Furthermore, existing studies have not consistently accounted for gestational age,16–18,20,21 despite embryonic features changing with advancing gestational age. Therefore, our objective was to identify routinely measured features of the early pregnancy sonogram and their gestational age–specific cut-points that are most predictive of subsequent risk of pregnancy loss.
Materials and Methods
Study design
This study was a secondary analysis of data obtained from the Effects of Aspirin in Gestation and Reproduction (EAGeR) trial, a block-randomized, double-blind, placebo-controlled trial conducted during 2006–2012. EAGeR was designed to study the effects of preconception low-dose aspirin on live birth in women with 1–2 prior pregnancy losses.23 Study design, methods, and inclusion and exclusion criteria have previously been described.23,24 Briefly, women 18–40 years of age with no known major medical conditions were recruited from 4 medical centers in the United States. Inclusion was limited to women having regular menstrual cycles and attempting to conceive with no prior diagnosis of infertility. Women were randomized to daily low-dose aspirin (81 mg/d) or placebo. Women were followed up to the completion of 6 menstrual cycles while trying to conceive, and through pregnancy for those who became pregnant, with study treatment continuing through week 36 of pregnancy. The institutional review board at each study site and data coordinating center approved the trial protocol (University of Utah #1002521; Scranton, Pennsylvania #HHSN275200403394; University at Buffalo #SPM0900107A; University of Colorado #08–0982). All participants provided written informed consent prior to enrollment. The data coordinating center and each site obtained institutional review board approvals. The trial was registered on ClinicalTrials.gov (#NCT00467363 27 April 2007).
Exposure: ultrasound parameters
For women with evidence of an ongoing pregnancy (a positive urine pregnancy test with no clinical evidence of subsequent pregnancy loss), an early ultrasound targeted for approximately 6.5 weeks of gestation was performed to clinically confirm the pregnancy by visualization of a gestational sac. Gestational age in weeks and days was recorded, as well as crown–rump length, yolk sac diameter, subchorionic hemorrhage, and fetal heart rate. Transvaginal ultrasounds were conducted by American Institute of Ultrasound in Medicine (AIUM)–certified sonographers in AIUM-certified centers and recorded for study purposes. M-mode was used to measure fetal heart rate, which was recorded as the average of 2 measurements. Fetal sex was unavailable at the early ultrasound.
Gestational age at ultrasound
For this analysis, gestational age at the time of the ultrasound was calculated based on the difference between the date of the ultrasound and last menstrual period (LMP). The LMPs were determined in this population from high-quality prospective bleeding dates reported in preconception daily diaries and fertility monitors. Accordingly, gestational ages for purposes of this analysis were not re-dated based on gestational age from the ultrasound.
Outcome: clinical pregnancy loss
Clinical pregnancy loss was defined as a loss identified by the participant or clinician following clinical confirmation of pregnancy by sonographic identification of a gestational sac. Pregnancy follow-up included clinical or telephone visits every 4 weeks up to week 36 of gestation. A postpartum telephone visit was conducted 6–8 weeks after delivery or pregnancy loss. Medical chart abstraction was completed on each participant in the study, from which clinical pregnancy losses were verified.
Sample derivation
Of the 1228 participants in the study,23 there were 732 (59.6%) clinically confirmed pregnancies. There was no ultrasound for 25 of these pregnancies (3.4%), of which 15 women experienced a clinical loss and 10 had a live birth. Of the remaining 707 unique ultrasound examinations, 674 (95.3%) occurred between 6 weeks days and 8 weeks 6 days (Figure 1) by LMP dating. Of those examinations, 617 (91.5%) had a detectable fetal heart rate.
FIGURE 1.
Participant inclusion in the ultrasound characteristics and clinical loss Effects of Aspirin in Gestation and Reproduction (EAGeR) trial analysis: number of women who were randomized, had clinically confirmed ultrasounds between 6 weeks 0 days and 8 weeks 6 days gestation, and a detectable fetal heart rate
Base predictors
In addition to ultrasound features, we sought to identify the most informative predictors of clinical pregnancy loss from multiple biological, clinical, and demographic covariates. At baseline, participants completed questionnaires on demographics, lifestyle habits, socioeconomic status, and reproductive history. Covariates were selected for model consideration based on identified a priori relationships with pregnancy loss.25,26 Baseline demographic covariates included study recruitment site (Scranton, PA; Denver, CO; Buffalo, NY; Salt Lake City, UT), treatment assignment (aspirin, placebo), age at study entry (years), and annual income (≥$100,000 USD; $75,000–99,000; $40,000–74,999; $20,000–39,999; ≤$19,999). Baseline clinical measures included gestational age at the first ultrasound, body mass index (kg/m2), systolic and diastolic blood pressure (mm Hg), exercise history (low, moderate, high), number of prior pregnancy losses (1, 2), number of prior live births (0, 1, 2), and time since the last pregnancy loss (≤4 months, 5–8 months, 9–12 months, >12 months).
Statistical analyses
Relationships between baseline measures and ultrasound parameters were examined using t tests, χ2, or Fisher exact tests, as appropriate.
Determination of cut-points for continuous ultrasound measurements
Using complete case data, cut-points were identified using the receiver operating characteristic (ROC) curve and the Youden index, a common summary measure of the ROC curve that indicates the maximum potential effectiveness of a biomarker.27 As part of this procedure, univariate logistic regression models were developed relating fetal heart rate, crown–rump length, or yolk sac diameter to clinical loss to produce ROC curves. For fetal heart rate and crown–rump length, this was conducted separately for gestational weeks 6, 7, and 8. Given the high proportion of missingness for yolk sac diameter (19.8%) and its relatively consistent median over gestational age, abnormal yolk sac diameter was defined for the full cohort without regard to gestational age. Given that both extreme high and low yolk sac measurements have been associated with elevated risk of pregnancy loss,28 yolk sac diameter was dichotomized at the median (4.0 mm) to separately define low and high cut-points.
Multiple imputation for missing data
Missing data for yolk sac diameter (n = 122), crown–rump length (n = 11), and subchorionic hemorrhage (n = 7) were addressed using multiple imputation to produce 10 imputed datasets. All other variables were completely observed.
Determination of variables most predictive of clinical pregnancy loss
Training sets (n = 462, 75%) of the 10 imputations were used for predictive feature identification, as this procedure requires complete data (Appendix 1). Each training set consisted of the same individuals, with between-imputation variation due to missing data imputation. The subset of individual measures (gestational age, low-dose aspirin randomization, maternal age, body mass index, diastolic and systolic blood pressures, annual income, exercise history, number of prior pregnancy losses, number of prior live births, and time since the last pregnancy loss) and ultrasound measurements (fetal heart rate, subchorionic hemorrhage, crown–rump length, abnormal yolk sac diameter) most predictive of clinical pregnancy loss were identified by recursive feature elimination with 10-fold cross-validation using the random forest algorithm, a machine learning algorithm.29 Recursive feature elimination explores all possible subsets of attributes to identify the subset of measures that yield the best model performance (ie, through minimizing cross-validation root mean squared error [RMSE]). In minimization of RMSE, the set of predictors common to all 10 imputations comprised gestational age, crown–rump length, and fetal heart rate. Three of 10 imputations identified maternal age as a predictor. Furthermore, 2 of 10 imputations included systolic and/or diastolic blood pressure as predictors.
Assessment of model performance
Test sets (n = 155, 25%) of the 10 imputations were used to evaluate model performance using the area under the receiver operating characteristic curve (AUC), which measures how well models distinguish between clinical pregnancy losses vs live birth. AUCs are reported as pooled summary measures across the 10 imputations using the PROC MIANALYZE procedure.
To illustrate the relative increase in prediction gained by using crown–rump length and fetal heart rate, a base model including only gestational age was compared to a model additionally including crown–rump length and fetal heart rate. In sensitivity analyses, additional ROC curves were compared to investigate whether further addition of maternal age and blood pressure (diastolic and systolic), addition of subchorionic hemorrhage and extreme yolk sac diameter, or continuous expression of fetal heart rate and crown–rump length changed model prediction of clinical loss.
Informed by recursive feature elimination, regression models were used to jointly estimate relationships with 95% confidence intervals between clinical pregnancy loss and low fetal heart rate and small crown–rump length in the imputed datasets using the PROC MIANALYZE procedure. Identity-binomial models estimated summary absolute risks and differences, and log-binomial models estimated summary relative risks (RRs).30Adjusted models accounted for gestational age.
Analyses were conducted in SAS version 9.4 (SAS Institute, Cary, NC) and R version 3.5.0 (2018-04-23); R Foundation for Statistical Computing, Vienna, Austria.
Results
Study sample characteristics
Of the 617 women in this analysis, 64 women experienced a clinical pregnancy loss following the first ultrasound (10.4%), 7 women were lost to follow-up (1.1%), and 546 (88.5%) women had a live birth. Women most frequently had an ultrasound during gestational week 7 (7 weeks 0 days to 7 weeks 6 days; n = 281, 45.5%), followed by week 6 (6 weeks 0 days to 6 weeks 6 days; n = 259, 42.0%) and week 8 (8 weeks 0 days to 8 weeks 6 days; n = 77, 12.5%). Median (interquartile range [IQR]) fetal heart rate (of viable fetuses) was 124 (118–132), 137 (122–146), and 144 (125–163) bpm; median (IQR) crown–rump length was 6.0 (4.5–8.0), 8.6 (6.0–10.6), and 10.0 (6.6–16.1) mm; and median (IQR) yolk sac diameter was 4.0 (2.9–4.5), 4.0 (3.0–5.0), and 4.7 (3.2–5.0) mm for gestational weeks 6, 7, and 8, respectively. There were 63 documented cases of subchorionic hemorrhage (10.3%) (Table 1).
TABLE 1.
Early ultrasound fetal characteristics by participant characteristics
| Characteristic | All (N = 617) | Low fetal heart ratea (n = 250) | Small crown–rump lengtha (n = 310) | Extreme yolk sac diameterb (n = 143) | Subchorionic hemorrhage (n = 63) |
|---|---|---|---|---|---|
| Age, y | 28.6 ± 4.6 | 28.2 ± 4.6 | 28.3 ± 4.7 | 28.5 ± 4.5 | 28.6 ± 4.2 |
| Body mass index | 25.4 ± 6.0 | 25.3 ± 5.4 | 25.7 ± 6.3 | 25.4 ± 6.9 | 24.6 ± 4.3 |
| Missing (n = 5) | |||||
| Systolic blood pressure | 111 ± 11.7 | 110 ± 11.2 | 111 ± 12.4 | 112 ± 11.0 | 110 ± 12.4 |
| Missing (n = 3) | |||||
| Diastolic blood pressure | 71.5 ± 9.1 | 70.9 ± 8.5 | 71.9 ± 8.9 | 72.4 ± 8.8 | 72.3 ± 8.3 |
| Missing (n = 4) | |||||
| Treatment | |||||
| Aspirin | 319 (51.7) | 134 (53.6) | 165 (53.2) | 85 (59.4)c | 35 (55.6) |
| Placebo | 298 (48.3) | 116 (46.4) | 145 (46.8) | 58 (40.6)c | 28 (44.4) |
| Annual income (USD) | |||||
| ≤19,999 | 40 (6.5) | 15 (6.0) | 22 (7.1) | 10 (7.0) | 5 (7.9) |
| 20–39,999 | 145 (23.5) | 62 (24.8) | 82 (26.5) | 37 (25.9) | 12 (19.1) |
| 40–74,999 | 89 (14.4) | 33 (13.2) | 39 (12.6) | 15 (10.5) | 5 (7.9) |
| 75–99,999 | 87 (14.1) | 35 (14.0) | 45 (14.5) | 16 (11.2) | 7 (11.1) |
| ≥100,000 | 256 (41.5) | 105 (42.0) | 122 (39.4) | 65 (45.5) | 34 (54.0) |
| Gestational age | |||||
| 6 wk 0 days–6 wk 6 days | 259 (42.0) | 121 (48.4)c | 132 (42.6) | 58 (40.6) | 25 (39.7) |
| 7 wk 0 days–7 wk 6 days | 281 (45.5) | 77 (30.8)c | 139 (44.8) | 66 (46.2) | 26 (41.3) |
| 8 wk 0 days–8 wk 6 days | 77 (12.5) | 52 (20.8)c | 39 (12.6) | 19 (13.3) | 12 (19.1) |
| Exercise | |||||
| Low | 163 (26.4) | 64 (25.6) | 84 (27.1) | 34 (23.8) | 13 (20.6) |
| Moderate | 250 (40.5) | 109 (43.6) | 123 (39.7) | 65 (45.5) | 30 (47.6) |
| High | 204 (33.1) | 77 (30.8) | 103 (33.2) | 44 (30.8) | 20 (31.8) |
| Prior loss number | |||||
| 1 | 412 (66.8) | 169 (67.6) | 205 (66.1) | 102 (71.3) | 38 (60.3) |
| 2 | 205 (33.2) | 81 (32.4) | 105 (33.9) | 41 (28.7) | 25 (39.7) |
| Prior live births | |||||
| 0 | 255 (41.3) | 114 (45.6) | 139 (44.8)c | 67 (46.9)c | 25 (39.7) |
| 1 | 242 (39.2) | 96 (38.4) | 126 (40.7)c | 43 (30.1)c | 27 (42.9) |
| 2 | 120 (19.5) | 40 (16.0) | 45 (14.5)c | 33 (23.1)c | 11 (17.5) |
| Time since last loss, mo | |||||
| ≤4 | 373 (61.9) | 162 (66.4) | 195 (64.8) | 85 (59.9) | 41 (66.1) |
| 5–8 | 104 (17.3) | 38 (15.6) | 50 (16.6) | 25 (17.6) | 9 (14.5) |
| 9–12 | 40 (6.6) | 15 (6.2) | 22 (7.3) | 9 (6.3) | 4 (6.5) |
| >12 | 86 (14.3) | 29 (11.9) | 34 (11.3) | 23 (16.2) | 8 (12.9) |
Continuous data are presented as mean ± standard deviation. Categorical data are presented as n (%).
USD, US dollars.
Low fetal heart rate (≤122, 123, and 158 bpm) and small crown–rump length (≤6.0, 8.5, and 10.9 mm) were separately defined for gestational weeks 6, 7, and 8, respectively, using the Youden index as the optimality criterion;
Abnormal yolk sac diameter was defined as measurements <2.3 mm (15th percentile) or >5.1 mm (87th percentile);
Significant differences (P≤.05) by t test compared to group without specified ultrasound characteristic (eg, pregnancies with low fetal heart rate were compared to pregnancies without low fetal heart rate).
Low fetal heart rate was more prevalent with younger gestational age. Women randomized to aspirin were more likely to have extreme yolk sac diameter measurements. Small crown–rump length and extreme yolk sac diameter were both associated with having no prior live births.
Principal Findings
Cut-points for predicting clinical losses using low fetal heart rate were found to be ≤122, 123, and 158 bpm for gestational weeks 6, 7, and 8, respectively. Cut-points for small crown–rump length were ≤6.0, 8.5, and 10.9 mm for gestational weeks 6, 7, and 8, respectively. Cut-points for abnormal yolk sac diameter were measurements of 2.3 mm (15th percentile) or greater than 5.1 mm (87th percentile).
Using these cut-points, of women who later experienced a clinical loss, 37 (59.7%) occurred among women having both low fetal heart rate and small crown–rump length, compared to 6 losses (9.7%) among women having neither characteristic. Following adjustment for gestational age, this corresponded to a 21% (15–27%) probability of clinical pregnancy loss among women having both characteristics compared to 5.0% (1.5–8.5%) probability with neither characteristic, equating to a 16% (9.1–23%) difference in adjusted absolute risk and a relative risk of 2.08 (95% CI, 1.24–2.91) (Table 2).
TABLE 2.
Adjusteda pregnancy loss risks, differences, relative risks, and 95% confidence intervals (CI) for early ultrasound characteristics: joint categorization by small crown–rump length and low fetal heart rate
| Small crown–rump length | Low fetal heart rate | Risk (95% CI) | Difference (95% CI) | Relative risk (95% CI) |
|---|---|---|---|---|
| No | No | 5.0% (1.5–8.5%) | Reference | Reference |
| No | Yes | 12% (2.1–22%) | 7.0% (−3.3% to 17%) | 1.33 (0.12–2.53) |
| Yes | No | 15% (8.3–23%) | 10% (2.5–18%)b | 1.79 (0.86–2.71) |
| Yes | Yes | 21% (15–27%) | 16% (9.1–23%)b | 2.08 (1.24–2.91)b |
Risk differences may not calculate exactly from risk column because of rounding percentages to the nearest whole number.
Gestational age (in weeks) at ultrasound since last menstrual period (continuous);
Values are significantly different from reference category.
In applying these cut-points to the test data, the pooled area under the AUC for prediction of pregnancy loss was 0.63 (95% CI, 0.47–0.79) for gestational age, which increased to 0.78 (0.65–0.90) after addition of low fetal heart rate and small crown–rump length (Table 3 and Supplemental Figure 1). The addition of age or blood pressure, or yolk sac diameter and subchorionic hemorrhage, or continuous expression of fetal heart rate and crown–rump length did not meaningfully improve prediction of clinical pregnancy loss (Table 3 and Supplemental Figure 2).
TABLE 3.
Pooled area under the receiver operating characteristic curves (AUCs) for gestational age at ultrasound (Base) compared to that with addition of low fetal heart rate and small crown–rump length (Primary), Primary compared to that with addition of maternal age (Age) or systolic and diastolic blood pressures (Blood pressure) or yolk sac diameter and subchorionic Hemorrhage (YSD + hemorrhage) or continuous expression of fetal heart rate and crown–rump length (Continuous) using 25% test sets
| ROC association statistics | Mann–Whitney |
|---|---|
| ROC model | Area (95% CI) |
| Base | 0.630 (0.471, 0.788)a |
| Primary | 0.775 (0.653, 0.898) |
| Age | 0.785 (0.662, 0.909) |
| Blood pressure | 0.804 (0.702, 0.905) |
| YSD + hemorrhage | 0.774 (0.646, 0.902) |
| Continuous | 0.794 (0.665, 0.923) |
| ROC contrast estimation | Difference in area (95% CI) |
| Primary − Base | 0.146 (0.00206, 0.294) |
| Age − Primary | 0.00993 (−0.0145, 0.0346) |
| Blood pressure − Primary | 0.0282 (−0.0321, 0.0885) |
| YSD + hemorrhage − Primary | −0.00145 (−0.271, 0.0242) |
| Continuous − Primary | 0.0185 (−0.0419, 0.0798) |
The area under the receiver operating characterisic curve (AUC) for each model is provided in ROC association statistics. Differences in area between models are provided in ROC contrast estimation. The increase in area under the AUC after adding fetal heart rate and crown–rump length is substantively improved, although not significantly, as shown by the ROC contrast estimation of a confidence interval (CI) that includes 0. There are no statistically significant or substantive differences in AUC with either addition of maternal age or blood pressure, or yolk sac diameter and subchorionic hemorrhage, or by continuous expression of fetal heart rate and crown–rump length.
All 10 imputations expectedly had identical estimates of 0.630 (0.471, 0.788), due to the absence of missingness for gestational age and clinical loss.
Comment
Principal findings
We identified gestational age specific cut-points for routinely measured ultrasound features that were most predictive of risk of pregnancy loss: low fetal heart rate ≤122, 123, and 158 bpm; small crown–rump length ≤6.0, 8.5, and 10.9 mm for gestational weeks 6, 7, and 8, respectively. Nonviability cut-points for low fetal heart rate and crown–rump length are lower than those that maximize predictive value. This study improves upon the prior allocation of cut-points for routinely measured features by basing them on clinical data for multiple sonographic characteristics, making them specific to gestational week, and determining them using methods that optimize prediction. These findings have implications for patient counseling, given that when both features are present, physicians could estimate a 15% increase in risk of clinical loss.
Comparison with other studies
While prediction of subsequent pregnancy loss could be improved through joint examination of multiple sonographic features, most existing studies investigated individual components of the first trimester ultrasound, including subchorionic hemorrhage,1–4 fetal heart rate,5–8 crown–rump length,9–11 and yolk sac diameter.12–15 The current findings differ from previous studies of subchorionic hemorrhage, which reported elevated risk of pregnancy loss with hemorrhage overall1,2 and with increasing size of hemorrhage,3,4 although subchorionic hematoma has not been shown to be an independent risk factor for pregnancy loss once vaginal bleeding has been taken into account.31 We were unable to further examine hemorrhage by size because of few observations with measured hemorrhage size (n = 48). Current findings also differ from studies of yolk sac diameter in which small,32 large,12,16,32 or extreme13–15,18 yolk sac measurements have previously been associated with increased risk of loss. In the current study, prediction was not improved beyond information provided by gestational age, low fetal heart rate, and small crown–rump length. This difference may be due to prior studies not having taken fetal heart rate and crown–rump length into account, which were more highly predictive of clinical loss in the current study. As expected, losses have been associated with lower heart rates5–8,16,18,32,33 and smaller crown–rump lengths,9–11,16,17,22,33 although when used, cut-points for these parameters are not based on prediction (fetal heart rate5,7,10,33 and crown–rump length11,33); rather, when data-defined, cut-points have been based on the distribution of the population examined.10,33 One study used data to define gestational week–specific cut-points8 and reported fetal heart rate cut-points of ≤115 bpm for 6–7 weeks and ≤145 bpm for 7–8 weeks for elevated risk of pregnancy loss (vs ≤122 for 6 to <7 weeks and 123 for 7 to <8 weeks in our study). Our findings expand on the prior data by using a large sample to define cut-points based on specific gestational week. We also combine multiple clinical measurements to improve pregnancy loss prediction, which has been performed on a limited basis.17,19–21 We are not aware of any prior studies that used clinical data to first identify features and gestational age specific cut-points most predictive of pregnancy loss.
Maternal age is acknowledged as 1 of the strongest risk factors for pregnancy loss, and was identified as independently associated with pregnancy loss in a prior ultrasound study,20 even after accounting for gestational age based on LMP, vaginal bleeding, gestational sac diameter, and presence or absence of a yolk sac. In the current study, prediction of clinical loss did not improve with addition of maternal age, after accounting for gestational age, small crown–rump length, and low fetal heart rate.
Furthermore, prior evidence is aimed toward distinguishing nonviability. Diagnosis of nonviability carries cut points for fetal heart rate (<85 bpm)5,34 and crown–rump length (≥7 mm and no heartbeat)35 that are intended to have 100% specificity. They are appropriately different from those identified in the current study that aimed to maximize prediction of subsequent loss, whereby sensitivity and specificity were considered equally important. However, the use of thresholds to determine nonviability with certainty does not allow for counseling many women who will go on to have a clinical pregnancy loss regarding their specific prognosis. In the current study, 95% of women who had neither low fetal heart rate nor small crown–rump length went on to have a live birth.
Clinical implications
Physicians use ultrasound indicators for patient counseling, and in some cases to decide upon the frequency of follow-up sonograms. These data can be used to reassure women without either low fetal heart rate or small crown–rump length, and to identify pregnant women with an elevated risk of pregnancy loss.
Strengths and limitations
This is the first study to identify routinely measured features and their gestational age–specific cut-points most predictive of subsequent risk of pregnancy loss. Because fetal heart rate and crown–rump length increase continuously with gestational age, cut-points require characterization on a gestational age–specific basis, and should be validated in other cohorts. These specified criteria should not be used to diagnose nonviability, as most pregnancies with low fetal heart rate and crown–rump length (79%) ended in live birth.
Analytic methods were used to reduce the potential impact of bias. To use all study observations and to improve efficiency, multiple imputations were used. Although rarely realized in practice, complete-case analysis assumes that the missing data are not related to missing or observed data. This assumption is relaxed in multiple imputation, which assumes that missing data can be related to observed data.
Although we used LMP dates, which may be considered unreliable in clinical obstetric populations, it is important to note that, in this setting, LMPs were determined prospectively from detailed follow-up using bleeding data from preconception daily diaries as well as fertility monitors. In addition, for trial inclusion, study participants were required to have regular menstrual cycle lengths between 21 and 42 days and to have discontinued contraception by the baseline visit. Furthermore, the use of LMP-based dating (rather than ultrasound-based gestational age dating) was necessary here to avoid the circular logic of calculating gestational week–specific cut-points by gestational ages determined by ultrasound measurements.
This study is limited by its generalizability to women who have an ultrasound between 6 weeks 0 days and 8 weeks 6 days since the beginning of their last menstrual period and have clinical confirmation with detectable fetal heart rate at this ultrasound. Furthermore, findings should be generalized only to women with a history of pregnancy loss. Although it should be noted that pregnancy loss is a relatively common event among women of reproductive age, and prior studies do not suggest that relationships between ultrasound parameters and risk of clinical loss differ by prior loss history despite a higher rate of subsequent losses among this population.36 Given the small number of study participants who experienced a clinical loss in the 25% test set, model performance is intended to be descriptive, and should not be considered a validation study. Further limitations are the relatively homogeneous, predominantly white population, the absence of data on causes of pregnancy loss such as phospholipid antibody syndrome, and insufficient data to examine genetic karyotype.
Conclusions
When both small crown–rump length and low fetal heart rate are present, physicians could estimate a 15% increase in risk of clinical loss. This study improves upon the prior allocation of cut-points for routinely measured features by basing them on clinical data for multiple sonographic characteristics, making them specific to gestational week, and determining them using methods that optimize prediction.
Supplementary Material
AJOG at a Glance.
Why was this study conducted?
Physicians use ultrasound indicators for patient counseling.
Key findings
Low fetal heart rate and small crown–rump length (≤122, 123, and 158 bpm; ≤6.0, 8.5, and 10.9 mm for gestational weeks 6, 7, and 8, respectively) were independent predictors of clinical pregnancy loss, with greatest risks observed for pregnancies having both characteristics.
What does this add to what is known?
This study improves upon prior cut-points for routinely measured characteristics by using clinical data from multiple sonographic characteristics, making them specific to gestational week, and optimizing prediction.
Acknowledgment
We thank Saima Rafique, MD (Howard University Hospital, Washington, DC) for contributing to a literature review.
Funding was provided by the Intramural Research Program of the Eunice Kennedy Shriver National Institute of Child Health and Human Development (National Institutes of Health, Bethesda, MD; contract numbers HHSN267200603423, HHSN267200603424, HHSN267200603426).
Footnotes
The results of this study were presented at the American Society for Reproductive Medicine October 9, 2018, Denver, CO.
The authors declare no conflict of interest.
References
- 1.Ball RH, Ade CM, Schoenborn JA, Crane JP. The clinical significance of ultrasonographically detected subchorionic hemorrhages. Am J Obstet Gynecol 1996;174:996–1002. [DOI] [PubMed] [Google Scholar]
- 2.Kurjak A, Schulman H, Zudenigo D, Kupesic S, Kos M, Goldenberg M. Subchorionic hematomas in early pregnancy: clinical outcome and blood flow patterns. J Maternal-Fetal Med 1996;5:41–4. [DOI] [PubMed] [Google Scholar]
- 3.Bennett GL, Bromley B, Lieberman E, Benacerraf BR. Subchorionic hemorrhage in first-trimester pregnancies: prediction of pregnancy outcome with sonography. Radiology 1996;200:803–6. [DOI] [PubMed] [Google Scholar]
- 4.Abu-Yousef MM, Bleicher J, Williamson R, Weiner C. Subchorionic hemorrhage: sonographic diagnosis and clinical significance. Am J Roentgenol 1987;149:737–40. [DOI] [PubMed] [Google Scholar]
- 5.Laboda L, Estroff J, Benacerraf B. First trimester bradycardia. A sign of impending fetal loss. J Ultrasound Med 1989;8:561–3. [DOI] [PubMed] [Google Scholar]
- 6.Rauch ER, Schattman GL, Christos PJ, Chicketano T, Rosenwaks Z. Embryonic heart rate as a predictor of first-trimester pregnancy loss in infertility patients after in vitro fertilization. Fertil Steril 2009;91:2451–4. [DOI] [PubMed] [Google Scholar]
- 7.Dede F, Ulucay U, Kose M, Dede H, Dilbaz S. Fetal loss in threatened abortion after demonstration of fetal cardiac activity in a low socioeconomic population. J Obstet Gynaecol 2010;30:622–5. [DOI] [PubMed] [Google Scholar]
- 8.Stefos TI, Lolis DE, Sotiriadis AJ, Ziakas GV. Embryonic heart rate in early pregnancy. J Clin Ultrasound 1998;26:33–6. [DOI] [PubMed] [Google Scholar]
- 9.Mukri F, Bourne T, Bottomley C, Schoeb C, Kirk E, Papageorghiou A. Evidence of early first-trimester growth restriction in pregnancies that subsequently end in miscarriage. BJOG 2008;115:1273–8. [DOI] [PubMed] [Google Scholar]
- 10.Abuelghar WM, Fathi HM, Ellaithy MI, Anwar MA. Can a smaller than expected crown–rump length reliably predict the occurrence of subsequent miscarriage in a viable first trimester pregnancy? J Obstet Gynaecol Res 2013;39:1449–55. [DOI] [PubMed] [Google Scholar]
- 11.Reljič M The significance of crown–rump length measurement for predicting adverse pregnancy outcome of threatened abortion. Ultrasound Obstet Gynecol 2001;17:510–2. [DOI] [PubMed] [Google Scholar]
- 12.Tan S, Tangal NG, Kanat-Pektas M, et al. Abnormal sonographic appearances of the yolk sac: which can be associated with adverse perinatal outcome? Medical Ultrasonog 2014;16:15–20. [DOI] [PubMed] [Google Scholar]
- 13.Lindsay DJ, Lovett IS, Lyons EA, et al. Yolk sac diameter and shape at endovaginal US: predictors of pregnancy outcome in the first trimester. Radiology 1992;183:115–8. [DOI] [PubMed] [Google Scholar]
- 14.Küçük T, Duru N, Yenen M, Dede M, Ergün A, Başer I. Yolk sac size and shape as predictors of poor pregnancy outcome. J Perinatal Medicine 1999;27:316–20. [DOI] [PubMed] [Google Scholar]
- 15.Bae S, Karnitis J. Triple ultrasound markers including fetal cardiac activity are related to miscarriage risk. Fertil Steril 2011;96:1145–8. [DOI] [PubMed] [Google Scholar]
- 16.Papaioannou GI, Syngelaki A, Maiz N, Ross JA, Nicolaides KH. Ultrasonographic prediction of early miscarriage. Hum Reprod 2011;26:1685–92. [DOI] [PubMed] [Google Scholar]
- 17.Stamatopoulos N, Lu C, Casikar I, et al. Prediction of subsequent miscarriage risk in women who present with a viable pregnancy at the first early pregnancy scan. Aust N Z J Obstet Gynaecol 2015;55:464–72. [DOI] [PubMed] [Google Scholar]
- 18.Wie JH, Choe S, Kim SJ, Shin JC, Kwon JY, Park IY. Sonographic parameters for prediction of miscarriage: role of 3-dimensional volume measurement. J Ultrasound Med 2015;34: 1777–84. [DOI] [PubMed] [Google Scholar]
- 19.Bottomley C, Van Belle V, Pexsters A, et al. A model and scoring system to predict outcome of intrauterine pregnancies of uncertain viability. Ultrasound Obstet 2011;37: 588–95. [DOI] [PubMed] [Google Scholar]
- 20.Guha S, Van Belle V, Bottomley C, et al. External validation of models and simple scoring systems to predict miscarriage in intrauterine pregnancies of uncertain viability. Hum Reprod 2013;28:2905–11. [DOI] [PubMed] [Google Scholar]
- 21.Bottomley C, Van Belle V, Kirk E, Van Huffel S, Timmerman D, Bourne T. Accurate prediction of pregnancy viability by means of a simple scoring system. Hum Reprod 2012;28: 68–76. [DOI] [PubMed] [Google Scholar]
- 22.Choong S, Rombauts L, Ugoni A, Meagher S. Ultrasound prediction of risk of spontaneous miscarriage in live embryos from assisted conceptions. Ultrasound in Obstet Gynecol 2003;22:571–7. [DOI] [PubMed] [Google Scholar]
- 23.Schisterman EF, Silver RM, Perkins NJ, et al. A randomised trial to evaluate the effects of low-dose aspirin in gestation and reproduction: design and baseline characteristics. Paediatr Perinat Epidemiol 2013;27:598–609. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Mumford SL, Silver RM, Sjaarda LA, et al. Expanded findings from a randomized controlled trial of preconception low-dose aspirin and pregnancy loss. Hum Reprod 2016;31:657–65. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Greenland S, Pearl J, Robins JM. Causal diagrams for epidemiologic research. Epidemiology 1999;10:37–48. [PubMed] [Google Scholar]
- 26.Hernan MA, Hernandez-Diaz S, Werler MM, Mitchell AA. Causal knowledge as a prerequisite for confounding evaluation: an application to birth defects epidemiology. Am J Epidemiol 2002;155:176–84. [DOI] [PubMed] [Google Scholar]
- 27.Perkins NJ, Schisterman EF. The inconsistency of “optimal” cutpoints obtained using two criteria based on the receiver operating characteristic curve. Am J Epidemiol 2006;163: 670–5. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Moradan S, Forouzeshfar M. Are abnormal yolk sac characteristics important factors in abortion rates? Int J Fertil Steril 2012;6:127. [PMC free article] [PubMed] [Google Scholar]
- 29.Svetnik V, Liaw A, Tong C, Wang T. Application of Breiman’s random forest to modeling structure-activity relationships of pharmaceutical molecules. International Workshop on Multiple Classifier Systems. New York: Springer; 2004; 2004:334–43. [Google Scholar]
- 30.Spiegelman D, Hertzmark E. Easy SAS calculations for risk or prevalence ratios and differences. Am J Epidemiol 2005;162: 199–200. [DOI] [PubMed] [Google Scholar]
- 31.Johns J, Hyett J, Jauniaux E. Obstetric outcome after threatened miscarriage with and without a hematoma on ultrasound. Obstet Gynecol 2003;102:483–7. [DOI] [PubMed] [Google Scholar]
- 32.Varelas FK, Prapas NM, Liang R-I, Prapas IM, Makedos GA. Yolk sac size and embryonic heart rate as prognostic factors of first trimester pregnancy outcome. Eur J Obstet Gynecol Reprod Biol 2008;138:10–3. [DOI] [PubMed] [Google Scholar]
- 33.Datta MR, Raut A. Efficacy of first-trimester ultrasound parameters for prediction of early spontaneous abortion. Int J Gynecol Obstetr 2017;138:325–30. [DOI] [PubMed] [Google Scholar]
- 34.Achiron R, Tadmor O, Mashiach S. Heart rate as a predictor of first-trimester spontaneous abortion after ultrasound-proven viability. Obstet Gynecol 1991;78:330–4. [PubMed] [Google Scholar]
- 35.Doubilet PM, Benson CB, Bourne T, et al. Diagnostic criteria for nonviable pregnancy early in the first trimester. N Engl J Med 2013;369: 1443–51. [DOI] [PubMed] [Google Scholar]
- 36.Magnus MC, Wilcox AJ, Morken N-H, Weinberg CR, Håberg SE. Role of maternal age and pregnancy history in risk of miscarriage: prospective register based study. BMJ 2019;364:l869. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.