A simple scoring system for the prediction of early pregnancy loss developed by following 13,977 infertile patients after in vitro fertilization

Abstract

Introduction

The first routine ultrasound scan is commonly arranged on day 28 after in vitro fertilization-embryo transfer (IVF-ET) in most reproductive centers in China to determine the location and viability of the embryo. Although the appearance of cardiac activity at first scan means a higher likelihood of ongoing pregnancy, the rate of subsequent miscarriage is between 2 and 16%[1], and there are some examples of pregnancies that were initially thought to be poorly developed but eventually developed normally[2]. Moreover, the mental and psychological pressure on women undergoing IVF-ET are significant, especially when the viability of a pregnancy is uncertain. Under these circumstances, the physician has an important role in rapidly and accurately predicting the pregnancy outcome in an evidence-based and professional manner, as this would be helpful for further consultations and in determining subsequent management.

Individual maternal factors, such as a high maternal age (MA)[[3], and abnormal ultrasound parameters such as embryonic bradycardia and excessively large or small yolk sac diameter (YSD) have been shown to be associated with pregnancy failures in previous studies[4][5]. Some previous mathematical models involve combinations of individual risk factors, with reasonable performance[2][6][7][8][9]. Although these models have achieved good prediction results, they also have some shortcomings, for example, the sample size of most studies is not large enough, some study groups are mixed (mixed natural and assisted pregnancies), the gestational age (GA) is inconsistent and some studies do not give a specific probability of ongoing pregnancy or miscarriage. In addition, many patients will return to the local hospital after pregnancy. What we want to build is a model by using ultrasound data obtained from real-time measurement and simple clinical indicators such as MA, so that we can quickly predict and judge pregnancy in different hospitals without knowing many assisted fertility indicators.

Therefore, the aim of this study was to investigate a convenient and accurate simple scoring system for the prediction of early pregnancy loss (EPL) based on simple demographics and the first routine ultrasound scan performed on days 27–29 after ET. 

Methods

Participants

This retrospective study was conducted at the Reproductive and Genetic Hospital of CITIC-Xiangya (Changsha, Hunan, People's Republic of China). The institutional review board approved this study (date of approval: 26 July 2019; reference number: LL-SC-2019-015; Changsha, China).

The infertile patients included in this study underwent IVF treatment between June 2016 and December 2017. One to 2 embryos with good quality were transferred at the day-3 or day-5 stage. The embryo morphology was scored according to the criteria by Hardarson et al.[10]. The first routine transvaginal sonography (TVS) scan using a 5-9 MHZ probe (GE VOLUSON 730 or E8, General Electric) was arranged on days 27–29 to observe the number of embryos and their locations and viability. Only intrauterine singleton pregnancies were included. The measurements referred to the ISUOG practice guidelines [11] and conformed to uniform standards: gestational sac diameter (GSD) was calculated as the mean value of 3 perpendicular diameters with the calipers placed at the inner edges of the trophoblast; YSD was calculated as the average of 3 perpendicular diameters with the calipers placed at the center of the yolk sac (YS) wall; embryonic length (EL) was measured as the greatest length of the embryo in the anterior to posterior dimension; and embryonic heart rate (EHR) was calculated from frozen M-mode images with electronic calipers by measuring the distance between two heart waves. The presence of intrauterine hematoma (IUH), a hypoechoic or anechoic crescent-shaped area between the chorionic membrane and the myometrium, was also noted. Data on clinical characteristics including day-14 (blastocysts on day 12), serum β-human chorionic gonadotropin (HCG) levels, MA, duration of infertility, infertility type and endometrial (EM) thickness on transfer day were also collected. The first-trimester pregnancy outcomes of these participants were noted at 12 weeks of gestation. Women with a continuing pregnancy for >12 weeks of gestation were classified as ongoing pregnancy, and women with spontaneous miscarriage before or at 12 weeks of gestation were classified as EPL.

Statistical analysis

Categorical variables are described using frequencies, and continuous variables are described using means and standard deviations. Pearson’s chi-squared test, Fisher’s exact test, T-test or the Wilcoxon rank sum test was used to compare categorical or continuous variables between women with ongoing pregnancies and women with EPL. The cases from the first year (year 2016) were used to generate the training set, and the cases from the second year (year 2017) were used as the verification set. Binary logistic regression (LR) models were used in the training set to identify the probable predictive factors of EPL. Points associated with each category of each risk factor were computed, and the risks associated with point totals were determined according to the “The Framingham Study risk score system”[12]. Based on the area under curve (AUC) and the clinical value, we evaluated the cutoff of the prediction model which had a relatively high risk of miscarriage and high prediction accuracy. Then, we used the verification set to evaluate and verify the simple scoring system. All statistical analyses were performed using SPSS version 24.0 software (SPSS, Inc.). The results of each test were considered significant when the two-sided P value did not exceed 0.05, except where otherwise specified.

Results

From June 2016 to December 2017, 23,929 infertile patients obtained clinical pregnancies via IVF-ET in our hospital. 14,118 women were identified with an intrauterine singleton pregnancy during this period and among them, 141 cases were lost to follow-up. Finally, a cohort of 13,977 women were enrolled, including 12,051 cases with ongoing pregnancies and 1,926 cases with EPLs.

Comparisons of women with ongoing pregnancies and EPLs

Compared with women with ongoing pregnancies, the MA, body mass index, infertility duration and transfer cycle were significantly higher, and the day-14 hCG and the EM thickness on transfer day were significantly lower in women with EPLs (p < 0.001). The infertility type, cause of infertility, and insemination methods were also significantly different between women with ongoing pregnancies and EPLs (p < 0.05). The number of embryos transferred was not significantly different (p = 0.44).

Based on the TVS measurements, the GSD (18.5±3.6 vs. 13.2±4.8 mm), EL (3.5±0.9 vs. 1.2±1.6 mm), YSD (3.6±0.4 vs. 2.6±1.5 mm) and EHR (114.5±2.2 vs. 42.4± 53.5 bpm) were significantly greater in women with ongoing pregnancies than those with EPLs (p < 0.001). The incidence of IUH (16.0% vs. 18.8%, P = 0.002) was also markedly higher in women with EPLs (Table 1).

Binary LR analysis

After stepwise screening, MA (p = 0.0001, OR 1.096, 95% confidence interval (CI) 1.073-1.119), GSD (p = 0.0001, OR 0.892, 95% CI 0.864-0.921), EL (p = 0.0020, OR 0.783, 95% CI 0.672-0.913), YSD (p = 0.0600, OR 0.853, 95% CI 0.723-1.007), EHR (p = 0.0001, OR 0.966, 95% CI 0.961-0.971) and EM on transfer day (p = 0.0030, OR 0.929, 95% CI 0.884-0.976) were found to be predictive factors of EPL and finally entered the scoring system.

The scoring system

The points associated with each of the categories of the risk factors were calculated and are shown in Table 2. Table 3 shows the scoring system derived from the binary LR model and the predictive value of each total point. The AUC of this scoring system was 0.884 (95% CI 0.870-0.899) in the training set. The score totals ranged from points -8 to 14. Points 5 and 6 offered the highest predictive accuracy (94.01% and 93.95%, respectively), and the predicted risks of miscarriage were 30.03% and 39.55%, respectively. Considering that IVF women were anxious, we chose point 5 corresponding to 30.03% risk of miscarriage as the cutoff value, with a sensitivity of 62.84%, specificity of 98.79%, positive predictive value (PPV) of 88.87% and negative predictive value (NPV) of 94.54%, for the prediction of EPL in the training set (Table 3).

In the verification set, the AUC of the scoring system was 0.890 (95% CI 0.878-0.903). Point 5 had the highest diagnostic accuracy, with 93.91% of the samples correctly predicted. The sensitivity, specificity, PPV and NPV were 64.69%, 98.78%, 89.87% and 93.62%, respectively (Table 4).

Discussion

In this study, we collected the demographic and ultrasound findings for days 27-29 after ET and constructed a simple scoring system for the prediction of EPL, with AUCs of 0.884 and 0.878 in the training set and verification set, respectively. Point 5 had the highest predictive accuracy and was recommended as the cutoff value for clinical practice.

The process of achieving pregnancy through IVF is usually very difficult for an infertile woman. Thus, when the woman becomes pregnant by IVF, she is usually very anxious about the development of the embryo, even after the detection of cardiac activity but especially if an empty gestational sac or only a YS are detected. Generally, for patients who have well-developed embryos at the first routine TVS examination on days 27-29 after ET, the next ultrasound scan will usually be scheduled on day 45 after ET; however, for patients with embryos of uncertain viability, the recommendation is usually to have another TVS scan 7-10 days after the first scan to assess the development and viability of the embryo.

If the pregnancy outcome can be predicted in advance, the anxiety of the pregnant woman can be greatly relieved. Many models have been constructed to predict pregnancy outcomes effectively[13][14][15]. Among them, LR analysis has been the most commonly used[15][16]. While what we want to construct is a simple and practical scoring system that is similar to the “Apgar score system” using simple demographic and ultrasound findings on days 27–29 after ET, as this would be easy for clinical application, especially for patients without sufficient assisted reproduction indicators when they return home after pregnancy, and this is why we performed this research. Through this system, we can inform patients with a specific probability of miscarriage after the first routine TVS examination, which may help reduce the patient’s anxiety and psychological burden and provide guidance for follow-up decisions.

The findings of this study agree with our current knowledge that miscarriage is more likely with increasing MA[17], low hCG level[18], low EM thickness on transfer day[19] and the presence of IUH[20] and that miscarriage is less likely after the visualization of embryonic cardiac activity[15]. Accordingly, in our scoring system, a greater MA was associated with a higher point, while greater GSD, YSD, EL and EHR were all related to lower points. However, both small and large YSDs corresponded to higher points. A higher point means a greater contribution to miscarriage. Thus, the possibility of miscarriage increased with the total score. Point 5 offered the highest predictive accuracy in both the training set and verification set, and the corresponding miscarriage risk exceeded 30%, which was not a low risk for anxious IVF patients. Therefore, we recommend using point 5 as a clinical threshold for warning patients of this risk  counselled more about miscarriage. In practical applications, we hope that the false positive rate (FPR) will be as low as possible, as this may lead to unnecessary medical treatment. A low FPR requires high specificity. In this system, the specificity of points -8 to 0 was not satisfactory. Thus, this is a scoring system for determining the probability of miscarriage that could allow doctors and patients to know the risks in advance, but because miscarriage cannot be prevented, the outcome cannot be changed by this scoring system. The advantage of this scoring system is that it is easily transferable to clinical use, where both maternal and ultrasound variables are easily available and the calculation is quite easy. For example, in a woman whose MA = 38 yr and who has an EM = 8 mm on transfer day, a GSD = 10 mm, a YSD =2.2 mm, and an EHR = 88 bpm by TVS on day 28 after ET, the miscarriage score is 9, and the estimated risk of miscarriage is 70.67%. Even though the results indicate a strong likelihood of EPL, there is still a significant risk that this is a false positive finding. Thus, it is imperative to repeat the ultrasound scan for this patient 7-10 days later to confirm the viability of the embryo before any medical interventions. In contrast, if a patient does not present with any symptoms and has a low total score (for example, point 0), no TVS scan is needed until day 45 after ET.

A previous study[21] constructed a similar scoring system to predict pregnancy viability and achieved an AUC of more than 0.90. However, this study focused on natural conceptions, and the ultrasound variables and clinical characteristics collected in that study were collected within a time period of GA <84 days. The measurement methods may also differ from ours. However, in our study, all ultrasound parameters were collected at the same time on days 27-29 after ET, and the specific EHR value was included, not just the presence or absence of embryonic cardiac activity. IUH was unexpectedly not included in the final system. We speculate that this may be because IUH occurs more frequently in pregnancies after IVF than in spontaneous conceptions[22], not only in women with EPLs but also with ongoing pregnancies.

We note that the sensitivity of this simple model was not ideal, which might indicate that for the IVF population, only including ultrasonic measurements and simple clinical indicators had limited predictive efficacy, and in further study, relevant indicators for assisted pregnancy should be added to improve the predictive efficacy. Another limitation was that since there was no information on bleeding, abdominal pain and smoking history in our hospital's electronic medical record system during the study period, these indicators were not included in our system, but might further improve the predictive performance if included. Since this system was derived from the IVF population, its application in the general population had yet to be validated.

In conclusion, we have developed a simple and practical scoring system that  provides a probability for EPL based on the simple demographic and ultrasound findings obtained on days 27–29 after ET. This system is easy and simple for clinical use. Point 5 is recommended as the clinical threshold for warning patients of an EPL risk. When the predictive result is a high risk of EPL, repeated scans are recommended 7-10 days later to confirm the viability of the embryo. When the predictive result is a low risk of EPL and patients have no symptoms, the next examination can be performed on day 45 after ET.

Abbreviations

EPL, early pregnancy loss; IVF-ET, in vitro fertilization-embryo transfer; CI,  confidence interval; PPV, positive predictive value; NPV, negative predictive value; MA, maternal age; YSD, yolk sac diameter; GA, gestational age; TVS, transvaginal sonography; GSD, gestational sac diameter; YS, yolk sac; EL, embryonic length; EHR, embryonic heart rate; IUH, intrauterine hematoma; HCG, human chorionic gonadotropin; 

EM, endometrium; LR, logistic regression; AUC, area under curve; FPR, false positive rate.

Declarations

Ethical Approval and consent to participate

This study was approved by the Ethics Committee of the Reproductive and Genetic Hospital of CITIC-Xiangya (date of approval: 26 July 2019; reference number: LL-SC-2019-015; Changsha, China). Informed consent was waived due to the retrospective nature.

Consent for publication

Not applicable.

Availability of data and materials

The data analysed during this study are included in the tables. The datasets used during the current study are available from the corresponding author on reasonable request.

Competing Interests

The authors have no relevant financial or non-financial interests to disclose.

Funding 

This work was funded by the Natural Science Foundation of Hunan Province (No. 20201955), the Research Grant of CITIC-Xiangya (No. YNXM-202220) and the Science and Technology Project of the Health, Family Planning Commission of Hunan Province (No. C20180289).

Authors’ contributions

XL and YO designed the study and all authors contributed to the study’s conception and design. Material preparation and data collection were performed by FG and YO. Data analysis were performed by SZ and YP. The first draft of the manuscript was written by YO. All authors read and approved the final manuscript.

Acknowledgements

The authors thank Qingqing Wu and Mingxiang Zheng for their help in data collection.

References

1. Rull K, Laan M (2005) Expression of β-subunit of HCG genes during normal and failed pregnancy. Hum Reprod. https://doi.org/10.1093/humrep/dei261
2. Yi Y, Li X, Ouyang Y, et al (2016) Discriminant analysis forecasting model of first trimester pregnancy outcomes developed by following 9,963 infertile patients after in vitro fertilization Abstract has been accepted for oral presentation (O-222) at the 71th America Society of Reproductive. Fertil Steril 105:. https://doi.org/10.1016/j.fertnstert.2016.01.033
3. Evaluation and treatment of recurrent pregnancy loss: A committee opinion. Fertil Steril. https://doi.org/10.1016/j.fertnstert.2012.06.048
4. Morin L, Cargill YM, Glanc P (2016) Ultrasound Evaluation of First Trimester Complications of Pregnancy. J Obstet Gynaecol Canada. https://doi.org/10.1016/j.jogc.2016.06.001
5. Spencer K, Cowans NJ, Avgidou K, Nicolaides KH (2006) First-trimester ultrasound and biochemical markers of aneuploidy and the prediction of impending fetal death. Ultrasound Obstet Gynecol. https://doi.org/10.1002/uog.3809
6. Stamatopoulos N, Lu C, Casikar I, et al (2015) Prediction of subsequent miscarriage risk in women who present with a viable pregnancy at the first early pregnancy scan. Aust New Zeal J Obstet Gynaecol. https://doi.org/10.1111/ajo.12395
7. Casikar I, Lu C, Reid S, Condous G (2013) Prediction of successful expectant management of first trimester miscarriage: Development and validation of a new mathematical model. Aust New Zeal J Obstet Gynaecol. https://doi.org/10.1111/ajo.12053
8. Shaamash AH, Aly HA, Abdel-Aleem M, Akhnowkh SN (2020) Clinical and Ultrasound Evaluation of Early Threatened Miscarriage to Predict Pregnancy Continuation up to 28 Weeks : A Prospective Cohort Study. J Ultrasound Med. https://doi.org/10.1002/jum.15282
9. Wang XH, Wang HJ, Deng XH, et al (2019) Predictive value of ultrasound-related scoring system on embryo development in early pregnancy after IVF/ICSI: An observation of embryonic quality. Taiwan J Obstet Gynecol. https://doi.org/10.1016/j.tjog.2019.05.013
10. Yi Y, Lu G, Ouyang Y, et al (2016) A logistic model to predict early pregnancy loss following in vitro fertilization based on 2601 infertility patients. Reprod Biol Endocrinol 14:. https://doi.org/10.1186/s12958-016-0147-z
11. Salomon LJ, Alfirevic Z, Bilardo CM, et al (2013) Erratum: ISUOG practice guidelines: Performance of first-trimester fetal ultrasound scan. Ultrasound Obstet Gynecol 2013; 41: 102-113. DOI: 10.1002/uog.12342. Ultrasound Obstet. Gynecol.
12. Hardarson T, Hanson C, Sjögren a, Lundin K (2001) Human embryos with unevenly sized blastomeres have lower pregnancy and implantation rates: indications for aneuploidy and multinucleation. Hum Reprod 16:313–318. https://doi.org/10.1093/humrep/16.2.313
13. Sullivan LM, Massaro JM, D’Agostino RB (2004) Presentation of multivariate data for clinical use: The Framingham Study risk score functions. Stat Med. https://doi.org/10.1002/sim.1742
14. Bottomley C, Van Belle V, Pexsters A, et al (2011) A model and scoring system to predict outcome of intrauterine pregnancies of uncertain viability. Ultrasound Obstet Gynecol. https://doi.org/10.1002/uog.9007
15. Guha S, Van Belle V, Bottomley C, et al (2013) External validation of models and simple scoring systems to predict miscarriage in intrauterine pregnancies of uncertain viability. Hum Reprod. https://doi.org/10.1093/humrep/det342
16. Lautmann K, Cordina M, Elson J, et al (2011) Clinical use of a model to predict the viability of early intrauterine pregnancies when no embryo is visible on ultrasound. Hum Reprod. https://doi.org/10.1093/humrep/der287
17. Khalil A, Syngelaki A, Maiz N, et al (2013) Maternal age and adverse pregnancy outcome: A cohort study. Ultrasound Obstet Gynecol. https://doi.org/10.1002/uog.12494
18. Pillai RN, Konje JC, Tincello DG, Potdar N (2016) Role of serum biomarkers in the prediction of outcome in women with threatened miscarriage: A systematic review and diagnostic accuracy meta-analysis. Hum. Reprod. Update
19. Kasius A, Smit JG, Torrance HL, et al (2014) Endometrial thickness and pregnancy rates after IVF: A systematic review and meta-analysis. Hum Reprod Update. https://doi.org/10.1093/humupd/dmu011
20. Tuuli MG, Norman SM, Odibo AO, et al (2011) Perinatal outcomes in women with subchorionic hematoma: A systematic review and meta-analysis. Obstet. Gynecol.
21. Bottomley C, Van Belle V, Kirk E, et al (2013) Accurate prediction of pregnancy viability by means of a simple scoring system. Hum Reprod. https://doi.org/10.1093/humrep/des352
22. Asato K, Mekaru K, Heshiki C, et al (2014) Subchorionic hematoma occurs more frequently in in vitro fertilization pregnancy. Eur J Obstet Gynecol Reprod Biol. https://doi.org/10.1016/j.ejogrb.2014.07.014

Tables

Table 1  Comparisons of the parameters between 2 groups

Table 2  Points associated with each of the categories of the risk factors

Points=βi (Wij-WiREF)/B    B=5*0.0920

Note: MA = maternal age; GSD = gestational sac diameter；CRL= crown-rump length; EHR = embryonic heart rate;YSD = yolk sac diameter; EM = endometrium.

Table 3  Scoring system derived from the binary LR model.

Note: LR = logistic regression; PPV = positive predictive value; 

NPV= negative predictive value; ACC = accuracy. 

Table 4 The classification of the results obtained using point 5 as the cutoff value 

Note: EPL: early pregnancy loss