Second-trimester pregnancy loss after in vitro fertilization: risk factors and risk reduction

Abstract

Objective

To study changes in the rate and potential causes of second-trimester loss (STL) after in vitro fertilization (IVF) after 2 decades marked by the gradual adoption of elective single embryo transfer.

Design

Retrospective cohort study

Subjects

All patients who conceived via IVF between 2015 and 2023 at an academic fertility center. Patients who underwent therapeutic termination were excluded.

Exposure

In vitro fertilization.

Main Outcome Measures

The primary outcome was the rate of STL after IVF. Secondary associated factors assessed included multiple pregnancy, race, uterine defects and tubal factor infertility. Secondary factors were compared with all live births (n = 853), uterine factors were compared with a subset with equivalent uterine structural evaluation. (n = 595). Outcomes were compared with our prior data from 1999–2002.

Results

A total of 971 patients with clinical pregnancies were included; 10.2% resulted in a first-trimester loss and a total of 14 patients in our group had a STL. These patients were compared with a control group of 853 patients. Adenomyosis was found to be associated with an increased risk of STL, 35.7% (5/14), compared with 6.4% (38/595) of patients in our control group. Additionally, intrauterine scarring was associated with a higher rate of STL at 35.7% (5/14) vs. 12.6% (75/595). Tuboperitoneal disease was positively associated with STL with 42.9% (6/14), compared with 13.6% (116/853) among the control group. Additionally, patients who identified as Black were associated with an 8.0% (6/75) risk of STL and patients who identified as Hispanic were associated with increased risk of STL at 4.1% (5/122). In univariate analysis, a significant association was found between patients who identified as Black and STL, with 20 times higher odds of STL than patients who identified as White. A significant association was also found between patients who identified as Hispanic or Latino and STL, with 7.90 times higher odds of STL than patients who identified as White. When adjusting for race and ethnicity, parity, and preimplantation genetic testing for aneuploidy (PGT-A) usage in the multivariate model, uterine factor was found to have a significant association with STL, with 32.95 times higher odds of STL in patients with uterine factor.

Conclusion

Factors associated with STL include Black race, Hispanic ethnicity, adenomyosis, tubal factor infertility, and intrauterine scarring, with the racial and ethnic disparities in STL largely unchanged over 2 decades. Further study is needed to tailor risk prevention for reducing STL after IVF in at-risk groups.

Second-trimester pregnancy loss (STL) is defined as pregnancy loss between 13 and 27 weeks of gestation and is estimated to complicate 0.1%–1% of documented pregnancies (1). Multiple studies have been performed to assess the rates and risk factors associated with STL in the general population. Previous studies have found increased risk of STL in women with increasing age, higher body mass index (BMI) and tobacco use (2). Chorioamnionitis was also found to be associated with increased risk of STL (3). Additionally, maternal genetic mutations, such as a prothrombin gene mutation and MTHFD1 gene polymorphisms, have been shown to be more prevalent in women with STL (4, 5).

There is very limited data evaluating the rate of STL and associated risk factors in patients who conceive via in vitro fertilization (IVF). The markedly elevated rate of preterm labor, however, is well established in IVF, and it shares many of the same risk factors found in STL. (6, 7, 8, 9). One study found that compared with White women, Black women experienced an increased odds for STL (9). Furthermore, the presence of uterine factors, specifically leiomyomas, was found to be associated with an increased risk of STL. Among the most common causes of STL were cervical incompetence and preterm premature rupture of membranes (PPROM). Additionally, of the cycles that resulted in an STL 48.3% were patients with multigestation (9).

The IVF practice has shifted toward the transfer of a single embryo. The American Society for Reproductive Medicine released an updated guideline in 2021, encouraging elective single embryo transfer (eSET) to reduce the risks associated with multigestational pregnancies (10). In 2003, we published our findings evaluating the risk factors associated with STL after IVF in patients who conceived between 1999 and 2002 (8). We found that Black and Hispanic patients, multiple gestation, and tubal infertility were associated with an increased risk of STL. Given the shift from multiembryo transfer to eSET, we sought updated data on the factors underlying STL in IVF patients.

This study aimed to assess risk factors associated with STL in IVF patients as an update to our prior data from 1999–2002. We hypothesized that multiple gestation would be markedly decreased, resulting in reduced STL, yet other risk factors associated with IVF may persist.

Materials and methods

We conducted a single-institution retrospective analysis with institutional review board approval. All women aged 18–42 years who underwent an IVF cycle with a fresh or frozen embryo transfer that resulted in a clinical pregnancy between 2015 and 2023 were included. Patients who underwent therapeutic termination because of fetal anomalies were excluded. A total of 5 patients had therapeutic terminations, with 2 because of aneuploidy and 3 because of euploidic anomalies and were therefore excluded from both the STL and control group. Patients who had a STL were then compared with a control group consisting of all patients during the same time frame who had a live birth. To analyze intrauterine pathology, the STL group was compared with a smaller subset of patients from the control group who had confirmed intrauterine imaging was performed. Furthermore, some patients in the control group were missing demographic variables, therefore were excluded from the denominator in these cases. Additionally, the STL group in our current cohort was then compared with the STL in our prior data set from 1999–2002 to provide an update to our prior study.

Patients underwent either an agonist suppression (oral contraceptive pill [OCP] lupron overlap), gonadotropin-releasing hormone antagonist or a gonadotropin-releasing hormone agonist flare protocol. Patients undergoing agonist suppression were administered at least 7 days of OCPs, followed by 10 units of leuprolide acetate subcutaneously, in combination with the OCP for 4 days. After this suppression, OCPs were discontinued, and withdrawal bleeding occurred. Once menses stopped, gonadotropin stimulation was started. Protocol and dosing of gonadotropins were based on doctor's preference and the patient’s prior IVF cycles.

Exogenous gonadotropins were administered daily until human chorionic gonadotropin (hCG) administration. Serial estradiol levels and transvaginal ultrasonography were used to monitor ovarian follicular development and determine the daily dose of gonadotropins, recombinant follicle-stimulating hormone and/or human menopausal gonadotropins. Controlled ovarian hyperstimulation and monitoring of ovarian response were continued until a dominant follicle reached a diameter of 18–20 mm. Human chorionic gonadotropin was administered in a dose of 2,500–10,000 IU based on the patient’s risk for ovarian hyperstimulation syndrome and weight. Thirty-six hours after hCG administration, egg retrieval was performed using transvaginal ultrasound guidance. Depending on the male factor, insemination was performed on an individual basis with either standard insemination or intracytoplasmic sperm injection. Fertilization was confirmed 20 hours later by the presence of 2 pronuclei. Embryos were kept in culture until day 3–4, then underwent laser-assisted hatching. Patients who underwent a fresh embryo transfer had it performed on day 5. For patients undergoing a subsequent frozen transfer, some or all day 5–7 blastocysts underwent trophectoderm biopsy, and all were subsequently frozen. Biopsy samples were sent to CooperGenomics (CooperSurgical Fertility Solutions, Livingston, NJ) for preimplantation genetic testing for aneuploidy (PGT-A) testing, which reported back next-generation sequencing data.

For all patients undergoing embryo transfer, luteal phase supplementation with intramuscular or vaginal progesterone was provided. Serum β-hCG levels were tested on days 9–12 after embryo transfer, and if positive, transvaginal sonography was performed on day 19 to confirm the presence of a gestational sac. Only pregnancies that resulted in sonographically confirmed gestational sacs were considered clinical pregnancies and therefore included in this study. Biochemical pregnancies were counted as IVF failure and not included in this study.

The primary outcome was the rate of STL after IVF. Secondary associated factors assessed included multiple pregnancy, race, ethnicity, parity, PGT-A usage, and infertility diagnoses. Race and ethnicity were reported by the patient at the initial fertility consultation as part of their standard intake.

All data were analyzed with Systat 13 software (Systat Software Inc., San Jose, CA). Statistical significance was set at P<.05. Continuous data were compared using the t test and Wilcoxon signed rank test when data were not normally distributed. Categorical data were compared using χ² or Fisher’s exact test for small totals less than 10. Logistic regression was performed to evaluate the association between STL and potential predictors, including race and ethnicity, parity, PGT-A usage, and infertility diagnoses. First, univariate models were fitted for each predictor, and crude odds ratios (ORs), 95% confidence intervals (CIs) and P values were reported. A multivariate model including all predictors was constructed, from which adjusted ORs, 95% CIs, and adjusted P values were reported. For the categorical predictors, reference categories were specified as follows: White race/ethnicity was used as the reference for race and ethnicity, and unexplained infertility was used as the reference for infertility diagnoses.

Results

Baseline characteristics

In our current data, a total of 971 patients had a clinical pregnancy after IVF and were included for analysis. Of these patients, 99 of 971 (10.2%) had a pregnancy that resulted in a first-trimester loss. In our current cohort, 14 patients had a pregnancy loss in the second trimester, 5 additional patients had therapeutic terminations for aneuploidies and were therefore excluded from analysis. Fourteen patients were included in the STL group for analysis and compared with the control group of 853 patients. There were no differences in the age between the STL and control group of 36.2 ± 4.5 and 35.7 ± 4.1, respectively, as shown in Table 1 (P=.45). There were differences in the proportion of patients who identified as each race in the STL and control group. There were no differences in BMI, gravida, gestation, PGT-A use, or infertility diagnosis overall between the STL and control groups. Regarding parity, there was a higher percentage of patients who were parous in the control group than the STL group, as shown in Table 1.

Table 1.

Patient baseline characteristics.

Parameter	STL group (N = 14) Mean (SD) (min, max)	Control group (N = 853) Mean (SD) (min, max)	P value
Age, y	36.2 (4.5) (25, 42)	35.7 (4.1) (25, 50)	.45
Race and ethnicity, n (%)			.02a
Asian	7.1% (1/14)	5.0% (43/853)
Black	42.9% (6/14)	8.0% (69/853)
Hispanic	35.7% (5/14)	13.7% (117/853)
White	14.3% (2/14)	58.3% (498/853)
Unknown	0% (0/14)	14.7% (126/853)
BMI	28.9 (4.5) (20, 37)	27.1 (5.8) (17, 52)	.1
Gravida, n (%)	10% (1/10)	47% (251/732)	.63
Nulligravid
Parity, n (%)			<.01a
Nulliparous	64.3% (9/14)	33.5% (245/732)
Parous	35.7% (5/14)	66.5% (487/732)
Gestation			.99
Singleton	92.9% (13/14)	91.9% (673/732)
Multigestation	7.1% (1/14)	8.1% (59/732)
PGT-A use	64% (9/14)	72% (527/732)	.98
Infertility diagnosis			.75
Male factor	14.2% (2/14)	35% (296/853)
Endometriosis	14.2% (2/14)	9.4% (80/853)
PCOS	0% (0/14)	0.8% (7/853)
Diminished ovarian reserve	0% (0/14)	14.4% (123/853)
Tubal	42.9% (6/14)	13.6% (116/853)
Uterine	64.3% (9/14)	29.3% (25/853)
Unexplained	14.2% (2/14)	12.7% (108/853)
Other	29% (4/14)	11.3% (96/853)

BMI = body mass index, N = total population size, PCOS = polycystic ovary syndrome, PGT-A = preimplantation genetic testing for aneuploidy, SD = standard deviation, STL = second-trimester loss.

^aStatistically significant.

Assessing causes of STLs

In our current cohort, of the 14 patients who had a pregnancy loss in the second trimester, 13 were because of PPROM and cervical incompetence (1.5%), and one was because of fetal demise (0.1%).

When assessing intrauterine pathology, neither fibroids nor uterine anomalies were associated with STL. To analyze intrauterine pathology, the STL group was compared with a smaller subset of patients from the control group who had intrauterine imaging performed. For the 14 patients with STL, 14.3% (2/14) had uterine fibroids compared with 7.2% (43/595) in the control group (P>.05). In the STL group, 14.3% (2/14) had uterine anomalies compared with 23.9% (142/595) in the control group (P>.05). However, adenomyosis was associated with an increased risk in 35.7% (5/14) of patients with a STL having adenomyosis compared with 6.4% (38/595) of patients in our control (P<.001). Additionally, intrauterine scarring was associated with a higher rate of STL at 35.7% (5/14) compared with our control group with 12.6% (75/595, P =.029).

Tuboperitoneal disease was positively associated with STL, with 42.9% (6/14) of our STL patients with tubal disease compared with 13.6% (116/853) among the control group (P =.0018).

Assessing race and ethnicity as factors in STL

Race and ethnicity were assessed within the context of our data, which demonstrate differences in the proportion of people of each race within each temporal cohort. For our patients who identified as Black, race was associated with an 8.0% (6/75) risk of STL (P<.0001). For ethnicity, patients who identified as Hispanic were associated with increased risk of STL at 4.1% (5/122) (P<.0001). There was no increased rate of STL in patients who identified as White (0.4%, 2/500) or Asian (2.2%, 1/44).

Logistic regression: univariate and summary of multivariate analysis of potential confounders

To assess for possible confounders, logistic regression was performed as shown in Table 2. In univariate analysis, a significant association was found between patients who identified as Black and STL, with 20 times higher odds of STL than patients who identified as White. A significant association was also found between patients who identified as Hispanic or Latino and STL, with 7.90 times higher odds of STL than patients who identified as White. Patients who identified as Asian and those with unknown race and ethnicity were not found to have a significant association with STL when compared with patients who identified as White. When adjusting for parity, PGT-A usage, and infertility diagnoses in multivariate analysis, patients who identified as Black and Hispanic or Latino were still found to have a significant association with STL compared with patients who identified as White.

Table 2.

Logistic regression: univariate and multivariate analysis of potential confounders.

Predictor	STL group (N = 14) Mean (SD)/n(%)	Control group (N = 732) Mean (SD)/n(%)	Univariate model OR (95% CI)	P value	Multivariate model adjusted OR (95% CI)	Adjusted P value
Race: Asian	1 (7.1%)	43 (5.0%)	6.54 × 10–7 (0–1.01 × 1046)	.99	2.16 × 10–7 (0–5.54 × 1093)	.99
Race: Black	6 (42.9%)	69 (8.0%)	20.21 (4.54–140.11)	<.001	17.96 (2.85–174.91)	.0038
Ethnicity: Hispanic or Latino	5 (35.7%)	117 (13.7%)	7.90 (1.52–57.57)	.018	9.11 (1.51–86.84) (1.91–272.11) 6	.024
Race: Unknown	0 (0%)	126 (14.7%)	3.84 (0.46–32.31)	.18	5.15 (0.56–51.26)	.13
Infertility diagnosis: endometriosis	14.2% (2/14)	9.4% (80/853)	2.89 (0.47–22.44)	.25	2.86 (0.24–67.03)	.42
Infertility diagnosis: male infertility	14.2% (2/14)	35% (296/853)	0.91 (0.19–6.43)	.91	1.24 (0.18–24.71)	.85
Infertility diagnosis: tubal	42.9% (6/14)	13.6% (116/853)	1.96 (0.23–16.72)	.51	1.69 (0.14–39.59)	.69
Infertility diagnosis: uterine	64.3% (9/14)	29.3% (25/853)	5.40 (0.24–61.43)	.18	32.95 (1.02–1157.61)	.031
Infertility diagnosis: PCOS	0% (0/14)	0.8% (7/853)	1.72 × 10–7 (0–1.15 × 10159)	.99	3.61 × 10–7 (0–Inf)	.99
Infertility diagnosis: DOR	0% (0/14)	14.4% (123/853)	1.72 × 10–7 (0–2.45 × 1028)	.99	5.22 × 10–8 (0–8.84 × 1056)	.99
Infertility diagnosis: other	29% (4/14)	11.3% (96/853)	0.47 (0.022–5.01)	.54	0.71 (0.027–18.88)	.81
Parous	35.7% (5/14)	66.5% (487/732)	0.38 (0.11–1.16)	.094	0.40 (0.11–1.34)	.14
PGT-A testing	64% (9/14)	72% (527/732)	0.68 (0.23–2.28)	.51	1.17 (0.32–4.88)	.82

CI = confidence interval, DOR = diminished ovarian reserve, n = total sample size, N = total population size, OR = odds ratio, PCOS = polycystic ovary syndrome, PGT-A = preimplantation genetic testing for aneuploidy, SD = standard deviation, STL = second-trimester loss.

In univariate analysis, infertility diagnoses of endometriosis, male infertility, tubal factor, uterine factor, polycystic ovary syndrome, diminished ovarian reserve, and other diagnoses were not found to have a significant association with STL. However, when adjusting for race and ethnicity, parity, and PGT-A usage in the multivariate model, the uterine factor was found to have a significant association with STL, with 32.95 times higher odds of STL in patients with the uterine factor. Parity and PGT-A use were not found to be significantly associated with STL in both the univariate and multivariate models. The data suggest lower odds of STL in parous patients compared with nulliparous, however, this association is not statistically significant.

Discussion

In this retrospective cohort study, STL after IVF was analyzed and compared with a well-matched control group; factors associated with STL include Black race, Hispanic ethnicity, adenomyosis, and tubal factor infertility. Additionally, this study was conducted as a follow-up to our initial study looking at STL in IVF patients from 1999–2002. When comparing the rate of STL from our current to our prior dataset, there were 14 (1.6%) of 872 in our current cohort, which is significantly lower than our prior data, where 17 (7.7%) of 220 had a STL. With an increase in eSET, the STL rate has been dramatically lowered. In our prior study, the most significant factor was multiple gestation, found in 58.8% of losses, followed by high-order multiple gestation, all triplets, found in 35.2% of losses. However, our current data only had one patient with multiple gestations who had an STL, therefore the rate of multifetal gestation was lower in our current group as compared with the prior cohort. Furthermore, the increased use of PGT-A, and the increase in transfer of known euploid or transferable mosaic embryos, has likely contributed to decrease in STLs as well.

When analyzing race and ethnicity, patients identifying as Black and Hispanic continued to suffer with higher rates of STL as compared with their White and Asian counterparts, which is consistent with our prior dataset as shown in Table 3. For our patients who identified as Black, race was associated with an 8.7% (6/69) risk of STL (P<.0001); this is a decrease compared with 14.5% in our old data, however, race still remains a highly significant risk factor. For ethnicity, patients who identified as Hispanic were associated with increased risk of STL at 4.3% (5/117) (P<.0001), although this is also lower than what we reported in the prior data, 7.8%, it remains significant. Furthermore, adenomyosis and intrauterine scarring were associated with a higher risk of STL. Tubal factor infertility remained associated with STL consistent with the prior dataset. Tuboperitoneal disease was positively associated with STL with 42.9% (6/14) of our STL patients with tubal disease, consistent with our prior data, although the rate was lower with 76.4% (13/17) of patients with STL having tubal factor previously.

Table 3.

Comparing rates of second-trimester loss among different ethnic and racial groups.

Race and ethnicity	Current data, 2015–2023 (n = number of patients within ethnic group with STL; N = total number of patients within ethnic group)		Prior data, 1999–2002 (n = 255)
Asian	2.3% (1/43)	ns	0%	ns
Black	8.7% (6/69)	P<.0001	14.5%	P<.01
Hispanic	4.3% (5/117)	P<.0001	7.8%	P<.01
White	0.4% (2/498)	ns	4.5%	ns

ns = not significant, STL = second-trimester loss.

In 2015, Bressler et al. (9) investigated STLs after IVF pregnancy in patients with a healthy uterine cavity. This study reported an STL of 2.1% which is consistent with our finding of an STL rate of 1.6%. Among STL, 48.3% of cycles were multigestation, which is consistent with our prior data. They found an increased odds of STL with leiomyomas, despite having a healthy intrauterine cavity on imaging. Tubal factor was not associated with an increased risk of STL. This is a deviation from our study, where tubal factor was associated with STL. In our study, tubal factor is likely associated with an increased risk of STL because of multiple reasons; patients with adenomyosis often have more endometriosis and thus more tubal factor. Furthermore, patients with pelvic inflammatory disease may have uterine and tubal inflammation, putting these patients at increased risk of STL. Regarding why the other study did not find that tubal factor was associated, it is possible that their patient group with STL was small enough (7/60) that it was not representative of the population as a whole and therefore represented a different population. Furthermore, the increase in incidence of STL among patients with adenomyosis and uterine adhesions may be secondary to many reasons. One consideration would be because of a history of corrective procedures that involve cervical dilation, such as operative hysteroscopy. Data show that repeat cervical dilation may increase the risk of cervical insufficiency, therefore STL (11). Although there may be other underlying reasons, the etiology is largely unknown; it is possible that these conditions are associated with an inflammatory process in the uterus, leading to early, silent dilation of the cervix, and thus very preterm labor.

Bressler et al. (9, 12) additionally found that compared with White women, Black women experienced an increased odds for loss, consistent with our findings, however their sample size was small (n = 48), therefore there may be residual confounding by leiomyomas. They found that ethnicity was not associated with increased odds of STL. However, their study population was 86.2% White, with only 2.8% of the study group identifying as Black and 3.5% as Hispanic highlighting the need for further studies on this subject with a diverse and representative patient population.

The baseline risk of STL for Black women has been found to be higher even in non-IVF populations. In Edlow et al. (1), it was reported that Black women comprised a greater proportion of the STL group, 27% (1). This is consistent with our results, which found that Black women continued to be at an increased risk of STL in our prior and updated study looking at IVF pregnancies. Furthermore, our study has found that Hispanic women remained at an increased risk of STL in both our prior and current study; however, we were unable to find a baseline risk in non-IVF populations. In non-IVF populations, our prior IVF population with a higher number of multiples, and our new IVF population with fewer multiples, Black and Hispanic women remain at an increased risk of STL, which highlights the need for further interventions to remedy this.

Our study has several limitations, the most significant of which is the small sample size. Although our total patient group is large (n = 971), because of the lower incidence of STL, our analysis of associated factors is within a much smaller group (n = 14), which may limit the generalizability of our findings. To generate confidence that these results are applicable to a broader population of patients who experience STL after an IVF pregnancy, a multicenter trial with a larger sample size is necessary. Despite a multivariate analysis to control for potential confounders, there are still limitations in controlling for confounders with retrospective data and a small sample size. The small sample size contributed to few observations in certain categorical predictors, which may have limited statistical power and resulted in unstable estimates in logistic regression, as reflected by extreme ORs and wide CIs.

Our study additionally has notable strengths; conducting the study at a single institution ensures consistency in the IVF process and makes it easier to track and monitor data. Additionally, the outcomes of our current study are able to be compared with our prior reported outcomes on STL during IVF from a different institution and patient base, which speaks to the consistency and generalizability of our data. Additionally, we were able to compare our study group with large control groups by comparing all STL patients to a control group of 853 patients and those with intrauterine cavity anomalies to a control group of 595 patients. Future prospective studies with larger sample sizes are warranted to further validate these findings. Additionally, further studies with a well-matched control group will be needed to eliminate bias.

In vitro fertilization continues to be associated with an increased risk of STLs. With increased eSET, this risk is far lower and is approaching that in the general population for White and Asian patients. However, Black and Hispanic patients, and those with adenomyosis, intrauterine scarring, and tubal factor infertility, remain at far too high a risk of STL and may benefit from early cervical length surveillance for IVF pregnancies (13). This update is necessary to develop subgroup-specific protocols to mitigate STL risk after IVF.

Conclusion

Factors found to be associated with STL include Black race, Hispanic ethnicity, adenomyosis, tubal factor infertility, and intrauterine scarring. The racial disparity in STL remains concerning and largely unchanged over 2 decades.

CRediT Authorship Contribution Statement

Sarah Rubin: Writing – review & editing, Writing – original draft, Visualization, Validation, Resources, Methodology, Formal analysis, Conceptualization. Olivia Nussbaum: Writing – review & editing, Writing – original draft, Validation, Methodology, Data curation, Conceptualization. Kaleb Noruzi: Software, Resources, Methodology, Formal analysis. Autumn Brewer: Supervision, Project administration, Conceptualization. Shmuel Sashitzky: Software, Resources, Data curation. Alexis Greene: Supervision, Project administration, Investigation. Martin Keltz: Writing – review & editing, Supervision, Software, Project administration, Methodology, Investigation, Formal analysis, Conceptualization.

Declaration of Interestss

S.R. has nothing to disclose. O.N. has nothing to disclose. K.N. has nothing to disclose. A.B. has nothing to disclose. S.S. has nothing to disclose. A.G. has nothing to disclose. M.K. has nothing to disclose.