The impact of a prior Cesarean delivery on embryo transfer: a prospective study

Abstract

Objective

To determine if a history of prior Cesarean delivery (CD) makes embryo transfer more difficult and impacts pregnancy outcomes.

Design

Prospective cohort study

Setting

Infertility practice in a Tertiary Care Military Facility

Patients

One hundred ninety-four patients with previous delivery undergoing IVF/ICSI-ET

Intervention

None

Main Outcome Measures

Live birth (primary), positive hCG, clinical pregnancy, and time to perform embryo transfer

Results

There was no statistically significant difference between patients with a history of only vaginal deliveries versus those with a history of CD for live birth (39% vs. 32%), positive hCG (56% vs. 53%), or clinical pregnancy (49% vs. 41%). Embryo transfers took longer in the history of CD group (157 vs. 187 seconds) and were more likely to have mucus (27% vs. 45%) or blood (8% vs. 21%) on the catheter.

Conclusion

Embryo transfers performed on patients with a prior CD took 30 seconds longer. They were also more likely to have blood or mucus on the catheter. Despite the apparently more difficult transfers, pregnancy outcomes were not different between the two groups.

INTRODUCTION

Over thirty percent of all deliveries in the United States today are Cesarean deliveries (CD) and this rate has remained relatively stable since 2010 (1). The majority are primary CD with an increasing number of elective primary CD, the long term implications of which may not be fully understood. Many complications of CD are well described, such as an increased risk for scarring of the pelvis and uterus, placental abnormalities and uterine rupture in future pregnancies. Multiple retrospective and observational studies also noted an increased risk of subfertility after cesarean delivery (2, 3). However, potential confounders exist and more recent data challenge this early association (4).

Potential causes of subfertility following CD include a prior history of infertility, increased age, obesity, or voluntary infertility (5). Another possible contributing factor is the impact of a cesarean section scar. A history of a CD leads to an anterior lower uterine segment defect in 42%–58% of patients (6, 7). These cases demonstrate a defect on routine transvaginal ultrasound or saline sonogram, with fluid often within the hysterotomy site. Pathologic evaluation of the hysterotomy site can demonstrate evagination of normal endometrium with the capacity to accumulate old blood (8). The role of the defect and how it affects the ease with which an embryo transfer can be performed, and the possible deposition of this trapped fluid on the catheter, has not been previously examined. Although fluid may be present in the hysterotomy site, the effect of ovarian stimulation on the amount of fluid and the potential leakage of this fluid into the endometrial cavity have not been examined. Furthermore, the hysterotomy scar fluid and old blood may impact embryo implantation and thus reduce pregnancy outcomes as has been found in patients with endometrial fluid with tubal factor infertility (9).

Embryo transfer (ET) is the final step of an assisted reproduction cycle and altered patient anatomy and surgical history may hinder transfer. Subjectively, multiple providers at our program have noted in several patients with prior CD, significant difficulty passing the embryo transfer catheter into the uterus. Under ultrasound guidance, the catheter has been noted to pass into the hysterotomy scar more readily than into the uterine cavity. Additional manipulation has often been required to enable threading the embryo transfer catheter past the hysterotomy scar, which could contribute to uterine irritation potentially impacting embryo implantation. The goal of this study was to evaluate the impact of a history of CD on ET and IVF outcomes. The long term effects of CD need to be more clearly elucidated to provide better information to patients and allow preparation for physicians performing ET.

MATERIALS AND METHODS

Study Design

The study design was a single site prospective cohort study of patients with a history of prior delivery. Patients were consecutively approached for enrollment while waiting in the holding area prior to transvaginal oocyte retrieval (TVOR). Each patient’s history was reviewed the morning of TVOR during routine preoperative counseling and if a patient confirmed a history of prior delivery, she was approached for consent for the study. Patients who consented to the study had the time of their embryo transfer recorded. The target enrollment was based on detecting a 15% difference in live birth rate from a baseline of 40% with a power of 80% and alpha of 0.05. To achieve this power, for a two-tailed hypothesis test 164 patients are required in each group, whereas for a one-tailed hypothesis test that CD live birth rates are lower than vaginal delivery rates, 132 patients are required in each group. This study was approved by the internal Institutional Review Board (IRB) at our institution.

Patients

Enrollment of patients occurred between March 2008 and May 2014 at a large tertiary care military facility. Patients met criteria for inclusion if they had a history of a previous delivery beyond 20 weeks gestation. For patients who had delivered more than once prior to enrollment, any history of CD led to being classified as having a history of CD. Patients with a single CD or multiple CDs were included in the study and assigned to the history of CD group. Only the first fresh embryo transfer was considered for patients who had multiple embryo transfers within the enrollment time period. Patients with missing embryo transfer timing were excluded from the analysis. Otherwise, all patients that met the inclusion criteria were incorporated in the analysis.

IVF Protocol

Patients underwent either a long luteal Lupron, microdose flare Lupron, or GnRH antagonist ovarian stimulation protocol after a course of three to five weeks on oral contraceptive pills, unless contraindicated. Patients were monitored with transvaginal ultrasound and blood estradiol levels. The endometrial thickness and follicle sizes for both ovaries were recorded during the ultrasound. When at least two lead follicles were 18 mm of larger, the trigger medication was given. Patients received human chorionic gonadotropin (hCG) 10,000 or 5,000 IU or leuprolide acetate 4 mg. Response to the trigger was assessed the subsequent day with serum hCG levels for the patients triggered with hCG or progesterone and LH levels for the patients triggered with leuprolide acetate. Patients underwent transvaginal oocyte retrieval (TVOR) 34–35 hours after trigger. Conventional fertilization or ICSI were performed following oocyte retrieval. Embryos were cultured for 3 or 5 days prior to transfer. Supernumerary embryos were cryopreserved on day 5 or 6. Patients received luteal support after TVOR in the form of intramuscular progesterone 50 mg/day or intravaginal progesterone 100mg three times daily. For patients receiving intravaginal progesterone, they were instructed to continue 100 mg three times daily until they have a positive pregnancy test, after which they are decreased to 100 mg twice daily.

Attending physicians and fellows in training performed all embryo transfers. The transferring providers were not blinded to the delivery history of the patient as this information was frequently noted on the IVF stimulation sheet. The transfers were performed using a Wallace transfer catheter with a full bladder under ultrasound guidance utilizing the afterload technique as previously described (10). The number of embryos transferred was in accordance with the ASRM guidelines. At our program it is routine for the transferring physician to record a subjective qualitative assessment of the difficulty of the transfer as “easy”, “medium”, or “difficult”. As there are no guidelines on how to distinguish between a “medium” versus “difficult” transfer and since the majority of transfers are recorded as “easy”, this variable was coded as either “easy” or “not easy” for the purpose of this analysis.

The time required for embryo transfer was recorded using a stopwatch. The nurse was instructed to start the timer when the empty transfer catheter was at the level of the external cervical os. The timer continued to run until the embryologist was done examining the transfer catheter for mucus, blood, and retained embryos. If there was a retained embryo, another attempt at transfer was made and the timer continued until the transfer resulted in a catheter clear of any retained embryos. If there was a switch from fellow to attending performing the transfer during the transfer, that time was also included in the final time for the embryo transfer.

The time taken to perform the embryo transfer along with the presence of mucus, blood, and whether there was a retained embryo was recorded in the patient’s chart at the time of the transfer. Blood and/or mucus were considered to be present if seen microscopically on the outside or inside of the catheter after flushing by the embryologist.

Serum hCG levels were drawn 14 days and 16 days after TVOR. Patients underwent a transvaginal ultrasound between 6–7 weeks gestation if they had an appropriately rising hCG. Subsequent pregnancy outcomes were determined from the military electronic medical records system or by contacting the patients by phone if they received care through a civilian provider.

Outcomes

The primary outcome of the study was live birth, which was defined as a pregnancy resulting in the birth of a viable infant after 24 weeks gestation. Secondary outcomes included positive serum hCG, clinical pregnancy, and time taken to perform embryo transfer. A positive serum hCG was defined as a value greater than 5 mIU/mL. Clinical pregnancy was defined as the presence of an intrauterine gestational sac seen on transvaginal ultrasound. The embryo transfer time was defined as the time from the transfer catheter reaching the level of the external cervical os to the time the transfer catheter is clear of retained embryos as described above.

Statistical Analysis

Variables were assessed for normality to determine if parametric or non-parametric statistical methods should be employed. Student’s t-test and the Mann-Whitney U test were used for comparing the central tendencies of parametric and non-parametric variables, respectively. Fisher’s Exact test was used to compare frequency data. Logistic regression was used to correct for confounding variables with respect to pregnancy outcomes. Confounding variables with a P-value < 0.2 were retained in the final adjusted models. Linear regression was used to assess the combined effects of other variables on embryo transfer time after applying a square root transform to mitigate the effects of the right-skewed distribution of the data. Power calculations were performed with G*Power version 3.1.9.2. All other statistical calculations were performed with IBM SPSS version 22. A P-value less than 0.05 was considered statistically significant.

RESULTS

The study discontinued enrolling patients after 5 years of attempting to reach the goal of 132 patients in each arm with valid outcome data. Enrollment was low because of absence of the principal and associate investigators due to military deployments. Out of all the patients approached for consent, only two patients declined participation in the study. Complete outcome data was available for 109 patients with a history of only vaginal deliveries and 85 patients with a history of at least one Cesarean delivery.

Baseline patient and cycle characteristics of the two groups are shown in Table 1. There were no statistically significant differences between the two groups with respect to baseline and cycle characteristics. Body mass index (BMI) was not tabulated because of the inconsistency in recording of this variable through the majority of the study period. The large number of microdose flare Lupron stimulation cycles at our program is due to equivalent outcomes of the protocol in good responders while minimizing the costs of delivering IVF care within the military medical system (11). The majority of cycles received hCG 10,000 IU as the trigger medication (vaginal delivery group 94%, history of CD group 92%; P=0.566). None of the antagonist cycles in either group received a GnRH agonist trigger in this cohort. There were no GnRH agonist triggers in the cohort because our program started using antagonist downregulation near the end of the study period and the total number of antagonist cycles were relatively small. None of those cycles were deemed to be at high risk for ovarian hyperstimulation syndrome, and therefore an hCG trigger was chosen.

Table 1.

Patient and cycle characteristics

		Vaginal Deliveries Only	History of CD	P-value
Number of Patients		109	85
Age (yrs)		35.3±4.2	35.6±4.3	0.649
Max FSH (IU/L)		7±2.4	6.8±1.8	0.556
Stimulation Protocol				0.601
Luteal Lupron		10/109 (9%)	10/85 (12%)
Microdose Flare		90/109 (83%)	65/85 (76%)
Antagonist		9/109 (8%)	10/85 (12%)
Total Gonadotropins (IU)		3719±1578	3689±1648	0.898
Endometrial Thickness at hCG Trigger (mm)		11.2±2.4	11.4±2.4	0.713
Oocytes Retrieved		12.5[9–18]	11[7.3–19]	0.501
ICSI		92/109 (84%)	65/85 (77%)	0.198
# of Embryos Transferred		1.95±0.74	1.96±0.76	0.922
Day of Transfer				0.885
3		71/109 (65%)	52/85 (61%)
5		38/109 (35%)	33/85 (39%)
Best Transfer Grade				0.629
Good		55/109 (51%)	37/85 (44%)
Fair		42/109 (38%)	37/85 (44%)
Poor		12/109 (11%)	11/85 (13%)
Fellow Transfer		69/109 (63%)	43/85 (51%)	0.081
Luteal Support				0.375
Vaginal Progesterone		71/109 (65%)	50/85 (59%)
IM Progesterone		38/109 (35%)	35/85 (41%)

Caption:

Values reported as mean±standard deviation, median [interquartile range], and fraction (%).

IM=intramuscular.

Table 2 shows the embryo transfer outcomes and cycle pregnancy outcomes. As expected, the time of embryo transfer was not normally distributed and right-skewed due to transfers that take longer to perform. The median time taken to perform the embryo transfer in the group with a history of CD versus the vaginal delivery only group was 30 seconds longer (P=0.002). Re-analysis of the median embryo transfer time excluding the retained embryos (1 in the vaginal delivery group and 2 in the history of CD group) results in the same 30 second median difference in transfer time (P=0.001), which is expected given that the median and corresponding non-parametric statistical analysis method (Mann-Whitney U test) are less sensitive to outliers than other statistical assessments of central tendency when dealing with right-skewed data. The embryo transfer catheters in the history of CD group were almost three times as likely to have blood (P=0.012) and almost twice as likely to have mucus on the catheter (P=0.010). There were no statistically significant differences in retained embryos, positive hCG, clinical pregnancy, or live birth between the two groups. Transfers in the vaginal delivery group were subjectively more likely to be assessed as “easy” compared to those in the CD group (95% versus 76%; P<0.001).

Table 2.

Embryo transfer characteristics and pregnancy outcomes

	Vaginal Deliveries Only	History of CD	P-value
Transfer time (sec)	157 [128–190]	187 [142–309]	0.002
Mucus on catheter	29/109 (27%)	38/85 (45%)	0.010
Blood on catheter	9/109 (8%)	18/85 (21%)	0.012
Retained embryo	1/109 (1%)	2/85 (2%)	0.582
Easy transfer	104/109 (95%)	62/82 (76%)	<0.001
Positive hCG	61/109 (56%)	45/85 (53%)	0.771
Clinical pregnancy	53/109 (49%)	35/85 (41%)	0.313
Live birth	42/109 (39%)	27/85 (32%)	0.366

Caption:

Values reported as fraction (%) and median [interquartile range]. Blood and mucus on catheter include microscopic blood and mucus contamination. Three of the charts in the history of CD group did not report the difficulty of the transfer.

Logistic regression was performed to determine the effect of history of CD for the outcomes of positive hCG, clinical pregnancy, and live birth while accounting for female age, day of transfer, number of embryos transferred, and best embryo grade transferred as confounders. The results of the analysis are presented in Table 3. There was no statistically significant impact of a history of CD on the odds of any of the pregnancy outcomes for both unadjusted and adjusted analyses. The confounding variables of female age, number of embryos transferred, day of embryo transfer, and best embryo transfer grade, when added to the logistic regression models, had a P-value < 0.2 and thus were retained in the models during the analysis. Other variables such as stimulation protocol, trigger medication dose, endometrial thickness on day of trigger, fertilization method, subjective difficulty of transfer, and luteal support had P-values > 0.2 and thus were not included in the adjusted model. Although the effect of including fellow transfer as a confounder did not meet the cutoff of P-value < 0.2 (P>0.5 or greater for all pregnancy outcomes), there has been some controversy over this factor in IVF cycles. Including the variable of whether a fellow performed the transfer in addition to the confounders with a stronger effect on the model (female age, day of transfer, number of embryos transferred, and best embryo grade transferred), the adjusted odds ratios of the pregnancy outcomes with respect to having a history of CD were as follows: positive hCG 0.898 (95% CI 0.483–1.67; P=0.733), clinical pregnancy 0.759 (95% CI 0.414–1.39; P=0.373), and live birth 0.789 (95% CI 0.421–1.481; P=0.461). This is in agreement with a recent study that found fellows participating in transfers do not impact pregnancy outcomes (12).

Table 3.

Effect of history of Cesarean delivery on pregnancy outcomes

	Unadjusted Odds Ratio	P-value	Adjusted Odds Ratio	P-value
Positive hCG	0.89 (95%CI 0.50–1.57)	0.675	0.91 (95%CI 0.49–1.69)	0.764
Clinical pregnancy	0.74 (95%CI 0.42–1.31)	0.302	0.76 (95%CI 0.41–1.39)	0.372
Live birth	0.74 (95%CI 0.41–1.35)	0.329	0.77 (95%CI 0.42–1.43)	0.410

Caption:

Adjusted for female age, day of embryo transfer, number of embryos transferred, and highest quality embryo transferred.

The effect of other variables on embryo transfer time was assessed by creating a linear regression model with the variables history of CD versus vaginal delivery, “easy” versus “not easy” transfer, and fellow versus attending as transferring physician. A square root transform was applied to the right-skewed distribution of embryo transfer time to normalize the data for linear regression. The adjusted effect of history of CD increased the transfer time by 33.4 seconds (95% CI 6.8–62.2; P=0.013). A transfer that was characterized as “not easy” increased the adjusted transfer time by 142 seconds (95% CI 93–196; P<0.001). There was a statistically insignificant increase in transfer time when a fellow performed the transfer of 15.7 seconds (95% CI −8.75–42; P=0.215). To determine the effect of a transfer classified as “not easy” when the patient had a history of CD, an interaction term between the two variables was added to the model. The only statistically significant finding of the model was for the interaction term of history of CD with “not easy” transfer which was an increase in the adjusted transfer time of 111 seconds (95% CI 11–234; P=0.026). The effect of history of CD alone failed to meet statistical significance, however, the trend and magnitude remained with an increase in transfer time of 23.8 seconds (95% CI −3.1–52.8; P=0.084) despite the addition of the interaction term. The impact of a “not easy” transfer alone also failed to meet statistical significance and the effect on transfer time was greatly reduced in the model with the interaction term [49.5 seconds (95% CI −26.4–144; P=0.221)]. The effect of a fellow performing the transfer was unchanged from the model without the interaction term.

DISCUSSION

This prospective observational study revealed no difference in pregnancy outcomes between patients with a history of CD versus those with only prior vaginal deliveries. Patients with a history of CD had ETs that took longer to perform and were more likely to have ET catheters with blood and/or mucus. Despite the increased occurrence of blood and mucus of the ET catheter, the pregnancy outcomes were not significantly different.

The finding of blood or mucus associated with the post-transfer examination of the catheter tip may reflect local trauma anywhere within the uterus. In 1998, Goudas et al. sought to quantify the amount and location of blood on the catheter at the time of embryo transfer (13). These transfers were not performed with ultrasound guidance but blood on the outside of the catheter significantly affected implantation rate. Similarly, in 2002, Tomas et al. included blood noted on the catheter as one of the criteria for a difficult transfer which had a lower pregnancy rate (14). Alvero et al. in 2003 reviewed 584 consecutive ETs with a significant decrease in clinical pregnancy rate from 49.7% to 31.7% if blood was noted (15). However, in 2010, Moragianni et al. in a randomized controlled trial evaluating clinical pregnancy rate in 364 fresh ETs noted no significant change in pregnancy rate if the delivery system had blood or mucus on review (16).

Some of the disagreement among prior studies regarding the presence of blood on the embryo transfer catheter may be explained by the role of the cesarean scar. In a prospective observational study of 70 individuals, Armstrong et al. were able to identify 100% of patients’ hysterotomy scars with ultrasound and noted fluid in 42% of the hysterotomy scars (6). The risk of having a cesarean scar defect was increased in patients with a history of laboring prior to CD and also with increasing number of CDs. Cesarean scar defects are thought to cause abnormal uterine bleeding due to the formation of an isthmocele in the lower uterine segment that traps menstrual blood which is later slowly released (17, 18). Further evidence that the isthmocele is the cause of abnormal bleeding is that hysteroscopic correction of the defect leads to improvement in symptoms in the majority of patients (18). Therefore, a plausible explanation for the increased incidence of blood on the transfer catheter in patients with a history of CD is due to the presence of an isthmocele depositing blood on the catheter. Since the source of the blood on the catheter would be the isthmocele and not due to endometrial trauma, an impact on pregnancy would not occur as we found in our study.

The type of uterine incision in the history of CD group may have impacted the difficulty of transfer. A low transverse scar may act like a pothole in a road that traps the transfer catheter as it moves up the cavity whereas a classical incision scar may act as a channel that the embryo transfer catheter follows toward the fundus. Since we did not collect that data from the patients in the study and cannot easily verify the type of the scar as most patients deliver at outside institutions, there is no way of knowing if the type of uterine scar impacted the results. The prevalence of patients with a history of classical incision is only about 1.5% (19). Given the relatively low prevalence of classical incisions along with the sample size of this study, it is unlikely that adjusting for incision type in any of the models would have an appreciable effect in this cohort. Performing a study with enough statistical power to examine the effects of the type of uterine incision on transfer difficulty would require a very large number of patients given the low prevalence of classical incisions, but may offer more insights into the fundamental reasons for the findings in our study.

A strength of our study is that patients were prospectively enrolled before embryo transfer at the time of oocyte retrieval. The prospective design of this study reduces the biases that affect retrospective studies. Furthermore, this study provides the highest level of evidence to answer this clinical question because a randomized controlled trial allocating patients to vaginal delivery or cesarean section to study their outcomes in a subsequent IVF cycle is not ethical.

One limitation of the current study is that it was stopped prior to enrollment of the planned number of patients. This ultimately reduced the power to detect the planned 15% difference in live birth. Post-hoc power analysis with the number of patients enrolled, assuming a 39% live birth rate in the vaginal delivery group, alpha of 0.05, and a power 80% results in a detectable difference of 19% for the two-tail hypothesis test and 17.5% for the one-tail hypothesis test that a history of CD results in a lower chance of pregnancy. Examined from the standpoint of power, the enrollment of this study achieves a power of 57% for the two-tail hypothesis test and 66% for the one-tail hypothesis test to detect a 15% difference in live birth outcome.

Another limitation of the study is that BMI was not included in the analysis due to inconsistencies in recording of this variable through the majority of the enrollment period. Although patients with a higher BMI may be more likely to have a CD, they are also more likely to have decreased live birth (20–22). Therefore, a difference in patient demographics favoring a higher BMI in the history of CD group would bias the results toward lower pregnancy. This would result in supporting the hypothesis of a lower occurrence of pregnancy in the history of CD group, which this study did not find. A confounding effect of BMI may have been seen in the secondary outcome of time to perform the ET. Higher BMI patients are generally more difficult to obtain clear transabdominal ultrasound images of during ET. A suboptimal ultrasound image may cause more difficulty in performing the transfer that translates into longer ET times.

A further limitation is that embryo transfer charting only noted the final transferring provider. This method of recording the level of experience of the transferring physician cannot capture the rare instances at our program of the attending taking over for the fellow during a difficult transfer. Although this may have happened in this cohort of patients, a recent study demonstrated no impact on pregnancy rates when the transfer is performed under ultrasound guidance with the afterload technique, which is how all transfers are performed at our program (12). The outcome of the logistic regression models including the fellow transfer variable further supports the lack of effect a transferring fellow has on pregnancy outcomes. Additionally, the transferring provider, whether a fellow or attending, was not blinded to the delivery history of the patient which could have potentially biased who was going to perform the transfer. Including the embryologist’s evaluation in the transfer time allowed for a well-defined endpoint to the transfer. Although multiple embryologists in our program assist with transfers, the difference in the technical speed with which they perform the post-transfer evaluation is minimal. Furthermore, any difference in speed between embryologists would not bias the towards one group over another since they rotate and are equally likely to participate in any given transfer. In fact, such differences in embryologist speed would bias the study towards a null result. Given that the study found a significant difference in transfer time between the two groups, the difference in embryologist speed was far less than that of the impact of a history of CD.

In conclusion, this study was unable to demonstrate a significant effect of a prior Cesarean delivery on pregnancy outcomes in IVF cycles. Further studies would be warranted to augment the population analyzed which may uncover a smaller potential effect. Notably, this study demonstrated that embryo transfers performed on patients with a history of CD are more difficult and require a significantly longer time to perform. Additionally, the embryo transfer catheters more often have blood and mucus in patients with a history of CD. As the final step in IVF, it is vital for providers to be appropriately prepared for difficult embryo transfer procedures in order to minimize trauma. Overall this study provides reassuring data to patients and providers that a prior CD will not significantly alter their IVF cycle pregnancy outcome.