Early β-hCG levels predict live birth after single embryo transfer

Abstract

Purpose

Specific serum beta human chorionic gonadotropin (β-hCG) parameters that can predict live birth after an embryo transfer have yet to be defined.

Methods

We performed a retrospective cohort study of 1,028 patients with a detectable β-hCG who underwent a single embryo transfer between 2002 and 2019 at a large academic center. Two β-hCG parameters were examined in relation to live birth: 1) “doubling” defined as β-hCG doubling over 48 h and 2) “reaching 100” defined as a β-hCG ≥ 100 mIU/mL by 15 days after oocyte retrieval (AOR).

Results

One thousand three hundred forty cycles involving a single embryo were analyzed. Two thirds were frozen embryos and 86% were blastocyst transfers. Preimplantation genetic testing was performed in almost 30% of cycles. When β-hCG levels “doubled,” a live birth occurred in 80.7% of cycles and when β-hCG levels “reached 100” by 15 days AOR, live birth occurred in 81.6% of cycles. When β-hCG levels both doubled and reached 100 by 15 days, AOR 85.4% cycles resulted in live birth. A multiple logistic regression model to control for patient and cycle level factors revealed a live birth odds ratio (OR) of 8.0 (95% CI 5.7–11.1) when β-hCG “doubled” and an OR of 21.2 (95% CI 14.3–31.5) when β-hCG “reached 100.” When both these latter parameters were met, the OR was 12.5 (95% CI 8.9–17.8).

Conclusion

β-hCG parameters of “doubling” and “reaching 100” by 15 days AOR are robust predictors of live birth and can aid in patient counseling regarding pregnancy outcomes soon after single embryo transfer.

Supplementary information

The online version contains supplementary material available at 10.1007/s10815-022-02606-w.

Introduction

Despite advances in assisted reproductive technology (ART) leading to improved pregnancy rates, less than 50% of oocyte retrievals result in a successful live birth following in vitro fertilization (IVF) [1]. Undergoing IVF is often stressful, not only because of the arduous nature of the process and financial investment, but also due to the uncertain outcomes [2–4]. Many couples experience anxiety throughout their reproductive journey with tension escalating during IVF and peaking around the time of embryo transfer [3–5]. Couples apprehensively await for a successful implantation, but given that the ultimate goal is a live birth, further seek the odds of longer-term success once initial testing confirms a pregnancy. Therefore, a timely and accurate predictor of live birth can be instrumental in counseling patients after an embryo transfer and may conceivably help decrease anxiety once implantation is confirmed. Currently, a widely accepted standard for the prediction of live birth relies on ultrasonographic detection of fetal cardiac activity, typically occurring around 6-week gestation [6]. However, as a pregnancy can be confirmed with serum testing around 4 weeks gestation, the opportunity exists to provide prognostic information weeks earlier.

When conception occurs, the syncytial trophoblast secretes human chorionic gonadotropin (β-hCG), which is detectable in maternal serum approximately 6–8 days post fertilization [7, 8]. Serum β-hCG can confirm early implantation but can also aid in identifying ectopic pregnancies or miscarriages [8, 9]. Low levels of β-hCG correlate with an increased likelihood of nonviable pregnancies, whereas higher levels usually confer viability [10]. β-hCG levels are also predictive when examined logarithmically as doubling time, the time interval in which β-hCG levels increase twofold [9, 11, 12]. Based on this evidence, it is common practice during IVF to trend sequential β-hCG levels after embryo transfer to determine the rate of β-hCG rise and, thus, predict whether a pregnancy is likely intrauterine. However, although serum β-hCG is frequently assayed early in gestation, there is a paucity of data on the precise trajectory of serum β-hCG levels after embryo transfer, particularly in the prediction of live birth. Numerous studies have sought to identify serum markers that can predict outcomes after IVF [7–9, 12–25]. Yet these prior studies, often of limited sample size, focus on disparate measurement days following oocyte retrieval and sometimes only examine specific IVF cycle types or include multiple embryo transfers. These discordant study designs preclude generalizable β-hCG criteria to interpret and apply clinically.

Notwithstanding that β-hCG levels can discern successful implantation, we aimed to evaluate the association between specific β-hCG parameters following single embryo transfer and live birth. We hypothesized that a doubling of serum β-hCG over 48 h would be associated with an increased likelihood of live birth. Additionally, we postulated that a β-hCG level of at least 100 mIU/mL at 15 days after oocyte retrieval would likewise be a reliable marker of live birth. Finally, we proposed that the likelihood of live birth would be the highest if both parameters were met (Fig. 1).

Fig. 1.

Study outline. Three serum β-hCG parameters are strong predictors of live birth following single embryo transfer. OR = odds ratio. Figure created from Canva.com. Embryo and oocyte images from eugenelabs.com

Methods

Institutional Review Board approval was obtained (STU00211722) for this retrospective cohort study. All fresh and frozen IVF cycles that occurred at Northwestern Medicine between January 2002 and January 2019 were reviewed. All individuals who underwent a single embryo transfer and were found to have a detectable initial serum β-hCG after embryo transfer (β-hCG > 0 mIU/mL) were included in the analysis. Given that our institution routinely measures β-hCG 1–2 days earlier than others following an embryo transfer, any detectable value was included to capture a comprehensive understanding of hCG rise in pregnancy. Patient demographics, IVF cycle details, β-hCG values, and pregnancy outcomes were collected from an internal database from which data is extracted for Society of Assisted Reproductive Technology (SART) reporting and routinely inspected for quality control. Demographic data included maternal age, body mass index (BMI), race, and infertility diagnosis/IVF indication. The following cycle data was also extracted: fresh versus frozen embryo transfer, blastocyst versus cleavage stage embryo transfer, utilization of pre-implantation genetic testing (PGT), and donor oocyte utilization. The decision to transfer a cleavage stage embryo was made if the quality or quantity of available embryos on day 3 post-retrieval was limited. Live birth was defined by SART criteria as the birth (delivery) of at least one fetus live born, i.e., showed signs of life after the complete expulsion or extraction from its mother or gestational carrier. Signs of life include breathing, heartbeat, pulsation of the umbilical cord, or definite movement of the voluntary muscles. Hence, any birth event in which an infant shows sign of life is counted as a live birth, regardless of gestational age at delivery. Due to the large sample size and retrospective nature of this study, a post hoc power analysis was deferred.

Quantitative serum β-hCG levels were assayed using a standardized Northwestern Hospital System laboratory protocol. The quantitative total β-hCG concentrations were measured using a sequential two-step chemiluminescent immunoenzymatic assay (Access Total BHCG 5th International Standard, Beckman Coulter, Brea, CA). The assay range is from 0.6 to 270,000 mIU/mL. The intra-assay and inter-assay coefficients of variation varied from 3.7 to 5.0% and from 4.0 to 5.3%, respectively. Following an embryo transfer, an initial serum β-hCG was scheduled 8 days after blastocyst single embryo transfer and 10 days after cleavage-stage single embryo transfer. Serum β-hCG was then repeated 48 h later. To compare β-hCG levels across IVF cycles where embryos were transferred at different developmental stages, β-hCG levels were determined in reference to days “after oocyte retrieval (AOR),” identical to the number of days after fertilization.

After reviewing prior literature on β-hCG trends following various embryo transfers, we established two specific β-hCG parameters to determine their relationship with live birth [7–10, 14, 17, 23, 26–29]. Our objective was to establish β-hCG parameters at the earliest possible time point following a single embryo transfer. The two following β-hCG parameters were assessed: 1) “reached 100” which was defined as a β-hCG level of ≥ 100 mIU/mL at 15 days AOR and 2) “doubled” which was defined as a β-hCG level that at least doubled over a 48-h period (Fig. 1). The primary outcome was these β-hCG criteria and their association with live birth. Secondary outcomes were the association between β-hCG trends and maternal BMI as well as IVF cycle details (fresh versus frozen embryo transfer, blastocyst versus cleavage stage embryo transfer, PGT status, and use of a donor oocyte).

Statistical analysis

Data were summarized descriptively, including within strata defined by live births versus other outcomes. Pairwise comparisons between variables for live birth cycles and those resulting in other outcomes were conducted using exact tests for categorical variables and nonparametric tests (Wilcoxon, Kruskal–Wallis) for continuous variables.

Primary analyses for the relationship between β-hCG trajectories and live birth were conducted using logistic regression with live birth (binary yes/no) as the outcome. A series of likelihood ratio tests and examination of variance inflation factors found that including β-hCG exceeding 100 mIU/mL at 15 days AOR and doubling from day 13 to 15 AOR in the same model provided a better fit despite their high correlation (Kendall’s τ = 0.61). Thus, we fit separate logistic regression models for the following dependent variables: (1) β-hCG exceeding 100 mIU/mL at 15 days AOR, (2) doubling from day 13 to 15 AOR, and (3) β-hCG both doubling and exceeding 100 mIU/mL at 15 days AOR. In addition, we fit a model including fixed effects for both β-hCG doubling and exceeding 100mIU/mL at 15 days AOR. Each of these logistic regression models adjusted for cycle and patient characteristics, including patient BMI, age, race/ethnicity, fresh versus frozen embryo transfer, blastocyst versus cleavage-stage embryo transfer, use of PGT, and use of a donor oocyte. These latter patient and cycle level covariates were selected for control due to protentional confounding influence on β-hCG from the literature [16, 19, 30, 31]. Inference focused on coefficients for relevant β-hCG variables (doubling, exceeding 100 mIU/mL, or both) with two-sided tests; we report results on the scale of odds ratios.

To examine the correlates of β-hCG trajectories, we modeled β-hCG levels as a function of days AOR. Due to the heavy skewness of β-hCG levels, log β-hCG was adopted as the dependent variable. We used mixed effects models with random cycle effects to account for the correlation of log β-hCG over time. Upon cursory visualizations, it became clear that while log β-hCG trajectories for live birth cycles were approximately linear, those for cycles resulting in other outcomes (not ending in a live birth) were not. Thus, we fit one model with all cycles (live births and other outcomes) using a piecewise polynomial basis expansion (B-spline of order 3) to account for nonlinear growth. This model included main effects for time (days AOR via the B-spline), live birth, and a time-live birth interaction, as well as the covariates used for adjustment in the models for live birth. To evaluate whether BMI is related to the trajectory of β-hCG levels for cycles resulting in live birth, we fit a separate, but similar, mixed effects model with random cycle effects and fixed effects for days AOR (without basis expansion), BMI, and BMI-days AOR interaction, while adjusting for relevant covariates listed above.

Because there was a nontrivial amount of data missing (rates ranging from < 1 to 28%), model-based analyses were run using multiple imputation. We generated m = 20 imputed datasets using predictive mean matching algorithms and did not exclude any patients. Analyses and imputations were conducted using the R programming language (version 4.1.0).

Results

A total of 1,340 IVF cycles from 1,028 patients where a single embryo transfer was performed, and the initial β-hCG level was detectable, were included in the analysis. A total of 76.5% of patients had only a single cycle of IVF included in the analysis while 17.7% of patients had two cycles, and 5.8% had three to five cycles. Baseline patient characteristics and IVF cycle details are illustrated in Table 1. The mean maternal age at the time of embryo transfer was 35.3 years. Two thirds of the cycles were frozen embryo transfers and 86% were blastocyst embryo transfers. PGT was performed in over a quarter of cycles and a donor oocyte was utilized in approximately 10% of cycles. The overall cycle live birth rate was 64% following a detectable β-hCG (Table 1).

Table 1.

Patient demographics and IVF cycle details

	Live birth	No live birth	All outcomes	P value
	(862 cycles)	(478 cycles)	(1,340 cycles)
Live birth rate	-	-	64.3%	-
Other outcomes rate (%)	-	-	35.7%	-
Spontaneous miscarriage	-	-	174 (36)	-
Stillbirth	-	-	2 (0.4)	-
Biochemical	-	-	271 (57)	-
Ectopic	-	-	15 (3.1)	-
Unknown	-	-	16 (3.3)	-
Age, years	35.2 ± 4.3	35.6 ± 4.6	35.3 ± 4.4	0.18
BMI, kg/m2	24.4 ± 5.1	25.1 ± 5.9	24.6 ± 5.4	0.15
Race (%)				0.60
White	464 (54)	262 (55)	726 (54.2)	-
Black	26 (3)	15 (3.1)	41 (3.1)	-
Hispanic/Latino	27 (3.1)	17 (3.5)	44 (3.3)	-
Asian	89 (10.3)	39 (8.2)	128 (9.6)	-
Multiple races/ethnicities	14 (1.6)	10 (2.1)	24 (1.8)	-
Unknown/not reported	240 (28)	134 (28.1)	374 (27.9)	-
Infertility indication (%)
Diminished ovarian reserve	177 (23)	112 (26)	289 (21.6)	0.33
PCOS	70 (9.2)	31 (7.2)	101 (7.5)	0.24
Endometriosis	19 (2.5)	26 (6.0)	45 (3.4)	0.004
Tubal factor	1 (0.1)	0 (0)	1 (0.1)	> 0.99
Recurrent pregnancy loss	28 (3.7)	11 (2.5)	39 (2.9)	0.32
Male Factor	180 (24)	87 (20)	267 (19.9)	0.17
Unexplained	155 (20)	95 (22)	250 (18.7)	0.55
Fresh embryo transfer (%)	291 (33.8)	167 (34.9)	458 (34.2)	0.72*
Frozen embryo transfer (%)	570 (66.1)	312 (65.1)	882 (65.7)
Cleavage stage transfer (%)	113 (12.6)	71 (14.9)	184 (13.7)	0.99*
Blastocyst transfer (%)	750 (87.1)	406 (85.4)	1,156 (86.2)
PGT Performed	257 (29.8)	120 (25.1)	377 (28.1)	0.33
Donor oocyte utilized	89 (10.3)	55 (11.5)	144 (10.7)	0.58

Study population characteristics examined by live birth, no live birth, and all outcomes. Age and body mass index (BMI) presented as mean ± standard deviation (SD). IVF cycle factors, race, infertility diagnosis/IVF indication, and other outcomes presented as N (% = percent cycles per each outcome group). P-values calculated between column groups of live birth versus no live birth using Wilcox rank sum test. *P-value calculated using chi-square comparing both groups (i.e., fresh versus frozen) and their relationship to live birth

In the cycles that resulted in live birth the median β-hCG level was 268.3 mIU/mL at 15 days AOR. In the cycles that did not end in live birth, the median β-hCG level was 56.8 mIU/mL at 15 days AOR (Table 2). The lowest initial β-hCG level at 15 days AOR to lead to a live birth was 9.1 mIU/mL following a blastocyst frozen embryo transfer. This β-hCG value doubled to 24.9 mIU/mL in 48 h and was 437 mIU/mL by 23 days AOR. When examined by maternal BMI, median β-hCG at 13 and 15 days AOR correlated negatively with BMI (both p < 0.001). Notably, median β-hCG level was also significantly lower for cleavage versus blastocyst at both day AOR 13 and 15 (both p = 0.02). The was no difference in median β-hCG concentrations at day AOR 13 and AOR 15 in the other subgroups: fresh versus frozen embryo transfer, PGT versus no PGT, and cycles that involved a donor oocyte versus an autologous oocyte.

Table 2.

Serum β-hCG values by outcome and cycle factor subgroups

	13 AOR		15 AOR
	β-hCG median (IQR)	P value	β-hCG median (IQR)	P value
Live birth	107.0 (75, 145)		268.3 (181.3, 375.0)
Other outcomes	36.8 (15, 69)		56.8 (18.9, 138.2)
Normal BMI < 25	116 (80, 153)	< 0.001	303 (195, 410)	< 0.001
Overweight BMI 25 to < 30	101 (67, 129)		240 (157, 338)
Obese BMI ≥ 30	91 (68, 112)		223 (137, 308)
Fresh IVF	114 (76, 146)	0.65*	280 (182, 381)	0.84*
Frozen IVF	107 (76, 146)		272 (185, 384)
Blastocyst	110 (78, 147)	0.02*	280 (187, 387)	0.02*
Cleavage	98 (56, 126)		249 (134, 340)
PGT-A	99 (76, 138)	0.07*	284 (184, 387)	0.38*
No PGT-A	112 (76, 149)		256 (184, 376)
Donor oocyte	120 (71, 165)	0.31*	338 (182, 451)	0.06*
Autologous oocyte	107 (76, 145)		270 (185, 379)

β-hCG concentration (mIU/mL) displayed as median with interquartile range (IQR). All subgroup rows except “other outcomes” are from pregnancies ending in live birth. P-values calculated between median BMI classes using Kruskal–Wallis rank sum test, one-way ANOVA. *P values calculated between subgroup median values (i.e., blastocyst versus cleavage stage) using Wilcoxon rank sum, Welch two sample t-test

When β-hCG levels “doubled” (N = 827) a live birth occurred in 80.7% of cycles and when β-hCG levels “reached 100” by 15 days AOR (N = 908), a live birth occurred in 81.6% of cycles. Meanwhile, 85.4% cycles for which β-hCG levels both doubled and reached 100 by 15 days AOR resulted in live birth (N = 731). When β-hCG levels did not “double” (N = 344), a live birth occurred in 22.7% of cycles and when β-hCG levels did not “reach 100” by 15 days AOR (N = 387), a live birth occurred in 22.2% of cycles. These results are displayed graphically in Fig. 2a. When β-hCG was 200 mIU/mL or higher by 15 days AOR, the chance of live birth was approximately 88% (Fig. 2b). Lastly, the fold increase of β-hCG from day AOR 13 to AOR 15 was 2.5 ± 0.5 for those who had a live birth and 1.7 ± 0.9 for those who did not have a live birth (mean ± SD, p < 0.001).

Fig. 2.

β-hCG and live birth rate. a Percent of live birth by β-hCG parameters. β-hCG parameters defined as β-hCG ≥ 100 mIU/mL by 15 days after oocyte retrieval (AOR), β-hCG level at least doubling over 48 h, and meeting both latter criteria. b Percent of live birth by β-hCG at 15 days AOR: range of serum β-hCG maximum levels (mIU/mL) by 15 days AOR

Multiple logistic regression models were employed to control for potential patient (maternal age, race/ethnicity, BMI) as well as cycle level factors (fresh/frozen embryo transfer, cleavage/blastocyst stage, PGT testing status, and donor oocyte utilization). To handle missing data, these models were fit on multiply imputed datasets and the results were pooled. These models estimated the odds ratio (OR) of a live birth of 8.0 (SE = 0.2, p < 0.0001) when β-hCG “doubled” (versus not doubling) and 21.2 (SE = 0.2, p < 0.0001) when β-hCG “reached 100” (Table 3). When both these latter parameters were met, the live birth OR was estimated to 12.5 (SE = 0.2, p < 0.0001). These ORs adjusted for the above-mentioned covariates and were based on logistic regression models applied to multiple imputed datasets (Table 3). Notably, in the model that included effects for β-hCG doubling and reaching 100 mIU/mL, we estimated odds ratios of 13.2 (SE = 3.8) for “reaching 100” and 3.3 (SE = 0.7) for “doubling” indicating that the relationship between β-hCG doubling and live birth is much weaker after adjusting for β-hCG reaching 100mIU/mL by 15 days AOR (Supplemental Table I).

Table 3.

Serum β-hCG parameters and the association with live birth

	Unadjusted			Adjusted
	Odds ratio	95% CI	P-value	Odds ratio	95% CI	P-value
Doubled	8.2	[6.3,10.9]	< 0.0001	8.0	[5.7, 11.1]	< 0.0001
Reached 100	19.7	[14.1, 27.4]	< 0.0001	21.2	[14.3, 31.5]	< 0.0001
Doubled and Reached 100	11.6	[8.7, 15.6]	< 0.0001	12.5	[8.9, 17.8]	< 0.0001

Imputation models adjusted for patient and cycle level covariates including maternal age, BMI, race/ethnicity, fresh/frozen IVF cycle, cleavage/blastocyst, PGT testing, and donor oocyte utilization. Adjusted odds ratio was calculated with 95% confidence intervals (CI). P-values calculated using Wald test for logistic regression coefficients

The correlation of β-hCG trajectory and live birth was further underscored by trajectory models as displayed in Fig. 3. Patients had a mean of 2.4 days between interval serum measurements (with a max of 6) following an embryo transfer. The mixed effects model using a B-Spline basis expansion to account for nonlinear trends in β-hCG as a function of days AOR found statistically significant differences in the initial β-hCG levels as well as subsequent rise among cycles resulting in live births versus those resulting in other outcomes (p < 0.0001 for main effects and interactions for time and live birth). Of clinical relevance, cycles not ending in live birth tended to have lower initial β-hCG levels and their trajectory appeared to diverge sharply around day 15 AOR. Regression coefficients for posterity are included in our supplemental data (Supplemental Table I).

Fig. 3.

Trajectory of serum β-hCG and live birth rate. Mixed effects model displaying log scale of total β-hCG (mIU/mL) over number of days after oocyte retrieval (AOR). B-Spline basis expansion used for nonlinear trends in β-hCG as a function of days AOR. Live birth versus other outcomes, P < 0.0001

Given the inverse trend of median β-hCG concentration with maternal BMI (Table 2), additional β-hCG trajectory and mixed effects analysis were performed. Among cycles resulting in live birth, BMI was related to trajectories of β-hCG based on both mixed effects models and descriptive statistics. These models found that individuals with higher BMI tended to have lower β-hCG levels at 13 days AOR (B =  − 0.01, SE = 0.005, p = 0.008) and their β-hCG growth appeared to be slightly slower (B =  − 0.002, SE = 0.002, p = 0.26). However, given the difference in these trajectory slopes did not vary by BMI, this was not a statistically significant interaction. Though these analyses are on the log scale, Fig. 4 shows median serum β-hCG levels by BMI among cycles that ended in live birth. For example, at day 13 AOR, individuals with BMI ≥ 30 whose cycle ended in a live birth had β-hCG values approximately 10 mIU/mL lower than those with BMIs 25 to < 30, and nearly 30 mIU/mL lower than those with a BMI < 25 (p < 0.001). By day 15 AOR, individuals with BMI ≥ 30 on average had β-hCG levels about 30 mIU/mL lower than those with BMI in the 25–30 range, and nearly 100 mIU/mL lower than those with BMI < 25 (p < 0.001).

Fig. 4.

Body mass index (BMI) and β-hCG in patients with live birth. Median serum β-hCG by day AOR 13 and 15. Boxes = interquartile range, IQR (25th percentile bottom, 75th percentile top), middle line = median (mIU/mL), whiskers = mean ± 2 standard deviations (SD), and dots = outliers. Medians between BMI groups at AOR 13 = P < 0.001 and AOR 15 = P < 0.001 by Kruskal–Wallis rank sum test; one-way ANOVA

Discussion

Despite the abundance of data on the rise of serum β-hCG to predict pregnancy outcomes, well-defined parameters to aid in accurate counseling, particularly following single embryo transfer after IVF, are limited. In this large retrospective analysis, β-hCG values that doubled over 48 h and reached 100 mIU/mL at 15 days AOR highly correlated with live birth (Fig. 1). This association persisted after controlling for both maternal and cycle level factors. When both specific parameters were met, the likelihood of live birth was over 85%. Our parameters highly correlated such that 90% of the patients whose β-hCG doubled also had levels exceed 100 mIU/mL by 15 days AOR. This strong correlation explains why there was no additional increase in OR in our imputation model when both parameters were achieved. Furthermore, our trajectory models further corroborate the significant association of β-hCG and live birth. In particular, our spline models provide greater power and accuracy to identify points where trajectories differ for cycles ending in live birth and other outcomes.

Our findings corroborate several recent studies in which the initial β-hCG, in addition to its rise, predicts live birth in IVF pregnancies [8, 12, 19, 32]. Prior studies revealed that mean initial β-hCG values ranging from 30 to 386 mIU/ml by 12 to 14 days AOR were associated with a 73 to 91% probability of live birth [8, 12, 14, 23, 24, 32]. However, many of these previous studies only focus on specific types of cycles (i.e., fresh or frozen), include multiple embryo transfers, and the majority examine an initial β-hCG over a wide range of days post embryo transfer. Moreover, our study is one of the few to examine PGT status (approximately 30% of our cycles underwent PGT), which is more reflective of IVF cycles nowadays [31, 33]. Furthermore, most studies focus on the comprehensive logarithmic rise of β-hCG in viable pregnancies rather than the initial doubling of β-hCG over 48 h. In sum, clinical interpretations of early β-hCG values on a specific day following an embryo transfer are ill-defined and lack clear criteria for patient counseling. Here we present discrete β-hCG parameters which can be easily applied and interpreted in a clinical setting.

When compared to spontaneous conception, β-hCG trends and doubling times in viable IVF pregnancies appear to be similar [17]. However, given the relatively known time of implantation following an embryo transfer, β-hCG can routinely be measured even earlier after IVF [17]. Nevertheless, among viable pregnancies conceived by IVF, multiple studies suggest variations of β-hCG levels based on cycle protocol [13, 16, 19]. Our study found no significant variation in serum β-hCG levels at day 13 and 15 AOR regardless of whether the embryo was transferred fresh or after being frozen. These findings are congruent with previous work by Reljic et. al. and Lawler et. al. which showed no difference in initial β-hCG at 13 and 15 days AOR after fresh versus frozen embryo transfer [13, 22]. In contrast, Grin et. al. and Zhao et. al. found overall higher β-hCG levels following a frozen compared to fresh embryo transfer in both nonviable and viable pregnancies, respectively [16, 19]. These discrepancies may be related to laboratory techniques or pre-transfer treatment protocol, thereby impacting trophoblastic development and endometrial maturation. One retrospective analysis opined that different culture mediums could impact maternal serum β-hCG levels given their effect on embryo growth and implantation rate [34]. Additionally, while pre-implantation trophectoderm biopsy could theoretically impact placental development and, hence, β-hCG production, we found no significant difference in β-hCG concentrations in viable pregnancies following PGT versus non-PGT cycles. This result was in concert to two recent retrospective studies suggesting that PGT does not affect serum β-hCG levels in early pregnancy [31, 33]. Overall, our findings highlight a consistent interpretation of early serum β-hCG levels to fortell live birth regardless of IVF cycle type or PGT performance.

Despite adjusting for the day of oocyte retrieval, we also found that β-hCG levels following cleavage-stage embryo transfers were significantly lower than blastocyst transfers at day AOR 13 and 15. This finding is consistent with reports by Kumback et. al. [35] and Oron et. al. [25] who both found higher initial blastocyst β-hCG levels compared to cleavage-stage transfers. Others have found conflicting results [36, 37] or even no difference by day AOR 17 [37]. However, most of these prior studies do not adjust for oocyte retrieval day which is paramount when comparing embryos at disparate developmental stages. Yet, given that patients undergoing a cleavage transfer may have a history of poor embryo quality or number, it is plausible that this contributes to overall lower β-hCG production. Furthermore, our findings may be related to different trophectoderm maturation over the two extra days in vivo for a cleavage (day 3 old) compared to a blastocyst (day 5 old) embryo. In other words, embryo growth in a laboratory is not identical to the complex milieu of the intrauterine environment. Likewise, these differences in β-hCG levels shortly following embryo transfer may eventually equalize by longer exposure to the same environment beyond AOR 17 [37]. Lastly, regarding oocyte origin, we found no difference in β-hCG levels following donor versus autologous oocyte transfer which is a novel addition to the literature.

In consideration of other confounding variables that affect serum β-hCG, maternal BMI has been demonstrated to inversely correlate with β-hCG levels in viable IVF pregnancies [29, 30, 38]. For instance, Mejia et. al. reported that a maternal weight increase of 100 lbs. was associated with a 30 to 50% reduction of early pregnancy β-hCG values [38]. This is consistent with our mixed effects model in which a higher BMI was associated with lower initial β-hCG levels at 13 and 15 days AOR in pregnancies that resulted in live birth. While the exact underlying mechanism is uncertain, the effect of maternal BMI on embryo development is likely multifactorial [30, 39, 40]. Some proposed explanations include a wider extracellular β-hCG distribution in obesity and the production of adipose-derived signaling molecules that hinder β-hCG section [41]. Nonetheless, our adjusted trajectory models signified that the strong association with β-hCG growth and live birth persisted despite BMI.

While many studies have analyzed the relationship between β-hCG levels and pregnancy outcomes after IVF, our study established highly significant parameters for the prediction of live birth after a single embryo transfer. Insofar as our institution routinely measures serum β-hCG about 8 days after an embryo transfer (which is notably 1–2 days earlier than many other facilities), we have the benefit to incorporate exceptionally early serum levels in our analysis. Our conclusions are bolstered by a large and well-described cohort, affording imputation models to adjust for confounding patient and cycle level factors. Likewise, our results are broadly generalizable as we incorporated all IVF cycle types (fresh, frozen, and cycles from donor oocytes), embryos that underwent PGT, and both cleavage- and blastocyst-stage embryos. Therefore, this is a unique addition to the literature as the majority of prior studies focus on one or two IVF types. The main limitation of this study is its retrospective nature which limits the ability to control for data collection at the time of patient care; our missing data rates ranged from < 1 to 28%. However, we accounted for this using an advanced imputation model which incorporated m = 20 imputed data sets in our final model. We also performed additional chart review to verify descriptive data. Another limitation is that the data came from a single institution. Although Northwestern Medicine serves an urban population with a large suburban catchment base, most of our patients are Caucasian. While our data on race/ethnicity was too small to form a substantial conclusion, we did not find an association of race on β-hCG trajectory and live birth. Therefore, race may still impact early serum β-hCG levels. For example, while there are no studies in the IVF population, Barnhart et. al. found that African American women had a more abrupt rise in β-hCG levels compared to other ethnicities in spontaneously conceived pregnancies [11].

In summary, initial serum β-hCG values and their trajectory are trustworthy and early predictors of live birth following a single embryo transfer. Despite the multitude of diverse studies that have characterized β-hCG levels and pregnancy outcomes, precise criteria to predict live birth and improve counseling after IVF are limited. While our results may not change clinical practice, using our parameters of “doubling” and “reaching 100” by 15 days AOR may improve forecasting pregnancy outcomes and assuage patient anxiety as early as possible after embryo transfer.

Supplementary information

Below is the link to the electronic supplementary material.

Abbreviations

β-hCG

Human chorionic gonadotropin

AOR

After oocyte retrieval

ART

Assisted reproductive technology

IVF

In vitro fertilization

PGT

Preimplantation genetic testing

BMI

Body mass index

SD

Standard deviation

SE

Standard error

B

Unstandardized regression coefficient

CI

Confidence interval

IQR

Interquartile range

OR

Odds ratio

Author contribution

Lydia Hughes and Lia Bernardi designed the study and collected the data. Adrienne Schuler also collected the data. Jacob Michael Schauer and Maxwell Sharmuk performed the data analysis and designed the figures. Lydia Hughes and Lia Bernardi wrote the manuscript. Adrienne Schuler, Jacob Michael Schauer, Maxwell Sharmuk, Mary Ellen Pavone, and Lia Bernardi assisted with manuscript editing. All authors participated in the discussion of the analysis and interpretation of data, in addition to reading and approving the final manuscript.

Funding

This study received statistical support for the data analysis by internal statistical support grants for the Biostatistics Collaboration Center at Northwestern University. The contribution from Jacob Michael Schauer was supported by this grant.

Declarations

Ethics approval

This research study was conducted retrospectively from data obtained for clinical purposes. We consulted with the Institutional Review Board (IRB) of Northwestern University who determined that our study did not need ethical approval. An IRB approval was granted to STU00213389 in September 2020.

Competing interests

The authors declare no competing interests.