Live births after low initial β-hCG in IVF cycles: a retrospective cohort study.

We conducted a single-centre retrospective cohort study of all pregnancies conceived following in vitro fertilization (IVF) between January 2019 and December 2022 at Mount Sinai Fertility, Toronto, Canada. Research ethics approval was granted by the Mount Sinai Hospital Research Ethics Board (23–0054-C). Patients who achieved pregnancy, defined by a positive β-hCG (≥ 3 mIU/mL), after a frozen or fresh embryo transfer cycle were included. Both autologous and donor oocyte conceptions were included. If patients had more than one pregnancy within the timeframe, each pregnancy was assessed and recorded individually. Exclusion criteria encompassed patients who did not undergo embryo transfer and those with an undetectable β-hCG post-transfer. Demographic data were collected from the patient’s electronic medical record, including age at oocyte retrieval, body mass index (BMI), reason for IVF treatment, gravidity, parity, and duration of infertility in months. Cycle characteristics collected were fresh or frozen embryo, method of fertilization (conventional IVF vs. intracytoplasmic sperm injection (ICSI), endometrial preparation protocol, number of embryos transferred, use of preimplantation genetic testing for aneuploidy (PGT-A), and developmental stage of embryo transfer (cleavage stage or blastocyst day 5–7). These variables were selected based on clinical relevance and existing literature demonstrating their association with live birth outcomes [ 17 , 18 ]. Serum β-hCG levels were drawn at 14 days post-fertilization. Measurements were largely performed internally at the hospital laboratory using the Roche Elecsys β-hCG assay on the Roche cobas e801 platform. External β-hCG testing was conducted at two other centres using the same assay, and infrequently at a third external laboratory using the Abbott Alinity Total β-hCG assay. A low initial serum β-hCG was defined as ≤ 50 mIU/mL, consistent with the lowest thresholds commonly reported in the literature [ 9 , 13 , 19 – 21 ]. Reproductive outcomes included biochemical pregnancy loss: the spontaneous decrease in β-hCG before confirmation by ultrasound or histopathology; clinical pregnancy loss: ultrasound or histopathologic confirmation of a gestational sac; ectopic pregnancy: implantation outside of the uterus; resolved pregnancy of unknown location: resolved elevated β-hCG without visualization of the gestation inside the uterus; therapeutic abortion: termination of pregnancy; and ongoing clinical pregnancy: pregnancy at 10 weeks’ gestation with ultrasound confirmation of fetal heartbeat. Birth data were sourced from the Better Outcomes Registry & Network (BORN) Ontario, with pregnancy outcomes categorized into live birth term delivery, live birth preterm delivery, late clinical pregnancy loss, and intrauterine fetal demise. Live births were defined as delivery of a fetus after 20 weeks’ gestation with confirmation of life, with term gestation reflecting birth after 37 weeks. Late clinical pregnancy loss was defined as spontaneous abortion after 10 weeks of gestation. Intrauterine fetal demise was defined as the loss of a fetus after 20 weeks’ gestation. Sonographic information was collected, including the number of gestational sacs, the number of embryos detected, and the number of fetal heartbeats. The primary outcome of the study was live birth. Patient and cycle characteristics were analyzed using means with standard deviations or medians and interquartile ranges for continuous variables, and frequencies and percentages for categorical variables. Univariable and multivariable logistic mixed-effects regression models were created to determine the association between patient and cycle factors and live birth outcomes, with results reported as odds ratios (OR) with 95% confidence intervals and p-values (significance at p  < 0.05). Receiver operating characteristic (ROC) curves were generated to evaluate the discriminatory performance of both the initial β-hCG levels and the trend of β-hCG levels for live birth outcomes. For each analysis, sensitivity, specificity, positive predictive value and negative predictive value were calculated, and cut-points were defined using the Youden index to optimize the balance of sensitivity and specificity, with an additional cut-point selected to maximize sensitivity. The area under the curve (AUC) with 95% confidence intervals was calculated to assess diagnostic accuracy, with an area of 1.00 indicating a perfect test. Additional sub-group analyses evaluated β-hCG dynamics in relation to ectopic pregnancy and embryo ploidy.

The median age at the time of egg retrieval was 34.0 [31.0, 37.0]. The median BMI was 24.5 kg/m 2 [21.6, 28.2]. Patients reported several reasons for seeking IVF treatment, the most common of which were male factor (28.7%), unexplained infertility (21.2%), and advanced maternal age (16.7%) (Table  1 ). Table 1 Characteristics of study population from MSF who underwent IVF between January 2019 and December 2022 Characteristic Categories Cohort (n) 2443 Patient's age at time of egg retrieval, median years [IQR] a 34.0 [31.0, 37.0] BMI, median kg/m 2 [IQR] b 24.5 [21.6, 28.2] Oocyte source, n (%) Autologous 2140 (87.6) Donor 274 (11.2) Reciprocal IVF 29 (1.19) Obstetric history, median n [IQR] Gravidity 1.00 [0.00, 2.00] Parity 0.00 [0.00, 1.00] Prior pregnancy losses 0.00 [0.00, 1.00] Duration of infertility, median months [IQR] 19.0 [9.00, 32.0] Reasons for seeking treatment, n (%) Male factor 702 (28.7) Unexplained infertility 608 (24.9) Advanced female age 518 (21.2) Diminished ovarian reserve 408 (16.7) Tubal factor 255 (10.4) Polycystic ovary syndrome (PCOS) 238 (9.74) Endometriosis 194 (7.94) No male partner 128 (5.24) Insemination method, n (%) Conventional IVF 806 (33.0) ICSI 1631 (66.8) IVF and ICSI 6 (0.25) Embryo transfer type, n (%) Frozen 1897 (77.7) Fresh 546 (22.4) Endometrial preparation, n (%) Ovulatory FET 367 (15.0) Programmed FET 1485 (60.7) Modified ovulatory FET 45 (1.84) Fresh embryo transfer 546 (22.3) PGT-A performed, n (%) 624 (25.5) PGT-A result of transferred embryo, n (%) Euploid 601 (96.3) Mosaic 13 (2.08) No DNA 8 (1.28) Inconclusive 2 (0.32) Stage of embryo transfer, n (%) Cleavage 88 (3.60) Blastocyst day 5 1678 (68.7) Blastocyst day 6 642 (26.3) Blastocyst day 7 21 (0.86) Blastocysts day 5 and 6 13 (0.53) Blastocysts day 6 and 7 1 (0.04) Embryos transferred, n (%) Single embryo transfer 2185 (89.4) Double embryo transfer 257 (10.5) Triple embryo transfer 1 (0.04) Initial serum β-hCG, median mIU/mL [IQR] 130 [65.0, 216] BMI Body mass index, FET Frozen embryo transfer, ICSI Intracytoplasmic sperm injection, IQR Interquartile range, IVF In vitro fertilization, PGT-A Preimplantation genetic testing for aneuploidy. a Missing age from 9 patients (0.37%). b Missing BMI from 129 patients (5.28%). Characteristics of study population from MSF who underwent IVF between January 2019 and December 2022 BMI Body mass index, FET Frozen embryo transfer, ICSI Intracytoplasmic sperm injection, IQR Interquartile range, IVF In vitro fertilization, PGT-A Preimplantation genetic testing for aneuploidy. a Missing age from 9 patients (0.37%). b Missing BMI from 129 patients (5.28%). In our cohort, the majority of patients ( n  = 2140; 87.6%) used autologous oocytes for embryo transfer, while 274 patients used donor oocytes (11.2%) and 29 patients used reciprocal IVF oocytes (1.19%). 1897 (77.7%) patients underwent frozen embryo transfer, and 546 (22.4%) underwent fresh transfer. Among transferred embryos, 1678 (68.7%) were day 5 blastocysts, 642 (26.3%) day 6 blastocysts, 21 (0.86%) day 7 blastocysts, and 88 (3.60%) were at the cleavage stage. There were 624 (25.5%) embryos screened with PGT-A before transfer. The full results are presented in Table  1 . The median initial β-hCG level was 130 mIU/mL [65.0, 216]. Initial β-hCG was intended to be collected at 14 days post-fertilization; the vast majority of the patient population followed this standardized collection time. Out of the 2443 embryo transfers, 1575 pregnancies (64.5%) were ongoing at or beyond 10 weeks’ gestation (Table  2 ). The remaining pregnancies resulted in biochemical pregnancy loss ( n  = 483, 19.8%), clinical pregnancy loss ( n  = 321, 13.1%), resolved pregnancy of unknown location (41, 1.68%), ectopic pregnancy ( n  = 18, 0.74%), or unknown ( n  = 5, 0.20%). Among the 1575 ongoing pregnancies, 1179 (74.6%) resulted in a live term delivery (≥ 37 weeks’ gestation), 133 (8.42%) in a live preterm delivery ( 10 weeks’ gestation) occurred in 64 pregnancies (2.62%), intrauterine fetal demise (> 20 weeks’ gestation) in 24 (0.98%), and 16 (0.66%) underwent therapeutic abortion. There were 132 (5.40%) with unknown outcomes beyond 10 weeks’ gestation, mainly due to incomplete delivery data in the BORN Ontario after discharge from the MSF clinic at 10 weeks or delivery outside Ontario, Canada. Pregnancy outcomes are reflected in Table  2 and Table  3 . Table 2 Pregnancy outcomes at 10 weeks and delivery for patients stratified by initial serum β-hCG levels Characteristic Categories All study pregnancies Pregnancies with initial serum β-hCG levels ≤ 50 mIU/mL Pregnancies with initial serum β-hCG levels > 50 mIU/mL (n) 2443 491 1952 Pregnancy outcome at 10 weeks’ gestation, n (%) Biochemical pregnancy loss 483 (19.8) 320 (65.2) 163 (8.35) Clinical pregnancy loss 321 (13.1) 70 (14.3) 251 (12.9) Ectopic pregnancy loss 18 (0.74) 13 (2.65) 5 (0.26) Ongoing pregnancy 1512 (61.9) 76 (15.5) 1436 (73.6) Resolved pregnancy of unknown location 41 (1.68) 12 (2.44) 29 (1.49) Vanishing twin ongoing pregnancy 10 (0.41) - 10 (0.51) Viable twin gestation 51 (2.09) - 51 (2.61) Viable triplet gestation 2 (0.08) - 2 (0.10) Unknown outcomes 5 (0.20) - 5 (0.26) Pregnancy outcome at delivery, n (%) Live birth term delivery (> 37 weeks) 1179 (48.3) 50 (10.2) 1129 (57.8) Live birth preterm delivery ( 10 weeks’ gestation,  20 weeks’ gestation) 24 (0.98) 3 (0.61) 21 (1.08) Therapeutic abortion 16 (0.66) 1 (0.20) 15 (0.77) Unknown outcomes 132 (5.40) 8 (1.63) 124 (6.35) Table 3 Delivery outcomes among singleton and multiple gestations with initial serum β-hCG > 50 mIU/mL Delivery Outcome Singleton gestations Multiple gestations (n) 1446 53 Live birth term delivery (> 37 weeks) 1112 (76.9) 17 (32.1) Live birth preterm delivery ( 10 weeks’ gestation,  20 weeks’ gestation) 19 (1.31) 2 (3.77) Therapeutic abortion 15 (1.04) - Unknown outcomes 117 (8.09) 7 (13.2) Pregnancy outcomes at 10 weeks and delivery for patients stratified by initial serum β-hCG levels Delivery outcomes among singleton and multiple gestations with initial serum β-hCG > 50 mIU/mL Among patients with an initial β-hCG between 3–50 mIU/mL ( n  = 491), the most frequent outcome was biochemical pregnancy loss, affecting 65.2% ( n  = 320) of the group (Table  2 ). Clinical pregnancy losses occurred in 70 patients (14.3%), resolved pregnancy of unknown location in 12 patients (2.44%) and ectopic pregnancy in 13 patients (2.65%). Final pregnancy outcomes were available for 98% ( n  = 483) of patients with a low initial β-hCG; of these, the prevalence of live birth was 12% with 58 patients delivering live infants. The prevalence of ectopic pregnancies was disproportionately represented in lower β-hCG ranges. Of the 18 total ectopic pregnancies, 13 (72%) had an initial β-hCG ≤ 50 mIU/mL, and 16 (89%) were < 100 mIU/mL. Additionally, 61% (11/18) of ectopic cases involved patients aged 35 years or older. Only one ectopic case involved a euploid embryo. Both univariable and multivariable logistic mixed-effect regression models were used to identify predictors of live birth. With each 10 unit (mIU/mL) greater initial serum β-hCG, there was a statistically significant increase in the odds of live birth by 3.5%, OR = 1.04, [95% CI 1.03, 1.04], after adjusting for age, BMI, stage of transfer, number of embryos transferred and use of PGT-A. A positive β-hCG after transfer of a euploid embryo conferred more than twice the odds of live birth compared to untested embryos, OR = 2.24 [95% CI 1.82, 2.76]. We examined early β-hCG dynamics in euploid and untested embryo transfers. Euploid transfers were associated with significantly higher initial β-hCG values (median 144 mIU/mL [IQR 87.0, 215]) compared to untested embryos (median 125 mIU/mL [IQR 57.6, 216]). As well as a faster early β-hCG rise in euploid transfers (median daily increase 74.8% [IQR 56.2, 96.3] vs untested transfers 67.7% [IQR 44.2, 89.0]). Slower growing blastocyst transfer (day 6 and day 7) significantly lowered odds of live birth compared to day 5 blastocysts, even after adjustment (Day 6: OR = 0.70 [95% CI 0.57, 0.85]; Day 7: OR = 0.26 [95% CI 0.09, 0.74]). Lastly, the number of embryos transferred was inversely correlated with live birth (OR = 0.44, [95% CI 0.39, 0.51]). Age at oocyte retrieval, BMI and cleavage-stage transfers were not significant predictors in the multivariable models ( p  > 0.05). Results from logistic mixed-effects models are presented in Table  4 . Table 4 Univariable and multivariable logistic mixed-effects models evaluating predictors of live birth Predictor Univariable Multivariable OR 95% CI OR 95% CI Initial serum β-hCG (per 10 units) 1.04 1.03, 1.04 1.04 1.03, 1.04 Age at egg retrieval (years) 0.99 0.97, 1.00 1.00 0.98, 1.01 BMI (kg/m 2 ) 0.97 0.96, 0.99 1.00 0.98, 1.01 Stage: Blastocyst day 5 (Reference) 1.00 1.00 Stage: Blastocyst day 6 0.76 0.63, 0.91 0.7 0.57, 0.85 Stage: Blastocyst day 7 0.26 0.09, 0.72 0.26 0.09, 0.74 Stage: Cleavage 0.64 0.41, 0.99 1.10 0.69, 1.74 Number of embryos transferred 0.37 0.28, 0.49 0.44 0.39, 0.51 PGT-A (Yes) 2.37 1.95, 2.89 2.24 1.82, 2.76 Univariable and multivariable logistic mixed-effects models evaluating predictors of live birth The ROC curve analysis was conducted to evaluate the discriminatory performance of initial β-hCG for live birth outcomes. The area under the curve (AUC) was 0.726, [95% CI 0.705, 0.746], indicating acceptable discrimination between live birth and no live birth. The optimal threshold that balances sensitivity and specificity using the Youden index was an initial β-hCG of 100.5 mIU/mL (sensitivity 0.79, specificity 0.57). A secondary logistic mixed-effects analysis using β-hCG above this threshold showed significantly increased odds of live birth, after adjusting for patient and cycle characteristics (OR = 5.18 [95% CI 4.14, 6.47]) (Supplementary data). This corresponds to a 418% increase. To characterize β-hCG values below which live birth was uncommon, we performed ROC analysis prioritizing maximal sensitivity (1.00). This analysis indicated that more than 99% of live births occurred above an initial β-hCG of 22.5 mIU/mL with a NPV of 97.6%. The rate of change from initial β-hCG provided greater discriminatory value among patients with low initial β-hCG (AUC 0.876 [95% CI 0.84, 0.91]) compared to those with an initial β-hCG > 50 mIU/mL (AUC 0.584 [95% CI 0.56, 0.61]). In the low initial β-hCG subgroup, the optimal threshold of 44% daily increase resulted in significantly stronger specificity (0.72) and negative predictive value (0.99), compared to the optimal threshold of 46% in the > 50 mIU/mL subgroup (specificity 0.25, negative predictive value 0.70). The ROC curves are presented in Fig.  1 . Fig. 1 ROC curves for initial β-hCG and β-hCG rise in predicting live birth. This figure presents four ROC curves illustrating the discriminatory ability of β-hCG measures for live birth outcomes. A ROC curve for live birth: initial β-hCG (full cohort) (Threshold = Youden optimal cut off). B ROC curve maximizing sensitivity for live birth: initial β-hCG (full cohort). C ROC curve for live birth: rate of change in β-hCG (initial ≤ 50 mIU/mL). D ROC curve for live birth: rate of change in β-hCG (initial > 50 mIU/mL) ROC curves for initial β-hCG and β-hCG rise in predicting live birth. This figure presents four ROC curves illustrating the discriminatory ability of β-hCG measures for live birth outcomes. A ROC curve for live birth: initial β-hCG (full cohort) (Threshold = Youden optimal cut off). B ROC curve maximizing sensitivity for live birth: initial β-hCG (full cohort). C ROC curve for live birth: rate of change in β-hCG (initial ≤ 50 mIU/mL). D ROC curve for live birth: rate of change in β-hCG (initial > 50 mIU/mL) In mixed-effects logistic regression models evaluating predictors of ectopic pregnancy, the daily rate of β-hCG change was not significantly associated with ectopic pregnancy in either univariable or multivariable analyses. In the primary model comparing ectopic pregnancy with intrauterine pregnancy, daily percent change in β-hCG showed no association with ectopic pregnancy (OR = 1.01 [95% CI 0.96, 1.07]) (Table  5 ). Higher initial β-hCG levels were associated with lower odds of ectopic pregnancy in univariable analysis (OR = 0.68 [95% CI 0.58, 0.80]), but this association was not significant after adjustment (Table  5 ). To improve clinical comparability, a restricted analysis was performed comparing ectopic pregnancy with ongoing singleton pregnancies meeting viability criteria (one gestational sac, one embryo, and one fetal heart) (Table  6 ). In this restricted model, the rate of β-hCG change remained non-significant (OR = 0.99 [95% CI 0.98, 1.00]). In contrast, each 10-unit increase in initial β-hCG was independently associated with an approximately 38% reduction in the odds of ectopic pregnancy (OR = 0.62 [95% CI 0.51, 0.75]). The transfer of a PGT-A tested embryo was also independently associated with lower odds of ectopic pregnancy in the restricted model (OR = 0.10 [95% CI 0.01, 0.96]). ROC curve analyses demonstrated poor discriminatory performance of β-hCG rise for ectopic pregnancy. The AUC was 0.54 [95% CI 0.37, 0.72] in the primary analysis and 0.55 [95% CI 0.38, 0.73] in the restricted analysis. Table 5 Univariable and multivariable logistic mixed-effects models evaluating predictors of ectopic pregnancy compared with intrauterine pregnancy Predictor Univariable Multivariable OR 95% CI OR 95% CI Initial serum β-hCG (per 10 units) 0.68 0.58, 0.80 0.04 0.00, 0.60 β-hCG change per day (%) 1.00 0.98, 1.01 1.01 0.96, 1.07 PGT-A (Yes) 0.16 0.02, 1.21 0 0.00, 0.00 Table 6 Univariable and multivariable logistic mixed-effects models evaluating predictors of ectopic pregnancy compared with ongoing singleton pregnancy (one gestational sac, one embryo, one fetal heart) Predictor Univariable Multivariable OR 95% CI OR 95% CI Initial serum β-hCG (per 10 units) 0.65 0.54, 0.77 0.62 0.51, 0.75 β-hCG change per day (%) 0.99 0.98, 1.01 0.99 0.98, 1.00 PGT-A (Yes) 0.12 0.02, 0.94 0.10 0.01, 0.96 Univariable and multivariable logistic mixed-effects models evaluating predictors of ectopic pregnancy compared with intrauterine pregnancy Univariable and multivariable logistic mixed-effects models evaluating predictors of ectopic pregnancy compared with ongoing singleton pregnancy (one gestational sac, one embryo, one fetal heart)

The emotional, physical, and financial burden of in vitro fertilization (IVF) often leaves patients in a state of stress surrounding the outcome of their treatment [ 1 ]. When compared to unassisted pregnancies, patients who conceive through IVF exhibit higher levels of anxiety and depression symptoms [ 2 – 5 ]. Anxiety tends to increase with each stage, with the highest levels observed after embryo transfer to the pregnancy test [ 6 , 7 ]. Effective counselling is imperative for patient well-being, particularly for patients undergoing fertility treatments [ 8 ]. Providing information on the prognosis can prepare patients for adverse outcomes and allow them to make an informed decision about the pregnancy. Serum beta-human chorionic gonadotropin (β-hCG) hormone is a well-supported biomarker used for pregnancy testing, monitoring, and prognosis. After a successful embryo transfer and implantation, the syncytial trophoblast of the growing placenta begins to secrete β-hCG, which can be detected in the patient’s serum to confirm the pregnancy. If positive, serial β-hCG levels will be evaluated for quantity and trend as these levels provide important information about the development of the pregnancy. It is widely acknowledged that higher β-hCG levels are correlated with positive pregnancy outcomes [ 9 – 14 ]. Clinicians are currently using β-hCG levels to counsel patients on the likelihood of a live birth, particularly when the initial level is high. Comparatively, numerous studies have described a β-hCG cut-off point below which the prognosis for live birth is poor [ 15 , 16 ]. The literature lacks well-defined statistics on the prevalence of live birth following a low initial serum β-hCG and how to interpret the trend of β-hCG rise in this context. This ambiguity presents a challenge in patient counselling and decision-making in IVF treatments. At Mount Sinai Fertility (MSF), patients with an initial β-hCG ≤ 50 mIU/mL are typically counselled that their pregnancy is unlikely to be viable and to monitor for signs and symptoms of pregnancy loss and ectopic pregnancy. Due to the lack of data published in the literature, patients cannot be provided with a reliable prognosis for a healthy pregnancy when their initial level is low. Therefore, our objective was to determine the prevalence of live births after IVF with a low initial β-hCG and to identify β-hCG patterns associated with a very low likelihood of live birth. This can inform patient counselling and prevent unnecessary intervention in viable and wanted pregnancies. Unlike existing studies that focus on a single cycle type, we considered both fresh and frozen cycles, as well as cycles that employed preimplantation genetic testing for aneuploidy (PGT-A). We identified patient and cycle characteristics that increase the likelihood of a positive pregnancy outcome.

A clear understanding of the prognosis associated with a low initial β-hCG is essential for early pregnancy counselling. In this retrospective analysis, among the 491 patients with an initial β-hCG level ≤ 50 mIU/mL, the prevalence of live births was 12%, resulting in 50 term and eight preterm births. The lowest level that resulted in a live birth was 6 mIU/mL, measured 14 days post-fertilization, which tripled in 48 h. Several studies have also documented live births from IVF despite very low initial β-hCG [ 22 , 23 ]. In one large retrospective cohort study, three cases of live births were reported with initial β-hCG levels of 5, 6, and 8 mIU/mL [ 14 ]. Our results demonstrate that pregnancies with a low initial β-hCG have the potential to result in a live birth. Interestingly, we found a higher proportion of ectopic pregnancies among the low β-hCG group compared to the high β-hCG group (2.65% vs 0.26%). However, none of these cases had significant risk factors for ectopic pregnancy (e.g. tubal disease, prior ectopic pregnancy). These findings underscore the importance of vigilant follow-up when early hormone levels are low. Higher initial β-hCG levels were associated with lower odds of ectopic pregnancy. However, the rate of β-hCG rise was not significantly associated with ectopic pregnancy and demonstrated poor discriminatory performance. Together, these findings suggest that while low initial β-hCG may warrant closer surveillance, early β-hCG trajectory alone does not reliably distinguish ectopic from intrauterine pregnancy in this cohort. PGT-A is widely recognized for its role in selecting viable embryos and reducing the risk of pregnancy loss [ 24 ]. Our study uniquely included PGT-A in the multivariable model for live birth. As our cohort only included pregnancies with a positive initial β-hCG, this analysis reflects outcomes after implantation rather than pre-transfer success rates. In our cohort, the use of a euploid embryo more than doubled the odds of live birth. Descriptive analyses similarly showed higher live birth rates among euploid transfers (67.9%) compared with untested transfers (46.5%). While the initial β-hCG remained a significant predictor, the transfer of a PGT-A-tested euploid embryo was independently associated with live birth after adjusting for early β-hCG levels. These results suggest that although euploid embryos demonstrate stronger early β-hCG dynamics, the advantage associated with PGT-A is not entirely captured by initial hormone levels alone. Similarly, our analysis revealed that day-5 blastocyst transfers were associated with increased odds of live birth compared with cleavage-stage or day-6 and day-7 blastocyst transfers, which is consistent with prior studies [ 10 , 20 , 25 ]. Of the 58 live births with a low initial β-hCG, over half of them were blastocyst day-5 transfers and only three cycles were transferred at the cleavage stage. In addition to the timing of transfer, we found that the increase in the number of embryos transferred was associated with a decrease in the odds of a live birth. Our result is likely attributable to confounding by indication; patients with advanced reproductive age or a history of recurrent implantation failure are more likely to undergo multiple embryo transfers, although they are also at a poorer prognosis [ 26 ]. Younger maternal age was associated with increased odds of live birth, although this association was not statistically significant. Patients with a low initial β-hCG were modestly older (34.40 ± 4.64 vs. 33.79 ± 4.60 years). In addition, advanced maternal age (≥ 40 years) was more prevalent among those with pregnancy loss at 10 weeks’ gestation, in both low and high initial β-hCG groups. This is in agreement with the literature on the impact of advanced maternal age [ 27 ]. A lower BMI appeared predictive of a live birth; however, after adjusting for other cycle and patient characteristics, its statistical significance was lost. In line with previous studies [ 28 , 29 ], among pregnancies that resulted in live births, we found that the mean maternal BMI was higher in those with a low initial β-hCG (28.22 ± 6.77 kg/m 2 ) compared with those with a high initial β-hCG (25.09 ± 5.07 kg/m 2 ). In our study, patient characteristics were associated with live birth rates in the univariable analysis, but only cycle characteristics remained significant in the multivariable model. As expected, the initial serum β-hCG was predictive of a positive pregnancy outcome; for each 10-unit greater initial β-hCG, there was a statistically significant increase in live birth rates, while adjusting for patient and cycle characteristics. Our ROC curve identified 100.5 mIU/mL (AUC 0.726) as the optimal cut-point with the greatest sensitivity and specificity for predicting live birth. Among patients with an initial β-hCG above this value, 66.4% delivered. The reliability of a high initial β-hCG in predicting positive pregnancy outcomes has been well supported [ 15 , 19 ]. The 100.5 mIU/mL level is comparable to that in the literature, with a similar sensitivity (~ 80%) [ 13 , 23 , 25 ]. The utility of a single time-point threshold for predicting positive pregnancy outcomes is limited because a range of β-hCG collection days is reported in the literature. Our study confirms live births can occur with β-hCG well below those thresholds. To better reflect the observed prevalence of live births after a low initial β-hCG, we explored a model that prioritized sensitivity. An initial β-hCG of greater than 22.5 mIU/mL captured over 99% of patients who ultimately had a live birth. Although this value should not be interpreted as a clinical threshold, it provides valuable context for counselling patients with low initial β-hCG. Live birth was extremely uncommon below this level, underscoring the low likelihood of viability. There is a lot of ambiguity regarding the prognosis of these pregnancies, with some recommending discontinuing luteal phase support if the initial β-hCG < 58.8 mIU/mL [ 21 ]. Our findings identify a β-hCG level with a wider catchment for live births to minimize the risk of incorrectly dismissing viable pregnancies. The trajectory of β-hCG rise also has prognostic relevance, beyond the absolute initial level. The doubling time of β-hCG is a well-established predictor of live birth [ 15 , 23 , 30 ]. Unique to previous studies, we performed an ROC subgroup analysis on two distinct ranges of β-hCG values. The rate of change from initial β-hCG was a stronger discriminator of live birth among patients with an initial serum β-hCG ≤ 50 mIU/mL (AUC 0.876 [0.84, 0.91]) than among those with an initial β-hCG > 50 mIU/mL (AUC 0.584 [0.56, 0.61]). In the low initial β-hCG subgroup, 99% of patients who did not meet the optimal 44% daily increase of β-hCG did not deliver a live birth. In comparison to the β-hCG > 50 mIU/mL subgroup, where only 70% of patients who did not meet the 46% threshold failed to deliver. There were no cases of live births after a decline in β-hCG. To our knowledge, this is the first study to demonstrate that the discriminatory performance of β-hCG rise is significantly greater in patients with initially low levels. Prior research has also highlighted the discriminatory value of β-hCG dynamics, with suggested thresholds of at least a 53% rise [ 15 ] or a 1.9-fold increase [ 21 ] in 48 h. Our findings align with the POPI-Plus study, a predictive clinical model for live birth [ 30 ]. The authors found that a greater 48-h rise was required to achieve the same live birth probability when the initial β-hCG level was lower, reflecting underlying physiologic variability. Our study has several limitations. Although initial β-hCG was intended to be collected at a standardized time point, minor deviations in the collection day occurred. As β-hCG rises quickly in early pregnancy, variation in sampling day can shift absolute values and may lead to misclassification around fixed thresholds, which can limit its interpretation. Accordingly, the 12% live birth prevalence reflects our cohort and may vary in centres using different collection protocols. This prevalence remains useful for counselling, acknowledging the daily rate-of-rise may be more broadly applicable as it is less sensitive to the timing of the initial sample. Most β-hCG measurements were performed using the same assay at our centre and major external laboratories, with a minority analyzed using a different assay. As the specific laboratory data of a patient’s test were not collected, inter-laboratory variation in absolute β-hCG values cannot be excluded. However, as trajectory analyses rely on changes over time, they are less likely to be affected. We also did not collect any data on the morphologic quality of embryos and their correlation with live birth outcomes. Finally, as a retrospective single-centre study, our findings may be influenced by practice patterns at our institution and should be validated across diverse clinical settings. These findings are specific to pregnancies conceived via IVF/ICSI and may not be generalizable to spontaneous conceptions or other forms of assisted reproductive technology.

Patients with detectable but low β-hCG levels represent a vulnerable population. There is less known about their prognosis because, despite increased prevalence of adverse outcomes, there are cases of live births after remarkably low levels. This poses a challenge to clinicians when counselling patients in this group. Patients often hold unrealistic expectations when starting IVF [ 31 – 33 ]. Managing expectations with objective information has been shown to reduce feelings of distress and anxiety, and aid in informed decision-making for the current cycle and subsequent cycles [ 34 ]. Our study demonstrates a 12% prevalence of live birth rates in this group, providing valuable information for setting patient expectations. By identifying 22.5 mIU/mL as a level that captured nearly all viable pregnancies, we were able to distinguish two subgroups of patients with a low initial β-hCG. This distinction supports more individualized prognostic counselling, as outcomes differ significantly between these groups. Notably, the use of a euploid embryo should inform counselling, as we saw the advantage of PGT-A is not entirely captured by the initial β-hCG level. Finally, we are able to demonstrate the unique impact of the rate of β-hCG rise on positive outcomes in this population. Our focus on live birth among patients with an initially low β-hCG was to increase transparency and address patients’ calls for greater clarity in this unfamiliar process [ 35 – 37 ]. In conclusion, our study contributes to growing evidence that a single low β-hCG value should be interpreted with careful consideration. Future directions should aim to validate our results in a multi-centre study.

Supplementary Material 1. Supplementary Material 1.