Patient population

Approval was obtained from the Partners Health Care Institutional Review Board. According to the Institutional Review Board at our center, use of discarded blood for research does not require patient consent. Records were reviewed for all women who underwent fresh or frozen autologous IVF/intracytoplasmic sperm injection (ICSI) cycles with a single blastocyst transfer between January 2017 and January 2018. Only fresh or frozen day 5 embryos were included, whereas blastocysts that had been frozen on day 6 in a prior cycle were excluded. Cycles were included only if the first serum hCG level was positive (>3 mIU/mL), and only 1 gestational sac was detected on ultrasound in cycles resulting in clinical pregnancy. Cycles resulting in any ultrasound evidence of multiple pregnancy (>1 gestational sac) and patients with current or prior gestational trophoblastic neoplasia were also excluded. Exclusion criteria also included body mass index >35 kg/m ² (as obese patients have lower total hCG levels), cycles involving transfer of multiple embryos, cleavage-stage embryos, and/or embryos biopsied for preimplantation genetic testing (all of which prior studies have suggested may affect initial hCG values), and those involving donor oocytes, gestational carriers, to be consistent with prior studies. Patients with recurrent pregnancy loss and uterine factor infertility were also excluded.

Stimulation protocols included those using gonadotropin-releasing hormone (GnRH) antagonists, downregulation protocols using GnRH agonists, and poor-responder protocols using low-dose GnRH agonist flare or estradiol priming. Frozen cycles were supported with oral estradiol (3 mg twice per day), titrated to a minimum serum level of 200 pg/mL. Luteal support was provided in fresh cycles using daily applications of vaginal micronized progesterone gel (Crinone 8%, Actavis Pharma, Inc, Parsippany, NJ), whereas frozen cycles were supported with 50 mg of intramuscular progesterone in sesame oil, adjusted for a target minimum serum level of 20 ng/mL per protocol at our center. Luteal support was continued in ongoing pregnancies through 10 weeks of gestation.

Clinical and laboratory data were extracted directly from our electronic medical record. Embryo quality parameters were among the data collected. Good-quality blastocysts (day 5 embryos) were defined as hatching or hatched blastocysts with fair- or good-quality inner cell mass and trophectoderm. Blastocysts were graded using a modified version of the algorithm published by Gardner and Schoolcraft ; blastocysts were scored on their expansion and hatching on a scale of 4–9. A score of 4 represents an early blastocyst, in which less than half of the embryo volume is occupied by the blastocoel cavity, whereas a score of 9 represents the most advanced blastocyst, in which the blastocyst has completely hatched from the zona pellucida. In embryos with a score of 5 (blastocoel cavity occupying more than half of the embryo volume) or more, the inner cell mass (which develops into the fetus), and trophectoderm, which develops into the placenta) were scored on a scale of A−D. For the inner cell mass, the cell number and density of packing decreased from A (many, tightly packed) to D (no evidence of an inner cell mass). Regarding the trophectoderm, the cell number and cohesion decreased from A (many, cohesive cells) to D (very few, if any trophectoderm cells). Embryos were scored by 1 of 6 embryologists who are regularly assessed for precise and accurate embryo assessment.

Biologic samples and laboratory testing

The total hCG assay was performed at Brigham and Women’s Hospital for all patients, on day 11 after embryo transfer. Per clinic protocol, 3 serial hCG measurements were obtained at 2-day intervals, followed by an ultrasound at approximately 8 weeks of gestation, at which point the patient’s care was transferred to an obstetrician.

The Brigham and Women’s clinical laboratory is inspected by the Joint Commission with Clinical Laboratory Improvement Amendments (CLIA) certification. The electrochemiluminescence immunoassay was run using the Cobas e602 platform (Roche Diagnostics, Mannheim, Germany). Inter- and intra-assay coefficients of variation for the hCG assay were <4%, and results were standardized against the fourth International Standard for Chorionic Gonadotropin from the National Institute for Biological Standards and Control (NIBSC) 75/589. hCG was reported as milli-international units per milliliter (mIU/mL). Discarded serum from eligible patients was collected by the Brigham and Women’s Specimen Bank, aliquoted, and frozen to −80 ^o C within 48 hours of collection. In the interim, samples were refrigerated. Blinded samples were sent in 3 batches to Quest Diagnostics (Quest Diagnostics, Nichols Institute, San Juan Capistrano, CA) by courier on dry ice. hCG-H was measured using an electrochemiluminescence assay in 96-well plates, which uses a hCG-H−specific antibody (B152) as the coating antibody with an hCG-β specific antibody as the labeled antibody, with a lower limit of detection of <1.0 μg/L with <1% cross-reactivity to hCG, luteinizing hormone, or follicle-stimulating hormone. Each of the 3 batches was run in a single plate, to reduce variability. This assay has coefficients of variation (intra- and interassay) of <12%. hCG-H was reported as nanograms per milliliter (ng/mL). Quality control measures (for low-, mid-, and high-range values) were included within each run, per the laboratory, and blinded controls (selected from among the study patients) were included across sets, to assess interassay (interday) variability. According to this laboratory, hCG-H levels in serum are reliable in samples at room temperature for up to 7 days, in refrigerated samples for up to 7 days, and in frozen samples for up to 10 months. Samples were frozen for a mean of 4.6 months (range, 0.8–9.7 months).

Statistical analysis

The primary outcomes of interest were hCG-H and hCG measured 11 days after blastocyst transfer. The ratio of hCG-H to hCG was also analyzed. The primary endpoint of interest was ongoing pregnancy at 8 weeks of gestation. Pregnancy outcomes also included clinical pregnancy loss (defined as failure to visualize a previously documented intrauterine gestational sac by ultrasound), biochemical pregnancy loss (ie, no documentation of an intrauterine pregnancy by ultrasound), and ectopic pregnancy (defined as negative endometrial sampling followed by persistent hCG elevation, ultrasound visualization of an extrauterine gestational sac, and/or surgical pathology confirming ectopic pregnancy).

Multivariable linear regression was used to assess differences in mean hCG and hCG-H among patients with ongoing pregnancy as compared to those with failed pregnancies, using robust “sandwich” standard errors. Differences were reported as linear regression β values with 95% confidence intervals (CI), representing differences in mean analyte values between endpoints. Poisson regression was used to compare differences in hCG-H:hCG ratios according to pregnancy endpoint and to generate risk ratios (RR) and 95% confidence intervals. Age of the patient at the time of oocyte retrieval was included in the models a priori. Covariates tested as potential confounders of the relationship between hCG-H and pregnancy endpoints included gravidity and parity, BMI, prior spontaneous abortion, race/ethnicity, infertility diagnosis, fresh vs frozen embryo transfer, use of ICSI, sperm source (ejaculated vs testicular biopsy), embryo quality and expansion, and inner cell mass and trophectoderm quality. The addition of these covariates separately to the base model did not change the effect estimates by more than 10% and were therefore not included in the final model.

Patient race/ethnicity (white, black, Asian, Hispanic, American Indian, or other), infertility diagnosis (male factor, oligoovulation, diminished ovarian reserve, unknown, other, and multiple diagnoses) were included as categorical variables. Prior spontaneous abortion, fresh vs frozen cycle, and use of ICSI were included as dichotomous variables. Patient age (years), BMI (kg/m ² ), gravidity, and parity were included as continuous variables. The sensitivity, specificity, positive predictive value (PPV) and negative predictive value (NPV) of various hCG-H cut-offs for ongoing pregnancy were calculated using raw analyte data. A Youden index, a reflection of performance of each cut-off, was calculated for each analyte, as (sensitivity – [specificity – 1]). An ideal dichotomous diagnostic test has a Youden index of 1. Receiver operating characteristic curves were generated for the accuracy of hCG-H in predicting pregnancy outcome.

The few studies investigating hCG-H as a predictor of pregnancy outcome have highly variable inclusion criteria and methods, with small sample sizes and limited applicability to the current study for the purposes of power calculations. Among these papers, the mean difference in hCG-H between groups with and without ongoing pregnancy ranged from 3000 to 5000 pg/mL (3.0–5.0 ng/mL per the units used in the present study). Assuming 80% power, a type 1 error of 5%, and an ongoing pregnancy rate of 50%, we were sufficiently powered with 115 patients to detect a significant difference within that range.

For the purposes of calculating inter- and intra-assay coefficients of variation for hCG-H, 15% of test samples (all of which were blinded) were sent in duplicate, within and among runs, to the test laboratory. The intra-assay coefficient of variation was calculated as 7.4%, whereas the interassay coefficient of variation was 9.1%.

Analyses were performed using Statistical Analysis Software version 9.3 (SAS Institute, Cary, NC).