Background
Preimplantation genetic testing is commonly performed by removing cells from the trophectoderm, the outer layer of the blastocyst, which subsequently forms the placenta. Because preimplantation genetic testing removes the cells that are destined to form the placenta, it is possible that preimplantation genetic testing could be associated with an increased risk for adverse outcomes associated with abnormal placentation. Despite the increasing utilization of preimplantation genetic testing, few studies have investigated the perinatal outcomes, with published studies yielding contradictory findings and using small sample sizes.
Objective
This study aimed to compare the perinatal outcomes of singleton pregnancies conceived following frozen embryo transfer of a single, autologous blastocyst either with or without preimplantation genetic testing.
Study Design
This was a retrospective analysis of autologous frozen embryo transfer cycles that led to singleton live births per the Society for Assisted Reproductive Technology Clinical Outcomes Reporting System, including cycles initiated between 2014 and 2015. The perinatal outcomes, including birthweight, Z-score, small for gestational age, large for gestational age, macrosomia, and preterm birth, were compared between pregnancies with or without preimplantation genetic testing. We conducted multivariable linear regression analyses for the birthweight and Z-score and logistic regression for the binary outcomes. A false discovery rate was adjusted to decrease the type I error from multiple hypothesis testing.
Results
Of the 16,246 frozen embryo transfers resulting in singleton births included in this analysis, 6244 involved the transfer of a single blastocyst that had undergone preimplantation genetic testing, and the remainder (n=10,002) involved the transfer of a single blastocyst that had not undergone a biopsy. When compared with the women from the nonpreimplantation genetic testing group, the average maternal age (35.8±4.1 vs 33.7±3.9; P <.001) and prevalence of prior spontaneous abortion (37.3% vs 27.7%; P <.001) were higher among women from the preimplantation genetic testing group. Bivariate analysis revealed a higher prevalence of small-for-gestational-age newborns (4.8% vs 4.0%; P =.008) and premature delivery (14.1% vs 12.5%; P =.005) and a lower prevalence of large-for-gestational-age newborns (16.3% vs 18.2%; P =.003) and macrosomia (11.1% vs 12.4%; P =.013) among the preimplantation genetic testing pregnancies. Multivariate regression analyses, adjusting for the year of transfer, maternal age, maternal body mass index, smoking status (3 months before the treatment cycle), obstetrical histories (full-term birth, preterm birth, and spontaneous abortion), infertility diagnosis, and infant sex suggested a significantly increased odds of preterm birth (adjusted odds ratio, 1.20; 95% confidence interval, 1.09–1.33; P <.001) from preimplantation genetic testing blastocysts. Birthweight (–14.63; 95% confidence interval, –29.65 to 0.38; P =.056), birthweight Z-score (–0.03; 95% confidence interval, –0.06 to 0.00; P =.081), and odds of small-for-gestational-age newborns (adjusted odds ratio, 1.17; 95% confidence interval, 0.99–1.38; P =.066), large-for-gestational-age newborns (adjusted odds ratio, 0.96; 95% confidence interval, 0.88–1.06; P =.418), and macrosomia (adjusted odds ratio, 0.96; 95% confidence interval, 0.85–1.07; P =.427) did not differ between the frozen transfer cycles with or without preimplantation genetic testing in the analysis adjusted for the confounders. Subgroup analysis of the cycles with a stated infertility diagnosis (n=14,285) yielded consistent results.
Conclusion
Compared with frozen embryo transfer cycles without preimplantation genetic testing, the frozen embryo transfer cycles with preimplantation genetic testing was associated with a small increase in the likelihood of preterm birth. Although the increase in the risk for prematurity was modest in magnitude, further investigation is warranted.
Why was this study conducted?
This study aimed to compare the perinatal outcomes of singleton pregnancies achieved via frozen embryo transfer with or without preimplantation genetic testing (PGT).
Key findings
Singletons conceived from PGT embryos have slightly increased risks of being delivered prematurely. PGT is not associated with a reduced birthweight or macrosomia.
What does this add to what is known?
This study carefully examined the perinatal outcomes after PGT and included adjustments for important confounders. Although the magnitude of the effect of PGT on perinatal outcomes appears to be small, further study is warranted, given the widespread utilization of PGT.
Introduction
The utilization of preimplantation genetic testing (PGT) has increased rapidly in the United States over the last decade, with approximately 40% of in vitro fertilization (IVF) cycles utilizing PGT as reported to the Society for Assisted Reproductive Technology (SART) in 2018. Although it was initially utilized to screen for sex-linked or monogenic disorders, 90% of PGT in contemporary practice is applied to deselect aneuploid embryos for transfer PGT for aneuploidy detection (PGT-A) because many unsuccessful pregnancies are associated with whole chromosome aneuploidies, and a significant portion of embryos created from IVF are aneuploid, particularly with increasing maternal age. The clinical use of PGT-A has increased in the past decade despite the fact that its advantages have not been corroborated in randomized trials and that limited data are available on the possible consequences of PGT on perinatal outcomes. Although PGT, in conjunction with IVF, has been performed for almost 3 decades, it is only in recent years that PGT has been used frequently enough to begin to examine whether or not there are any effects on perinatal outcomes.
During the 3 decades of clinical experience with preimplantation embryo biopsies, the timing of the biopsy and genetic analysis methods have evolved. The most contemporary application of PGT utilizes a trophectoderm biopsy from a blastocyst. The number of cells removed appears to be quite important because high-cell-count biopsies are associated with failed implantation, miscarriage, and monozygotic twinning. If a biopsy can affect the immediate embryo transfer outcomes, there is a possibility that performing a biopsy may also affect the pregnancy process and subsequent perinatal health outcomes. Prenatal and perinatal morbidities, including premature rupture of membranes, preterm birth, and intrauterine growth restriction, have been attributed to aberrant implantation or placentation as documented in the existing literature. , It is therefore logical to hypothesize that removing a portion of the placental precursor could theoretically increase the risk of abnormal placentation.
To date, a limited number of studies have evaluated the perinatal outcomes from PGT pregnancies. , The limitations of many of these studies include small sample sizes, a lack of control for confounders now known to affect perinatal outcomes such as fresh vs frozen transfer, and older PGT techniques that are less common in the contemporary practice of IVF. Given the increasing utilization of PGT, more perinatal outcome data are clearly needed. This study aims to compare the initial neonatal outcomes of singleton pregnancies conceived via frozen transfer cycles of previously biopsied blastocysts vs blastocysts that had not undergone a biopsy.
Materials and Methods
Study population
This retrospective cohort study was performed using the SART Clinical Outcomes Reporting System (CORS) database, which contains data that were collected and verified by the SART and reported to the Centers for Disease Control and Prevention in compliance with the Fertility Clinic Success Rate and Certification Act of 1992 (Public Law 102-493). The SART CORS data are validated annually with some clinics having on-site visits for chart review on the basis of an algorithm for clinic selection. During each visit, data reported by the clinic are compared with the information recorded in the patients’ charts. Of 11 data fields selected for validation, 10 were found to have discrepancy rates of ≤5%.
In this study, only frozen embryo transfers (FETs) that occurred in the period between 2014 and 2015 and in which only 1 embryo created with autologous oocytes was transferred and a singleton pregnancy was delivered, were included. Embryos created using donor oocytes were excluded. We focused on FETs on the basis of the documented increased risk of low birthweight in fresh compared with frozen transfers. , All cycles in our study sample were blastocyst transfers. We limited the analysis to singleton live births because multiple gestation has a profound independent effect on the birthweight and gestational age at delivery. We did not restrict the analysis to cycles designated as “elective single embryo transfer,” which is a field that a fertility program is supposed to select if there is more than 1 embryo suitable for transfer, but the program has chosen to transfer only 1 embryo. We included all single-embryo transfers (not only elective) because we aimed to maximize the number of cycles available for analysis and because we are unaware of any data that suggest that the perinatal outcome should differ between elective and nonelective single-embryo transfers.
Among these FET cycles, cycles with PGT were not further restricted by the particular PGT indications; thus, cycles indicated for aneuploidy, single-gene disorder, sex selection, and human leukocyte antigen (HLA) determination were all eligible. Cycles were excluded if it was unclear whether the embryo transferred underwent PGT, or if the newborn sex, birthweight, and gestational age at delivery were unknown because of the inability to calculate a Z-score if these data were missing. In addition, if potential confounders such as maternal age, gravidity, or body mass index (BMI) were unknown, that cycle was excluded. Cycles with a gestational age outside of those in the study by Talge et al were excluded because of the inability to calculate a Z-score. Cycles with implausible Z-scores, defined as a Z-score of <–4 or >4, were excluded as suggested in the study by Baker et al, which also analyzed cycles from the SART CORS. This cutoff identified cycles with an implausible combination of birthweight and gestational age. The final cohort included 16,246 frozen cycles.
Predictor and key outcomes
The status of whether an embryo underwent PGT was the predictor and coded as a dichotomized variable (yes vs no). The key outcomes included continuous birthweight in grams and Z-score. The birthweight Z-score was calculated using a US sex-specific birthweight reference, in accordance with previously described methods, as a measure of the deviation of the birthweight of an individual from the median value of the reference population, divided by the standard deviation of the reference population, to control for both sex and gestational age at delivery. Dichotomized outcomes included small-for-gestational-age (SGA) newborn (Z-score of ≤–1.28 vs >1.28), large-for-gestational-age (LGA) newborn (Z-score of ≥1.28 vs <1.28), macrosomia (birthweight of ≥4000 g vs <4000 g), and preterm birth (gestational age at delivery of <37 vs ≥37 weeks).
Covariates
Considering the potential confounding effects on the association between the perinatal outcomes and PGT, we selected the following factors as the covariates: year of transfer (2015 vs 2014), maternal age at cycle start (continuous), maternal BMI (continuous), smoking in 3 months before the treatment cycle (yes vs no), obstetrical history as full-term birth (nulligravida or no full-term birth vs having had a full-term birth), preterm birth (nulligravida or no preterm birth vs having had a preterm birth), previous experience of spontaneous abortion (nulligravida or no spontaneous birth vs having had a spontaneous abortion), and infertility (male factor, endometriosis, polycystic ovarian syndrome [PCOS], diminished ovarian reserve, tubal factors, uterine, unexplained, ≥2 diagnoses, or other). Because the obstetrical history was only obtained for women who had ever been pregnant, we incorporated gravidity in each binary indicator of obstetrical history.
Data analysis
For the cases with missing information about the smoking status (12.09%) and obstetrical history (full-term birth, 0.01%; preterm birth, 0.19%; spontaneous abortion, 0.03%), the most frequent category for each variable (ie, smoking: no; full-term birth, preterm birth, and spontaneous abortion: nulligravida or no such experiences) was imputed.
We initially conducted a descriptive analysis comparing the cohort characters and outcome distributions by PGT status. The continuous variables were compared by using a Student t test or Wilcoxon rank-sum test contingent on normality and equal variance. The categorical variables were compared using a chi-square test. A separate regression model was constructed for each outcome.
We subsequently performed bivariate and multivariable linear regressions across the cohort for the continuous outcomes (birthweight and Z-score), and logistic regressions for the dichotomized outcomes (SGA, LGA, macrosomia, and preterm birth). Different sets of confounders were adjusted for in the statistical models depending on the outcome being examined as the dependent variable. In addition to a list of the core covariates (year of transfer, maternal age, BMI, smoking, obstetrical history, and infertility diagnosis), the gestational age at delivery and infant sex were additionally adjusted for in the regression models analyzing birthweight (g) and macrosomia; for the model with preterm birth as the outcome, the infant sex was controlled for in addition to the shared core covariates. Considering the differential effects of the offspring’s sex on the birthweight, macrosomia, and preterm birth, we tested the interaction between PGT and infant sex to assess if sex modified the associations. Because none of these interactions was statistically significant based on our overall data analyses, we did not include the interaction term in the final regression model for each of these outcomes.
We further performed subanalyses to confirm the initial findings. Specifically, we identified cycles with a stated infertility diagnosis (ie, male factor, endometriosis, PCOS, diminished ovarian reserve, tubal factors, uterine factors, unexplained, or ≥2 diagnoses) and those without, and tested hypotheses within each subset of the study cycles ( Supplemental Table 1 for cycles without stated infertility diagnosis). In the multivariable logistic regression model for the outcome SGA among cycles without a stated infertility diagnosis, we did not adjust for the smoking status because none of the smokers (n=52) delivered an SGA. To account for the increased likelihood of obtaining a P value of less than .05 from multiple statistical tests (n=18), we applied the Benjamini-Hochberg procedure to control for a false discovery rate at a value of .05. The hypothesis testing was only considered statistically significant if the P value was less than the corresponding Benjamini-Hochberg critical value.
Among the 16,246 cohort cycles, 69 women each had completed 2 cycles. Considering a correlation between the repeated cycles, we additionally excluded these 138 cycles from the analyses, and the results were consistent with the main findings ( Supplemental Tables 2 , 3 , and 4 ). We further compared the perinatal outcomes from the PGT cycles for aneuploidy with those exclusively for single-gene screening ( Supplemental Table 5 ) because women from the latter group were more likely to be fertile and possibly had better pregnancy outcomes. A comparison between the outcomes from the PGT cycles with a stated infertility diagnosis and those without were also conducted ( Supplemental Table 6 ).
Two-sided P values less than .05 were considered statistically significant. Statistical analysis was conducted using Stata SE 15.1 (StataCorp LLC, College Station, TX).
Ethics
The study was approved by the Stanford University Institutional Review Board (Stanford IRB-30443).
Results
A total of 19,642 assisted reproductive technology (ART) frozen cycles that were recorded in the SART CORS were initially included. These cycles were conducted in the period between 2014 and 2015 and utilized a single-embryo transfer, which resulted in a singleton live birth, with complete information on the PGT status, gestational age, infant birthweight, and infant sex. Of these, 3396 cycles were excluded from the analysis owing to missing values on the maternal BMI (n=3320), gravidity (n=5), and gestational duration outside of the reference range provided in the study by Talge et al (n=18). Of the remaining cycles, singletons with an implausible Z-score (ie, a Z-score of <–4 or >4) (n=53) were additionally excluded. The final analytical sample included 16,246 eligible cycles ( Figure ).
The study sample characteristics are summarized in Table 1 . Of the eligible 16,246 cycles, more than one-third (38.4%, 6244/16,246) underwent PGT. Among the PGT cycles, 5.6% (350/6244) underwent the biopsy exclusively for single-gene disorder screening. Women from the PGT group were older at the initiation of the treatment cycle (35.8±4.1 vs 33.7±3.9; P <.001). The BMI was within the reference range for both groups, but women who underwent PGT had a slightly lower mean BMI (24.2±4.8 vs 24.8±5.2; P <.001). Although fewer women from the PGT group reported smoking (1.9% vs 4.1%; P <.001), most of the women in the cohort did not smoke. Approximately one-third (5101/16,246) of all the women had not previously been pregnant. A significantly lower proportion of women from the PGT group had 1 to 3 pregnancies before the present treatment cycle (53.9% vs 64.6%; P <.001). Approximately one-third of the women with prior pregnancies had full-term births (37.0%) or spontaneous abortions (31.4%), with full-term births being more prevalent among the PGT group (39.8% vs 32.7%; P <.001) and spontaneous abortion being more common among the non-PGT group (37.3% vs 27.7%; P <.001). The prevalence of preterm birth history was relatively low and more significant among the non-PGT group (11.1% vs 6.3%; P <.001). Regardless of the PGT status, the male factor (22.5%) was a common reason for IVF treatment, and patients commonly had at least 2 infertility diagnoses (21.0%). The etiology of the infertility overall differed between the PGT and non-PGT groups ( P <.001).
| Characteristics | Overall (N=16,246) | PGT (–) (n=10,002) | PGT (+) (n=6244) | P value |
|---|---|---|---|---|
| Treatment year, n (%) | ||||
| 2014 | 7486 (46.1) | 5094 (50.9) | 2392 (38.3) | <.001 |
| 2015 | 8760 (53.9) | 4908 (49.1) | 3852 (61.7) | |
| Maternal age (y), mean±SD | 34.6±4.1 | 33.7±3.9 | 35.8±4.1 | <.001 |
| BMI (kg/m 2 ), mean±SD | 24.5±5.0 | 24.8±5.2 | 24.2±4.8 | <.001 |
| Smoking in 3 mo before treatment cycle, n (%) | ||||
| No | 15,720 (96.8) | 9596 (95.9) | 6124 (98.1) | <.001 |
| Yes | 526 (3.2) | 406 (4.1) | 120 (1.9) | |
| Obstetrical history, n (%) | ||||
| Gravidity | ||||
| 0 | 5101 (31.4) | 2956 (29.6) | 2145 (34.4) | <.001 |
| 1–3 | 9824 (60.5) | 6460 (64.6) | 3364 (53.9) | |
| >3 | 1321 (8.1) | 586 (5.9) | 735 (11.8) | |
| Full-term birth | ||||
| Nulliparous or no prior full-term birth | 10,229 (63.0) | 6024 (60.2) | 4205 (67.3) | <.001 |
| Yes | 6017 (37.0) | 3978 (39.8) | 2039 (32.7) | |
| Preterm birth | ||||
| Nulliparous or no prior preterm birth | 14,742 (90.7) | 8890 (88.9) | 5852 (93.7) | <.001 |
| Yes | 1504 (9.3) | 1112 (11.1) | 392 (6.3) | |
| Spontaneous abortion | ||||
| Nulliparous or no prior spontaneous abortion | 11,144 (68.6) | 7232 (72.3) | 3912 (62.7) | <.001 |
| Yes | 5102 (31.4) | 2770 (27.7) | 2332 (37.3) | |
| Infertility diagnosis, n (%) | ||||
| Male factor | 3654 (22.5) | 2613 (26.1) | 1041 (16.7) | <.001 |
| Endometriosis | 516 (3.2) | 383 (3.8) | 133 (2.1) | |
| PCOS | 1833 (11.3) | 1273 (12.7) | 560 (9.0) | |
| DOR | 1375 (8.5) | 492 (4.9) | 883 (14.1) | |
| Tubal factors | 797 (4.9) | 608 (6.1) | 189 (3.0) | |
| Uterine | 277 (1.7) | 149 (1.5) | 128 (2.0) | |
| Unexplained | 2415 (14.9) | 1637 (16.4) | 778 (12.5) | |
| ≥2 diagnoses | 3418 (21.0) | 2139 (21.4) | 1279 (20.5) | |
| Other | 1961 (12.1) | 708 (7.1) | 1253 (20.1) | |
| Gestational age (wk), median (IQR) | 39.0 (38.0–39.0) | 39.0 (38.0–39.0) | 38.6 (37.9–39.2) | .698 a |
| Newborn sex, n (%) | ||||
| Male | 8745 (53.8) | 5181 (51.8) | 3564 (57.1) | <.001 |
| Female | 7501 (46.2) | 4821 (48.2) | 2680 (42.9) | |
| Birthweight (g), mean±SD | 3387.7±577.9 | 3399.2±582.2 | 3369.4±570.5 | .001 |
| Z-score, mean±SD | 0.4±1.0 | 0.4±1.0 | 0.3±1.0 | <.001 |
| SGA, n (%) | ||||
| No | 15,549 (95.7) | 9606 (96.0) | 5943 (95.2) | .008 |
| Yes | 697 (4.3) | 396 (4.0) | 301 (4.8) | |
| LGA, n (%) | ||||
| No | 13,412 (82.6) | 8187 (81.9) | 5225 (83.4) | .003 |
| Yes | 2834 (17.4) | 1815 (18.2) | 1019 (16.3) | |
| Macrosomia, n (%) | ||||
| No | 14,308 (88.1) | 8759 (87.6) | 5549 (88.9) | .013 |
| Yes | 1938 (11.9) | 1243 (12.4) | 695 (11.1) | |
| Preterm birth, n (%) | ||||
| No | 14,116 (86.9) | 8750 (87.5) | 5366 (85.9) | .005 |
| Yes | 2130 (13.11) | 1252 (12.5) | 878 (14.1) |
a A Wilcoxon rank-sum test was used to determine the P value.
For the PGT conceptions, boys were more common than girls (57.1% vs 42.9%; P <.001). Compared with the cycles without PGT, the mean birthweight for infants conceived via PGT cycles was lower (3369.4±570.5 vs 3399.2±582.2; P =.001) and the birthweight Z-score was also lower (0.3±1.0 vs 0.4±1.0; P <.001). In the univariate analysis, PGT was associated with a higher percentage of infants who were SGA (4.8% vs 4.0%; P =.008) and a lower percentage was LGA (16.3% vs 18.2%; P =.003) and macrosomic (11.1% vs 12.4%; P =.013). Although the gestational age (weeks) was comparable between the groups (38.6; interquartile range [IQR], 37.9–39.2 vs 39.0; IQR, 38.0–39.0; P =.698), preterm births were more prevalent among the PGT pregnancies (14.1% vs 12.5%; P =.005).
The multivariate analysis is presented in Tables 2 and 3 . After adjusting for confounders, including year of transfer, maternal age at cycle initiation, maternal BMI, smoking, obstetrical history (full-term birth, preterm birth, spontaneous abortion), infertility diagnosis, gestational age, and infant sex, there was no significant difference in the birthweight between singleton infants born after PGT compared with those born without PGT (birthweight difference, –14.63; 95% confidence interval [CI], –29.65 to 0.38). The birthweight Z-score did not differ between the PGT vs non-PGT pregnancies (–0.03; 95% CI, –0.06 to 0.00) when controlling for the same covariates except for gestational age and infant sex, which were already controlled for in the Z-score calculation. There did not seem to be a difference in the odds of SGA newborns (adjusted odds ratio [aOR], 1.17; 95% CI, 0.99–1.38), LGA newborns (aOR, 0.96; 95% CI, 0.88–1.06), and macrosomia (aOR, 0.96; 95% CI, 0.85–1.07) after controlling for the essential confounders between the PGT and non-PGT pregnancies. Note that these odds ratios do not examine whether LGA is increased with FETs overall, but instead examine the PGT vs non-PGT pregnancies. We observed a 20% increase in the odds of preterm birth when comparing singletons conceived from PGT embryos with those without PGT (aOR, 1.20; 95% CI, 1.09–1.33), and the association remained significant after adjusting the P value to minimize the type I error from multiple hypothesis tests.
| Outcomes | Unadjusted point estimate (95% CI) | P value | Adjusted point estimate (95% CI) | P value |
|---|---|---|---|---|
| Continuous outcomes (coefficient) | ||||
| Birthweight (g) a | –29.76 b (–48.02 to 11.49) | .001 | –14.63 (–29.65 to 0.38) | .056 |
| Birthweight (Z-score) c | –0.08 d (–0.11 to –0.05) | <.001 | –0.03 (–0.06 to 0.00) | .081 |
| Dichotomized outcomes, odds ratio | ||||
| SGA c | 1.23 b (1.05–1.43) | .008 | 1.17 (0.99–1.38) | .066 |
| LGA c | 0.88 b (0.81–0.96) | .003 | 0.96 (0.88–1.06) | .418 |
| Macrosomia a | 0.88 e (0.80–0.97) | .013 | 0.96 (0.85–1.07) | .427 |
| Preterm birth f | 1.14 b (1.04–1.25) | .005 | 1.20 d , g (1.09–1.33) | <.001 g |
a Model adjusted for the core covariates in addition to smoking, gestational age, and infant sex
c Model adjusted for the core covariate in addition to smoking
f Model adjusted for the core covariates in addition to smoking and infant sex
g Adjusted statistics and P values indicate a statistical significance after controlling for a false discovery rate.
Stay updated, free articles. Join our Telegram channel
Full access? Get Clinical Tree
