Factors associated with large-for-gestational-age infants born after frozen embryo transfer cycles.

The number and proportion of FET cycles increased from 21,245 FETs(20% of all transfers) in 2004 to 121,521 FETs (74% of all transfers) in 2018 ( Fig. 1 A). In contrast, the percentage of frozen transfers resulting in singleton LGA infants declined from 18% in 2004 to 12% in 2018 ( Fig. 1 B). Similarly, the incidence of LGA infants after fresh embryo transfer decreased from 11%–9% during this period. The mean birth weight for singletons conceived via FET cycles declined from a mean of 3,364 g in 2004 to a mean of 3,310 g in 2018 (data not shown). FIGURE 1 (A) Trends in the percentages of all embryo transfer cycles that used fresh vs. frozen embryos, 2004–2018. (B) Trends in the percentages of singleton births that were large-for-gestational-age resulting from fresh embryo transfers vs. frozen embryo transfers, 2004–2018. FET = frozen embryo transfer. (A) Trends in the percentages of all embryo transfer cycles that used fresh vs. frozen embryos, 2004–2018. (B) Trends in the percentages of singleton births that were large-for-gestational-age resulting from fresh embryo transfers vs. frozen embryo transfers, 2004–2018. FET = frozen embryo transfer. Among the 127,525 FET cycles during 2016–2018 that resulted in singleton live births, 16,374 (12.8%) cycles resulted in LGA infants and 111,151 (87.2%) cycles resulted in non-LGA infants. There were differences in the characteristics of the cycle and the ART patient with the cycle between FET cycles resulting in LGA and non-LGA infants ( Table 1 ). Compared with the non-LGA infant cycles, cycles resulting in LGA infants were more likely to include non-Hispanic (NH) White patients (45.0% vs. 38.0%) and less likely to include NH Black (2.4% vs. 3.1%) or NH Asian, Native Hawaiian, or Pacific Islander (5.8% vs. 11.0%) patients and were more likely to include patients with a BMI of ≥25 kg/m 2 (43.4% vs. 34.6%) and higher gravidity and parity. Cycle factors associated with LGA included reason for ART (more likely to be male factor infertility or polycystic ovary syndrome and less likely to be diminished ovarian reserve), cycle started in 2016 or 2017, and cycles performed in clinics located in the South and Midwestern regions. TABLE 1 Assisted reproductive technology patient, cycle, and infant characteristics of frozen embryo transfer cycles resulting in LGA vs. non-LGA singleton infants, NASS, 2016–2018. Characteristic LGA Non-LGA P value N = 16,374 12.8% N = 111,151 87.2% No. % No. % Patient age at cycle start (y)  16–24 97 0.6% 801 0.7% <.0001  25–29 1,483 9.1% 10,350 9.3%  30–34 5,631 34.4% 37,815 34.0%  35–39 6,165 37.7% 40,109 36.1%  40–44 2,353 14.4% 16,628 15.0%  ≥45 645 3.9% 5,448 4.9% Patient race and ethnicity a  Hispanic 600 3.7% 4,206 3.8% <.0001  NH American Indian or Alaska Native 32 0.2% 173 0.2%  NH Asian, Native Hawaiian, Pacific Islander 950 5.8% 12,183 11.0%  NH Black 393 2.4% 3,392 3.1%  NH White 7,365 45.0% 42,201 38.0%  Missing 7,034 43.0% 48,996 44.1% Maternal body mass index (kg/m 2 ) a  <18.5 207 1.3% 2,874 2.6% <.0001  18.5–24.9 6,396 39.1% 52,812 47.5%  25.0–29.9 3,784 23.1% 22,597 20.3%  30.0–34.9 1,864 11.4% 9,533 8.6%  ≥35.0 1,457 8.9% 6,286 5.7%  Missing 2,666 16.3% 17,049 15.3% Patient smoking history a  Yes (within 3 mo of cycle start) 253 1.6% 1,542 1.4% .27  No 14,419 88.1% 97,964 88.1%  Missing 1,702 10.4% 11,645 10.5% Patient maximum FSH level a (mIU/mL)  <4.0 340 2.1% 2,680 2.4% <.0001  4.0–9.9 8,189 50.0% 54,309 48.9%  ≥10 1,933 11.8% 14,936 13.4%  Missing 5,912 36.1% 39,226 35.3% Parity  0 8,376 51.2% 68,189 61.3% <.0001  1 5,980 36.5% 32,209 29.0%  ≥2 1,746 10.7% 9,196 8.3%  Missing 272 1.7% 1,557 1.4% Gravidity  0 4,839 29.6% 39,807 35.8% <.0001  1 4,941 30.2% 31,748 28.6%  ≥2 6,322 38.6% 38,039 34.2%  Missing 272 1.7% 1,557 1.4% Reason(s) for ART b  Diminished ovarian reserve 3,415 20.9% 25,903 23.3% <.0001  History of endometriosis 1,226 7.5% 7,890 7.1% .07  Male factor infertility 5,565 34.0% 35,642 32.1% <.0001  Polycystic ovary syndrome 2,608 15.9% 16,537 14.9% .0005  Tubal factor 1,736 10.6% 11,840 10.7% .86  Uterine factor 932 5.7% 6,894 6.2% .01  Recurrent pregnancy loss 604 3.7% 4,091 3.7% .95  Other factors 3,862 23.6% 27,290 24.6% .01  Unexplained 2,125 13.0% 14,721 13.2% .35 Year of FET cycle start  2016 4,621 28.2% 30,812 27.7% .0003  2017 5,596 34.2% 36,737 33.1%  2018 6,157 37.6% 43,602 39.2% Clinic region  Northeast 4,430 27.1% 30,572 27.5% <.0001  Midwest 2,898 17.7% 16,761 15.1%  South (including Puerto Rico) 4,899 29.9% 31,052 27.9%  West 4,154 25.4% 32,766 29.5% No. of thawed embryos transferred  1 13,032 79.6% 88,593 79.7% .9389  2 3,208 19.6% 21,646 19.5%  ≥3 134 0.8% 912 0.8% Oocyte source  Patient 14,487 88.5% 96,674 87.0% <.0001  Donor 1,887 11.5% 14,477 13.0% Gestational carrier .03  Yes 1,022 6.2% 6,469 5.8%  No 15,352 93.8% 104,682 94.2% Assisted hatching .006  Yes 16,048 98.0% 108,527 97.6%  No 294 1.8% 2,356 2.1%  Missing 32 0.2% 268 0.2% Intracytoplasmic sperm injection a <.0001  Yes 12,026 73.4% 84,513 76.0%  No 1,982 12.1% 13,334 12.0%  Missing 2,366 14.4% 13,304 12.0% Preimplantation genetic testing a <.0001  Yes 8,414 51.4% 61,278 55.1%  No 7,926 48.4% 49,661 44.7%  Missing 34 0.2% 212 0.2% Stimulation protocol c  No GnRH protocol 931 5.7% 7,747 7.0% <.0001  GnRH agonist suppression 1,030 6.3% 6,901 6.2%  GnRH agonist flare 598 3.7% 4,472 4.0%  GnRH antagonist suppression 8,559 52.3% 62,145 55.9%  Missing 5,256 32.1% 29,886 26.9% No. of oocytes retrieved c  0–4 394 2.4% 3,441 3.1% <.0001  5–9 1,602 9.8% 12,674 11.4%  10–19 4,868 29.7% 34,874 31.4%  20–29 2,805 17.1% 19,919 17.9%  ≥30 1,481 9.0% 10,632 9.6%  Missing 5,224 31.9% 29,611 26.6% Sperm source c  Partner 10,133 61.9% 75,381 67.8% <.0001  Donor/male patient/mixed 959 5.9% 5,736 5.2%  Missing 5,282 32.3% 30,034 27.0% Semen collection method c  Ejaculation 10,702 65.4% 78,244 70.4% <.0001  Other 395 2.4% 2,904 2.6%  Missing 5,277 32.2% 30,003 27.0% Embryo quality a .09  Good 9,992 61.0% 68,698 61.8%  Fair 3,019 18.4% 19,603 17.6%  Poor 170 1.0% 1,171 1.1%  Missing 3,193 19.5% 21,679 19.5% No. of fetal heartbeats <.0001  1 15,784 96.4% 106,163 95.5%  ≥2 307 1.9% 2,739 2.5%  Missing 283 1.7% 2,249 2.0% Infant sex  Male 8,551 52.2% 57,106 51.4% .04  Female 7,823 47.8% 54,045 48.6% Mean gestational age (wk) 39.0 38.8 <.0001 Mean infant birth weight (g) 4,098 3,210 Note: ART = assisted reproductive technology; FET = frozen embryo transfer; FSH = follicle-stimulating hormone; GnRH = gonadotropin-releasing hormone; LGA = large-for-gestational age; NH = non-Hispanic. a Data obtained from both transfer and retrieval cycles. b Reasons for ART are not mutually exclusive. c Data from retrieval cycles only, thus not available for transfer cycles that did not link to retrieval cycle. Assisted reproductive technology patient, cycle, and infant characteristics of frozen embryo transfer cycles resulting in LGA vs. non-LGA singleton infants, NASS, 2016–2018. Note: ART = assisted reproductive technology; FET = frozen embryo transfer; FSH = follicle-stimulating hormone; GnRH = gonadotropin-releasing hormone; LGA = large-for-gestational age; NH = non-Hispanic. Data obtained from both transfer and retrieval cycles. Reasons for ART are not mutually exclusive. Data from retrieval cycles only, thus not available for transfer cycles that did not link to retrieval cycle. Large-for-gestational-age infants were less likely to have resulted from cycles that used preimplantation genetic testing. There was no observed difference in the quality of embryos that had been transferred between cycles resulting in LGA and non-LGA infants. Fewer infants with LGA were from cycles that resulted in pregnancies with >1 heartbeat at first obstetric ultrasound. Infants with LGA had slightly higher mean gestational age (39.0 vs. 38.8 weeks) at delivery. There was a small yet significant difference in the percentage of male infant sex, with 52.2% men in the LGA cohort and 51.4% men in the non-LGA cohort ( P =.04). The mean infant birth weights were 4,098 and 3,210 g in the LGA and non-LGA cohorts, respectively. In the adjusted model, several variables were significantly associated with LGA among FET cycles ( Table 2 ). The risk of LGA directly increased with increasing maternal BMI (aRR of 1.31 [95% CI, 1.26–1.36] for a BMI of 25.0–29.9 kg/m 2 ; aRR of 1.48 [95% CI, 1.41–1.55] for a BMI of 30.0–34.9 kg/m 2 ; and aRR of 1.68 [95% CI, 1.59–1.77] for a BMI of >35.0 kg/m 2 compared with a BMI of 18.5–24.9 kg/m 2 ) and increased parity (aRR of 1.36 [95% CI, 1.31–1.41] for 1 prior pregnancy and aRR of 1.38 [95% CI, 1.30–1.46] for ≥2 prior pregnancies compared with none). The use of a gestational carrier and that of donor sperm increased the risk of LGA (aRRs of 1.25 [95% CI, 1.16–1.34] and 1.17 [95% CI, 1.10–1.25], respectively). TABLE 2 Relative risk of large-for-gestational-age infants conceived from frozen embryo transfer cycles resulting in singleton live births. Variable Unadjusted Adjusted a RR 95% CI aRR 95% CI Patient age at cycle start (y)  16–24 0.86 (0.70–1.04) 0.88 (0.72–1.07)  25–29 Ref. Ref.  30–34 1.04 (0.98–1.09) 1.02 (0.96–1.07)  35–39 1.06 (1.01–1.12) 1.03 (0.97–1.09)  40–44 0.99 (0.93–1.05) 1.00 (0.93–1.06)  ≥45 0.85 (0.78–0.93) 0.87 (0.79–0.97) Patient race and ethnicity b  Hispanic 0.84 (0.78–0.91) 0.83 (0.76–0.89)  NH American Indian or Alaska Native 1.06 (0.77–1.46) 1.07 (0.78–1.48)  NH Asian, Native Hawaiian or Pacific Islander 0.49 (0.46–0.52) 0.55 (0.51–0.59)  NH Black 0.70 (0.64–0.77) 0.64 (0.59–0.71)  NH White Ref. Ref. Maternal body mass index (kg/m 2 ) b  <18.5 0.62 (0.54–0.71) 0.65 (0.57–0.74)  18.5–24.9 Ref. Ref.  25.0–29.9 1.32 (1.27–1.37) 1.31 (1.26–1.36)  30.0–34.9 1.51 (1.44–1.58) 1.48 (1.41–1.55)  ≥35.0 1.74 (1.65–1.83) 1.68 (1.59–1.77) Patient maximum FSH level (mIU/mL) b  <4.0 0.86 (0.78–0.96) 0.88 (0.80–0.98)  4.0–9.9 Ref. Ref.  ≥10 0.87 (0.84–0.92) 0.95 (0.90–0.99) Parity  0 Ref. Ref.  1 1.43 (1.39–1.47) 1.36 (1.31–1.41)  ≥2 1.45 (1.39–1.53) 1.38 (1.30–1.46) Gravidity  0 Ref. Ref.  1 1.24 (1.20–1.29) 1.02 (0.97–1.07)  ≥2 1.32 (1.27–1.36) 1.02 (0.97–1.07) Reasons for ART (yes vs. no)  Diminished ovarian reserve 0.88 (0.85–0.91) 1.00 (0.96–1.05)  Male factor infertility 1.08 (1.05–1.11) 1.05 (1.02–1.09)  Polycystic ovary syndrome 1.07 (1.03–1.11) 0.99 (0.95–1.03)  Uterine factor 0.93 (0.87–0.98) 0.93 (0.88–1.00) Year of cycle start  2016 Ref. Ref.  2017 1.02 (0.98–1.06) 1.02 (0.98–1.06)  2018 0.96 (0.93–0.99) 0.95 (0.92–0.99) Clinic region  Northeast 0.86 (0.82–0.90) 0.93 (0.89–0.97)  Midwest Ref. Ref.  South 0.92 (0.88–0.96) 0.98 (0.94–1.03)  West 0.76 (0.73–0.80) 0.88 (0.84–0.93) Oocyte source  Patient vs. donor 1.13 (1.08–1.18) 1.16 (1.08–1.24) Gestational carrier (yes vs. no) 1.07 (1.01–1.14) 1.25 (1.16–1.34) Assisted hatching used (yes vs. no) 1.16 (1.04–1.29) 1.06 (0.95–1.18) Preimplantation genetic testing used (yes vs. no) 0.88 (0.85–0.90) 0.98 (0.95–1.01) Stimulation protocol c  No GnRH protocol 0.89 (0.83–0.95) 0.95 (0.89–1.02)  GnRH agonist suppression 1.08 (1.01–1.15) 1.03 (0.97–1.10)  GnRH agonist flare 0.97 (0.90–1.05) 0.99 (0.92–1.08)  GnRH antagonist suppression Ref. Ref. Number of oocytes retrieved c  0–4 0.84 (0.76–0.92) 0.91 (0.83–1.01)  5–9 0.92 (0.87–0.97) 0.93 (0.89–0.99)  10–19 Ref. Ref.  20–29 1.00 (0.96–1.05) 1.01 (0.97–1.06)  ≥30 1.00 (0.94–1.05) 1.02 (0.96–1.08) Sperm source c  Partner Ref. Ref.  Other (donor, male patient, or mixed) 1.21 (1.14–1.29) 1.17 (1.10–1.25) Semen collection method c  Ejaculation Ref. Ref.  Other 1.00 (0.91–1.10) 0.96 (0.88–1.06) Number of fetal heartbeats  1 Ref. Ref.  ≥2 0.78 (0.70–0.86) 0.76 (0.68–0.84) Infant sex  Male 1.03 (1.00–1.06) 1.04 (1.02–1.07)  Female Ref. Ref. Note: ARR = adjusted relative risk; ART = assisted reproductive technology; CI = confidence interval; FSH = follicle-stimulating hormone; GnRH = gonadotropin-releasing hormone; NH = non-Hispanic; Ref. = referent; RR = relative risk. a The multivariate model included age, race and ethnicity, body mass index, maximum FSH, parity, gravidity, reason for ART, year of cycle start, clinic region, oocyte source, use of a gestational carrier, assisted hatching, preimplantation genetic testing, stimulation protocol, number of oocytes retrieved, sperm source, semen collection method, number of fetal heartbeats, and infant sex. b Data obtained from both transfer and retrieval cycles. c Information obtained from retrieval cycles only, thus not available for transfer cycles that did not link to prior retrieval cycle. Relative risk of large-for-gestational-age infants conceived from frozen embryo transfer cycles resulting in singleton live births. Note: ARR = adjusted relative risk; ART = assisted reproductive technology; CI = confidence interval; FSH = follicle-stimulating hormone; GnRH = gonadotropin-releasing hormone; NH = non-Hispanic; Ref. = referent; RR = relative risk. The multivariate model included age, race and ethnicity, body mass index, maximum FSH, parity, gravidity, reason for ART, year of cycle start, clinic region, oocyte source, use of a gestational carrier, assisted hatching, preimplantation genetic testing, stimulation protocol, number of oocytes retrieved, sperm source, semen collection method, number of fetal heartbeats, and infant sex. Data obtained from both transfer and retrieval cycles. Information obtained from retrieval cycles only, thus not available for transfer cycles that did not link to prior retrieval cycle. There were several factors associated with a decreased relative risk of LGA among FET cycles. The risk of LGA was lower among cycles in which the patient race/ethnicity was NH Asian, Native Hawaiian, or Pacific Islander (aRR, 0.55; 95% CI, 0.51–0.59); NH Black (aRR, 0.64; 95% CI, 0.59–0.71); or Hispanic (aRR, 0.83; 95% CI, 0.76–0.89) than that among cycles that involved NH White patients. The risk of LGA infants was also reduced among cycles in which the maternal BMI was lower than normal (<18.5 vs. 18.5–24.9) (aRR, 0.65; 95% CI, 0.57–0.74). The risk of LGA was lower for cycles performed at clinics in the Northeast (aRR, 0.93; 95% CI, 0.89–0.97) and West (aRR, 0.88; 95% CI, 0.84–0.93) regions of the United States than in the Midwest and for cycles in 2018 (aRR, 0.95; 95% CI, 0.92–0.99) than in 2016.

In conclusion, although the number and proportion of FET cycles dramatically increased from 2004–2018, the incidence of LGA after both fresh embryo transfer and FET decreased during the same time period. The incidence of LGA after FET was greater than the incidence after fresh embryo transfer. Among singleton infants resulting from FET in 2016–2018, the patient factors of maternal BMI, parity, and race/ethnicity were the strongest independent risk factors for LGA. Cycle factors that increased the risk of LGA were donor sperm and gestational carrier use. Because many of these factors are also known to increase the risk of LGA in non-ART pregnancies, the cause for the increased risk of LGA in FET cycles remains unclear. Focus on modifiable factors (e.g., BMI) may be helpful to reduce the risk of LGA in FET cycle patients; however, continued research is needed to uncover the pathophysiologic cause of this relationship.

Data for this study were obtained from the Centers for Disease Control and Prevention’s NASS, which was mandated by the Fertility Clinic Success Rate and Certification Act (Fertility Clinic Success Rate and Certification Act of 1992, Public Law No. 102-493, October 23, 1992) and, as a result, includes information on nearly all (98%) ART cycles performed in US fertility clinics ( 2 ). Since 1995, the NASS has collected cycle-level data on patient characteristics, infertility diagnoses, reproductive history, clinical parameters for ART procedures, and resultant pregnancies and births. In this analysis, the subject and unit of analysis is the FET or fresh embryo transfer cycle. The characteristics of the ART patient involved with the cycle are referred to as patient factors. For the subanalysis, we restricted our analyses to FET cycles that resulted in singleton live birth. We analyzed 1,993,742 embryo transfer cycles reported to the NASS from 2004–2018 to examine the trends in the proportions of FET vs. fresh embryo transfers. We then examined the rate (%) of LGA among 575,107 cycles resulting in singleton live births conceived from frozen vs. fresh embryo transfer during this time period. Large-for-gestational-age was defined as an infant birth weight (grams) for gestational age (weeks) of >90th percentile according to a gestational age–specific population-based reference by Talge et al. ( 31 ). This birth weight reference is based on singleton live births between 22 and 44 weeks of gestational age to US resident women in 2009–2010. Non-LGA was defined as an infant birth weight for gestational age of ≤90th percentile.

Our population-based study using the NASS data found that from 2004–2018, the proportion of FET cycles increased from 20%–74% of all transfers performed, whereas the rate of LGA singletons born after FET decreased from 18%–12%. The rate of LGA singletons after fresh embryo transfer also decreased from 11%–9% during this period. In a subanalysis limited to FET cycles during 2016–2018 that resulted in singleton infants, the factors associated with the increased risk of LGA included increasing maternal BMI and parity of ≥1. In contrast, the cycle characteristics associated with the lower risk of LGA included low maternal BMI (<18.5 kg/m 2 ) and cycles involving NH Asian, Native Hawaiian, or Pacific Islander; NH Black; or Hispanic ART patients. Our study confirmed previous findings that maternal BMI is an independent risk factor for having an LGA infant after FET-associated conceptions, with the risk increasing directly with increasing BMI ( 21 ). Although weight loss before FET cycles has been hypothesized to improve live-birth rates in certain patients ( 34 ), to our knowledge, no study has investigated the effect of patient weight loss on infant weight after FET cycles. The risk of LGA was lower after FET performed at clinics in the Northeast and West than that after FET cycles performed at clinics in the Midwest. This finding may reflect the lower prevalence of obesity in these regions ( 35 ) or other unmeasured factors. Our study found a small increased risk of LGA (aRR, 1.04; 95% CI, 1.02–1.07) in cycles involving male infants after FET. This is consistent with previous reports that suggest that male infants are more likely to be LGA than female infants regardless of IVF or spontaneous pregnancy ( 14 , 36 ). Several previous studies that assessed the LGA risk in FET cycles did not report patient race and ethnicity, often instead reporting country of origin. Most of these studies were performed in Scandinavian countries with predominantly White populations ( 12 , 27 , 37 ). However, an analysis of all births in California in 2007 found that Hispanic, Asian, and Black women had a lower risk of delivering infants with macrosomia than White women ( 23 ). Our study demonstrated that compared with cycles involving NH White ART patients, cycles involving NH Asian, Native Hawaiian, or Pacific Islander; NH Black; and Hispanic patients had a lower risk of LGA after FET. Our subanalysis only included FET cycles; therefore, it is unclear whether the relationship observed between race and ethnicity and LGA risk in our study is specific to FET-associated deliveries. Several IVF and FET cycle factors were included in our subanalysis, including oocyte source (donor vs. patient), use of a gestational carrier, source of sperm (donor vs. partner), and number of fetal heartbeats noted on the first obstetric ultrasound. Although cycles involving gestational carriers and the use of donor sperm were associated with a statistically significant increased risk of LGA, the clinical significance of these differences is unclear. Interestingly, having >1 fetal heartbeat on the first obstetric ultrasound (which ultimately became a singleton gestation) was associated with a decreased likelihood of LGA. One possible explanation for this finding is that “vanishing twin” phenomenon is associated with smaller placentas and the increased likelihood of anatomic pathology, such as velamentous course insertion, which can contribute to fetal growth restriction ( 38 ). Unfortunately, the NASS data did not include FET cycle protocols or regimens (i.e., natural vs. stimulated cycle, endometrial preparation regimen, and cryopreservation technique) as part of the reporting information; therefore, our study could not assess the association of these cycle characteristics on the risk of LGA. A recent retrospective cohort study conducted by Wang et al. ( 37 ) analyzed the endometrial preparation protocols in more than 9,000 singleton live births and found that singleton newborns conceived after programmed cycle (i.e., with exogenous estradiol and progesterone) FET were more likely to be LGA than those born after natural cycle FET or stimulated cycle FET. These data, coupled with earlier studies that noted associations between factors related to the freezing of embryos (cryopreservation protocols, culture media exposure, and absence of a corpus luteum), suggest that the FET protocols, along with maternal factors, influence fetal birth weight and increase the risk of LGA in FET cycles ( 4 , 5 , 6 , 39 ). Our analysis of the NASS data showed that the LGA rates after FET decreased from 18% in 2004 to 12% in 2018. Interestingly, a recently published retrospective cohort study from a single US fertility clinic by Shah et al. ( 40 ) demonstrated that between August 1995 and October 2019, changes in the IVF protocols coincide with the decreased rates of LGA infants after FET. This study analyzed several different time points in the evolution of IVF regimens, including vitrification at the blastocyst stage (2011), the use of benchtop incubators (2012), and single-step embryo culture media (2013). The percentage of LGA infants decreased at each time point. Because the NASS data comprise 98% of all IVF cycles in the United States, our study represents one of the largest and most comprehensive studies to evaluate the risk factors for LGA after FET cycles. Limitations include the retrospective nature of this study, risk of selection bias using retrospective cohort design methodology, large percentages of missing data on BMI and race and ethnicity, and inability to link 27% of FET cycles to their preceding egg retrievals or ovarian stimulation cycles. Missing data are a large concern for studies that use large data sets and may limit generalizability of results ( 41 ). Techniques including available case analysis, several types of data imputation, and the use of the missing data as a subvariable are used to treat these missing data ( 42 ). We chose to use missing data as a subvariable in our analysis, and with the lack of large clinical differences in missing data between cohorts, we did not further seek specific data imputation. In addition, several important cycle variables (e.g., estradiol or progesterone levels on day of trigger) were unavailable in the NASS. The NASS data, similar to the use of data in other large databases, may also have included repeated treatment cycles by the same individual. Moreover, limiting our study to cycles that resulted in singleton deliveries may have resulted in selection bias by the predominant inclusion of good prognosis patients. Our data demonstrate that approximately 79% of all patients received a single embryo transfer, which is commonly selected for patients with a favorable prognosis for live birth ( 43 ). However, by excluding multiple pregnancies, we aimed to minimize confounding given the known growth abnormalities associated with multiple pregnancies ( 44 ). Studies using large databases often find significant results for variables that may have limited clinical significance ( 41 ). Similarly, our analysis found several statistically significant associations that may or may not be clinically significant, such as differences among regions of the United States or by infant sex.

To examine factors associated with the risk of LGA after FET, we then conducted a retrospective cohort subanalysis of FET cycles performed during 2016–2018 that resulted in singleton live births (n = 135,512). We excluded cycles that used donated embryos (n = 2,301) and those with unknown infant sex (n = 933) or birth weight (n = 4,125). In addition, we excluded cycles with gestational age of 44 weeks (n = 378). We further excluded cycles with implausible infant birth weights (n = 250) according to previously published criteria by Alexander et al. ( 32 ). Our final analytic cohort was 127,525 FET cycles resulting in singleton live births. To identify the potential risk factors for LGA after FET, we compared the characteristics of ART patients (age, race/ethnicity, BMI, smoking status, and maximum follicle-stimulating hormone level), maternal reproductive history (parity, gravidity, and reason for ART/infertility diagnosis), IVF and transfer cycle characteristics (year of cycle start, clinic region, number of thawed embryos transferred, oocyte source, stimulation protocol used for oocyte retrieval, number of oocytes retrieved, use of intracytoplasmic sperm injection, use of assisted hatching, preimplantation genetic testing, embryo quality, and sperm source and collection method), and cycle pregnancy factors (number of fetal heartbeats and infant sex) between LGA and non-LGA outcomes. Because certain IVF cycle characteristics (e.g., stimulation protocol, intracytoplasmic sperm injection, and preimplantation genetic testing) collected during oocyte retrieval are not always carried forward to subsequent FET cycles in the NASS, we used the date of retrieval to link FET cycles to their original oocyte retrieval cycles to obtain relevant data related to retrieval. To identify cycle characteristics independently associated with LGA after FET, we performed a multivariate modified Poisson regression analysis, including patient factors (age, race and ethnicity, BMI, maximum follicle-stimulating hormone level, parity, gravidity, and reason for ART) and IVF and transfer cycle factors (year of cycle start, clinic region, oocyte source, use of a gestational carrier, use of assisted hatching, use of preimplantation genetic screening, IVF stimulation protocol, number of oocytes retrieved, sperm source, sperm collection method, number of fetal heartbeats, and infant sex). A missing category was created for variables with missing observations to retain observations in the multivariate model. Although most variables had <2% missing data, a larger proportion of missing data were noted for some variables, including patient race/ethnicity (44%) and maternal BMI (15%), as well as for those variables obtained from the initial retrieval cycle. The χ 2 and Fisher’s exact tests were used to compare categorical variables between the LGA and non-LGA outcomes for FET cycles. We calculated the means for gestational age and infant birth weight for each of the 2 outcomes. Factors that were selected a priori or that were significantly associated with LGA in univariate analyses were included in the multivariate analysis. We estimated the adjusted relative risks (aRRs) and 95% confidence intervals (CIs) to determine cycle-level factors associated with the increased risk of LGA after FET ( 33 ). All statistical analyses were performed using SAS statistical software (version 9.4; SAS Institute, Cary, NC), and the results were considered significant at a P value of <.05. Epidemiologic research using NASS data is approved by the institutional review board at the Centers for Disease Control and Prevention.