The impact of reference growth standards on small- and large-for-gestational age outcomes among pregnancies conceived by fresh and frozen embryo transfers.

Sunidhi Singh: Study design, Interpretation of data, Article drafting, Final version approval. Pietro Bortoletto: Study design, Acquisition of data, Analysis of data, Interpretation of data, Article drafting, Final version approval. Blair J. Wylie: Study design, Interpretation of data, Article drafting, Final version approval. Alexis P. Melnick: Study design, Interpretation of data, Article drafting, Final version approval. Malavika Prabhu: Study design, Interpretation of data, Article drafting, Final version approval.

A total of 2,567 frozen and 2,931 fresh ET cycles resulted in singleton live births during the study period ( Table 1 ). Among women who underwent fresh ET cycles, the mean age was 35.4 ± 4.3 years, and the most common reasons for infertility were diminished ovarian reserve (44.4%) and male factor (37.4%). Among women who underwent frozen ET cycles, the mean age at retrieval was 35.1 ± 4.1 years, and the most common reasons for infertility were diminished ovarian reserve (40.9%) and male factor (30.9%). Most fresh ETs were at cleavage stage (68.2%), whereas most frozen ETs were at the blastocyst stage (95.7%). The median numbers of embryos transferred were 2 for fresh ET (interquartile range [IQR], 1–3) and 1 for frozen ET (IQR, 1–1). Table 1 Patient and cycle characteristics. Characteristics Frozen ET cycle N = 2,567 Fresh ET cycle N = 2,931 P value Age, y, mean (SD) 35.1 (4.1) 35.4 (4.3) .001 Gravidity, median (IQR) 1 (0–2) 1 (0–2) .057 Parity, median (IQR) 0 (0–0) 0 (0–1) .422 BMI, median (IQR) 22.4 (20.6–25.0) 22.7 (20.8–25.7) .001 Race, n (%) <.001  White 1,529 (59.6) 1,627 (55.5)  Asian 449 (17.5) 398 (13.6)  Black 80 (3.1) 68 (2.3)  Other/unknown 509 (19.8) 838 (28.6) Infertility diagnosis, n (%)  Idiopathic 202 (7.8) 321 (11.0) <.001  Anovulatory 298 (11.6) 283 (9.7) .019  Diminished ovarian reserve 1,051 (40.9) 1,302 (44.4) .009  Tubal factor 327 (12.7) 408 (13.9) .199  Uterine factor 158 (6.2) 144 (4.9) .044  Endometriosis 191 (7.4) 298 (10.2) <.001  Male factor 793 (30.9) 1,096 (37.4) <.001 AMH (ng/mL), median (IQR) 1.7 (0.8–3.5) 2.7 (1.5–4.8) <.001 Stimulation protocol, n (%) <.001  Antagonist 2,279 (88.8) 2,348 (80.1)  Agonist 143 (5.6) 361 (12.3)  Antagonist + CC/LTZ 145 (5.7) 222 (7.6) Frozen transfer type, n (%)  Natural 1,744 (67.9) --  Programmed 823 (32.1) -- Developmental stage, n (%) <.001  Cleavage 110 (4.3) 1,999 (68.2)  Blastocyst 2,457 (95.7) 932 (31.8) No. of embryos transferred, median (IQR) 1 (1–1) 2 (1–3) <.001 Trophectoderm biopsy, n (%) 1,343 (52.3) -- AMH = antimüllerian hormone; CC = clomiphene citrate; ET = embryo transfer; IQR = interquartile range; LTZ = letrozole; SD = standard deviation. Patient and cycle characteristics. AMH = antimüllerian hormone; CC = clomiphene citrate; ET = embryo transfer; IQR = interquartile range; LTZ = letrozole; SD = standard deviation. The median birth weights and gestational ages at birth among fresh ET cycles were 3,289 (IQR, 2,977–3,600) g and 39.4 (IQR, 38.6–40.3) weeks, respectively ( Table 2 ). Those among frozen ET cycles were 3,399 (IQR, 3,065–3,685) g and 39.4 (IQR, 38.7–40.1) weeks, respectively. Table 2 Neonatal outcomes. Outcomes Frozen ET cycle N = 2,567 Fresh ET cycle N = 2,931 P value Gestational age (wk), median (IQR) 39.4 (38.7–40.1) 39.4 (38.6–40.3) .221 Birth weight (g), median (IQR) 3,399 (3,065–3,685) 3,289 (2,977–3,600) <.001 Neonatal sex at birth, n (%) .644  Male 1,345 (52.4) 1,554 (53.0)  Female 1,222 (47.6) 1,377 (47.0) ET = embryo transfer; IQR = interquartile range. Neonatal outcomes. ET = embryo transfer; IQR = interquartile range. The frequency of SGA and LGA among fresh and frozen ETs varied significantly on the basis of the reference standard used ( Table 3 ). The proportion of SGA neonates among fresh ET cycles ranged from 5.8% (95% CI, 5.0%–6.7%) to 13.4% (95% CI, 12.2%–14.6%) ( P < .001). Among women who underwent frozen ET cycles, the frequency of neonates classified as SGA varied from 3.5% (95% CI, 2.9%–4.3%) to 8.7% (95% CI, 7.7%–9.9%) ( P < .001). Variation in the frequency estimates for LGA was also noted. Among women who underwent fresh ET cycles, the proportion of LGA infants ranged from 5.3% (95% CI, 4.5%–6.2%) to 14.3% (95% CI, 13.0%–15.6%) ( P < .001), and that among women who underwent frozen ET cycles ranged from 6.6% (95% CI, 5.7%–7.7%) to 21.2% (95% CI, 19.6%–22.8%) ( P < .001). The INTERGROWTH-21 international reference standard had the highest rate of LGA classification and lowest rate of SGA classification. When evaluating the frequency of SGA < 3rd percentile, the point estimates ranged from 1.5% (95% CI, 1.0%–2.0%) to 3.2% (95% CI, 2.6%–3.9%) ( P < .001) among fresh ETs and from 0.9% (95% CI, 1.0%–2.0%) to 2.6% (95% CI, 2.0%–3.3%) ( P < .001) among frozen ETs ( Table 4 ). The INTERGROWTH-21 curve had the lowest proportion of SGA neonates in this analysis. Analysis stratified by neonatal sex is available in Supplemental Tables 2 and 3 . Table 3 Proportion and 95% confidence interval meeting criteria for small for gestational age (90%) by individual growth curve. Outcomes Fenton INTERGROWTH-21 WHO Duryea Oken P value SGA (<10%)  Fresh 12.6 (11.4–13.9) 5.8 (5.0–6.7) 10.1 (9.0–11.2) 13.3 (12.1–14.6) 13.4 (12.2–14.6) <.001  Frozen 7.9 (6.9–9.1) 3.5 (2.9–4.3) 7.7 (6.7–8.8) 8.4 (7.3–9.5) 8.7 (7.7–9.9) 90%)  Fresh 6.1 (5.2–7.0) 14.3 (13.0–15.6) 7.8 (6.9–8.9) 6.9 (6.0–7.9) 5.3 (4.5–6.2) <.001  Frozen 8.2 (7.2–9.4) 21.2 (19.6–22.8) 10.8 (9.7–12.1) 9.0 (8.0–10.2) 6.6 (5.7–7.7) <.001 LGA = large for gestational age; SGA = small for gestational age; WHO = World Health Organization. Table 4 Proportion and 95% confidence interval meeting criteria for small for gestational age (<3%) by individual growth curve. Outcome Fenton INTERGROWTH-21 Oken P value SGA (<3%)  Fresh 3.2 (2.6–3.9) 1.5 (1.0–2.0) 3.1 (2.5–3.8) <.001  Frozen 2.4 (1.8–3.0) 0.9 (0.6–1.4) 2.6 (2.0–3.3) <.001 SGA = small for gestational age. Proportion and 95% confidence interval meeting criteria for small for gestational age (90%) by individual growth curve. LGA = large for gestational age; SGA = small for gestational age; WHO = World Health Organization. Proportion and 95% confidence interval meeting criteria for small for gestational age (<3%) by individual growth curve. SGA = small for gestational age.

All patients who underwent controlled ovarian hyperstimulation between January 1, 2010, and December 31, 2020, at the Ronald O. Perelman and Claudia Cohen Center for Reproductive Medicine were reviewed. Patients were included if they underwent fresh or frozen ET resulting in a singleton live birth between the gestational ages of 36 weeks and 0 days and 42 weeks and 6 days. If more than 1 birth was recorded per patient during this period, only the first live birth was included. Patients with a vanishing twin, birth weight of <500 g, and gestational age of ≥43 weeks, as well as those whose gestational age, neonatal sex, or birth weight were not recorded, were all excluded. Ovarian stimulation and ET protocols at this center have previously been reported ( 33 ). The primary outcomes were SGA, defined as lower than the 10th percentile for gestational age and neonatal sex at birth, and LGA, defined as greater than the 90th percentile for gestational age and neonatal sex at birth. For this analysis, the following birth weight standards were included: Fenton ( 34 ); World Health Organization (WHO) ( 35 , 36 ); INTERGROWTH-21 ( 37 , 38 ); Duryea et al. ( 39 ); and Oken et al. ( 40 ). Fenton was chosen because it is commonly used in newborn nurseries throughout the United States, INTERGROWTH-21 and the WHO Multicentre Growth Charts provided international reference standards for neonatal and postnatal growth in healthy pregnancies, and Duryea et al. ( 39 ) and Oken et al. ( 40 ) publish contemporary US birth weight standards. The curves are further described in Supplemental Table 1 (available online). For each growth curve, the proportion and 95% confidence interval of neonates being classified as SGA or LGA were calculated. The results were stratified by fresh vs. frozen ET. As a secondary outcome, the frequency estimates for SGA < 3rd percentile were calculated where data were available from the reference curves ( 34 , 37 , 38 , 40 ). Infants < 3rd percentile are most likely to have adverse health outcomes associated with SGA, and we were interested to test whether at the extremes, the curves identified a similar fraction of at-risk neonates ( 1 ). Differences in the frequency estimates were evaluated across each birth weight standard using the chi-square test for SGA < 10th percentile, SGA < 3rd percentile, and LGA. This study was approved by the Weill Cornell Medical College Institutional Review Board (study protocol number 19-06020283).

This study is a novel analysis that provides evidence that the frequency estimates for SGA and LGA among neonates conceived by ART are driven, in part, by the reference standard used. Reassuringly, most fetuses are born at a normal birth weight regardless of reference standard used. Although this study cannot adjudicate which curve may be closest to a gold standard, it is important for clinicians to thoughtfully select the reference growth standard that is most similar to their population. Misclassification of infants with either overdiagnosis or underdiagnosis of SGA and LGA on the basis of the selected reference curve impacts the perceived risks of ART, clinical care provision at the time of birth, and comparability across studies.

In this large retrospective cohort study, the frequency estimates for SGA and LGA among neonates conceived by assisted reproductive technology (ART) are, in part, driven by the reference standard used. The frequency estimates for SGA infants ranged from 4% to 13%, and those for LGA infants ranged from 5% to 21%, the ranges which are of clinical relevant in the neonatal period. Misclassification of infants at risk when they are appropriate for gestational age, or underdiagnosis of SGA or LGA, could lead to infants either having additional unnecessary interventions of glucose surveillance, calcium surveillance, and polycythemia evaluation or inadequate surveillance altogether. In contrast to the risk of misclassification at the 10th and 90th percentiles, the range of risk estimates for neonates meeting criteria for SGA < 3rd percentile was much lower, suggesting that different reference curves identify a similar fraction of the highest risk neonates. The findings of this study suggest that although there is no single best reference birth weight standard, selecting a standard that most closely mirrors the patient population may help minimize underdiagnosis or overdiagnosis of SGA/LGA. Standards for the assessment of fetal growth and birth weight are essential for providing appropriate clinical care, especially because growth abnormalities have short- and long-term health consequences. Significant differences in the derivation of each of the reference curves used in this study are important to consider to interpret the findings. The Fenton reference standard, which is used in several newborn nurseries, was derived from large population-based studies of postnatal growth of preterm infants from high-resource countries. Data are extrapolated between 36 and 50 weeks on the basis of observed preterm growth until 36 weeks of gestational age and smoothed to meet the WHO reference standards of postnatal growth at 50 weeks ( 34 ). Mother-infant pairs sampled in the WHO Multicentre Growth Reference Study had no economic or environmental constraints on growth, including no cigarette smoking ( 35 ). The WHO curves are not specific for gestational age at birth; the standard begins at “month 0,” the gestational ages of all births were between 37 and 42 weeks, and any significant morbidity in newborns were excluded ( 35 , 36 ). In contrast, the INTERGROWTH-21 birth weight standards are both sex-specific and gestational age–specific and aim to serve as a global reference point for normal fetal growth ( 37 ). Fetal growth was assessed in 8 different international urban populations sampling healthy, low-risk women receiving adequate prenatal care and nutrition for pregnancy ( 38 ). As there is no universally accepted US reference for birth weight, 2 recent publications were selected as comparators demonstrating different inclusion criteria for neonates. Duryea et al. ( 39 ) included all singleton births recorded in the US National Center for Health Statistics in 2011 and only excluded infants with a documented anomaly or birth weight of 6,000 g. The inclusion criteria of Oken et al. ( 40 ) were even more generous, including singletons between the gestational ages of 22 and 44 weeks born to US resident mothers in 1999 and 2000. Because of the different populations included and methods used to generate reference standards, we note that the INTERGROWTH-21 reference standard classifies the lowest proportion of neonates born after fresh and frozen ETs as SGA and the largest fraction as LGA. This finding is important because, of the reference growth standards included in this study, the infants born in the highly selected INTERGROWTH-21 cohort are the most likely to be healthy births and have gestational age and neonatal sex-specific values as points of comparison. Thus, if this is an appropriate reference standard to use, only 3.5% and 5.8% of women having pregnancies conceived by frozen and fresh ETs, respectively, have SGA neonates. In contrast, if either the Fenton or US reference standard is used, an almost doubled number of neonates are now classified as SGA and, thus, possibly “at risk.” Choosing an appropriate birth weight reference curve is critical to obstetric and pediatric management alike and a challenge faced for all pregnancies, not just those after ART. Obstetricians need to understand the risks of ART interventions for appropriate preconception counseling and prenatal care management with appropriately targeted interventions for screening for fetal growth abnormalities that are commensurate with the risk incurred in the pregnancy. If the risk of SGA neonates is at the baseline population risk, for instance, routine screening with fundal height measurements is indicated. In contrast, if the risk of SGA is significantly elevated, a third trimester growth ultrasound may be indicated. At birth, pediatricians should be able to identify those neonates requiring additional observation and monitoring for hypoglycemia, hypothermia, and hyperbilirubinemia. Selection of a reference growth standard could result in either overdiagnosis or underdiagnosis of infants at risk of short-term adverse health outcomes. For the practicing clinician, it is a useful reminder to be circumspect about applying a risk estimate derived from a study to an individual patient without first considering whether the patient population in the study mirrors the patient at hand. Finally, reproductive epidemiologists should also understand the impact of selecting a reference growth standard whose population mirrors the sample in question’s sociodemographic characteristics to be able to describe the true frequency of growth abnormalities associated with ART interventions and ultimately allow for the identification of possible risk mitigation strategies. This study has several strengths. Because of the sample size, the cohort was able to be stratified by fresh and frozen ETs, with frequency estimates with narrow confidence intervals. The limitations of this study include the lack of granular data on maternal comorbidities that may additionally be associated with SGA or LGA neonates and a population that is largely Caucasian and, thus, may differentially match the populations of some reference growth curves. There are also no neonatal outcome data or ultrasound-generated estimated fetal weights to be correlated with the proportion identified as SGA or LGA, which may influence obstetric management before delivery, and the proportion of SGA/LGA as identified by the curve to any adverse neonatal outcomes. Finally, it is important to recognize that no reference growth standard, including customized growth standards, can identify whether any individual neonate met its in utero growth potential ( 41 ).

S.S. has nothing to disclose. P.B. has nothing to disclose. B.J.W. has nothing to disclose. A.P.M. has nothing to disclose. M.P. has nothing to disclose.