Introduction

Assisted Reproductive Technology (ART) is an infertility therapy that involves the handling of both gametes in the laboratory to achieve pregnancy. ART-conceived singletons have an increased risk for low birth weight (LBW) compared with singletons in the general population.[1–9] LBW has long been used as an indicator for child health, however, its interpretation is unclear because LBW may be related to short gestation, small newborn size or their combination.[10–12] Therefore, indicators that distinguish LBW infants resulting from short gestation or small newborn size provide a more informative measure of risk by reducing this confounding. Two such indicators are small for gestational age (SGA) and birth-weight-z-score.

The definition of SGA varies across studies and may include infants whose birth weight is below the 10^th or 5^th percentile (SGA/10^th or SGA/5^th) for gestational age or whose birth weight is ≥2 standard deviations below the mean birth weight for gestational age. Two recent studies did not detect an increased risk of SGA/10^th for ART compared with non-ART singletons[13, 14], whereas, other studies reported a significantly increased risk of SGA/10^th among ART singletons, with odds ratios ranging from 1.22–2.29.[8, 15–17] Using the 5^th percentile, ART singletons were found to have a 40% higher odds for SGA birth compared with their non-ART counterparts.[18] These conflicting results may be attributed to the variety of SGA definitions, or to other sources of heterogeneity such as sample size and/or the approach to potential confounders e.g. plurality, social factors, ART and infertility characteristics.

Birth-weight-z-score, constructed as a continuous measure, allows the comparison of newborn size across gestational ages, sexes and birth weights, representing the same relative birth weight for infants, rather than their absolute weight. Compared with LBW, newborn size measured as birth-weight-z-score has rarely been used to investigate birth outcomes in ART populations.[19]

Previous reports suggested differential birth weight for ART singletons conceived with fresh versus cryopreserved embryos, with smaller newborn size for the former. [20–22] In contrast, one cohort study associated lower birth weight with ART singletons from cryopreserved embryos transfer. [23] The inconsistent findings across studies of ART and newborn size warrant additional investigation. The purpose of this study was to use data from the States Monitoring Assisted Reproductive Technology (SMART) Collaborative to examine: 1) whether ART singletons, born following a fresh or cryopreserved embryos transfer, are at higher risk of small newborn size, measured by both SGA/10^th, SGA/5th and birth-weight-z-score, compared with singletons in the general population; and 2) whether an association between ART and newborn size is driven by infertility source (female, male, combined, or unexplained infertility).

Materials and Methods

Study Population

We used a population-based dataset of birth certificates from three states linked to the National ART Surveillance System (NASS) by the SMART Collaborative project. The SMART Collaborative was established by the Centers for Disease Control and Prevention (CDC) and public health agencies of Florida, Massachusetts and Michigan to evaluate maternal and perinatal outcomes of ART and to improve state-based ART surveillance.[24] The sample included all live births in Michigan from 2000–2009 and in Florida and Massachusetts from 2000–2010, linked to ART cycles in the respective states using a probabilistic linkage method with a high linkage rate (87.8%) and good validity.[25]

We restricted the sample to singletons born to mothers aged 15–60 between 22 and 44 weeks’ gestation. We then excluded records with implausible combinations of birth weight and gestational age using the approach described in Table 1 of Alexander et al [26] along with recently published criteria. [27, 28] Finally, we excluded singletons, conceived with donor oocytes from the ART group and created two strata of ART singletons, born following a fresh or cryopreserved embryos transfer. After all exclusions, the final dataset included a total of 4,292,792 singleton live births, of which 4,263,846 (99%) were non-ART and 28,946 (1%) were ART related births. ART singletons included 25,054 infants from fresh embryos transfer and 3,879 from cryopreserved embryos transfer (Fig 1).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Flowchart of Participants in the States Monitoring Assisted Reproductive Technology (SMART) dataset: Florida and Massachusetts 2000–2010 and Michigan 2000–2009.

https://doi.org/10.1371/journal.pone.0169869.g001

The study received approval from the Institutional Review Boards of Florida, Massachusetts, Michigan and the CDC.

Results

After excluding infants with implausible combinations of birth weight and gestational age, the final sample included 4,292,792 infants. Frequencies of maternal and infant characteristics were similar in samples with and without the excluded births. In descriptive analyses (Table 1), ART mothers were significantly older, more educated and more likely to be Non-Hispanic white and primiparous compared with non-ART mothers. Infants born to non-ART and ART mothers had a similar sex distribution, but ART infants’ mean birth weight was 24 grams lower. Compared with non-ART singletons, all three measures of newborn size (SGA/10^th, SGA/5^th, and birth-weight-z-score), were similar among ART singletons born following a fresh embryos transfer (p>0.05), but different for ART births from cryopreserved embryos transfer.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Maternal and infant characteristics for fresh and frozen autologous ART cycles and non-ART singleton live births in Massachusetts and Florida 2000–2010 and in Michigan 2000–2009: After exclusion of implausible birth weight for gestational age.

https://doi.org/10.1371/journal.pone.0169869.t001

The associations between ART treatment and newborn size (SGA/10^th and SGA/5^th) for the pooled ART group and for each infertility subgroup, stratified by fresh and cryopreserved embryos transfer, are presented in Table 2. In adjusted analyses, the odds of an SGA/10^th or SGA/5^th infant from fresh embryos transfer were significantly greater in the ART combined group than in the non-ART group [adjusted odds ratio (aOR) 1.19 (95% CI 1.14, 1.25) and aOR 1.17 (95% CI 1.10, 1.25), respectively]. Each ART subgroup had increased odds of delivering a SGA/10^th infant relative to the non-ART group. Similar results were observed in adjusted SGA/5^th models with the exception of male infertility diagnosis. Conversely, ART infants born following a cryopreserved embryos transfer were less likely to be SGA/10^th or SGA/5^th compared with non-ART singletons.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Associations between newborn size and ART singletons born following fresh or cryopreserved embryos transfer, using non-ART as the referent group.

Population-based data of all singleton live births in Florida and Massachusetts 2000–2010 and in Michigan 2000–2009.

https://doi.org/10.1371/journal.pone.0169869.t002

After excluding all preterm births from the cohort, the statistically significant association between ART and small newborn size remained. Full term ART singletons from fresh embryos transfer were more likely to be SGA/10^th, while full term singletons from cryopreserved embryos transfer were less likely to be born SGA/10^th, relative to full term non-ART singletons [aOR 1.22 (95% CI 1.17, 1.28)] and [aOR 0.60 (95% CI 0.51, 0.71)] respectively. Similar results were observed when each ART subgroup was compared with the non-ART group (Table 3).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Associations between full-term newborn size and ART singletons born following fresh or cryopreserved embryos transfer, using non-ART as the referent group.

Population-based data of all singleton live births in Florida and Massachusetts 2000–2010 and in Michigan 2000–2009.

https://doi.org/10.1371/journal.pone.0169869.t003

Next, newborn size was modeled as a continuous variable (birth-weight-z-score) with non-ART singletons as the referent group (Table 4). We observed that all ART singleton births from fresh embryos had a negative mean birth-weight-z-score, i.e., their mean birth weight was below the mean of the reference population. In contrast, non-ART singleton and ART singletons from cryopreserved embryos transfer, had mean birth-weight-z-scores above that of the reference population.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 4. Associations between birth weight z-score and ART singletons born following fresh or cryopreserved embryos transfer, using non-ART as the referent group.

Population-based data of all singleton live births in Florida and Massachusetts 2000–2010 and Michigan 2000–2009.

https://doi.org/10.1371/journal.pone.0169869.t004

Discussion

We investigated the association between ART and newborn size using one of the largest datasets of ART data linked to birth records, involving over four million newborns. SGA and birth-weight-z-scores are informative outcomes and more specific to newborn size compared with other frequently used measures, e.g. LBW and mean birth weight. Our findings suggest that ART singletons from fresh embryos transfer had increased odds of being SGA/10^th regardless of whether infertility was diagnosed in the female patient, male partner, both, or was unexplained. Conversely, ART singletons, resulted from cryopreserved embryos transfer, had lower odds of being SGA/10^th compared with singletons in the general population.

Our observed SGA/10^th odds for ART singletons born after fresh embryos transfer relative to non-ART singletons were more modest in comparison to a recent Australian study that reported a 1.5-fold higher SGA/10^th risk among ART singletons from fresh embryos transfer compared with non-ART singleton infants. Our results for SGA/10^th among ART singletons from cryopreserved embryos transfer compared with non-ART singletons were aligned with their findings. [16] In other studies, when ART singletons from fresh embryos transfer were selected as the reference group, ART singletons from cryopreserved embryos transfer had lower [20, 23] or the same odds [31] for SGA/10^th. This is consistent with reports that large for gestational age and macrosomic singletons have been associated with transfer of cryopreserved embryos compared with singletons from fresh embryos transfer or non-ART singletons. [31, 32]

Inconsistent results across studies of ART and newborn size of singletons from fresh or cryopreserved embryos transfer might be explained by multiple heterogeneous elements, such as SGA definitions, SGA/10^th, or SGA< 2 standard deviations below a growth standard of a reference population, [16, 31] selection of different control groups, [16, 23] or lack of adjustment for important confounders, such as parity or maternal age. [20] In this analyses, we included parity as a confounder to allow inclusion of females with primary infertility as well as women with potentially secondary infertility, who otherwise would have been excluded from this study. Thus, the current effects reflect a comprehensive and diverse population of subfertile ART users.

Post hoc univariate analyses showed that although all covariates were statistically significant, only ‘maternal education’ and ‘maternal age’ contributed largely to the risk of small newborn size among ART singletons. These factors were previously reported as risk factors for small size at birth.[33, 34]

Although birth-weight-z-scores are recommended for reporting perinatal outcomes among ART populations [19, 35] such studies are rare. In 2008, Shih et al, used birth-weight-z-scores based on British growth reference data to examine whether newborn size was associated with different types of ART treatment.[21] While their findings suggested lower mean birth-weight-z-scores among infants born to couples who used ART with fresh embryos, -0.163 (SD = 1.004) compared with infants of non-ART couples -0.061 (SD = 1.099), both groups had mean birth-weight-z-scores below the expected mean. In this analysis we used two indicators, SGA and birth-weight-z-score, to measure sex-specific newborn size both categorically and continuously. To minimize measurement errors, both indicators were carefully constructed after exclusion of birth records with implausible birth weight and gestational age combinations using established and recently published criteria and algorithms.[26–28] In contrast to Shih, [21] our results showed that the non-ART group had mean birth-weight-z-scores slightly above that of the reference population, whereas the ART group’s means of singletons from fresh embryos transfer were below the referent.

The differential newborn size of ART singletons from fresh or cryopreserved embryos transfer is poorly understood, and may be partially explained by aspects related to ART procedure and patient profiles. ART cycles involving cryopreserved embryos transfer are associated with reduced ovarian stimulation, thus, improved endometrial receptivity, no oocyte retrieval and transfer of higher quality embryos that survived the freezing-thawing process. [23, 36] However, recent data suggest that the freezing-thawing process complicates ART pregnancies with preeclampsia. [37, 38]

In this study design and analyses we tried to address limitations of previous studies, however, some limitations remain. Although SGA is a better indicator of newborn size than LBW, it still represents a heterogeneous group, i.e. those who are constitutionally small and those with pathologically small newborn size. In addition, fetuses who begin as appropriate size and then experience poor growth may not meet the 10^th percentile cutoff at birth, thus using SGA as an outcome could result in some misclassification of small newborn size for infants above and below the cut point. [39] SGA is constructed using population-based references of birth weight while excluding fetal size of unborn fetuses. As a result, SGA formulation for preterm infants may be prone to bias resulting from birth weight differences between preterm infants (often smaller) and those who remain in-utero. In an attempt to minimize this observational bias, we repeated the analysis using only full term infants. The results indicated a similar risk of smaller newborn size for ART singletons, from fresh or cryopreserved embryos transfer, versus non-ART infants born at term. Potential confounders such as BMI, smoking and an early loss of a co-twin were not included in the primary analyses. BMI information was not available for all states and years. Although smoking is associated with small newborn size [40], it was excluded from the primary analyses due to its low prevalence (1%) among ART mothers and its low reliability in birth records. [41, 42] A secondary analysis with smoking included in the adjusted regression models produced similar effect estimates among all ART subgroups. Our results were also unaffected when analyses were restricted to singleton births originating from single fetal pregnancies. Treatment information on culture media, freezing techniques, day of embryo transfer and pre-implantation genetic diagnosis (PGD) were not available for the years included in this study. In 2011–2012, PGD was reported in 4.5% of fresh embryos transfer [43] and is therefore unlikely to have a significant impact on our findings.

The SMART dataset includes two distinct populations, ART and non-ART couples. We recognize that there are unidentified couples with infertility among non-ART users. These subfertile couples may have conceived using non-ART therapy to treat subfertility or achieved a spontaneous pregnancy. The resulting misclassification could have attenuated the association between ART births and small newborn size. We also acknowledge that data quality with respect to timing of conception may vary by the mode of conception, and is more accurate for the ART population. To create the SMART dataset, a probabilistic method was used to link birth certificates and ART surveillance data. While highly successful, with reported linkage rate of 87.8% and a good validity, this method is not free of matching errors.[25] In 2010, it was estimated that ART surveillance data represents more than 97% of ART cycles performed in the US, as only few small clinics did not provide their data to NASS.[44]

Conclusions

We found that the risk of smaller newborn size was higher in ART singletons from fresh embryos transfer and lower in ART singletons from cryopreserved embryos transfer compared with non-ART singleton infants. Although statistically significant, the magnitude of excess risk among ART singletons from fresh embryos transfer was small, which is reassuring. Similarly, the excess risk of smaller newborn size within ART subgroups defined by infertility source (male, female), was not large. Greater subgroup heterogeneity in newborn size may be detected if assessed by underlying infertility causes and ART therapy subtypes; our future work will investigate this potential heterogeneity.

Acknowledgments

The findings and conclusions in this report are those of the authors and do not necessarily represent the official position of the Centers for Disease Control and Prevention.