The growth of assisted reproductive treatment-conceived children from birth to 5 years: a national cohort study

Background Birth weight and early child growth are important predictors of long-term cardiometabolic disease risk, in line with the Developmental Origins of Health and Disease hypothesis. As human assisted reproductive technologies (ARTs) occur during the sensitive periconceptional window of development, it has recently become a matter of urgency to investigate risk in ART-conceived children. Methods We have conducted the first large-scale, national cohort study of early growth in ART children from birth to school age, linking the register of ART, held by the UK’s Human Fertilisation and Embryology Authority, to Scottish maternity and child health databases. Results In this study of 5200 ART and 20,800 naturally conceived (NC) control children, linear regression analysis revealed the birthweight of babies born from fresh embryo transfer cycles is 93.7 g [95% CI (76.6, 110.6)g] less than NC controls, whereas babies born from frozen embryo transfer (FET) cycles are 57.5 g [95% CI (30.7, 86.5)g] heavier. Fresh ART babies grew faster from birth (by 7.2 g/week) but remained lighter (by 171 g), at 6–8 weeks, than NC babies and 133 g smaller than FET babies; FET and NC babies were similar. Length and occipital-frontal circumference followed the same pattern. By school entry (4–7 years), weight, length and BMI in boys and girls conceived by fresh ART and FET were similar to those in NC children. Conclusions ART babies born from fresh embryo transfer grow more slowly in utero and in the first few weeks of life, but then show postnatal catch up growth by school age, compared to NC and FET babies. As low birth weight and postnatal catch-up are independent risk factors for cardiometabolic disease over the life-course, we suggest that further studies in this area are now warranted. Electronic supplementary material The online version of this article (10.1186/s12916-018-1203-7) contains supplementary material, which is available to authorized users.


Background
More than six million children have been born worldwide since 1978 using assisted reproductive technologies (ART) [1]. While the majority of these appear to be healthy, a higher incidence of some congenital abnormalities, pre-term birth and low birth weight (LBW) for gestational age has been noted among ART singletons [2][3][4][5]. The latter is of increasing concern, since birth weight is a surrogate for fetal growth and a strong predictor of cardiometabolic disease (CMD) risk across the life-course [6] (the Developmental Origins of Health and Disease (DOHaD); https://dohadsoc.org/)).
Studies in both human populations [7] and animal models [8,9] have clearly identified the periconceptional period immediately before and after fertilisation as a unique developmental window during which the long-term health of the individual may be programmed. This has raised significant concerns about ART children as it coincides with the period during which gamete and embryo manipulations take place in the nonphysiological ART environment, with specific embryo manipulations associated with altered fetal growth and birth weight [9].
Postnatal infant and child growth is arguably an even more important indicator of long-term disease risk; in particular, catch-up growth from LBW increases the risk of obesity and CMD [10,11]. Owing to the lack of routinely collected data, only a limited number of small (≤ 500 children) studies have been carried out on ART child growth and these generally show conflicting results. In three small European cohort studies, children born from intra-cytoplasmic sperm injection (ICSI) were found to be lighter than their target weight (for age and height) at age 3 years, but not at age 5 years [12]. A longitudinal study [13] noted that ART children had lower weight at birth and 3 months, with similar growth rates to age 3, compared with naturally conceived (NC) subfertile controls. Given the implications of altered early life growth for health risks later in life, it is clearly important to collect definitive data from larger groups of ART children. It is also far from clear which specific ART factors might be causing changes in birth weight, although extending the length of time embryos are kept in culture before (blastocyst) transfer, the type of culture medium and the use of frozen embryo transfer (FET) have been implicated [5,[14][15][16][17]. Without this essential information on modifiable risk factors, it will be impossible to improve the ART process and reduce long-term risk to children.
The UK pioneered ART 40 years ago with the first successful birth in Oldham in July 1978 and thus is home to the first cohort of children now approaching middle age. However, the UK has been among the slowest countries to conduct long-term ART follow-up studies, because the original Human Fertilisation and Embryology Act (1990) did not envisage this. In line with shifts in public opinion, the Act was changed in 2010 and under the principle of presumed consent, the UK now possesses the world's oldest and largest database of ART available for research, with approximately 110,000 children from 1991 to 2009. However, so far, this has only been exploited for a single cancer follow-up study [18]. There have been no large studies of ART child growth or causal factors in the UK or anywhere in the world. Therefore, the primary aim of this work was to assess the role of ART treatment factors on growth from birth to age of school entry, in a large UK national study. A secondary aim was to provide comparative data for NC babies.

Data sources
A cohort of ART data from the Human Fertilisation and Embryology Authority (HFEA) register was linked to a series of routinely collected birth and child health records held by the NHS National Services Scotland (NHS NSS) (Fig. 1).

HFEA Assisted Reproductive Technologies Register
The UK has a uniquely comprehensive record of all women who have had ART since August 1991, when it has been compulsory for every UK fertility clinic to report details of all treatments to the HFEA. Up to October 2009, data on 289,000 women and 110,000 children born after treatment has been collected. Data on treatment cycles after October 1, 2009, was not included as, from this time, those undergoing ART were asked to give "consent to disclosure" of identifying information, enabling their records to be used for research. The data relating to children born prior to April 1997 was not included in the study, as they were over the age of 16 at the time of data release and we took a cautious approach as to whether consent had been provided. The HFEA register does not contain a single unique patient identifier, but does contain names and dates of birth.

Scottish Morbidity Record-SMR02
The SMR02 is primarily a source of data on obstetric outcomes, collated from hospital administration records following a woman's episode of care in an obstetric facility (inpatient and day cases, but not home births). Birth weight, gestational age, Apgar score, occipital-frontal circumference (OFC), length, weight and admission to a neonatal intensive care unit (NICU) have been recorded consistently over time. There is also data on maternal lifestyle with selfreported smoking being consistently recorded. The child's Community Health Index (CHI) number allows linkage to child data in other datasets.

Scottish child health programme
Data on child growth is available from a number of routine screening programmes. The data consistently collected at 6-8 weeks and on entry to school (generally collected between 4 and 7 years) comprise height (length and OFC at 6-8 weeks), weight, body mass index (BMI) and read-coded health concerns.

Data linkage
In the absence of a unique identifier, HFEA and SMR02 data were linked using probabilistic matching based on maternal names and maternal/child dates-of-birth as described in Additional file 1.

Statistical analyses
Deprivation was assessed from quintiles of the 2012 Scottish Index of Multiple Deprivation [19]. Maternal smoking history was self-reported.
Birth weight was adjusted for gestational age, gender and (when available) parity (1st baby vs. > 1st baby) using the GROW formula [20].
Growth velocities were defined as the average change in weight per week (between birth and 6-8 week measurements) or month (between 6 and 8 weeks and school entry).
Numerical outcomes were modelled using linear regression while NICU admission and sex ratio (ratio of female to male births) were modelled using logistic regression. Bootstrapped standard errors (using 1000 replications) are presented. Details of all the models considered are given in Additional file 1: Table S4.
For analyses comparing ART and naturally conceived infants, the primary covariates were deprivation, smoking status, maternal age (fitted as a quadratic function) and an indicator of whether the embryos were from fresh or frozen transfer cycles, plus the matching variables, year and month of delivery and gender. For within ART analyses, we additionally included paternal age (as a quadratic function), use of ICSI, number of previous ART cycles and infertility causes (tubal, endometrial, ovulatory, male and idiopathic as binary indicators) plus treatment centre and year of treatment. As data relating to embryo culture time and number of embryos created were only available on a subset of cases (generally not available for FET cycles), these were not included in the primary analyses, but a secondary analysis including these was conducted. For the analysis of child health programme (CHP) outcomes, we additionally controlled for age at the time In the analysis of NICU admission, the primary analysis was unadjusted for birth weight and gestational age. A secondary analysis was also conducted in which both raw birth weight and gestational age were controlled for.
Using tables of age-and gender-specific L, M and S values [21], we transformed school-entry weight into deciles of age-standardised z-scores.
All analyses were conducted using STATA (version 14), within the National Safe Haven of Scotland, managed by NHS NSS.

Data linkage and analysis datasets
The linkage process and derivation of the analysis datasets is summarised in Fig. 1.
The resultant dataset comprised 4127 singleton births following fresh ART and 1091 singleton births from FET cycles, along with 20,879 matched NC children. See Additional file 1: Tables S1 and S2 for study characteristics and main outcome variables.
The proportions and characteristics of the infants contributing to each of the assessments did not vary between conception groups (Additional file 1: Table S3).
Analyses conducted, outcomes and covariates considered and tables in which corresponding results are presented are summarised in Additional file 1: Table S4.

Newborn health: ART vs. naturally conceived (NC)
To quantify the impact of ART on newborn health, we first compared all ART children to their NC counterparts. Controlling for deprivation, maternal age and maternal smoking status during pregnancy, there was a significant association between each of birth weight (Table 1), OFC and crown-heel length and ART conception (Additional file 1: Table S5).
Babies conceived from fresh transfers, but not FET, had a higher rate of admission to NICU [OR = 1.24; 95% CI (1.10, 1.40) for fresh and OR = 1.00; 95% CI (0.80, 1.26) for FET]. However, this effect did not remain when we adjusted for gestation and birth weight (Additional file 1: Table S5).

Newborn health: ART treatment variables
Within the ART cohort, we examined the effect of ART treatment factors on birth outcomes (Table 2; Additional file 1: Table S6), controlling for differences in infertility diagnosis, treatment and parity in addition to those factors considered above.
The difference in (adjusted) birth weight between babies from FET, compared to fresh transfer, was estimated as 127. 4 Table S7).
Analysis of ART factors (Additional file 1: Table S9) shows that babies from FET cycles are significantly heavier at 6-8 weeks than those from fresh cycles by 133 g [95% CI (75, 190)] and they tend to remain heavier at school entry, although significance is lost [by 258 g; 95% CI (-116, 651)]. FET babies are significantly taller at school entry [by 0.56 cm; 95% CI (0.07, 1.11)], but no significant difference was seen in BMI. No significant difference in the growth parameters was seen with any of the other ART treatment/diagnosis factors.

Infant growth rates from birth to 6-8 weeks and school entry
We have restricted this analysis to a consideration of weight, which we have for all three time points. Table 3 shows the linear regression models for growth rate, controlling for gender, deprivation, maternal smoking and age and type of feeding. Babies born from fresh ART grew significantly faster between birth and 6-8 weeks than NC babies, by 7.2 g/week [95% CI (2.1, 12.7)], while babies born from FET were not significantly different. ART-conceived babies grew at a greater rate than NC babies over the period 6-8 weeks to school entry [Fresh vs. NC, 0.25 g/week, 95% CI (-0.34, 0.88); FET vs. NC, 0.74 g/week, 95% CI (-0.39, 1.90)].
No ART-related factors were significantly associated with either growth rate (Additional file 1: Table S10).

Catch-up growth
Of particular interest is the proportion of infants who show accelerated ('catch-up') or decelerated ('catchdown') growth (weight) from birth to school entry. In Table 4, we show the numbers who crossed more than one decile in the age/gender adjusted growth charts in the three conception groups. Significantly, more fresh ART babies show catch-up growth from birth to school entry, compared to NC (27% vs. 24%; P = 0.004), but this is not seen in just the smaller babies (lower 3 deciles). Significantly, less catch-down growth is seen in the fresh ART cohort (vs. NC, 23% vs. 27%; P = 0.002) and this is also seen when considering only the largest babies (43% vs. 48%; P = 0.014). FET babies generally show more similar rates of catch-up and catch-down growth to NC babies than to fresh ART babies.

Discussion
Birth weight and early child growth are important predictors of long-term CMD risk [6]. Although it is well established that ART children are at increased risk of low birth weight, there is little understanding of the causal factors in the ART process and there have been no large studies of postnatal growth in these children. Using the UK ART data register held by the HFEA, linked to maternity and child health data, we conducted a large-scale analysis of more than 5000 singleton ART children. Our key findings are that not only do ART babies born from fresh embryo transfers have lower weights, head circumference and length at birth than NC children, they grow more quickly and catch up by school age. Babies from FET cycles, by contrast, have greater weight, head circumference and length at birth and show similar growth to NC children.
The major strength of this study is the registry-based national cohort design. The loss to follow-up is a concern, although we believe most of this loss is administrative and we could detect no strong biases in those with  Table S4 Quoted effect sizes represent the difference from the indicated reference category. The effect size for maternal smoking represents the difference between mothers who smoked and those who did not smoke during pregnancy. The effect size for ICSI represents the difference between ART cycles in which ICSI was used and those in which it was not. The effect size for infertility causes represents the difference between the stated cause being identified within the couple and not being identified P values are calculated from an omnibus chi-squared test that parameter estimates are simultaneously zero Note that for ease of interpretation, the effect sizes for the linear and quadratic components of maternal/paternal age represent the change in outcome for a 10-year change in maternal/paternal age and without data. The data on NC infants did not include parity and, therefore, we were not able to include this in the birth weight adjustment. It is likely that more NC babies were born to parous women than ART-conceived babies, which would lead to a small, but unquantifiable, over-estimation of birth weight in this group, in turn reducing the effect in ART-Fresh cycles while increasing the effect in FET cycles. Within-ART comparisons were parity-adjusted. Postnatal growth estimates for both ART-conceived and NC children are based on changes and are therefore largely unaffected by parity. Newborn infants conceived by fresh embryo transfer were smaller than NC infants, with significant differences in birth weight, length and OFC. Some infants were small because they were born early, at gestational ages of < 37 or < 32 weeks, and there were more admissions to the NICU, probably due to conditions associated with prematurity. Other infants were small for gestational age (SGA), which may be due to a slowing down of intrauterine growth, implicating a reduction in placental transfer of nutrients. FET infants by comparison had a higher birth weight, length and OFC than NC infants. The birth weight findings are consistent with previous studies [2,4,22], and the weight differences are remarkably consistent across studies, e.g. in a recent prospective RCT in China, FET babies were 161 g larger than fresh ART babies [16] and in a large Japanese registry study, FET babies were 91 g larger than fresh ART, and 40 g larger than NC babies [23], very similar findings to ours. At 6 to 8 weeks of age, the weight, length and OFC of the FET infants and NC infants were similar, while weight, length and OFC in the fresh ART-conceived infants remained smaller. At school entry, weight, length and BMI in the three study groups were similar, showing that low birth weight fresh ART infants, when free from the constraints on intrauterine growth, resume their growth trajectory postnatally and attempt to achieve their genetic potential by catch-up growth. OFC measurements were not available; however, brain growth and somatic growth are likely to be in harmony. The BMI values at school entry were consistent with those reported previously at a similar age. Previous studies of ART child growth have been small (≤ 500 children) and unable to adequately control for confounding factors, and show conflicting results [12]. In three small European cohort studies, children born from intra-cytoplasmic sperm injection (ICSI) were found to be lighter than their target weight (for age and height) at age 3 years, but not at age 5 years [12]. In a longitudinal study of approximately 200 ART children, they had lower weight at birth and 3 months, with no difference in early growth rate (velocity) compared to a subfertile NC control group; however, ART children then showed greater growth velocity from 3 months to 12 months, followed by similar childhood growth rates to age 3 [13]. Small studies of IVF and ICSI children measured at birth and up to age 18 months in Taiwan [24], 3 years in the USA [25], 4 years in the Netherlands [26] and 10-12 years in the UK [27] showed few differences in growth [12]. Data from a UK child screening programme [28] showed that ART children were less likely to be overweight at 5 years old than controls, with significantly lower BMIs at 7 years of age than controls consisting of births after normal conception or to subfertile couples. Green et al. [29] controlled for embryo freezing as a confounding factor, and showed that fresh  Table 3 Growth rates of singleton births from fresh ART transfers, frozen transfers (FET) and naturally conceived (NC) babies between birth and 6-8 weeks, and 6-8 weeks and primary school entry Results from multiple linear regression models for growth (change in weight) between birth and 6-8 weeks (average weekly growth rate) and 6-8 weeks and primary school entry (average monthly growth rate), adjusting for type of conception, Scottish Index of Multiple Deprivation (IMD), maternal age and smoking status during pregnancy, feed type at 10 days and the ART-NC matching variables: gender, year and month of delivery and delivery hospital. Full model details are available in Additional file 1: Table S4 Quoted effect sizes represent the difference from the indicated reference category. The 'Constant' coefficient represents the average rate of growth for a naturally conceived boy, breast-fed at 10 days by a non-smoking mother who resided in an area belonging to the least deprived quintile P values are calculated from an omnibus chi-squared test that parameter estimates are simultaneously zero Note that for ease of interpretation, the effect sizes for the linear and quadratic components of maternal age represent the change in outcome for a 10-year change in maternal age Catch up/down is defined as a change of more than one decile in the standardised growth charts (between birth and primary school entry) P values are from a χ 2 test comparing, separately, ART fresh with naturally conceived children and FET with naturally conceived children Numbers and proportions are shown for all children experiencing catch-up (first row) and catch-down (second row) growth: then catch-up growth for those babies who were small at birth (defined here as 1st-3rd deciles -third row) and catch-down growth for those babies who were large at birth (defined here as 8th-10th deciles -fourth row) transfer IVF children, while lower birthweight than FET or NC children, did not show catch up weight growth in childhood; however, ART conceived girls were taller than matched controls after fresh IVF and also FET [29,30].
Our data extend this current literature significantly by virtue of the large sample size and careful adjustment for confounding factors, by use of accurate growth data recorded by maternity units including length and OFC in addition to birth weight and, most importantly, by following growth of children to age of school entry. It is currently unknown what alters fetal and child growth trajectories for children conceived through ART; however, it is noteworthy that the magnitude of impact of ART on birth weight is nearly as great as that of maternal smoking in pregnancy. Catch-up or catch-down growth (centile crossing) after birth are both significant predictors of long-term disease risk. Most LBW children show catch-up growth in the first few months, with accelerated growth associated with increased risk of obesity and CMD risk [11,31,32]. Our study shows that fresh ART babies grow faster, from birth, over the first few weeks, than their NC counterparts with further catch-up growth then seen in fresh ART children by school entry age. This is associated, in NC children, with increased risk of obesity and CMD in adulthood [10-12, 33, 34] and, in ART children, with greater weight gain during childhood, associated with increased cardiovascular risks [13]. Reassuringly, however, catch-up was not seen in the smallest fresh ART birth weight centiles in our study, whereas the increased rate of catch-down growth observed was strongest for the largest birth weight centile babies.
Our major finding that children born from FET cycles differ significantly in birth weight and child growth from their fresh ART counterparts, strongly implicates some aspect of the ART treatment process itself rather than parental subfertility, as does the animal literature showing that adverse outcomes can be phenocopied in fertile mice subjected to ART [9]. However, the mechanisms involved are unknown. One plausible hypothesis is that the uterine environment is hormonally dysregulated following ovarian hyperstimulation, such that in fresh cycles fetal growth is restricted due to impaired placental function [5,35]. Alternatively, the ART environment may act via a direct epigenetic effect on the embryo; however, this would predict that the impaired fetal growth trajectory would be continued into postnatal life, and so is not supported by our data. Fetal growth in FET cycles may also be affected by some aspect of the embryo cryopreservation process, consistent with our previous finding that frozen/thawed embryos show altered gene expression [5] and our finding (and from others [23]) that FET babies have a consistently slightly higher birth weight compared to NC babies.
Other than the effects of embryo cryopreservation/ FET, we failed to detect any significant associations between newborn or child weight, length and OFC and other ART patient or treatment factors available. However, the HFEA database lacks information on factors such as the type of culture medium embryos were grown in, an important omission in light of recent evidence that culture medium is associated with altered embryo gene expression [15], epigenetics [9] and fetal and postnatal growth following ART [15], resulting in calls for action from professional societies and leading scientific journals [5].

Conclusions
These longitudinal growth data from a national cohort of more than 5000 ART children identify an increased risk profile for non-communicable disease in later life. ART treatments are constantly evolving, and these studies will need to be replicated in more recent cohorts to capture changes in technology and practice. There is clearly a need to further the research agenda in this area and to take steps to reduce risks from treatment. We call for more detailed studies of the impact of modifiable aspects of IVF on early child growth to reduce risk to the next generation of ART children.

Availability of data and materials
The data that support the findings of this study are available from The Human Fertilisation and Embryology Authority and NHS National Services Scotland but restrictions apply to the availability of these data, which were used under licence for the current study, and so are not publicly available.
Authors' contributions DRB (principal investigator) and SAR and NM (co-applicants) wrote the application to secure the funding. DRB, SAR and MH undertook negotiations with the HFEA and NHS NSS to obtain access to the data and affect its transfer to the National Safe Haven. MH and SAR undertook all analyses with expert clinical direction from DRB and PC. MH, SAR, DRB and SDS wrote the manuscript: all authors critically reviewed the manuscript. All authors read and approved the final manuscript.

Consent for publication
Not applicable.