Impact of maternal age on birthweight in frozen embryo transfer cycles: an analysis including 12565 singleton newborns

Background: Previous studies have investigated the effect of maternal age on assisted reproductive technology (ART) success rates. However, little is known about the relationship between maternal age and neonatal birthweight in frozen embryo transfer (FET) cycles. Does maternal age have an impact on singleton birthweight in FET cycles? Methods: This retrospective study was conducted at a tertiary care centre, involving singleton live births born to women undergoing frozen-thawed embryo transfer during the period from January 2010 to December 2017. A total of 12565 women who fullled the inclusion criteria were enrolled and were grouped into four groups according to the maternal age: <30, 30–34, 35-39, and ≥ 40 years old. Maternal age <30 years old was taken as a reference group. Singleton birthweight was the key outcome measure. A multivariable linear regression analysis was conducted to reveal the relationship between maternal age and neonatal birthweight with controlling for a number of potential confounders. Results: A modest decrease but no signicant difference in birthweight and gestational age- and gender-adjusted birthweight (Z-scores) was observed in maternal age over 35 years old as compared with group with <30years old. The highest proportions of LBW (4.1%), HBW (1.2%), PTB (5.9%) and very PTB (0.9%) were found in group over 40 years old, but no signicant difference was observed among four groups. Additionally, the group with 35-39 years old had the highest very LBW (0.6%) , while the group with 30-34 years old had the lowest SGA (2.7%). However, multivariate analyses revealed that neonatal outcomes including PTB, LBW and SGA were similar between the different maternal age groups. Conclusion: Grouping with different maternal age was not associated with mean birthweight and Z-scores of singletons resulting from FET.

problems [10]. However, little is known regarding the in uence of maternal age on birthweight in vitri edthawed embryo transfer cycles.
To our knowledge, only few studies have examined the effect of maternal age on birthweight with ART, containing different kinds of treatments without considering enough confounders [11][12][13][14]. Further, the published data, except for Lin's study [14], exclusively focused on fresh IVF cycles, without ruling out the possibility of adverse fetal growth caused by a hypoestrogenic milieu. Of note, supraphysiological estrogen levels during ovarian stimulation can create a suboptimal peri-implantation environment for implantation and placentation, thus causing abnormal fetal growth including LBW and SGA [15,16].
Unlike fresh ovarian stimulation cycles, FET seems to provide a more physiological uterine environment for early fetal development [17]. Thus, the current study aims to explore the effect of maternal age on the birthweight of newborns conceived by embryo transfers during FET cycles.

Study design and population
This retrospective study involved women who had undergone FET during the period from January 2010 to December 2017, which was performed at the Department of Assisted Reproduction of the Ninth People's Hospital of Shanghai Jiao Tong University School of Medicine. Women who met the following inclusion criteria were involved in the study: BMI <30 kg/m 2 , the transfer of embryos resulting in a live singleton birth. In all FET cycle, no more than 2 embryos can be transferred. Further, only the rst live birth IVF/ICSI cycles were retained for the women wo had more than one delivery during the study period. The exclusion criteria were as follows: vanishing twin syndrome, congenital uterine malformations, the presence of submucosal broids or polyps and intramural broids > 4 cm determined by ultrasound or hysteroscopy. And exclude women with gestational diabetes, pregnancy-induced hypertension and preeclampsia, as these pregnancy-related factors may have bad effect on intrauterine fetal growth. This study was approved by the Institutional Review Board of the hospital and was carried out in accordance with the Helsinki Declaration. Informed consent was not required result from the retrospective nature, and patients' data were used anonymously.

Laboratory protocols
The procedures of ovarian stimulation, oocyte retrieval and IVF/ICSI have been described in previous studies [18,19]. In brief, IVF or ICSI was conventionally carried out according to semen parameters and previous fertilization histories. For IVF, oocytes were inseminated in human tubal uid (HTF; Irvine Scienti c) which was supplemented with 10% serum substitute supplement (SSS; Irvine Scienti c) and 300 000 progressively motile spermatozoa. For ICSI, oocytes were transferred into dishes immediately after microinjection with HTF+10% SSS. The assessment of fertilization was performed 16-18 h after insemination/injection. A dish containing preequilibrated culture medium was then prepared for the transfer of zygotes. Before 2013, embryos were cultured in early cleavage medium (Irvine Scienti c) before Day 3 and then in multiblast medium (Irvine Scienti c). However, a continuous single culture medium (Irvine Scienti c) was introduced after January 2013. All embryos were cultured under mineral oil and grew in the incubator at 37 ℃, under 5% O 2 and 6% CO 2 concentration (the balance gas was nitrogen). Except for the change of culture medium types, no change was made for the other laboratory conditions and IVF protocols throughout the study period..

Endometrial preparation and vitri cation
Protocols of endometrial preparation for FET have been previously described [20]. Brie y, a natural cycle FET was suitable for women having regular menstrual cycles with the use of hCG for triggering ovulation. Arti cial cycles were offered for women with irregular cycles according to the discretion of treating physicians. The procedure of vitri cation and thawing were previously described [19]. In short, Cryotop carrier system with dimethylsulfoxide-ethylene glycol-sucrose was used as cryoprotectants for embryo vitri cation. Dilution solution in a sequential manner (1 mol/L to 0.5 mol/L to 0 mol/L sucrose) was used for embryos thawing. All embryos were thawed on the day of transfer.

Maternal age
Maternal age at the birth of the child was the key explanatory variable, which was divided into the following categories: <30, 30-34, 35-39, and ≥40 years old. The age group <30 years old was set as the reference category in our analyses.

Outcome measures
The primary neonatal outcomes focused on live singleton birthweight, including Z-scores, birthweight categories and birthweight percentiles. Secondary outcome measures were associated with neonatal outcomes, including gestational age at birth and newborn gender. The de nition for live singleton birth was a delivery of a singleton viable infant after the 24th gestational week. Gestational age (GA) in FET cycles was calculated from the day of embryo transfer (Day 17 for cleavage-stage embryo transfer and Day 19 for blastocysts embryo transfer ) [21]. The de nitions for PTB and very PTB were live births at <37 and <32 completed gestational weeks, respectively. Z-scores were calculated according to gestational age and newborn gender on birthweight based on the national birthweight reference as previously described [22,23]. A birth weight <2500 g was de ned as LBW and <1500 g was de ned as very LBW. Birthweight of infant was divided as follows: LBW (<2500 g), very LBW (<1500 g), high birthweight (HBW) (>4500 g) and normal birthweight. Birthweight percentiles were also based on the national birthweight reference [23] and were divided into the following categoriesas: SGA de ned as birthweight <10th percentile, very SGA de ned as birthweight <3rd percentile, large for gestational age (LGA) de ned as birthweight >90th percentile and very LGA de ned as birthweight >97th percentile.
One-way analysis of variance was performed for continuous data, while Pearson's chi-squared test or Fisher's exact test were applied for categorical data. A post hoc Bonferroni correction was performed for multiple comparisons. The association between maternal age and neonatal outcomes was detected by multivariable logistic regression analysis, while the independent effect of maternal age on neonatal outcomes was analyzed by a multiple linear regression.
The multivariable analyses included the following confounders: maternal BMI, paternal age, parity, infertility cause and duration, insemination method, the type of endometrial preparation, endometrial thickness, the year of treatment and newborn gender. In multivariable models, continuous covariates (maternal BMI, paternal age, infertility duration, endometrial thickness and the year of treatment) and categorical covariates were indicated in Table 1. Maternal age <30 years old was taken as a reference group in multivariable analyses. For the development of IVF techniques over time [24], a sensitivity analysis was performed on treating the year of treatment as a categorical variable. All analyses were conducted with SPSS Statistics (version 21.0) and P <0.05 was considered to be statistically signi cant.

Results
The nal dataset included 12565 women who ful lled the inclusion criteria, with no loss to follow-up.
Baseline demographic and cycle characteristics are presented in Table 1. Comparison between the reference group and other three groups revealed signi cant difference for maternal BMI, infertility cause, parity, infertility duration, FET cycle rank, fertilization method, the number of embryo transferred, FET endometrial preparation, endometrial thickness and year of treatment. Infertility duration and embryo developmental stage at transfer did not differ signi cantly among maternal age categories.
Neonatal outcomes strati ed by maternal age are listed in Table 2. There was no signi cant difference on gestational age, mean birthweight and gestation-adjusted Z-scores varied signi cantly according to maternal age categories. The group with 30-34 years old had the highest birthweight and Z-scores (3355.8±483.3 g, 0.38±1.03). With the increase of maternal age, a modest decrease of birthweight was observed, and group over 40 years old had the lowest birthweight values as 3321.6±503.9 g. Additionally, no signi cant differences were found between any two groups by Post hoc analysis on birthweight and Zscores. The highest proportions of LBW (4.1%), HBW (1.2%), PTB (5.9%) and very PTB (0.9%) were found in group over 40 years old, but no difference was observed among four groups. Interestingly, the group with 35-39 years old had the highest very LBW (0.6%) , while the group with 30-34 years old had the lowest SGA (2.7%). Furthermore, there was no difference in very LGA, LGA, very SGA, and newborn gender between groups.
In multivariate analyses (Table 3), the neonatal outcomes including PTB, LBW, HBW, SGA and LGA were similar between the different maternal age groups. The odds of PTB and LBW was lower in group over 40 years old as compared with the reference group, not reaching signi cant difference. Although the analysis of very PTB (<32 weeks), very LBW(<1500 g) were performed, the number of cases in the two categories was too small to make any meaningful comparisons. In addition, no signi cant difference were found on birthweight percentiles categories between the reference and other three groups.
Multiple linear regression analyses were conducted to assess the relationship between maternal age and birthweight (

Discussion
Recent years, much attention has been paid to the impact of maternal age on ART success rates, and several studies have revealed that a high maternal age (over 40 years old) has a negative effect on pregnancy outcomes [11,13]. However, our study showed that no signi cant association existed between maternal age and singleton birthweight in FET cycles with considering related confounders. Furthermore, linear regression indicated that maternal age was not an independent predictor of singleton birthweight in FET cycles.
Four studies, as we know, have been interested in the potential relationship between maternal age and neonatal birthweight. Wennberg et al. investigated maternal age in uencing adverse maternal and neonatal outcomes following ART treatment, suggesting that the risk of LBW and very LBW was signi cantly higher in ART than in spontaneous conception (SC) singletons in all ages up to maternal age 40 years (LBW: aORs 1.44-2.35; VLBW: aORs 1.67-3.44). Additionally, when the analysis was restricted to maternal age >35 years, an increased risk of LBW existed for SC pregnancies, but not for ART pregnancies [11]. Due to medical, educational and socioeconomic reasons, women who conceive through ART with >35 years old may pay more attention to their state of health and seek medical assistance more often than SC women, which could result in increased detection of complications and decreased risk of LBW. Secondly, Moaddab et al. found that maternal age did not predict newborns' birthweight in pregnancies with maternal age grouping as <40, 40-44, 45-49, ≥50 years old [12]. However, maternal age group with <40 years old per se have an inadequate analysis on birthweight and miss important information. There may be some interesting ndings among groups at ages <25, 26-30, 31-34, 35-40 years old and the increased risk of LBW may appear at age of 35 years old, but this study seemed to miss these potential information. Many studies have set up more detailed groups with maternal age under 40 years old to assess the in uence of maternal age on neonatal outcomes and gained the more credible results [13,14]. Another study reported that the risk of LBW was increased only at maternal ages over 40 years old (6 percentage points, 95% CI: 0.2, 12) with medically assisted reproduction (MAR) compared to the risk of LBW at ages 30-34 years old [13]. However, a limited number of confounders were included in the study, and the effect of different kinds of MAR treatments could not be reliably distinguished, which included less invasive treatments such as ovulation induction only that were less strongly associated with adverse birth outcomes [25]. A recent study based on 4,958 infertile women using a freeze-all strategy observed that maternal age grouping was not related with increased risks of LBW, very LBW, preterm LBW and macrosomia [14]. Yet only 1450 singleton live birth was involved for the analysis of LBW, and the 44-50 year old group of singleton live births was very small (n=9), which may limit the power of statistical analyses between groups.
Our study aimed to improve on the aws of abovementioned studies, and focused on the exact role of maternal age in the singleton birthweight after FET cycles. The current study based on 12565 singleton newborns born after FET cycles, demonstrated that maternal age itself had no impact on singleton birthweight and neonatal outcomes including PTB, LBW and SGA were similar between the different maternal age groups. Due to so many confounding factors, direct comparability across different age group has very limited clinical signi cance in table 2. Because of strict exclusion criteria, generality of this nding may be, to some extent, restricted. The reason why no signi cant correlation existed between maternal age and birthweight in our study is likely complex. It is generally known that with aging comes a reduction in ovarian function, resulting in the decrease of ovarian response to ovulation promoting drugs and the low number of oocytes retrieved [26]. Additionally, the decreased quality of oocytes [27, 28], abnormal endometrial function and degeneration of multiple organs function will appear in women with advanced maternal age [22,29]. All the above mentioned factors would affect the development of embryo and cause adverse effects on the newborn, leading to LBW. However, with the popularization of education, many women tend to choose late marriage and late childbearing and enjoy a simple single life before marriage. In this kind of life, they are less stressed and have more opportunities to get in touch with life freedom than the women who are married and have children. Meanwhile, these knowledge women tend to choose a healthy and regular life, and possess good habits, physical quality and economic conditions, thus having a better choice on ART treatment [11]. Aging leads to an irreversible decline in fertility, forcing older women to pay more attention on pregnancy and to seek medical help more actively than the young. In addition, the spouse's income also increase with age to guarantee maternity. Most importantly, the development of ART has well ful lled the reproductive needs of women with different ages to improve the quality of newborns.
In this study, results from multiple linear regression analysis indicated that maternal BMI, embryo developmental stage at transfer, parity, number of embryos transferred, endometrial thickness, year of treatment, gestational age and newborn gender, were the independent predictors for neonatal birthweight, which was consistent with previous results [22,30,31]. Z-scores were calculated and compared across the four groups in order to reduce bias caused by newborn gender and gestational age, and no signi cant difference on Z-scores was found among different maternal age groups. In addition, there were signi cant differences between the maternal age groups in baseline and cycle characteristics including infertility duration, infertility cause and fertilization method. However, these confounders had no impact on neonatal birthweight based on the linear regression model.
There are limitations in our study. The biggest one is its retrospective design, so we strictly checked the database with strict criteria. Secondly, due to personal privacy restrictions, we were unable to obtain the education and economic background of patients. Thirdly, there are many confounding factors strongly associated with birthweight in linear regression analysis, data bias during the experimental design cannot be all corrected by regression equations. Further, embryo quality and paternal BMI were important factor that may affect neonatal outcomes, yet these data were missed in this manuscript. However, The large number of singleton live births from a single centre can assure the practice consistency, which is the main strength of the current study. Additionally, aside from the change of culture medium types, all other laboratory conditions and protocols remained invariant throughout the study period. Further, maternal age was recorded according to the identi cation card, and endometrial thickness was measured by the same trained sonographers, reducing the recorder variability. Importantly, a number of potential confounders were included in our study, which may minimize their impact on the ndings.

Conclusions
This study expands the current knowledge about association between maternal age and neonatal outcomes, especially indicated that maternal age was not associated with mean birthweight and Zscores. This important nding should be adequately applied for women over 40 years old prior to FET and strengthen their con dence. A large prospective study, of course, are needed to verify our ndings in future.

Declarations
Ethics approval and consent to participate

Consent for publication
Not applicable.

Availability of data and materials
The data that support the ndings of this study are available from the Ninth People's Hospital of Shanghai Jiao Tong University School of Medicine, but restrictions apply to the availability of these data, which were used under guidelines for the current study, and so are not publicly available.

Competing Interest
The authors have no con icts of interest to declare. Funding the National Natural Science Foundation of China (81774352, 81573755, 81503604, 81703874).
Authors' contributions C.Q.Y and Y.P.K. conceived and designed this study. Z.X.N. and S.S. contributed to data acquisition, analysis and interpretation and drafted the manuscript. W.C., J.Y., D.Z, D.Y.Z., and Z.L.C. were responsible for the collection of data. All authors interpreted the data.