Women’s Childbearing Age Associated with Optimal Fetal and Infant Outcomes: a population-based study

Background: Both teenage and delayed childbearing were associated with adverse pregnancy outcomes. However, the optimal maternal age that is associated with the lowest risk of fetal and infant mortalities and morbidity is unclear. Methods: In this retrospective cohort, we pooled nulliparous singleton births from the U.S. Cohort Linked Birth-Infant Death Data, Natality and Fetal Death Data from 2011 to 2017. The study population included non-obese (body mass index < 30 kg/m2) nulliparous women with singleton gestation and greater than 20 weeks of gestation. The optimal maternal age at rst childbirth for the lowest risk of fetal death, infant mortality, and neonatal morbidity (early preterm birth and severe small-for-gestational-age) was identied by using a second-degree fractional polynomial regression, stratied by race/ethnicity. Results: A total of 5,285,126 singleton births were analyzed for the outcomes of infant mortality and morbidity, and 4,386,757 were analyzed for fetal mortality rate. Non-Hispanic White, Hispanics and Asian American women shared similar maternal age patterns. Maternal age at 26-31 years was associated with the lowest risk of adverse fetal and infant outcomes. However, non-Hispanic Black had the optimal age 24 years earlier than other race/ethnic groups, namely 24-27 years old. Conclusions: Maternal age for the optimal pregnancy outcomes is around 24 to 31 years old for the majority of rst-time mothers. Our results may provide useful information for counseling women regarding optimal timing of childbearing.


Plain English Summary
Previous studies have shown that women at older age are more likely to be infertile. There is also a higher risk of adverse pregnancy outcomes, such as gestational diabetes, pregnancy hypertension and preterm birth. The adverse consequencies of teenage pregnancy have also been identi ed. However, few studies have attempted to answer what is the optimal age to give birth, especially for rst-time mothers. Our study found that maternal age for the optimal pregnancy outcomes is around 24 to 31 years old for the majority of rst-time mothers. However, our results should not be used in the clinical settings in a proscriptive way. Rather, it may provide useful information for counseling women regarding optimal timing of pregnancy, according to their own contextual situation.

Background
Delayed childbearing has become a global phenomenon in the past several decades. For instance, in the United States, the mean maternal age at rst birth increased from 24.9 years in 2000 to 26.3 years in 2014; the proportion of rst births to women over 35 years rose from 7.4-9.1% during the same period. 1 The consequences are signi cant as advanced maternal age is associated with increased risks of infertility and adverse pregnancy outcomes. [2][3][4] For instance, female fecundability declines precipitously after 35 years of age 5 while risks of miscarriage and preeclampsia increase. 6 Advanced maternal age is a signi cant risk factor for diabetes (including gestational diabetes), pregnancy hypertension and preterm birth. 7,8 These complications, in turn, contribute to higher neonatal and infant morbidity and mortality in older women. 9 The nancial and social costs of the sequelae of delayed childbearing are signi cant but have not been well evaluated. 10 Numerous studies have examined the consequences of pregnancies in teenagers and advanced maternal age women, but few tried to directly answer what is the optimal maternal age associated with the best perinatal outcomes especially for rst-time mothers. Furthermore, selection bias due to decreasing fecundability with advancing woman's age and over-adjustment of mediation factors may be important methodological de ciencies that may have masked the total effects of advanced maternal age on pregnancy outcomes and child health. More speci cally, less fertile women may be unable to achieve pregnancy with advancing age. Since low fecundability is associated with poor perinatal outcomes, 11 the exclusion of infertile women would lead to an underestimation of age effect on adverse pregnancy outcomes. 12 Besides, many studies adjust for gestational age and pregnancy complications (e.g. gestational diabetes mellitus) in examining the association between maternal age and perinatal outcomes and infant mortality. 7,13 As these complications are age-related, adjusting for these mediators can underestimate the total effects of maternal age on the outcomes of interest. Our study uses population-based data to explore the optimal maternal age that is associated with the lowest risk of adverse fetal and infant outcomes. following women were excluded for the current analysis: 1) race/ethnicity other than non-Hispanic White, non-Hispanic Black, Hispanics or Asian; 2) multiple gestation; 3) multiparous; 4) pregnancies less than 20 weeks of gestation; 5) had pre-pregnant obesity (BMI at or greater than 30 kg/m 2 ). We limited to nonobese women as the effect of pre-pregnancy obesity on perinatal outcomes differs by maternal age and race/ethinicity. 15 For example, it was after 20 years old that obesity began to be positively associated with preterm birth in non-Hispanic White women, while the age of turning point for non-Hispanic Black women was 30 years old.

Page 4/12
The current study measures optimal fetal and infant outcomes based on the following parameters: fetal death, infant death, and neonatal morbidity which was composed of early preterm birth and severe smallfor-gestational-age. Fetal death includes antepartum and intrapartum fetal deaths with 20 weeks of gestation or more. Infant death is death of live-born baby within 1 year of age. Early preterm birth refers to a livebirth born at or after 20 weeks but before 34 weeks of gestation. Severe small-for-gestational-age is de ned as a weight below the 3rd percentile for the gestational age in our dataset.
All analyses were strati ed by maternal race/ethnicity (Non-Hispanic White, Non-Hispanic Black, Non-Hispanic Asian and Hispanic).
To provide a more exible parameterization for maternal age, a second-degree fractional polynomial regression was performed to t a non-linear relationship between continuous maternal age and each outcome of interest. The model restricted powers in a given set of (-2 -1 -0.5 0 0.5 1 2 3), with Χ 0 denoting logarithm of covariate (ln(X)). If a second-degree fractional polynomial model was selected with powers (p 1 , p 2 ), FP function would be logit(π i ) = , where π i was the probability of mortality/morbidity for individual i. The coe cients of maternal age ( ) were generated. The predicted probability for each infant outcome was then calculated by the formula: odds/(1+odds), where odds= . In order to be consistent with the expression of mortality rates (death per 1,000 live births), the predicted probability was expressed as per 1,000. The plot of predicted probability against maternal age was then drawn to visually show the pattern of each outcome by maternal age, by using R studio (Version 1.0.136). All analyses were performed using Stata/IC 15.1.
The maternal age associated with the lowest mortality and morbidity rates was calculated by rst estimating the nal FP model using logistic regression. To derive the optimal maternal age, the rst derivative of the estimated FP function was set to be 0 and used to solve maternal age. For example, FP function as logit (p) = has rst derivative of , and thus the optimal maternal age (minimum value of X) would be . The procedure NLCOM in STATA was used to calculate estimates and 95% con dence intervals (95% CI). Figure 1 shows the selection of the study population. A total of 5,285,126 nulliparous singleton births from the Linked Birth Cohort were included for the analysis of infant mortality and morbidity rates, and 4,386,757 from the combined Natality and Fetal Death datasets for the analysis of fetal mortality rate. Figure 2 and Table 1  ) mortality risks in non-Hispanic Black were generally higher than those in other race/ethnic groups; and the nadirs were all shifted toward 2-4 years younger, i.e., at 24-27 years old. a. The outcome of morbidity combined early preterm delivery and severe small-for-gestational-age.

Discussion
With combined data from the US Linked Birth Cohort Data, Natality Data and Fetal Death Data, our study shows that there is a J-/U-shaped relationship between maternal age and adverse fetal and infant outcomes, i.e., the younger (< 20 years) and older ages (> 35 years) are associated with higher risks of adverse outcomes. Our ndings suggest that the maternal age for the optimal infant outcomes is around 24 to 31 years old for non-obese rst-time mothers in general. Non-Hispanic Black women had the lowest mortality/morbidity rates 2-4 years younger than other race/ethnic groups.
The association between maternal age and adverse infant outcomes results from a mixture of physiological and socioeconomic factors. Physiologically, after puberty, women reach sexually maturity and are prepared to undergo pregnancy. However, young women who are undergoing growth may compete with the developing fetus for intrauterine resources (e.g. nutrients), contributing to restricted fetal growth and, subsequently, place the fetus at the risk for certain health problems. 16 Biological immaturity may also predispose young mothers to infection-mediated preterm delivery. 17 The risk of adverse pregnancy outcomes gradually rises as the reproductive system mature; increased myometrial sclerotic lesions with advancing age may cause uteroplacental under-perfusion and, consequently, increase the risk of stillbirth. 18 Furthermore, chronic diseases (e.g. hypertension, diabetes, renal disease) associated with advancing maternal age, potentially increase the risk of adverse outcomes. 19 Heffner et al. suggested that maternal age between 25 and 35 would be ideal for childbearing in westernized societies. 20 One retrospective cohort study with 203,517 mothers of singleton gestation found that maternal age of 25-30 years old was associated with the lowest risks of low birth weight (< 2500 g), admission to neonatal intensive care unit and perinatal mortality, after controlling for race, parity, insurance type, BMI, pre-existing medical conditions, substance use, and clinical site. 21 Another study using hospital inpatient stay data with 4,109,297 weighted sample, also found that maternal age of 25-29 was related to the lowest odds for most of maternal complications (e.g. chorioamnionitis, preeclampsia, eclampsia, pregnancy-related hypertension) with the adjustment of demographics and comorbidities (e.g. gestational diabetes, cardiac disease). 13 However, as adjustment of age-related medical conditions may undermine the total effects of maternal age on pregnancy outcomes, we did not adjust for medical conditions.
Our study is population-based, involving the whole U.S. birth population and examined race/ethnicityspeci c patterns. Overall, maternal age associated with the optimal fetal and infant outcomes was 24-31 years old for the vast majority of non-obese rst-time women. The optimal age for African American mothers was 2-4 years earlier than other race/ethnic groups. One possible explanation is the cumulative exposure to socioeconomic disadvantage that leads to Black women to experience earlier health deterioration, what Geronimus proposed as "whethering hypothesis". 22 In addition, adaptation to such chronic stress results in wear and tear on the body, which is referred as "allostatic load". 23 Lu and Halfon suggested that allostatic load may affect reproductive health in Black women, which partly explains the racial/ethinic disparities in birth outcomes. 24 The current study has several strengths. It included a population-based large sample size, which allowed the analysis of rare outcomes such as fetaland infant mortality. Statistically, we applied a exible method to t the non-linear and asymmetric relationship between maternal age and each pregnancy outcome, letting the data determine the functional form of the equation relating maternal age and outcome. Use of the fractional polynomial regression preserves the continuous nature of maternal age and the non-linear nature of the association between maternal age and adverse outcomes, enabling us to nd an optimal age rather than an age group.
Nonetheless, our study has some limitations. First, we did not identify the optimal maternal age for maternal morbidity (e.g. hemorrhage, postpartum sepsis, pre-eclampsia/eclampsia), maternal mortality and severe neonatal morbidity because previous studies found that maternal and neonatal morbidity information had substantial underreporting in the birth certi cate data. 25 We, therefore, used severe small-for-gestational-age and early preterm birth as a combined morbidity index. While severe neonatal outcomes have more clinical implications, less severe morbidity may also have long-term concerns for adult chronic diseases. Second, we could not provide a single index to identify the optimal maternal age, as the incidences of mortality (in per 1,000 births) and morbidity (in per 100 births) differed greatly and the datasets used for fetal and infant mortalities were different; instead, we used three indexes. As the maternal age patterns for mortalities and morbidity were very similar, the conclusion is reassured. Third, the ndings in the study apply to non-obese women only. Finally, as we did not have good information on socioeconomic status, the study may be susceptible to residual confounding.

Conclusions
For non-obese rst-time mothers, 24-31 years of age was associated with optimal fetal and infant mortality and morbidity, depending on mothers' race/ethnicity. Our results should not be used in the clinical settings in a proscriptive way. Rather, it may provide useful information for counseling women regarding optimal timing of pregnancy, according to their own contextual situation. Authors' contributions: XZ contributed to the acquisition, analysis and interpretation of data, and drafting of the manuscript; UMR interpreted the data and revised the manuscript substantively. JZ conveived and designed the study, interpreted the data and critically reviewed tha manuscript. All authors reviewed the nal version of the manuscript and gave a nal approval of the version to be submitted.

Figure 1
Flow chart of Retrospective Cohort.

Figure 2
Predicted mortality and morbidity rates plotted against maternal age in different group.