We conducted a nationwide retrospective cohort study using the Korean National Health Insurance Service (K-NHIS) database, which covers 50 million people (approximately 99% of the South Korean population), from 2004 to 2020 [9]. The K-NHIS database represents the entire population of South Korea and contains national records of all covered inpatient and outpatient visits, procedures, and prescriptions. The K-NHIS also includes data from annual or biennial health screening exams for adults and health examinations based on age group for infants provided free of charge by the Ministry of Health and Welfare.

Because we used 2004 as the washout period and 2020 as the follow-up period, our cohort included all pregnancies resulting in live births from January 1, 2005 to December 31, 2019, identified with procedure codes of obstetric delivery. We included all liveborn infants who were linked with their mothers and restricted the pregnancy cohort to women at delivery. We included first-delivery women in the dataset (n=3,685,817). Furthermore, for the sensitivity analysis, we restricted patients who had infant health screening exam 4 years prior to the baseline, and we used the weight of baby from the health screening exam.

The need for informed consent was waived as this study was conducted using anonymized claims data. This study was approved by the Institutional Review Board of our institution (SMC 2021-08-107).

The K-NHIS data comprise individual-level demographics and all records of diagnosis and healthcare utilization (e.g., drug prescription and medical procedure), provided through inpatient, outpatient, and emergency department visits. The NHIS claims for inpatient and outpatient visits, procedures, and prescriptions were coded using the International Classification of Diseases (ICD), 10th revision [10]. The exposed group was age at initial birth. The study endpoints were chromosomal abnormalities, major congenital malformations, adverse short-term neonatal outcomes, and long-term developmental infant outcomes for the offspring. Chromosomal abnormalities and major congenital malformations were identified by diagnostic records, according to the ICD-10 codes defined by the European surveillance of congenital anomalies classification [11]. Chromosomal abnormalities were categorized into autosomal trisomy and others. Major congenital malformations were further categorized into 12 types of organ-specific malformations (Supplementary Table 1): 1) nervous system; 2) eye; 3) ear, face, and neck; 4) heart; 5) respiratory system; 6) oral cleft; 7) digestive system; 8) abdominal wall; 9) urinary system; 10) genital organs; 11) limb; and 12) other malformations [12]. Outcomes, including preterm birth (O60.1 and O60.3), placenta previa (O44 and O69.4), preeclampsia (O15), small for gestational age (O36.5), and large for gestational age (O36.6) were also investigated according to the ICD-10 codes. The short-term neonatal outcome was NICU admission and composite morbidity, which included any of the following: sepsis, transient tachypnea, respiratory distress syndrome, necrotizing enterocolitis, intraventricular hemorrhage, and bronchopulmonary dysplasia (Supplementary Table 1). The long-term outcome was defined according to pre-specified neurological and neurodevelopmental diagnoses as any one or more of the following: autism, attention-deficit/hyperactivity syndrome (ADHD), cerebral palsy, any developmental delay including motor or cognitive delay, epileptic and febrile seizures, and tics and stereotypic behavior as identified by diagnostic records according to the ICD-10 codes (Supplementary Table 1). As the NHIS routinely audits the claims, such data are considered reliable and used in numerous peer-reviewed publications [13,14]. In a validation study comparing our database and electronic medical records, the overall positive predictive value of diagnosis records was 82% [15].

We considered a broad range of covariates as potential confounders or proxies of potential confounders. Maternal comorbidities, including history of congestive heart failure, diabetes mellitus, renal disease, and cancer within the past year prior to birth were summarized using the Charlson’s index [16,17]. We also assessed hypertension (ICD-10 code: I10-I13, and I15), hypertensive disorders during pregnancy (ICD-10 code: O14, O11, O15, O13, O16, I10, and O10), gestational diabetes (ICD-10 code: O244 and O249), and overt diabetes (ICD-10 code: O240, O241, O242, O243, E10, E11, E12, E13, and E14). To control the confounding factors, we adjusted for Charlson’s index, chronic hypertension, hypertensive disorders during pregnancy, gestational diabetes, overt diabetes, cesarean delivery, income, neonatal sex, birth year, and multifetal pregnancies.

In addition, we obtained the birth weight for infants from the infant health screenings exam, which was provided by the K-NHIS and based on age group.

For major congenital malformations and short-term outcomes, odds ratios (ORs) with 95% confidence intervals (CIs) were calculated using logistic regression. For long-term outcomes, infants were followed from birth to event, death, or end of the study period (December 2020), whichever came first. Person-time was calculated from the date of birth to the date of the incidence of event, death, or last follow-up. We used the Cox proportional hazards regression models to estimate the hazard ratio (HR) with 95% CI to compare the control groups. We examined the proportional hazards assumption using plots of the log (-log) survival function and Schoenfeld residuals. Considering multifetal pregnancies, including twins, triplets, and higher-order pregnancies, could be correlated within a mother, and we performed a mixed model analysis to adjust for this correlation in all the analyses. We also modeled maternal age as a continuous variable using restricted cubic splines with knots at the 5th, 35th, 65th, and 95th percentiles of the sample distribution to visualize maternal age-dependent change in long-term infant outcomes.

Furthermore, we conducted a quantitative bias analysis based on the probabilistic method among unexposed pregnancies without malformation where (P00) was defined as 80% [18]. Following this, the probability of live birth among unexposed pregnancies with malformations where (P10) was considered to be 60%, based on a previous study [19]. In addition, a five-time-higher frequency of stillbirth was considered in the older group compared with that in the general population based on previous literature [20]. Considering a difference in probability of live births between the older and younger groups, the corrected relative risks were calculated as follows: corrected OR and observed OR× (P₁₀×P₀₁/P₁₁×P₀₀).

All analyses were two-sided and P-values <0.05 were considered statistically significant. Statistical analyses were performed using SAS version 9.2 (SAS Institute Inc, Cary, NC, USA) and R software version 3.3.2 (R Foundation for Statistical Computing, Vienna, Austria).

The median (inter-quartile) age of study participants was 32 years (29-35). Of all participants, 4.2%, 22.9%, 47.1%, 21.4%, 4.1%, and 0.3% were the first birth at age <25 years, 25-29 years, 30-34 years, 35-39 years, 40-44 years, and >44 years, respectively, during the total study period. The changes in maternal age over the study period are depicted in Fig. 1. In detail, 15.96% of pregnant women in 2005 and 30.44% in 2019 were between the ages of 35 and 39. Only 2.06% of pregnant women were between the ages of 40 and 45 in 2005; however, this proportion increased to 7.47% in 2019.

Table 1 summarizes the characteristics of study population, including maternal and neonatal outcomes according to the maternal age group. Advanced maternal age, defined as 35 years old and older, was associated with higher family income, decreased residency in rural areas, and increased Charlson index. As for pregnancy outcomes, there was a correspondingly higher prevalence of hypertension and diabetes during pregnancy according to advanced maternal age. The rate of cesarean delivery was also significantly increased as follows: 29.5%, 34.7%, 40.5%, 52.5%, 65.3%, and 74.0% in ages <25 years, 25-29 years, 30-34 years, 35-39 years, 40-44 years, and >44 years, respectively. Of note, the rate of multifetal pregnancies peaked in the group with mothers aged 35-39 years (6.0%), whereas preterm delivery rate showed a stepwise increase with advancing maternal age, peaking at 8.3% in nulligravida women above 44 years. Supplementary Table 2 shows additional analyses on some perinatal complications, including placenta previa, preeclampsia, and small and large for gestational age based on maternal age groups. Placenta previa exhibited an age-dependent increase, whereas preeclampsia and small for gestational age showed a U-shaped curve. Of note, the adjusted odds ratios (aOR) placenta previa also showed a stepwise increase, with the highest risk in women aged over 44 years (OR, 3.74; 95% CI, 3.28-4.27).

The absolute risks of offspring with a chromosomal abnormality were 0.06%, 0.04%, 0.06%, 0.09%, 0.21%, and 0.86% in ages <25 years, 25-29 years, 30-34 years, 35-39 years, 40-44 years, and >44 years, respectively. The aORs for chromosomal abnormalities increased in the advanced maternal age group, with a particularly significant increase in autosomal trisomy (Table 2). Of note, the group <25 years also had a higher risk for autosomal trisomy compared with that of the reference group (age, 25-29 years).

The absolute risks for major congenital malformations were 2.9%, 3.0%, 3.4%, 4.1%, 4.8%, and 6.0% in ages <25 years, 25-29 years, 30-34 years, 35-39 years, 40-44 years, and >44 years, respectively. Overall, the aORs of most congenital malformations tended to increase in the advanced maternal age group after adjusting for potential confounders. The most significant dose-response relationship between maternal age and congenital malformations was observed in infant heart defects with the highest risk in women aged over 44 years (OR, 1.46; 95% CI, 1.31-1.62). The highest risk among congenital malformations was found in oral cleft in women aged over 44 years (OR, 2.66; 95% CI, 1.81-3.91). Although maternal age and the risk of oral cleft also showed a dose-response relationship, the youngest age group (<25 years) also exhibited a higher risk (OR, 1.24; 95% CI, 1.05-1.46) compared with that of the reference group (age, 25-29 years), resulting in a U-shaped prevalence. Similarly, the OR of abdominal wall defects was also elevated in the youngest age group (under <25 years) (OR, 1.59; 95% CI, 1.22-2.07) compared with that of the reference group (Table 2).

The aOR of preterm birth was elevated with increasing maternal age, with the highest risk observed in women aged over 44 years (OR, 1.85; 95% CI, 1.71-2.01) as shown in Table 3. Women aged <25 years also had a slightly increased risk for preterm birth (OR, 1.16; 95% CI, 1.13-1.20). As expected, the proportion of admittance to the NICU within a year showed similar trends with preterm birth. The composite outcome of infants showed a stepwise increase according to advanced maternal age. Among them, transient tachypnea, respiratory distress syndrome, and bronchopulmonary dysplasia were remarkable (Table 3).

During the follow-up period (median, 10.4 years), 0.49, 0.24, 0.24, 0.32, 0.43, and 0.73 infants per 1,000 person-years passed away in age groups <25 years, 25-29 years, 30-34 years, 35-39 years, 40-44 years, and >44 years, respectively, manifesting a U-shape curve as shown in Table 4. For long-term developmental problems, the risk of autism, ADHD, cerebral palsy, and cognitive developmental delay increased with advancing maternal age, while offspring of women aged less than 25 years also had a relatively higher risk, and demonstrating a U-shaped prevalence (Table 4). Notably, the risk of epileptic and febrile seizures in offspring increased in the group aged <25 years; this group also showed the lowest adjusted risk for motor developmental delay and tic and stereotypic behavior (Table 4). When the result was represented as a restricted cubic spline curve (Fig. 2), death, ADHD, cerebral palsy, developmental delay, and seizures showed a non-linear association (P for non-linearity <0.05). As neonatal birth weight alone can influence long-term outcomes, a further regression model analysis was performed by adding neonatal weight as a confounding variable. However, not all infants have their birth weights recorded; therefore, the analysis was conducted based on the weights that were registered in the health screening exam. Consequently, the effect of maternal age on the overall developmental problems of offsprings were similar but attenuated as presented in Supplementary Table 3.