Background
Preterm birth (PTB), defined as births before 37 completed gestational weeks, is an important global public health issue. It is estimated that 14.84 million newborns were born to preterm in 2014 worldwide, accounting for 10.6% of all live births [
1]. Apart from being the leading cause of neonatal and under-5 child mortality worldwide [
2], PTB also increases the risk of short-term and long-term morbidities, such as poor growth, acute morbidity, respiratory illnesses, neurocognitive disorders, and chronic disease in adulthood, which could lead to high social and economic burdens [
3,
4].
Reasons for the rising trend of PTB around the world in recent years [
5‐
7] remain elusive. Upward shift in maternal age, a known risk factor of PTB [
8,
9], may partially explain the increasing trend in PTB [
9,
10]. Maternal early life factors, such as childhood hardships [
11] and being born in a household with lower socio-economic status [
12,
13] have been associated with preterm delivery. A classic age-period-cohort analysis, simultaneously examining maternal age (age effect, reflecting biological changes and social processes) [
14], delivery year (period effect, capturing change given a specific period such as advance in medical services, health policies), and the birth year of mothers (cohort effect, reflecting unique experience/exposure and contextual factors experienced by the cohort over their life time) [
15], is therefore a useful tool to interpret trends of health outcomes [
14,
16,
17]. Most previous studies examining the association of maternal age and delivering period with incidence of PTB did not take into account the role of maternal birth cohorts [
5,
6,
18]. Only one study in the United States has explored the interaction among maternal age, period, and birth cohort and found a strong effect of maternal age and period on PTB and a weak effect of cohort among older African American mothers [
11].
Maternal age at childbirth has been increasing and the distribution of parity has altered over the past decades in China due to the relaxation of the one-child policy [
10,
12]. Recent studies have suggested higher risk of PTB in primiparas than that in multiparas among mothers with advanced age [
13,
19], indicating that the changes in distribution of parity may modify the contribution of maternal age to PTB incidence at a population level. Understanding the mutual effects of parity and mothers’ childbearing age on longitudinal trends in PTB incidence is particularly important in areas where the distributions of such factors have changed remarkably. However, information on the effects of the changing PTB trends at a population level has been lacking.
The present study aimed to examine the contribution of these 3 time-dependent factors to the long-term trends in PTB from 2001 to 2016 in Guangzhou, China. We also assessed whether the complex effects were modified by parity.
Methods
Data sources and population
Data on all singleton live births (
n = 2,553,803) at 20–43 gestational weeks from 2001 to 2016 were retrospectively obtained from the Guangzhou Perinatal Health Care and Delivery Surveillance System [
7,
20] in Guangzhou, China. The data included maternal age, parity, delivery mode, gestational age at birth, newborn’s sex, and birth weight. We excluded mothers who were < 15 or ≥ 45 years of age (
n = 1947), born either before 1961 or after 2000 (
n = 411), non-Mainland Chinese (
n = 8976) and with missing data on maternal age (
n = 1881), parity (
n = 2815), newborn’s sex (
n = 2050) and birth weight (
n = 723), resulting in 2,535,000 singleton live births in the present study. The study was approved by the institutional ethical committee board of Guangzhou Women and Children’s Medical Center (No.201924801).
Age, period, and cohort
Maternal age at delivery (age) was grouped into seven intervals: 15–19, 20–23, 24–27, 28–31, 32–35, 36–39, and 40–44 years. Delivery years (period) were grouped into four 4-year categories (2001–2004, 2005–2008, 2009–2012, and 2013–2016). Ten 4-year maternal birth cohorts were obtained by subtracting age from period and labelled as the midpoint of the birth years (1961, 1965, 1969, 1973, 1977, 1981, 1985, 1989, 1993 and 1997). For example, those delivered during the period 2001–2004 at the age of 20–23 were born between 1978 and 1984, thus the midpoint, 1981, was used to represent their birth cohort.
Outcome measurements
Gestational age was confirmed based on an ultrasound examination in the first or second trimester [
21]. When the examination was unavailable, gestational age was calculated according to the last menstrual period. PTB was defined as delivery before 37 gestational weeks.
Statistical analysis
Poisson regression models with robust variance were used to estimate the change per year of PTB incidence and 95% confidence intervals (CIs). We also examined the interaction effects between maternal delivery year and parity to test whether the change of PTB incidences among primipara and multipara differed over the 16-year period using Poisson regression model. Age-specific PTB incidences were plotted using heatmaps by maternal delivery period and birth cohort for both parity, respectively.
The age-period-cohort (APC) analysis models were used to investigate the effects of maternal age, delivery year and birth year of mothers on the change trends of PTB incidence. The method described by Carstensen was used to solve the exact linear dependency (i.e.,
age = period – cohort) of 3 time-related factors [
22]. The effect of the “drift” variable [
22‐
24] (a combination of linear period effects and cohort effects) was estimated for the overall linear trend of PTB incidence, which cannot be attributable uniquely to period or cohort effects. Deviations from linearity (termed curvature), which are uniquely attributable to period or cohort effects and not dependent on any model constraint, were then estimated as period and cohort effects. To detect non-linear effects on PTB incidence trend, we fitted the parametric smooth functions based on natural splines with 3–6 knots for age, period, and cohort. The full models were carried out as a priori and submodels sequentially were fitted with age by adding drift, period and cohort variables. The deviance of the models was used to measure the goodness of fit. The statistical significance of the difference in deviances between the subsequent models was assessed by a log-likelihood ratio test. A smaller deviance value of the model indicates a better goodness fit to the data.
We performed a sensitivity analysis by restricting to births before 32 completed gestational weeks to explore the influence of the 3 time-related factors on extremely (< 27 weeks) or very (28–31 weeks) preterm delivery, which is an important indicator of improvement of modern intensive care for the premature infant.
The age-period-cohort models were performed separately by primiparas and multiparas, using the apc.fit function in the Epi package [
25] in R (version 3.5.2). The period of 2001–2004 and the year of 1981 (the median maternal birth year among PTB cases) were chosen as the reference groups.
Discussion
In a large sample of 2.5 million pregnant women from Guangzhou, China, we observed that the PTB incidence increased linearly from 2001 to 2016. We also found there were strong effect of maternal age, moderate effect of birth cohort and no or weak effect of delivery period on PTB. A J-shaped and a V-shaped relationship were found between maternal age and PTB among primiparous and multiparous mothers, respectively. Among first delivery mothers, the incidence of PTB was higher among those born in more recent birth cohorts than those born earlier whereas an inverse U-shaped association between maternal birth cohort and risk of PTB was found in multiparous mothers. We observed that weak decreasing period effects on the trend of PTB among multiparous mothers during the period of 2001–2012 and that the risks of extremely or very PTB among both groups in 2009–2012 were the lowest compared with other periods.
Our study builds on previous studies regarding the association of maternal age and delivering period with the incidence of PTB, and further explored the effect of birth cohort and the modification effect by parity. To our knowledge, there has been only one study examining the maternal age-period-cohort effect on PTB in the American population [
11]. While our finding on the association between maternal age and PTB is in agreement with that observed in the US study, Ananth and colleagues found that year of delivery had strong effects on PTB and weak effects for mother’s birth cohort [
11], which was different from our findings. Interestingly, different patterns of maternal age and maternal birth cohort effect were observed between primiparous and multiparous mothers in the present study, which had not been explored in previous related studies. A comprehensive understanding of interaction effects of maternal age, delivery period and maternal birth cohort on PTB and modification effects of parity should inform the design and implementation of public health interventions.
Consistent with previous study [
8] both younger and older maternal age increased the risks of PTB regardless of parity. The increasing trend in preterm birth was mainly driven by upward shift in maternal childbearing age due to experiencing high levels of economic and mental stress, especially in primiparous mothers [
26]. Also, the proportion of advanced-age of multiparous mothers has risen owing to change of family child planning policy, [
12] contributing to the increasing trend of PTB. Our study also showed that primiparous mothers who had higher risk of PTB were younger than multiparous mothers in older age. Larger increase in the prevalence of chronic conditions by maternal age in primipara than multipara [
13,
27] may explain, at least partly, the trend towards younger maternal age with PTB among primipara.
Our study showed that lower risks of extremely or very PTB among both primiparous and multiparous mothers who delivered their children in the period around 2009–2012. Increasing level of educational attainment, enhancement of perinatal care service provision and technological advance in PTB prevention over the past decade may have contributed to the decreasing trend of extremely or very PTB. In our population, we found higher education level was associated with lower risk of preterm delivery, especially among primiparous mothers (Data not shown). However, other risk factors of PTB, such as health problems triggered by urbanization and increasing use of assisted reproductive technology, may counteract the effects of improvement of medical care services, which may partially explain the effects of periods of extremely or very PTB tended to be null in the more recent years [
28,
29]. In Guangzhou, we found higher incidence of preterm birth in the central area than that in the suburban area over time (Data not shown).
Our findings showed that mothers born in more recent era had a higher risk of PTB among primiparous mothers as well as among multiparous mothers who born before 1981. Mothers born in more recent birth cohort experience remarkable lifestyle change, increased environmental pollution, and higher stress due to rapid change of economy and society, which may pose profound impacts on women’s reproductive system [
30‐
32]. For example, prevalence of obesity in both children and adult female increased rapidly over past two decades in China [
28,
33]. Evidence has shown that obesity is associated with inflammatory up-regulation and higher risk of metabolic disorders, such as insulin resistance and lipotoxicity, which may disturb the function of reproductive system during their lifetime and ultimately increase the risk of PTB [
34]. However, a decreased cohort effect was also observed among multiparous mothers who born after 1981 but not in primiparous mothers.
We observed that the effects of maternal age or maternal birth cohorts on PTB differed by parity. Although the mechanisms are still not fully understood, it might be partly due to the difference in maternal health conditions, socio-economic background and psychological pressure and so on at each age group between primiparous and multiparous mothers. Further studies are warranted to determine which factors are relevant to our observed effects of maternal age and birth cohorts on PTB and how they might interact with each other.
Although a large sample size included in the analysis enabled us to estimate robust results, several limitations should be considered in the study. Firstly, data on some explanatory variables, such as cigarette smoking, a known risk factor of PTB, were not available in the electronic surveillance database of birth registration; thus we were unable to elucidate the detailed mechanisms of observed maternal age and birth cohort on PTB in this study. However, it is worth noting that in China, the prevalence of cigarette smoking among pregnant women is lower than 1% [
35,
36]. Secondly, although it is currently standard practice to confirm gestational age using ultrasound examination at first or early second trimester in Guangzhou [
21], the last menstrual period based method was also used in some rural areas, especially during the early delivery period, which might have resulted in misclassification of PTB. Thirdly, although age-period-cohort models can provide an exploratory descriptive tool for examining population health patterns, they cannot establish causal inference of disease incidence. Fourthly, the data in our study was from a single city in China, which might limit the generalization of the results. However, Guangzhou is a mega city with a large population (14.5 million) [
37] which is comparable to or even larger than the nationwide population size of many countries around the world.
From a public health perspective, this study suggests that young and advanced-age mothers have higher risk of PTB, and primiparous mothers at higher risk of PTB were younger than multiparous mothers in older age. Thus, prenatal care and interventions should be targeted at young and older pregnant women, particularly for primiparous mothers. Since the implementation of universal 2-child policy in January 2016, there has been a substantial rise in the number of advanced-age pregnant women and premature infants. Allocation of resources for reversing the rising trend of PTB and preventing and treating PTB-related complications in children are warranted. Besides, as Guangzhou is one of the most developed cities in China, our study will provide evidence for the future change of PTB in other developing regions.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.