An Epidemiologic First Principles Approach to Assessing Changes in the Frequency of Early Term Birth

In this issue of Paediatric and Perinatal Epidemiology, Berman et al. [1] address the methodological challenge involved in evaluating temporal changes in population rates of early term birth (37–38 weeks' gestation). Early term births have received increasing attention since the relatively recent recognition that these newborn infants experience respiratory and other complications at a higher rate than newborn infants at full‐term gestation (39–41 weeks) [2, 3]. Berman et al. [1] describe the dilemma of quantifying the effect of interventions to reduce rates of preterm and early term birth, as the proportion of early term birth can be influenced by changes in the frequency of births at preterm gestation. After an empirical assessment of options, Berman et al. [1] highlight the importance of ‘… considering the broader context in analyses’ and recommend that ‘Researchers should examine all births < 39 + 0 weeks rather than just early term births, or use nominal logistic regression’.

We agree with the insights and discussion points that Berman et al. offer: evaluating changes in early term birth rates requires a comprehensive perspective. As for the solutions Berman et al. [1] recommend, estimating changes in births at < 39 weeks' gestation addresses the problem only in part because quantifying changes in this proportion effectively ‘averages’ changes in preterm and early term birth rates (which could potentially be quantitatively or qualitatively different). Multinomial logistic regression yields the odds of preterm birth and early term birth relative to the odds of birth ≥ 39 weeks' gestation (or the corresponding probabilities), thereby providing more context than an isolated focus on births at early term gestation. Finally, the segmented quantile approach quantifies temporal changes in the gestational age distribution, though it does not exactly reflect changes in the frequency of preterm and early term birth.

In this commentary, we briefly review the historical development of concepts related to preterm birth and gestational duration, and discuss an epidemiologic first principles approach to assessing changes in the frequency of early term birth. Data from the United States fetal death and period birth cohort/infant death files were used to recreate birth cohorts for the years 2016 and 2023 in order to illustrate the proposed methods. The cohorts were restricted to singleton live births and stillbirths at 22–43 weeks' gestation.

1. Concepts Related to Preterm Birth and Gestational Duration

In the mid‐twentieth century, perinatal health surveillance focused on indices such as preterm birth and low birthweight rates. These indices were identified as robust predictors of the key indicators of perinatal population health, namely, perinatal mortality and serious neonatal morbidity rates. Subsequent recognition of the need to distinguish between short gestational duration and restricted fetal growth led to a focus on preterm birth and small‐for‐gestational age (SGA) rates (in addition to morbidity and mortality rates) as measures of population perinatal health [4]. In fact, the preterm birth rate came to be regarded as one of the most important predictors of perinatal outcomes in the population.

Despite being a strong predictor of population perinatal health, the preterm birth rate is slightly flawed as an epidemiologic measure of frequency: it is not a bona fide incidence measure since it is restricted to live births. This design feature of the preterm birth rate is a relic of history, and a witness to the past when stillbirths were accorded a lesser social and legal status than live births [5]. Although the exclusion of stillbirths from the preterm birth rate calculation appears inconsequential (since stillbirths are relatively infrequent), it has had the insidious effect of discouraging the use of fundamental epidemiologic measures for quantifying perinatal phenomena. Thus, perinatal epidemiologists in the latter half of the twentieth century rarely invoked incidence measures for birth and related pregnancy and perinatal phenomena.

The first break from this tradition occurred a few decades ago, when Yudkin et al. [6] proposed using ‘fetuses at risk’ as the appropriate denominator for calculating gestational‐age‐specific stillbirth rates. Under this formulation, gestational duration is viewed as survival time from a fetal perspective [7]. This development, and the subsequent extension of this concept to gestational age‐specific perinatal mortality rates [7], effectively introduced incidence measures for quantifying fetal, pregnancy, and other perinatal outcomes. The incidence measures of relevance to the discussion on early term birth include the incidence of birth, medically‐indicated birth and perinatal mortality.

2. Traditional Preterm Birth and Early Term Birth Rates

The preterm birth rate among singletons in the United States increased from 7.9% in 2016 to 8.6% in 2023, whereas the rate of early term birth increased from 25.1% to 29.7% over the same period (Table S1). These calculations are similar to those carried out by Berman et al. [1], with both estimates restricted to live births.

3. Foetuses at Risk Incidence of Birth and Medically‐Indicated Birth

The cumulative incidence (henceforth referred to as incidence) of birth at < 37 weeks' gestation (including live births and stillbirths) increased from 8.3% in 2016 to 8.9% in 2023, whereas the incidence of early term birth increased from 27.3% to 32.5% from 2016 to 2023. The incidence of medically‐indicated birth (i.e., the incidence of labour induction and/or caesarean delivery) at < 37 weeks' gestation increased from 4.4% in 2016 to 5.5% in 2023, whereas the incidence of medically‐indicated early term birth increased from 12.3% in 2016 to 18.2% in 2023.

Figure 1A shows the weekly (cumulative) incidence of birth from 22 to 39 weeks' gestation. Birth rates were essentially unchanged or showed small temporal changes from 22 to 33 weeks' gestation, but showed larger increases at 34 and later gestational weeks. This pattern of temporal change was similar to the pattern of change evident in the weekly incidence of medically‐indicated birth (Figure 1B).

FIGURE 1.

Observed weekly cumulative incidence rates of birth (A), medically‐indicated birth (B) and perinatal mortality (C) among all singletons from 22 to 39 weeks' gestation, United States, 2016 and 2023. Panels A and B also show the change in the spline‐modelled birth rate per week and the spline‐modelled medically‐indicated birth rate per week (secondary Y‐axis). The rate of change shown at 23 weeks refers to the change between Weeks 22 and 23, the change shown at 37 weeks refers to the change between Weeks 36 and 37, etc.

4. Changes in the Incidence Rate of Birth and Medically‐Indicated Birth by Gestational Week

Figure 1A,B also show changes in the weekly incidence rates of birth and medically indicated birth (obtained by fitting a spline to the birth rate and calculating its first derivative). The weekly change in birth rates showed a temporal increase at 37 weeks' gestation (indicating a temporal increase in the rate of birth between 36 and 37 weeks' gestation), whereas the weekly change in the rate of medically‐indicated birth showed very small temporal increases at 34 and 36 weeks (indicating minor temporal increases in the rate of such births between 33–34 and 35–36 weeks), a moderate temporal decrease at 38 weeks (i.e., a decrease in the rate of such births between 37 and 38 weeks), and fairly large temporal increases in medically‐indicated births at 37 and 39 weeks' gestation (i.e., substantial temporal increases in the rate of such births between 36–37 and 38–39 weeks). Figure S1A,B show the weekly incidence of birth and medically‐indicated birth from 22 to ≥ 42 weeks' gestation.

5. Changes in the Incidence of Perinatal Mortality

Figure 1C shows the weekly incidence of perinatal mortality in 2016 and 2023. Stable rates were seen at 28 and 32 weeks, small temporal decreases were evident at 29, 33, 34, 35, 36, 38 and 39 weeks, and small increases were observed at 30, 31, and 37 weeks' gestation. Figure S1C shows the weekly incidence of perinatal mortality from 22 weeks to ≥ 42 weeks' gestation.

6. Changes in the Incidence of Birth and Perinatal Mortality Among Women With Diabetes

The incidence of birth at < 37 weeks' gestation among singleton births to women with pre‐pregnancy diabetes increased from 26.3% in 2016 to 27.4% in 2023, whereas the incidence of early term birth increased from 42.5% to 49.2%. Medically‐indicated birth at < 37 weeks' gestation increased from 20.9% in 2016 to 23.1% in 2023, whereas the incidence of medically‐indicated early term birth increased from 46.5% to 58.2% (Table S2). Perinatal mortality rates among singleton births to women with pre‐pregnancy diabetes decreased from 26.6 in 2016 to 19.6 per 1000 total births in 2023.

Figure 2 shows the weekly incidence of birth, medically‐indicated birth and perinatal mortality from 22 to 39 weeks' gestation among singleton births to women with pre‐pregnancy diabetes in 2016 and 2023. Modest temporal declines in birth and medically‐indicated birth rates were observed at 28 and 29 weeks' gestation, whereas moderately large increases were evident at 34, 36, 37 and 38 weeks' gestation. Perinatal mortality rates decreased at all gestational ages from 28 to 39 weeks. Figure S2 shows these weekly incidence rates from 22 to ≥ 42 weeks' gestation.

FIGURE 2.

Observed weekly cumulative incidence rates of birth (A), medically‐indicated birth (B) and perinatal mortality (C) among singletons born to women with pre‐pregnancy diabetes from 22 to 39 weeks' gestation, United States, 2016 and 2023. Panels A and B also show the change in the spline‐modelled birth rate per week and the spline‐modelled medically‐indicated birth rate per week (secondary Y‐axis). The rate of change shown at 23 weeks refers to the change between Weeks 22 and 23, the change shown at 37 weeks refers to the change between Weeks 36 and 37, etc.

7. An Epidemiologic First Principles Perspective

The use of incidence measures by gestational week provides a comprehensive and detailed account of temporal changes in the population frequency of birth at preterm and early term gestation. The role of obstetrical intervention in influencing birth and perinatal mortality rates is highlighted, especially among women with pre‐pregnancy diabetes. These results also support the proposition that preterm birth rates have evolved from being key predictors of population perinatal health, and now also reflect the magnitude of fetal surveillance and obstetrical intervention at preterm and early term gestation, at least in high‐income countries [8]. A more complete analysis of the causal factors underlying the observed temporal changes in birth rates at preterm and early term gestation could be carried out using the multivariable extension of the survival analysis framework [9, 10].

We thank Berman et al. [1] for highlighting a methodological issue central to assessment of population perinatal health. Although we agree with their propositions and insights, which were based on traditional perinatal metrics, we believe the analysis of temporal changes in early term birth can benefit from the adoption of an epidemiologic first principles approach that focuses on incidence rates.

Author Contributions

O.S.J. and K.S.J. discussed the conceptual issues involved, and carried out the data analysis. All authors discussed the results and O.S.J. drafted the commentary. All authors critically reviewed the commentary, revised its intellectual content, and approved the final version of the manuscript.

Funding

The authors have nothing to report.

Conflicts of Interest

The authors declare no conflicts of interest.

Linked Articles

This article is linked to Berman et al. paper. To view this article, visit https://doi.org/10.1111/ppe.70153.

Supporting information

Table S1: Numbers and rates of live births, total births, preterm births, early term births, and perinatal mortality among singletons at 22–43 weeks' gestation, United States, 2016 and 2023.

Table S2: Numbers and rates of live births, total births, preterm birth, early term birth, and perinatal mortality among singletons of women with pre‐pregnancy diabetes at 22–43 weeks' gestation, United States, 2016 and 2023.

Figure S1: Observed weekly cumulative incidence rates of birth (A), medically‐indicated birth (B) and perinatal mortality (C) among all singletons from 22 to 43 weeks' gestation, United States, 2016 and 2023. Panels A and B also show the change in the spline‐modelled birth rate per week and the spline‐modelled medically‐indicated birth rate per week (secondary Y‐axis). The rate of change shown at 23 weeks refers to the change between Weeks 22 and 23, the change at 37 weeks refers to the change between Weeks 36 and 37, etc.

Figure S2: Observed weekly cumulative incidence rates of birth (A), medically‐indicated birth (B) and perinatal mortality (C) among singletons born to women with pre‐pregnancy diabetes from 22 to 43 weeks' gestation, United States, 2016 and 2023. Panels A and B also show the change in the spline‐modelled birth rate per week and the spline‐modelled medically‐indicated birth rate per week (secondary Y‐axis). The rate of change shown at 23 weeks refers to the change between Weeks 22 and 23, the change at 37 weeks refers to the change between Weeks 36 and 37, etc.

Data Availability Statement

The data that support the findings of this study are available in Vital Statistics Data Online Portal at https://www.cdc.gov/nchs/data_access/vitalstatsonline.htm. These data were derived from the following resources available in the public domain: Fetal Death Data Files, https://www.cdc.gov/nchs/data_access/vitalstatsonline.htm. Period/Cohort Linked Birth‐Infant Death Data Files, https://www.cdc.gov/nchs/data_access/vitalstatsonline.htm.