Methods:

Study design and data sources

In this nationwide population-based cohort study, we used data from the Swedish Medical Birth Register (MBR), covering about 99% of all births in Sweden, and we included women with singleton births between January 1, 1982, and December 31, 2012.
The obstetric history of each woman, including data on comorbidity (such as diabetes, hypertension, and cardiovascular diseases), was obtained from the MBR and the Swedish National Patient Register (NPR). Data from these registers were linked using a unique Swedish personal identification number20. All diagnoses and complications during pregnancy or delivery are classified according to the Swedish version of the International Classification of Diseases (ICD), using the ICD-8 until 1986, the ICD-9 (1987 to 1996), and the ICD-10 since 1997. This study is reported as per the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guideline (S1 Table).

Study population

The study cohort consisted of women who had their first two births between 1982 and 2012. Women who had singleton birth with available data on the mode of delivery on the 1st delivery were included. We excluded records where the first pregnancy was a multiple gestation or resulted in stillbirth. Ethical approval was obtained from Regional Ethical Review Board in Stockholm.

Mode of delivery

Mode of birth was categorised into VB and CD. We further subclassified this as follows: (1) spontaneous VB [reference group], (2) assisted delivery (vacuum/forceps), (3) pre-labour CD (before the onset of labour), (4) in-labour CD (after the onset of labour), and (5) unspecified CD, where data on the type of CD were not available. The variable was defined this way to assess the effect of specific types of mode of birth in the 1st birth on the risk of stillbirth in the second pregnancy.
Then we considered mode of birth in both first and second births in order to assess the impact of VBAC on the risk of stillbirth, in which mode of delivery was grouped as follows: (1) repeat CD (CD in both pregnancies, [reference group]), (2) VBAC, (3) first VB and subsequent CD, and (4) VB in both pregnancies. It should be noted that mode of delivery recorded in the MBR is the ’final’ mode of delivery, which may differ from the woman’s intended mode of delivery. For example, a woman may attempt a trial of labour after a CD and fail, ending up with an in-labour CD. Thus, our study groups represent a good reflection of what happens in ’real-world’ situations.

Outcome

The outcome of interest was stillbirth (antepartum and intrapartum fetal death). We used data from the Swedish MBR to identify stillbirths in the second pregnancy. This was defined as fetal death after 28 completed weeks (until June 2008) and fetal death after 22 completed weeks since July 2008.
We classified stillbirth into explained and unexplained stillbirth using an adapted version of the ReCoDe (RElevant COndition at DEath) classification system. The ReCoDe classification system is designed in a hierarchical manner to organise relevant clinical conditions associated with death in utero. It contains nine main categories, starting fromA (conditions affecting the fetus) to I(unclassified), and each category is divided into several subgroups22. These categories include a wide range of clinical conditions related to the fetus, the placenta, the mother, and intrapartum conditions. On the other hand, unexplained cases are divided into two subcategories; cases with irrelevant conditions despite information or cases lacking available information22.
We used the diagnosis variables from the MBR and NPR to classify stillbirth according to the underlying conditions of the ReCoDe classification system. All diseases and complications during pregnancy or delivery were classified according to the Swedish version of ICD codes. In addition, small for gestational age (SGA) was defined as a birthweight below 2 standard deviations (SDs) of the population mean birthweight adjusted for sex-specific and gestational age distributions or according to ICD codes (codes are shown in Table S2).

Statistical analysis

Maternal and birth characteristics are presented according to stillbirth and mode of delivery (Tables S3-S5) using frequency and percentages. Logistic regression models were performed to evaluate all associations using crude ORs and aORs, along with 95% CIs. First, we estimated the odds of stillbirth in the second birth following a CD in the first delivery, compared with the outcome of second deliveries following a prior VB. Second, we expanded the exposure variable to evaluate the impact of specific types of mode of delivery, specifically pre-labour CD and in-labour CD, on the associations. Finally, in a third model, we considered the mode of delivery in the first and second births to estimate the odds of stillbirth in women with a VBAC compared to women with a repeat CD. For completeness, we report the results for all the mode of delivery combinations in the first and second births.
Adjusted models included maternal age, BMI, smoking, education, country of origin, year of delivery, comorbidities (diabetes, chronic hypertension, cardiovascular disease), and pregnancy-related disorders (gestational diabetes, gestational hypertension, and preeclampsia) in the first pregnancy. We added a missing data category to control for missing data on BMI and smoking.
We undertook subgroup analysis based on ReCoDe classification categories, including a specific cause of stillbirth for explained stillbirths (cases with a known condition for death). We conducted separate analyses for causes of death restricted to stillbirths caused by: (a) lethal congenital anomaly, (b) SGA\fetal growth restriction, (c) any cord issue, (d) placental abruption, (e) any placental abnormalities, (f) any maternal conditions, (g) uterine rupture, (h) intrapartum asphyxia/birth trauma. We additionally evaluated the association between a previous CD and explained stillbirth (including any relevant condition), unexplained stillbirth (including cases without relevant condition), and finally relevant versus no relevant condition (Table S7).
We did sensitivity analyses according to birth defects, preeclampsia, gestational diabetes, preterm birth, SGA, and time period. We also calculated the population attributable fraction (AF) (details are shown in Appendix page 2).
Statistical analyses were performed using Stata version 16.1 (StataCorp, Texas, USA), and all tests were two-sided with a 5% significance level.