Design and setting
We obtained data from China’s NMNMSS between 2012 and 2018. The system covers 438 hospitals in 326 districts or counties throughout 30 provinces, each of which manages more than 1,000 deliveries annually[11]. Data collected included sociodemographic characteristics, obstetric history, pregnancy complications, and pregnancy outcomes of all pregnant or postpartum women in a hospital. Doctors in each hospital were trained to collect data prospectively from admission to discharge. Quality assurance was ensured by staff from county-level, municipal-level and provincial-level maternal and child health hospitals 1–2 times a year. At the same time, the National Office for Maternal and Child Health Surveillance verified the quality of the records by selecting 6–8 hospitals randomly in each province once a year[23].
Study population
We restricted the analysis to pregnant women with singleton births who delivered at or after 28 complete weeks of gestation. Women with multiple pregnancies were not included because they are prone to pregnancy complications[24] and UR[25]. Women lacking information on delivery method, history of caesarean section, or gravidity were excluded. We also excluded women with an unlikely combination of gravidity and parity. This left a total of 9,454,239 women for the study.
Variable definition
UR was defined as uterine or lower uterine dehiscence during late pregnancy or delivery[26]. According to the degree of dehiscence, UR can be divided into complete UR (tearing in all layers of the uterine wall) and incomplete UR (tearing in the muscular layers)[10]. Common clinical manifestations of UR include foetal distress, sudden tearing uterine pain, cessation of uterine contractions and abnormal vaginal bleeding[27]. UR was diagnosed by a health professional with imaging techniques (magnetic resonance imaging or ultrasound examination)[28]; or during emergency caesarean delivery; or peripartum hysterectomy or laparotomy after vaginal birth[3]. Unfortunately, UR is captured as a dichotomous variable (yes/no) in the NMNMSS, and the type of rupture is lacking. We identified five pregnancy complications related to previous caesarean delivery for analysis: preeclampsia, gestational diabetes mellitus, placental abruption, placenta previa and placenta percreta[13, 14]. Preeclampsia included pregnancies with preeclampsia, eclampsia or HELLP (haemolysis, elevated liver enzymes and low platelets) syndrome, as well as chronic hypertension with superimposed preeclampsia. Gestational diabetes was diagnosed by an oral glucose tolerance test (OGTT) during pregnancy. Placental abruption was defined as the premature separation of the implanted placenta before delivery. Placenta previa was defined as the placenta covering the internal os of the cervix. Placenta percreta, as the most severe grade of the placenta accreta spectrum disorders, occurs when the chorionic villi penetrate the uterine serosa[29].
We selected variables that may be related to the occurrence of UR, including region, hospital level, education level (none, primary school, middle school, high school, college or higher), maternal age at delivery (<20, 20–24, 25–29, 30–34, 35–39 and ≥40 years), the number of antenatal visits (none, 1–3, 4–6, 7–9, ≥10), gravidity (1, 2–3, ≥4), parity (0, 1, ≥2), number of previous caesarean deliveries (0, 1, ≥2), foetal presentation (cephalic and other abnormal lies), gestational age, birthweight, and mode of delivery (vaginal delivery and caesarean section). We divided China’s regions into three categories (eastern, central and western) and classified hospitals into three levels (the first level represents the smallest hospital) according to standard definitions[23]. Gestational age was defined based on ultrasound measurement results or estimated from the date of the last menstrual period and classified as early preterm (28–33 weeks), late preterm (34–36 weeks), or term (≥ 37 weeks). Large for gestational age (LGA) was defined as a gestational age-adjusted birth weight above the 90th percentile[30]. Other factors thought to be associated with UR included gestational hypertension, chronic hypertension, heart disease, hepatic disease, severe anaemia (haemoglobin concentration lower than 70 g/L), infection (excluding abortion-related infection, puerperal infection and abdominal incision infection), thrombophlebitis, renal disease, lung disease, and connective tissue disorders.
Statistical Analysis
Primary analysis
We expressed the UR rate as the number of pregnant women with UR per 1,000 pregnant women. Since some women giving birth in township hospitals were not included in the NMNMSS, we weighted the UR rate for the sampling distribution of the population according to the 2010 census of China, as detailed elsewhere[23]. Moreover, a history of caesarean section is a major risk factor for UR in subsequent pregnancies[1, 3], and the risk increases with the number of previous caesarean deliveries[4, 5]. Thus, we calculated the previous caesarean deliveries adjusted rate of UR in women by using the margins command in Stata[31].
We classified pregnancy complications into mutually exclusive categories: preeclampsia, gestational diabetes mellitus, placental abruption, placenta previa, and placenta percreta. We used Poisson regression with a robust variance estimator to assess the association of UR with pregnancy complications, reporting the results from two models. The reference for each model was women without any of the five pregnancy complications. Model 1 describes the adjusted relative risk (aRR) and 95% confidence Interval (CI), taking into account the sampling distribution of the population and birth clustering within hospitals, medical institutions, and pregnant women's sociodemographic and clinical factors that might contribute to the observed associations. Model 2 adjusted for the covariates in Model 1 as well as the number of previous caesarean deliveries (0, 1, ≥2) and LGA (yes/no). We did not adjust for gestational age or final mode of delivery because they included consequences of UR (i.e., laparotomy due to UR). To identify the most robust and stable model, we investigated both multicollinearity and model goodness-of-fit.
Due to a history of caesarean section related to both pregnancy complications[1, 3] and UR[12–14], pregnancy complications may be only an intermediate factor in the causal chain between a history of caesarean section and the risk of UR. We repeated the association analysis of pregnancy complications with UR only in women without previous caesarean delivery.
Secondary analysis
Restricting to a group of women
To investigate the association between pregnancy complications and UR without potential maternal confounding factors (advanced maternal age[1] and multiple gravidities[32]), we performed sensitivity analyses excluding women with advanced maternal age (≥35 years) and/or multiple gravidities (≥4). Given the possible impacts of abnormal foetal presentation and macrosomia on the occurrence of UR[33], we restricted the association analysis to women with offspring having a cephalic lie and a birth weight of less than or equal to 4000 g.
Co-occurrence of pregnancy complications and UR risk
Pregnancy complications may co-occur in a given pregnancy. We therefore repeated model 1 and 2 testing for the associations between having at least two or more pregnancy complications and the risk of UR. Because the numbers were too small to assess unique combinations of pregnancy complications, we modelled the variables “no pregnancy complications”, “any one pregnancy complication”, and “any two or more pregnancy complications” in a single model.
Risk of UR in different gestational age groups
To explore the risk of UR with pregnancy complications in different gestational age groups, we compared the UR rates in women for each pregnancy complication and gestational age against those in women without pregnancy complications at 28–33 weeks of gestational age, using model 1 and 2.
Role of large for gestational age
Because LGA is associated with gestational diabetes mellitus[34], we repeated model 1 and 2 testing to analyse a possible effect of LGA foetuses on the risk of UR among women with gestational diabetes mellitus.
Trends over time in UR rates
To examine trends over time in UR rates among women with pregnancy complications, we repeated model 1 and 2 by including the year of study period as a continuous variable.
Statistical analysis was performed using Stata (version 16.0, Stata Corp LP., College Station, United States of America). P<0.05 (2-sided) was considered statistically significant.
Patient involvement
Informed consent from the patients was waived by the Ethics Committee, as the data used in our study were obtained from a national routine surveillance system established by the government. Data use was authorized by the National Health Commission, and data provided to us were deidentified.