Pregnancy outcomes of women with previous caesarean sections: Secondary analysis of World Health Organization Multicountry Survey on Maternal and Newborn Health

Secondary analysis of World Health Organization Multicountry Survey on Maternal and Newborn Health (WHOMCS) was undertaken among 173,124 multiparous women to assess the association between previous caesarean sections (CS) and pregnancy outcomes. Maternal outcomes included maternal near miss (MNM), maternal death (MD), severe maternal outcomes (SMO), abnormal placentation, and uterine rupture. Neonatal outcomes were stillbirth, early neonatal death, perinatal death, neonatal near miss (NNM), neonatal intensive care unit (NICU) admission, and preterm birth. Previous CS was associated with increased risks of uterine rupture (adjusted Odds Ratio (aOR); 7.74; 95% confidence interval (CI) 5.48, 10.92); morbidly adherent placenta (aOR 2.60; 95% CI 1.98, 3.40), MNM (aOR 1.91; 95% CI 1.59, 2.28), SMO (aOR 1.80; 95% CI 1.52, 2.13), placenta previa (aOR 1.76; 95% CI 1.49, 2.07). For neonatal outcomes, previous CS was associated with increased risks of NICU admission (aOR 1.31; 95% CI 1.23, 1.39), neonatal near miss (aOR 1.19; 95% CI 1.12, 1.26), preterm birth (aOR 1.07; 95% CI 1.01, 1.14), and decreased risk of macerated stillbirth (aOR 0.80; 95% CI 0.67, 0.95). Previous CS was associated with serious morbidity in future pregnancies. However, these findings should be cautiously interpreted due to lacking data on indications of previous CS.

During the study period, 45,232 (26.1%) women underwent CS. Of the 33,267 women with previous CS, 26,930 (80.9%) underwent repeated CS (Table 2). Table 3 presents the maternal and neonatal characteristics, medical comorbidities, and obstetric complications of all women and cross-tabulated with the history of previous CS.
Almost all facilities (90%) had blood bank service. An intensive care unit was available in approximately 56% of participating facilities. The majority of countries of participating facilities (51.7%) were reported to have a moderate maternal mortality ratio (MMR). Approximately 17.2% had high and 24.1% very high MMRs of countries of participating facilities 17 . Association between previous Cs and maternal outcomes. Table 4 presents the adverse maternal outcomes in this study. Overall, the common maternal adverse events were severe maternal outcome (SMO) of 6.5 per 1,000 livebirths, maternal near miss (MNM) of 5.5 per 1,000 livebirths, and placenta previa of 5.1 per 1,000 women. There were 161 maternal deaths, accounting for an overall ratio of 0.9 per 1,000 livebirths.
Association between number of previous Cs and pregnancy outcomes. Table 6 presents association between number of previous CS and adverse pregnancy outcomes. The risks of MNM, SMO, placenta praevia, and morbidly adherent placenta in subsequent pregnancy increased markedly with the increasing number of previous CS. There were no marked differences in the magnitude of association between the risk of adverse neonatal outcomes and number of previous CS. Approximately 19.2% of the study population reported a history of previous CS. Rates of previous CS, however, varied widely ranging from only 3.7% in Afghanistan to 37.5% in Paraguay. This reflects a wide variation of previous CS rates across the participating countries. After controlling the potential confounders at both individual and facility levels, it was observed that there were significant association of various adverse pregnancy outcomes among pregnant women with a previous CS. A previous CS is significantly associated with increased risks of uterine rupture, morbidly adherent placenta, MNM, SMO, and placenta previa. There was no significant difference, however, in rate of maternal death between pregnant women with a previous CS and those who had none. For neonatal outcomes, a previous CS was significantly associated with increased risks of NICU admission, NNM, www.nature.com/scientificreports www.nature.com/scientificreports/ preterm birth, and decreased risk of macerated stillbirth. There were no significantly increased risks of, fresh stillbirth, early neonatal death, perinatal death, low Apgar score, and low birthweight between the two comparison groups.

Discussion
In the present analysis, uterine rupture was the outcome with the strongest association with a previous CS. The risk of uterine rupture was nearly 7 times higher for women with previous CS when compared to those without previous CS. Uterine rupture during pregnancy and delivery is amongst the most devastating obstetric complications as it frequently results in life-threatening maternal and foetal compromises. Previous secondary analysis of the WHOMCS dataset reported that uterine rupture was significantly associated with increases in MNM (aOR, 45. 25 7 . The present study observed a significant association between previous CS and the risk of MNM and SMO. MNM and SMO cases are 1.9 times and 1.8 times more frequent in women with previous CS. There was no significant difference, however, between the two comparison groups in risk of maternal death which might be secondary to a remarkably low rate of maternal death in the WHOMCS data. Women being survivors of a life-threatening condition that is a maternal near-miss case, share several clinico-pathologic factors in common with those who died. The WHO near-miss approach recommends incorporating the near-miss cases into enquiries of maternal deaths to better assess the quality of care and its associated factors 18 . www.nature.com/scientificreports www.nature.com/scientificreports/ In line with literature, this study showed association of abnormal placentation among women with previous CS 8,19,20 . Previous CS was associated with increased risks of morbidly adherent placenta (2.6 times higher) and placenta previa (1.8 times higher) compared to those without previous CS. This finding is clinically important. A previous cohort study in UK reported that morbidly adherent placenta constituted approximately 40% of the causes leading to peripartum hysterectomy 21 . In addition, 20% of postpartum haemorrhage cases requiring massive blood transfusion were secondary to placenta 22 praevia.
A previous study reported that infants born to mothers with a previous CS were significantly more likely to require respiratory support and have prolonged hospitalisation for more than 7 days when compared to those infants born to mothers with prior vaginal delivery. There was no difference, however, in neonatal mortality rate between the groups 23 . In the present study, previous CS was associated with increased risk of NICU admission and marginally increased risk of NNM.    www.nature.com/scientificreports www.nature.com/scientificreports/ Interestingly, this analysis observed a statistically significant association between previous CS and preterm birth. The pragmatic nature of a cross-sectional study per se makes this analysis unable to determine the causal relationships between previous CS and this neonatal outcome. Previous study undertaken in a large cohort of pregnant women in the United States also noted a small, but statistically significant association between previous CS and preterm birth. Pregnant women with previous CS were 14% more likely to experience a preterm birth in the second pregnancy (95% CI, 1.12 to 1.16) 24 . These findings highlight that efforts to limit unnecessary CS may lead to lowering the preterm birth rate.

No. of Previous caesarean section
This study analysed the WHOMCS dataset that used pretested, standardized data collection forms collected by well-trained research assistants in institutes with experience from the previous WHO Global Survey to strengthen the quality of data. In term of available knowledge, this study is the most recent and largest multi-country survey data analyzed to determine the impact of previous CS on the risks of key adverse maternal and perinatal outcomes in subsequent pregnancies. The large sample size in this survey yields a notably high precision of summary measures. In addition, this dataset offered outcome measures as per the most recent approach recommended by WHO.
Nonetheless, some limitations of this study are worthy of note. First, although data were extracted by intensively trained research assistants using standardized recorded forms, applying medical records as the primary data source,   www.nature.com/scientificreports www.nature.com/scientificreports/ means that the possibility of missing data and/or errors in these records cannot be discarded. Secondly, the WHOMCS was conducted in 359 health facilities from 29 countries in Africa, Asia, Latin America, and the Middle East. All of these countries except Japan are in low or low-middle income countries (LMICs). In addition, the WHOMCS data were obtained primarily from secondary and tertiary health facilities which were likely to have an over-representation of complicated pregnancies. Extrapolation of these findings to facilities of different backgrounds, i.e. high income   Table 5. Associations between previous caesarean sections and pregnancy outcomes among multiparous pregnant women. Multiparous women without previous caesarean section is the reference. Note: * Adjusted for individual level: maternal age, marital status, years of maternal school attendance, parity, indirect causes (anaemia, malaria/dengue, HIV, heart disease, lung disease, renal disease, hepatic disease, and chronic hypertension), pre-eclampsia/eclampsia, fetal presentation, birthweight; facility level: blood bank, adult intensive care unit, neonatal intensive care unit. ** Excluded birthweight from the model. CI = confidence interval, NICU = neonatal intensive care unit, aOR = adjusted OR.
www.nature.com/scientificreports www.nature.com/scientificreports/ countries or primary health facilities, may be limited. Thirdly, this study attempted an adjustment for potential confounders to demonstrate a possible independent association of previous CS on adverse pregnancy outcomes. However, information on some other variables that might be associated with pregnancy outcomes, including adequacy of antenatal care, indications of previous CS, nutritional status, smoking, obesity and diabetes were not available. This was because this WHOMCS focused on the assessment of maternal death and maternal near miss. Data collection on potential confounding factors was therefore not very comprehensive. The absence of data on indication of previous CS including whether the CS was medically indicated, or whether the indication for the original CS still exists for subsequent pregnancies. This limitation should be cautiously considered in the interpretation of findings.
Based on the heightened risks of MNM and SMO, women with a history of previous CS should be regarded as a group of high-risk pregnancies requiring a specialised care bundle from multidisciplinary trained providers. Similarly, the greater risks of neonatal near miss and NICU admission associated with previous CS women indicates the need of high quality of cares including an availability of a NICU to minimize long-term neonatal consequences. Primary CS should be performed only when medically necessary. Future researches to delineate an effective strategy for curbing rising CS rates are mandatory to avoid unfavourable pregnancy outcomes that are associated with a previous CS.
The present study did suggest some general trends in that a previous CS is associated with various adverse maternal and neonatal outcomes. These findings however should be cautiously interpreted due to lacking data on indication of previous CS.
Methods study design and setting. The design of the WHOMCS was described in detail elsewhere 25 . Briefly, WHOMCS was a facility-based survey conducted to determine pregnancy outcomes among health facilities in Africa, Asia, Latin America, and the Middle East from May 1, 2010 to December 31, 2011. WHOMCS used a stratified, multistage cluster sampling approach to obtain 359 health facilities in 29 countries. Health facilities were only recruited if they dealt with at least 1000 deliveries per year and had the capacity to provide caesarean section.
Data collection took place at two levels including the individual and facility levels. At the individual level, data of management and the pregnancy outcomes for the women were included in the study, and their respective newborn records, were retrieved from the medical records of the participating facilities by trained data collectors. Data were recorded in the pre-established form at hospital discharge, transfer, or death. Data collectors did not contact the women included in the study; however, data clarification was occasionally required from facility staff. At the facility level, data of characteristics of each health facility, including infrastructure, obstetric and intensive care services, availability of blood bank service, and their ability to treat severe complications, were collected through a specific survey among the healthcare professionals responsible for the participating facilities. Data were collected from the time of admission to death, discharge or 7 days postpartum (whichever came first).

Crude analysis
Adjusted analysis *  www.nature.com/scientificreports www.nature.com/scientificreports/ The duration of data collection depended on the number of deliveries per annum of each facility. To reduce variation in cluster size, data were collected for 2 months if they had 6000 deliveries or more per year, and for 3 months if the health facility had fewer than 6000 deliveries per year. If the anticipated sample size for a country was fewer than 3000 women, the data collection period was 4 months in all health facilities.
Online data entries were carried out in each country, either at the health facility or at a central level, depending on the logistics and available infrastructure using the web-based data management system developed by the Centro Rosarino de Estudios Perinatales (CREP), Rosario, Argentina. Monitoring of data validity and consistency were carried out by Data Managers from CREP and Thailand. This study was approved by the WHO Ethics Review Committee and relevant ethics clearance bodies in participating countries. study population. The study population was all consecutive multiparous pregnant women who gave birth to singleton babies, either live births or stillbirths, at the 359 participating facilities. Figure 1 displays the population flow chart of subjects and exclusion criteria in the present analysis. We additionally excluded women with at least three previous CS who gave birth vaginally as this is unusual practice at present.
Variables and definitions. Multiparous pregnant women who had a history of previous CS were considered the exposure group. The comparison group was pregnant women without history of previous CS.
Pregnancy adverse outcomes were categorized into maternal and neonatal outcomes. The maternal outcomes were MNM, maternal death, and SMO which is a combination of MNM and maternal death. The neonatal outcomes included stillbirth, early neonatal death, perinatal death, NNM, NICU admission, an Apgar score at 5 min of less than 7, low birthweight (<2500 g), and preterm birth (<37 weeks of gestation).
Potential confounding factors were abstracted from both facility and individual levels. Potential confounding factors at the facility level included the availability of a blood bank, an adult intensive care unit for severe maternal conditions, and a NICU for severe neonatal conditions. Potential confounding factors at an individual level were maternal demographic and labor characteristics. Maternal demographic information included marital status, maternal education in years of school attendance, parity, and indirect causes (aneamia, malaria/dengue, HIV, heart disease, lung disease, renal disease, hepatic disease, and chronic hypertension). Labour characteristics included fetal presentation, and mode of delivery. However, we did not have information about indications of these previous CS. statistical analysis. Frequencies were used to describe country groups, baseline maternal and neonatal characteristics, and maternal and neonatal outcomes. Details of maternal and neonatal outcomes were also stratified by prior history of CS. The association between previous CS and maternal and neonatal outcomes were analyzed using a two-level logistic regression model by package lme4 of R software 26 . This procedure was carried out to account for clustering effects of health facilities and confounding effect at both health facility and individual levels for estimating risks of previous CS for each outcome. Women who never had CS represented the reference group.
Associated risks of previous CS for individual outcomes were presented by adjusted odds ratios (aORs), with corresponding 95% confidence intervals. Adjusted confounders in the models were chosen based on literature review. Statistical analyses were performed using R software. ethical approval. The HRP Specialist Panel on Epidemiological Research reviewed and approved the study protocol for technical content. This study was approved by the World Health Organization Ethical Review Committee and the relevant ethical clearance mechanisms in all countries (protocol ID A65661; 27 October 2009). This study adhered to the principles of the Declaration of Helsinki. Informed consent was formally waived by the WHO Ethical Review committee. Therefore, written consent from individual women was not required as data collectors extracted data from medical records and did not contact the individual women.

Data Availability
The datasets generated and/or analyzed during the current study are not publicly available due to they belonged to Department of Reproductive Health and Research, The World Health Organization but could be available from WHO on reasonable request.