A growing interest in the long-term effects of cesarean delivery has been fueled by rising rates of cesarean delivery in many countries,1 together with the trends toward “informed choice” for pregnant women and clinicians making decisions in partnership with women.2
Although abnormal placentation and vaginal birth after cesarean delivery have been studied extensively, few population studies have examined other obstetric and perinatal outcomes, adjusted for confounders, in pregnancies subsequent to cesarean delivery. Analysis of these outcomes is confounded by indication for the first cesarean,3 and complicated by the need to analyze elective and emergency cesarean deliveries separately, both for the first and subsequent cesarean deliveries.4
In 2003, Smith et al5 reported increased risks for unexplained stillbirth in the birth subsequent to cesarean delivery compared with vaginal birth. This population cohort study was conducted to determine if there was evidence to support that finding and also to test the hypothesis that South Australian women who have a cesarean first birth also have increased risks for other adverse obstetric and perinatal outcomes in their second birth, compared with women who have a vaginal first birth.
MATERIALS AND METHODS
The study included all second, singleton births in the South Australian perinatal data collection from 1998 to 2003, excluding late terminations of pregnancy. This collection has been mandated by legislation since 1986.6 A validation study has shown overall 96.7% agreement with hospital medical records for all variables examined in the data collection.7 Kappa values ranged from 0.77 to 1.00 for the variables used in this study, indicating excellent agreement beyond chance. Two variables introduced after the validation study, smoking and the date of the end of the last pregnancy, have not been validated. The data are reported to the Pregnancy Outcome Unit of the Department of Health, by midwives, neonatal nurses, and hospital information services staff on the Supplementary Birth Record. The collection includes maternal sociodemographic, pregnancy, birth and neonatal details of all live births and stillbirths of at least 20 weeks of gestation or 400-g birth weight in South Australia, including late terminations of pregnancy.6 There was no unique identifier for women during the study period to enable us to link births longitudinally to the same woman. Each perinatal death is reviewed and the cause of death classified by a multidisciplinary, clinical committee according to Whitfield and other classifications.8
The sample size was adequate for detecting a twofold difference (relative risk [RR]=2) between the cohorts with 80% power at the .05 level of significance for outcomes with incidence of 0.18% or greater in the vaginal birth cohort.
Statistical analysis was carried out using STATA Statistical Software 8 (StataCorp, College Station, TX). Multvariable logistic regression models were based on backward elimination of nonsignificant variables and adjusted for the following predictors: age, indigenous status, smoking, pregnancy interval (there may have been a miscarriage, termination of pregnancy, or ectopic pregnancy between the first and second births), medical complications (hypertension, diabetes, asthma), obstetric complications, hospital category, patient type (public/private), and gestation and history of ectopic pregnancy, miscarriage, stillbirth or termination of pregnancy. There were two pregnancy interval categories for the interval between the end of the last pregnancy and the second birth: less than 15 months and 15 or more months. Hosmer-Lemeshow goodness-of-fit tests were performed for all models. The number needed to harm was calculated for statistically significant outcomes using adjusted odds ratios (OR) or RRs except for rare outcomes where differences in proportions between the cohorts were used.9 It was not possible to adjust for indication for cesarean delivery in the first birth; only the indication for the current cesarean delivery was collected.
Outcomes are defined when first mentioned in the text. Specific definitions are not given for outcomes which are fields on the Supplementary Birth Record (eg, abruption).
The risk for prolonged labor (greater than 18 hours in duration) was analyzed using only those women in each cohort who labored. Women who had an elective cesarean delivery were excluded from analysis of the risk for emergency cesarean (Fig. 1). Adjusted RRs were calculated for the common outcomes of elective and emergency cesarean delivery.
For uterine rupture, placenta accreta, and maternal death, where data were sparse, Log Exact Statistical Package (Cytel Software Corporation, Cambridge, MA) was used to generate ORs and 95% confidence intervals (CIs). Cases of uterine rupture were confirmed by checking hospital morbidity data (Integrated South Australian Activity Collection) and by reviewing hospital case records to identify partial and complete uterine ruptures. Placenta accreta, increta, and percreta have been grouped together using one code for accreta, with or without hemorrhage.
Causes of stillbirth were extracted from the perinatal mortality data using Whitfield cause-specific classifications.8 Time-to-event analyses were conducted for fetuses at risk of stillbirth and unexplained stillbirth, using gestation as the timescale and ongoing pregnancies as the denominator. Hazard ratios were calculated using Cox Proportional Hazards models allowing for covariates. The proportional hazards assumption was tested with Grambsch and Therneau tests. Log rank tests were calculated to detect the gestation at which there was a statistically significant difference of risk between the two cohorts.
During the study period, the cesarean delivery rate in South Australia increased from 23.5% in 1998 to 30% in 2003.6 There were 36,038 second, singleton births: 27,313 (75.8%) in the previous vaginal birth cohort and 8,725 (24.2%) in the previous cesarean delivery cohort. See Figure 1 for subcategory numbers for birth outcomes in the two cohorts: vaginal birth, elective cesarean delivery, emergency cesarean delivery (before or during labor). and trial of labor.
Compared with the vaginal birth cohort, larger proportions of women in the cesarean delivery cohort were aged more than 30 years, private patients or gave birth in private, rather than public or country hospitals, and smaller proportions were smokers or had a pregnancy interval less than 15 months (Table 1).
The significance probability of the Hosmer-Lemeshow goodness-of-fit tests for the unconditional logistic regression models ranged from .2 to more than .9. The range of the number of primary cesarean deliveries needed to harm included 134 to produce one additional preterm infant, up to 1,536 to produce one additional placenta accreta (Tables 2 and 3).
The cesarean cohort had significantly increased risks for placenta previa, antepartum hemorrhage, malpresentation (other than vertex before birth), prolonged labor, cesarean delivery (elective and emergency), uterine rupture, and placenta accreta compared with the vaginal birth cohort (Table 2). The increased risks for placental abruption and maternal death in the cesarean delivery cohort were not statistically significant.
The incidence of uterine rupture for all women in the study was 0.53 per 1,000 confinements. All 19 cases of uterine rupture were from the cesarean delivery cohort. Eighteen were complete ruptures through all layers of the uterine wall, and one was a partial rupture through some layers of the uterine wall. Fifteen of these women had a trial of labor (4 did not labor), giving a RR of 4.92 (95% CI 1.63–14.80) for uterine rupture with trial of labor (absolute risk 0.40%) compared with no labor. Uterine rupture was associated with two hysterectomies and one late maternal death from amniotic fluid embolism.8
Five of the six women in the cesarean delivery cohort who had placenta accreta had hysterectomies, and one of these women died from hemorrhage.8 The risk for peripartum hysterectomy for all women in the two cohorts could not be calculated because a specific code for hysterectomy was not used during the study period.
The cesarean delivery cohort infants had small increased risks for preterm birth (less than 37 weeks of gestation) as well as very preterm birth (less than 32 weeks of gestation), of being small for gestational age (birth weight less than the 10th percentile for gestational age according to the Australian national birth weight percentiles10), of low birth weight (less than 2,500 g), stillbirth and a higher risk of unexplained stillbirth (Table 3). There was no association between previous cesarean delivery and neonatal death.
Time-to-event analysis generated a hazard ratio for stillbirth of 1.55 (95% CI 1.07–2.23, P=.020) from the Cox proportional model adjusted for antepartum hemorrhage, race, and patient type. A Grambsch and Therneau test for variation in the RR showed no significant departure from the proportional hazards assumption (P=.252). The data were consistent with a slight increase in RR from 20 to 44 weeks of gestation, but our study lacked the power to detect a departure from the proportional hazards assumption. The log-rank test detected a significant increase (P=.023) in risk for stillbirth from 39 weeks of gestation (Fig. 2).
The hazard ratio for unexplained stillbirth from the Cox proportional model adjusted for diabetes was 2.69 (95% CI 1.51 – 4.79, P=.001). The Grambsch and Therneau test for variation in the RR was nonsignificant (P=.711). The associated diagnostic plot showed no support for linear variation in the RR with increasing gestation. The log rank test detected a significant increase (P<.001) in risk for unexplained stillbirth only from 40 weeks of gestation (Fig. 3). The cause of death was unexplained in a higher proportion of stillbirths in the cesarean birth cohort (48.8%) compared with the vaginal birth cohort (26.6%). Autopsies were performed for 70.8% of all stillbirths and 76.1% of the unexplained stillbirths.
An interval of 15 months or more between the end of the last pregnancy and the second birth was associated with small decreases in risk for antepartum hemorrhage, preterm birth, and low birth weight.
This study using a validated population database adds to the evidence that primary cesarean delivery is associated with increased risks in the next birth for adverse obstetric and perinatal outcomes. The absolute risks remain small (except for a 47% risk of emergency cesarean delivery with trial of labor) but remain relevant to women at risk of future pregnancy.
Some increased risks may be due to confounding factors related to the indication for the first cesarean delivery. While first-birth information available to us is accurate, we cannot be sure whether outcomes in the second birth are caused by confounding factors in the first birth, due to our inability to link births longitudinally to the same woman. During the study period, the indications for first, singleton cesarean deliveries (up to two can be specified) were failure to progress (49%), fetal distress (31.8%), malpresentation (17%), hypertension (5%), antepartum hemorrhage (2.7%), intrauterine growth restriction (1.3%), and other (10.8%).
The women in the two cohorts were different (Table 1); however, an observational study such as this is unable to assess other important ways in which the women may have been different, such as body mass index, education level, values, beliefs, and attitudes toward birthing.
Several of the increased risks are associated with abnormal placentation, which is consistent with previous studies. 3,11–18
The proportion of women choosing a trial of labor after cesarean delivery varies widely and is multifactorial: feelings and beliefs regarding birth, family considerations, the need for a shorter recovery, parity, clinician bias, knowledge, and ethnicity. 19–21 Women who have experienced a previous vaginal birth are more likely to attempt vaginal birth after cesarean delivery (VBAC).20,21 No women in the cesarean delivery cohort had experienced vaginal birth, which may be a factor in the low proportion of those who opted for a trial of labor (43.3%) and the low success rate (53.1%). In South Australia, there has been a decline in VBAC rates after primary cesarean delivery from 30.4% in 1998 to 19.7% in 2003.6 The data are unable to distinguish between women in the cesarean birth cohort who were scheduled to have an elective cesarean but commenced labor before the scheduled date and had an “emergency” cesarean delivery and women who had a trial of labor which ended in emergency cesarean delivery. Thus, the true proportion of women who had a trial of labor may have been lower than 43.3%, and the VBAC success rate may have been slightly higher.
Women in the cesarean birth cohort who had a trial of labor were much more likely than the vaginal birth cohort to have an emergency cesarean delivery. Although we know the indication for the second cesarean delivery, we do not know the indication or type of incision performed for the first cesarean delivery; therefore, we cannot interpret the increased risk. (During the study period, the proportion of classical incision cesarean delivery was 0.8%, compared with 99.2% lower segment cesarean delivery, at the biggest teaching hospital in the state.)
A cohort study which linked women longitudinally and matched for maternal age and the baby’s sex also found increased risks for the little-studied complications of malpresentation and prolonged labor.22 Multivariate studies in which the indication for the first cesarean delivery is known will be required to confirm these risks. Previous breech presentation has shown a fourfold increase in the risk of recurrence in the next pregnancy, suggesting recurring specific causal factors of genetic or environmental origin.23 Previous cesarean delivery may compound existing factors by a mechanism similar to that for placenta previa: impaired growth of the lower uterine segment due to scarring, resulting in its being less accommodating for the fetal head.
Primary cesarean delivery for failure to progress or prolonged labor has been shown to increase the risk of repeat cesarean delivery.2,22 Failure to progress is the major indication for first cesarean delivery in South Australia and is likely to be a confounder for repeat cesarean delivery. There is also potential for bias from varying definitions of prolonged labor, subjective estimation of the length of labor, and closer monitoring of women with a history of cesarean delivery. Women giving birth in private hospitals were less likely to have prolonged labor than public patients in public hospitals (OR 0.24, 95% CI 0.11–0.50), possibly reflecting differences between public and private patients or in clinical practices.
The incidence and OR for uterine rupture in the cesarean birth cohort were higher than those reported in another larger Australian study,24 which may be due to several factors; for example, our OR 95% CI was wide, and their use of the Inpatient Statistics Collection alone to ascertain uterine rupture (not available in their perinatal data collection) may have under ascertained the true number of uterine ruptures. We found our coded hospital morbidity data detected 16 of the 19 uterine ruptures. Other studies have reported under ascertainment from hospital morbidity data.25 Reported risks for uterine rupture in a cesarean birth cohort have usually compared a trial of labor with elective repeat cesarean delivery.26 In our study, the risk for uterine rupture was increased almost fivefold for women who had a trial of labor. The absolute risk of 0.4% in the trial of labor cohort remains small and is consistent with previous studies.25,26
This study adds evidence to support the association between previous cesarean delivery and stillbirth,5,27 preterm birth, and small for gestational age infants.5 Smith et al5 found an increased risk for antepartum stillbirth from 34 weeks of gestation after previous cesarean birth, with unexplained stillbirths accounting for most of the increase in risk. Our study was consistent with an increase in risk for stillbirth and unexplained stillbirth, but with a smaller data set may have been underpowered to detect an effect at gestations earlier than 39 and 40 weeks, respectively.
Preterm birth, small for gestational age, and low birth weight can recur, making interpretation of the risk of cesarean delivery for these outcomes also difficult.
These study findings will assist women and clinicians in making informed choices, balancing the risks and benefits of cesarean delivery in the first and future births. The decision for an elective primary cesarean delivery for no reported medical indication should be carefully considered for its impact in future births.
1. Walker R, Turnbull D, Wilkinson C. Strategies to address global cesarean section rates: a review of the evidence. Birth 2002;29:28–39.
2. Raheem M, Salloum M. Clinical audit: obstetric performance after caesarean section. J Obstet Gynaecol 2003;23:503–6.
3. Hemminki E, Merilainen J. Long-term effects of caesarean sections: ectopic pregnancies and placental problems. Am J Obstet Gynecol 1996;174:1569–74.
4. Murta EFC, Nomelini RS. Is repeated caesarean section a consequence of elective caesarean section? Lancet 2004;364:649–50.
5. Smith GCS, Pell JP, Dobbie R. Caesarean section and risk of unexplained stillbirth in subsequent pregnancy. Lancet 2003;362:1779–84.
6. Chan A, Scott J, Nguyen A, Sage L. Pregnancy Outcome in South Australia, 1998-2003. Adelaide (Australia): South Australian Department of Health; 1999-2005. Available at: http://www.dh.sa.gov.au/pehs/pregnancyoutcome.htm
. Retrieved November 1, 2006.
7. McLean A, Scott J, Keane RJ, Sage L, Chan A. Validation of the 1994 South Australian perinatal data collection form. Adelaide (Australia): Pregnancy Outcome Unit, South Australian Department of Health; 2001.
8. Maternal, Perinatal and Infant Mortality Committee. Maternal, Perinatal and Infant Mortality in South Australia 1998–2003. Adelaide (Australia): South Australian Department of Health; 1999–2005. Available at: http://www.dh.sa.gov.au/pehs/pregnancyoutcome.htm
. Retrieved November 1, 2006.
9. Centre for Evidence Based Medicine. Oxford (United Kingdom): Institute of Health Sciences; 2006. Available at: http//www.cebm.net/nnts.asp
. Retrieved November 1, 2006.
10. Roberts CL, Lancaster PAL. Australian national birthweight percentiles by gestational age. Med J Aust 1999;170:114–8.
11. McMahon MJ, Li R, Schenck AP, Olshan AF, Royce RA. Previous cesarean birth. a risk factor for placenta previa? J Reprod Med 1997; 42:409–12.
12. Rasmussen S, Irgens LM, Dalaker K. A history of placental dysfunction and risk of placental abruption. Paediatr Perinat Epidemiol 1999;13:9–21.
13. Hendricks MS, Chow YH, Bhagavath B, Singh K. Previous cesarean section and abortion as risk factors for developing placenta praevia. J Obstet Gynaecol Res 1999;25:137–42.
14. Taylor VM, Kramer MD, Vaughan TL, Peacock S. Placenta previa and prior cesarean delivery: how strong is the association? Obstet Gynecol 1994;84:55–7.
15. Lydon-Rochelle M, Holt VL, Easterling TR, Martin DP. First-birth caesarean and placental abruption or previa at second birth. Obstet Gynecol 2001;97:765–9.
16. Ananth CV, Smulian JC, Vintzileos AM. The association of placenta previa with history of caesarean delivery and abortion: a metaanalysis. Am J Obstet Gynecol 1997;177:1071–8.
17. Clark SL, Koonings PP, Phelan JP. Placenta previa/accreta and prior cesarean section. Obstet Gynecol 1985; 66:89–92.
18. Nielsen TF, Hagberg H, Ljungblad U. Placenta previa and antepartum hemorrhage after previous cesarean section. Gyncol Obstet Invest 1989;27:88–90.
19. Davies GAL, Hahn PM, McGrath MMJ. Vaginal birth after cesarean. physicians’ perceptions and practice. J Reprod Med 1996;41:515–20.
20. Eden KB, Hashima JN, Osterweil P, Nygren P, Guise JM. Childbirth preferences after cesarean birth: a review of the evidence. Birth 2004;31:49–60.
21. Dodd JM, Crowther CA, Huertas E, Guise JM, Horey D. Planned elective repeat caesarean section versus planned vaginal birth for women with a previous caesarean birth (Cochrane Review). In: The Cochrane Library, Issue 4, 2004. Oxford: Update Software.
22. Hemminki E. Pregnancy and birth after cesarean section: a survey based on the Swedish birth register. Birth 1987;14:12–7.
23. Albrechtsen S, Rasmussen S, Dalaker K, Irgens LM. Reproductive career after breech presentation: subsequent pregnancy rates, interpregnancy interval, and recurrence. Obstet Gynecol 1998;92:345–50.
24. Taylor LK, Simpson JM, Roberts CL, Olive EC, Henderson-Smart DJ. Risk of complications in a second pregnancy following caesarean section in the first pregnancy: a population-based study. Med J Aust 2005;183:515–9.
25. Guise JM, McDonagh MS, Osterweil P, Nygren P, Chan BKS, Helfand M. Systematic review of the incidence and consequences of uterine rupture in women with previous caesarean section. BMJ 2004;329:1–7.
26. Appleton B, Targett C, Rasmussen M, Readman E, Sale F, Permezel M, VBAC Study Group. Vaginal birth after caesarean section: an Australian multicentre study. Aust NZ J Obstet Gynaecol 2000;40:87–91.
27. Smith GCS, Pell JP, Cameron AD, Dobbie R. Risk of perinatal death associated with labor after previous caesarean delivery in uncomplicated term pregnancies. JAMA 2002;287:2684–90.