The institutional ethics committee granted ethics approval for this study. All data were de-identified for analysis and presentation. The study was a matched case–control study of women with placenta accreta and placenta praevia over the 15-year period from January 1993 to May 2008. The setting was a tertiary referral centre in Perth, WA. The hospital is the only referral centre for high-risk pregnancies in the state, delivering over 6000 births per year. The data analysis was conducted using a variably matched set, defined by specific criteria detailed below. The analysis examined the effects of the mode of the primary caesarean section on the subsequent development of placenta accreta, or its variants, in a pregnancy complicated by placenta praevia. A prospective design was considered; however, given the relative rarity of the disease being studied, and the time it would require to gather sufficient women for a meaningful analysis, a retrospective design was deemed more appropriate by the authors.
The primary outcome was the influence of the indication for a woman's primary caesarean section (elective without labour or emergency after some labour) on her subsequent risk of developing a placenta accreta. The secondary outcomes were the risk of a placenta accreta given the following variables: parity; gravidity; uterine surgery prior to the primary caesarean section; and vaginal delivery prior to and following the primary caesarean section. The morbidity associated with the disease at this institution over time was also examined.
Placenta accreta was defined primarily as a histopathology report of an adherent placenta. In a few cases preoperative ultrasound assessment and a clinical diagnosis, based on an operative report of a difficult manual and/or piecemeal removal of the placenta during surgery, with heavy bleeding from the placental site requiring hysterectomy or transfer to another institution for further surgical or radiological intervention, and requiring subsequent intensive care admission, was used. Placenta praevia was defined as a placenta situated over the internal cervical os or in close proximity. Primary caesarean section was defined as the first caesarean section delivery, regardless of whether the woman had a vaginal delivery before or after the first caesarean section.
All cases meeting the criteria of placenta accreta with or without praevia over the study period were analysed to determine the changing trend of the disease over time and its associated morbidity.
In the subsequent matched analysis designed to answer the primary question, a subset of cases were included if they had a placenta praevia, a co-existing placenta accreta, and a history of one or more caesarean sections. Nulliparous women and those with an accreta in the absence of a prior caesarean section delivery were excluded.
All cases were identified from three separate but overlapping sources: the hospital electronic database of all births (STORK), the state hospital morbidity data system (HMDS), and the handwritten birth register. The paper medical record files of all cases and controls were reviewed by an obstetric resident and a maternal fetal medicine specialist to confirm the diagnosis, and to abstract demographic, obstetric, peripartum, and neonatal data.
The collection of HMDS data commenced in 1970. It includes all in-patient episodes for patients admitted to public, private, and freestanding day hospitals in WA. At present, the database contains nearly 60 data items, including diagnosis and procedure codes, and currently receives information for almost 700 000 in-patient episodes per year. KEMH is a public hospital and is required to forward at least 80% of their morbidity information for inclusion in the HMDS within 2 weeks of the discharge of the patient from hospital. The remaining 20% is required within 8 weeks. Once received, they are subject to editing, validation, and corrections, entered within ten working days. Morbidity information for all admissions from 1 July 2006 used the International Statistical Classification of Diseases and Health Related Problems, Tenth Revision, Australian Modification, (ICD-10-AM). The ninth revision was used for the study period preceding this date.
Stork is a Microsoft sql application that captures perinatal data about women giving birth at King Edward Memorial Hospital for Women in WA. It is accessible on the hospital's local area network to over 40 ward and office locations. It stores data for the 50 000 births that have occurred at the hospital since December 1997. Registered midwives enter the data within hours of key events occurring in a woman's pregnancy, birth, and postnatal care. These data include the antenatal booking visit, admission for birth, the following birth, and the time of discharge from the hospital. The Western Australian Midwives Notification System receives this data as mandated by the Western Australian Health Act 1911. Stork data are used to report on unit activity and clinical outcomes, as well as quality audit and research.
In addition, the handwritten birth register for the period was reviewed for women with invasive placentation, peripartum hysterectomy, or massive peripartum haemorrhage to ensure the near complete ascertainment of cases.
Potential controls were identified using the same sources. The final selection of appropriate controls was performed by manual review of patient paper records to determine eligibility and compliance with matching criteria. Eligibility criteria for controls included: placenta praevia; delivery during the study period at the study hospital; and a prior history of at least one caesarean section. All controls were matched with cases by the number of previous caesarean sections, placenta praevia, and maternal age to within 2 years. This reduced to a minimum the confounding effects of age and previous caesarean sections on the risk exerted by the primary caesarean section.
Where possible, each case was matched to two controls. Given that the rate of placenta accreta rises significantly with each delivery by caesarean section, it was not possible to obtain a complete matched set of two controls per case.
A sample size of 100 cases and 200 matched controls was estimated to attain an 80% power to detect an increase in risk of placenta accreta associated with a primary elective caesarean section, with an odds ratio of 2.0.
Medians and ranges were used to summarise continuous data, and frequency distributions were used to summarise categorical data. Unmatched univariate comparisons of outcomes between cases and controls were conducted using Mann–Whitney U-tests and Fisher's exact tests for continuous and categorical data, respectively. Conditional logistic regression accounting for the matching factors and variable matching design was used to examine the effects of the first caesarean section on the risk of accreta. The effects of a primary elective caesarean section on placental invasion and other risk factors were summarised using odds ratios (ORs) and 95% confidence intervals (95% CIs). Statistical analysis was performed using sps 18 for windows (SPSS Inc., Chicago, IL, USA) and logxact 8.1 (Cytel Software Corporation, Cambridge, MA, USA). All hypothesis tests were two-sided and P < 0.05 was considered statistically significant.