Caesarean section and subsequent ectopic pregnancy: a systematic review and meta-analysis

Authors


Correspondence: Ms. SM O'Neill, National Perinatal Epidemiology Centre, 5th Floor, Cork University Maternity Hospital, Wilton, Cork, Ireland. Email sinead.oneill@ucc.ie

Abstract

Background

Caesarean section rates are increasing worldwide, and the long-term effects are unknown.

Objective

To evaluate the risk of subsequent ectopic pregnancy in women with a previous caesarean section, compared with vaginal delivery.

Search strategy

Systematic review of the literature using CINAHL, the Cochrane Library, Embase, Medline, PubMed, SCOPUS and Web of Knowledge, published from 1945 until 17 July 2011.

Selection criteria

Cohort and case–control designs reporting on the mode of delivery and subsequent ectopic pregnancy. Two reviewers independently assessed the titles, abstracts, and full articles to identify eligible studies, using a standardised data collection form, and also assessed the study quality. Reference lists of the studies included were also cross-checked.

Data collection and analysis

Odds ratios (ORs) were combined using a random-effect model to estimate the overall association between caesarean section delivery and the risk of subsequent ectopic pregnancy.

Main results

Thirteen studies were included, which recruited a total of 61 978 women. Five studies reported adjustment for confounding factors, and the pooled OR of subsequent ectopic pregnancy following a caesarean section was 1.05 (95% CI 0.51–2.15). The removal of one study that reported outlier results yielded a pooled OR of 0.82 (95% CI 0.42–1.61). The pooled crude OR for all 13 studies was 1.36 (95% CI 0.99–1.88).

Author's conclusions

This review found no evidence of an association between prior caesarean section delivery and the occurrence of a subsequent ectopic pregnancy, but the studies included were of poor or variable quality, and only a small number adjusted for potential confounding factors. Further research of a higher methodological quality is required to assess any potential association between mode of delivery and subsequent ectopic pregnancy.

Introduction

Ectopic pregnancy is defined as a pregnancy that implants outside the uterine cavity, and is one of the leading causes of morbidity and mortality for pregnant women.[1-3] In many parts of the world, including the UK, there has been a doubling in its incidence over recent decades,[4, 5] and it remains a significant cause of preventable maternal death.[6] Approximately 2% of all pregnancies in the USA are diagnosed as an ectopic pregnancy,[7, 8] and incidence rates have increased by greater than three-fold since the 1970s.[9, 10]

Several risk factors for ectopic pregnancy have been identified, including smoking at the time of conception,[2, 11, 12] a history of pelvic inflammatory disease (PID),[13-15] previous ectopic pregnancy,[2, 13, 16] previous pelvic surgery,[2, 14, 15] and the use of intrauterine devices (IUDs).[17-19] Having a previous caesarean section has also been implicated as a risk factor for subsequent ectopic pregnancy[20-23]; however, to date, the evidence has been conflicting.[24, 25] The potential underlying mechanisms for an association between caesarean section and subsequent ectopic pregnancy are unclear, and may relate to placental bed disruption, infection, or adhesion formation, which in turn may be influenced by the indication for the caesarean section. These mechanisms seem credible in light of the links between caesarean section and subsequent ectopic pregnancy, placenta praevia, and placental abruption.[21, 26]

In the context of the sustained increase in caesarean section rates in developed and developing countries over the past three decades,[27] the aim of this systematic review was to examine the association between caesarean section and subsequent risk of developing an ectopic pregnancy.

Methods

Search strategy

We carried out a systematic review and meta-analysis in accordance with the Meta-analysis of Observational Studies in Epidemiology (MOOSE) guidelines of published work, without any language restrictions.[28] We selected relevant studies published between 1945 and 17 July 2011, from CINAHL, the Cochrane Library, Embase, Medline, PubMed, SCOPUS, and Web of Knowledge databases with the following combined text and medical subject headings (MeSHs), including the exposure, outcome, female gender only, and study design (#Caesarean section AND #Ectopic Pregnancy AND #Female AND #Case control OR #Cohort study; Table S1). We supplemented our search by cross-checking the reference lists of all relevant articles. We included studies with published quantitative estimates (including variability) of the association between mode of delivery and subsequent ectopic pregnancy. The eligibility criteria for inclusion in the meta-analysis were:

  1. Reported data were from an original study (i.e. no review articles).
  2. Cohort or case–control study in which mode of delivery was reported and subsequent ectopic pregnancy was the end point.
  3. Reporting of relative risk (RR), odds ratio (OR) or hazard ratio (HR) (or sufficient data in order to calculate these parameters), of the mode of delivery associated with subsequent ectopic pregnancy.

Study and data collection processes

Potentially eligible titles and abstracts of identified studies from the search strategy were reviewed using the inclusion and exclusion criteria. The full-text article was obtained for all potentially eligible articles.

Data abstraction

Using a standardised data collection form, two reviewers (S.M.O.N. and P.M.K.) independently abstracted data on study design, year of study, characteristics of the study participants, risk factors, ectopic pregnancy, and potentially confounding variables, including smoking, history of surgery, history of miscarriage or ectopic pregnancy, and use of fertility medications. Inconsistencies in data abstraction between reviewers were reviewed and resolved through consensus.

Statistical analysis

Our primary analysis was a comparison of the risk estimates of subsequent ectopic pregnancy in women with a previous caesarean delivery versus previous vaginal delivery. Pooled estimates across studies were obtained by means of random-effects models using the adjusted estimates from the included studies to calculate a pooled OR. We estimated the percentage of variability between studies attributable to between-study heterogeneity with the I2 statistic. Studies were weighted according to an estimate of statistical size defined as the inverse of the variance of the OR. We repeated the primary analysis excluding one study that reported outlier results (a very high OR). All studies with adjusted estimates were case–control studies of multiparous women, and therefore no sensitivity analysis by study design or parity were conducted. We derived a funnel plot of the overall OR and its standard error (SE) to assess for the possibility of publication bias. Statistical analyses were performed using sas 9.2 (SAS Institute Inc.), excel 2010 (Microsoft), and revman 5.0 (The Cochrane Collaboration).

Quality assessment

Two reviewers (S.M.O.N. and P.M.K.) measured the quality of the included studies using a predefined estimation of six items of potential bias: selection, exposure, outcome, analytical, attrition, and confounding (Table S2). This bias assessment instrument has been described in detail elsewhere.[29]

Results

Summary of study characteristics

We identified 8028 non-duplicated articles (Figure 1), 24 of which included data for the association between mode of delivery and subsequent ectopic pregnancy, and we then reviewed the full text of these articles. The most common reasons for exclusion of studies were failure to report the outcome or the exposure of interest, study designs that were not included in the eligibility criteria, and non-original study reports (i.e. letters, reviews, or editorials). Eight studies met the inclusion criteria, and five further eligible articles were identified from cross-checking of the reference lists, yielding a total of 13 articles for inclusion in the review: four cohort studies[21, 22, 24, 25]; and nine case–control studies.[1, 15, 20, 23, 30-34] The studies included were conducted in the UK,[24, 25] Sweden,[22, 33] Finland,[21, 34] Greece,[23] the USA,[1, 31, 32] Italy,[15, 20] and Turkey (Table 1).[30]

Table 1. Characteristics of studies included in the systematic review and meta-analysis of previous caesarean section and subsequent ectopic pregnancy
Study (year)RegionPopulation samplingSettingExposed (n)Unexposed (n)

LaSala et al. (1987)

Retrospective case–control

New York, USAAll women having an index caesarean section in 1978, followed until 1981 for next birth; age-matched vaginal deliveriesDaily obstetric logbook records in hospital291279

Hemminki (1991)

Retrospective cohort

SwedenAll Swedish primiparas who had a caesarean section in 1973 or 1976; age-matched vaginal deliveriesNational birth and hospital discharge registries1973 = 2578 1976 = 38221973 = 2578 1976 = 3822

Hemminki et al. (1996)

Retrospective cohort

FinlandAll women having an index caesarean section between 1987 and 1993; age-matched vaginal deliveriesNationwide birth and hospital registers74748999

Bastianelli et al. (1998)

Retrospective case–control

Rome, ItalyMultiparous women diagnosed with an ectopic pregnancy between 1991 and 1996; matched vaginal deliveriesClinical Institute of Obstetrics & Gynaecology Hospital, Rome213213

Tower et al. (2000)

Prospective cohort

Nottingham, UKAll primiparous women delivering between 1992 and 1993; age-matched vaginal deliveriesHospital Maternity Information system576576

Mollison et al. (2005)

Prospective cohort

Aberdeen, UKWomen who delivered their first child between 1980 and 1997; matched vaginal deliveriesAberdeen Maternity Hospital363621735

Barnhart et al. (2006)

Retrospective case–control

Pennsylvania, USAMultiparous women diagnosed with an ectopic pregnancy between 1990 and 1999; vaginal deliveries, women with miscarriage controlsUniversity Hospital Database3671659

Karaer et al. (2006)

Prospective case–control

Ankara, TurkeyMultiparous women diagnosed with an ectopic pregnancy between 2003 and 2005; three control groupsZekai Tahir Burak Research Hospital225375

Nordenskjold et al. (1991)

Retrospective case–control

Lund, SwedenMultiparous women diagnosed with an ectopic pregnancy between 1982 and 1983; age-matched vaginal deliveriesDepartment of Obstetrics & Gynaecology, Lund Hospital119119

Parazzini et al. (1992)

Prospective case–control

Milan, ItalyMultiparous women with a diagnosis of ectopic pregnancy between 1989 and 1991; matched vaginal deliveriesClinica Mangiagalli, Milan Hospital120209

Kendrick et al. (1996)

Prospective case–control

Georgia, USAMultiparous black non-Hispanic women aged 18–44 years who had a surgically confirmed ectopic pregnancy between 1988 and 1990; matched vaginal deliveriesMetropolitan Hospital Admission Log138842

Tuomivarra et al. (1988)

Prospective case–control

Not stated (author based in Finland)116 consecutive multiparous women diagnosed with an ectopic pregnancy (no year stated); two age-matched control groups (planned and induced abortion)Hospital (face-to-face interview)116116

Michalas et al. (1992)

Prospective case–control

Athens, GreeceMultiparous women diagnosed with an ectopic pregnancy between 1988 and 1990; term delivery controlsAlexandra Maternity Hospital361420
Figure 1.

Selection of studies for inclusion in the systematic review.

The diagnosis of ectopic pregnancy was confirmed in the studies through one or more of the following methods: hospital database;[1, 20, 24, 25] patient charts;[23, 31-34] diagnostic tests;[1, 30, 33] interviews with mothers;[1, 31, 32] surgery;[15, 20, 33] or nationwide registers.[21, 22] Only two of the studies cited used the WHO International Classification of Disease (ICD) codes.[21, 22] Only one study differentiated whether the caesarean section was performed in an emergency or electively.[24] For the purpose of this review, a caesarean section is assumed to refer to either an emergency or elective delivery in all of the other studies included.

Ten studies used matched comparison groups (including age, parity, and date of delivery),[15, 20-24, 31-34] and five studies adjusted for potential confounding factors such as age, history of previous pregnancy loss, marital status, and parity.[1, 15, 23, 30, 31] Adjustment for other potential confounding factors (i.e. history of surgery, infertility, and pelvic inflammatory disease [PID]) varied between the studies. None of the studies reported a sample size or power calculation.

Primary analysis

From the 13 studies included, data were available on 61 978 women, of which 1131 (1.8%) were diagnosed with ectopic pregnancy. Of these, only five studies adjusted for confounding factors, and so these are presented in the overall meta-analysis. The primary analysis therefore includes data on 4716 women and 490 events. The random-effects model is reported, as there was evidence of significant heterogeneity between the studies in the fixed-effects model (I2 = 88%, P = <0.00001). The pooled OR of having an ectopic pregnancy among women with prior caesarean section delivery versus vaginal delivery was 1.05 (95% CI 0.51–2.15; Figure 2). The crude pooled OR (including the crude estimates of the five studies that adjusted for confounding factors) was 1.04 (95% CI 0.52–2.05). One study that reported outlier results was removed from the overall meta-analysis (Figure 3),[15] producing a pooled OR of 0.82 (95% CI 0.42–1.61). When all 13 eligible studies were meta-analysed using crude estimates, a pooled OR of 1.36 (95% CI 0.99–1.88) was yielded (data not shown). Visual inspection of the funnel plot (Figure 4) suggested that there was no evidence of any publication bias.

Figure 2.

Random-effects model of the risk of subsequent ectopic pregnancy associated with caesarean section compared with vaginal delivery from five published studies, including 4716 participants and 490 events.

Figure 3.

Sensitivity analysis using a random-effects model of the risk of subsequent ectopic pregnancy associated with caesarean section, compared with vaginal delivery, after the removal of one study reporting an extreme outlier result (high OR value).

Figure 4.

Funnel plot assessing publication bias in the risk of subsequent ectopic pregnancy associated with caesarean section, compared with vaginal delivery.

Quality assessment

The quality of the studies included was based on six items for the evaluation of possible bias (Tables 2, 3). Further details on the instrument used to assess for study quality can be provided on request (Table S2). The overall likelihood of bias was considered to be moderate in seven studies (where no adjustment for confounding factors was reported), and minimal in five studies. The quality of the studies overall all was considered to be satisfactory.

Table 2. Quality assessment based on evaluation of bias in cohort studies investigating the risk of subsequent ectopic pregnancy with previous caesarean section delivery
StudySelection biasExposure biasOutcome assessment bias*Confounding factor biasAnalytical biasAttrition biasOverall likelihood of bias
  1. NR, not reported; NA, not applicable. *Confounding factor bias: evaluation of each study's assessment of potential confounders by four methods: adjustment with regression; matching; assessment of potential confounders on univariate analyses that were not significantly different between groups; assessment of potential confounders on univariate analyses that were different between groups and not controlled for.

Mollison et al. (2005)Minimal (select group, but eligibility explained)Low (assessment from data set)Low (assessment from data set)Moderate (no adjustment for confounding factors)Minimal (sample matched, no sample size calculation)Minimal (no loss to follow-up)Moderate
Hemminki (1991)Minimal (nationwide register)Low (assessment from nationwide register)Low (assessment from nationwide register)Moderate (NR whether assessed for confounding factors)Minimal (sample matched, no sample size calculation)Minimal (no loss to follow-up)Moderate
Hemminki et al. (1996)Minimal (two nationwide registers)Low (from nationwide register)Low (from nationwide register)Moderate (NR whether assessed for confounding factors)Minimal (sample matched, no sample size calculation)Minimal (none)Moderate
Tower et al. (2000)Minimal (select group, but eligibility explained)Low (assessment from data set)Low (assessment from data set)Moderate (NR whether assessed for confounding factors)Minimal (sample matched, no sample size calculation)Minimal (no loss to follow-up)Moderate
Table 3. Quality assessment based on evaluation of bias in case control studies investigating the risk of subsequent ectopic pregnancy with previous caesarean section delivery
StudySelection biasExposure biasOutcome assessment bias*Confounding factor biasAnalytical biasAttrition biasOverall likelihood bias
  1. NR, not reported. *Assessment of confounding factor bias: evaluated each study's assessment of potential confounders by four methods: adjustment with regression; matching; assessment of potential confounders on univariate analyses that were found not to be significantly different between groups; assessment of potential confounders on univariate analyses that were different between groups and not controlled for.

Bastianelli et al. (1998)Minimal (select group studied, but eligibility criteria explained)Minimal (assessment from medical records)Minimal (hospital records, diagnostic tests)Moderate (no adjustment for confounding stated)Minimal (matched analysis but no sample size calculation)Minimal (no loss to follow up)Moderate
Barnhart et al. (2006)Low (select group of population studied; university hospital)Minimal (assessment from data set)Minimal (diagnostic tests)Minimal (adjusted for age, gravity, parity, STDs, surgery)Minimal (matched analyses, but no sample size calculations)Minimal (no loss to follow up)Minimal
Karaer et al. (2006)Minimal (eligibility criteria and rationale of subjects explained)Minimal (face-to-face interview of mothers)Minimal (histopathological examination and interview)Minimal (adjusted for SES, smoking, obstetric history, STDs, sexual partner)Minimal (no sample size)Minimal (no loss to follow up)Minimal
Nordenskjold et al. (1991)Minimal (eligibility criteria and rationale of subjects explained)Minimal (medical records)Minimal (hospital records)Moderate (no adjustment for confounding stated)Minimal (analyses appropriate, no sample size calculation)Minimal (all subjects accounted for)Moderate
Parazzini et al. (1992)Minimal (eligibility criteria and follow-up explained)Minimal (direct questioning of mothers)Minimal (assessment from hospital records and mothers)Minimal (age, smoking, SES and BMI adjusted for)Minimal (analyses appropriate but no sample size calculated)Minimal (all subjects accounted for)Minimal
Kendrick et al. (1996)Low (select group, black women in Georgia USA)Minimal (direct questioning)Minimal (assessment from direct questioning of mother)Minimal (assessed for common confounding factors, inc. maternal age, parity, smoking, obstetric historyLow (all available eligible patients included, no sample size calculation)Minimal (no loss to follow up)Minimal
Tuomivaara et al. (1988)Moderate (NR where study took place, but case rationale explained)Minimal (face-to-face interview with a doctor)Minimal (hospital records and interview with mother)Moderate (matched for age and parity only. NR of confounding factor adjustmentModerate (only stated that Chi2 tests were used)Minimal (no loss to follow up)Moderate
Michalas et al. (1992)Low (select group, Greek women)Minimal (hospital records)Minimal (hospital records)Minimal (adjusted for age, parity, history of abortion, infertility)Minimal (matching, adjustment for confounding factors, regression)Minimal (no loss to follow up)Minimal
LaSala et al. (1987)Minimal (all women giving birth in the New York Hospital in 1978)Minimal (recorded from hospital chart)Minimal (assessment from hospital records)Moderate (no adjustment for confounding stated)Moderate (analyses not accounting for common statistical adjustment, no sample size calculation reported)Moderate (>20% attrition but reasons for loss to follow up explained)Moderate

Heterogeneity assessment

The characteristics of the included studies are shown in Table 1. Statistical heterogeneity (I2) was considered high, ranging from 86 to 88%, and is reported at the bottom left of each forest plot (Figures 2, 4).

Discussion

The overall findings based on available evidence suggest that women who previously delivered by caesarean section are not at increased risk of subsequent ectopic pregnancy. Crude analysis of all 13 included studies showed a 36% increased risk of subsequent ectopic pregnancy; however, this finding may be a result of confounding, as only five out of the 13 studies included adjusted for relevant risk factors.

In addition, caesarean section deliveries have many potential benefits, including a reduction in pelvic floor problems such as urinary incontinence,[35, 36] as well as in emergency situations such as fetal distress, prematurity, and dystocia.[37, 38] Given the large number of women worldwide having caesarean sections, any potential causal association between caesarean section delivery and subsequent ectopic pregnancy could have a profound public health and economic effect, with major implications for health policy in the future. The results of this review suggest that no such association exists.

Still, caesarean section delivery has a profound economic impact on the healthcare system. A caesarean section, even without complications, has been shown to cost almost twice that of a spontaneous vaginal delivery to health service providers.[39] Approximately half of the caesarean sections currently performed in the USA have no obstetric indication, resulting in increased postpartum recovery time and maternal morbidity, and a cost of over $1 billion per annum.[40, 41] However, there is a dearth of methodologically robust economic analyses of caesarean section versus alternative modes of delivery.

Strengths and limitations of the review

The strengths of this review include the thoroughness of the literature search, which examined multiple databases using a comprehensive search strategy, as well as the use of multiple assessors to identify the relevant studies. In addition, this is the first comprehensive systematic review on the risk of subsequent ectopic pregnancy following caesarean section delivery. Heterogeneity between studies was investigated with sensitivity analyses, and the pooled results were derived to measure the effect of caesarean section delivery on subsequent ectopic pregnancy; however, some of the studies included in the review encompass several inherent limitations.

Firstly, many of the studies had data of poor or variable quality, and the coding of the exposure or the outcome varied. Only two studies reported using the ICD codes and nationwide registers with almost complete coverage.[21, 22] The other studies used patient records, interviews, or surgical confirmation, which may be subject to errors in the form of miscoding or information bias. The power of the studies included is also questionable because of small or inadequate sample sizes, with many studies including a very limited number of women who experienced a subsequent ectopic pregnancy; the largest effect estimates were based on small sample sizes, as well as heterogeneous entry criteria. No study reported a priori sample size calculations. Most studies failed to adjust for potential confounding factors. In addition, only one of the studies included differentiated emergency versus elective caesarean section delivery. Subanalyses by type of caesarean section may provide interesting results on this relationship where research is limited. The quality of the outcome data may also be biased because of under-reporting of events in hospital records, recall bias, and measurement bias. The selection of an appropriate, unbiased control group is also an issue in the case–control studies. We have also not included studies published in the grey literature (i.e. those produced on all levels without peer review: i.e. from government, academic, business, and industry publications in paper print and electronic formats not widely disseminated and controlled by commercial publishers) in this review.

Conclusions and implications

In conclusion, compared with women who delivered vaginally, women who delivered by caesarean section did not have an increased risk of subsequent ectopic pregnancy; however, further research of a higher methodological quality is required to confirm this finding, as many of the studies in this review included small sample sizes and did not adjust for confounding factors.

Undoubtedly, there is a multitude of medical and personal factors affecting mode of delivery, and caesarean section delivery should be performed when medically required; however, the potential for subsequent adverse effects from caesarean section delivery should serve as a fundamental element in clinical decision-making. Ideally, women should be counselled regarding the range of potential short-term and long-term adverse outcomes associated with caesarean section delivery, especially with the introduction of the recent National Institute for Health and Clinical Excellence (NICE) guidelines in the UK, which recommend that women be given the right to request a caesarean section delivery (even when there is no obstetrical reason to do so).[42]

Disclosure of interests

None to declare.

Contribution to authorship

SMON, PMK, LCK, ASK, RAG, JEL, and TBH conceived and designed the study. SMON and PMK acquired, analysed and interpreted the data. SMON drafted the article. PMK, LCK, ASK, RAG, JEL, and TBH critically revised the article for important intellectual content. All authors agreed on the final article and approved its submission for publication. SMON will act as guarantor for the article.

Details of ethics approval

None required.

Funding

This research was supported by the National Perinatal Epidemiology Centre of Ireland as well as the Health Research Board Ireland through PhD scholarship funding.

Acknowledgements

We would like to acknowledge the staff of the National Perinatal Epidemiology Centre and the Anu Research Centre for their support on this project. We would also like to acknowledge Prof. John Browne of the Department of Epidemiology and Public Health in University College Cork for his didactic ‘Systematic Reviews in Healthcare’ course.

Ancillary