To investigate the association between caesarean section and later endometriosis.
To investigate the association between caesarean section and later endometriosis.
A prospective cohort study.
The Swedish Patient Register (PAR) and the Swedish Medical Birth Registry (MBR).
Women who were delivered in Sweden between 1986 and 2004.
Women with the diagnosis of endometriosis, defined as codes 617 (International Classification of Diseases, ninth revision, ICD–9) or N80 (ICD–10), were retrieved from the PAR. Obstetric outcome was assessed through linkage with the MBR. Out of 709 090 women, 3110 were treated as inpatients with a first diagnosis of endometriosis after their first delivery. Women with a diagnosis of endometriosis before their first delivery were excluded. Cox analyses were performed to obtain hazard ratios for endometriosis and adjusted for maternal age at first delivery, body mass index, maternal smoking, and years of involuntary childlessness at study entry. Kaplan–Meier estimates were performed to calculate the risk according to time elapsed.
In-hospital diagnosis of endometriosis.
The Cox analyses yielded a hazard ratio of 1.8 (95% CI 1.7–1.9) for endometriosis in women who had had a previous caesarean section compared with women with vaginal deliveries only. The risk of endometriosis increased over time: one additional case of endometriosis was found for every 325 women undergoing caesarean section within 10 years. No increase in risk could be seen after two caesarean deliveries. The risk of caesarean scar endometrioma was 0.1%.
In addition to the recognised risk of scar endometrioma, we found an association between caesarean section and general pelvic endometriosis. Further studies are needed to confirm our findings.
The increasing rate of caesarean sections has raised concerns about both the short- and the long-term maternal consequences of the procedure.[1-6] The literature on endometriosis as a sequel of caesarean section has focused on the risk for scar endometrioma,[7-9] which has also been described in women who delivered vaginally after an episiotomy. Scar endometriomas usually present as a lump in the incision, sometimes mistaken for an incisional hernia, and occasionally with cyclic pain, making surgery necessary. However, it is possible that the development of pelvic endometriosis after caesarean section is overlooked because symptoms of chronic pelvic pain, dysmenorrheal, and dyspareunia are less specific. Endometriotic lesions have been found following laparoscopic subtotal hysterectomy, but the risk of endometriosis developing within the pelvis after caesarean section has not been extensively explored. Endometriosis is associated with substantial morbidity, being associated with a reduction in health-related quality of life, an increased risk for ovarian cancer, and possibly even malignant transformation within scar endometrioma.
A previous caesarean section is known to be associated with an increased risk of chronic pelvic pain. Whether this is related to endometriosis is unknown. The pathophysiology behind endometriosis is uncertain, but one theory is that degenerating endometrial components undergo metaplastic transformation when introduced into the abdominal cavity, and this may theoretically happen when the uterine cavity is opened in the course of abdominal delivery. Therefore, the objective of this study was to investigate the association, if any, between in-hospital diagnosis of endometriosis and caesarean section.
In a prospective cohort study the Swedish Patient Register (PAR), kept by the National Board of Health and Welfare, Stockholm, was used to identify women with a diagnosis of endometriosis (International Classification of Diseases, ninth revision, ICD9, 617; ICD10, N80). For scar endometriosis the diagnosis of 617G (ICD9) and N80.6 (ICD10) was used. Using the personal identification number assigned to each resident in Sweden, the data were linked to the Swedish Medical Birth Registry (MBR), which is also kept by the National Board of Health.
The PAR contains information on diagnoses (1987–1996, ICD9; 1997 and onwards, ICD10) and operation codes of all inpatients admitted to any Swedish hospital. The MBR is also kept by the National Board of Health and Welfare, and contains medical information on nearly all deliveries in Sweden (with a coverage of about 99%). Standardised record forms are used at all antenatal clinics, all delivery units, and at all paediatric examinations of newborn infants in the maternity ward. Information on maternal smoking and body mass index (BMI, kg/m2) are prospectively recorded by the midwife at the first visit to the antenatal centre. Copies of the standardised record forms are sent to the National Board of Health and Welfare, where they are computerised.
Women were included if they gave birth to their first child between 1986 and 2004, a period that was covered by the PAR register. Cases were excluded if they had a reported diagnosis of endometriosis before their first delivery, and for women who had been diagnosed more than once, only the first diagnosis was counted. The time for entry to the study was set to the date of the first delivery. The time for the study exit was set at the date of the first diagnosis of endometriosis, the date of the 55th birthday, or on 31 December 2004 (when the data set was retrieved), depending on which event happened first. In the descriptive tables, women with vaginal births before their first caesarean section were displayed only once, and are presented in the caesarean section group only. However, in the analyses, women with vaginal births before their first caesarean section contributed with person-months to the vaginal births group before the first caesarean section. The main hypothesis was that a caesarean section may increase the risk for endometriosis several years after the operation. Thus, after a woman's first caesarean section, all her person-months were designated to the caesarean section group, irrespectively of whether the caesarean section was followed by vaginal deliveries or not.
Cox analyses were performed in order to obtain hazard ratios (HRs) for endometriosis after the first caesarean section versus the first vaginal labour. The time at risk for each woman and time to diagnosis, respectively, was defined as explained above. Adjustments were made for maternal age at first delivery (continuous variable, quadratic model), BMI (linear, continuous), maternal smoking (yes/no), and years of involuntary childlessness (linear, continuous) at study entry (the first delivery). Kaplan–Meier estimates were computed to produce a graph illustrating the percentage of women who had ever been diagnosed with endometriosis by the time elapsed from the first delivery, and delivery mode. The COX- and the Kaplan–Meier analyses were performed using gauss. A comparison between the descriptive characteristics of women with caesarean section and women with vaginal births only was evaluated using chi-square analyses: P < 0.05 was considered to be significant.
Table 1 shows the descriptive characteristics at first delivery, of women who had at least one caesarean section or vaginal deliveries only, respectively, during the study period. Women who had any caesarean section were older, were more often overweight, smoked more, and had longer periods of involuntary childlessness, than had women with vaginal births only. Therefore, all these factors were considered to be possible confounders in the analyses of a possible association between caesarean section and endometriosis.
|Study group||P a|
|At least one caesarean section||Vaginal births only|
|n (%)||n (%)|
|<20 years||4832 (3.7)||29 155 (5.0)||<0.001|
|20–24 years||30 525 (23.4)||171 536 (29.6)|
|25–29 years||47 483 (36.4)||227 057 (39.2)|
|30–34 years||32 399 (24.9)||115 656 (20.0)|
|35–39 years||12 205 (9.4)||30 458 (5.3)|
|40+ years||2861 (2.2)||4923 (0.9)|
|Maternal BMI (kg/m2)b|
|<20||11 236 (12.3)||65 765 (16.0)||<0.001|
|20–24||50 357 (55.1)||248 377 (60.6)|
|25–29||20 960 (22.9)||73 637 (18.0)|
|30+||8829 (9.7)||22 026 (5.4)|
|Not known||38 923||168 980|
|No smoking||97 557 (80.6)||441 702 (81.3)||<0.001|
|<10 cigaretes/day||15 757 (13.0)||68 908 (12.7)|
|≥10 cigaretes/day||7782 (6.4)||32 673 (6.0)|
|Not known||9209||35 502|
|No||11 5854 (88.9)||533 439 (92.2)||<0.001|
|1–2 years||6739 (5.2)||25 606 (4.4)|
|3–4 years||3852 (3.0)||10 754 (1.9)|
|5+ years||3855 (3.0)||8957 (1.5)|
|1||44 089 (33.8)||186 442 (32.2)||<0.001|
|2||61 467 (47.2)||286 991 (49.6)|
|3||19 318 (14.8)||87 393 (15.1)|
|4+||5431 (4.2)||17 959 (3.1)|
|Minimum follow-up timed|
|At least 5 years||86 831 (66.6)||438 449 (75.8)||<0.001|
|At least 10 years||54 388 (41.7)||310 908 (53.7)|
|At least 15 years||21 096 (16.2)||136 468 (23.6)|
Table 2 displays the percentage of women with a diagnosis of endometriosis reported to the PAR register in 1987–2005 by mode of delivery and length of follow-up. From Table 2 it is evident that the estimated endometriosis rate is heavily dependent on the length of follow-up. The crude rates indicate that the incidence of endometriosis was higher among women who underwent at least one caesarean section than among women who had vaginal deliveries only.
|At least one caesarean section||Vaginal births only|
|n (%)||n||n (%)||n|
|Total n||749 (0.6)||130 305||2361 (0.4)||578 785|
|Follow-up of at least 5 years||715 (0.8)||86 831||2288 (0.5)||438 449|
|Follow-up of at least 10 years||608 (1.1)||54 388||2042 (0.7)||310 908|
|Follow-up of at least 15 years||311 (1.5)||21 096||1211 (0.9)||136 468|
Cox analyses were performed in order to adequately consider the different lengths of follow-up. These analyses yielded an almost doubled risk for endometriosis in women who had had at least one caesarean section (HR 1.8, 95% CI 1.66–1.94), compared with women who had vaginal deliveries only (Table 3). The estimate was only marginally altered when adjustments for maternal age, BMI, smoking, or years of involuntary childlessness were made. The results from the Cox analysis, in combination with the absolute risk for endometriosis in the current population, concluded that the numbers needed to harm within 10 years were 325, that is one additional case of endometriosis was found for every 325 women undergoing caesarean section within 10 years. We did not detect any increased risk of endometriosis in women with two caesarean sections, compared with women with one caesarean section (HR 0.77, 95% CI 0.49–1.22). The risk for scar endometriosis in all women delivered by caesarean section was 0.1%. In women with the diagnosis of endometriosis after caesarean section the proportion with scar endometriosis was 7/749 (9%).
|Factor||Univariate models||Multivariate modela|
|HR||95% CI||HR||95% CI|
|5–year increase, linear term||0.59||0.45–0.78||0.74||0.56–0.98|
|Quadratic term||1.06||1.03–1.08 Simultaneousb P = 4.0 × 10−9||1.03||1.00–1.06 Simultaneousb P = 0.073|
|Body mass index (kg/m 2 )|
|Yes versus no||1.22||1.13–1.33||1.20||1.11–1.30|
|Years of involuntary childlessness|
|1-year step increase, linear term||1.11||1.09–1.13||1.09||1.07–1.11|
|Any caesarean section|
|Yes versus no||1.82||1.69–1.97||1.79||1.66–1.94|
Figure 1 shows the Kaplan–Meier estimates (with 95% CIs) for endometriosis by years after the first vaginal delivery or the first caesarean section, respectively. For both groups, the proportion of women who had ever had a diagnosis of endometriosis increased linearly with the time elapsed from the first vaginal or first abdominal delivery, respectively. The curve showing the proportion of women who had ever had a diagnosis of endometriosis among women who underwent caesarean section had a considerably steeper slope than the corresponding slope for women who had vaginal deliveries only.
In this large population-based prospective cohort study, we have found an association between in-hospital diagnosis of any endometriosis and at least one previous caesarean section. The risk was related to time of follow-up. No increased risk could be shown after two caesarean sections. The risk for abdominal endometriosis was higher than for scar endometrioma, which is not previously known (any endometriosis 0.6%, scar endometriosis 0.1%).
The main advantage is the large size of this study. The Swedish registries contain information on practically all women delivered, and validation has shown that the data are reliable. All Swedish citizens obtain health care funded publicly, and few seek care outside the system.
We were unable to control for all confounding factors. The diagnosis of endometriosis is often delayed, and undiagnosed endometriotic lesions may have been present prior to the first caesarean section. However, this could also be the case in the vaginal delivery control group. We excluded women who had had the diagnosis of endometriosis before the first delivery, but as the PAR started in 1987, women with an in-hospital diagnosis of endometriosis before that time could not be excluded. Also, no reliable register on outpatients is available in Sweden for the period studied. If the diagnosis of endometriosis was made before 1987, or merely as an outpatient diagnosis, this and the ensuing infertility may have contributed to the indication for caesarean section, and thus may have distorted the results. On the other hand, adjustment for years of involuntary childlessness only marginally altered our results.
Breastfeeding has been reported to be less frequent after caesarean section. Women with previous infertility may be less inclined to use hormonal contraceptives. Both may decrease the risk for endometriosis by inhibiting ovulation. Likewise, previous abdominal surgery has been linked to an increased risk for endometriosis. We lacked this information also, which could influence the results. We did not have access to information on marital status or educational level; however, previous experience from the Swedish Medical Birth Register has shown that maternal smoking, parity, age, and BMI are the factors that are most strongly linked with delivery mode. Despite this, adjusting for these factors no more than marginally changed the HR for endometriosis (caesarean section versus vaginal birth). Therefore, we do not think that the analysis would benefit from including more possible confounding factors.
Endometriosis is clinically important in increasing the risk for pelvic pain, infertility, and cancer; however, it is unclear from our study whether the observed HR of 1.8 is merely an association, or is truly indicative that caesarean section increases the risk of developing endometriosis. Two caesarean sections did not increase the risk of being diagnosed with endometriosis in our study. This absence of a dose–response effect may, however, reflect the fact that women who develop endometriosis after a first caesarean section may be less likely to have another delivery because of pain symptoms and possible subfertility. Furthermore, women diagnosed with endometriosis after first caesarean section are no longer at risk of developing endometriosis for the first time.
Theoretically, pregnancy and caesarean section could increase the risk for endometriosis. The pathogenesis of endometriosis is not fully known, but immunological factors as well as metaplasia are considered to be important. Pregnancy is a state of altered immune response. Metaplasia can occur when endometrial cells are spread in the abdominal cavity after entry into the uterus. One recent study has shown that scar endometriomas are more frequent after unlaboured caesarean sections. The authors conclude that this may elucidate the pathogenesis of endometriosis. Immunological changes occur when labour starts. These changes may contribute to a lower risk of endometriosis in laboured caesarean sections.
Further studies should focus on investigating the pathogenesis of endometriosis in relation to pregnancy and delivery, as well as studies minimising confounding factors such as diagnosis of endometriosis before the first delivery, previous surgery, breastfeeding, and hormonal contraception. At present, the information from this study cannot be used clinically.
There are no conflicts of interest to declare.
EA took an active role in the conception, planning, carrying out, analysing, and writing-up of this study. MT and KK planned, carried out, analysed data, and wrote up the study.
The study was approved by the Research Ethics Committee at Karolinska Institutet, Stockholm, Sweden, no. 2005/89-31, date is 16 February 2005.
Financial support was provided by the Karolinska Institutet and Martin Rind foundations, and by the Swedish Council for Work Life and Social Research.