Previous induced abortions and the risk of very preterm delivery: results of the EPIPAGE study
Dr C. Moreau, INSERM U569, Hôpital de Bicêtre, 82 avenue du Général Leclerc, 94276 Le Kremlin Bicêtre Cedex, France.
Objectives To evaluate the risk of very preterm birth (22–32 weeks of gestation) associated with previous induced abortion according to the complications leading to very preterm delivery in singletons.
Design Multicentre, case-control study (the French EPIPAGE study).
Setting Regionally defined population of births in France.
Sample The sample consisted of 1943 very preterm live-born singletons (<33 weeks of gestation), 276 moderate preterm live-born singletons (33–34 weeks) and 618 unmatched full-term controls (39–40 weeks).
Methods Data from the EPIPAGE study were analysed using polytomous logistic regression models to control for social and demographic characteristics, lifestyle habits during pregnancy and obstetric history. The main mechanisms of preterm delivery were classified as gestational hypertension, antepartum haemorrhage, fetal growth restriction, premature rupture of membranes, idiopathic preterm labor and other causes.
Main outcome measures Odds ratios for very preterm birth by gestational age and by pregnancy complications leading to preterm delivery associated with a history of induced abortion.
Results Women with a history of induced abortion were at higher risk of very preterm delivery than those with no such history (OR + 1.5, 95% CI 1.1–2.0); the risk was even higher for extremely preterm deliveries (<28 weeks). The association between previous induced abortion and very preterm delivery varied according to the main complications leading to very preterm delivery. A history of induced abortion was associated with an increased risk of premature rupture of the membranes, antepartum haemorrhage (not in association with hypertension) and idiopathic spontaneous preterm labour that occur at very small gestational ages (<28 weeks). Conversely, no association was found between induced abortion and very preterm delivery due to hypertension.
Conclusion Previous induced abortion was associated with an increased risk of very preterm delivery. The strength of the association increased with decreasing gestational age.
Induced abortion is a relatively common event in France, occurring in 15 out of every 1000 women aged 15 to 44 years per year. This rate has been stable over time and is comparable to that from other European countries.1 There is some concern about the consequences of induced abortion on subsequent childbearing because induced abortion generally occurs at the beginning or middle of a woman's reproductive life. In France, 60% of women who have an induced abortion are younger than 30 years old.2 Therefore, many women probably intend to have a child after one or several induced abortions.
There has been much debate about the impact of induced abortion on subsequent pregnancy outcome as results of earlier studies were inconsistent and failed to draw clear conclusions regarding the obstetric risks following induced abortion.3–5 More recent studies have tended to show an excess risk of preterm delivery associated with induced abortion, with adjusted odds ratios between 1.3 and 2. These studies also suggested that the risk increases with the number of previous abortions.6–9
When analysing the risk factors for preterm delivery, it is important to differentiate the risk according to gestational age. This issue is a major public health concern as infants born before 33 weeks of gestation have a higher risk of morbidity and mortality than more mature babies. In addition, risk factors for very and moderately preterm births have been found to be similar, but the strength of the associations is different, especially for obstetric history.10 However, information concerning the impact of induced abortion on the risk of subsequent very preterm births remains limited. In a study conducted in Australia, Lumley8 found the risk to be higher for extremely preterm delivery (<28 weeks of gestation) than for more mature delivery (≥28 weeks). However, given the infrequency of very preterm delivery, only a few studies have focussed on this particular issue.7,8
The mechanisms by which induced abortions affect the outcome of subsequent pregnancies have rarely been explored. The role of infection and cervical trauma following surgically induced abortion has been discussed.3,11 Several studies have indirectly addressed this question by distinguishing between the risk of spontaneous preterm delivery after induced abortion, expected to be related to infectious mechanisms, and the risk of indicated preterm delivery after induced abortion, which is more often related to vascular complications during pregnancy.7,12,13 However, as suggested by Ancel et al.,14 the distinction between spontaneous and induced delivery only very partially reflects the underlying infectious or vascular complications responsible for preterm delivery.
The present study had two main objectives. The first was to quantify the risk of very preterm birth associated with previous induced abortion in a large population-based case–control study conducted in France in 1997. The second was to investigate differences in the risk of very preterm delivery according to the main pregnancy complications leading to the delivery, with the hypothesis that induced abortion could increase the risk of a subsequent very preterm birth due to infectious11,14–16 or mechanical processes,14,17 but not the risk of very preterm delivery due to vascular causes, especially hypertension.18
This study was undertaken using data from the EPIPAGE study, designed to explore very preterm births, their risk factors and their consequences. Recruitment took place in 1997 in all maternity wards in nine French regions, covering about one-third of all births in France.19 Cases were defined as all births, including stillbirths and late terminations of pregnancies for medical reasons, occurring between 22 and 32 completed weeks of gestation.
Two other groups were also selected in the same regions:
- 1A reference group of infants born alive at 39 or 40 weeks of gestation. Infants born during a one-week period were selected with the following criteria: one live birth at 39–40 weeks out of four was randomly selected from all maternity units in university hospitals and all maternity units with at least 2000 births per year; for all other maternity units, one maternity out of four was randomly selected, and all live births at 39 or 40 weeks from these selected maternities were included.
- 2A group of infants (stillbirths and live births) born at 33 and 34 completed weeks of gestation was used to estimate the variation of risks around 32 completed weeks of gestation. All eligible infants of this group born during a two-month period were included.
Gestational age at delivery was defined as the number of completed weeks of amenorrhea, based on the date of the last menstrual period, confirmed by ultrasonography in almost all cases.19
Women were enrolled during their postpartum stay in the hospital. Information was obtained partly from the medical records and partly by interviewing the mothers during their hospital stay. Data extracted from the medical records included information on maternal age, reproductive and medical history, care and complications of the current pregnancy, delivery and infant's health. The interviews of the mothers provided information on social and demographic characteristics, lifestyle and smoking habits before and during pregnancy.
This study was limited to pregnancies resulting in single births, alive at the beginning of the labour; 717 late terminations and 529 fetal deaths before labour were excluded. The initial sample included 3056 women (<33 weeks: n+ 2103; 33–34 weeks: n+ 294; 39–40 weeks: n+ 659). Women were excluded if one of the key pieces of information from the medical record was missing: 46 with missing information on previous induced abortions, eight with missing information on maternal age, two with missing information on parity and 167 with missing information on outcome of previous pregnancies. Thus, this study included 2837 women (<33 weeks: n+ 1943; 33–34 weeks: n+ 276; 39–40 weeks: n+ 618), of whom 405 (14.3%) had had at least one previous induced abortion.
We first compared the social demographic and obstetric characteristics of the women in the three groups: very preterm births (<33 weeks), moderately preterm births (33–34 weeks) and reference group (39–40 weeks of gestation). The following characteristics, known as risk factors for preterm or very preterm births,10,12,13,20,21 were examined: maternal age, parity and history of preterm delivery, marital status, level of education, employment during pregnancy, smoking habits during pregnancy, weight before pregnancy. Parity and history of preterm delivery were taken into account in a single variable defined in three categories: nulliparous, multiparous with no history of preterm delivery, multiparous with a history of preterm delivery. We looked for associations between previous induced abortion and very preterm delivery, and tested the possibility that the risk of preterm delivery increased with the number of previous induced abortions. Interaction between a history of preterm delivery and induced abortion was examined and the analysis was also performed after excluding women with a history of preterm delivery. We explored the differences in risk of preterm delivery associated with previous induced abortion according to the length of gestation, categorised in three groups as follows: 22–27 weeks, 28–32 weeks, 33–34 weeks. Finally, we estimated the risk of very preterm delivery related to previous induced abortions according to the main complications leading to preterm delivery. For this analysis, we defined six subgroups of women according to the main complication leading to very preterm delivery. Women were assigned to either one of these six categories, designed to be mutually exclusive, according to the following hierarchic order:
- 1hypertension or HELLP syndrome
- 2antepartum haemorrhage, including placenta praevia, placental abruption and blood loss before delivery without any other complication
- 3intrauterine fetal growth restriction without hypertension, placenta praevia or placental abruption
- 4PROM defined as premature spontaneous rupture of the membranes at least 12 hours before the onset of labour without placenta praevia, placental abruption, hypertension or fetal growth restriction
- 5idiopathic spontaneous preterm labour, defined as the onset of labour without hypertension, placenta praevia, placental abruption, fetal growth restriction or PROM
- 6all other cases
The analysis was carried out using univariate and multivariate logistic regression models providing unadjusted and adjusted estimates of the odds ratios of very preterm delivery according to the history of previous induced abortion. Polytomous regression models were used to estimate the odds ratios of very preterm birth by gestational age and by pregnancy complication leading to preterm delivery, with 95% confidence intervals. In all models, the reference group consisted of women whose current pregnancy ended at 39 or 40 weeks of gestation. Multivariate models included the following risk factors for preterm delivery20: maternal age, parity and history of preterm delivery, marital status, level of education, employment during pregnancy, weight before pregnancy and smoking during pregnancy. For several reasons, mainly linked to early neonatal deaths and region of inclusion, some of the women did not complete the interview, thus resulting in missing information on marital status, level of education, employment during pregnancy, smoking during pregnancy and weight before pregnancy. To limit the selection bias and to avoid excluding these women from the analysis, missing data were taken into account in the analysis by creating an ‘unknown’ category for each of these variables.
Statistical analyses were performed using the Stata software.22 The study received the approval of the CNIL (Commission Nationale de l'Informatique et des Libertés).
The social and demographic characteristics and the obstetric history of the women in the three groups are presented in Table 1. Mothers of preterm infants (very preterm or moderately preterm) were younger, had a lower educational level, and were more likely to live alone, to be unemployed and to smoke at the time of the current pregnancy than were mothers in the reference group. They were also more frequently nulliparous and were more likely to report a previous preterm delivery.
Table 1. Characteristics of mothers of very preterm, moderate preterm and control neonates. Values are presented as n (%),a unless otherwise indicated.
|<25 years||540 (27.8)||69 (25.0)||121 (19.6)|
|25–34 years||1098 (56.5)||164 (59.4)||418 (67.6)|
|≥35 years||305 (15.7)||43 (15.6)||79 (12.8)|
|% missing informationb||0.0||0.0||0.0|
|Parity = 0||1002 (51.6)||141 (51.1)||281 (45.5)|
|Parity > 0 with no previous preterm birth||575 (29.6)||86 (31.2)||310 (50.2)|
|Parity > 0 with a history of preterm birth||366 (18.8)||49 (17.7)||27 (4.4)|
|% missing informationb||0.0||0.0||0.0|
|Married||878 (52.6)||101 (45.1)||339 (59.4)|
|Unmarried cohabiting||599 (35.9)||92 (41.1)||191 (33.5)|
|Single||191 (11.5)||31 (13.8)||41 (7.2)|
|% missing informationb||14.2||18.9||7.6|
|Level of education|
|Primary school||112 (7.3)||14 (6.7)||20 (3.6)|
|Secondary||701 (45.4)||97 (46.2)||229 (41.2)|
|High school||312 (20.2)||44 (21.0)||112 (20.1)|
|University||418 (27.1)||55 (26.2)||195 (35.1)|
|% missing informationb||20.6||23.9||10.0|
|Employment during pregnancy|
|Yes||892 (54.2)||136 (61.3)||367 (65.5)|
|No||754 (45.8)||86 (38.7)||193 (34.5)|
|% missing informationb||15.3||19.6||9.4|
|Weight before pregnancy|
|≥45 kg||1500 (92.0)||206 (94.1)||537 (94.7)|
|<45 kg||131 (8.0)||13 (5.9)||30 (5.3)|
|% missing informationb||16.1||20.7||8.3|
|Smoking during pregnancy|
|0 cigarette per day||1022 (64.3)||132 (62.6)||418 (76.0)|
|<10 cigarettes||269 (16.9)||45 (21.3)||71 (12.9)|
|≥10 cigarettes||298 (18.8)||34 (16.1)||61 (11.1)|
|% missing informationb||18.2||23.6||11.0|
Women with a history of induced abortion were at higher risk of very preterm delivery than those with no such history (OR + 1.6, 95% CI 1.2–2.1). Similar results were obtained when controlling for maternal characteristics (Table 2) and when the analysis was performed without adjusting for the history of preterm deliveries (OR + 1.5, 95% CI 1.1–2.0). The association remained unchanged after excluding women with a history of preterm delivery (OR + 1.5, 95% CI 1.1–2.1). The risk tended to increase with the number of past induced abortions (Table 2).
Table 2. Risk of very preterm delivery according to the number of previous induced abortions. Values are presented as n (%), unless otherwise indicated.
|0||1644 (84.6)||555 (89.8)||1|| ||1|| |
|≥1||299 (15.4)||63 (10.2)||1.6||1.2–2.1||1.5||1.1–2.0|
|1||238 (12.2)||56 (9.1)||1.4||1.1–1.9||1.3||1.0–1.8|
|>1||61 (3.1)||7 (1.1)||2.9||1.3–6.5||2.6||1.1–5.9|
The adjusted risk for preterm delivery associated with a history of induced abortion tended to be higher for extremely preterm deliveries (22–27 weeks of gestation) than for other very preterm deliveries (28–32 weeks of gestation); no difference in risk was observed between 28–32 weeks and 33–34 weeks of gestation (Table 3).
Table 3. Risk of very preterm delivery by length of gestation associated with previous induced abortions.
|Reference group (39–40 weeks)||618||10.2||1|| ||1.0|| |
The main complications leading to very preterm births differed according to gestational age (Table 4); PROM and idiopathic spontaneous preterm labour were more frequently observed among the extremely preterm births (22–27 weeks) than among less extreme preterm births (28–32 weeks). Conversely, hypertension and fetal growth restriction were more common among the group of infants born between 28 and 32 weeks than among infants born before 28 weeks. A history of induced abortion significantly increased the risk of very preterm delivery associated with premature rupture of the membranes (OR + 1.7) and placenta praevia (OR + 2.4), but the associations were strongest for extremely preterm delivery (22–27 weeks). The risk of idiopathic spontaneous preterm labour related to past induced abortion (OR + 1.8) was higher in the extremely preterm group than in the other groups. An association between past induced abortion and very preterm delivery related to fetal growth restriction was observed in the group of infants born between 28 and 32 weeks of gestation (OR + 1.7). Finally, no significant associations were found with maternal hypertension (OR + 1.1).
Table 4. Risk of very preterm delivery associated with induced abortion according to the aetiological context leading to delivery.
|Reference group (39–40 weeks)||618|| ||1|| ||618|| ||1|| ||618|| ||1|| |
|Hypertension and HELLP syndrome||77||17||1.4||0.6–2.9||418||28.1||1.0||0.7–1.6||495||25.5||1.1||0.7–1.7|
|Other haemorrhages||0||0.0|| || ||11||0.7||0.8||0.1–6.9||11||0.6||0.8||0.1–6.4|
|Isolated fetal growth restriction||11||2.4||0.6||0.1–5.9||124||8.3||1.7||1.0–3.0*||135||6.9||1.7||1.0–2.8*|
|Idiopathic spontaneous preterm labour||157||34.7||1.8||1.0–3.0*||303||20.3||1.0||0.7–1.6||460||23.7||1.3||0.9–1.9|
|Other||7||1.6|| || ||50||3.4||1.7||0.7–3.8||57||2.9||1.5||0.7–3.2|
We found that a history of induced abortion was associated with a higher risk of very preterm delivery, especially for very preterm births (22–32 weeks) following PROM or placenta praevia, and for extremely preterm births (22–27 weeks) after idiopathic spontaneous preterm labour. Conversely, induced abortion was not associated with an increased risk of very preterm delivery after hypertensive complications.
This study is one of the first to explore the risk of very preterm delivery (<33 weeks of gestation) associated with previous induced abortion. Furthermore, it is the first to study this relationship according to the pregnancy complications leading to very preterm delivery, an important issue that needs to be considered to improve our understanding of the mechanisms by which induced abortion can affect subsequent pregnancies. As the study was geographically defined, it was possible to recruit all very preterm deliveries that occurred in the nine regions included, thus limiting selection bias. Nevertheless, some methodological points need to be discussed.
As commonly reported in fertility surveys based on women's reports, it is likely that induced abortion was under-reported. The extent of under-reporting varies between 40% and 65% in the literature.23 However, in this study, data on previous induced abortion were taken from hospital records that were filled in prior to enrolment, which reduces the risk of differential recall according to gestational age. In a specific study addressing this question in relation to cancer, Tang et al.24 found no differential reporting of induced abortion between cancer cases and controls. In addition, under-reporting varies according to women's social and demographic background and is more common among older women and women living alone, with a low educational level.25–27 Thus, we would expect more under-reporting among cases than among controls, and consequently, an under-estimation of the association.
In the EPIPAGE study, the control group was defined as live births occurring at 39–40 weeks of gestation, instead of a larger definition of deliveries at term, usually defined as 37 to 41 weeks of gestation. Due to this definition, it is possible that women in the control group had a lower level of most risk factors for preterm deliveries than the general population, including a lower frequency of previous induced abortions. Therefore, the relationship between previous induced abortions and very preterm delivery might have been over-estimated. However, using data from the 1998 French national perinatal survey, we checked that the use of controls at 39–40 weeks of gestation rather than ≥37 weeks, only slightly over-estimated the association between induced abortion and very preterm delivery (Ancel, personal communication). In addition, our findings are consistent with other studies.6,7,28
Due to early neonatal deaths, severe complications and difficulties in organising data collection in some units, some women did not complete the postpartum interview, thus resulting in missing information concerning their social and demographic background. The proportion of missing values was higher among cases than controls, but was not linked to a history of induced abortion. Treating missing data of a variable as a category allowed us to include all women in the analysis, thus increasing statistical power. The impact of such an approach is difficult to evaluate. However, in our study, the confounding effect of the characteristics we took into account seems limited as crude and adjusted ORs were very similar. Also, we performed sensitivity analyses by systematically assigning women with missing data either to the highest level of risk (heavy smoker, single, non-active, low weight…) or, on the contrary, to the lowest level of risk: the adjusted OR remained unchanged (OR + 1.5).
The association observed between very preterm delivery and previous induced abortion raises the question of possible residual confounding. However, the main socio-demographic characteristics (level of education, employment, marital status) and lifestyle characteristics (smoking) known as risk factors of preterm delivery have been taken into account in the analysis. Other behavioural characteristics which may be more common among women with a history of induced abortion, such as high risk sexual habits, have been suggested to influence preterm delivery.29 However, studies addressing this question give inconsistent results, showing either no or only weak associations between sexual behaviour and prematurity.
In our analysis, we adjusted for previous preterm deliveries, known to be a major risk factor for a subsequent preterm delivery.20 However, as no information concerning the chronology of the previous pregnancies was available, some of the previous preterm deliveries may also have been linked to an induced abortion that occurred earlier in life, thus resulting in over-adjustment of our model. However, the same analysis without adjusting for previous preterm deliveries gave similar results. Also, the association remained significant after excluding women with a previous preterm delivery.
Our findings are consistent with recent results that gave adjusted odds ratios between 1.3 and 2.7,28 As described by Lumley in Australia,8 we found a tendency for a stronger association with extremely preterm deliveries (22–27 weeks of gestation) than with less extreme gestational ages (28–32 or 33–34 weeks of gestation). Pregnancy complications vary with gestational age, which may explain some of the differences observed. However, very few studies have explored the underlying mechanisms by which induced abortion increases the risk of subsequent preterm delivery (<37 weeks),28 and none have addressed this question concerning very preterm delivery (<33 weeks). Spontaneous preterm delivery, which is expected to be related to infectious mechanisms, and indicated preterm delivery, which is more frequently related to vascular complications, are sometimes preceded by the same pregnancy complications.28 In our study, delivery was induced in 50% of women with placenta praevia and 24% of women with other types of antepartum haemorrhage. Our results improve our understanding of the relationship between induced abortion and very preterm delivery as we tested the association among different subgroups of very preterm births, based on the main complications leading to preterm delivery. Previous studies have suggested that infectious diseases following induced abortion account for the increase in the risk of preterm delivery.15,30 Krohn et al.11 found that women who had previously undergone an induced abortion were at higher risk than other women of intra-amniotic infection, which is a risk factor for PROM.16 This association may result from the revival of latent local infectious processes caused by surgery at the time of the abortion,15 or reveal a mechanical adverse outcome caused by dilatation, leading to cervical incompetence31 with a possible risk of upper genital tract infections.
Cervical instrumentation has also been suggested to increase the risk of endometrial damage, in such a way as to impair trophoblastic invasion and migration, thus increasing the risk of placenta praevia, a major cause of antepartum haemorrhage.17,32 In accordance with our initial hypothesis that an infectious or a mechanical mechanism was involved, we found that induced abortion increased the risk of very preterm delivery due to PROM and placenta praevia and the risk of extremely preterm delivery (<28 weeks) due to idiopathic spontaneous preterm labour. Conversely, no association was found between induced abortion and very preterm delivery due to hypertension. These findings are consistent with the results of a recent multicentre case–control survey conducted in 10 European countries.28 They all suggest that induced abortion could produce cervical and uterine abnormalities, responsible for an increased risk of subsequent preterm delivery. As medical abortion is expected to reduce mechanical injuries, it would be interesting to assess if it also reduces the risk of subsequent preterm delivery. We also found an association between induced abortion and very preterm delivery after fetal growth restriction that was not described in the European study. This needs further investigation.
Our results show that a history of induced abortion increases the risk of very preterm births, particularly extremely preterm deliveries. Both infectious and mechanical mechanisms may be involved, as we found that past induced abortions were more specifically associated with a higher risk of very preterm delivery following PROM and placenta praevia.
The authors would like to thank all women who participated in this survey. The authors also thank all those who contributed to data collection. The study was financially supported by INSERM (National Institute of Health and Medical Research), Merck-Sharp, Dohme-Chibret, la Fondation de la Recherche Médicale (The Medical Research Foundation) and la Direction Génerale de la Santé du Ministère des Affaires Sociales (French Ministry of Health).
Appendix: Composition of the EPIPAGE Group.
INSERM U149: B Larroque (national coordinator), PY Ancel, B Blondel, G Bréart, M Dehan, M Garel, M Kaminski, F Maillard, C du Mazaubrun, P Missy, F Sehili, K Supernant.
ALSACE: M Durand, J Matis, J Messer, A Treisser (Hôpital de Hautepierre, Strasbourg).
FRANCHE-COMTE: A Burguet, L Abraham-Lerat, A Menget, P Roth, J-P Schaal, G Thiriez (CHU St Jacques, Besançon).
HAUTE-NORMANDIE: C Lévêque, S Marret, L Marpeau (Hôpital Charles Nicolle, Rouen).
LANGUEDOC-ROUSSILLON: P Boulot, J-C Picaud (Hôpital Arnaud de Villeneuve, Montpellier), A-M Donadio, B Ledésert (ORS Montpellier).
LORRAINE: M André, J Fresson, JM Hascoët, JL Boutroy, (Maternité Régionale, Nancy).
MIDI-PYRENEES: C Arnaud, S Bourdet-Loubère, H Grandjean (INSERM U558, Toulouse), M Rolland (Hôpital des Enfants, Toulouse).
NORD-PAS-DE-CALAIS: C Leignel, P Lequien, V Pierrat, F Puech, D Subtil, P Truffert (Hôpital Jeanne de Flandre, Lille).
PAYS-DE-LOIRE: G Boog, V Rouger-Bureau, J-C Rozé (Hôpital Mère-Enfant, Nantes).
PARIS-PETITE-COURONNE: PY Ancel, G Bréart, M Kaminski, C du Mazaubrun (INSERM U149, Paris), M Dehan, V Zupan (Hôpital Antoine Béclère, Clamart), M Vodovar, M Voyer (Institut de Puériculture, Paris).