Adverse obstetrical outcomes after treatment of precancerous cervical lesions: a Belgian multicentre study

Authors


Dr M Arbyn, Unit of Cancer Epidemiology, Scientific Institute of Public Health, J. Wytsmanstreet 14, B-1050 Brussels, Belgium. Email marc.arbyn@wiv-isp.be

Abstract

Please cite this paper as: Simoens C, Goffin F, Simon P, Barlow P, Antoine J, Foidart J, Arbyn M. Adverse obstetrical outcomes after treatment of precancerous cervical lesions: a Belgian multicentre study. BJOG 2012;119:1247–1255.

Objective  To assess the impact of cervical intraepithelial neoplasia (CIN) treatment on the risk of (spontaneous) preterm delivery (PD) and small for gestational age (SGA) at birth.

Design  A multicentre cohort study.

Setting  Maternity wards of four academic hospitals in Belgium.

Population  Ninety-seven exposed pregnant women (with a CIN treatment history) and 194 nonexposed pregnant women (without a history of CIN treatment).

Methods  A questionnaire and check of obstetrical files included socio-demographic characteristics, risk factors for PD, obstetrical history for all women and characteristics of the CIN treatment for exposed women. Pregnancy outcomes were recorded after delivery. The influence of previous treatment of CIN on pregnancy outcomes, adjusted for confounding variables, was assessed by Cox regression and lifetables (for the outcome gestational age at birth) and by logistic regression (for the outcomes PD and SGA at birth).

Main outcome measures  Occurrence of PD and SGA at birth.

Results  Seventy-nine per cent of the women in the database were multiparous; 16.3% of women with a previous excisional treatment spontaneously delivered preterm, compared with 8.1% of unexposed women [odds ratio (OR), 2.19; 95% confidence interval (CI), 0.97–4.99]. When adjusting for confounding factors (ethnicity, HIV status, education, age, smoking and parity), the OR for PD was 2.33 (95% CI, 0.99–5.49). Excisional treatment did not have an impact on SGA at birth (OR, 0.94; 95% CI,0.41–2.15). The depth of the cone was >10 mm in 63.5% of the documented cases. Large cones, more than 10 mm deep, were associated with a significantly increased risk of PD (adjusted OR, 4.55; 95% CI, 1.32–15.65) compared with untreated women, whereas smaller cones (≤10 mm) were not significantly associated with PD (OR, 2.77; 95% CI, 0.28–27.59). The associations seen for PD with respect to the cone size did not hold for SGA at birth.

Conclusions  There was an increased risk of (spontaneous) PD after excision of CIN, in particular when the cone depth exceeded 10 mm.

Introduction

Cervical cancer is still an important health problem in Europe, with 52 000 cases and 27 000 deaths per year.1 In Belgium, 667 new cases (world age-standardised incidence rate, 9.4/100 000 women-years) were registered and approximately 275 women died (world age-standardised mortality rate, 2.7/100 000 women-years) from this disease in 2008.2 The incidence and mortality have dramatically decreased in developed countries as a result of cytological screening3,4 and the effective treatment of precancerous cervical lesions (CIN, cervical intraepithelial neoplasia). Currently, large loop excision of the transformation zone (LLETZ) is the most frequently used procedure to treat CIN in Belgium, as well as in other industrialised countries.5,6 Most of the women with high-grade CIN are of reproductive age.7 In Belgium, the peak incidence of conisation or LLETZ is noted in the 30–34-year age group, and 44% of treatments take place in women under the age of 35 years.8 As these women often show a definite desire to have children, it is important not to compromise future pregnancies by surgical interventions on the cervix. Meta-analytical work has shown evidence of pregnancy-related morbidity, such as preterm delivery (PD) (<37 weeks), low birthweight (<2500 g) and preterm spontaneous rupture of membranes after cold-knife conisation, LLETZ and laser conisation (all excisional treatments), but has also found substantial inter-study heterogeneity.9 In a more recent meta-analysis targeting severe adverse pregnancy outcomes, and including larger population-based studies, it was found that only cold-knife conisation was significantly associated with perinatal mortality, severe PD (<32/34 weeks) and extremely low birthweight (<1500 g).10 Both meta-analyses confirmed that ablative procedures (cryotherapy or laser ablation) are safe with respect to obstetrical outcomes.

Although there is evidence that incomplete excision, especially at the endocervical margins, is associated with a greater treatment failure rate,11,12 the overall conditions that determine success or failure of CIN treatment and, especially, the negative effects on obstetrical outcomes are not well known. The association between adverse pregnancy outcomes and previous treatment of CIN is most often assessed in retrospective studies. Therefore, a research initiative was set up comprising two cohorts. In the first cohort, women were recruited at the hospitals at which CIN was treated and the major outcome, treatment success or recurrence of CIN, was assessed over time in relation to patient, lesion and treatment characteristics. In the second cohort, which is the object of the current paper, obstetrical safety is assessed.

Methods

The present study was a multicentre clustered cohort study involving women delivering in four Belgian academic hospitals (‘Hôpital de la Citadelle’ and ‘Centre Hospitalier du Bois de l’Abbaye (CHBAH)’ in Liège, ‘Hôpital St. Pierre’ and ‘Hôpital Erasme’ in Brussels). A centralised web-based database was generated, which encoded the identity of enrolled women. From September 2008 onwards, women with a history of previous CIN treatment were selected (exposed group) among all women presenting at the labour ward for delivery. Following the recruitment of an exposed woman, the next two women that met the criteria for nonexposure were chosen (admittance in the same maternity ward, who never had treatment for CIN or a history of CIN). By November 2010, 104 exposed pregnant women delivering in one of the university maternity wards and 229 clustered unexposed women were recorded in the database.

The study included only singleton pregnancies. Exposed women (and their clustered nonexposed women) with insufficient data on previous treatment (n = 5), without a nonexposed cluster (n = 1), lacking pregnancy outcome data (n = 1) and unexposed women in excess (n = 21) were excluded, resulting in a final study population containing 97 exposed and 194 nonexposed women.

A questionnaire and check of obstetrical medical files included socio-demographic characteristics (maternal age, ethnicity, education level, HIV status and smoking behaviour), obstetrical history [gravidity, parity, risk factors for prematurity (gestational hypertension, uterine malformation, cervical incompetence, body mass index, diabetes, hydramnios, intrauterine growth restriction, urogenital infection at the end of the first trimester, previous PD for multiparous women), management of the pregnancy, duration of the pregnancy (in postmenstrual weeks), mode of delivery and neonatal parameters (birthweight in grams) for all women, and characteristics of the treatment of CIN for exposed women (type of treatment, characteristics of the lesion, margin status and cone depth). Pregnancy outcomes were recorded after delivery. The primary endpoint was the number of (spontaneous) PDs [before 37 weeks of gestation, but more than 20 weeks (miscarriages were not included)] in exposed versus nonexposed women. Secondary endpoints were small for gestational age (SGA) and low birthweight below 2500 g. SGA was defined as having a weight below the 10th percentile for a given gestational age, taking into account the gender of the baby and the parity of the mother using updated standard tables from the Netherlands.13

Qualitative variables were reported as absolute and relative frequencies, whereas quantitative variables were described in terms of the mean, minimum and maximum range, 95% confidence interval (95% CI) or median and interquartile range (for non-normally distributed variables). Comparisons between exposed and nonexposed women were carried out by Student’s t-test (to compare two continuous variables), Pearson’s chi-squared (to compare categorical variables between exposed and nonexposed women) or Wilcoxon rank-sum test (to compare non-normally distributed continuous variables). Analyses were completed with univariable and multivariable logistic regression for the dichotomous endpoints to obtain odds ratios (ORs) and 95% CIs for the association with the exposed (all treatments, excisional treatments, LLETZ, laser conisation and different cone depths) versus nonexposed population, adjusted for potential confounding factors (ethnicity, HIV status, education level, maternal age, smoking behaviour and parity). The influence of previous excisional treatment of CIN on pregnancy outcomes, adjusted for confounding variables, was assessed by Cox regression and lifetables (for the outcome gestational age at birth). Clustered analyses were performed [Cox regression using the triplet (one exposed pregnant woman and the two next nonexposed pregnant women) as a statistical cluster]. A subgroup analysis was performed on multiparous women to take previous obstetric history (occurrence of previous PD) into account as a confounding factor. All deliveries and a subgroup of the spontaneous deliveries only were included in the univariable and multivariable logistic regression, Cox regression and lifetables. For all analyses, statistical significance was defined at P < 0.05. Statistical analysis was performed using STATA version 10 (StataCorp, College Station, TX, USA).

Results

Socio-demographic data

The most important socio-demographic data are summarised in Table 1. The age of the women exposed to previous treatment ranged from 21 to 42 years, with a mean age of 32.9 years, which was significantly higher (P = 0.01) than the mean age of 31.0 years of the nonexposed population (range, 16–44 years). The majority of the women in the database were white European, but relatively more women were black, HIV seropositive and more highly educated (P = 0.03, 0.001 and 0.03, respectively) in the exposed versus the nonexposed group. The overall smoking status did not differ significantly between exposed and nonexposed women (P = 0.24), although there were more nonsmokers in the nonexposed population relative to the exposed group (P = 0.05).

Table 1.   Socio-demographic characteristics of the exposed versus nonexposed group
VariableExposed (treated for CIN) (N = 97)Nonexposed (no treatment) (N = 194) P
Mean (95% CI)RangeMean (95% CI)Range
  1. 95% CI, 95% confidence interval; CIN, cervical intraepithelial neoplasia.

  2. *Overall chi-squared.

Age (years) 32.9 (31.9–33.8)21–4231.0 (30.2–31.8)16–440.01
  n % n %  
Ethnicity     0.03*
White European6061.911358.20.56
Black1414.4136.70.03
Other2323.76835.10.05
HIV positivity 77.210.50.001
Smoking status     0.24*
Nonsmokers6466.014976.80.05
Ex-smokers99.384.10.08
Current smokers1919.63116.00.44
Unknown55.263.10.39

Treatment of CIN

Of the 97 treated women, 81 received an excisional treatment, eight an ablative treatment and eight an excisional treatment followed by ablation (mixed). The most frequently used excisional treatments were LLETZ (59.5%), followed by laser conisation (27.0%) (see Table 2). Almost all ablative treatments were performed by laser vaporisation (6/8, 75.0%). The excisional and mixed treatments resulted in reliable specimens for histology, with 78.6% (70/89) confirmed as CIN2/3 and 19.1% (17/89) as positive margins. More detailed histological outcomes can be found in Table 2. The mean depth of the removed cones was 14.5 mm (ranging from 2 to 30 mm; 95% CI, 12.9–16.1 mm). The depth of the cone was >10 mm in 63.5% (47/74) of the documented cases.

Table 2.   Type of treatment and histological results of the excised cones, performed in exposed women
  n %
  1. CIN, cervical intraepithelial neoplasia; CKC, cold knife conisation; LLETZ, large loop excision of the transformation zone.

  2. *Mixed treatments included (n = 8).

Type of CIN treatment (N = 97)
Excision* 8991.8
LLETZ5359.5
Laser2427.0
CKC89.0
Unknown44.5
Ablation (only) 88.2
Laser vaporisation675.0
Electrocoagulation112.5
Cryotherapy112.5
Histology of the cones (N = 89)
CIN11011.2
CIN22730.3
CIN34348.3
In situ adenocarcinoma11.1
In situ squamous carcinoma11.1
Unknown77.9

Obstetrical anamnesis, obstetrical outcomes, univariable and multivariable analysis

Neither gravidity nor parity differed between the exposed and nonexposed population (see Table 3). Gestational hypertension (10.3% versus 2.6%, P = 0.01) and premature labour (18.3% versus 7.7%, P = 0.02) were significantly more frequent in exposed women. Neither tocolytic drugs nor cerclage were used more often in exposed women than in the nonexposed group. There was no difference with regard to the rate of caesarean delivery, being 22.7% (22/97) in the exposed cohort and 23.2% (45/194) in the nonexposed population.

Table 3.   Obstetrical anamnesis in exposed versus nonexposed pregnant women
VariableExposed (treated for CIN)Nonexposed (no treatment) P
Median(Interquartile range)Median(Interquartile range)
  1. CIN, cervical intraepithelial neoplasia.

Gravidity 2(1–3)2(1–3)0.29
Parity 1(0–1)1(0–1)0.40
  n/N % n/N %  
Antepartum complications 27/9030.023/17713.00.0008
Risk of premature delivery
Previous preterm delivery10/7812.89/1516.00.07
Urogenital infection 1st trimester8/7810.37/1544.50.09
Diabetes6/797.619/15812.00.30
Gestational hypertension8/7810.34/1562.60.01
Intrauterine growth restriction5/796.36/1593.80.38
Management of pregnancy
Premature labour15/8218.312/1557.70.02
Tocolysis9/8111.112/1547.80.40
Cerclage3/803.81/1560.60.08

Duration of pregnancy, (spontaneous) PD

Exposed women delivered, on average, after 37.8 weeks (95% CI, 37.0–38.5) of gestation, compared with 39.0 weeks (95% CI, 38.7–39.2) in nonexposed women (P = 0.0002). The mean birthweight of the infants of exposed women was 3021 g (95% CI, 2869–3174), compared with 3298 g (95% CI, 3216–3381) of nonexposed women (P = 0.0003).

Data are presented on all deliveries and a subgroup of spontaneous deliveries only (Table 4A,B). The overall rate of PD after excisional treatment of CIN was 21.6% and the rate of spontaneous PD was 16.3%. There were no substantial differences in the ORs or hazard ratios (HRs) between the two groups and the spontaneous deliveries are further elaborated below.

Table 4.   Risk of adverse pregnancy outcomes and impact of cone depth in women treated by excision (all versus spontaneous deliveries only), LLETZ or laser conisation versus nontreated women. Adjusted odds ratios (ORs) were obtained by logistic regression, including ethnicity, HIV status, education, age, smoking and parity
VariableExposed (excisional treatment)Nonexposed (no treatment)Crude OR (95% CI)Adjusted OR (95% CI)
n % n %
  1. 95% CI, 95% confidence interval; LLETZ, large loop excision of the transformation zone.

(A) All deliveries
Excision (N = 88) (N = 176)   
Preterm delivery (<37 weeks)1921.6169.12.75 (1.34–5.67)2.82 (1.32–6.00)
Severe preterm delivery (<32 weeks)55.710.610.54 (1.21–91.67)12.93 (1.43–117.18)
Small for gestational age1011.42715.30.71 (0.33–1.54)0.76 (0.34–1.68)
Low birthweight (<2500 g)1719.3158.52.57 (1.22–5.43)2.76 (1.25–6.07)
LLETZ (N = 52) (N = 104)   
Preterm delivery (<37 weeks)1223.165.84.90 (1.72–13.96)5.63 (1.85–17.15)
Small for gestational age47.71615.40.46 (0.15–1.45)0.51 (0.15–1.70)
Low birthweight (<2500 g)1019.243.85.95 (1.77–20.05)6.99 (1.91–25.63)
Laser conisation (N = 24) (N = 48)   
Preterm delivery (<37 weeks)520.8510.42.26 (0.59–8.75)2.52 (0.62–10.25)
Small for gestational age28.3714.60.53 (0.10–2.79)0.65 (0.11–3.78)
Low birthweight (<2500 g)520.8612.51.84 (0.50–6.79)2.36 (0.57–9.80)
Cone depth ≤10 mm (N = 26) (N = 52)   
Preterm delivery (<37 weeks)311.547.71.57 (0.32–7.58)1.97 (0.30–12.71)
Small for gestational age27.7815.40.46 (0.09–2.33)0.53 (0.09–3.11)
Low birthweight (<2500 g)311.547.71.57 (0.32–7.58)1.65 (0.26–10.36)
Cone depth >10 mm (N = 47) (N = 94)   
Preterm delivery (<37 weeks)1225.577.44.26 (1.55–11.72)4.91 (1.67–14.50)
Small for gestational age714.91617.00.85 (0.32–2.24)0.94 (0.35–2.54)
Low birthweight (<2500 g)1021.355.34.81 (1.54–15.04)5.50 (1.62–18.64)
(B) Spontaneous deliveries only
Excision (N = 80) (N = 160)   
Spontaneous preterm delivery (<37 weeks)1316.3138.12.19 (0.97–4.99)2.33 (0.99–5.49)
Spontaneous severe preterm delivery (<32 weeks)33.800.0
Small for gestational age1012.52314.40.85 (0.38–1.89)0.94 (0.41–2.15)
Low birthweight (<2500 g)1215127.52.18 (0.93–5.09)2.34 (0.96–5.71)
LLETZ (N = 45) (N = 90)   
Spontaneous preterm delivery (<37 weeks)715.633.35.34 (1.31–21.77)6.77 (1.52–30.16)
Small for gestational age48.91314.40.58 (0.18–1.89)0.71 (0.20–2.49)
Low birthweight (<2500 g)613.322.26.77 (1.31–35.04)8.17 (1.42–46.91)
Laser conisation (N = 23) (N = 46)   
Spontaneous preterm delivery (<37 weeks)417.4510.91.73 (0.42–8.53)1.99 (0.45–8.67)
Small for gestational age28.76130.63 (0.12–3.42)0.72 (0.12–4.30)
Low birthweight (<2500 g)417.4510.91.73 (0.42–7.16)2.68 (0.53–13.49)
Cone depth ≤10 mm (N = 24) (N = 48)   
Spontaneous preterm delivery (<37 weeks)28.324.22.09 (0.28–15.84)2.77 (0.28–27.59)
Small for gestational age28.3714.60.53 (0.10–2.79)0.57 (0.09–3.46)
Low birthweight (<2500 g)28.336.31.36 (0.21–8.76)1.62 (0.20–12.83)
Cone depth >10 mm (N = 43) (N = 86)   
Spontaneous preterm delivery (<37 weeks)920.967.03.53 (1.17–10.69)4.55 (1.32–15.65)
Small for gestational age716.31315.11.09 (0.40–2.97)1.35 (0.47–3.88)
Low birthweight (<2500 g)716.333.55.38 (1.32–21.98)9.12 (1.69–49.31)

The frequency of spontaneous PD (>20 weeks, <37 weeks) in the exposed group (14.9%, all treatment types) was higher than in the group of nonexposed women (7.5%), with a corresponding crude OR of 2.18 (95% CI, 0.96–4.92). There were three cases of severe spontaneous preterm birth (<32 weeks) in the exposed population (3.4%) versus none in the nonexposed cohort. The occurrence of SGA was not significantly different between exposed and nonexposed women (crude OR, 0.74; 95% CI, 0.31–1.74). Twelve low-birthweight infants (<2500 g) were observed in the treated group (13.8%) and 12 in the untreated group (6.9%), corresponding to a crude OR of 2.16 (95% CI, 0.93–5.03).

The analyses were also performed separating excisional and ablative treatments. None of the treated women with an ablative technique delivered a baby with a birthweight below 2500 g or delivered spontaneously before 37 weeks of gestation. The risks of adverse obstetrical outcomes for women treated with an excisional procedure are shown in Table 4 (A, all deliveries; B, spontaneous deliveries only). Excisional treatment was always associated with a risk of adverse obstetrical outcomes, but no influence was seen on SGA. The risk was not lower when adjusted for confounding factors in the multivariable analysis: adjusted ORs of 2.33 (95% CI, 0.99–5.49), 0.94 (95% CI, 0.41–2.15) and 2.34 (95% CI, 0.96–5.71) for spontaneous PD, SGA and low-birthweight infants, respectively. When only focusing on treatments with LLETZ or laser conisation, the ORs for spontaneous PD and low birthweight were 6.77 (95% CI, 1.52–30.16) and 8.17 (95% CI, 1.42–46.91) for LLETZ, and 1.99 (95% CI, 0.45–8.67) and 2.68 (95% CI, 0.53–13.49) for laser conisation. Again, no negative effect was shown for these treatment types on SGA (OR, 0.71; 95% CI, 0.20–2.49 and OR, 0.72; 95% CI, 0.12–4.30 for LLETZ and laser conisation, respectively).

The rate of spontaneous PD for women with a cone smaller or equal to 10 mm in depth was 8.3% (2/24), whereas, in the deeper cones (>10 mm), the rate was 20.9% (9/43). The adjusted OR for spontaneous PD associated with less deep excision (≤10 mm) versus no treatment was 2.77 (95% CI, 0.28–27.59), whereas the adjusted OR associated with deeper treatment (>10 mm) was 4.55 (95% CI, 1.32–15.65). The same was seen for low birthweight: 8.3% (2/24) of women with small cones, compared with 16.3% (7/43) of women with deeper cones, had a low birthweight infant in a subsequent pregnancy after an excisional treatment. The respective adjusted ORs were 1.62 (95% CI, 0.20–12.83) and 9.12 (95% CI, 1.69–49.31) compared with the nonexposed population.

The Kaplan–Meier curve (Figure 1) and lifetable (Table 5) show the proportion of births by gestational age for exposed and nonexposed women. Exposed women delivered earlier than the nonexposed group, with 20.4% and 6.5% of cumulative deliveries before 37 weeks of gestation, respectively.

Figure 1.

 Kaplan–Meier graph comparing the cumulative proportion of births over gestational age for exposed (excisional treatments only) and nonexposed women. Only spontaneous births were taken into account.

Table 5.   Lifetable of the cumulative proportion of births over gestational age (GA) (weeks) for exposed (excisional treatments only) and nonexposed women. Only spontaneous births were taken into account
Interval GA (weeks)Beg. totalBirths exposedBirths nonexposedCumulative proportion (%)95% CI (%)
2627126100.620.094.30
2728161101.230.314.85
3132160101.850.605.63
3233159011.850.605.63
3334158213.101.307.28
3435155113.731.698.10
3536153023.731.698.10
3637151526.933.9012.17
37381445510.226.3916.15
383913471215.1310.3121.91
3940115123025.3218.7833.61
404173195456.1645.1467.79
41420100.00  

The Cox proportional hazards regression, clustered on the triplet, confirmed the findings of the logistic regression. Excisional treatments were clearly associated with spontaneous delivery at an earlier gestational age (HR, 1.32; 95% CI, 1.05–1.67); this was also observed for LLETZ (HR, 1.33; 95% CI, 1.00–1.77), and showed the same trend for laser conisation (HR, 1.28; 95% CI, 0.73–2.24). Large cones (>10 mm) were also clearly associated with an increased risk of spontaneous delivery at an earlier gestational age in the Cox regression, with an HR of 1.45 (95% CI, 1.11–1.89), whereas small cones (≤10 mm) were not (HR, 1.11; 95% CI, 0.52–2.37). In multiparous women, adjusted for previous PD, the association between excisional treatment, in particular large cones, and spontaneous PD was confirmed. However, the HRs were only marginally significant (HR, 1.21; 95% CI, 0.94–1.57 for excisional treatment and HR, 1.38; 95% CI, 1.05–1.82 for large cones).

Discussion

In our study, the incidence of (spontaneous) PD was more than twofold higher in women delivering after treatment of CIN than in women without any treatment for CIN or CIN history. However, it has been described in the literature that excisional treatments, in particular, result in adverse obstetrical outcomes, whereas ablative treatments do not (except for radical diathermia).9,10,14 Our study showed that the incidence of spontaneous PD was more than twofold higher in women with excisional treatment compared with the nonexposed group. The OR of 2.33 (delivery before 37 weeks of gestation) is in line with two meta-analyses,9,10 and with more recent articles discussing adverse obstetrical outcomes after treatment of CIN.15–17 The number of ablative treatments was limited in our study, and so no conclusions can be drawn. However, meta-analytical findings have revealed that women treated by ablation are not exposed to adverse obstetrical outcomes.

In our study, large cones (>10 mm in depth) showed a significantly higher risk of spontaneous PD (OR, 4.55) compared with the untreated population, whereas smaller cones (≤10 mm) were not significantly related to adverse outcomes. The preponderance of deep cones (the depth of the cone was >10 mm in 63.5% of the documented cases) in this study can be explained by a tendency to prioritise oncological safety rather than potential obstetrical adverse outcomes by Belgian gynaecologists. After all, the main objective of conisation remains the total excision of the transformation zone with the aim of negative margins. Four studies18–21 have stratified the relative risk of PD in pregnant women with previous excisional treatment compared with nontreated pregnant women according to the depth of the cone (≤10 mm and >10 mm, respectively). In all four studies, the risk was consistently and significantly higher when the depth was >10 mm: the pooled relative risk (RR) was 2.39 (95% CI, 1.55–6.69). When the depth of excision was ≤10 mm, no increased risk of PD was observed, except in one study (pooled RR, 1.32; 95% CI, 0.58–2.95). A recent study by Ang et al.11 has suggested that cones smaller than 10 mm in depth should be the aim in women of reproductive age (≤35 years) to avoid potential adverse outcomes in future pregnancies, and this without the fear of increasing the risk of recurrent disease in this group of women, as treatment with larger cones did not lower the recurrence rate. This can only be advised under circumstances in which the whole lesion is clearly visible on the ectocervix and lower endocervical canal, there is no suggestion of invasive disease and the woman has not received previous treatment to the cervix that may have altered the topography of the transformation zone.22

In this article, we also evaluated SGA. In most studies related to obstetrical outcomes after CIN, only PD and low birthweight are addressed. However, low birthweight is defined as an infant with a birthweight of <2500 g, regardless of gestational age at the time of birth, whereas SGA infants are those who are smaller in size than normal for the infant’s sex, gestational age and parity of the mother. SGA, reflecting fetal growth restriction, is defined as a weight below the 10th percentile for the gestational age. Our study showed that previous treatment of CIN does not affect intrauterine growth, and that low birthweight is only the consequence of PD.

We assessed all deliveries and the subgroup of spontaneous deliveries. The latter subgroup is more appropriate to address treatment effects as iatrogenic preterm deliveries are removed. This subgroup should preferentially be used in future research. The results in the subgroup of spontaneous deliveries were comparable with those for all deliveries, but somewhat less pronounced. In order not to exclude women with preterm prelabour rupture of membranes (PPROM), data on all deliveries were also shown, as data on PPROM were not available in our database. However, PPROM is a possible important effect of CIN treatment (as a result of chorioamnionitis and other pathophysiological changes), and these women are often induced and deliver preterm.19 Restriction to spontaneous preterm deliveries only would have excluded this important population.

The majority of women in this study were multiparous, thus allowing a previous PD to be taken into account as a confounding factor. However, in this dataset, it could not be distinguished whether earlier pregnancies happened before or after the treatment of CIN (the dates of earlier pregnancies were not available). Therefore, previous PD could not be controlled for fully in our study, which is a clear limitation in the assessment of possible confounding.

Another limitation of our study may be the comparison group used: women in the general population without a history of treatment for CIN. Other factors associated with CIN could cause the adverse obstetrical effects observed rather than the treatment itself. The fact that the adjusted ORs were not lower than the unadjusted relative ORs does not corroborate this hypothesis; however, residual confounding by unrecorded or unknown factors cannot be excluded. For instance, periodontal disease has been shown to be associated with an increased incidence of spontaneous preterm birth, an effect which can be controlled for by adequate dental care.23 Other recent studies have compared adverse pregnancy outcomes in treated versus untreated women with CIN,24,25 and have found no increased risk of adverse pregnancy outcomes associated with treatment. Indeed, these data show that the occurrence of PD and low birthweight may be caused by factors associated with the disease rather than with treatment. However, here, an opposite bias can be suspected. Untreated women with CIN, in particular if CIN3, may belong to a category of patients that ignores treatment advice and may share risk factors for adverse obstetrical outcomes. Another argument that corroborates the hypothesis of an association between treatment and adverse obstetrical outcomes is the dose–effect relationship observed in our study, as well as in other published studies.26 Only a few studies have documented this dose–effect relationship together with other potentially confounding data. Pooled analyses of individual records from databases of good quality and a high level of completeness may be helpful to provide better estimates of the obstetrical and oncological risks associated with treatment and patient characteristics.27,28

Conclusions

Our data support the conclusion that women with a history of excisional treatment of CIN have an increased risk of (spontaneous) PD relative to women without such a history. This suggests that the use of excisional treatment should be limited to high-grade lesions (CIN2+), and the use of surgery should be avoided for CIN1 lesions. Large cones (>10 mm in depth) are associated with an increased risk of adverse obstetrical outcomes. If smaller cones should be proven to be oncologically safe, the depth of the cones should be limited in young reproductive women whenever possible.

Disclosure of interests

There are no conflicts of interest to declare.

Contribution to authorship

The respective contributions of the authors were as follows: conception and design (FG, PS, JMF, MA); acquisition of the data (FG, PS, PB); statistical analysis and interpretation of the data (CS, MA, JA); drafting of the manuscript (CS, MA, FG); critical revision and editing of the manuscript (PS, PB, JA, JMF).

Details of ethics approval

As this study was observational (not a single intervention was performed during this study), no ethical committee approval was required.

Funding

Financial support was received from: FNRS (le Fonds National de la Recherche Scientifique), through TELEVIE, Brussels, Belgium (ref 7.4.628.07.F) (FG is ‘MD postdoctoral fellow’ of the FRS-FNRS); the Belgian Foundation Against Cancer, Brussels, Belgium; and the European Commission (Directorate of SANCO, Luxembourg, Grand-Duchy of Luxembourg), through the ECCG project (European Cooperation on Development and Implementation of Cancer Screening and Prevention Guidelines, International Agency for Research on Cancer, Lyon, France) and the 7th Framework Programme of DG Research of the European Commission through the PREHDICT project (grant no. 242061, coordinated by the Vrije Universiteit Amsterdam, the Netherlands).

Acknowledgements

We would like to thank Colette Gerday for management of the study, and all study nurses and participating gynaecologists for data collection.

Ancillary