Maternal History of Cervical Surgery and Preterm Delivery: A Retrospective Cohort Study

Introduction

Preterm birth occurs in about 10% of pregnancies in the United States and affects ∼15 million pregnancies worldwide. ¹ Although the etiology of preterm delivery is multifactorial, cervical conization and/or loop electrosurgical excision procedure (LEEP) have been cited as risk factors for preterm delivery in prior studies. ^{2 –8}

In the United States, human papillomavirus (HPV) is the primary cause of cervical dysplasia and the most common sexually transmitted infection, with about 40% of women (aged 18–59) being infected with any genital HPV. ⁹ Maternal HPV infection in the pre- and early pregnancy period has been associated with adverse pregnancy outcomes, such as pregnancy-induced hypertension, preterm delivery, and spontaneous abortion. ^{10 –14} Whether the higher rate of preterm birth among women with a history of cervical conization or LEEP is due to the procedure or HPV infection remains unclear.

The primary aim of our study was to investigate whether maternal history of cervical conization and/or LEEP increases the risk of preterm delivery independently of maternal HPV status. Secondarily, our study aimed to determine whether maternal infection with HPV modified the effect of cervical conization and/or LEEP on preterm delivery.

Materials and Methods

We conducted a retrospective cohort study of most recent singleton deliveries, including all modes of delivery that occurred at the University of Virginia Hospital between 2010 and 2015. Ethical approval was obtained from the Institutional Review Board for Health Sciences Research at the University of Virginia with a waiver of consent. Women aged 16–49 years were included. All modes of delivery were included to account for comorbidities relating to preterm delivery that would precipitate a medical induction of labor or cesarean delivery.

All variables, including cervical cytology and/or HPV genotyping results, were abstracted from electronic medical records by a database administrator that was not involved in data analysis. The primary exposure of interest in this study was maternal history of cervical conization, including both cold knife and laser conization, and/or LEEP. The primary outcome of interest in this study was preterm delivery, which was defined as delivery between 24- and 37-weeks' gestation. Gestational age was based on most recent ultrasound dating for women who received prenatal care at our institution, or based on self-report for women who did not receive prenatal care at our institution.

Pre- or early pregnancy maternal infection with HPV was assessed as an effect modifier in this study. HPV infection was based on the patient's most recent cervical cytology and/or HPV genotyping results within 3 years before delivery. Cervical cytology and genotyping samples were obtained using a ThinPrep liquid-based Pap test. Women were considered HPV positive if their most recent HPV genotyping results were positive for high risk types (types 16, 18, 31, 33, 35, 39, 45, 51, 52, 56, 58, 59, and 68). For women without HPV genotyping results, women were considered positive if their most recent two consecutive cervical cytology results had high- or low-grade squamous intraepithelial lesions (HSILs or LSILs), atypical squamous cells cannot rule out high-grade lesions (ASC-H), or atypical squamous cells of undetermined significance (ASCUS). Notably, no patients were included who had ASCUS without accompanying results for HPV testing. Remaining patients were considered HPV negative if their last two consecutive cervical cytology results were both negative for HSIL, LSIL, ASC-H, and ASCUS. Women were also considered negative if their most recent HPV genotyping results were negative combined with at least one negative cervical cytology result. This diagnostic criteria was intended to optimize sensitivity, which was reported to be 94.6% for HPV DNA testing and 55.4% for single cervical cytology testing. ¹⁵ In addition, the American Society for Colposcopy and Cervical Pathology's most recent screening guidelines for the prevention and early detection of cervical cancer included cervical cytology testing every 3 years in low-risk women aged 21–65 years. ¹⁶

Maternal sociodemographic and insurance information as well as smoking and drug use during pregnancy were self-reported during regular prenatal visits. Maternal body mass index (BMI) was calculated based on prenatal height and weight at the time of delivery and categorized into normal/underweight (BMI ≤24.9), overweight (BMI 25–29.9), obese (BMI 30–39.9), and morbidly obese (BMI ≥40). All other maternal medical history, pregnancy diagnoses and outcomes, and neonatal outcomes were based on ICD-9 diagnoses for the year of delivery.

Potential risk factors, including maternal history of cervical conization and/or LEEP, demographic information, HPV infection, and medical and obstetric history, were compared between women with and without preterm delivery using univariate logistic regression. A Mantel–Haenszel test for homogeneity of odds ratios (ORs) was used to assess whether maternal HPV infection modified the effect of cervical conization and/or LEEP on preterm delivery during univariate analysis. Potential risk factors that displayed a potential association with both the primary outcome and cervical conization and/or LEEP were considered potential confounders and adjusted for in multivariable analysis. Based on this assessment, ultimately pregnancy-related hypertension, preterm premature rupture of membranes (PPROM), placenta previa, bacterial vaginosis infection, history of preterm delivery, ethnicity, maternal BMI at delivery, parity, and smoking or drug use during pregnancy were adjusted for in multivariable analysis.

Logistic regression was used for multivariable analysis. Multicollinearity was assessed during intermediate multivariable models. Likelihood ratio tests were used to determine a p-value for dependent variables with more than two categories. An interaction test was used to assess whether maternal HPV infection modified the effect of maternal history of cervical conization and/or LEEP on preterm delivery. Adjusted OR and 95% confidence intervals (CIs) were also reported stratified by maternal HPV status.

A power analysis revealed that our study required 1050 participants based on a sample size calculation assuming a 10.0% and 16.0% prevalence of preterm delivery women with and without a history of cervical conization and/or LEEP, respectively, an alpha of 0.05, and beta of 0.20. All statistical analyses were conducted using STATA 14.1 (College Station, TX).

Results

Of the 3933 women who delivered at our institution who met inclusion/exclusion criteria and had complete data for the multivariable analysis, 19.8% (n = 792) delivered prematurely. Of these women, 9.1% (n = 362) had a history of cervical surgery, including cervical conization and/or LEEP. We identified 2153 of the 3933 women included that also had available cervical cytology and/or HPV genotyping results within 3 years before delivery for the purposes of our secondary objective. Among included women, 1324 (61.5%) and 829 (39.5%) were HPV negative and positive, respectively, within the 3 years before delivery.

The majority of women in our cohort were white, non-Hispanic, overweight or obese at delivery, multiparous, and had Medicaid (Table 1). The mean maternal age was 28.8 (standard deviation ±5.2). The overall prevalence of smoking or drug use was 11.2% (n = 242). Less than 1% (n = 35) of women reported a history of preterm delivery, and 4.9% (n = 257) of women had PPROM.

Cervical conization and/or LEEP was associated with preterm delivery in univariate analysis (Table 1, p < 0.001). Similarly, maternal infection with HPV was associated with preterm delivery (p = 0.04). However, cervical conization and/or LEEP was not associated with maternal HPV perhaps indicating that the procedure removed all dysplastic tissue that would otherwise be found during a Pap smear (p < 0.05). During univariate analysis, pregnancy-induced hypertension, PPROM, bacterial vaginosis infection, placenta previa, placental abruption, ethnicity, maternal BMI, multiparity, and smoking or drug use during pregnancy were also associated with preterm delivery as expected (p < 0.05), whereas a history of preterm delivery had a p-value of 0.06.

Table 2 shows that cervical conization and/or LEEP remained associated (OR = 1.75, 95% CI: 1.31–2.33) with preterm delivery when adjusting for other covariates (p < 0.001). Other variables in the model that displayed an association with preterm delivery included pregnancy-induced hypertension, PPROM, placenta previa, bacterial vaginosis infection, ethnicity, maternal BMI, and nulliparity (p < 0.05).

Maternal HPV infection was not included in the final multivariable model because of current infection (based on cervical cytology and/or HPV genotyping) and was not associated with cervical conization and/or LEEP (p = 0.99, p = 0.95, respectively). However, when maternal HPV status was added into the final model, cervical conization and/or LEEP remained associated (OR = 2.08, 95% CI: 1.22–3.57) with preterm delivery (p = 0.01). Maternal HPV infection also appeared associated (OR = 1.70, 95% CI: 1.03–2.78) with preterm delivery, independent of other covariates. Maternal infection with HPV was not observed to modify the effect of cervical conization and/or LEEP on preterm delivery (Test for homogeneity of ORs: p = 0.95). After adjusting for other confounders, maternal infection with HPV did not modify the effect of cervical conization and/or LEEP (likelihood ratio test: p = 0.91) during multivariable analysis. Table 3 displays the crude and adjusted association between cervical conization and/or LEEP and preterm delivery, stratified by maternal HPV status.

Discussion

Based on our data, we conclude that maternal history of cervical conization and/or LEEP increases the risk of preterm delivery independent and irrespective of maternal HPV status. In our study, maternal HPV status based on most recent Pap smear results was not associated with a history of cervical conization or LEEP. This finding may be due to the removal of dysplastic or infected tissue from the cervix in women with a prior history of cervical surgery, thereby resulting in a negative Pap smear.

In previous studies, the association between cervical conization and/or LEEP and preterm delivery has been controversial. ⁶ In 1993, a Swedish study reported that women with a history of cervical conization had an increased risk of preterm birth, whether conization preceded or followed delivery. ² Since then, another published study has also concluded that cervical conization increases the risk for preterm delivery. ³ However, Kristensen et al. concluded that the observed association between cervical conization and preterm delivery may result from other factors. ² Studies examining the association between LEEP and preterm delivery have reported similar findings. ^4,5,8,17 More recent investigations have not determined whether the increased risk of preterm delivery results from the procedure or other confounding variables, such as sexually transmitted infections. ⁶ For example, HPV infection has been associated with preterm delivery in prior studies that did not account for cervical conization or LEEP. ^11,12

The mechanism of action for how cervical surgery increases the risk of preterm delivery is not well established. Previous investigators have hypothesized that removal of healthy tissue during excisional procedures may alter cervical functioning and possibly explain the observed increased risk of preterm birth. ¹⁸ Increasing cone depth was found to be an independent risk factor for preterm birth, further supporting this hypothesis. ^18,19 Another hypothesis is that excisional procedures of the cervix could lessen mechanical support. ²⁰ Cervical excisional procedures have also been hypothesized to allow microbial pathogens to ascend through the cervix into the uterine cavity, leading to an inflammatory response that stimulates the release of prostaglandins with subsequent cervical ripening and uterine contractions. ²⁰

Strengths of our study include a large sample size at a hospital that serves a diverse patient population in a large catchment area. In addition, we based a diagnosis of HPV on cervical cytology and HPV genotyping results based on the most recent guidelines of the American Cancer Society. ^15,16 However, the retrospective study design at a single center introduces certain limitations. For example, this study design limited the ability to test women coming in for labor—thus, the study can only refer to HPV positivity within the last 3 years of pregnancy for most study participants rather than known status at the time of pregnancy and delivery. Another limitation of our study was lack of clinical detail in some of the records and the reliance on patient recall for pertinent medical and surgical history for patients referred into our system after their initial prenatal visits. For example, details regarding size of cervical cone resection was not consistently available and included as a variable in our analysis, but could be an area of further research. Our study included women with any mode of delivery for the purposes of accounting for deliveries that would have resulted in a preterm delivery but had an intervention such as induction of labor or cesarean delivery. However, this may miscode certain deliveries. Notably, our study site had a higher preterm birth rate than the national average. However, as displayed in Table 1 of our study, there is a high frequency of risk factors for preterm birth within the study site's patient population.

We found that cervical surgery increases the risk of preterm birth independent of current HPV infection. In addition, we found that current HPV infection based on the most recent Pap smear was not associated with a history of cervical surgery. However, the role of cervical length measurements in these women is not clear and the Society for Maternal-Fetal Medicine does not currently recommend serial cervical length sonograms in women with a history of cervical surgery. ²¹ Similarly, the role of cervical cerclage to prevent preterm birth in these women is controversial. ²² Prospective randomized trials of serial cervical length sonograms with or without cerclage placement are needed to determine if these measures may improve pregnancy outcomes and decrease the risk of preterm birth.