Transvaginal Cervical Length and Tobacco Use in Appalachian Women: Association with Increased Risk for Spontaneous Preterm Birth

Abstract

Introduction

The rate of preterm births has risen in most industrialized countries, increasing from 9.5% in 1981 to 12.7% in 2005 in the United States.¹ Preterm births account for 75% of perinatal mortality and more than half the long term morbidity.² Births following spontaneous preterm labor and preterm premature rupture of membranes (PPROM) are grouped together as Spontaneous Preterm Births (SPB). Preterm labor is thought to be a syndrome initiated by complex mechanisms.³ Infection and tobacco exposure play an important role in PPROM leading to SPB.⁴ West Virginia ranks first in the nation for women who smoke while pregnant. According to West Virginia Vital Statistics data (2013), 26.3% of women reported they smoked during pregnancy, which is double the national rate of 13%.⁵ Furthermore, 38.5% of women enrolled in Medicaid use tobacco and 49.3% of all births are financed by Medicaid in West Virginia.⁶ Maternal demographic characteristics also play a role, with low socioeconomic and educational status, low and high maternal age, marital status and body mass index being associated with preterm birth.^7-9

Identification of patients at risk may allow clinical interventions. Previous studies suggest cervical ripening as part of the common final pathway of human partutition.^9,10 Investigators have therefore evaluated ultrasonographic examination of the cervix for the prediction of SPB.^11-21 While there are several diagnostic methods used to predict SPB, sonographic transvaginal cervical length (CL) measurements have proven to be the most reliable.^22,23 Risk of preterm delivery has been found to be inversely correlated with CL.^13,19 A previous study by Moroz, et al 2012 found no added effects of tobacco abuse on cervical length and preterm birth.²⁴ ACOG’s previous recommendations, however, did not advise routine screening for short cervix in low risk population, citing lack of proven treatment or interventional methods.²⁵ ACOG’s most recent publication on preterm birth now recognizes the possible use of a screening strategy to identify cervical length in the management and prevention of preterm labor due to new therapies for preterm delivery prevention.²⁶ Vaginal progesterone is recommended for women with a cervical length ≤20 mm. Cerclage is recommended for women with prior spontaneous preterm birth who are already receiving progesterone supplementation and CL is <25 mm. Because of our high incidence of tobacco use among pregnant patients, we decided to study tobacco use as an associated increased risk for preterm birth with cervical shortening. Since our study took place prior to these new ACOG recommendations, we had been routinely recording ultrasound (US) parameters including cervical length in an US database to identify patients who were candidates for the only intervention at the time, cervical cerclage.

Methods

WVU/CAMC Institutional Review Board reviewed and approved this research study. A HIPPA waiver was obtained prior to the study and no consents were deemed necessary. This is a retrospective cohort study analyzing consecutive pregnant women from January 2007 to June 2008 who received routine measurements of CL by transvaginal US between 15-26 weeks gestation at Charleston Area Medical Center (CAMC) Women and Children’s Hospital, a tertiary medical center outpatient clinic. Patients who missed US appointments, had no CL measurements recorded, or were transferred into the clinic late into pregnancy were not included in analysis. In addition, patients who did not give birth at CAMC Women and Children’s Hospital were excluded due to lack of birth outcome data. A total of 1116 patients met inclusion/exclusion criteria. Primary outcome for this study was spontaneous preterm birth before 35 weeks gestational age, defined as birth that resulted directly from spontaneous preterm labor.

Transvaginal ultrasound of the cervix was performed using a GE Volusan E8 USG machine with multi frequency vaginal probe (5 - 8 MHz) and Aloka Prosound alpha – 10 with vaginal probe (1 – 15 MHz). CLs were measured three times in standard fashion.²⁷ With an empty maternal bladder, the transvaginal probe was placed in the interior fornix of the vagina. A faint line of echodensity between internal and external os was measured while avoiding pressure on the cervix that might increase its apparent length. The average measurement was recorded in the US database that contained all data from ultrasounds performed. Each ultrasound image was individually reviewed by the authors (JF or BCC), and resulted in 42 patients excluded through the quality control process. Thus, 1074 patients were included in this study.

Each patient’s medical record, data from the US database, and birth certificate record were reviewed to collect variables found in previous studies to have an association with SPB <35 weeks (low socioeconomic status, hypertension, depression, thyroid dysfunction, polyhydramnios, BMI, sexually transmitted infection, smoking, previous cervical surgery, drug abuse, maternal age <18 or >45, previous abortions, diabetes mellitus, placenta previa).²⁸ Maternal height and weight from first antenatal visit were recorded to calculate BMI. Data were analyzed using the statistical package SPSS (version 17; Chicago, IL). “Short” and “normal” CL groups were defined using CL threshold values determined by a Receiver Operator Characteristic (ROC) curve (Fig. 1). On this curve, we calculated the Youden index (J) ²⁹ using the formula (J = maximum sensitivity + specificity −1) to identify the optimal threshold value. The Youden index assumes that sensitivity and specificity are equally important.³⁰ J is defined as the maximum vertical distance between the ROC curve and the diagonal line or the point on the curve farthest from chance.³¹ We investigated other reported cut off values, 20, 25 and 30mm. Rates of SPB and variables associated with SPB from above were compared using Pearson’s Chi Square, Student’s T test, or Fisher’s Exact test when appropriate to determine if there was a significant difference between short and long CL. Finally, stepwise backward logistic regression was utilized to evaluate these variables, as well as second trimester CL, as predictors of SPB.

Open in a separate window

Figure 1

ROC curve cut-off values for SPB (Area Under the Curve 76.6) p<0.001; cutoff value = 34mm.

Results

Of the 1,074 patients studied, 48.7% (523/1074) delivered by spontaneous vaginal delivery and 11.6% (127/1074) delivered vaginally following induction of labor. The remaining (39.5% (424/1074)) were delivered by cesarean section. The average gestation age at the time of CL measurement was 19.0 ± 2.0 weeks. Three percent (3.0%) of our sample (32/1074) delivered by spontaneous vaginal delivery at less than 35 weeks gestational age. Mean second trimester CL by transvaginal ultrasound for our sample was 41 mm, ranging from 12 mm - 70 mm. The median CL measurement was 40 mm and the 25^th, 50^th and 75^th percentile values were 35 mm, 40 mm and 45 mm, respectively.

A ROC curve was created using our patients’ CL data, resulting in a cut-off value of 34 mm (Area Under the Curve (AUC) 76.6, p<0.001) (Figure 1). This cut-off was used to create the short CL group (≤34 mm) resulting in 224 patients (20.9%) and the normal CL group (>34 mm) resulting in 850 patients (79.1%). Examining the occurrence of SPB and short vs. normal CL, SPB occurred in 1.4% (12/850) of patients with a CL >34 mm, while 8.9% (20/224) of the patient’s with CLs ≤ 34 mm experienced SPB (p<0.001) (Figure 2). Rate of SPB was calculated for lower cut-off values and is shown in Table 1. As illustrated in Table 2, the >34mm cut-off provided the highest accuracy, while producing a sensitivity of 62.5% with a drop off in specificity to 80.4%. We found with BMI≤30, there were significantly more cases of SPB (4.9%) than when BMI>30 (1.8%); p=0.005. For heavy smoking (>10 cigarettes, we found significantly more SPB (5.1%) compared to (2.2%) for light/nonsmokers; p=0.022. Two had undergone an intervention to prevent SPB, cervical cerclage of which one resulted in SPB.

Open in a separate window

Figure 2

Spontaneous Preterm Delivery by Cervical Length using ROC cut-off value.

Variables (shown in Table 3) were evaluated using univariate analyses to determine significant differences for each variable between the normal (>34 mm) and shortened (≤ 34 mm) CL groups. Significant differences were found between the two groups (normal vs. shortened CL) for maternal age (25.6 vs. 24.7 years) and BMI (33.9 vs. 31.3).

Stepwise backward logistic regression was then utilized to evaluate each of the variables found to be significant through univariate analysis as independent predictors of SPB. Body Mass Index (BMI), CL, and tobacco usage were analyzed as categorical variables rather than continuous variables. While 36 percent of patients used tobacco, those who smoked ≥ 10 cigarettes per day, exhibited 2.3 times the risk of SPB (95% CI 1.08-4.78, p=0.030) than those who smoked < 10 cigarettes per day. Similarly, BMI ≤30) doubled the risk for SPB (95% CI 1.07-4.77, p=0.034). Those with CLs ≤30 mm were 12.0 times more at risk for SPB (95% CI 4.86-29.42, p<0.001) than those with CLs >34 mm. Women with CLs between 30 mm and 34 mm were 3.6 times more at risk for SPB (95% CI 1.48-8.84, p=0.005) (Table 4) than those with CLs >34 mm.

Further analyses explored the additive effect of shortened CL and heavy smoking status (>10 cigarettes per day) on the risk of SPB. In Figure 3, the effects of heavy smoking groups are illustrated on the left side of the graph and the low/non-smoking group on the right. Shortened CL and low BMI yielded a SPB rate of 5.1% in those with low and nonsmoking status, while heavy smoking status increased this rate to 12.3% in the presence of these other risk factors. The data suggest that heavy smoking status doubles, and almost triples, the risk of SPB in patients with shortened CL and normal to low BMI (<30) respectively.

Open in a separate window

Figure 3

Percentage of SBP by Risk Factors

S=smoker, NS=nonsmoker, N=normal cervix, s=short cervix, L=low BMI, H=high BMI,

At a cut-off level for cervical length of 3.4 cm in our study with 1,074 scans, we would identify 20 possibly preventable preterm births, treat 11 patients for each preterm birth prevented, and screen 53 patients to identify one patient to treat for preterm birth. Using cost analysis based on the 53 screening transvaginal scans (cost of approximately $250 per scan based on ICD9 RVU’s) and 11 patients treated with vaginal progesterone (approximately $2,500 per patient for 21 weeks-www.drugstore.com 2011), the approximate cost to prevent one preterm birth would be about $40,750 [(53 scans × $250) + (11 × $2,500)]. However, using cost data on PTB, we find that a very low-birth (VLBW) weight neonate (< 28 weeks or 1,500 gms) costs 28 times that a term delivery costs ($280,146 versus $9,803)^42-45 and a low-birth weight neonate (28-34 weeks or < 2,500 gms) costs 4 times what a term neonate costs ($38,367 versus $9,393)^42-45 In our analysis, by preventing only 1 very low birth weight neonate (< 28 weeks or <1,500 gms) in our care pathway, we save over $239,000 in hospital costs alone, and, by preventing only 1 low birth weight neonate (28-34 weeks or <2,500 gms), we essentially are at a breakeven point for cost of screening with therapy. Of the 220 with shortened cervix, 92 (41.1 %), smoked and 9 had SPB. Integrating the biofeedback of cervical length into the current tobacco risk reduction counseling in the clinic would not involve any additional cost and has the potential of reducing 9 SPBs to 4 if the intervention is 44% effective, resulting in a savings of 4 × $239,000, which is close to a million dollars over an 18 month time period.

Discussion

For our study, the cutoff value was determined to be 34 mm, indicating increased risk for SPB in a population previously thought to have normal cervices. Through logistic regression, we were able to identify shortened CL as the single most powerful predictor of preterm birth, a finding that is in agreement with previous studies.^27-28,31 Additionally, we found that smoking status was closely linked to SPB in our sample as it has been widely recognized as a significant contributor to SPB.^24,32-35 It is thought that smoking may induce preterm labor by increasing the concentration of inflammatory mediators in the amniotic fluid, such as platelet activating factor.³⁴ As previously mentioned, 36 percent of patients used tobacco, and those who smoked ≥ 10 cigarettes per day exhibited 2.19 times the risk of SPB compared to those who smoked < 10 cigarettes per day. Smoking in our study, had an additive effect in patients with a shortened cervix by transvaginal ultrasound, further increasing the risk of SPB in these individuals. A previous study examined cervical shortening and tobacco use associations with spontaneous preterm birth without evidence of effect by tobacco.²⁴ However, to our knowledge, our study is the first to describe smoking’s additive effect on the risk of SPB in patients with shortened cervices.

Studies investigating the utility of transvaginal CL to predict preterm labor and birth have recommended using much lower cutoff values to avoid unnecessary interventions.¹⁹ Our study sample had an average CL of 40 mm while previous studies have found an average CL of 35-38 mm when measured in the second trimester,^12,20 There are a few potential explanations for this finding. First, our “longer” average CL may be due to the aforementioned relative racial homogeneity of our sample, as one previous dissertation study has suggested that Caucasian patients have longer CLs when compared to African-American patients. ³⁶ Anum et al., 2010 reported that Black patients in their study population exhibited twice the risk of cervical insufficiency compared to White patients when controlling for all other confounding variables.³⁷ Another potential reason for our study’s “longer” CL measurements is the method by which CL imaging was interpreted. Though individual ultrasonographic CL images were obtained in a fashion similar to previous studies, the CLs that we recorded in our data set were the average of the measurements from the clearest image for each of our patients. This differs from many previous studies in that they typically recorded the shortest CL obtained from a clear sonographic image for each patient.¹²

Identification of patients with tobacco abuse, as in our study, may lead to possible therapeutic interventions for shortened cervix. In 2011, Hassan et al. showed that asymptomatic women with a sonographic short cervix (10-20 mm) identified in the second trimester who were treated with vaginally administered micronized progesterone gel, experienced a 45% reduction in the rate of preterm birth before 33 weeks gestational age when compared to those treated with placebo.³⁹ Hassan’s, et al 2011 study provided not only an effective treatment for those with a sonographically identified short cervix, but it also strengthened the argument supporting the use of routine CL screening by transvaginal ultrasound in the second trimester. Such studies^39-41 led to the more recent ACOG Practice Bulletin in 2012 recognizing the utility of routine second trimester cervical length evaluation even in low risk populations.²⁶

In the results section, we mentioned that in preventing only 1 very low birth weight neonate (< 28 weeks or <1,500 gms) in our care pathway, we save over $239,000 in hospital costs alone, and, by preventing only 1 low birth weight neonate (28-34 weeks or <2,500 gms), we essentially are at a breakeven point for cost of screening with therapy. However, the progesterone treatment is only 45% effective. Thus, if treated, most likely you could prevent preterm birth in half of these cases. Therefore, in the context of our high tobacco use population, routine cervical length screening may not make good financial sense. However, it makes clinical and financial sense to use shortened cervical length status (even at the cut-off value of 3.4 mm) in educating women who smoke of their increased risk of preterm birth. Getting women to reduce or quit smoking is difficult. By screening our patients for cervical length, we will be able to target a specific health message of doubling risk for preterm birth with both a short cervix and practice of smoking.

In the local context of smoking families, often our patient population is surrounded by examples of friends and family members who have smoked during pregnancy and experienced no preterm birth. By targeting the patients with both risk factors, short cervix and smoking, we can provide the biomedical feedback of the shortened cervix so that the patients can better understand their increased risk. This targeted education would show the comparison of their increased risk to women with a normal size cervix. Our hope is that the shortened cervical length status feedback to patients will be another tool in the repertoire to counsel our patients to quit smoking during pregnancy.

There are some limitations to our study. First, it has a retrospective design and is thus subject to the inherent bias associated with selecting rather than assigning study group membership. Next, the precision of our study results are diminished by the low number of women who experienced SPB. Also, it is unclear that our technique and methods for measuring CL (average of 3 clear images) will produce the same result as a single shortest CL. Also, our results may also be limited because patients were lost to follow up or did not meet inclusion criteria. Further, there is presently no data to support the use of progesterone in asymptomatic patients with a cervical length greater than 20 mm even in the context of tobacco. Finally, these findings may not be generalized to a population with a much lower tobacco abuse rate.

Our study confirms that women with a short cervical length at increased risk of SPB. It also identifies that women with lengths up to 3.4 cm could be at increased risk as well. Furthermore, our study suggests a strategy to target smoking cessation counseling. We propose incorporating the biofeedback of CL measurement and educating the patient about the increase risk for SPB. We think this will result in a more meaningful and effective smoking cessation counseling for women with shortened CL.

Acknowledgements

Contributor Information

Joseph Findley, Charleston Area Medical Center, Dept. of Obstetrics & Gynecology, WVU, Charleston.

Dara J. Seybold, Center for Health Services & Outcomes Research, Charleston Area Medical Center Health Education and Research Institute, Charleston.

Mike Broce, Center for Health Services & Outcomes Research, Charleston Area Medical Center Health Education and Research Institute, Charleston.

Dolly Yadav, Center for Health Services & Outcomes Research, Charleston Area Medical Center Health Education and Research Institute, Charleston.

Byron C. Calhoun, Charleston Area Medical Center, Dept. of Obstetrics & Gynecology, WVU, Charleston.

References