Cervical length screening in women with a history of prior preterm birth is recommended by the American College of Obstetricians and Gynecologists.1 However, it remains controversial whether universal second-trimester cervical length screening should be used as a strategy among women without prior preterm birth. Proponents of universal screening point to the results of two large randomized trials that demonstrated vaginal progesterone reduces the risk of preterm birth in women with a short cervix.2,3 Further support is garnered by two cost-effective analyses, each demonstrating universal cervical length screening to be a cost-effective strategy to prevent preterm birth.4,5
Opponents of universal cervical length screening point to the relative rarity of a short cervix that has been found in actual clinical care and suggest that universal screening may be cost-effective only in higher risk cohorts.6,7 For example, the point estimate of the frequency for a short cervix used in the cost-effectiveness analyses references the frequency observed in randomized trials. However, 13–15% of women included in these trials had a prior preterm birth; also, the overall populations may not reflect low-risk populations in the community.2,3 Accordingly, the point estimate of the frequency of a short cervix in these trials was higher than the observed frequency of a short cervix in a low-risk cohort.6,7 Thus, a program of universal screening in a low-risk population may require a large number of women to be screened to prevent one preterm birth, may be onerous from a health care delivery perspective, and may not be as cost-effective as previously thought. Conversely, if women who are at highest risk of having a short cervix could be reliably identified, such women could be selected for targeted screening, whereas those women highly unlikely to have a short cervix could avoid unnecessary testing altogether.
The objective of this study was to estimate whether there are demographic or clinical risk factors that are associated with a short cervix and whether these factors can be utilized in an algorithm to optimize the efficiency of a cervical length screening program.
MATERIALS AND METHODS
This is a cohort study of women with a singleton gestation without a history of spontaneous preterm birth and who underwent routine second-trimester transvaginal cervical length screening between 18 0/7 and 23 6/7 weeks of gestation from December 2010 until January 2014 because this was a routine part of the ultrasonogram at the time of the fetal anatomic survey during the study period at our institution. Women were included in the study if they were at least 18 years of age, had a singleton gestation, and delivered at Northwestern Memorial Hospital. Women were excluded if they had a multiple gestation or had a history of a spontaneous preterm delivery. Cervical length measurements were obtained transvaginally by staff ultrasonographers, all of whom had undergone formal training and certification in the context of multicenter studies and trials, and read by the attending sonologist.8 If multiple examinations were performed, the shortest cervical length obtained before 23 6/7 weeks of gestation was recorded.
Data abstracted from the medical records of all eligible patients were downloaded and reviewed. For data points unavailable through this method or for missing variables, medical charts were manually reviewed. Demographic characteristics and clinical history that have been associated with spontaneous preterm birth were abstracted, including maternal age, race–ethnicity, smoking status, body mass index (calculated as weight (kg)/[height (m)]2), obstetric history, prior cervical excisional procedure, and conception by in vitro fertilization. Charts of patients with missing data were rereviewed and, if the variable continued to be missing, the individual variable was omitted from analysis. Women were categorized according to whether they had a short cervix, which has been defined by multiple investigators as a cervical length 2.5 cm or less.9,10 Bivariable analyses were used to compare demographic characteristics and clinical data between women with and without a short cervix. Comparisons were performed using Student's t test, χ2 analysis, or Fisher's exact tests, as appropriate. Variables significantly associated with a short cervix were then included in a multivariable logistic regression to identify characteristics that remained independently statistically significantly associated with a short cervix.
Prediction models for short cervix, based on the number of risk factors present, were then developed. Test characteristics (sensitivity and specificity) of cervical length assessment to identify a short cervix were calculated for different risk-based approaches with universal screening considered to be the “gold standard” for complete ascertainment (ie, 100% sensitivity) of women with a short cervix. The binomial proportions method was used to calculate 95% confidence intervals (CIs) around each.
Finally, the mean gestational age at delivery, risk of preterm birth before 37 weeks of gestation, and risk of preterm birth before 34 weeks of gestation were calculated for all women with a short cervix as well as for the entire population if the risk-based models were to be applied. Because this analysis began before the routine use of progesterone among women with a short cervix, we assumed that the observed preterm birth rate approximated the background rate without screening and treatment. Assuming a 40% risk reduction of preterm birth with the use of vaginal progesterone and 100% compliance with the prescription of progesterone, the number needed to screen to prevent one preterm birth was calculated for each screening strategy arm by dividing the number of women screened in a given strategy by the reduction in the number of preterm births that would be achieved as a result of progesterone use after identification of women with a short cervix. The 95% CI for each was calculated using the binomial proportions method. To assess whether the assumption of a vaginal progesterone-naïve population affected our results, a sensitivity analysis was performed in which half of the women who had preterm birth in the context of a short cervix were assumed to have received progesterone.
P<.05 was used to define statistical significance and all tests were two-tailed. Stata 13.1 was used to perform the statistical analyses. This study was approved by the Northwestern University institutional review board before its initiation.
During the study period, 18,250 women with a singleton gestation without a history of spontaneous preterm birth underwent routine second-trimester transvaginal cervical length screening. The mean gestational age at screening was 20.3±0.9 weeks. Twenty-nine women underwent a follow-up ultrasonogram for transvaginal cervical length as a result of the finding of a cervical length near the threshold that would lead to a recommendation for treatment. Of all women who underwent cervical length screening, 1,156 (6.3%) delivered preterm. One hundred sixty-four women (0.9%) were found to have a short cervix (ie, 2.5 cm or less) during routine screening. In bivariable analysis, race–ethnicity, current tobacco use, obesity, a prior indicated preterm birth, and a prior cervical excisional procedure (ie, loop electrosurgical excisional procedure or cold knife cone) all were associated with a short cervix (Table 1). In multivariable logistic regression, black (adjusted odds ratio [OR] 3.77, 95% CI 2.42–5.87) and Hispanic (adjusted OR 1.73, 95% CI 1.10–2.74) race–ethnicity, current tobacco use (adjusted OR 3.67, 95% CI 1.56–8.62), prior indicated preterm birth (adjusted OR 2.26, 95% CI 1.26–4.05), and prior cervical excisional procedure (adjusted OR 2.96, 95% CI 1.86–4.70) all remained independently associated with a short cervix (Table 2).
A total of 6,832 (37.4%) women had at least one identifiable risk factor for a short cervix and 649 (3.6%) had at least two risk factors. Both the frequency of preterm birth and of a short cervix increased in the presence of an increasing number of risk factors (Table 3). Test characteristics of different risk-based models that use these factors to select patients for cervical length screening are shown in Table 3. Although using this risk-based approach improves specificity, there is a corresponding reduction in sensitivity. Selecting women for cervical length assessment based on their number of risk factors results in 65 women with a short cervix missed (ie, 39.6% of all women with a short cervix) if at last one clinical risk factor were to be used as a criterion and 140 (ie, 85.4% of all women with a short cervix) missed if at least two clinical risk factors were to be used as a criterion.
Table 3 illustrates the implications such risk-based approaches have for preterm birth prevention. In our population, 30.5% of women with a short cervix delivered preterm. If the population were to be defined on the basis of risk factors, 36.4% of women with a short cervix and at least one risk factor and 54.2% of women with a short cervix and at least two risk factors delivered preterm. Thus, having a short cervix in the presence of risk factors for a short cervix increases the chance of preterm birth compared with women who have a short cervix alone. A similar pattern was observed for preterm birth before 34 weeks of gestation. Assuming a 40% risk reduction afforded by vaginal progesterone, these differences translate into a number needed to screen to prevent preterm birth in a universal screening, at least one risk factor, and at least two risk factor approaches of 913 (95% CI 591–1,494), 474 (95% CI 291–892), and 125 (95% CI 56–399) women, respectively. In a sensitivity analysis in which we had assumed that half of women with preterm birth had been exposed to vaginal progesterone, the number needed to screen to prevent one preterm birth would be 675 (95% CI 465–1,025) for the universal strategy, 356 (95% CI 230–597) for the one-factor model, and 94 (95% CI 45–230) for the two-factor model.
The implementation of a policy of universal second-trimester cervical length assessment remains a contentious topic given that its potential benefit of a reduction of preterm births is countered by the additional burden it places on the health care system and the concern that it may only be cost-effective in settings of a relatively high frequency of a short cervix. Alternative strategies such as using obstetric history to select women who would then undergo screening have been proposed,6 although obstetric history is only one of many risk factors associated with a short cervix. This study has attempted to identify additional risk factors in an effort to determine whether a particularly high-risk population could be identified that would benefit most from screening and whose targeting would make a screening program most cost-effective. Such targeted screening exists for many other types of screening modalities such as for Pap tests, mammography, and colonoscopy.
Our results show that a risk-based selection system could significantly reduce the number of women who need to undergo cervical length screening. There is an increased, albeit arguably marginal, incremental burden of a screening transvaginal ultrasonogram at the time of the anatomic survey (including patient discomfort, additional time required for preparation and scanning, and ultrasonographer time for sterilization of the probes) that could be avoided with risk-based screening. In addition, the use of a risk-based system can identify women at an even higher risk of preterm birth if a short cervix is found. However, by using a risk-based system instead of universal cervical length screening, a reasonably large number of women with a short cervix and a preterm birth will not be identified. For example, if women with at least one risk factor were screened, 63% of transvaginal ultrasonograms could be avoided, but 40% of women with a short cervix will not be identified and would miss the corresponding chance of averting a preterm birth through progesterone treatment.
Assuming progesterone reduces the chance of preterm birth by 40%, the consequences of such a risk-based system for selection can be placed into context. In our population of 18,250 women, routine universal examination accompanied by universal treatment would result in approximately 20 fewer preterm births occurring than if no screening and treatment had been performed. The preterm birth rate of 6.33% would then be 6.22%. Using the one-factor system, more than 60% fewer transvaginal ultrasonograms would be performed and the preterm birth rate would be 6.26%; the use of a two-factor system would result in more than 97% fewer ultrasonograms but a preterm birth rate of 6.31%. This analysis cannot determine what the optimal tradeoff is, although additional cost-effective analyses may give useful insight.
The strength of these data derives from the fact that they were obtained from a large unselected population of women undergoing routine cervical length screening. The risk factors for short cervical length that were identified such as black race are consistent with known risk factors for spontaneous preterm birth as well. Of note, a prior indicated preterm birth was found to be associated with a short cervix in a subsequent pregnancy. This finding supports recent data that have shown that women with a prior indicated preterm birth are at increased risk of subsequent spontaneous preterm birth, presumably as a result of a common pathophysiologic etiology.11–13
Although the frequency of a cervical length 2.5 cm or less was only 0.9%, recent studies support the finding that among a general population, the actual frequency of a short cervix is much lower than originally estimated in study populations.6,7,9 The validity of this low prevalence in our population is further buttressed by the observed frequency of preterm birth. If the low frequency were the result of a systematic overmeasurement of cervical length in our population, we would anticipate that those with an identified short cervix would be observed to have a higher incidence of preterm birth than that reported in prior studies. The 31% preterm birth rate in women identified to have a short cervix mirrors that of other studies.2,3 The lower frequency of a short cervix in our population underscores the importance of determining the actual cost-effectiveness of a universal screening strategy among an actual unselected population.
One limitation of these results is the assumption that progesterone is equally efficacious in each subgroup. It is possible that these phenotypic differences correspond to a differential effect of progesterone that would influence screening recommendations. However, existing data do not support any differential effect of progesterone based on either race–ethnicity or prior preterm birth.10 Also, we used a cervical length of 2.5 cm to define a short cervix, because this measure has been advocated by some as the threshold for treatment, although there is as yet no direct evidence that progesterone is efficacious in low-risk populations at cervical length measurements beyond 2.0 cm. Similarly, we defined preterm birth as occurring at less than 37 weeks of gestation although in one trial, vaginal progesterone reduced preterm birth only at lower gestational ages. If a lower cervical length threshold or gestational age were to be considered, the number needed to screen would rise considerably. Also, some women with a short cervix in our population may have received vaginal progesterone after the publication of randomized trials and the corresponding American College of Obstetricians and Gynecologists committee opinion supporting its use.1–3 The certainty regarding this use for a given woman could not be established given it is administered and recorded as an outpatient. Nevertheless, our assumption that our observed population was progesterone-naïve changes the specific number needed to screen for each strategy but not the gestalt of the results.
In summary, these data support a growing body of literature suggesting that universal cervical length screening may not be as cost-effective as previously believed or inevitably the best strategy once all cost and effectiveness considerations are incorporated to prevent preterm birth. The risks of missed opportunities for intervention must be weighed against the costs of screening when most women are found to have a normal cervical length. Future development of prediction models for a short cervix along with formal cost-effectiveness analyses of these various risk-based screening strategies may inform policy.
1. Prediction and prevention of preterm birth. Practice Bulletin No. 130. American College of Obstetricians and Gynecologists. Obstet Gynecol 2012;120:964–73.
2. Fonseca EB, Celik E, Parra M, Singh M, Nicolaides KH; Fetal Medicine Foundation Second Trimester Screening Group. Progesterone and the risk of preterm birth among women with a short cervix. N Engl J Med 2007;357:462–9.
3. Hassan SS, Romero R, Vidyadhari D, Fusey S, Baxter JK, Khandelwal M, et al.. Vaginal progesterone reduces the rate of preterm birth in women with a sonographic short cervix: a multicenter, randomized, double-blind, placebo-controlled trial. Ultrasound Obstet Gynecol 2011;38:18–31.
4. Cahill AG, Odibo AO, Caughey AB, Stamilio DM, Hassan SS, Macones GA, et al.. Universal cervical length screening and treatment with vaginal progesterone to prevent preterm birth: a decision and economic analysis. Am J Obstet Gynecol 2010;202:548.e1–8.
5. Werner EF, Han CS, Pettker CM, Buhimschi CS, Copel JA, Funai EF, et al.. Universal cervical-length screening to prevent preterm birth: a cost-effectiveness analysis. Ultrasound Obstet Gynecol 2011;38:32–7.
6. Orzechowski KM, Boelig R, Nicholas SS, Baxter J, Berghella V. Is universal cervical length screening indicated in women with prior term birth? Am J Obstet Gynecol 2015;212:234.e1–5.
7. Facco FL, Simhan HN. Short ultrasonographic cervical length in women with low-risk obstetric history. Obstet Gynecol 2013;122:858–62.
8. Grobman WA, Thom EA, Spong CY, Iams JD, Saade GR, Mercer BM, et al.. 17 alpha-hydroxyprogesterone caproate to prevent prematurity in nulliparas with cervical length less than 30 mm. Am J Obstet Gynecol 2012;207:390.e1–8.
9. Iams JD, Goldenberg RL, Meis PJ, Mercer BM, Moawad A, Das A, et al.. The length of the cervix and the risk of spontaneous premature delivery. National Institute of Child Health and Human Development Maternal Fetal Medicine Unit Network. N Engl J Med 1996;334:567–72.
10. Romero R, Nicolaides K, Conde-Agudelo A, Tabor A, O'Brien JM, Cetingoz E, et al.. Vaginal progesterone in women with an asymptomatic sonographic short cervix in the midtrimester decreases preterm delivery and neonatal morbidity: a systematic review and metaanalysis of individual patient data. Am J Obstet Gynecol 2012;206:124.e1–19.
11. Savitz DA, Dole N, Herring AH, Kaczor D, Murphy J, Siega-Riz AM, et al.. Should spontaneous and medically indicated preterm births be separated for studying aetiology? Paediatr Perinat Epidemiol 2005;19:97–105.
12. Ananth CV, Getahun D, Peltier MR, Salihu HM, Vintzileos AM. Recurrence of spontaneous versus medically indicated preterm birth. Am J Obstet Gynecol 2006;195:643–50.
© 2015 by The American College of Obstetricians and Gynecologists. Published by Wolters Kluwer Health, Inc. All rights reserved.
13. Laughon SK, Albert PS, Leishear K, Mendola P. The NICHD Consecutive Pregnancies Study: recurrent preterm delivery by subtype. Am J Obstet Gynecol 2014;210:131.e1–8.