To investigate the efficacy of vaginal progesterone to prevent early preterm birth in women with sonographic evidence of a short cervical length in the midtrimester.
To investigate the efficacy of vaginal progesterone to prevent early preterm birth in women with sonographic evidence of a short cervical length in the midtrimester.
This was a planned, but modified, secondary analysis of our multinational, multicenter, randomized, placebo-controlled trial, in which women were randomized between 18 + 0 and 22 + 6 weeks of gestation to receive daily treatment with 90 mg of vaginal progesterone gel or placebo. Cervical length was measured with transvaginal ultrasound at enrollment and at 28 weeks of gestation. Treatment continued until either delivery, 37 weeks of gestation or development of preterm rupture of membranes. Maternal and neonatal outcomes were evaluated for the subset of all randomized women with cervical length < 28 mm at enrollment. The primary outcome was preterm birth at ≤ 32 weeks.
A cervical length < 28 mm was identified in 46 randomized women: 19 of 313 who received progesterone and 27 of 307 who received the placebo. Baseline characteristics of the two groups were similar. In women with a cervical length < 28 mm, the rate of preterm birth at ≤ 32 weeks was significantly lower for those receiving progesterone than it was for those receiving the placebo (0% vs. 29.6%, P = 0.014). With progesterone, there were fewer admissions into the neonatal intensive care unit (NICU; 15.8% vs. 51.9%, P = 0.016) and shorter NICU stays (1.1 vs. 16.5 days, P = 0.013). There was also a trend toward a decreased rate of neonatal respiratory distress syndrome (5.3% vs. 29.6%, P = 0.060).
Vaginal progesterone may reduce the rate of early preterm birth and improve neonatal outcome in women with a short sonographic cervical length. Copyright © 2007 ISUOG. Published by John Wiley & Sons, Ltd.
Despite medical efforts towards its prevention, the rate of preterm birth, defined as birth occurring prior to 37 weeks of gestation, continues to rise among certain populations, occurring in up to 15% of pregnancies in the developed world and 12.7% in the United States1, 2. Preterm birth is associated with a high prevalence of severe neurological deficits and developmental disabilities and is a leading cause of infant and neonatal mortality in the United States3–5. Preterm neonates are at increased risk of developing respiratory distress syndrome, sepsis, intraventricular hemorrhage, necrotizing enterocolitis and disorders related to gestational age at birth6.
A short cervix is a known risk factor for preterm birth; in fact, data support an inverse relationship between cervical length and preterm delivery. In a prospective study of 2915 women that investigated the relationship between short cervical length and preterm delivery, researchers found that even a small decrease in cervical length between the 24th and 28th weeks of gestation was associated with an increased risk of preterm birth (relative risk, 2.03; 95% CI, 1.28–3.22)7. At 24 weeks, when compared with women whose cervical length was above the 75th percentile, women who had a cervical length on or below the 25th percentile (30 mm) had a relative risk of preterm delivery of 3.79 (95% CI, 2.32–6.19), and those on or below the 10th percentile (26 mm) had a relative risk of 6.19 (95% CI, 3.84–9.97)7.
Progesterone administration has been advocated for the prevention of preterm birth in women considered to be at high risk8, 9, although the primary focus has been on patients with a prior history of preterm birth10–13. There has been increasing interest in the efficacy of vaginally administered progesterone to prevent preterm birth in women at especially high risk of preterm delivery, those with a short sonographic cervical length. The use of progesterone to treat all women at risk for preterm birth does not have uniform support at present, but it has rapidly become accepted as a prophylactic measure to prevent preterm birth in women with a documented history of spontaneous preterm birth. Several authors have expressed the need for adequately designed, randomized trials in larger populations to identify the ideal progesterone formulation and dosage, and to demonstrate whether progesterone administration results in a decline in preterm births before 37 weeks and a reduction in perinatal morbidity and mortality8, 14–16.
The primary objective of our prospective, randomized, double-blind, placebo-controlled, multicenter study, described in an accompanying article17, was to assess the efficacy and safety of 90-mg progesterone vaginal gel compared with placebo to decrease the rate of early preterm birth (≤ 32 + 0 weeks) among women with a prior history of spontaneous preterm delivery. In that study, we found no difference between placebo and progesterone in reducing the frequency of preterm birth among women selected only by a documented history of preterm birth. The objective of this planned secondary analysis was to evaluate the efficacy of daily vaginal progesterone gel to prevent preterm birth in women with a short cervix in the midtrimester.
This was a planned, but modified, secondary analysis of the prospective, randomized, double-blind, placebo-controlled trial of progesterone to prevent preterm birth in high-risk women, reported in the accompanying article in this issue of the Journal17. Prior to the trial, we planned to analyze data and report outcomes separately for the women enrolled without a history of preterm birth but with a short cervical length alone (≤ 25 mm). Due to an insufficient number of subjects recruited into this arm of the trial (n = 9, 1.3% of the study population), the separate analysis of these patients was not meaningful. Therefore, we modified the planned analysis of women enrolled based on a short cervix only to include women enrolled with a prior preterm birth who also had a short cervix at enrollment. To accomplish this, we divided the study population into quartiles based on cervical length. We then subdivided the lowest quartile (≤ 32 mm) sequentially and analyzed the primary and secondary outcomes for women with cervical lengths of ≤ 30 mm and < 28 mm at the time of enrollment. There were an insufficient number of patients with even shorter cervical lengths to allow further analysis.
Women with a documented history of spontaneous preterm birth (< 35 weeks) in a singleton pregnancy in the immediate preceding pregnancy, regardless of cervical length, and women without a history of preterm birth but with a short cervix (≤ 25 mm) in the midtrimester of the current pregnancy were screened by the investigator or study coordinator between 16 + 0 and 22 + 6 weeks of gestation. Subjects meeting the study criteria were offered enrollment into the study at 18 + 0 to 22 + 6 weeks of gestation to receive daily treatment with 90 mg of vaginal progesterone gel (Prochieve® 8%/Crinone® 8%) or placebo (Replens®), both provided by Columbia Laboratories, Inc. (Livingston, NJ, USA). Cervical length was measured with transvaginal ultrasound at enrollment and at 28 weeks of gestation. Treatment was continued until either delivery, 37 weeks of gestation or development of preterm rupture of membranes.
Details of the study period, exclusion criteria, randomization and drug/placebo treatment procedures, as well as the definition of preterm labor are provided in the accompanying report of the randomized clinical trial17.
The outcomes for this study were defined a priori. The primary outcome was preterm birth at ≤ 32 weeks. Secondary outcomes included: preterm birth at < 37 weeks, ≤ 35 weeks, and ≤ 28 weeks; treatment-related adverse events; hospital admission for preterm labor; prolongation of latency period after treatment for preterm labor; change in cervical length between randomization and 28 weeks; and neonatal morbidity and mortality.
Statistical power was calculated a priori for the entire study population based on the primary outcome: a 50% reduction in the rate of preterm birth at ≤ 32 weeks. The methods used to calculate power for the primary study are detailed in the accompanying article17. Because performing retrospective power analyses can be considered statistically inappropriate, a supplemental power calculation for this secondary analysis was not undertaken18, 19. Data were analyzed based on the intent-to-treat (ITT) principle using the SAS 9.1 statistical software package (SAS Institute Inc., Cary, NC, USA). Chi-square or Fisher's exact tests were used for categorical data and ANOVA was used for continuous data. Pregnancy prolongation was evaluated using life-table analysis and the Kaplan–Meier method to estimate the time-to-delivery event curve for treatment groups. The log-rank and Wilcoxon tests were used to evaluate Kaplan–Meier curves. A P-value of < 0.05 was considered statistically significant.
There were 53 study sites participating in the full trial. The 17 US study sites initially recruited enrolled women with either a history of spontaneous preterm birth (< 35 weeks) or a short cervix only (≤ 25 mm), while the 36 sites subsequently recruited worldwide enrolled only women with a history of spontaneous preterm birth (< 35 weeks), because the primary objective of the trial was to evaluate the effect of progesterone in patients with a previous preterm birth. A total of 711 women gave written informed consent, of whom 42 were not randomized. The most common reasons for exclusion after consent were planned cerclage, comorbid conditions and failure to document a previous spontaneous preterm birth, as determined by review of medical charts of previous pregnancies. A record of screened patients was constructed when consent was obtained. The exact number of those prescreened is not available; an estimate of 1500 prescreened subjects was obtained by querying the study sites after trial completion. One patient was lost to follow-up prior to randomization, so a total of 668 women at high risk for preterm birth participated in the trial. Outcome data were available for 620 of these women (92.8%): 313 who received progesterone (309 enrolled with a prior preterm birth at < 35 weeks and four enrolled with only a short cervix ≤ 25 mm) and 307 who received the placebo (302 enrolled with a prior preterm birth and five enrolled with only a short cervix). Patients who took at least one dose of study medication and for whom information was available regarding the date of delivery were included in the ITT population. Patients without a record of delivery date were considered lost to follow-up.
Using a cervical length of < 28 mm (the 9th percentile of the ITT population) to expand the criteria for short cervix resulted in a cohort of 46 women: 19 who received progesterone and 27 who received placebo (Figure 1). The randomization provided treatment groups in which the baseline demographic and obstetric characteristics were similar (Table 1).
|Progesterone (n = 19)||Placebo (n = 27)|
|Maternal age (years, mean (SD))||27.4 (4.9)||25.4 (4.8)||0.18|
|Race/ethnicity (n (%))|
|Caucasian||9 (47.4)||10 (37)||0.55|
|African-American||3 (15.8)||11 (40.7)||0.10|
|Asian/Pacific Islander||0||4 (14.8)||0.13|
|Other||6 (31.6)||2 (7.4)||0.05|
|Body mass index (mean (SD))||28.5 (8.3)||26.9 (6.7)||0.52|
|Prior preterm births (n, mean (SD))||1.2 (0.5)||1.4 (0.8)||0.26|
|> 1 prior preterm birth (n (%))||7 (37)||5 (19)||0.80|
|Prior cervical surgery (n (%))||3 (16)||6 (22)||0.70|
|Prior spontaneous miscarriages (n, mean (SD))||0.8 (1.4)||0.4 (0.7)||0.22|
|GA at randomization (weeks, mean (SD))||20.4 (1.3)||20.4 (1.6)||0.98|
The original planned secondary analysis involved evaluating the efficacy of progesterone to prevent preterm birth in women enrolled with only a short cervix (≤ 25 mm) at baseline. The frequency of preterm birth at ≤ 32 weeks' gestation in these women was 0% in the progesterone group (0/4) compared with 40% in the placebo group (2/5).
Because cervical length is a stronger predictor of preterm birth than is obstetric history20, an evaluation of outcomes based on the baseline cervical length of all randomized patients in the main trial was added to the analysis. Kaplan–Meier curves comparing the progesterone and placebo groups for time to delivery were prepared for all women enrolled in the trial with a baseline cervical length measurement (n = 609)17, for women with the lowest quartile of cervical length (≤ 32 mm; n = 172; Figure 2a) and for women with the highest three quartiles combined (> 32 mm; n = 437; Figure 2b). Women in the lowest quartile of cervical length (≤ 32 mm) who were treated with progesterone did not demonstrate a significant delay in preterm delivery when compared to placebo patients (Wilcoxon's P = 0.37, log-rank P = 0.34; Figure 2a), nor did women in the upper three quartiles (> 32 mm; Figure 2b). Women with cervical lengths in the upper three quartiles had a Kaplan–Meier curve for time to delivery no different from all randomized women in the main trial17.
The maternal/fetal outcomes of the 172 women enrolled with a cervical length ≤ 32 mm did not differ significantly between treatment groups. Of these women, 83 received progesterone and 89 received the placebo. The rate of preterm birth at ≤ 32 weeks of gestation for this group of women was 7.2% for those treated with progesterone compared with 13.5% for those treated with placebo (P = 0.21). The rates of preterm birth at < 37, < 35 and < 28 weeks were 44.6% vs. 51.7%, 22.9% vs. 30.3% and 1.2% vs. 6.7%, respectively, for the progesterone and placebo groups (P-values were 0.36, 0.30 and 0.12, respectively). The number of neonates admitted to the neonatal intensive care unit (NICU) did not differ significantly for progesterone compared with placebo groups (13 vs. 21, P = 0.25), nor did the length of NICU stay (13.0 vs. 32.7 days, P = 0.14). Neonatal morbidities of intraventricular hemorrhage (1.2% vs. 2.4%, P = 1.0), respiratory distress syndrome (7.2% vs. 13.5%, P = 0.21) and necrotizing enterocolitis (1.2% vs. 1.1%, P = 1.0) did not differ between the groups.
Kaplan–Meier curves were also generated for subsets of all randomized women within the lowest quartile of cervical lengths (≤ 30 mm and < 28 mm). Figure 3 shows that there was a statistical trend towards delay in preterm delivery for progesterone-treated patients relative to placebo-treated patients in the 116 women with a baseline cervical length ≤ 30 mm (Wilcoxon's P = 0.043, log-rank P = 0.057). Among women with a baseline cervical length of ≤ 30 mm, progesterone treatment was associated with a significant reduction in NICU days when compared with the placebo results (2 vs. 12 days, P = 0.026). The progesterone vs. placebo groups were not significantly different with respect to the number of neonates admitted to the NICU (16% vs. 32%, P = 0.077), total number of neonatal hospital days (7 vs. 14 days, P = 0.095) and occurrence of neonatal respiratory distress syndrome (7% vs. 19%, P = 0.09).
Figure 4 demonstrates Kaplan–Meier curves for the 46 randomized women with a baseline cervical length of < 28 mm. This subgroup of women with a short cervix experienced a significant reduction in the frequency of preterm birth for the primary end point of this study, ≤ 32 weeks of gestation (0% vs. 29.6%, P = 0.014) (Table 2). There were decreasing rates of preterm birth for secondary outcomes, < 37 weeks, ≤ 35 weeks and ≤ 28 weeks; however, these differences did not reach statistical significance (Table 2). In women with a cervical length < 28 mm the number of admissions to the NICU was lower in those receiving progesterone than it was in those receiving placebo (15.8% vs. 51.9%, P = 0.016) and the length of NICU stay was shorter in the progesterone group than it was in the placebo group (1.1 vs. 16.5 days, P = 0.013). There was a trend, although it was not statistically significant, towards a reduction in total neonatal hospital days (5.8 vs. 18.2 days, P = 0.055) and decreased occurrence of neonatal respiratory distress syndrome (5.3% vs. 29.6%, P = 0.060) with progesterone therapy (Table 3).
|Progesterone (n = 19)||Placebo (n = 27)|
|GA at birth (weeks, mean (SD))||36.3 (2.4)||34.6 (4.6)||0.160|
|Preterm birth (n (%))|
|< 37 weeks||8 (42.1)||16 (59.3)||0.370|
|≤ 35 weeks||7 (36.8)||13 (48.1)||0.551|
|≤ 32 weeks*||0||8 (29.6)‡||0.014§|
|≤ 28 weeks||0||3 (11.1)||0.257|
|Cervical length at enrollment (mm)|
|Mean (SD)||24 (0.2)||22 (0.5)||0.07|
|Median (range)||25 (19–27)||25 (11–27)|
|Cervical length at 28 weeks (mm, mean (SD))||25 (0.8)||22 (0.8)||0.27|
|Change in cervical length (mm, mean (SD))||2 (0.9)||0 (0.9)||0.70|
|Admission for preterm labor (n (%))||6 (31.6)||7 (25.9)||1.0|
|Latency period to delivery after tocolysis for preterm labor (days, mean (SD))||42.7 (52.3)||10.0 (18.0)||0.287|
|Compliance (% (SD))†||93.9 (9.77)||94.7 (13.03)|
|Progesterone (n = 19)||Placebo (n = 27)|
|Birth weight (g, mean (SD))||2726 (645)||2290 (937)||0.1|
|Hospital days (n, mean (SD))||5.8 (9)||18.2 (25.5)||0.055|
|NICU admission (n (%))||3 (15.8)||14 (51.9)||0.016|
|Days in NICU per admission (n, mean (SD))||1.1 (2.7)||16.5 (24.9)||0.013|
|Respiratory distress syndrome (n (%))||1 (5.3)||8 (29.6)||0.060|
|Intraventricular hemorrhage (n (%))|
|Grade 1||0||2 (7.4)||0.5|
|Necrotizing enterocolitis (n (%))|
|Proven sepsis (n (%))||1 (5.3)||3 (11.1)||1.0|
|Neonatal death (n (%))||0||1 (3.7)||1.0|
In the main trial, there was no difference between the vaginal progesterone gel group and the placebo group with respect to the frequency of adverse events overall (81.3% vs. 83.2%) or the frequency of serious adverse events17. Complaints about vaginal discharge occurred in 9.2% of placebo patients and 8.4% of progesterone patients, but for only 4.4% and 4.0%, respectively, was the vaginal discharge attributed to the vaginal gel. No miscarriages occurred among the trial participants, although 53% of all subjects began therapy prior to 20 weeks of gestation. There was also no difference in the occurrence of congenital anomalies between the progesterone and placebo groups. Finally, there was no increase in fetal/infant mortality overall or in any of the subgroups analyzed. Among women with a cervical length ≤ 30 mm at enrollment, there were four fetal/infant deaths in the placebo group ((1) 23 weeks of gestation, 590 g, prematurity; (2) 26 weeks of gestation, 620 g, intrauterine fetal demise; (3) 35 weeks of gestation, aspiration pneumonia; (4) term, expired at 11 months of age, sudden infant death syndrome (SIDS)) and two fetal/infant deaths in the vaginal progesterone group ((1) 21 weeks of gestation, 250 g, intrauterine fetal demise; (2) term, expired at 6 months of age, gastroenteritis). In the subgroup of women enrolled with a cervical length of < 28 mm, two fetal/infant deaths occurred in the placebo group ((1) term, expired at 11 months of age, SIDS; (2) 35 weeks of gestation, aspiration pneumonia) and there were none in the vaginal progesterone group.
This study is a secondary analysis of the largest randomized trial conducted to date to evaluate the efficacy of progesterone for the prevention of preterm birth17. While progesterone did not reduce the frequency of preterm birth in women at high risk based on history alone, an objective screening tool, transvaginal sonographic determination of cervical length, was useful to identify patients responding to therapy. Our analysis suggests that progesterone may prevent early preterm birth (≤ 32 weeks of gestation) and improve neonatal outcome in women with short cervical lengths < 28 mm identified between 18 + 0 and 22 + 6 weeks of gestation.
The cohort of women evaluated in this study included a population of high-risk pregnancies with at least one of two well-known risk factors for preterm birth: a history of preterm birth and a short cervix in the midtrimester. According to data from To et al.20, short cervical length is the best single predictor for preterm birth, predicting 61.2% of cases, while the combination of cervical length and a history of preterm birth predicts only an additional 4.4% (65.6% total). Data from the Preterm Prediction Study has shown that the 10th percentile of cervical length in a low-risk population is 26 mm7. The measurement of baseline cervical length in all patients and exclusion of women with a plan for cervical cerclage placement may have resulted in the exclusion of most of the subjects we expected to enroll with a short cervix in the midtrimester. Although the number of patients recruited into the short-cervix-only population (≤ 25 mm) was small, the rate of preterm birth in those receiving the placebo was 40%. In comparison, the preterm birth rate for the placebo group of women with a cervical length < 28 mm at enrollment (including a large proportion of patients with both a short cervix and a previous preterm birth) was 30%, with no apparent increase. These results lead us to speculate that the ‘responders’ in the previously published trials of progestin administration in women with a history of preterm birth12, 13 may well have been those women with early-onset cervical shortening. Neither of the two largest previous trials in the population of women with a history of preterm birth included a measurement of baseline cervical length12, 13.
Cervical length is the best predictor of preterm birth according to data previously published from the research centers of Iams7, Nicolaides20 and Berghella21, 22. This fact was noted by the American College of Obstetricians and Gynecologists in ACOG Practice Bulletin Number 3123 and was cited in the Institute of Medicine Report24, ‘Preterm Birth: Causes, Consequences, and Prevention.’ In addition to providing improved prognostication, this objective screening tool for preterm birth risk offers other advantages compared with screening by historical factors. Any population better defined by useful objective characteristics is more likely to share a similar pathophysiology to the end point of preterm birth, which has multiple etiologies. In well-defined populations, an intervention may be better directed toward interrupting a specific mechanism or mechanisms of disease.
Several studies have addressed the mechanism with which progesterone might reduce the rate of cervical shortening or ‘ripening’24. Cervical ripening can be characterized as a reduction in total collagen content by an increase in collagen solubility and collagenolytic activity. Collectively, these activities cause a remodeling of the extracellular matrix of the cervix25. A number of hormones, including progesterone, influence this activity. Estrogen stimulates collagen degradation in vitro and progesterone blocks estrogen-induced collagenolysis in vitro26. Progesterone also down-regulates interleukin 8 production by the cervix27. Likewise, administration of a progesterone-receptor antagonist induces cervical ripening in the first trimester28. Taken together, these observations suggest that progesterone inhibits the cervical ripening that is known to precede labor.
The efficacy of progesterone in our study population is noteworthy and encouraging. Although we do not believe that every patient will respond identically, in women with a short cervix, no treated patients delivered at less than 32 weeks of gestation. This treatment effect was not seen in the overall study population or in the remaining 75% of trial participants who were at high risk for preterm birth based only on a previous preterm birth earlier than 35 weeks' gestation. Unfortunately, for women with longer cervical lengths, other strategies to prevent preterm birth using other objective criteria for risk assessment are needed.
A recent trial by Fonseca et al.29 also demonstrated that women with a short cervical length identified in the midtrimester by transvaginal sonography are less likely to deliver preterm if they are treated with vaginal progesterone. In that trial women were screened between 20 and 25 weeks of gestation and offered randomization to 200-mg vaginal progesterone suppositories or placebo from 24 to 34 weeks if the midtrimester cervical length was ≤ 15 mm. Progesterone treatment was associated with a significant reduction in preterm birth at < 34 weeks (relative risk, 0.56; 95% CI, 0.36–0.86).
To our knowledge, ours is the only trial of a progestin in the prevention of preterm birth to be associated with a significant improvement in neonatal outcome. Rates of admission to the NICU30, 31 and NICU days32 are recognized measures of infant morbidity, as they relate to other neonatal complications. The intravaginal formulation of natural progesterone utilized in our trial had an adverse event profile similar to the placebo, and its route of administration was not associated with miscarriage. There were no safety issues identified with natural progesterone in this trial, or in a previous study when it was used as luteal support during in-vitro fertilization cycles33.
The results from our study and that of Fonseca et al.29 suggest that objective criteria to determine indications for prophylactic treatment with progesterone, specifically sonographic evidence of a short cervix, should be utilized to prevent preterm birth. We found that women with second-trimester sonographic evidence of a short cervical length (< 28 mm) have a reduction in the risk of early preterm birth (≤ 32 weeks) and neonatal morbidity when treated with vaginal progesterone gel. Given the findings of this trial and those of Fonseca et al.29, we suggest that midtrimester cervical length assessment by transvaginal ultrasound be incorporated into routine clinical practice, as prophylactic vaginal progesterone supplementation benefits asymptomatic high-risk pregnancies with early-onset cervical shortening.
This trial was funded by Columbia Laboratories, Inc.
J. M. O'Brien is a consultant and has received honoraria from Cook Biotech, Inc. G. W. Creasy is an employee of Columbia Laboratories, Inc.
W. Hansen (Lexington, KY), M. Newman (Baton Rouge, LA), B. Rosenn (New York, NY), S. Dabak (Pune, India), L. Parker, (Winston-Salem, NC), J. Stern (Memphis, TN), L. Bayer-Zwirello (Boston, MA), L. Cousins (San Diego, CA), A. Kekre (Vellore, India), R. McDuffie, (Denver, CO), J. Schucker (Danville, PA), C. Barrera (Santiago, Chile), C. Goldberg (Tucson, AZ), A. Jiratko (Zlinska, Czech Republic), K. Swenson (Austin, TX), A. Evans (Lubbock, TX), G. Gross (St. Louis, MO), M. Short (Baltimore, MD), S. Sunderji (Toledo, OH), R. Artal (St. Louis, MO), M. Binstock (Bedford, OH), J. Hibbard (Chicago, IL), R Kelly (Odessa, TX), X. Sandovol-Lopez (San Salvador, El Salvador), L. Smith (Livingston, NJ), M. Stitley (Morgantown, WV), E. Wang (Chicago, IL), M. Beall (Torrence, CA), J. Carvajal (Santiago, Chile), V. Rappaport (Albuquerque, NM), L. Wilkins-Haug (Boston, MA), B. Sibai (Cincinnati, OH).