Effect of progesterone on cervical shortening in women at risk for preterm birth: secondary analysis from a multinational, randomized, double-blind, placebo-controlled trial




To determine whether progesterone supplementation alters cervical shortening in women at increased risk for preterm birth.


We performed a planned secondary analysis from a large, multinational preterm birth prevention trial of daily intravaginal progesterone gel, 90 mg, com- pared with placebo in women with a history of spontaneous preterm birth or premature cervical shortening. Transvaginal cervical length measurements were obtained in all randomized patients at baseline (18 + 0 to 22 + 6 weeks' gestation) and at 28 weeks' gestation. For this secondary analysis, the difference in cervical length between these time points was compared for the study population with a history of spontaneous preterm birth and for a population with premature cervical shortening (≤ 30 mm) at randomization. Differences between groups in cervical length for the 28-week examination were analyzed using ANCOVA, including adjustment for relevant clinical parameters and maternal characteristics.


Data were analyzed from 547 randomized patients with a history of preterm birth. The progesterone-treated patients had significantly less cervical shortening than the placebo group (difference 1.6 (95% CI, 0.3–3.0) mm; P = 0.02, ANCOVA). In the population of 104 subjects with premature cervical shortening at randomization, the cervical length also differed significantly on multivariable analysis, with the treatment group preserving more cervical length than the placebo group (difference 3.3 (95% CI, 0.3–6.2) mm; P = 0.03, ANCOVA), with adjustment for differences in cervical length at screening. A significant difference was also observed between groups for categorical outcomes including the frequency of cervical length progression to ≤ 25 mm and a ≥ 50% reduction in cervical length from baseline in this subpopulation.


Intravaginal progesterone enhances preservation of cervical length in women at high risk for preterm birth. Copyright © 2009 ISUOG. Published by John Wiley & Sons, Ltd.


Prevention strategies for preterm birth have included assessing risk factors, with intervention based on obstetric history or other clinical factor such as the presence of a multiple pregnancy or uterine anomaly. A shortcoming with such strategies has been the lack of objective, contemporaneous data to define an indication for therapy and the inability to follow an objective marker that could identify treatment effect. The lack of objective criteria, and past reliance on clinical or historical population characteristics, has resulted in inconsistencies in observations regarding interventions, limited sensitivity in identifying those at risk, and an inability to tailor an intervention directed toward ongoing pathophysiology1, 2.

Clinical research observations in drug trials can suggest where in the reproductive tract further efforts should be directed to elucidate predominant mechanism(s) of action. Defining the mechanism(s) of action of a drug could ultimately validate candidates for this intervention and optimize treatment. Two recent studies have demonstrated a reduction in the rate of preterm birth with intravaginal progesterone administration to women with premature cervical shortening3, 4. Progesterone is known to affect several cellular physiologies in each tissue of the reproductive tract including the cervix5–21.

The present investigation is an analysis assessing cervical length differences in women enrolled in the largest multinational, randomized, double-blind, placebo-controlled trial investigating progesterone for the prevention of singleton preterm birth that has been performed to date. The primary and secondary outcomes for this trial were the frequency of preterm birth at ≤ 32 weeks and neonatal measures. These results have been reported previously and, although women with a history of preterm birth alone did not achieve a prolongation of pregnancy or improved neonatal outcome with progesterone therapy, a successful treatment response was observed in women with premature cervical shortening4, 22. These findings suggest that clinical history alone may not sufficiently discriminate the patient population with a pathophysiology who could benefit from treatment. In this study, we attempt to better understand the relationship between cervical length and progesterone therapy in the mid-trimester by comparing the change in cervical length over time in women at high risk of preterm birth treated with vaginal progesterone gel vs. placebo.


This was a planned secondary analysis from a multinational, industry-sponsored clinical trial, which was performed at 53 centers on five continents (Clinical Trials number NCT00086177). No outcome differences were identified by study location. Patients were candidates for inclusion if they had a documented history of spontaneous singleton preterm birth between 20 and 35 weeks for the delivery immediately preceding the index pregnancy or if they had evidence of premature cervical shortening ≤ 25 mm at 18–22 weeks' gestation. Details of the methods for this trial have been described elsewhere4, 22. The institutional review board approved the protocol at each institution before enrollment.

Patients enrolled in this study were to have an initial transvaginal sonographic examination at between 18 + 0 and 22 + 6 weeks' gestation. A second transvaginal examination was to be performed per protocol at 28 weeks' gestation. The cervical change between these time points was calculated by subtraction of the 28-week value from that obtained at randomization. The primary outcome for this secondary analysis was change in cervical length between enrollment and 28 weeks. Secondary outcomes included length of the cervix at the 28-week examination, the number of women with a cervical length ≤ 25 mm or ≤ 15 mm at the 28-week examination, and the frequency of ≥ 50% cervical length reduction from the baseline value.

As described previously, 620 patients comprised the intention-to-treat population and had as risk factors for preterm birth either a history of spontaneous preterm birth (n = 611) or sonographic evidence of premature cervical shortening (n = 9). These patients were randomized in a 1 : 1 ratio to receive 90 mg intravaginal progesterone gel, Prochieve® 8%, or identical placebo in a double-blinded scheme. Patients with a history of preterm birth and a transvaginal ultrasound examination of the cervix at baseline and at 28 weeks were included in the ‘history of preterm birth’ population (n = 547).

Because women with a history of preterm birth alone did not have an identified benefit of prolongation of pregnancy with treatment in our study, whereas women with a prematurely shortened cervix ≤ 30 mm at randomization did, we also analyzed cervical change in this latter subgroup4. Analysis of other subgroups (< 28 mm or ≤ 32 mm) was not performed because ≤ 30 mm is the cut-off at which we first observed a treatment effect on latency to delivery. This group comprised both patients with a history of spontaneous preterm birth (n = 101) and those without (n = 9). One hundred and four patients in this population had serial sonographic examinations and this cohort was termed the ‘premature cervical shortening’ population. However, it should be noted that 92% of these patients defined as having premature cervical shortening also had a history of preterm birth so there was a large percentage of patients present in both study groups. A 7.4% disparity in the frequency of serial cervical length measurements was noted between the placebo and treatment groups, and this difference was solely attributable to the number of extremely preterm births in each group. To correct for this difference, a limited adjusted analysis was also performed with women who delivered before 28 weeks being assigned a cervical length of zero for the follow-up examination. This group was defined as the ‘modified premature cervical shortening’ group.

Continuous variables included cervical length at randomization, cervical length at 28 weeks' gestation and cervical change. Categorical outcome variables derived from assessment of the cervix at 28 weeks' gestation included the frequency at which the cervical length measurement was ≤ 25 mm or ≤ 15 mm and the frequency of a ≥ 50% reduction in cervical length compared with the length at randomization. Statistical analysis was performed utilizing Fisher's exact test, t-test and ANCOVA. A multivariable analysis aimed at assessing differences in cervical length from screening to the 28-week examination between groups was performed. A model for the 28-week cervical length was constructed including the following clinically relevant covariates: maternal age, race, body mass index, investigator site, treatment group, gestational age at previous preterm birth, gestational age at randomization and cervical length at randomization. Statistical analysis was performed utilizing SAS and JMP statistical software (SAS Inc., Cary, NC, USA). P < 0.05 was considered significant.


A flow diagram describing the overlapping populations for analysis is presented in Figure 1. The demographic characteristics of the history of preterm birth population are shown in Table 1. The treatment and placebo groups were similar with respect to important baseline demographic and historical characteristics, and the mean cervical length at randomization was the same in both groups. The demographic characteristics for the smaller population with premature cervical shortening are also shown in Table 1. In this subpopulation, the mean cervical length at randomization was significantly different between groups; cervical length in the placebo group was 2 mm shorter than that in the progesterone group, which was adjusted for in subsequent analysis. Twelve randomized patients delivered at < 28 weeks (seven placebo, five treated) and as such these patients did not have a measurable outcome (i.e. cervical length measurement at 28 weeks' gestation); of these, six patients had premature cervical shortening (≤ 30 mm) at randomization (five placebo, one treated).

Figure 1.

Flow diagram describing the populations under analysis. Note that the final subgroups may include shared patients creating a total greater than that shown in the previous box. PTB, preterm birth.

Table 1. Demographic characteristics of the study populations
ParameterHistory of preterm birth (n = 547)Premature cervical shortening at randomization (n = 104)
Progesterone group (n = 273)Placebo group (n = 274)Progesterone group (n = 54)Placebo group (n = 50)
  1. Values are mean ± SD or n (%). No significant differences were observed with the exception of cervical length at randomization for the population with premature cervical shortening (P = 0.02).

Maternal age (years)27.3 ± 5.827.4 ± 5.627.0 ± 5.326.1 ± 5.0
 African–American or African100 (36.6)107 (39)20 (37)25 (50)
 Caucasian101 (37.0)94 (34.3)14 (26)10 (20)
 Asian/Pacific Islander53 (19.4)58 (21.2)15 (28)15 (30)
Body mass index (kg/m2)26.6 ± 6.326.3 ± 6.926.3 ± 7.125.1 ± 7.1
Number of previous preterm births1.3 ± 0.71.3 ± 0.71.3 ± 0.51.5 ± 0.8
Previous cervical surgery19 (7.0)25 (9.1)7 (13)4 (8)
Cervical length at randomization (mm)37 ± 737 ± 828 ± 326 ± 5
Gestational age at randomization (weeks)20.0 ± 2.220.2 ± 3.520.2 ± 1.620.1 ± 1.9

As expected, the cervical length of both the placebo-treated and progesterone-treated patients with a history of preterm birth decreased significantly over the interval from randomization to 28 weeks (mean ± SD, −6 ± 9 mm for the placebo group vs. −5 ± 9 mm for the progesterone group) (Figure 2). Importantly, however, the extent of cervical change was significantly different between the progesterone- and placebo-treated groups. Patients treated with placebo had greater cervical shortening than those treated with progesterone. The difference between groups was 1.6 (95% CI, 0.3–3.0) mm, which was statistically significant using both univariate and multivariate analysis (P = 0.02, ANCOVA).

Figure 2.

Cervical length differences over the observation interval from randomization (a) to 28 weeks' gestation (b) for the ‘history of preterm birth’ population (n = 547), comprising 273 patients treated with progesterone and 274 with placebo. Horizontal lines represent mean values and boxes represent 1 SD. There was a significant difference for both groups over the interval. At the 28-week examination, the difference in adjusted cervical length means of 1.6 (95% CI, 0.3–3.0) mm between groups was also significant (P = 0.02, ANCOVA).

On multivariable analysis, the premature cervical shortening population also demonstrated a significant difference in cervical length between patients administered placebo and those treated with progesterone, although no significant difference was found on univariate analysis. Once adjusted for clinically relevant covariates, this difference in preservation of cervical length was greater than that observed in the history of preterm birth population. With progesterone treatment, more cervical length was preserved on average than in the placebo group (3.3 (95% CI, 0.3–6.2) mm; P = 0.03, ANCOVA) (Figure 3). Of note, the mean cervical length in the progesterone group did not change over the observation interval, although there was also no significant change in cervical length for the placebo group.

Figure 3.

Cervical length differences over the observation interval from randomization (a) to 28 weeks' gestation (b) in the subpopulation with premature cervical shortening (≤ 30 mm) at randomization (n = 104), comprising 54 patients treated with progesterone and 50 with placebo. Horizontal lines represent mean values and boxes represent 1 SD. Following adjustment for clinically relevant covariates, a significant difference in adjusted mean cervical length at the 28-week examination was observed between groups of 3.3 (95% CI, 0.3–6.2) mm (P = 0.03, ANCOVA).

The proportion of patients who underwent marked cervical shortening in the studied interval was also assessed (Table 2). In the history of preterm birth population, nine patients in the progesterone group (3.3%) and 19 patients in the placebo group (6.9%) had a cervical length ≤ 15 mm at the second transvaginal examination (P = 0.079). Forty-six patients (17%) and 63 patients (23%) had a cervical length ≤ 25 mm at the second assessment in the progesterone and placebo groups, respectively (P = 0.087).

Table 2. Cervical length at the 28-week examination
Cervical lengthProgesterone groupPlacebo groupPOdds ratio (95% CI)
  • Values are n (%).

  • *

    Patients with cervical length ≤ 30 mm at randomization at 18–22 weeks (most patients in this population also had a history of preterm birth).

  • Assigning cervical length at 28-week scan as 0 mm for those who delivered after randomization but before 28 weeks.

  • Fisher's exact test.

≤ 25 mm    
 History of preterm birth (n = 547)46/273 (17)63/274 (23)0.0870.73 (0.52–1.03)
 Premature cervical shortening* (n = 104)14/54 (26)21/50 (42)0.100.62 (0.35–1.08)
 Modified premature cervical shortening (n = 110)15/55 (27)26/55 (47)0.0480.58 (0.35–0.96)
≤ 15 mm    
 History of preterm birth (n = 547)9/273 (3)19/274 (7)0.0790.48 (0.22–1.03)
 Premature cervical shortening* (n = 104)3/54 (5.6)7/50 (14)0.190.40 (0.11–1.45)
 Modified premature cervical shortening (n = 110)4/55 (7.3)12/55 (22)0.0560.33 (0.11–0.97)
≥ 50% decrease    
 History of preterm birth (n = 547)27/273 (10)35/274 (13)0.350.77 (0.48–1.24)
 Premature cervical shortening* (n = 104)1/54 (1.9)5/50 (10)0.100.19 (0.02–1.53)
 Modified premature cervical shortening (n = 110)2/55 (3.6)10/55 (18)0.0290.20 (0.05–0.87)

To evaluate the frequency of relative change in cervical length, we also assessed whether a ≥ 50% decrease in cervical length occurred over the observation interval from randomization to 28 weeks. This percentage cervical change was considered more representative of a true progression for a worsening cervical status in the women with premature cervical shortening at baseline. Within this subpopulation, we noted one patient in the progesterone group (1.9%) and five in the placebo group (10%) with this degree of cervical change (P = 0.10). Three progesterone-treated patients (5.6%) in this population had evidence of cervical change to ≤ 15 mm at 28 weeks compared with seven (14%) of those given placebo (P = 0.19).

Finally, the adjusted analysis for the modified premature cervical shortening group attempted to compensate for the increased frequency of extremely preterm birth in the placebo-treated group by assigning these patients a value of zero for follow-up. With this adjustment, a significant difference was observed in the frequency of women having a ≥ 50% change between the interval from randomization to 28 weeks or delivery: 3.6% in the progesterone group vs. 18% in the placebo group (P = 0.029) (Table 2).


We identified a significant treatment effect with preservation of cervical length at 28 weeks in women administered intravaginal progesterone. This effect was greater as an absolute value and as a percentage of initial cervical length in those with premature cervical shortening. We also noted that fewer treated patients progressed to a cervical measurement indicative of marked cervical shortening. This pharmacodynamic effect may explain, in part, the drug's mechanism for efficacy. These data are complementary to those in our previous analysis, which found significantly improved outcomes with progesterone therapy occurring in women with a shortened cervix in the mid-trimester. Therefore, our findings support the use of transvaginal cervical ultrasound imaging not only to identify a patient population that may benefit from treatment, but also this tool may potentially identify a treatment response or failure.

What is remarkable regarding the current pharmacodynamic finding is the ability to identify significant differences between groups even though the majority of patients delivered at term. Therefore, despite the absence of a pathophysiology in the majority of patients, an identifiable treatment effect was still appreciated and the magnitude of this effect may be greater if a different population was assessed. This evidence, however, cannot definitely prove that preservation of cervical length is primarily responsible for improved outcomes in those at risk for preterm birth treated with progesterone. We observed that approximately 5% of progesterone-treated patients with a history of preterm birth progressed to a cervical length ≤ 15 mm at 28 weeks' gestation and approximately 5% of treated patients with premature cervical shortening at randomization had progressive premature shortening ≥ 50% of their original cervical length. Finally, although these observations were more frequent for placebo-treated patients, differentiating progressive cervical shortening associated with a pathophysiology vs. normal physiological cervical change presumably will be more difficult later in gestation and may require adjuvant biochemical testing such as fetal fibronectin or markers for inflammation.

Cervical length in the present study was assessed at two time points over an approximate 8-week interval between the middle of the mid-trimester and the start of the third trimester. An accelerated rate of change in cervical length during the mid-trimester has been associated with an increased risk for preterm birth in women undergoing transvaginal cervical surveillance and therefore we believe our observations are important23. Reducing the rate of cervical change during this interval presumably reduces the frequency of subsequent myometrial activation leading to preterm birth.

This mechanism for benefit with treatment of preserving cervical length is consistent with the observations of Facchinetti et al., who assessed periodic cervical change after exposure to supplemental progestin and noted a significant reduction in the rate of cervical change with treatment5. These investigators performed a randomized trial in symptomatic women who were successfully tocolyzed, mostly in the third trimester. They noted a reduced rate of cervical shortening and an increased latency to delivery in patients treated with progestin after initial tocolysis, most of whom had a cervical length ≤ 25 mm at randomization. Preservation of cervical length in either the second or early third trimesters being associated with a greater latency to delivery is the common observation between these investigations. Further studies are needed, however, to validate these findings.

Cervical ripening and effacement is influenced by multiple factors including nitric oxide synthesis, prostaglandin synthesis and the presence of cytokines. Inflammatory mediators can act as neutrophil chemoattractants and act to alter the activity of matrix metalloproteinases, whose combined actions can prematurely distort the cervical anatomy24. Several investigations have provided evidence of the anti-inflammatory effect of progesterone inhibiting COX-2 gene expression as well as its ability to downregulate proteolytic activity or upregulate mechanisms enhancing cervical stromal integrity6, 7. Detailed histological evidence regarding cervical response to progesterone antagonists confirms that progesterone exerts marked control over this process of cervical stromal collagen degradation in several species8, 9. From an evolutionary biological perspective, a preferential difference between the cervix and myometrium in response to a functional progesterone withdrawal has an intuitive rationale. With cervical ripening being the preface to active labor, a circumstance is created in which labor can occur more efficiently, which has survival advantages. Numerous differences in gene expression between the cervix, lower segment and fundal myometrium are associated with the onset of labor25.

Xu et al. and Marx et al. have raised the question of whether the cervix is the predominant site of supplemental progesterone action, and we believe that these clinical data along with basic scientific findings suggest that the answer may be yes6, 7. However, this hormonal strategy may have several other mechanisms that improve outcome (Table 3). Only large prospective studies repeatedly evaluating the numerous possible end-effects for supplemental progesterone, including markers for inflammation, biomarkers for decidua–membrane disruption and myometrial activity, in addition to serial assessment of cervical composition and length, will be able to definitively identify the most predominant mechanism(s) underlying this treatment strategy.

Table 3. Potential sites and mechanisms of action in the reproductive tract for supplemental progesterone to reduce the frequency of preterm birth
Site of actionMechanism of actionPartial reference list
  1. CRH, corticotropin releasing hormone.

CervixReduce cervical stromal degradation5,6,7,8,9
 Alter barrier to ascending infection/inflammation10
MyometriumReduce contraction frequency11,12,13,14
DeciduaAttenuate response to hemorrhage/inflammation15,16
Fetal membranes/placentaSuppress prostaglandin synthesis, reduce apoptosis16,17
 Alter CRH gene expression18,19
 Alter estrogen synthesis20
FetusAlter fetal endocrine-mediated effects21

The landmark study by Fonseca et al. identified a significant reduction in preterm birth at < 34 weeks with intravaginal progesterone in patients treated based on premature cervical shortening as the indication for therapy3. The findings from our study are consistent with this observation and the present investigation suggests that intravaginal progesterone therapy limits further cervical shortening in these patients4. Progesterone therapy, indicated for premature cervical shortening, has the potential to more substantially affect the rate of preterm birth compared with treatment indicated for history alone26, 27. Obstetric history may best be used to modify thresholds for intervention. We believe that utilizing objective data and monitoring cervical length will yield more consistent and rewarding results for preventing preterm birth, especially if we understand the pharmacodynamic effects of treatment and the frequency with which patients may not respond to this promising intervention.


This study was funded by Columbia Laboratories, Inc., Livingston, NJ, USA, the manufacturers of Prochieve® 8% progesterone vaginal gel.


Collaborators in the Progesterone Vaginal Gel Study Group:

S. Fusey (India), P. Soma-Pilla (South Africa), K. Porter (Mobile, AL, USA), R. Schackis (South Africa), D. Eller (Atlanta, GA, USA), Y. Trivedi (India), G. VanBuren (Cleveland, OH, USA), M. Khandelwal (Camden, NJ, USA), K. Trofatter (Greenville, SC, USA), D. Vidyadhari (India), J. Vijayaraghavani (India), J. Weeks (Louisville, KY, USA), B. Dattel (Norfolk, VA, USA), E. Newton (Greenville, SC, USA), C. Chazotte (Bronx, NY, USA), G. Valenzuela (Colton, CA, USA), P. Calda (Czech Republic), W. Hansen (Lexington, KY, USA), M. Newman (Baton Rouge, LA, USA), B. Rosenn (New York, NY, USA), S. Dabak (India), L. Parker, (Winston-Salem, NC, USA), J. Stern (Memphis, TN, USA), L. Bayer-Zwirello (Boston, MA, USA), L. Cousins (San Diego, CA, USA), A. Kekre (India), R. McDuffie, (Denver, CO, USA), J. Schucker (Danville, PA, USA), C. Barrera (Chile), C. Goldberg (Tucson, AZ, USA), A. Jiratko (Czech Republic), K. Swenson (Austin, TX, USA), A. Evans (Cincinnati, OH, USA), G. Gross (St Louis, MO, USA), M. Short (Baltimore, MD, USA), S. Sunderji (Toledo, OH, USA), R. Artal (St Louis, MO, USA), M. Binstock (Bedford, OH, USA), J. Hibbard (Chicago, IL, USA), R. Kelly (Odessa, TX, USA), X. Sandovol-Lopez (El Salvador), L. Smith (Livingston, NJ, USA), M. Stitley (Morgantown, WV, USA), E. Wang (Chicago, IL, USA), M. Beall (Torrence, CA, USA), J. Carvajal (Chile), V. Rappaport (Albuquerque, NM, USA), L. Wilkins-Haug (Boston, MA, USA), B. Sibai (Cincinnati, OH, USA).