Study design and participants

This multicenter, open-label RCT took place from November 2014 through October 2018 in 20 centers in France. Women with an asymptomatic monochorionic or dichorionic twin pregnancy between 16 ^+0d and 24 ^+0d weeks’ gestation and a cervical length (CL) <35 mm (ie, the 25th percentile at 22 weeks’ GA ) were asked to participate and were randomly allocated to standard care with or without an Arabin pessary in a 1:1 ratio. A vaginal swab was used to collect samples for bacteriologic examination in both groups. An accredited sonographer performed all CL measurements by transvaginal sonography. Accreditation required sonographers to provide 3 CL measurements to verify their quality according to the criteria described by Iams et al. Exclusion criteria included monoamniotic twins, placenta previa, uterine malformation, previous conization, chronic maternal diseases (type 1 and type 2 diabetes, ongoing heparin therapy, chronic kidney disease, and hypertension), cerclage, progesterone treatment at inclusion, twin-to-twin transfusion syndrome, major fetal abnormalities, and fetal weight discordance exceeding 40%. All women consented to participate in writing after information was provided. Because it is required by French law, the consent of fathers was sought for neonatal data collection. The appropriate French ethics committee (2014-A00616-41) approved this study. The trial was registered under NCT02328989 with Clinicaltrials.gov .

Randomization and masking

We used a web-based randomization system with varying random block sizes, stratified by GA (2 periods: 16 ^+0d to 19 ^+6d weeks and 20 ^+0d to 24 ^+0d weeks). The allocation sequence was generated by a statistician and concealed from the obstetricians and clinicians enrolling the patients. Given the nature of the intervention, participants and obstetricians were aware of the allocated intervention. Neonatal outcome assessors and data analysts were, however, blinded to it.

Study procedures

We used a single size Arabin (ASQ type) pessary (Dr Arabin GmbH and Co KG, Witten, Germany) of 65×25×32 mm (ie, lower larger diameter, height, and upper smaller diameter, respectively). Before pessary insertion, a vaginal swab was used to collect samples for bacteriologic examination. An obstetrician inserted the pessary in an outpatient procedure. A video about this insertion, produced by the PECEP-Twins trial Group at the Val d’Hebron, was provided to participating obstetricians to view before the study started. Investigators also received written guidance on pessary insertion, management, and removal, which was done during a vaginal examination at 36 weeks’ gestation in asymptomatic patients. Indications for earlier removal were severe patient discomfort, persistent signs of preterm labor despite tocolysis, active vaginal bleeding, and any sign of chorioamnionitis. Preterm premature rupture of membranes (pPROM) was not an indication itself for pessary removal.

Women in both groups received standard care based on the 2009 French College of Obstetrics and Gynecology guidelines for the management of twin pregnancies. Any adverse effects were recorded. Preterm labor was defined as painful uterine contractions with a shortened CL. Tocolysis and corticosteroids were considered before 34 weeks’ gestation; atosiban was the first-line tocolytic treatment for preterm labor. Neither vaginal progesterone nor cervical cerclage after randomization was permitted under the protocol but they were systematically recorded in the data set if they occurred. All outcome measurements were collected at delivery and during the neonatal period, extracted from the hospital records, and entered into a standardized electronic case report form. The centers were monitored regularly throughout the study. A clinical research associate (CRA) performed quality control of screening and data processing and verified protocol compliance and adverse event reporting during regular visits. These took place during the study at a frequency that depended on the number of inclusions and at study conclusion. Moreover, the records of all women who gave birth before 34 weeks’ gestation were reviewed by the CRA to determine whether the birth was spontaneous or medically indicated. Spontaneous births included those for whom labor began spontaneously and those with PROM, regardless of whether labor was induced. The surveillance of the pregnant women and of the infants ended, respectively, at birth and at hospital discharge.

Study endpoints

The primary endpoint was a composite of any one of perinatal death (ie, medical termination of pregnancy, intrauterine fetal death, intrapartum death), neonatal death (within 28 days post-theoretical term), and significant neonatal morbidity until hospital discharge, defined as one or more of the following complications: bronchopulmonary dysplasia, intraventricular hemorrhage grade III to IV, periventricular leukomalacia, necrotizing enterocolitis grade II or higher, culture-proven sepsis, and retinopathy of prematurity requiring treatment. The primary endpoint, defined as at least 1 infant with an adverse perinatal outcome, was determined for each mother. Secondary endpoints included preterm birth (PTB) before 34 weeks’ gestation, time to delivery after inclusion in days, and pessary side effects (vaginal discharge, vaginal or uterine infection, early removal of pessary, pain with pessary, or cervical tear).

Sample size

Sample size calculations were based on the expected proportions of women with at least 1 infant with an adverse perinatal outcome. When we assumed a proportion of 25% and 12% in the control and pessary arms, respectively, based on the Dutch ProTWIN trial, 308 patients were required to have an overall power of 0.80 with a 2-sample binomial test using a continuity correction to consider the normal approximation to the binomial, with an equal sample size for both arms and an overall 2-sided type I error of 0.05.

Statistical analysis

Statistical analyses were based on the intention-to-treat (ITT) principle after multiple imputation of missing data. These analyses were complemented by a per-protocol (PP) analysis, excluding patients with prespecified major protocol violations, defined as violation of at least 1 eligibility criterion (pessary placement for patients randomized to the control group, no pessary placement for patients randomized to the pessary arm, cerclage and vaginal progesterone use, and women found not eligible after randomization). The PP analysis was based on the available data (without imputation). To test several assumptions about missing values, we performed unadjusted sensitivity analyses of the ITT population based on the best-case scenario (ie, each missing outcome treated as no adverse perinatal outcome) and the worst-case scenario (ie, each missing outcome treated as an adverse perinatal outcome for at least 1 infant). A post hoc analysis of the primary outcome was also performed on the subpopulation of pregnancies with a CL of <25 mm.

Preliminary analysis of missing data among the intent-to-treatsample

For 34 women (10.9% of enrolled pregnancies), the composite perinatal outcome was missing for both infants. Moreover, for 2 women, an adverse perinatal outcome was observed for 1 infant, whereas there were no data for the other twin in those pregnancies; these women were classified as having an adverse perinatal outcome ( Figure 1 ). Most often (27/34), missing information was because of a failure to obtain paternal consent to collect the neonatal data.

Flowchart of the PESSARONE trial

Maternal chronic disease, defined as diabetes type 1 or 2, ongoing heparin treatment, chronic kidney disease, or high blood pressure. The per-protocol analysis comprises patients without major protocol deviations and with data available for the primary endpoint

GA , gestational age (in weeks of gestation); PESSARONE , pessary to prevent prematurity in twins in case of short cervix.

Groussolles et al. Pessary, twin pregnancies, short cervix, preterm birth, adverse neonatal outcomes. Am J Obstet Gynecol 2022.

Analyses of the effects of the missing value mechanism were applied at a 10% level of significance because of their exploratory nature. Chi-square tests were used to examine the characteristics of patients with a missing primary endpoint (n=34) in comparison with patients (n=281) with known perinatal outcomes. We then explored whether factors that were unbalanced between individuals with complete and incomplete primary endpoints differed by randomization group. Specifically, we used logistic regression models including an interaction term the between randomization group and covariate as the independent variables and missing primary endpoint (yes/no) as the dependent variable. Finally, factors that were unbalanced between patients with a complete and those with incomplete primary endpoints were entered in the multiple imputation model as auxiliary variables.

Three factors were associated with a missing primary endpoint—mode of conception (85/281 vs 3/32 pregnancies obtained by in vitro fertilization [IVF]), history of late miscarriage between 16 and 22 weeks (6/281 vs 3/32), and steroids after inclusion (178/281 vs 25/29) ( Supplemental Table 1 ). GA at delivery also differed between the 2 groups: no patient with a missing primary endpoint gave birth before 24 weeks of gestation, thus all women delivering <24 weeks had a known adverse neonatal outcome or primary endpoint. Moreover, there was a significant interaction between GA and randomization groups in the logistic regression model predicting the probability of a missing primary endpoint.

Statistical approach

We described, by randomization group, the demographic and clinical characteristics of all randomized patients, the treatments administered after inclusion (cerclage, progesterone, steroids, and tocolysis), and distribution of the outcomes at birth and the neonatal morbidities among live births.

Log binomial models were used to estimate risk ratios (RRs), comparing the pessary and control arms with their 95% confidence intervals (CIs). Analysis of the multiple imputed data estimated the RRs between arms for each imputed data set and pooled them according to Rubin’s rules.

We estimated the crude (ie, unadjusted) RR for an adverse perinatal outcome for patients allocated to the pessary and control arms. RRs were then adjusted for GA at inclusion (stratification variable, 16–19 weeks and 20–23 weeks) and CL at inclusion, which were considered to be strongly related to the primary endpoint. Because unbalanced use of steroids and tocolytics after inclusion between the randomization groups in this unblinded trial could impact the estimate of the difference between the interventions, this model was further adjusted in a secondary analysis for administration of corticosteroids and/or tocolysis after inclusion (categorized into 3 groups: no corticosteroids [regardless of tocolytic status], corticosteroid therapy but no tocolysis, or both corticosteroid therapy and tocolysis). CL was dichotomized by using the cutoff (15 mm or 25 mm) that produced the model that fit the data best according to the P value of the overall model to test that all coefficients (excluding the constant) were equal to zero. The interaction terms between CL at inclusion and randomization group and between randomization group and corticosteroid administration or tocolysis after inclusion were tested.

Secondary endpoints were then analyzed. The proportion of preterm deliveries before 34 weeks’ gestation was compared between randomization groups. The proportion of births after spontaneous labor and before 34 weeks’ gestation was also compared between groups after exclusion of patients who gave birth after labor was induced before 34 weeks’ gestation for a non-PROM reason. These outcomes were compared between randomization groups by using log binomial models. Finally, the duration of pregnancy was described using Kaplan-Meier curves that showed the cumulative distribution of pregnancy durations and Cox proportional hazards models with delivery as the event. The proportion of preterm deliveries and of spontaneous preterm deliveries before 28 weeks and before 37 weeks were also reported by randomization group. Unadjusted analyses were completed by analyses adjusted for GA at inclusion and CL at inclusion (<25 mm or ≥25 mm). Interactions between CL and randomization group were tested for secondary outcomes.

Significance was defined as a P value of <.05.

Primary endpoint

At the infant level, an adverse perinatal outcome was observed for 45 infants in the pessary group (45/291, 23 missing infant-level outcomes) and for 54 infants in the control group (54/269, 47 missing infant-level outcomes).

The proportion of adverse perinatal outcomes for at least 1 infant was 16.8% in the pessary arm (95% CI, 10.7–22.9) and 22.5% in the control arm (95% CI, 15.7–29.4) (RR, 0.69; 95% CI, 0.39–1.23; P =.210) ( Table 2 ). In the model adjusted for GA at inclusion (16–19 weeks or 20–23 weeks), CL at inclusion (<25 mm vs ≥ 25 mm), and administration of corticosteroids and/or tocolysis after inclusion, no significant difference in adverse perinatal outcomes was observed (RR, 0.63; 95% CI, 0.34–1.19; P =.153). When restricted to the subgroup of pregnancies with a CL of <25 mm at inclusion, the randomization groups did not differ significantly: 23.6% of adverse perinatal outcomes for at least 1 infant (95% CI, 12.2–34.9) in the pessary group vs 32.4% (95% CI, 20.9–44.0) in the control group (RR, 0.64; 95% CI, 0.29–1.44).

Table 2

Analysis of the primary outcome composite of perinatal death ^a or significant neonatal morbidity until hospital discharge ^b , by randomization group, in the intention-to-treat population in the PESSARONE trial

Population	Pessary	Control
ITT analysis of complete data, n/N	25/146	32/135
%	17.1	23.7
ITT analysis (N=315, m=10)	n=157	n=158
%	16.8	22.5
95% CI	10.7–22.9	15.7–29.4
RR Pessary vs Control	0.69
95% CI	0.39–1.23
P value	.210
RR Pessary vs control adjusted c	0.73
95% CI	0.40–1.32
P value	.295
PP analysis (n=256), n/N	24/139	26/117
%	17.3	22.2
95% CI	11.4–24.6	15.1–30.8
RR Pessary vs Control	0.78
95% CI	0.47–1.28
P value	.320

Only gold members can continue reading. Log In or Register to continue

Share this:

• Click to share on Twitter (Opens in new window)
• Click to share on Facebook (Opens in new window)

Related

Related posts:

1. More than treatment refusal: a National Cancer Database analysis of adjuvant treatment refusal and racial survival disparities among women with endometrial cancer
2. Differences and similarities in endothelial and angiogenic profiles of preeclampsia and COVID-19 in pregnancy
3. Predictors of same-day discharge following benign minimally invasive hysterectomy
4. Mechanism of neonatal hypoglycemia after late preterm steroids: are fetal metabolic effects responsible?

Stay updated, free articles. Join our Telegram channel

Join