2.1.3 Exclusion criteria
1) Non-English or Chinese literatures; 2) only the most recent article
was included when two or more articles reporting a same study; 3)
studies unavailable in full-text; 4) studies whose data were inadequate
or insufficient to estimate the outcomes; 5) studies involving patients
at risk for premature rupture of membranes, intrauterine infection,
monochorionic twin-twin transfusion syndrome, fetal anomalies or
chromosomal abnormalities, or those who had regular contractions and
unexplained active vaginal bleeding; and 6) studies involving patients
who had allergy or contraindications to the cervical interventions used.
2.2 Study selection and
data extraction
Study selection and data extraction was conducted by two reviewers
independently. First, they screened the title and abstract for
potentially relevant studies. Next, they read through the entire study
to evaluate its eligibility for inclusion. Then, resultant studies from
the two reviewers were compared; any discrepancies between the two
reviewers were resolved through discussion with a third reviewer. Data
were extracted from each study according to the data extraction table.
If there was any missing data, the corresponding author was contacted
via email or telephone call to obtain the required information. The data
extracted comprised the following: 1) basic information such as article
title, study type, name of the first author and year of publication; 2)
subject baseline characteristics and different
interventions;
3) study design and essential parameters for risk of bias assessment;
and 4) outcome data.
2.3 Assessment of risk of
bias
Two reviewers independently assessed the quality of each included study,
and any disagreements between the two reviewers were resolved by
consensus. The Cochrane risk-of-bias tool [6]was
utilized to evaluate the included RCTs. For non-RCTs, the methodological
index for non-randomized studies (MINORS) [8]was
used to assess risk of bias.
2.4 Statistical analysis
We used Review Manager 5.3 (The Cochrane Collaboration, Oxford, United
Kingdom) to perform meta-analysis. Continuous variables were presented
as mean differences (MD), while categorical variables were reported as
relative risks (RR). For each outcome measure, point estimates and 95%
confidence intervals (95% CI) were also calculated. Heterogeneity
across studies was evaluated using the I2 statistic.
Statistical significance was set at α = 0.1. If P> 0.1 and
I2< 50%, the heterogeneity among studies
was not significant, so that a fixed effect model was used; otherwise, a
random-effect model was applied. Sensitivity analysis was used to detect
the source of heterogeneity by excluding each study based on the
leave-one-out method. In all meta-analyses, figures with a p-value less
than 0.05 were considered statistically significant.
3. Results
3.1 Study selection
A total of 322 relevant studies were initially retrieved. After careful
screening, 9 studies [9-17] met the eligibility
criteria including 2 RCTs [9,11] and 7 non-RCTs[10, 12-17] .Totally 1174 patients with the threat
of preterm birth were included for analysis, 693 in the cervical pessary
group and 481 in the cervical cerclage group. The search process and the
number of studies obtained in each step were presented in a flow chart
in Figure 1-1.
3.2 Basic characteristics and
quality assessment
Table 1-1 presented the basic characteristics of the included studies.
The methodological quality assessment of non-RCTs (including cohort
studies) was presented in Table 1-2. And Table 1-3 presented the
assessment of risk of bias for the included
RCTs.
3.3 Results of meta-analysis