Amnioinfusion vs. standard management for the second trimester PPROM: a systematic review and meta-analysis of observational studies and RCTs

Abstract Objective This meta-analysis aims to review the effect of serial transabdominal amnioinfusion (TAI) on short-term and long-term perinatal outcomes in mid-trimester preterm premature rupture of membranes (PPROM). Methods Literature searches of PubMed, Web of Sciences, Scopus, and Cochrane Library were performed from their inception to April 2022. Studies comparing conventional treatment with serial TAI in women with proven PPROM at less than 26 + 0 weeks of gestation with oligohydramnios were included. Studies that included oligohydramnios due to other reasons such as fetal growth retardation or renal anomalies were excluded. Risk of bias in observational studies was assessed using the tool of the Cochrane Review group identified as risk of bias in non-randomized studies – of interventions. The risk of bias assessments for RCTs were performed according to the Cochrane risk-of-bias tool for randomized trials. An I2 score was used to assess the heterogeneity of included studies. The analyses were performed by using random-effect model, and the results were expressed as relative risk (RR) or mean difference with 95% confidence intervals (CIs). Results Overall, eight relevant studies including five observational studies (n = 252; 130 women allocated to the intervention) and three RCTs (n = 183; 93 women allocated to the intervention) were eligible. The pooled latency period was 21.9 days (95% CI, 13.1–30.8) and 5.8 days (95% CI, −11.6–23.2) longer in the TAI group in the observational studies and RCTs, respectively. The perinatal mortality rate reduced in the intervention group when tested in observational studies (RR 0.68; 95% CI, 0.51–0.92), but not in RCTs (RR 0.79; 95% CI, 0.56–1.13). The rate of long-term healthy survival was higher in the children whose mothers were treated with the TAI (35.7%) than those were treated with the standard management (28.6%) (RR 1.30, 95% CI 0.47–3.60, “best case scenario”). Conclusions The efficacy of serial TA on early PPROM associated morbidity and mortality is not attested. Additional randomized control trials with adequate power are needed.


Introduction
Mid-trimester preterm premature rupture of membranes (PPROM) is an occasional clinical condition, occurring in approximately 0.3-0.4% of pregnancies [1].Delivery usually occurs in 38-70% of pregnancies with PPROM within the first week of PPROM, leading to the birth of periviable fetuses [2,3].The ongoing pregnancies with oligohydramnios at the second trimester led to impairment of fetal lung development, which may eventually result in pulmonary hypoplasia [4].
The clinical management options for women with PPROM include either expectant management, including the combination of prophylactic antibiotics with corticosteroids, or immediate delivery.Prolongation of latency and advanced gestational age may lead to a reduction in the potential risk of neonatal death and life-threatening morbidity associated with PPROM [1].Despite the development in perinatal medicine, the rate of healthy survival remains low in PPROM before 25 weeks of gestation, reported as around 10% [2].
In the last decade, attempts at restoring the amniotic fluid to reduce the rates of neonatal mortality and morbidity in women with PPROM were undertaken with various methods, such as the sealing of membranes or amnioinfusion [5].Evidence from observational studies suggests that replenishing the amniotic fluid volume is a potential management option that reduces the risks associated with reduced amniotic fluid, prolongs the pregnancy period, and ultimately increases the rate of neonatal survival [6][7][8][9].However, placental abruption, cord prolapse, and fetal damage due to puncture are all risks associated with the transabdominal amnioinfusion (TAI) procedure.Surprisingly, TAI was shown to be somewhat beneficial and had no obvious effects on the histologic characteristics of the amnion or umbilical cord [10].
Recent meta-analysis on the efficacy of amnioinfusion in women with PPROM found that the rate of perinatal mortality and pulmonary hypoplasia decreased in the observational studies, but no differences in RCTs [11].However, this meta-analysis included heterogeneous studies involving protocols regarding the interventions performed at the various gestational ages.Despite the comparatively high prevalence of the condition, controversy has remained with respect to the optimal approach.The additional findings emerging from both observational studies and RCTs provided more information in the light of this controversy.Therefore, this meta-analysis aims to assess the effects of amnioinfusion for mid-trimester PPROM on perinatal morbidity and mortality.

Methods
The meta-analysis was conducted in accordance with a prospectively prepared protocol and reported in accordance with Preferred Reporting Items for Systematic Reviews and Meta-Analyses statement [12].The study is registered with PROSPERO (no: CRD42021262620).The PRISMA Checklist is provided as a supplementary file.

Data sources and search strategy
A comprehensive literature search was conducted from the inception of the project to May 2022 with the assistance of an experienced librarian.The search was performed using PubMed, the Web of Sciences, Scopus, MEDLINE, ClinicalTrials, and the Cochrane Library as electronic databases.The trials were identified by using a combination of the following MESH terms: fetal membrane, premature rupture, premature rupture of membranes, amnioinfusion, amniotic, and amniotic fluid.First, the reviewers (E.C., S.G.C., and R.M.) independently reviewed abstracts and the titles of relevant full texts for their eligibility.Any conflicts were resolved by consensus.No language restrictions were applied.

Study selection and data collection
Both comparative observational studies and RCTs conducted on comparison of TAI and standard management in pregnant women with PPROM and oligohydramnios <26 þ 0-weeks' gestation were included.Case reports, case series, and abstract publications were not considered eligible for inclusion.Studies with confirmed PPROM-related oligohydramnios were included.Studies that included oligohydramnios due to other reasons (fetal growth retardation, renal anomalies) were excluded.Studies used transabdominal catheters for amnioinfusion were excluded [13][14][15].
Two reviewers (S.G.C. and A.B.Y.) independently evaluated the published reports against inclusion criteria by using a standardized data collection form.Agreement about potential relevance was reached by consensus, and full-text copies of that research were obtained.Two reviewers (S.G.C. and A.B.Y.) independently extracted data regarding study characteristics and outcomes.Disagreements were resolved by consensus by the other reviewers after discussion (E.C. and R.M.).

Outcome measures
The primary outcomes of interest were the latency period (interval from PPROM until delivery) and perinatal mortality.Secondary outcomes were pulmonary hypoplasia, neonatal death, gestational age at birth, chorioamnionitis, early-onset neonatal sepsis, bronchopulmonary dysplasia (BPD), intraventricular hemorrhage (IVH), necrotizing enterocolitis (NEC), neonatal composite outcomes, and cesarean delivery.
Long-term outcomes, including respiratory problems and neurodevelopmental delays, were reported by two RCTs [16,17].The AMIPROM trial performed the follow-up assessment of infants at 6, 12, and 18 months of age by using Bayley's Scale of Infant Development II (BSID-II) assessment for neurodevelopmental delay [16].The combination of validated respiratory questionnaires with infant lung function tests was used for assessment of respiratory morbidity [16].In the AMIPROM trial [16], BSID-II assessment was performed ranging from ages of 2 years 3 months to 3 years 3 months.The PROMEXIL-III trial for follow-up assessment up to 5 years of corrected age, used Bayley-III-NL or Wechsler Preschool and the Primary Scale of Intelligence-3rd edn.Dutch version for neurodevelopmental delay and a parental respiratory questionnaire defining the symptoms interfering with daily activities for respiratory problems [18].Two authors (E.C. and A.B.Y.) independently summarized the relevant data from selected studies.

Assessment of risk of bias
Three investigators (A.B.Y., E.C., and S.G.C.) independently assessed the risk of bias.Risk of bias in observational studies was assessed using the tool of the Cochrane Review group identified as risk of bias in non-randomized studiesof interventions (ROBINS-I) [19].For observational studies, the domains of assessment included these sections: confounding, selection, classification, and deviation of interventions, data, outcome assessment; and selection.The risk of bias assessments for RCTs were performed according to the Cochrane risk-of-bias tool for randomized trials, which included five domains: the randomization process, deviation from the intended intervention, missing outcome data, measurement of outcomes, and selection of reported results [20].Any disagreement regarding the quality assessments was resolved by consensus.

Statistical analysis
Data analyses were performed by using the RevMan software (version 5.4.1;The Nordic Cochrane Centre, Cochrane Collaboration, Copenhagen, Denmark).The observational studies and RCTs were separately assessed.Statistical heterogeneity between studies was assessed using Higgins I 2 statistics [21].The random effects model as proposed by the Cochrane Handbook for Systematic Reviews of Interventions was used [22].Dichotomous outcome measures are presented as Mantel-Haenszel adjusted relative risk (RR) with 95% confidence interval (CI).Continuous outcome measures are presented as a mean difference with a 95% CI.

Study selection and characteristics
The process of identification and selection of studies is presented in Figure 1.A total of 2026 citations were yielded by the initial searches.After screening titles and abstracts, 25 articles were retrieved for full-text review.Six studies were excluded because gestational age was more than 26 weeks' gestation at the time of the intervention [23][24][25][26][27][28].Five studies were excluded because either the inclusion criteria were not fulfilled, or the port was used for amnioinfusions [13][14][15]25,29].Six studies were excluded due to non-matched case and control groups [6,8,[30][31][32].The remaining eight studies, including a total of 464 women with PPROM, met the inclusion criteria; 212 women (122 in the observational studies and 90 in the RCTs) were allocated to the standard management, and 252 women (130 in the observational studies and 93 in the RCTs) were treated with a combination of TAI and standard management [7,9,10,16,17,[33][34][35].In the quasi-randomized study, the patients were admitted by chance to one of two departments as one department provided the standard management, and the other department provided a combination of standard management with serial TAI after consenting patients [34].Therefore, after discussion among us, we decided to include the quasi-randomized trial with the other RCTs.
The baseline characteristics of the included studies in the systematic review are presented in Supplementary Table 1.The diagnosis of rupture of membranes was made in all studies by sterile speculum examination with observation of pooling of amniotic fluid in the posterior fornix.Additionally, the nitrazine test was used in three studies [7,9,17], Amnisure or Amniocator in two study [17,33] and the demonstration of oligohydramnios by ultrasonography was used for the confirmation of PPROM in all studies [7,9,16,17,[32][33][34][35].Hospital bed rest, prophylactic antibiotic therapy, and corticosteroid administration were used as the standard management in all studies [7,9,10,16,17,[32][33][34][35].In three studies, tocolysis was administered when uterine contractions occurred in women with no sign of clinical chorioamnionitis or placental abruption [7,9,10,16,17,[32][33][34][35].Prophylactic tocolysis was administered as a preventive measure for all patients, regardless of the presence of uterine contractions in two studies [34,35].Targeted antibiotic therapy according to cervical and vaginal cultures was administered in three studies [7,9,10,16,17,[32][33][34][35].The average number of procedures per patient varied from 2 to 3 in PROMEXIL-III trial, and Ogunyemi and Thompson, ranging from 0 to 9 in Vergani and AMIPROM trial [7,9,16,17].In Ogunyemi and Thompson trial, the results of one patient were excluded from the analysis of each arm because these patients were recruited after 26 þ 0 weeks' gestation [7].De Santis et al. and De Carolis et al. indicated a mean number of procedures as 3.8 and 4.0, respectively [34,35].The failure rate per procedure was 2.73 ranging from 1.2 to 4.6 in the AMIPROM trial [16].The procedure was unsuccessful in four cases of 147 amnioinfusions in De Santis et al. and in one case in PROMEXIL-III studies [17,34].
Six minor maternal complications occurred during the procedure in the PROMEXIL-III trial: a small amount of fluid infusion into the myometrium in one case, painless contractions leading to the cessation of the procedure in one patient, severe abdominal pain in one patient after the procedure, and postprocedural vaginal bleeding in three patients [17].Two studies reported that onset of labor occurred within postprocedural 24 h in one patient [34,35].Vergani et al. and De Santis et al. reported on the cord prolapses that occurred in one patient and two patients, respectively [9,34].
Two studies reported on fetal complications that occurred during procedure [7,17].The PROMEXIL-III trial reported on four fetal complications: a fetal death less than 30 min following amnioinfusion and fetal puncture without postnatal sequelae in three cases: two cases with intracutaneous and one case with intraperitoneal amniotic fluid injection [17].Ogunyemi and Thompson also reported on two fetal injuries during the procedure: a 2 cm leg laceration that was sutured after birth and a 0.5 cm superficial chest scar that was recovered spontaneously [7].
In the AMIPROM trial, six women did not receive the intervention: one had a termination of pregnancy; one declined amnioinfusion in the post-randomization; the deepest pool level was sustained at around 2 cm throughout pregnancy in four women [16].In PROMEXIL-III trial, seven women did not have the intervention: two women delivered before the first procedure; two women had a termination of pregnancy due to lethal anomaly; one woman had septic shock before the intervention; and one woman had a failure of the procedure due to technical problems [17].Two women in each group were opted for termination of pregnancy [17].

Risk of bias of included studies
Assessment of qualities for observational studies and RCTs is presented in Supplementary Figure 1, respectively.Five included observational studies were regarded as high risk of bias because of recruitment of participants and deviations from intended intervention.The quality assessment of the evidence is presented in Supplementary Table 2. GRADE evidence profiles, including both the quality of the evidence and the results of the evidence synthesis for each outcome, were created using GRADEprofiler (GRADEpro) software (Version 3.6) and presented in Supplementary Table 3.

Primary outcomes
The latency period was prolonged, but it did not reach a statistically significant level (MD 5.82; 95% CI, À11.58 to 23.23; heterogeneity: I 2 ¼ 91%; p value ¼ .51; Figure 3(A)) [16,17,34].There was not a statistically significant difference in overall perinatal mortality rate between the groups (three studies; RR Figure 2. Effects of amnioinfusion on neonatal outcomes for observational studies.Forest plot of the results of the meta-analysis of observational studies for (A) latency period length, which displays the difference in latency period lengths between the amnioinfusion and standard management groups; (B) perinatal mortality rates refer to the risk ratio."if all children lost to follow up are unhealthy".

Long-term outcomes
The rate of healthy children without either respiratory problems or neurodevelopmental delays (defined as "long-term healthy survivor") was 16.1% (10/56) in the intervention group and 12.5% (7/56) in the standard management group (two studies: RR 1.37; 95% CI, 0.55-3.4;heterogeneity: I 2 ¼ 0%; p value ¼ .5; in best case scenario, defined as if all children lost to followup are healthy) [16,18].The rate of long-term healthy survivor in the "worse-case scenario" (defined as if all children lost to follow-up are unhealthy) was not different between groups (two studies: RR 3.  1).

Principle findings
The findings from this meta-analysis demonstrated that either the perinatal mortality rate or latency period did not improve in women who underwent the TAI in RCT.However, the pregnant women with early PPROM who underwent successful amnioinfusion had a prolongation of the latency period and lower perinatal mortality rate when tested in the observational studies.The intervention group also experienced less pulmonary hypoplasia in the observational studies.

Strengths and limitations
The strength of this meta-analysis is that it is the first meta-analysis, including the studies conducted on amnioinfusion performed at an early gestational age, which is a serious condition due to its poor prognosis.
Given that most of the data on amnioinfusion in PPROM arises from observational studies, our metaanalysis is relatively homogenous due to the application of strict inclusion criteria.The limitations of this review consist of heterogeneity in the quality of studies regarding the high risk of biases and assessment of outcomes.Another limitation of this review is the low number of participants, both in observational studies and RCTs.Finally, there was heterogeneity in terms of the protocol and assessment of outcomes.

Comparison with existing literature
Mid-trimester PPROM remains a challenging clinical problem in the management of patients.PPROM at mid-trimester of pregnancy is a key contributing factor to perinatal mortality and adverse perinatal outcomes [36].While therapeutic aims in women with PPROM concentrate on decreasing neonatal mortality and morbidity, early PPROM is a management and counseling dilemma due to the high risk of pulmonary hypoplasia.Data from a meta-analysis of RCTs found that the intervention did not differ the pulmonary hypoplasia perinatal and neonatal mortality rates; thus, the amnioinfusion is not a reasonable option as routine management.
It is anticipated that infection/inflammation will lead to the activation of uterine contractions during labor [37].Three mechanisms are hypothesized for amnioinfusion's potential benefits: dilution of preexisting intra-amniotic bacteria, cleaning through recirculation of fluid, diluting the inflammatory mediators, increasing amniotic fluid volume, and intraamniotic pressure.The addition of fluid is presumed to have a beneficial effect in prolonging the latency by diluting the preexisting intraamniotic bacteria and inflammatory cells [27].Choriodecidual infection may be an underlying etiology or consequence of PPROM.Recent studies demonstrated that histological chorioamnionitis is associated with the risk of adverse neurodevelopmental outcomes, particularly in extremely preterm infants [38,39].In our meta-analysis, there were trends toward a decline in the frequency of chorioamnionitis in women who underwent amnioinfusion, in accordance with the previous meta-analysis [11].
Although the rate of healthy children was roughly four times higher in the intervention group than in the standard treatment group, it was found to be insignificant, which is possible because there were few absolute numbers of children in the follow-up assessments.Hatfield et al. reported that chorioamnionitis had the potential to change the path of brain development in children born prematurely, resulting in long-term adverse outcomes [40].It has been assumed that flushing inflammatory cells and preexisting bacteria may have the potential effect of reducing neurologic damage.By virtue of the small number of participating children, we are cautious while interpreting the results.Thus, large studies with adequate power in which long-term follow-up is assessed are desirable to demonstrate the impact of intrauterine infection, particularly regarding neurological development and outcomes.
The results of observational studies demonstrated a longer latency between PPROM and delivery (nearly 22 days) in the intervention group.Data from RCTs also found that the latency period was nearly þ5 days longer in the intervention group, albeit with no statistically significant difference.Nonetheless, a week increment in latency increases the rate of neonatal survival by nearly 10-20% in periviable fetuses [41].This is of primary importance because increased latency may allow a fetus to reach the viability threshold determined by the local hospital to be managed.Given the limited options for the management of PPROM, most patients desperately opt for termination of pregnancy.However, existing evidence is still limited to offer this intervention as a standard management option for such patients.

Conclusions and implications
The results of meta-analysis from observational studies assert that amnioinfusion ameliorated the short-term pregnancy outcomes, but data from RCTs did not show any significant differences in the short-term analysis.This meta-analysis demonstrated that there were no significant differences in perinatal mortality rate and latency period when compared to women treated with standard care in RCTs.Given limited options for the management of PPROM, most patients desperately opt for termination of pregnancy.The clinicians should be cautious while interpreting the findings from this metaanalysis.If stratified large RCTs according to gestational age identifies a threshold in which the amnioinfusion is superior to the standard management for benefits, this intervention would be an option for such patients.The findings warrant an appropriately powered RCT with multicenter collaboration to indicate the utility of such intervention in routine care.

Figure 1 .
Figure 1.PRISMA 2020 flow diagram of selection process for included studies.

Figure 3 .
Figure3.Effects of amnioinfusion on neonatal outcomes for randomized controlled trials.Forest plot of the results of the metaanalysis of randomized control trials for (A) latency period length, which displays the difference in latency period lengths between the amnioinfusion and standard management groups; (B) perinatal mortality rates refer to the risk ratio.

Table 1 .
The effects of intervention on the secondary outcomes.
GA: gestational age; PPROM: preterm premature rupture of membranes; N/A: not applicable; RR: risk ratio.Ã refers to P value < .05. a Defined as: no long-term neurodevelopmental delay or no respiratory problems.Worst case scenario defined as: