Evaluation of Preventive Antibiotic Regimens for Preterm Premature Rupture of Membranes: Efficacy and Safety of Triple Therapy versus Cephalosporin Monotherapy Authors: Chenjie Luo, Caihua Yang, Tianyue Zhou, Ping Zheng, Taoyan Lin, Xiangying Deng, Chao Sheng, Yilei Li Published: 2026 ABSTRACT PurposeControversy remains regarding the optimal prophylactic antibiotic regimen for preterm premature rupture of membranes (PPROM). This study compared the efficacy and benefits of cephalosporin monotherapy with a triple-therapy regimen comprising a cephalosporin, azithromycin, and clindamycin.Patients and MethodsWe conducted a single-center cohort study at a tertiary hospital in China from January 2024 to June 2025, prospectively enrolling pregnant women receiving triple therapy and retrospectively selecting those receiving cephalosporin monotherapy as controls. Maternal analyses included 32 patients in the triple-therapy group and 60 in the monotherapy group; neonatal analyses included 36 and 65 infants, respectively (twins analyzed separately). Inverse probability of treatment weighting was used to adjust for baseline confounding, propensity score matching and multiple regression were applied for sensitivity analysis. And subgroup analysis was stratified by gestational age.ResultsAfter adjusting, no significant between-group differences were observed in the primary maternal composite outcome (P=0.510) or neonatal respiratory distress syndrome (P=0.227). In exploratory secondary analyses, triple therapy was associated with lower observed risks of bronchopulmonary dysplasia(BPD) (P=0.010) and intraventricular hemorrhage(IVH) (P<0.001). In the 24–34-week gestation subgroup, it also prolonged gestational age versus the monotherapy regimen. No other secondary outcomes differed significantly. The conclusions of sensitivity analyses were basically consistent with the main analysis.ConclusionTriple therapy was not associated with significant differences in the primary maternal or neonatal outcomes compared with cephalosporin monotherapy. Exploratory findings suggested possible associations with a longer latency interval among women with PPROM at 24 - 34 - week of gestation subgroup and with lower observed risks of BPD and IVH. Given the limited sample size, small numbers of events, and nonrandomized design, these findings should be interpreted cautiously and confirmed in larger prospective studies. Introduction Preterm premature rupture of membranes (PPROM) refers to the spontaneous rupture of the fetal membranes that occurs before 37 weeks of gestation and prior to the onset of labor.1 Its incidence rate ranges from 2% to 4%.2–5 It is the most common cause of preterm birth6 and poses a serious threat to the health of both the mother and the fetus.7,8 For pregnant women, the primary risks are concurrent intrauterine infections such as chorioamnionitis.9 For neonates, PPROM frequently results in preterm birth, which in turn increases the risk of respiratory distress syndrome (RDS),2 and other prematurity-related complications, including bronchopulmonary dysplasia (BPD), intraventricular hemorrhage (IVH), neonatal sepsis, and infections. At present, the role of prophylactic antibiotic therapy in the management of PPROM has been widely recognized.10,11 Prophylactic use of antibiotics can significantly prolong of the latency interval from membrane rupture to delivery, reduce the incidence of maternal chorioamnionitis, and decrease the risk of neonatal infectious diseases.2,12 However, there remain significant controversies and evidence gaps in the selection of specific solutions.13 For example, the recommended protocol of the Maternal - Fetal Medicine Units Network in the United States is to administer ampicillin and erythromycin intravenously for 48 hours, followed by oral administration of amoxicillin and erythromycin for 5 days.9 The Society of Obstetricians and Gynaecologists of Canada recommends the use of macrolides (erythromycin, azithromycin, or clarithromycin) alone; or the combined use of amoxicillin/ampicillin and macrolides.14 In recent years, these controversies have predominantly arisen from shifts in microbial epidemiology. In the study by Wolf et al,15 among 1133 cases with positive culture results of postpartum placenta and amniotic membrane, 740 cases (65.3%) were identified as Gram - negative Enterobacteriaceae. The increasing proportion of Gram - negative bacterial infections poses challenges to the efficacy of traditional treatment regimens. Sgayer et al16 found through research that in the positive specimens of placental or amniotic fluid cultures from patients with PPROM, the maternal infection incidence rate in the ampicillin group (17.7% vs 6.5%; P value = 0.048) and the composite adverse neonatal outcomes (88.7% vs 74.2%; P value = 0.03) were both higher than those in the cefuroxime group. Enterobacteriaceae cultures were identified in 68.6% of the cultures in the ampicillin group and 43.2% of the cultures in the cefuroxime group. Schuler L et al17 proposed that Escherichia coli and coagulase-negative staphylococci are mainly resistant to aminomycin, and amoxicillin no longer appears to be the optimal treatment option. They recommended azithromycin instead. Additionally, some scholars18,19 suggested that a the triple regimen (such as ceftriaxone, clarithromycin, and metronidazole) with a broader spectrum of pathogens may yield better therapeutic outcomes. As a regional critical maternal care center in Guangzhou, our hospital manages a large number of complex and high-risk pregnancies. Based on local clinical practice, and antimicrobial-resistance patterns, our institution adopted a triple therapy regimen based on cephalosporins, which has a complementary antibacterial spectrum. Azithromycin and clindamycin were included to complement cephalosporin coverage against atypical, particularly Ureaplasma species, and anaerobic genital tract pathogens, respectively. In addition to its antimicrobial activity, azithromycin may have immunomodulatory effects that could influence the inflammatory environment associated with PPROM, providing an additional rationale for evaluating maternal and neonatal outcomes (such as BPD, IVH). However, direct comparative evidence for cephalosporin–azithromycin–clindamycin triple therapy remains limited, and whether its broader antimicrobial coverage translates into improved maternal and neonatal outcomes is uncertain. We therefore conducted a real-world comparative effectiveness study using a prospectively assembled triple-therapy exposure cohort and a retrospectively selected contemporaneous cephalosporin-monotherapy cohort. This hybrid design was adopted because the triple-antibiotic regimen was newly implemented at our institution: exposed patients were enrolled prospectively to ensure standardized collection of treatment and outcome data, whereas contemporaneous patients receiving conventional therapy were retrospectively identified as a pragmatic real-world comparator. And because gestational age at membrane rupture influences the competing risks of prematurity and intrauterine infection, exploratory subgroup analyses were performed across gestational-age categories. We hypothesized that the triple regimen might provide broader antimicrobial coverage and be associated with improved maternal and neonatal outcomes compared with cephalosporin monotherapy, while maintaining an acceptable safety profile. Materials and Methods Participants This study was conducted at Guangzhou Nanfang Hospital from January 2024 to June 2025, adopting a prospective exposure cohort design with a retrospectively identified contemporaneous comparison cohort. During the predefined study period, all eligible and consenting women receiving the triple-antibiotic regimen after implementation of the institutional treatment protocol were enrolled prospectively, whereas all eligible patients receiving cephalosporin monotherapy during the same period were consecutive retrospectively identified and included from the electronic medical records. Both groups were screened according to the same predefined inclusion and exclusion criteria. A total of 107 eligible pregnant women with PPROM received prenatal care at the hospital during the study period, and 92 cases were finally included after screening based on the inclusion and exclusion criteria. The inclusion criteria were as follows: (1) Pregnant women diagnosed with PPROM via clinical or auxiliary examinations; (2) Voluntary participation in this study with signed informed consent. The exclusion criteria were as follows: (1) No receipt of single-agent cefuroxime antibiotic therapy before delivery; (2) Concomitant other severe pregnancy complications, such as severe preeclampsia, HELLP syndrome, acute fatty liver of pregnancy, etc; (3) Failure to provide signed informed consent; (4) Missing key information in medical records that prevents evaluation of primary outcomes; (5) Loss to follow-up. Antibiotic Regimens During the study period, our hospital followed a standardized clinical pathway for PPROM management. Both groups received cefuroxime sodium for injection at a dose of 1.5 g by intravenous infusion twice daily. The triple-therapy group additionally received azithromycin for injection at a dose of 0.5 g by intravenous infusion once daily and clindamycin hydrochloride injection at a total daily dose of 0.9 g by intravenous infusion, divided into four administrations. The planned duration of antibiotic treatment was 7–10 days. Data Collection Demographic characteristics collected at baseline included maternal age, body mass index (BMI), reproductive history (including parity and history of abortion), gestational age at the time of PPROM, amniotic fluid depth, complete blood count at admission, C-reactive protein (CRP) and procalcitonin (PCT) levels at admission, preexisting medical conditions, and whether antenatal corticosteroids were used. Data on primary and secondary outcomes for both mothers and neonates during follow-up were extracted from the hospital’s electronic medical record system and laboratory information management system, encompassing gestational age at delivery, prolongation of the latency interval, post-antibiotic treatment complete blood count, occurrence of new infections, placental pathology results, and mode of delivery. Neonatal adverse events were also recorded, including RDS, BPD, IVH, sepsis, neonatal infections, severe neonatal morbidity, and Apgar scores documented at birth. Maternal follow-up extended from antibiotic initiation through delivery to maternal discharge, and neonatal follow-up from birth to neonatal discharge or in-hospital death. The same assessment windows were applied to both groups; follow-up was recorded prospectively in the triple-therapy cohort and retrieved from contemporaneous medical records in the comparison cohort. Indicators for Evaluation The primary endpoints were the composite maternal infection outcome and neonatal RDS, with the composite outcome comprising maternal new-onset infections and placental pathological infections. Neonatal RDS was selected as the primary neonatal outcome because it is a major prematurity-related complication of PPROM and is relatively frequent in this population. Although neonatal sepsis is also clinically important, its lower expected incidence made it less suitable as a standalone primary outcome and it was assessed as a secondary outcome. Secondary endpoints were evaluated across multiple safety dimensions, including pregnancy and delivery complications as well as laboratory parameters. Among these, maternal secondary outcomes consisted of gestational age at delivery, prolongation of the latency interval, mode of delivery, post-treatment white blood cell count, and post-treatment neutrophil percentage. Neonatal secondary outcomes included neonatal infection, severe disease, BPD, IVH, sepsis, and Apgar scores at 1, 5, and 10 minutes after birth. Statistical Analysis This study is a single-center cohort study, and the inverse probability of treatment weighting (IPTW) method based on propensity score was adopted to adjust for confounding bias.20 Specifically, a logistic regression model incorporating multiple clinically relevant baseline covariates (eg, maternal age, pre-pregnancy BMI, and parity), as well as other characteristics listed in Table 1, was applied to estimate the propensity score of each pregnant woman receiving triple antibiotic therapy, and the IPTW weights were then calculated.21 The standardized mean difference (SMD) was adopted to evaluate the balance of baseline characteristics between the two groups after IPTW adjustment. An absolute SMD value below the threshold of 0.2 is widely regarded as indicative of adequate balance.22,23Table 1Baseline Maternal Characteristics and Standardized Mean Differences Between Two GroupsCharacteristicUnadjusted AnalysisAfter IPTWExperimental GroupControl GroupSMDP valueExperimental GroupControl GroupSMDP valueMaternal age, years30.75 ± 4.4229.94 ± 4.760.1770.39330.61 ± 4.2830.19 ± 4.850.0920.689BMI24.95 ± 2.9026.31 ± 4.280.3740.06025.61 ± 2.7925.74 ± 4.340.0370.864Gravidity2.03 ± 1.082.05 ± 1.180.0160.9372.05 ± 1.042.02 ± 1.220.0290.900Prior_Abortion0.42 ± 0.770.60 ± 0.840.2270.2720.59 ± 0.850.54 ± 0.810.0630.809GA_at_PROM28.83 ± 3.8831.68 ± 5.900.5700.00429.10 ± 3.6729.97 ± 6.990.1560.498AF_Depth3.46 ± 1.494.15 ± 2.240.3650.0663.66 ± 1.553.80 ± 2.230.0720.751WBC_Adm11.24 ± 3.3310.79 ± 3.840.1260.53811.09 ± 2.9911.04 ± 4.060.0160.939Neu_Pct_Adm77.87 ± 8.0574.24 ± 7.560.4640.03077.59 ± 7.3974.91 ± 7.140.3690.078CRP_Adm10.16 ±13.9411.96 ± 15.130.1240.54912.80 ±16.2511.77 ± 15.330.0650.817PCT_Adm0.06 ± 0.060.09 ± 0.080.3680.0690.06 ± 0.060.08 ± 0.080.2740.161Preexisting_Dis20 (55.6%)38 (58.5%)0.0590.94251 (53.9%)59 (57.2%)0.0670.788ACS_Use27 (75.0%)37 (56.9%)0.3890.11256 (58.2%)62 (60.6%)0.0490.850 Table 1Baseline Maternal Characteristics and Standardized Mean Differences Between Two GroupsCharacteristicUnadjusted AnalysisAfter IPTWExperimental GroupControl GroupSMDP valueExperimental GroupControl GroupSMDP valueMaternal age, years30.75 ± 4.4229.94 ± 4.760.1770.39330.61 ± 4.2830.19 ± 4.850.0920.689BMI24.95 ± 2.9026.31 ± 4.280.3740.06025.61 ± 2.7925.74 ± 4.340.0370.864Gravidity2.03 ± 1.082.05 ± 1.180.0160.9372.05 ± 1.042.02 ± 1.220.0290.900Prior_Abortion0.42 ± 0.770.60 ± 0.840.2270.2720.59 ± 0.850.54 ± 0.810.0630.809GA_at_PROM28.83 ± 3.8831.68 ± 5.900.5700.00429.10 ± 3.6729.97 ± 6.990.1560.498AF_Depth3.46 ± 1.494.15 ± 2.240.3650.0663.66 ± 1.553.80 ± 2.230.0720.751WBC_Adm11.24 ± 3.3310.79 ± 3.840.1260.53811.09 ± 2.9911.04 ± 4.060.0160.939Neu_Pct_Adm77.87 ± 8.0574.24 ± 7.560.4640.03077.59 ± 7.3974.91 ± 7.140.3690.078CRP_Adm10.16 ±13.9411.96 ± 15.130.1240.54912.80 ±16.2511.77 ± 15.330.0650.817PCT_Adm0.06 ± 0.060.09 ± 0.080.3680.0690.06 ± 0.060.08 ± 0.080.2740.161Preexisting_Dis20 (55.6%)38 (58.5%)0.0590.94251 (53.9%)59 (57.2%)0.0670.788ACS_Use27 (75.0%)37 (56.9%)0.3890.11256 (58.2%)62 (60.6%)0.0490.850 Baseline Maternal Characteristics and Standardized Mean Differences Between Two Groups Following baseline adjustment using IPTW, univariate linear regression and logistic regression models were employed to assess the effects of antibiotic treatment regimens for continuous and binary outcome variables, respectively. Missing data (The missing rate is shown in Appendix Table 1) were addressed via the predictive mean matching method within the multiple imputation framework.24 For comparative purposes, unadjusted analysis results are presented, wherein between-group comparisons for continuous variables were conducted using two-sample t-tests, while categorical variables were analyzed using chi-square or Fisher’s exact tests. All statistical analyses were performed using R version 4.5.0 in conjunction with the R packages “survey”, “MatchIt”, and “EValue”. Sensitivity Analyses In this study, we employed IPTW to adjust for baseline covariates such as age and BMI; however, unmeasured confounders may persist. To demonstrate the robustness of the principal analysis (IPTW) results, we utilized two additional confounding adjustment methodologies, namely multivariate regression analysis and propensity score matching (PSM) analysis. For continuous outcomes, a multiple linear regression model was applied; for binary outcomes, either multiple logistic regression or Firth regression models (for rare events) were used, with adjustments for baseline covariates such as age and BMI. In the PSM analysis, a 1:1 matching was performed between the experimental and control groups through the nearest-neighbor matching method, enabling a comparison of outcomes between pregnant women using triple-therapy and those administered single-agent antibiotic therapy matched in a 1:1 ratio. Furthermore, to quantify the potential impact of unmeasured confounding bias, we calculated the E-value. The E-value indicates25,26 the minimum strength of association required for an unmeasured confounder to be jointly associated with both treatment assignment and outcome, such that the observed treatment-outcome association would no longer be statistically significant. A larger E-value implies that a stronger unmeasured confounder would be necessary to negate the current findings, thereby indicating greater robustness of the study results against potential confounding.27 If the conclusions derived from the aforementioned analyses are broadly consistent with the primary IPTW results, they collectively support the robustness of the present findings under different model assumptions. Subgroup Analysis This study incorporated one subgroup analysis. Given that the potential benefits of antibiotics in prolonging latency interval may be gestational age-dependent,28 a subgroup analysis stratified by gestational age at the time of membrane rupture was conducted. Based on guidelines and consensus evidence,28–30 three subgroups were defined as follows: (1) gestational age at membrane rupture ≤24 weeks; (2) gestational age at membrane rupture between 24 and 34 weeks; and (3) gestational age at membrane rupture >34 weeks. Result The initial cohort of this study comprised 107 pregnant women diagnosed with premature rupture of membranes. Based on predefined inclusion and exclusion criteria, 95 women (delivering 101 neonates) were enrolled in the follow-up cohort. Ultimately, 92 participants completed the follow-up, as illustrated in Figure 1. For the sensitivity analysis using PSM, a 1:1 matching between the two groups was performed based on propensity scores, resulting in a matched dataset of 32 subjects each in the control and experimental groups. Figure 1Flow diagram of participant selection and analysis. The diagram shows participant screening, reasons for exclusion, final cohort sizes, and the populations included in the IPTW and PSM analyses.The flowchart illustrates the selection and analysis process for a study involving pregnant women with preterm premature rupture of membranes. Initially, 107 women receiving antenatal care at Nanfang Hospital from January 2024 to June 2025 were included. Twelve were excluded: eight not using antibiotics, three not using cephalosporin antibiotics and one with severe pregnancy complications. This left 95 women in the follow-up cohort. Sixty were included in the control group and 35 in the experimental group. Three from the experimental group did not give informed consent, leaving 32. Propensity score matching resulted in 32 women in both control and experimental groups. Inverse Probability of Treatment Weighting analysis included 32 in the experimental group and 60 in the control group.Flowchart of participant selection and analysis for a study on pregnant women with premature rupture of membranes. Flow diagram of participant selection and analysis. The diagram shows participant screening, reasons for exclusion, final cohort sizes, and the populations included in the IPTW and PSM analyses. Demographic Characteristics of Participants As shown in Table 1, the unadjusted data analysis revealed that the control group exhibited a significantly higher mean gestational age than the experimental group (31.68 ± 5.90 vs 28.83 ± 3.88, p = 0.004), along with a reduced neutrophil count (74.24 ± 7.56 vs 77.87 ± 8.05, p = 0.030). No statistically significant differences were observed in other baseline indicators such as patient age, BMI, and number of preterm births between the two groups. Following IPTW, most baseline variables achieved balance an SMD below the threshold of 0.2. However, residual imbalance remained for admission neutrophil percentage (SMD = 0.369) and admission procalcitonin level (SMD=0.274), indicating that covariate balance was improved but incomplete. This residual imbalance was considered when interpreting the adjusted and sensitivity analyses. After PSM, no statistically significant differences were observed between the groups (P > 0.05), indicating that the matching procedure effectively improved the balance of baseline characteristics between the two cohorts to a certain extent (Appendix Table 2). Maternal Outcomes Table 2 presents the primary and secondary outcomes for pregnant women in the control and experimental groups. Furthermore, survival curves, violin plots, E-value plots, and forest plots (Figures 2–4) were employed to visualize the results.Table 2Comparison of Maternal Outcomes Before and After IPTWOutcomeUnadjusted AnalysisIPTW AnalysisExperimental GroupN=32Control GroupN=60OR_MD (95% CI)P valueExperimental GroupN=32Control GroupN=60OR_MD (95% CI)P valueComposite Outcome14 (38.89%)19 (29.23%)1.54 (0.66, 3.59)0.32138.2 (39.61%)33.1 (32.27%)1.38 (0.53, 3.59)0.510Placental_Pathologya11 (30.56%)18 (27.69%)1.16 (0.47, 2.77)0.74430.2 (31.34%)31.6 (30.82%)1.02 (0.38, 2.79)0.961New_Infection3 (8.33%)1 (1.54%)4.49 (0.71, 47.74)0.1128.0 (8.27%)1.5 (1.46%)6.10 (0.58, 64.20)0.130Delivery_Modeb20 (55.56%)31 (47.69%)1.36 (0.61, 3.08)0.45459.5 (61.73%)45.3 (44.19%)2.04 (0.82, 5.06)0.124GA_at_Delivery31.81 ± 4.2732.83 ± 5.28−1.03 (−3.07, 1.01)0.32131.77 ± 4.2731.48 ± 5.280.29 (−2.25, 2.83)0.822Extension_Time21.47 ± 32.617.02 ± 25.7714.46 (2.76, 26.16)0.01618.10 ± 32.619.21 ± 25.778.89 (−3.93, 21.71)0.172WBC_Post11.96 ± 2.0711.55 ± 3.410.42 (−0.82, 1.66)0.50511.91 ± 2.0711.52 ± 3.410.39 (−0.71, 1.49)0.483Neu_Pct_Post72.74 ± 7.4971.96 ± 7.240.78 (−2.24, 3.80)0.60972.08 ± 7.4971.86 ± 7.240.22 (−3.21, 3.64)0.900Notes: a: Placental pathology was defined as histopathological evidence of placental inflammatory or infectious changes documented in the pathology report and was coded as 0 for absence and 1 for presence. b:Mode of delivery was coded as 0 for vaginal delivery and 1 for cesarean delivery. Figure 2E-value analysis for the primary maternal and neonatal outcomes. (A and B) present the E-values for the maternal infection-related composite outcome and neonatal RDS, respectively.Two line graphs showing E-value curves for composite infectious outcome and respiratory distress syndrome. Figure 3Comparison of the latency interval between the experimental and control groups. (A) Kaplan–Meier curves showing the probability of remaining undelivered, with the corresponding number-at-risk table. (B) Violin plots showing the distribution of latency intervals in the two groups.Image A shows a Kaplan Meier curve with Control and Experimental groups. The x-axis is Latency Interval (Days) from 0 to 150 and the y-axis is Probability of Undelivered from 0.00 to 1.00, with p < 0.0001 noted. The Control curve drops sharply near day 0 to 0.10, nearing 0.00 by day 10 and stays near 0.00 until day 150. The Experimental curve declines gradually, staying above Control, reaching 0.20 by day 30, 0.10 by day 70 and near 0.00 by days 120-150. Risk table: 0 days - Control 65, Experimental 36; 50 days - Control 2, Experimental 5; 100 days - Control 2, Experimental 2; 150 days - Control 0, Experimental 0. Image B features a violin plot with points and a box plot for both groups. The y-axis is Days from 0 to 150, with P = 0.021. Control points cluster near 0 days, some around 50 and 150 days. Experimental distribution is broader, with many points between 0 and 30 days, extending beyond 50 days to above 100 days, peaking near 150 days.A Kaplan Meier plot and a violin plot comparing latency interval between control and experimental groups. Figure 4Subgroup analysis of the latency interval according to gestational age at membrane rupture. Effect estimates are presented as mean differences with 95% confidence intervals.A forest plot with a table and three subgroups under ′GA at PROM′ includes columns: Subgroup, Experimental group (n/N), Control group (n/N), Mean Difference (95% CI) and P for interaction. Subgroups: <24 weeks (experimental 5/36, control 8/65, mean difference 12.73, 95% CI -65.87 to 91.32), 24-34 weeks (experimental 31/36, control 30/65, mean difference 12.37, 95% CI 5.06 to 19.68), 34-<37 weeks (experimental 0/36, control 27/65, mean difference 0.00, 95% CI 0.00 to 0.00). The plot features a vertical reference line at 0, horizontal confidence interval lines and square markers at mean differences. The x-axis ranges from -30 to 30, labeled with ticks at -30, -20, -10, 0, 10, 20, 30 and text indicating ′Unfavorable Factor′ on the left and ′Favorable Factor′ on the right. The y-axis lists the subgroups. P for interaction is 0.466.A forest plot of latency interval mean differences by gestational age at membrane rupture subgroup. Table 2Comparison of Maternal Outcomes Before and After IPTWOutcomeUnadjusted AnalysisIPTW AnalysisExperimental GroupN=32Control GroupN=60OR_MD (95% CI)P valueExperimental GroupN=32Control GroupN=60OR_MD (95% CI)P valueComposite Outcome14 (38.89%)19 (29.23%)1.54 (0.66, 3.59)0.32138.2 (39.61%)33.1 (32.27%)1.38 (0.53, 3.59)0.510Placental_Pathologya11 (30.56%)18 (27.69%)1.16 (0.47, 2.77)0.74430.2 (31.34%)31.6 (30.82%)1.02 (0.38, 2.79)0.961New_Infection3 (8.33%)1 (1.54%)4.49 (0.71, 47.74)0.1128.0 (8.27%)1.5 (1.46%)6.10 (0.58, 64.20)0.130Delivery_Modeb20 (55.56%)31 (47.69%)1.36 (0.61, 3.08)0.45459.5 (61.73%)45.3 (44.19%)2.04 (0.82, 5.06)0.124GA_at_Delivery31.81 ± 4.2732.83 ± 5.28−1.03 (−3.07, 1.01)0.32131.77 ± 4.2731.48 ± 5.280.29 (−2.25, 2.83)0.822Extension_Time21.47 ± 32.617.02 ± 25.7714.46 (2.76, 26.16)0.01618.10 ± 32.619.21 ± 25.778.89 (−3.93, 21.71)0.172WBC_Post11.96 ± 2.0711.55 ± 3.410.42 (−0.82, 1.66)0.50511.91 ± 2.0711.52 ± 3.410.39 (−0.71, 1.49)0.483Neu_Pct_Post72.74 ± 7.4971.96 ± 7.240.78 (−2.24, 3.80)0.60972.08 ± 7.4971.86 ± 7.240.22 (−3.21, 3.64)0.900Notes: a: Placental pathology was defined as histopathological evidence of placental inflammatory or infectious changes documented in the pathology report and was coded as 0 for absence and 1 for presence. b:Mode of delivery was coded as 0 for vaginal delivery and 1 for cesarean delivery. Comparison of Maternal Outcomes Before and After IPTW Notes: a: Placental pathology was defined as histopathological evidence of placental inflammatory or infectious changes documented in the pathology report and was coded as 0 for absence and 1 for presence. b:Mode of delivery was coded as 0 for vaginal delivery and 1 for cesarean delivery. E-value analysis for the primary maternal and neonatal outcomes. (A and B) present the E-values for the maternal infection-related composite outcome and neonatal RDS, respectively. Comparison of the latency interval between the experimental and control groups. (A) Kaplan–Meier curves showing the probability of remaining undelivered, with the corresponding number-at-risk table. (B) Violin plots showing the distribution of latency intervals in the two groups. Subgroup analysis of the latency interval according to gestational age at membrane rupture. Effect estimates are presented as mean differences with 95% confidence intervals. Primary Outcome - Composite Maternal Infection Outcome As shown in Table 2, unadjusted data analysis identified a total of 33 cases of the composite maternal infection outcome, with 14 cases (38.89%) in the experimental group and 19 cases (29.23%) in the control group. After balancing baseline confounders using IPTW, the odds ratio (OR) between the two groups was 1.38 (39.61% vs 32.27%; 95% CI, 0.53–3.59; P = 0.510). No statistically significant difference was observed between two groups in the prevention of the composite maternal infection outcome. Furthermore, for the secondary indicators of the composite outcome—including placental pathology and new infections—no statistically significant differences were detected in either the unadjusted analysis or the IPTW results. In the additional multivariable regression analysis, the maternal infection-related composite outcome also did not differ significantly between the groups (Appendix Table 3). In the sensitivity analysis, the E-value for the composite maternal infection outcome was 2.16 (Figure 2), suggesting that an unmeasured confounder would need to demonstrate a relatively strong association with both the exposure and the outcome to alter the conclusions drawn in this study. PSM was employed to assess baseline covariate balance, thereby strengthening the internal validity of the observed associations with the composite outcome (Appendix Table 4). These results yielded the same overall conclusion as the primary IPTW analysis. Secondary Outcomes Our analysis demonstrates that in the unadjusted data, the experimental group exhibited a significantly longer prolongation of the latency interval compared to the control group (21.47 ± 32.61 vs 7.02 ± 25.77; P = 0.016). In the IPTW analysis, the mean difference (MD) between the two groups was 8.89 (18.10 ± 32.61 vs 9.21 ± 25.77; P = 0.172). Figure 3 visually presents the distribution of gestational age prolongation through Kaplan–Meier survival curves and violin plots. Further subgroup analysis results are presented in Figure 4. Prolongation of the latency interval was assessed based on the number of days from initiation of antibiotic treatment to delivery. In the 24–34-week subgroup, the experimental regimen was associated with a longer latency interval than the control regimen (MD = 12.37 days; 95% CI, 5.06–19.68). In the “less than 24 weeks” subgroup, while the point estimate (MD = 12.73 days) suggests a potential benefit, the extremely wide confidence interval that crosses zero, attributable to the small sample size, precludes a definitive conclusion due to insufficient statistical power. In the “greater than 34 weeks” subgroup, no events occurred in the experimental group (0/36), rendering the calculation of MD and 95% CI unfeasible. These exploratory subgroup findings suggest a possible association between the triple regimen and a longer latency interval among women with PPROM at 24–34 weeks; however, this observation should be interpreted cautiously and requires confirmation in larger studies. No statistically significant differences were observed in secondary outcomes, including gestational age at delivery, mode of delivery, and post-treatment white blood cell counts along with neutrophil percentages (Table 2). The corresponding PSM and multivariable regression sensitivity analyses are presented in Appendix Tables 3 and 4, respectively. Given the remaining covariate imbalance and the limited numbers of some events, these secondary findings should be interpreted cautiously. Neonatal Outcomes Primary Outcome - Respiratory Distress Syndrome A total of 101 newborns were included in the study. As presented in Table 3, the unadjusted data showed that 30 cases of neonatal RDS were observed, with 14 cases (38.89%) occurring in the experimental group. After applying IPTW to account for confounding by baseline covariates, the odds ratio (OR) for neonatal RDS was 1.82 (45.84% vs 31.69%; 95% CI, 0.68–4.86, P = 0.227). This study did not identify a statistically significant difference in the incidence of neonatal RDS between the triple regimen and the monotherapy regimen. The additional multivariable regression analysis yielded a similar conclusion for neonatal RDS (Appendix Table 3).Table 3Comparison of Neonatal Outcomes Before and After IPTWOutcomeUnadjusted AnalysisIPTW AnalysisExperimental GroupN=36aControl GroupN=65aOR_MD (95% CI)P valueExperimental GroupN=36aControl GroupN=65aOR_MD (95% CI)P valueRDS14 (38.89%)17 (26.15%)1.79 (0.75, 4.23)0.18644.2 (45.84%)32.5 (31.69%)1.82 (0.68, 4.86)0.227BPD1 (2.78%)9 (13.85%)0.25 (0.03, 1.16)0.0811.3 (1.40%)20.8 (20.30%)0.06 (0.01, 0.49)0.010IVH0 (0.00%)3 (4.62%)0.24 (0.00, 2.63)0.2810.0 (0.00%)5.8 (5.68%)0.00 (0.00, 0.00)0.000Sepsisb2 (5.56%)1 (1.54%)3.12 (0.40, 35.05)0.2715.2 (5.37%)1.5 (1.46%)3.84 (0.32, 46.39)0.286Neonatal_Infectionc11 (30.56%)14 (21.54%)1.60 (0.64, 3.98)0.31126.2 (27.18%)29.3 (28.58%)0.93 (0.34, 2.57)0.892Severe_Neonatal_Dzd15 (41.67%)21 (32.31%)1.49 (0.65, 3.44)0.34643.6 (45.25%)39.5 (38.55%)1.32 (0.51, 3.42)0.568Apgar18.39 ± 1.138.78 ± 0.70−0.40 (−0.76, −0.04)0.0328.22 ± 1.138.64 ± 0.70−0.42 (−1.05, 0.21)0.193Apgar59.69 ± 0.679.88 ± 0.38−0.18 (−0.39, 0.02)0.0819.62 ± 0.679.82 ± 0.38−0.20 (−0.56, 0.15)0.258Apgar109.81 ± 0.479.91 ± 0.29−0.10 (−0.25, 0.05)0.1799.77 ± 0.479.86 ± 0.29−0.09 (−0.34, 0.16)0.481Notes: a: Twins count as two cases of data. b:Neonatal sepsis was defined as a systemic infectious syndrome occurring within the first 28 days after birth, with compatible clinical manifestations and supporting laboratory and/or microbiological evidence of bacterial, viral, or fungal infection. c:Neonatal infection was defined as any confirmed localized or systemic bacterial, viral, or fungal infection diagnosed before neonatal discharge on the basis of microbiological and/or accepted clinical diagnostic evidence, including intrauterine infection diagnosed after birth, pneumonia, and other documented site-specific infections. d:Severe neonatal disease was a composite outcome defined as the occurrence of at least one of the following before neonatal discharge: patent ductus arteriosus requiring treatment, Bell stage II or higher necrotizing enterocolitis, stage III or higher retinopathy of prematurity, respiratory distress syndrome requiring both mechanical ventilation and surfactant therapy, or severe cerebral injury. Table 3Comparison of Neonatal Outcomes Before and After IPTWOutcomeUnadjusted AnalysisIPTW AnalysisExperimental GroupN=36aControl GroupN=65aOR_MD (95% CI)P valueExperimental GroupN=36aControl GroupN=65aOR_MD (95% CI)P valueRDS14 (38.89%)17 (26.15%)1.79 (0.75, 4.23)0.18644.2 (45.84%)32.5 (31.69%)1.82 (0.68, 4.86)0.227BPD1 (2.78%)9 (13.85%)0.25 (0.03, 1.16)0.0811.3 (1.40%)20.8 (20.30%)0.06 (0.01, 0.49)0.010IVH0 (0.00%)3 (4.62%)0.24 (0.00, 2.63)0.2810.0 (0.00%)5.8 (5.68%)0.00 (0.00, 0.00)0.000Sepsisb2 (5.56%)1 (1.54%)3.12 (0.40, 35.05)0.2715.2 (5.37%)1.5 (1.46%)3.84 (0.32, 46.39)0.286Neonatal_Infectionc11 (30.56%)14 (21.54%)1.60 (0.64, 3.98)0.31126.2 (27.18%)29.3 (28.58%)0.93 (0.34, 2.57)0.892Severe_Neonatal_Dzd15 (41.67%)21 (32.31%)1.49 (0.65, 3.44)0.34643.6 (45.25%)39.5 (38.55%)1.32 (0.51, 3.42)0.568Apgar18.39 ± 1.138.78 ± 0.70−0.40 (−0.76, −0.04)0.0328.22 ± 1.138.64 ± 0.70−0.42 (−1.05, 0.21)0.193Apgar59.69 ± 0.679.88 ± 0.38−0.18 (−0.39, 0.02)0.0819.62 ± 0.679.82 ± 0.38−0.20 (−0.56, 0.15)0.258Apgar109.81 ± 0.479.91 ± 0.29−0.10 (−0.25, 0.05)0.1799.77 ± 0.479.86 ± 0.29−0.09 (−0.34, 0.16)0.481Notes: a: Twins count as two cases of data. b:Neonatal sepsis was defined as a systemic infectious syndrome occurring within the first 28 days after birth, with compatible clinical manifestations and supporting laboratory and/or microbiological evidence of bacterial, viral, or fungal infection. c:Neonatal infection was defined as any confirmed localized or systemic bacterial, viral, or fungal infection diagnosed before neonatal discharge on the basis of microbiological and/or accepted clinical diagnostic evidence, including intrauterine infection diagnosed after birth, pneumonia, and other documented site-specific infections. d:Severe neonatal disease was a composite outcome defined as the occurrence of at least one of the following before neonatal discharge: patent ductus arteriosus requiring treatment, Bell stage II or higher necrotizing enterocolitis, stage III or higher retinopathy of prematurity, respiratory distress syndrome requiring both mechanical ventilation and surfactant therapy, or severe cerebral injury. Comparison of Neonatal Outcomes Before and After IPTW Notes: a: Twins count as two cases of data. b:Neonatal sepsis was defined as a systemic infectious syndrome occurring within the first 28 days after birth, with compatible clinical manifestations and supporting laboratory and/or microbiological evidence of bacterial, viral, or fungal infection. c:Neonatal infection was defined as any confirmed localized or systemic bacterial, viral, or fungal infection diagnosed before neonatal discharge on the basis of microbiological and/or accepted clinical diagnostic evidence, including intrauterine infection diagnosed after birth, pneumonia, and other documented site-specific infections. d:Severe neonatal disease was a composite outcome defined as the occurrence of at least one of the following before neonatal discharge: patent ductus arteriosus requiring treatment, Bell stage II or higher necrotizing enterocolitis, stage III or higher retinopathy of prematurity, respiratory distress syndrome requiring both mechanical ventilation and surfactant therapy, or severe cerebral injury. In the sensitivity analysis, the E-value for neonatal RDS was 1.69 (Figure 2). And the PSM (Appendix Table 5) and multivariable regression analyses (Appendix Table 3) yielded the same overall nonsignificant conclusion. Secondary Outcomes Table 3 outlines the secondary neonatal outcomes, including BPD, IVH, sepsis, neonatal infection, severe neonatal morbidity, and Apgar score at birth. Based on unadjusted data, the experimental group and the control group showed no significant differences in the incidence of IVH or BPD. However, IPTW analysis showed that the experimental regimen was associated with lower odds of BPD compared with the control regimen (1.40% vs 20.30%; OR = 0.06; P = 0.010), although the number of BPD events was limited. Furthermore, a lower observed incidence of IVH was noted in the experimental group than in the control group (0.00% vs 5.68%; p < 0.001). No statistically significant differences were observed between the two groups regarding other secondary outcomes such as neonatal sepsis, severe neonatal morbidity, or Apgar scores. The PSM and additional multivariable regression results are presented in Appendix Tables 3 and 5, respectively. Because BPD, IVH, and neonatal sepsis involved small numbers of events, estimates from the reduced or Firth regression models were imprecise and should be regarded as exploratory. Discussion In the context of PPROM, intrauterine infection may manifest as either clinically overt infection (such as clinical chorioamnionitis) or as histologic chorioamnionitis (HCA) with no typical symptoms, both representing distinct phenotypes within the same disease spectrum.31,32 From a mechanistic perspective, ascending infection and inflammatory responses may initially emerge at the placental level (detected through positive histopathological or inflammatory biomarkers) before becoming apparent systemically or during delivery as clinical manifestations in the mother.33 Therefore, relying solely on clinical or histologic criteria may underestimate the true incidence of intrauterine infection. In this study, the composite endpoint comprising “clinical chorioamnionitis/maternal infectious events (new-onset infections)” and “histological or placental pathological infection” is employed to more comprehensively evaluate the preventive efficacy of antibiotic regimens across the entire spectrum of infection. The findings of this study indicate that, after adjusting for potential confounding factors, no statistically significant differences were observed between the two antibiotic regimens regarding the primary outcomes (the composite maternal infection outcome and neonatal RDS). This finding is consistent with prior comparative studies evaluating prophylactic antibiotic regimens for PPROM. A prospective randomized trial34 demonstrated no significant differences in primary outcomes, such as the composite neonatal morbidity rate, between patients receiving cephalosporins combined with macrolides and those administered cephalosporins alone. Nevertheless, the present investigation revealed several exploratory between-group differences in certain secondary outcomes, further underscoring the absence of consistent superiority among different antibiotic regimens.12,35 In the secondary neonatal outcomes, the adjusted analyses showed that the experimental regimen was associated with lower odds of BPD and IVH. This finding is consistent with the results reported by Tanaka et al,36 which also demonstrated that for patients with PPROM at 22–27 weeks of gestation, the regimen of ampicillin-sulbactam combined with azithromycin significantly reduces the incidence of moderate-to-severe BPD compared to regimens using piperacillin or cefmetazole combined with clindamycin (18.2% vs 90.9%). However, the very large BPD effect estimate in our study should not be regarded as a precise quantitative estimate because it was based on a small number of events and may be affected by sparse-data bias and random variation. One possible underlying explanation may be related to the activity of azithromycin in combating Ureaplasma infection. Ureaplasma represents one of the most prevalent microorganisms in the intrauterine cavity following PPROM and is associated with fetal inflammatory response syndrome.37 It may contributing to adverse neonatal outcomes such as BPD and IVH.38,39 Azithromycin achieves high concentrations and exhibits prolonged persistence in placental tissue,40 while demonstrating dual efficacy through anti-Ureaplasma activity41 and immunomodulatory effect.42 These properties may facilitate activity against Ureaplasma, thereby mitigating inflammatory stimuli at their origin, which in turn reduces damage to immature lung parenchyma and injury to developing cerebral microvasculature. These mechanisms may represent one possible pathophysiological explanation for the observed associations with BPD and IVH. However, this study did not directly test the status of Ureaplasma urealyticum infection, inflammatory biomarkers, or placental pathological changes. Therefore, although the above mechanisms have certain biological rationality, they remain speculative in the context of this study. This study provides complementary real-world evidence from a Chinese tertiary hospital on a cephalosporin–azithromycin–clindamycin regimen. Although the regimen differed from that evaluated by Tanaka et al, our study included a larger cohort and extended previous observations on azithromycin-containing regimens to another clinical setting. Regarding maternal secondary outcomes, prolongation of the latency interval was greater in the experimental group in the unadjusted analysis, but the difference was no longer statistically significant after IPTW adjustment. In an exploratory subgroup analysis, triple therapy was associated with a longer latency interval among women with PPROM at 24–34 weeks of gestation (MD = 12.37 days; 95% CI, 5.06–19.68). The reason may be attributed to the dual mechanism of “synergistic coverage” and “anti-inflammatory” actions inherent to the triple regimen. On one hand, it expands the antimicrobial spectrum against aerobic, anaerobic, and atypical pathogens; on the other hand, it enhances anti‑inflammatory and immunomodulatory effects. Taking clindamycin as an example, it provides effective coverage against anaerobic bacteria – significant pathogens in pelvic infections43,44 – thus aiding in the control of ascending infections45 and suppressing the inflammatory cascade,46 thereby prolonging the latency interval. A multicenter retrospective cohort study47 reached a similar conclusion, indicating that the combination of β-lactams with macrolides significantly extends the latency period compared to macrolide monotherapy. It is noteworthy that 24–34 weeks of gestation represents a critical window for a marked increase in preterm survival rates (eg, survival exceeds 90% after 28 weeks),48,49 yet complication rates remain high.50–52 Prolongation of the latency interval by ≥12 days may provide additional time for fetal lung maturation and neurological development,29,53 creating a therapeutic window for antenatal corticosteroids to enhance lung maturity,54 ultimately improving neonatal outcomes. Management after 34 weeks often favors delivery rather than prolonged expectant management,55,56 which may partly explain the limited representation of triple therapy in this subgroup. However, the overall IPTW-adjusted result and the treatment-by-gestational-age interaction were not significant (P for interaction = 0.466); therefore, the finding in the 24–34-week subgroup should be regarded as exploratory and hypothesis-generating. Furthermore, in the unadjusted analysis of the triple-therapy group, the 1-minute Apgar score of newborns was significantly lower than that in the monotherapy group. The Apgar score is influenced by multiple factors,57 and dynamic assessment is critical. For instance, changes in scores from 1 minute to 5 minutes to 10 minutes more accurately reflect the neonatal status.58,59 Following IPTW adjustment, no significant difference in the 1-minute Apgar score was observed between the two groups. Given that no differences were found between the groups in terms of gestational age at delivery, 5-minute Apgar, and 10-minute Apgar scores, this subtle discrepancy may be attributed to the physiological condition of the fetus in utero rather than the antibiotic regimen. Overall, the triple-antibiotic regimen was not associated with significant differences in the primary maternal infection-related composite outcome or neonatal RDS compared with cephalosporin monotherapy. The lower observed risks of BPD and IVH and the longer latency interval in the 24–34-week subgroup were exploratory and should be interpreted cautiously given the limited sample size, sparse events, and multiple comparisons. Accordingly, these findings do not support routine replacement of recommended regimens or a general change in clinical practice. However, individualized use of a cephalosporin–azithromycin–clindamycin regimen may be considered in clinically stable women with PPROM at 24–34 weeks who are eligible for expectant management when mixed or atypical genital pathogens are suspected or detected, infection is persistent or recurrent, or local resistance patterns suggest inadequate routine coverage, within multidisciplinary assessment and antimicrobial stewardship. Further multicenter prospective studies are needed to confirm these observations. Advantages and Limitations The strengths of this study include the use of complementary analytical methods, including inverse probability of treatment weighting (IPTW), propensity score matching (PSM), and multivariable regression, to reduce measured confounding and improve between-group comparability. In addition, this study provides complementary real-world evidence from a Chinese tertiary-care population regarding an cephalosporin–azithromycin–clindamycin regimen for PPROM. The exploratory subgroup analysis suggested a possible association between the triple regimen and a longer latency interval among women with PPROM at 24–34 weeks, providing preliminary support for further investigation of gestational age–stratified management. Moreover, the observed differences in selected neonatal complications generate hypotheses regarding whether broader antimicrobial coverage may influence neonatal outcomes, although these findings require confirmation in larger prospective studies. Nevertheless, several limitations warrant mention. Several limitations should be acknowledged. First, the nonrandomized, single-center design and retrospective selection of the comparison cohort may have introduced residual confounding and selection bias despite multiple adjustment methods. Second, the relatively small sample size, particularly the limited numbers of BPD and IVH events, reduced the precision of effect estimates. Third, secondary and subgroup findings should be considered exploratory because of multiple comparisons. In addition, microbiological profiles, inflammatory biomarkers, and standardized placental histopathology were unavailable; therefore, the proposed biological mechanisms remain speculative. Conclusion In this study, a triple-antibiotic regimen was not associated with significant differences in the primary maternal infection-related composite outcome or neonatal RDS compared with cephalosporin monotherapy. Exploratory analyses suggested possible associations with lower observed risks of BPD and IVH and a longer latency interval among women with PPROM at 24–34 weeks of gestation. However, the single-center nonrandomized design, limited sample size, small numbers of neonatal events, and multiple secondary and subgroup analyses preclude conclusions regarding treatment superiority or routine clinical replacement. Larger multicenter prospective studies incorporating Ureaplasma status, and inflammatory biomarkers assessments are needed to confirm these findings.