This systematic review contains studies that met the following inclusion criteria: (1) Research type: This research included randomized controlled trials (RCTs) published both domestically and internationally, regardless of whether blinding was used. (2) Research subjects: All the parturients selected within this study were those who took oxytocin to prevent PPH. (3) Intervention measures: RCTs evaluating different doses of oxytocin. Oxytocin administration was initiated either after delivery of the anterior shoulder, following placental expulsion, or immediately after cesarean closure, depending on individual trial protocols. (4) Primary outcome measure: Incidence of PPH. Secondary outcome measure: Adverse reactions. For consistency, “low-dose” oxytocin was defined as standard prophylactic regimens commonly recommended in international guidelines ($\le$10 IU IM [Intramuscular]/IV [Intravenous] or equivalent), while “high-dose” oxytocin referred to regimens exceeding this threshold ($\ge$20–80 IU IV/IM, depending on trial). Studies with non-Chinese or non-English literature; repetitive literature, experience summaries, case studies, reviews, studies with insufficient or incomplete information, meeting minutes; and literature with less than 20 cases were all excluded.

Literature retrieval was conducted through PubMed, China National Knowledge Infrastructure (CNKI), Embase, Web of Science, Cochrane Library, Ovid Technologies (OVID) and Wanfang databases. The literature search time frame spanned from the database’s establishment to April 2025. The search strategy combined subject terms and free terms. The English search strategy used the keywords, “oxytocin”, “dosage”, “postpartum hemorrhage”, “random”, “randomized controlled trials (RCTs)”. The Chinese search strategy used the keywords “oxytocin”, “postpartum hemorrhage”, “dose” and “randomized control”. Thresholds were defined following WHO (2022) and ACOG (2023) recommendations, where a high dose exceeds 10 IU and a low dose is $\le$10 IU for prevention protocols.

Two researchers independently screened the literature through title and determine the eligibility from the extracted data. Disagreements regarding the inclusion of extracted data from the literature were resolved by consensus with the third researcher. The EndNote software 2025.3.1 (Clarivate, Philadelphia, PA, USA) was used for literature screening and deduplication. Secondary screening of the selected literature meeting the inclusion criteria was conducted by reviewing abstract, followed by full-text reading, to determine the final literature to be included. The content extracted from the literature mainly includes the author’s name, publication dates, sample sizes and other relevant details. The literature search was completed in March 2025, with inclusion of one early online publication from April 2025. All data were obtained directly from published RCTs; no additional author’s contact was required.

The quality of the included studies was assessed by applying the Cochrane Collaboration’s risk of bias assessment tool. This included (1) random sequence generation methods (computer random number generator, random number table); (2) hidden distribution (central distribution, opaque sealed envelopes); (3) blinding of researchers and participants (single-blind or double-blind); (4) blinded evaluation of the research outcomes (whether blinding was adopted or not); (5) integrity of outcome data and selective reporting bias in study results; (6) other biases.

Data were analyzed using Review Manager (RevMan), version 5.3 (The Cochrane Collaboration, London, UK). For measurement outcomes, effect sizes were expressed as mean difference (MD) with 95% confidence intervals (CI). For categorical outcomes, odds ratios (OR) with 95% CI was calculated. Heterogeneity was assessed using I² statistic and Chi-square test. Results were considered homogeneous when p-value was $\ge$0.1 and I² was 50%; therefore, a fixed-effect model was applied. A random-effect model was used when heterogeneity was detected (p-value 0.1 or I² $\ge$50%) and sources of heterogeneity were explored if appropriate. Publication bias was evaluated using Egger’s test and funnel plots.

A total of 746 records were identified through database searches (PubMed, CNKI, Embase, Web of Science, Cochrane Library, OVID and Wanfang). After removing 158 duplicates, 588 unique records were screened. Following screening, 546 records were excluded for not fulfilling the eligibility criteria. After full text screening of 42 records, 36 records were further excluded from the study. Therefore, 6 RCTs met the inclusion criteria and were included in the meta-analysis (Fig. 1).

Fig. 1.

PRISMA Flowchart. PRISMA, Preferred Reporting Items for Systematic Reviews and Meta-analyses.

Among the 6 records included in the analysis, 2 were Chinese articles and 4 were foreign language articles, all of which were RCTs. There were 3615 subjects among them (1705 receiving high-dose oxytocin group and 1910 receiving low-dose oxytocin group). The studies included were published between 2012 and 2025 (Table 1, Ref. [9, 10, 11, 12, 13, 14]). The literature was evaluated using the Cochrane quality evaluation standard (Fig. 2).

Fig. 2.

Risk of bias assessment of included RCTs. RCT, randomized controlled trial. For each bias domain, the symbol “+” denotes low risk of bias, whereas “?” denotes unclear risk of bias, based on the Cochrane Collaboration risk-of-bias tool.

Table 2 (Ref. [9, 10, 11, 12, 13, 14]) summarizes how each trial defined high-dose and low-dose oxytocin including the route and timing of administration and whether participants delivered vaginally or via cesarean section (C-section). The details allow for a transparent comparison of dose categorizations across studies and reduce interpretive heterogeneity.

Six studies [9, 10, 11, 12, 13, 14] reported the incidence of PPH. The analysis of the 6 studies showed no significant heterogeneity in statistics (p = 0.17, I² = 36). After treatment, the results indicated that high-dose oxytocin was associated with a lower incidence of PPH (OR = 0.44, 95% CI: 0.32–0.62, p 0.00001, Fig. 3).

Fig. 3.

Forest plot comparing the incidence of PPH in the two groups of participants. PPH, prevention of postpartum hemorrhage.

Three literature sources [10, 12, 14] reported the incidence of adverse reactions. Analyzing the statistics of the 3 studies revealed no heterogeneity (p = 0.88, I² = 0%), and a fixed-effect model was applied. After treatment, there was no statistically significant difference in the incidence of adverse reactions between the two groups (p = 0.33 $>$ 0.05, Fig. 4). Only three studies reported adverse events, which limits the generalizability of these findings.

Fig. 4.

Forest plot comparing the incidence of adverse reactions.

Subgroup analysis stratified by mode of delivery showed that the preventive effect of high-dose oxytocin was consistent for both vaginal and cesarean births, although the magnitude of effect was greater in cesarean deliveries. Sensitivity analysis excluding smaller studies (n 100 per arm) did not materially change the pooled effect estimates, supporting the robustness of the findings.

In addition to pooled estimates, we qualitatively summarized adverse reactions reported in the included trials. The most common adverse events were hypotension, nausea/vomiting, and tachycardia. No trial reported severe arrhythmias or maternal deaths attributable to oxytocin. A supplementary forest plot (Fig. 4) illustrates the distribution of adverse outcomes across studies, showing no significant imbalance between high- and low-dose groups.

To investigate the publication bias in this study, 6 records comparing the incidence of PPH with different doses of oxytocin were selected to draw funnel plots. The results of the funnel plots are shown in Fig. 5. The funnel plot was visually symmetrical, and Egger’s test confirmed no evidence of publication bias (p = 0.256), suggesting that the pooled effect is unlikely to be driven by selective reporting. We therefore consider publication bias to be unlikely for the primary outcome. However, as most studies originated from Asia and Africa, caution is required when generalizing results to populations in high-income countries.

Fig. 5.

Funnel plot of publication bias in the incidence.

A prespecified subgroup analysis was performed by both dose range (10–20 IU vs. $>$40 IU) and route of administration (IV vs. IM). The preventive effect of high-dose oxytocin remained statistically significant in both subgroups (p 0.05), confirming robustness of the pooled estimate despite heterogeneity in dosing protocols. The effect size was greater for IV administration, likely reflecting faster onset of uterotonic activity.

Based on the synthesized evidence, clinicians should consider using high-dose oxytocin regimens within the ranges supported by international guidelines (10–40 IU depending on mode of delivery and administration route). This strategy reduces the risk of PPH without increasing adverse events. Standardization of dose definitions in clinical protocols and future trials will further strengthen the quality of evidence and improve maternal outcomes.

The main strength of this meta-analysis is that it included only RCTs, which represent the highest level of evidence for clinical interventions. A robust methodology was employed, including comprehensive database searches, standardized quality assessment using the Cochrane risk-of-bias tool, and a pooled analysis of more than 3600 participants. Moreover, the consistency of findings across multiple geographic regions and delivery modes is also a major strength of this meta-analysis.

However, several limitations of this study must be acknowledged. Firstly, the number of included trials was relatively small (n = 6), which may limit generalizability. Secondly, there was variation in how individual studies defined “high-dose” and “low-dose” oxytocin, potentially contributing to heterogeneity. Thirdly, most studies were conducted in limited regional contexts, reducing applicability to all health-care settings. Finally, subgroup analysis (e.g., vaginal vs. cesarean delivery) were limited by available data, and future studies should explore these differences more comprehensively.

The results of this meta-analysis provide clinically meaningful evidence that high-dose oxytocin is more effective in preventing PPH than low-dose regimens, without additional safety concerns. This supports the adoption of standardized high-dose protocols within national and international guidelines to reduce maternal morbidity and mortality. Clinicians should evaluate dose optimization with individual patient risk factors, but these findings suggest that concerns over adverse events may be less evident within recommended dose ranges.

For future research, harmonizing dose definitions across trials will be essential to reduce heterogeneity and improve comparability. Large, multicentre RCTs conducted in diverse populations could strengthen the evidence base, particularly for cesarean versus vaginal deliveries. Additional work is needed to report adverse outcomes more systematically, enabling clearer risk–benefit assessments.