Results

The database search identified 6,975 records across three databases: MEDLINE (n = 3,726), CENTRAL (n = 1,468), and Web of Science (n = 1,781). After 1,531 duplicates were removed, 5,444 records underwent title and abstract screening. Ultimately, 72 studies were included in this systematic review and network meta-analysis, with the study selection process displayed in Fig. 1. Overall, 8,728 patients were enrolled across the 72 included studies. Probiotics, prebiotics, and synbiotics were evaluated in 32, 20, and 20 studies, respectively. Placebo and usual care were used as comparators in 42 and 24 studies, respectively. The characteristics of the included studies are summarized in Table 1. Risk of bias assessments for the 33 studies contributing to the network meta-analysis of gastrointestinal symptoms are presented in eFig. S1.

Fig. 1: Study selection.

Study (year) | Country | ICU population | Setting | Sample size (n) | Intervention | Control | Duration of intervention | Oligosaccharides included | Risk of bias for GI symptoms
Mahmoodpoor 2023 | Iran | Surgical | Single-center | 40 | Probiotics | Placebo | 7–13 days | No | Low
Johnstone 2021 | Mainly Canada | Mixed | Multicenter | 2650 | Probiotics | Placebo | 7–13 days | No | Low
Anandaraj 2019 | India | Medical | Single-center | 146 | Probiotics | Placebo | < 7 days | No | Not assessed
Wongseree 2023 | Thailand | Medical | Single-center | 24 | Synbiotics | Usual care | 7–13 days | Yes | Some concerns
Cook 2016 | Canada and USA | Mixed | Multicenter | 150 | Probiotics | Placebo | ≥ 14 days | No | Not assessed
Habib 2020 | Egypt | Trauma | Single-center | 65 | Probiotics | Placebo | ≥ 14 days | No | Not assessed
Prasoon 2022 | India | Mixed | Single-center | 120 | Probiotics | Usual care | Not reported | No | High
Aghababaee 2024 | Iran | Medical | Single-center | 70 | Synbiotics | Placebo | 7–13 days | Yes | Not assessed
Nazari 2020 | Iran | Trauma | Multicenter | 150 | Probiotics | Placebo | Not reported | No | Not assessed
Ataollahi 2025 | Iran | Mixed | Single-center | 40 | Probiotics | Placebo | ≥ 14 days | No | Not assessed
Tan 2011 | China | Trauma | Single-center | 52 | Probiotics | Usual care | ≥ 14 days | No | Not assessed
Zeng 2016 | China | Mixed | Multicenter | 250 | Probiotics | Usual care | 7–13 days | No | Not assessed
Kooshki 2018 | Iran | Mixed | Multicenter | 60 | Prebiotics | Usual care | Not reported | No | High
Rohith 2023 | India | Medical | Single-center | 76 | Synbiotics | Placebo | ≥ 14 days | No | Low
Morrow 2010 | USA | Mixed | Multicenter | 138 | Probiotics | Prebiotics | Not reported | No | Low
Chen 2021 | China | Medical | Single-center | 46 | Prebiotics | Usual care | Not reported | No | Some concerns
Abbaszadeh 2024 | Iran | Trauma | Multicenter | 46 | Probiotics | Placebo | ≥ 14 days | No | Not assessed
Park 2025 | USA | Medical | Single-center | 90 | Prebiotics | Placebo | 7–13 days | No | Some concerns
Gatt 2010 | UK | Surgical | Single-center | 50 | Synbiotics | Placebo | ≥ 14 days | Yes | Some concerns
Yagmurdur 2016 | Turkey | Medical | Single-center | 120 | Prebiotics | Usual care | < 7 days | Yes | Some concerns
Tsilika 2022 | Greece | Trauma | Multicenter | 112 | Probiotics | Placebo | ≥ 14 days | No | Low
Lu 2024 | China | Mixed | Single-center | 24 | Probiotics | Usual care | Not reported | No | High
McNaught 2005 | UK | Mixed | Single-center | 103 | Probiotics | Usual care | 7–13 days | No | Not assessed
Spindler-Vesel 2007 | Slovenia | Trauma | Single-center | 132 | Prebiotics | Synbiotics | ≥ 14 days | No | Some concerns
Tzikos 2022 | Greece | Trauma | Multicenter | 112 | Probiotics | Placebo | ≥ 14 days | No | Not assessed
Dehghani 2023 | Iran | Surgical | Single-center | 105 | Synbiotics | Usual care | 7–13 days | Yes | Some concerns
Xi 2017 | China | Mixed | Single-center | 166 | Prebiotics | Usual care | < 7 days | No | Some concerns
Barraud 2010 | France | Medical | Single-center | 167 | Probiotics | Placebo | ≥ 14 days | No | Low
Abbasi 2023 | Iran | Trauma | Multicenter | 80 | Synbiotics | Placebo | 7–13 days | Yes | Not assessed
Seifi 2022a | Iran | Medical | Single-center | 42 | Synbiotics | Placebo | 7–13 days | Yes | Not assessed
Seifi 2022b | Iran | Medical | Single-center | 38 | Synbiotics | Placebo | 7–13 days | Yes | Not assessed
Wang 2021 | China | Medical | Single-center | 61 | Probiotics | Placebo | Not reported | No | Some concerns
Seifi 2022 | Iran | Medical | Single-center | 38 | Synbiotics | Placebo | 7–13 days | Yes | Some concerns
Malik 2016 | Malaysia | Mixed | Single-center | 60 | Probiotics | Placebo | 7–13 days | No | Not assessed
Litton 2021 | Australia | Mixed | Multicenter | 221 | Probiotics | Placebo | ≥ 14 days | No | Not assessed
Giamarellos-Bourboulis 2009 | Greece | Surgical | Multicenter | 72 | Synbiotics | Placebo | ≥ 14 days | No | Not assessed
Alberda 2007 | Canada | Mixed | Single-center | 28 | Prebiotics | Synbiotics | 7–13 days | Yes | Not assessed
Majid 2014 | UK | Mixed | Multicenter | 47 | Prebiotics | Placebo | 7–13 days | Yes | High
Ferrie 2011 | Australia | Mixed | Single-center | 36 | Synbiotics | Prebiotics | 7–13 days | No | Some concerns
Shimizu 2018 | Japan | Mixed | Multicenter | 77 | Synbiotics | Usual care | ≥ 14 days | Yes | Some concerns
Mahmoodpoor 2019 | Iran | Surgical | Multicenter | 120 | Probiotics | Placebo | ≥ 14 days | No | Some concerns
van der Spoel 2007 | Netherlands | Mixed | Multicenter | 308 | Prebiotics | Placebo | < 7 days | No | Low
Jain 2004 | UK | Surgical | Single-center | 90 | Synbiotics | Placebo | 7–13 days | Yes | Not assessed
Chittawatanarat 2010 | Thailand | Surgical | Single-center | 34 | Prebiotics | Usual care | 7–13 days | Yes | Some concerns
Rushdi 2004 | Egypt | Mixed | Single-center | 30 | Prebiotics | Usual care | < 7 days | No | High
Tuncay 2018 | Turkey | Medical | Single-center | 68 | Prebiotics | Usual care | ≥ 14 days | Yes | High
Knight 2009 | UK | Mixed | Single-center | 300 | Synbiotics | Placebo | ≥ 14 days | No | Some concerns
Frohmader 2010 | Australia | Mixed | Single-center | 45 | Probiotics | Placebo | 7–13 days | No | Some concerns
Kotzampassi 2006 | Greece | Trauma | Multicenter | 65 | Synbiotics | Placebo | ≥ 14 days | No | Not assessed
Kwon 2015 | USA | Medical | Single-center | 70 | Probiotics | Usual care | 7–13 days | No | Not assessed
Forestier 2008 | France | Mixed | Single-center | 208 | Probiotics | Placebo | < 7 days | No | Not assessed
Caparrós 2001 | Spain | Mixed | Multicenter | 220 | Prebiotics | Usual care | ≥ 14 days | Yes | Some concerns
Dobb 1990 | Australia | Mixed | Single-center | 91 | Prebiotics | Usual care | ≥ 14 days | No | High
Falcão de Arruda 2004 | Brazil | Trauma | Single-center | 20 | Probiotics | Usual care | 7–13 days | No | Not assessed
Koutelidakis 2010 | Greece | Trauma | Multicenter | 65 | Synbiotics | Placebo | ≥ 14 days | No | Not assessed
Schultz 2000 | USA | Mixed | Single-center | 44 | Prebiotics | Placebo | < 7 days | No | High
Karakan 2007 | Turkey | Medical | Single-center | 30 | Prebiotics | Usual care | 7–13 days | Not reported | Not assessed
Lee 2016 | South Korea | Mixed | Single-center | 22 | Prebiotics | Usual care | 7–13 days | No | Not assessed
Spapen 2001 | Belgium | Medical | Single-center | 25 | Prebiotics | Usual care | 7–13 days | No | High
Fazilaty 2018 | Iran | Trauma | Single-center | 40 | Prebiotics | Placebo | ≥ 14 days | No | Not assessed
Bleichner 1997 | France | Mixed | Multicenter | 128 | Probiotics | Placebo | 7–13 days | No | Some concerns
Klarin 2008 | Sweden | Mixed | Multicenter | 44 | Probiotics | Placebo | < 7 days | No | Not assessed
Klarin 2005 | Sweden | Mixed | Single-center | 15 | Probiotics | Usual care | 7–13 days | No | Not assessed
Sanaie 2014 | Iran | Mixed | Single-center | 40 | Probiotics | Placebo | 7–13 days | No | Not assessed
Naslowski 2025 | Brazil | Mixed | Single-center | 70 | Synbiotics | Placebo | 7–13 days | Yes | High
Ebrahimi-Mameghani 2013 | Iran | Surgical | Single-center | 40 | Probiotics | Placebo | 7–13 days | No | Not assessed
Kasiri 2023 | Iran | Mixed | Single-center | 80 | Synbiotics | Placebo | ≥ 14 days | Yes | Some concerns
Sanaie 2013 | Iran | Mixed | Single-center | 40 | Probiotics | Placebo | 7–13 days | No | Not assessed
Kamel 2021 | Egypt | Trauma | Single-center | 85 | Probiotics | Usual care | < 7 days | No | Not assessed
Rayes 2002 | Germany | Surgical | Single-center | 95 | Synbiotics | Prebiotics | 7–13 days | No | Some concerns
Vahdat Shariatpanahi 2018 | Iran | Mixed | Single-center | 32 | Prebiotics | Placebo | 7–13 days | Yes | High
Mao 2022 | China | Medical | Single-center | 60 | Synbiotics | Probiotics | 7–13 days | No | High

Primary outcome: gastrointestinal symptoms

Overall, 33 studies enrolling 5,073 patients reported gastrointestinal symptoms and were included in the network meta-analysis of the primary outcome. Study-level event counts are presented in eTable S2. The network comprised four nodes representing the control, prebiotic, probiotic, and synbiotic groups, all connected within a single network map, which permitted valid indirect comparisons across treatments (Fig. 2). In the network meta-analysis, synbiotics and prebiotics significantly reduced the risk of gastrointestinal symptoms compared with control (synbiotics: risk ratio [RR] 0.52; 95% confidence interval [CI]: 0.35 to 0.77 and prebiotics: RR 0.66; 95% CI: 0.49 to 0.87), whereas probiotics showed a directionally similar but statistically nonsignificant estimate (RR 0.81; 95% CI: 0.60 to 1.10) (Fig. 3). No statistically significant differences were observed among the active interventions (Table 2). Exploratory SUCRA rankings are presented in eTable S3. Conventional pairwise meta-analyses yielded results consistent with the network estimates (eTable S4), and neither global nor local inconsistencies were detected (eTables S5 and S6). However, Egger’s regression test using comparison-adjusted effect sizes suggested potential small-study effects (p = 0.035) (eFig. S2). For diarrhea, 27 RCTs involving 4,786 patients were included. Prebiotics and synbiotics were associated with a lower risk than control (prebiotics: RR 0.65; 95% CI: 0.48 to 0.88, synbiotics: RR 0.52; 95% CI: 0.31 to 0.87), whereas probiotics were not (RR 0.84; 95% CI: 0.61 to 1.14). For increased GRV or feeding intolerance, nine RCTs involving 826 patients were included. Prebiotics and probiotics were associated with a lower risk than control (prebiotics: RR 0.61; 95% CI: 0.48 to 0.79, probiotics: RR 0.51; 95% CI: 0.31 to 0.84), whereas synbiotics were not (RR 0.84; 95% CI: 0.51 to 1.40). No statistically significant differences were observed among the active interventions in either symptom-specific analysis.

Fig. 2: Network map for gastrointestinal symptoms.Overall, 33 studies were included in the network. Each node represents a treatment, and node size is proportional to the number of randomized patients. Line thickness is proportional to the number of studies contributing to each comparison.

Fig. 3: Forest plot of network meta-analysis for gastrointestinal symptoms: prebiotics, probiotics, and synbiotics vs control.Forest plot comparing prebiotics, probiotics, and synbiotics with controls for gastrointestinal symptoms. The results were expressed as risk ratios with 95% confidence intervals (CIs).

 | control | prebiotics | probiotics | synbiotics
control | – | 0.66 (0.49, 0.87) | 0.81 (0.60, 1.10) | 0.52 (0.35, 0.77)
prebiotics | 1.52 (1.15, 2.03) | – | 1.24 (0.89, 1.74) | 0.79 (0.50, 1.24)
probiotics | 1.23 (0.91, 1.66) | 0.81 (0.57, 1.13) | – | 0.63 (0.39, 1.02)
synbiotics | 1.94 (1.31, 2.88) | 1.27 (0.81, 2.00) | 1.58 (0.98, 2.56) | –

Secondary outcomes

The numbers of RCTs and patients contributing to each secondary outcome were as follows: mortality, 43 RCTs (6,064 patients); ICU mortality, 20 RCTs (4,698 patients); ICU length of stay, 37 RCTs (6,109 patients); duration of mechanical ventilation, 21 RCTs (4,668 patients); infectious complications, 45 RCTs (6,624 patients); VAP, 22 RCTs (5,032 patients); and adverse events, 15 RCTs (3,876 patients). The network map for each outcome is shown in eFig. S3. Conversion from medians to means was required for 4 of the 37 studies reporting ICU length of stay and 1 of the 21 studies reporting duration of mechanical ventilation. The same validated conversion methods were applied consistently to all studies requiring conversion.

Mortality

In the network meta-analysis, none of the interventions demonstrated a statistically significant reduction in mortality compared with the control: prebiotics (RR 0.79; 95% CI: 0.62 to 1.02), probiotics (RR 0.94; 95% CI: 0.85 to 1.04), and synbiotics (RR 0.83; 95% CI: 0.66 to 1.05) (eFigure S4). No significant global or local inconsistencies were detected, and the comparison-adjusted funnel plot showed no marked asymmetry (eTables S5 and S6; eFigure S5). Regarding ICU mortality, the trend was similar to that of mortality.

Length of ICU stay

Prebiotics (MD -1.70 days; 95% CI: −3.27 to −0.14) and probiotics (MD −1.52 days; 95% CI: −2.75 to −0.29) significantly reduced ICU length of stay compared with control, whereas synbiotics did not (MD −1.48 days; 95% CI: −3.11 to 0.16) (eFigure S4). No global or local inconsistency was detected (eTables S5 and S6).

Mechanical ventilation duration

Synbiotics significantly shortened mechanical ventilation duration relative to control (MD, -2.73 days; 95% CI: -5.43 to -0.03), whereas prebiotics (MD, -2.18 days; 95% CI: -5.39 to 1.02) and probiotics (MD, -1.03 days; 95% CI: -2.91 to 0.86) did not reach statistical significance (eFigure S4). No global or local inconsistency was detected, and the comparison-adjusted funnel plot demonstrated no marked asymmetry (eTables S5 and S6; eFigure S5).

Infectious complications

Significant global inconsistency was detected in the network meta-analysis using the design-by-treatment interaction model (p = 0.02). Local inconsistency was also identified in the node-splitting analyses for the comparisons of prebiotics versus control and synbiotics versus control. In the direct pairwise meta-analyses, prebiotics, probiotics, and synbiotics were each associated with a lower risk of infectious complications than control (prebiotics: RR 0.62; 95% CI: 0.42 to 0.91, probiotics: RR 0.64; 95% CI: 0.51 to 0.79, synbiotics: RR 0.65; 95% CI: 0.48 to 0.87, respectively).

For VAP, probiotics (RR 0.64; 95% CI: 0.47 to 0.85) and synbiotics (RR 0.59; 95% CI: 0.37 to 0.94) were also associated with a lower risk than control; conversely, prebiotics were not (RR 0.85; 95% CI: 0.45 to 1.60) (eFigure S4). No significant differences were observed among the active interventions (eTable S7). No evidence of global or local inconsistency was detected, and the comparison-adjusted funnel plot showed no marked asymmetry (eTables S5 and S6; eFigure S5).

Adverse events

No significant differences in adverse events were observed between the intervention and control groups, although the estimates were imprecise (eFigure S4). Detailed results for all secondary outcomes are provided in the Supplementary Material.

Subgroup and sensitivity analyses

In the prespecified subgroup and sensitivity analyses conducted within the random-effects network meta-analytic framework, treatment effects and corresponding SUCRA rankings varied across clinical contexts. Among surgical ICU patients, probiotics and synbiotics were preferred over controls for reducing gastrointestinal symptoms, whereas prebiotics were not. Conversely, among medical ICU patients, prebiotics were favored over control; additionally, probiotics and synbiotics showed no clear benefit. Regarding intervention duration, the evidence was largely concentrated in the 7–13-day treatment range; within this subgroup, synbiotics remained beneficial compared with control, whereas prebiotics and probiotics were not. In trials incorporating oligosaccharide supplementation, both prebiotics and synbiotics were favored over control, with no clear distinction between them. In the sensitivity analysis excluding studies judged to be at high risk of bias, synbiotics were associated with a lower risk of gastrointestinal symptoms than control, whereas prebiotics were not. However, significant global inconsistency was detected in this restricted network (p = 0.04), with evidence of local inconsistency for the comparisons of prebiotics versus control and probiotics versus control (eTables S5 and S6). Another sensitivity analysis excluding studies with fewer than 50 participants included 21 RCTs involving 4,678 participants. No evidence of funnel plot asymmetry was detected (Egger’s test, p = 0.19; eFigure S8). In this analysis, synbiotics were associated with a significantly lower risk of gastrointestinal symptoms than control (RR 0.53; 95% CI, 0.36 to 0.76), whereas prebiotics and probiotics were not (eFigure S7). In contrast to the primary analysis, statistically significant differences were observed between synbiotics and the other active interventions. Exploratory treatment rankings placed synbiotics highest, with a SUCRA value of 1.00, followed by prebiotics (0.50), probiotics (0.30), and control (0.10) (eTable S3). In the sensitivity analysis restricted to trials with a limited control group (28 RCTs, 4,723 patients), the findings were similar to those of the primary analysis. Detailed results of subgroup and sensitivity analyses are provided in the Supplementary Material (eFigures S6–S8 and eTable S7).