Enteral nutrition interruption in critically ill patients: a systematic review and meta-analysis of prevalence, causes, and clinical outcomes Xiaojing Guo, Libing Jiang, Shouyin Jiang, Xiaoxia Huang, Jiaying Tang, Xiaoli Liang, Xiuqin Feng, Shanxiang Xu Abstract IntroductionEnteral nutrition interruption (ENI) is common in critically ill patients and may compromise nutritional delivery, but its prevalence, causes, and clinical consequences remain uncertain. This systematic review and meta-analysis synthesized evidence on ENI prevalence, contributing factors, clinical outcomes, and potential interventions.MethodsPubMed, Embase, CINAHL, Web of Science, Cochrane Library, CNKI, and WanFang Data were searched from inception through December 2025. Studies of critically ill adults receiving enteral nutrition were eligible. The review followed PRISMA guidance and was registered in PROSPERO (CRD420251266654). Random-effects models, subgroup analyses, sensitivity analyses, and GRADE assessments were performed.ResultsForty-one studies involving 8,260 critically ill patients were included. Crude ENI prevalence ranged from 11.2% to 100%, with a median of 68.5% (IQR, 57.1%-96.6%). The estimated pooled prevalence was 79.2%, although heterogeneity was extreme (I2 = 99%) and publication bias was evident (Egger’s test, p < 0.001). Common causes included airway management, radiological or imaging procedures, gastrointestinal intolerance, and surgical procedures or preparation. ENI was associated with reduced energy adequacy (mean difference, -4.21 percentage points; 95% CI, –7.71 to –0.71; p = 0.02). No statistically significant association was observed with ICU length of stay (mean difference, –0.41 days; 95% CI, –2.05 to 1.23; p = 0.62). Short-term mortality was not statistically significantly associated with ENI (risk ratio, 1.86; 95% CI, 0.67–5.11), but the estimate was imprecise and heterogeneous (I2 = 88%).DiscussionENI appears common and is predominantly procedure-related. It is associated with reduced energy adequacy, whereas its effects on ICU length of stay and short-term mortality remain uncertain because of inconsistent ENI definitions, substantial heterogeneity, publication bias, and limited outcome data. Standardized definitions, shorter peri-procedural fasting, improved gastrointestinal management, and multidisciplinary nutrition protocols may reduce avoidable interruptions. High-quality prospective studies and randomized trials are needed.Systematic review registrationhttps://www.crd.york.ac.uk/PROSPERO/view/CRD420251266654. Introduction Adequate nutritional support is a cornerstone of management in critically ill patients (), with early enteral nutrition (EN) recommended by international guidelines to preserve gut integrity, modulate immune responses, and improve clinical outcomes (, ). However, despite strong guideline endorsement, the delivery of EN in the intensive care unit (ICU) remains suboptimal, 60–75% of patients in the ICU (, ), do not reach the target feeding volume. Research has found that the feeding deficiency rates were 54 and 15% (p < 0.001) on trial days with and without ENI (), respectively, indicating a positive correlation between ENI and insufficient feeding. The incidence of ENI is relatively high in the ICU, ranging from 54 to 84% (). Although ENI has been increasingly recognized as an important clinical issue, existing studies are highly heterogeneous in terms of definitions, reported prevalence, contributing factors, and outcome measures. Reported ENI rates vary widely across studies, partly due to inconsistent definitions of interruption duration and differences in clinical practice. Moreover, while several studies have explored the causes of ENI, there is no comprehensive synthesis of these factors across different ICU settings. In addition, evidence regarding clinical and biochemical indicators associated with ENI remains limited and inconsistently reported, and the effectiveness of proposed interventions has not been systematically evaluated. Therefore, a comprehensive synthesis of the current evidence is needed to better understand the burden, determinants, and potential clinical consequences of ENI in critically ill patients. This systematic review and meta-analysis aims to (1) estimate the pooled prevalence of ENI, (2) identify the main contributing factors, (3) summarize reported clinical outcomes (including energy adequacy, intensive care unit length of stay (LOS), and short-term mortality) and biochemical indicators, and (4) synthesize evidence on potential intervention strategies to reduce interruptions and optimize nutritional delivery in the ICU. Methods This study was performed according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) checklist (). This systematic review was registered prospectively with the PROSPERO database (registration number: CRD420251266654) on 16 December 2025. Search strategy We performed a systematic search of the PubMed, Embase, CINAHL, Web of Science, Cochrane Library, CNKI, and WanFang Data electronic databases. References were also traced. Only original studies published in English or Chinese were included, and the search spanned from database inception to December, 2025. The search employed a combination of Medical Subject Headings (MeSH) and free-text keywords related to: “Intensive Care Units,” “critically ill patients,” “ICU,” “Enteral Nutrition,” “tube feeding,” “nutrition support,” “Treatment Interruption,” “feeding interruption,” “enteral nutrition interruption.” The whole search strategy is provided in . Inclusion and exclusion criteria This review included studies that met the following PICOS criteria: Population (P): Critically ill adult patients (aged ≥18 years) admitted to intensive care units receiving enteral nutrition support. Intervention (I): Clinical situations featuring ENI, including all types, durations and root causes of feeding suspension. Comparator (C): Critically ill adult patients receiving continuous enteral nutrition without any recorded ENI episodes. However, considering that this article is a meta-analysis of ENI incidence rates, the absence of a control group is also acceptable. Outcome (O): Primary outcome: Pooled prevalence of ENI and energy adequacy rate; Secondary outcomes: ICU length of stay, short-term all-cause mortality, serum nutritional and inflammatory biochemical markers. Study design (S): Observational studies (cohort, cross-sectional), randomized controlled trials and their secondary analyses published in Chinese or English. Study exclusion criteria included inaccessible full text, duplicate publications or studies based on the same patient cohort, incomplete or unextractable study data, reviews, editorials, conference abstracts, case reports, and studies with low-quality results assessed by critical appraisal tools. Study selection and data extraction Based on the search results, the literature was first imported into the reference manager EndNote X9 for deduplication. Two researchers then independently screened the literature according to the inclusion and exclusion criteria. The screening process involved initially reviewing the titles and abstracts, followed by reading the full texts to determine eligibility. Additionally, the references of included articles were manually searched to identify further relevant studies. In cases where consensus could not be reached, a third researcher was consulted. Extracted data from included studies comprised study identifier, country, department, sample size, number of centers, age, study design, incidence of ENI, definition of ENI, feeding route, biochemical markers, causes of ENI, and recommended interventions. Risk of bias assessment Two reviewers independently performed risk of bias assessments for the included studies. All reviews underwent cross-verification procedures, with disagreements resolved through a third reviewer. Cohort and case–control studies were assessed using the Newcastle–Ottawa Scale (NOS), which has three dimensions: participant selection, group comparability, and outcome assessment, with a total score of 9. Study quality scores were defined arbitrarily as poor (0–3), fair (4–6), or good (7–9) (). Cross-sectional studies used the 11-item AHRQ checklist (1 = compliant, 0 = unclear/non-compliant), categorized as high (8–11), moderate (4–7), or low quality (0–3). RCTs and their secondary analyses were evaluated with Cochrane ROB 2.0, rating seven bias domains to determine “low risk,” “some concerns,” or “high risk.” Two reviewers independently assessed quality, resolving disagreements via discussion or third-party arbitration (). Certainty of evidence assessment The certainty of evidence related to our primary outcomes was evaluated using the Grading of Recommendations Assessment, Development, and Evaluation (GRADE) () approach to categorize the certainty of evidence as “high,” “moderate,” “low,” or “very low.” As part of the GRADE assessment, evidence was evaluated based on risk of bias, inconsistency, indirectness, imprecision, and other biases including publication bias and small study effects (by examining funnel plot asymmetry). Any discrepancies between author assessments were resolved by consensus. Data analysis Statistical analyses were performed using Review Manager (RevMan) version 5.4 (The Cochrane Collaboration, Copenhagen, Denmark) and Stata version 17.0 (StataCorp LLC, College Station, TX, USA). A single-group rate meta-analysis with logit transformation was conducted to estimate the pooled prevalence of ENI across all included studies. Heterogeneity across studies was evaluated using the I2 statistic, with I2 > 50% indicating substantial heterogeneity, and a random-effects model was used for all analyses. Subgroup analyses were performed to explore sources of heterogeneity, stratified by study design, ICU type, country development level, ENI time threshold definition, feeding route, and feeding mode. Sensitivity analysis was carried out by sequentially omitting one study at a time to assess the stability of the pooled effect estimates. Publication bias was visually assessed using funnel plots and quantitatively assessed using Egger’s linear regression test, with a two-sided p < 0.05 considered statistically significant. Results Selection of trials The systematic search initially retrieved 5,113 citations from 7 databases and registers. After removing 826 duplicates, 4,218 irrelevant studies were excluded via title/abstract screening, leaving 69 records for abstract assessment. Of these, 31 proceeded to full-text review, where 38 were excluded (8 for topic mismatch, 15 for inconsistent subjects, 10 for inappropriate study type, 2 for unavailable full text, 3 for inconsistent data types). An additional 10 eligible studies were identified from other sources, resulting in 41 included studies. The selection process is detailed in . Trial characteristics The basic characteristics of the included studies are presented in . A total of 41 studies (, ) involving 8,260 critically ill patients receiving EN were included, with sample sizes ranging from 22 to 1,495 participants (, ). The included patients had a median age between 51 and 64 years across studies, with some studies reporting that most patients were under 60 years old. Only a few studies were multicenter investigations, including Davies et al. () involving 40 centers, Zheng et al. () involving 90 centers, and the large international observational study by Saran et al. () involving 403 centers, as well as the international multicenter study reported by Czapran et al. (). Quality evaluation of the literature Among the 41 included studies, 35 observational studies were assessed using the NOS, with scores ranging from 7 to 8, indicating generally high methodological quality (). Four cross-sectional studies (, , , ) were evaluated using the AHRQ checklist, with scores ranging from 9 to 10, suggesting moderate-to-high quality (). Two RCTs (, ) or secondary analyses of randomized trials were assessed using the Cochrane ROB 2.0 tool and were judged to have some concerns of bias mainly due to the lack of participant blinding (). Prevalence of ENI The unpooled crude incidence of ENI demonstrated tremendous between-study fluctuation, ranging from 11.2 to 100%. The median ENI prevalence across individual investigations was 68.5%, with an interquartile range of 57.1–96.6%. These findings indicate that the pooled global estimate should not be interpreted as a uniform prevalence applicable across healthcare systems, particularly given the marked differences between developed and developing countries. A logit-transformed single-rate meta-analysis of all 41 included studies estimated the prevalence of ENI among critically ill patients. The pooled OR for ENI was 3.82 (95% CI: 2.52–5.78), corresponding to an overall estimated prevalence of approximately 79.2%. Extreme and largely unexplained heterogeneity was present (I2 = 99%, p < 0.00001). Subgroup analyses only partially explained the sources of heterogeneity. The observed variability among studies likely reflects fundamental differences in case mix, definitions of ENI, clinical practice, and measurement methods, which cannot be adequately overcome by meta-analytic pooling (). Readers should interpret this global pooled proportion with extreme caution, and it cannot be generalized to all critical care environments. Subgroup analysis by study design revealed that the pooled OR was 5.05 (95% CI: 3.20–7.98), with extremely high heterogeneity (I2 = 99%, p < 0.00001). Specifically, RCTs (2 studies) showed a non-significant OR of 0.54 (95% CI: 0.10–2.91, p = 0.47, I2 = 99%), cross-sectional studies (4 studies) had an OR of 4.39 (95% CI: 2.11–9.11, p < 0.00001, I2 = 87%), prospective observational studies (29 studies) had an OR of 5.65 (95% CI: 3.02–10.56, p < 0.00001, I2 = 99%), and retrospective observational studies (7 studies) had an OR of 11.69 (95% CI: 3.83–35.69, p < 0.00001, I2 = 93%). The test for subgroup differences was statistically significant (χ2 = 9.18, df = 3, p = 0.03, I2 = 67.3%), confirming that study design was a major contributor to heterogeneity (). Subgroup analysis by country development level indicated that studies from developed countries (21 studies) had an OR of 8.38 (95% CI: 4.12–17.06, p < 0.00001, I2 = 97%), studies from developing countries (17 studies) had an OR of 2.83 (95% CI: 1.39–5.74, p = 0.004, I2 = 99%), and international multicenter studies (2 studies) had a non-significant OR of 0.39 (95% CI: 0.04–3.73, p = 0.42, I2 = 99%). The test for subgroup differences was statistically significant (χ2 = 9.08, df = 2, p = 0.01, I2 = 78%), confirming that national level is a major source of heterogeneity (). Subgroup analysis by enteral feeding route showed that gastric/intestinal mixed feeding studies (16 studies) had an OR of 2.18 (95% CI: 1.26–3.78, p = 0.005, I2 = 99%), gastric-only feeding studies (18 studies) had an OR of 7.32 (95% CI: 2.26–23.74, p = 0.0009, I2 = 99%), and studies without reported feeding route (7 studies) had an OR of 5.68 (95% CI: 1.99–16.23, p = 0.001, I2 = 95%). The test for subgroup differences was not statistically significant (χ2 = 4.89, df = 2, p = 0.09, I2 = 59.1%), there was insufficient statistical evidence to conclude that feeding route explained the observed heterogeneity, although a borderline trend was noted (p = 0.09) (). As clearly listed in the Definition column of , operational time thresholds for ENI differed substantially across 41 included studies. The number of studies, total sample and crude prevalence range within each stratum are summarized in . Thirty-two studies failed to report a standardized minimal interruption duration; only three adopted a ≥ 10 min threshold, three used a ≥ 15 min cut-off, and three applied strict standards of ≥30 min or ≥1 h. Crude ENI prevalence varied markedly across strata, the subgroup without unified definitions ranged from 11.2 to 100%, while studies with strict ≥30 min criteria reported a minimum incidence as low as 18.6%, which clearly demonstrates that the stringency of ENI definition directly alters the detected frequency of ENI. Subgroup analyses based on clinically meaningful ENI durations yielded numerically different pooled effects without statistical intergroup differences (χ2 = 3.55, df = 3, p = 0.31, I2 = 15.4%). This indicates inconsistent duration criteria only explain a small portion of the extreme overall heterogeneity (I2 = 99%), with patient case mix and institutional protocols as major confounders. Similarly, subgroup analyses stratified by ICU type and feeding mode also failed to resolve substantial between-study heterogeneity (). The global pooled prevalence of 79.2% is subject to definitional bias and should be interpreted with reference to duration subgroups rather than alone. To further verify the stability of prevalence estimates restricted to clinically meaningful sustained feeding holds, we performed stratified leave-one-out sensitivity analyses separately for three groups with explicit ENI duration cut-offs (). For the mild clinical threshold (ENI ≥ 10 min), the random-effects pooled OR remained statistically significant (OR = 2.93, 95%CI 1.23–6.99, p = 0.02). After removing the outlier study within this subgroup, between-subgroup heterogeneity was eliminated (I2 = 0%), and consistent significant pooled effects were observed under both fixed- and random-effects models. For the moderate threshold (ENI ≥ 15 min), the pooled OR remained significantly elevated (OR = 20.04, 95%CI 1.13–353.94, p = 0.04); exclusion of the outlier study fully eliminated subgroup heterogeneity without altering the positive association. For the strict long-duration definition (ENI ≥ 30 min, representing prolonged interruptions linked to substantial caloric deficits), the pooled effect turned non-significant (OR = 1.04, 95%CI 0.18–6.05, p = 0.97), which was largely attributable to the limited number of eligible studies in this stratum. Collectively, stratified sensitivity analyses focusing on clinically meaningful ENI durations further confirmed that the high overall pooled prevalence of 79.2% was largely driven by studies counting all trivial brief feeding pauses without standardized time thresholds. Sensitivity analysis using a leave-one-out approach confirmed the robustness of the results, with no single study materially influencing the pooled estimate (). Funnel plot asymmetry suggested potential publication bias and small-study effects (). This was supported by Egger’s regression test (intercept = 9.29, SE = 1.26, p < 0.001), indicating significant asymmetry. Consistently, the Egger’s regression plot showed clear deviation from the horizontal axis (). Together, these findings suggest possible publication bias and a potential overestimation of the pooled effect size. Reasons for ENI In this study, the reasons for ENI were categorized into nine main groups: airway management procedures, bedside diagnostic/therapeutic procedures, medical care procedures, gastrointestinal factors, hemodynamic factors, device-related problems, radiological procedures, surgical procedures/preparation, and other or undocumented causes (). The most frequent causes were radiological procedures, airway management, and surgical procedures. Radiological procedures were reported in 12 studies, with a median incidence of 15.0% (IQR, 8.8–27.4%), airway management in 11 studies had a median of 27.1% (IQR, 16.2–43.4%), and surgical procedures or preoperative preparation in 10 studies had a median of 10.4% (IQR, 4.9–19.6%). Other notable causes included gastrointestinal intolerance (16.2%, IQR 11.6–29.4%, n = 14), device-related problems (5.8%, IQR 3.0–9.6%, n = 8), hemodynamic factors (6.0%, IQR 0.9–9.6%, n = 5), and bedside diagnostic/therapeutic procedures (9.4%, IQR 5.2–21.8%, n = 8). Three studies reported other or undocumented causes exceeding 40% of interruptions, reflecting variability in reporting practices. Overall, these findings indicate that procedures both inside and outside the ICU, airway management, and gastrointestinal intolerance are the predominant contributors to ENI, suggesting potential targets for interventions to reduce interruption frequency. Impact of ENI on clinical outcomes Three studies (, , ) reported data on energy adequacy (). We pooled the energy adequacy rates using a fixed-effects model, as between-study heterogeneity was low (I2 = 24%; p = 0.27). The pooled MD was −4.21 (95% CI, −7.71 to −0.71), indicating that ENI was associated with a significantly lower energy adequacy rate compared with non-ENI (p = 0.02) (). Three studies (, , ) reported ICU LOS data. Given the low between-study heterogeneity (I2 = 28%; p = 0.25), a fixed-effects model was applied in accordance with standard meta-analytic practice. Indicating the pooled estimate did not reach statistical significance, yet the 95% CI [−2.05, 1.23] was wide and could not exclude either prolonged or shortened ICU stay related to ENI (p = 0.62) (). Four studies (, , , ) reported data on short-term mortality (including 28-day, 30-day, and in-ICU mortality). Given the substantial between-study heterogeneity (I2 = 88%; p < 0.0001), a random-effects model was employed in accordance with standard meta-analytic practice. Although the point estimate suggested a potential increase in mortality risk (RR = 1.86), the wide 95% CI indicated high imprecision and low statistical power. Therefore, we cannot rule out a clinically meaningful causal relationship between ENI and short-term mortality due to severe imprecision of the pooled estimate (). Sensitivity analysis was performed by excluding the study 26 with the most extreme effect size, which markedly reduced heterogeneity (I2 = 34%; p = 0.22). The updated pooled RR was 1.00 (95% CI, 0.67–1.49; Z = 0.00; p = 1.00), confirming that the primary finding of no significant association remained robust (). Certainty of evidence GRADE assessments for primary outcomes explored in meta-analyses are presented in . Given that most included studies were observational in design, the initial certainty of evidence was rated as low. The certainty was further downgraded according to risk of bias, inconsistency, indirectness, imprecision, and publication bias. For energy adequacy, the certainty of evidence was rated as low due to moderate risk of bias and limited number of studies. For ICU length of stay, the certainty of evidence was also rated as low, mainly driven by imprecision and limited sample size. For short-term mortality, the certainty of evidence was rated as very low, due to substantial heterogeneity (I2 = 88%), serious imprecision, and evidence of publication bias. Overall, the certainty of evidence for all clinical outcomes ranged from low to very low, indicating that further high-quality studies are highly likely to change the estimates. Qualitative results Among the 41 included studies, 13 reported biochemical markers related to ENI (). Eight studies (, , , , , , , ) all mentioned this biomarker, serum albumin, primarily reflecting nutritional status. Inflammatory markers were reported in five studies (, , ), most commonly C-reactive protein (CRP), sometimes combined with albumin or prealbumin. Three studies evaluated metabolic or pancreatic-related markers, including blood glucose (, ) and amylase/lipase (). Other less frequently reported indicators included prealbumin (, , ), transferrin (), white blood cell count (), serum lactate (), and magnesium (). Overall, biochemical markers were inconsistently reported across studies, with no standardized measurement approaches or consistent associations with ENI identified. Most studies also proposed interventions aimed at reducing ENI and improving nutritional delivery. These recommendations generally focused on four key aspects. First, several studies (, , ) emphasized the importance of standardized EN protocols, including clear criteria for ENI and resumption, standardized gastric residual volume thresholds, and structured feeding algorithms. Second, optimization of peri-procedural fasting management was frequently recommended, such as minimizing unnecessary fasting before diagnostic or therapeutic procedures and restarting feeding promptly after procedures (, , ). Third, improved management of gastrointestinal (GI) intolerance-induced interruptions and procedure-related operational withholding of EN was highlighted, as these two distinct root causes require differentiated targeted intervention bundles; corresponding measures include the rational use of prokinetic agents, appropriately elevated gastric residual volume thresholds, and strengthened tube fixation and maintenance (, , ). Finally, several studies advocated multidisciplinary nutritional management, including the involvement of nutrition support teams, regular staff training, and formal implementation of volume-based feeding as a core compensatory strategy to offset caloric deficits accumulated during repeated ENI (, , ). Collectively, these tiered, cause-tailored strategies aim to reduce avoidable interruptions and improve the adequacy of EN delivery in critically ill patients. Discussion This systematic review and meta-analysis of 41 studies including 8,260 critically ill adult patients was conducted to determine the pooled prevalence, contributing reasons, key outcome indicators, and evidence-based interventions for ENI in the ICU. The pooled ENI prevalence of 79.2% cannot be treated as a universal benchmark for ICU nutrition practice. Extreme inter-study heterogeneity (I2 = 99%) could only be partially explained by study design and national level. Broad discrepancies in patient populations, inconsistent ENI duration cutoffs, variable procedural workflows and uneven clinical documentation standards collectively cause ENI crude rates to range from 11.2 to 100%. Retrospective observational studies showed a higher pooled effect than prospective designs, likely due to more complete capture of interruption events (, ). Developed countries reported higher ENI prevalence than developing countries, possibly related to more frequent procedures and stricter monitoring (, ). Gastric feeding alone was associated with higher ENI risk than combined gastric/intestinal feeding, mainly due to frequent GRV monitoring and intolerance (, ). In addition, detected publication bias likely inflates the pooled proportion, overstating the overall burden of ENI worldwide. To further quantify the bias induced by divergent ENI duration definitions, we performed stratified leave-one-out sensitivity analyses restricted to clinically meaningful sustained ENI with explicit time thresholds. This stratified sensitivity evidence directly confirms that the inflated global pooled prevalence of 79.2% is largely driven by studies that count all trivial, transient short feeding pauses without standardized minimum duration criteria, rather than prolonged, clinically impactful interruptions that carry substantial risks of cumulative caloric deficits. Instead of relying solely on the single global estimate, clinicians should select subgroup data matching their local ICU context: ENI prevalence is generally higher in high-income countries and units using exclusive gastric feeding, while lower rates are observed in international multicenter cohorts and mixed gastroduodenal feeding protocols. Standardized ENI operational definitions and large prospective multicenter surveillance are required to generate more generalizable incidence data in future research. The leading causes of ENI were airway management procedures (median 27.1%), radiological/imaging procedures (15.0%), patient-related gastrointestinal factors (16.2%), and surgical procedures with preoperative preparation (10.4%). Airway-related procedures were identified as the most frequent cause of ENI. Although such procedure-related interruptions may be clinically unavoidable, their duration and frequency could still be minimized through standardized protocols, optimized peri-procedural fasting strategies, and prompt resumption of enteral feeding after procedures (, ). This also indicates the importance of optimizing peri-procedural feeding strategies, such as minimizing fasting duration and promoting early resumption of feeding. Gastrointestinal intolerance was reported across 14 studies, indicating the presence of modifiable factors that may be addressed through improved feeding protocols and pharmacological interventions. Notably, approximately 13.3% of interruptions were undocumented or unexplained, reflecting inconsistencies in clinical documentation and suggesting the presence of potentially avoidable interruptions. This underscores the need for standardized recording practices and quality improvement initiatives. Our meta-analysis showed that ENI was significantly associated with lower energy adequacy in critically ill patients, supporting the view that ENI directly compromise nutritional delivery and may contribute to cumulative energy deficits. Previous studies (, ) have suggested that inadequate or delayed enteral nutrition may be associated with worse clinical outcomes. However, evidence regarding the timing and intensity of early nutrition should be interpreted according to the phase of critical illness. In a large multicentre prospective cohort study (), early nutrition support within 48 h of ICU admission was independently associated with increased 28-day mortality, longer ICU lengths of stay, and higher crude mortality, even after multivariable and propensity score adjustments. This finding does not necessarily contradict the present study’s overall thesis. Rather, it suggests that in the very early acute phase, particularly among patients with greater baseline illness severity, hemodynamic instability, shock, or evolving organ dysfunction, ultra-early or aggressive nutritional support may reflect disease severity and may not confer immediate benefit. In this phase, ENI may sometimes be clinically necessary for safety reasons. By contrast, after initial stabilization and during the post-acute or recovery phase, repeated and avoidable ENI may become an important contributor to underfeeding. Insufficient energy delivery during this period can lead to cumulative caloric and protein deficits, which have been consistently associated with adverse outcomes in critically ill patients. The goal is not to promote indiscriminate uninterrupted feeding during all phases of critical illness, but to minimize preventable interruptions and optimize nutritional delivery when EN is clinically appropriate and safe. These considerations provide a more nuanced interpretation of the relationship among early nutrition, ENI, energy adequacy, and prognosis. For short-term mortality, the pooled RR was 1.86, suggesting a potential increase in risk. However, the 95% CI was extremely wide (0.67–5.11), indicating high imprecision and low statistical power. Similarly, the pooled mean difference for LOS crossed the null value (MD = −0.41, 95% CI −2.05 to 1.23), with overlapping intervals for both shortened and prolonged ICU admission, also reflecting limited ability to draw firm conclusions. Neither result constitutes evidence of no true causal association. These findings should be interpreted with extreme caution for three key reasons. First, only a small number of studies reported data on ICU LOS and short-term mortality separately, severely limiting overall statistical power and the precision of pooled effect estimates. Second, substantial heterogeneity existed for both outcomes, especially mortality (I2 = 88%), reflecting meaningful clinical and methodological discrepancies across trials that could mask underlying true associations. Third, included work varied widely in enrolled patient cohorts, baseline illness severity, ICU setting, EN operational definitions, and confounder adjustment strategies; short-term mortality endpoints were also inconsistent (in-ICU, 28-day, 30-day timeframes). Varied follow-up windows introduce additional clinical heterogeneity that may dilute any independent link between ENI and hard endpoints. This contradictory pattern across individual studies is largely attributable to methodological variation. Research detecting harmful outcomes tended to enroll patients with recurrent, long-duration ENI and severe cumulative calorie deficits, or more homogeneous high-risk cohorts where nutritional impairment exerted clearer prognostic effects. In contrast, null findings often stemmed from small sample sizes, overly broad ENI criteria counting trivial transient pauses, or highly mixed ICU populations where multiple competing risk factors diluted the independent impact of ENI. ICU mortality and LOS are multifactorial outcomes driven by numerous non-nutritional variables; rigorous standardization of ENI exposure, confounder adjustment and clinical endpoints is required to isolate the specific prognostic role of ENI. Taken together, our pooled results reflect substantial evidentiary uncertainty rather than proof that ENI has no clinical impact. Large prospective cohorts and RCTs with unified ENI and outcome definitions are necessary to clarify whether minimizing avoidable ENI improves patient-centered clinical outcomes. Nutritional and biochemical indicators included in this systematic review mainly comprised serum albumin, serum prealbumin, serum total protein, serum hemoglobin, serum transferrin, and serum cholesterol as nutrition-related biomarkers as well as serum alanine transaminase (ALT), serum aspartate aminotransferase (AST), serum total bilirubin, and serum amylase as organ function-related biomarkers. Significant heterogeneity existed in the definition, measurement timing, cutoff values, and reporting methods of the above indicators across studies, with a lack of standardized measurement and reporting criteria. Most studies proposed interventions to cut ENI and optimize nutrition delivery. For higher clinical practicability, future integrated intervention packages should adopt targeted, actionable measures against common avoidable interruptions. Roll out standardized enteral nutrition protocols with clear rules to start or pause feeds and nurse-led feeding workflows. Apply volume-based feeding to make up calorie gaps from repeated feeding breaks and hit daily nutrition targets steadily. Shorten peri-procedural fasting by setting procedure-specific fasting limits and resuming feeds quickly once patients are stable post-procedure. Abolish unnecessary routine GRV checks. Manage GI complications with rational prokinetics, post-pyloric feeding and secure tube care to stop unplanned interruptions. Engage ICU dietitians and multidisciplinary nutrition teams to monitor real-time feeding status, spot preventable pauses and adjust feeding plans timely. These layered, cause-oriented tactics can effectively lower unnecessary ENI and raise feeding sufficiency for critically ill patients. This study comprehensively searched seven databases, including both English and Chinese literature, and used logit-transformed single-arm meta-analysis to ensure robustness. Subgroup and sensitivity analyses were rigorously performed, and causes and interventions were systematically summarized. However, several limitations should be acknowledged. First, extreme and largely unexplained heterogeneity (I2 = 99%) was detected across included studies, which was only partially explained by study design and feeding route. Such extensive heterogeneity reflects fundamental differences in case mix, definitions, and clinical practices across studies and cannot be resolved by quantitative pooling, significantly limiting the robustness and clinical utility of the pooled prevalence estimates. Extreme between-study heterogeneity was present for both prevalence and mortality outcomes and could not be fully explained, which further limits the validity and clinical usefulness of our pooled findings. Second, significant publication bias was observed (Egger’s test p < 0.001), which may lead to overestimation of the pooled prevalence of ENI. Third, the definition of ENI varied substantially across studies, which limited direct comparability and contributed to heterogeneity. Fourth, the number of studies reporting clinical outcomes was relatively small, with only 3–4 studies included in each meta-analysis of energy adequacy, ICU LOS, and short-term mortality. Fifth, high heterogeneity was observed in the mortality analysis (I2 = 88%), and most included studies were observational, which limits causal inference between ENI and clinical outcomes. In addition, biochemical marker data were limited, inconsistently reported, and lacked standardized thresholds. Sixth, for short-term mortality, the pooled effect size suggested a potential increase in risk (RR = 1.86), but the wide confidence interval indicated high imprecision and low statistical power. Therefore, the finding reflects uncertainty rather than evidence of no association. Furthermore, our analysis was primarily limited to energy delivery, reflecting the reporting patterns in the included studies. We acknowledge a notable limitation regarding the lack of commentary on protein adequacy. Given recent advancements in critical care nutrition, protein provision is increasingly recognized as being at least as critical as energy delivery for preserving muscle mass and modulating clinical outcomes. Future research should prioritize the rigorous reporting of both energy and protein intake to provide a more comprehensive assessment of nutritional support in the ICU. These findings have important implications for clinical practice and future research. In clinical practice, targeted interventions should be implemented to reduce the occurrence of ENI, including the establishment of standardized management protocols, reasonable shortening of peri-procedural fasting time, standardized monitoring of gastric residual volume, timely identification and intervention of feeding intolerance, and strengthening the standardized maintenance of feeding tubes. From a research perspective, future studies should aim to improve methodological consistency and evidence quality. The use of standardized definitions of ENI (e.g., ≥10 min) is essential to enhance comparability across studies. Future studies should adopt uniform outcome measures for energy adequacy, ICU LOS, and short-term mortality to reduce heterogeneity. Large-scale prospective cohort studies are needed to better characterize interruption patterns and associated outcomes, while RCTs should evaluate whether reducing ENI can improve nutritional adequacy and clinical outcomes. Furthermore, the exploration of objective biomarkers may help guide individualized feeding strategies, and future research should also focus on developing and evaluating structured quality improvement strategies to reduce avoidable interruptions. Conclusion ENI incidence varied widely across critical care investigations, with a median crude prevalence of 68.5% (IQR: 57.1–96.6%). The global pooled prevalence of 79.2% remains highly uncertain due to extreme heterogeneity and publication bias. Key causes included airway management, radiological procedures, gastrointestinal intolerance, and surgical preparation. ENI was associated with reduced energy adequacy. However, the association between ENI and ICU length of stay as well as short-term mortality remains highly uncertain due to limited outcome data, substantial heterogeneity and wide confidence intervals. Given the limited number of studies and substantial heterogeneity, the impact of ENI on these hard clinical outcomes remains uncertain. Standardized ENI definitions, optimized peri-procedural fasting, improved gastrointestinal management, and multidisciplinary nutrition teams may reduce unnecessary interruptions. Future high-quality studies with unified designs and outcome measures are warranted to optimize EN delivery in critical care.