External validation of the feeding intolerance prediction model (NOFI) in critically ill patients: A post hoc analysis of a large-scale randomized controlled trial

Feeding intolerance (FI)—the inability to tolerate enteral nutrition because of gastrointestinal symptoms and inadequate feed delivery—is common in critically ill patients and often complicates early nutrition therapy in the intensive care unit (ICU). When FI develops, clinicians face a difficult balancing act: pushing enteral feeding too quickly may worsen intolerance and lead to interruptions, while progressing too slowly may delay adequate nutrition and potentially worsen outcomes. Despite FI’s clinical importance, many ICUs still rely primarily on experience-based judgment to anticipate which patients are most likely to develop intolerance.

A new study reports the independent external validation of NOFI, a simplified, bedside prediction tool designed to identify FI risk early after ICU admission. Using only three routinely available clinical variables—primary diagnosis, APACHE II score, and acute gastrointestinal injury (AGI) grade—NOFI aims to support timely, individualized enteral nutrition decisions. This study led by Dr. Lu Ke and Dr. Dong Zhang was made available online on February 02, 2026 in the journal Journal of Intensive Medicine.

Study design and population

The investigators conducted a post hoc analysis using individual participant data from the NEED trial, a multicenter, cluster-randomized controlled study carried out across 97 ICUs in China. For this external validation, the analysis included 1,545 critically ill adults who started early enteral nutrition within 48 hours of ICU admission and received enteral feeding for at least 3 days. In this cohort, feeding intolerance occurred in 856 patients, reflecting the high clinical burden of FI in routine ICU practice.

Model performance and risk stratification

The study evaluated NOFI on three fronts that matter for real-world clinical use:

• Discrimination (who will vs will not develop FI): NOFI demonstrated moderate discriminatory performance with an AUROC of 0.723, suggesting the model can meaningfully separate higher-risk from lower-risk patients.
• Clinical utility (does the model help decision-making?): Beyond traditional accuracy metrics, the analysis assessed decision-analytic measures, including decision curve analysis and clinical impact curves, to estimate whether using NOFI could provide net benefit across clinically relevant decision thresholds.
• Risk grouping (turning probabilities into actionable categories): To enhance usability at the bedside, patients were stratified by predicted FI probability into three risk tiers: high risk (> 0.85), middle risk (0.30–0.85), and low risk (< 0.30). This structure is intended to help clinicians tailor monitoring intensity and feeding strategies according to risk.

Association with 28-day mortality

Importantly, the study examined whether NOFI-defined FI risk groups also tracked with short-term outcomes. Using multivariable Cox regression, the analysis found that patients in the high-risk group experienced substantially higher 28-day mortality compared with those in the low-risk group (HR 2.28). Mortality risk was also higher for the high-risk group relative to the middle-risk group (HR 1.54). While these findings do not prove that predicted FI risk causes death, they suggest that a high FI-risk profile may serve as an early marker of overall severity and vulnerability.

Why it matters clinically

Because NOFI relies on only three commonly documented variables, it is feasible for rapid implementation without specialized equipment, biomarkers, or complex calculations. In practice, a bedside risk estimate could help ICUs:

• Target closer monitoring for patients at high FI risk (e.g., more frequent assessment of GI symptoms and feed tolerance).
• Guide feed advancement more safely, avoiding overly aggressive escalation in those likely to fail early progression.
• Support earlier, individualized nutrition strategies, such as proactive tolerance management (where clinically appropriate), reassessment of feeding route, and consideration of formula adjustments in patients with persistent intolerance.
• Inform broader risk management, as high predicted FI risk may align with heightened short-term mortality risk and thus prompt earlier multidisciplinary evaluation.

Caution and next steps

The authors emphasize that this work represents an external validation and a post hoc analysis. The tool is intended to aid clinical judgment, not replace it, and treatment decisions should always consider the full clinical context. Future prospective studies could test whether NOFI-guided feeding strategies improve patient-centered outcomes compared with usual care, and whether integrating additional bedside data further improves performance without sacrificing simplicity.
Bottom line: In a large multicenter ICU cohort, NOFI—a three-variable bedside model—showed practical predictive performance for feeding intolerance and stratified patients by 28-day mortality risk, supporting its potential as a simple, implementable tool for early nutrition decision-making in critical care.

***

Reference

DOI: https://doi.org/10.1016/j.jointm.2025.12.004

Funding information

This study was supported by the Special Fund for Clinical Research of Wu Jieping Medical Foundation (320.6750.2022-13-2), National Natural Science Foundation of China (82200666), and Natural Science Foundation of Jilin Province, China (YDZJ202201ZYTS015).