Clarifying Geriatric Nutrition Risk Index Classification in Critically Ill Patients with Severe Coronavirus Disease 2019

With great interest, we read the multicenter study by Yeo et al. [1], which investigated the prognostic significance of the geriatric nutritional risk index (GNRI) in patients with severe coronavirus disease 2019 (COVID-19) and respiratory failure admitted to the intensive care unit (ICU). Their study found that lower GNRI scores (<98) were independently associated with higher in-hospital mortality in patients over 19 years (hazard ratio [HR], 1.64) and those aged more than 65 years (HR, 1.66), underscoring the importance of nutritional assessment in this population. While their study is robust, we would like to offer some comments that might enhance its clinical applicability.

The GNRI, first introduced by Bouillanne et al. [2], is a widely adopted nutritional screening tool for elderly patients. Numerous studies have demonstrated a stepwise correlation between these GNRI categories and adverse clinical outcomes, including increased mortality among critically ill and elderly populations [3-5]. Despite its broad use, the literature reflects a variety of approaches to applying GNRI, often using different cutoff values. These include the original four-category classification, thresholds determined through receiver operating characteristic (ROC) curve analysis, and patient distribution-based methods. In some studies, GNRI is analyzed as a continuous variable.

The four-category GNRI classification (major risk, <82; moderate risk, 82 to <92; low risk, 92 to <98; and no risk, ≥98) is widely used in clinical research. For example, Zhao et al. [3] have found that GNRI scores are independently associated with in-hospital mortality in critically ill patients with acute kidney injury, showing stepwise increasing odds ratios (ORs) of 1.41, 2.15, and 3.34 in low-, moderate-, and high-risk groups, respectively, compared to the no-risk group. Similarly, elderly trauma patients showed stepwise increased mortality risk in low-, moderate-, and high-risk GNRI groups (ORs=2.3, 2.9, 7.4, respectively) [4]. Liu et al. [5] have also confirmed that GNRI is an independent mortality risk factor in older ICU trauma patients. Compared to patients with GNRI >98, those with GNRI <82 carried a significantly higher mortality rate and length of hospital stay [5].

Besides the four-category classification, many studies have used a simplified two-group GNRI model for easier clinical use and greater statistical power. However, cutoff points for dichotomizing GNRI are different depending on disease and cohort. GNRI cutoffs can be determined in several ways:

(1) Original classification scheme: A GNRI of 98 is commonly used to identify high vs. low nutritional risk in elderly patients with COVID-19 [6], elderly inpatients [7], and elderly patients with colorectal cancer [8]. A GNRI of 92 has also been applied in hemodialysis patients to assess bone fracture risk [9];

(2) ROC curve analyses: Cutoffs derived through ROC analyses include GNRI of 91.96 in elderly patients with cancer cachexia [10], 104.25 in advanced rectal cancer [11], 89.5 in non-small cell lung cancer [12], and separate cutoffs for survival outcomes (e.g., 101 and 91 for overall and recurrence-free survival, respectively) in colorectal cancer [13]. Different cutoffs have also been reported in patients with hematologic malignancies [14] and stroke [15];

(3) Patient distribution-based methods: Some studies have divided populations by median or quantiles. Tanaka et al. [16] have split 1,489 dialysis patients at the GNRI median (87.86), finding that higher GNRI is linked to lower all-cause and infection-related mortality. Tanaka et al. [17] have categorized 1,588 hypertensive patients into tertiles, observing that lower GNRI is associated with kidney events and mortality. Matsukuma et al. [18] have used quartiles in 3,436 Japanese hemodialysis patients, showing that GNRI could independently predict mortality. GNRI has also been analyzed as a continuous variable. Meta-analyses have revealed stepwise associations between GNRI and outcomes in patients with peripheral artery disease [19] and those undergoing hemodialysis [20].

These findings highlight a key issue: optimal GNRI cutoff values for patient stratification often differ across study populations and clinical outcomes. Additionally, insights from critically ill COVID-19 patients are valuable. Thus, we recommend that Yeo et al. [1] could consider including further analyses using the original four-category GNRI classification or alternative stratifications based on ROC-derived or population-based cutoffs. This approach could enhance the study’s relevance, facilitate comparisons across studies, and support efforts to standardize GNRI use in diverse patient groups. Standardizing, or at least contextualizing, GNRI thresholds might promote a more consistent and practical application of this essential nutritional assessment tool in clinical practice.