This is a cross-sectional study in the part of the Fukushima Diabetes, Endocrinology, and Metabolism cohort (Fukushima DEM cohort). The DEM cohort recruited people with diabetes mellitus or high risk at diabetes who had visited the Department of Diabetes, Endocrinology, and Metabolism, Fukushima Medical University Hospital. The study protocol was approved by the Fukushima Medical University Ethics Committee (Number 29118). This study was conducted according to the Ethical Guidelines for Medical and Health Research Involving Human Subjects enacted by MHLW of Japan (https://www.mhlw.go.jp/file/06-Seisakujouhou-10600000Daijinkanboukouseikagakuka/0000069410.pdf and http://www.mhlw.go.jp/file/06-Seisakujouhou-10600000-Daijinkanboukouseikagakuka/0000080278.pdf) in line with the principles of the Declaration of Helsinki. The inclusion criteria of the current study were people with T2DM in the DEM cohort who had been recruited between January 2018 and December 2019. Among 795 patients who gave written informed consent, 240 patients who were either non-diabetic, had type 1 diabetes mellitus, or had secondary diabetes mellitus were excluded from the study (Supplementary Figure 1). After excluding missing data, 479 patients (265 men and 215 women) were included for a full analysis set. Various patient parameters, such as age, sex, history of diabetes, family and social history, medical checkup history, complications, medications, laboratory data, and all dates, were obtained from their paper and/or electrical medical records.

Patients visited the hospital at 1–3 month intervals and continued receiving standardized treatment by endocrinologists/diabetologists. Trained staff measured the height, body weight, blood pressure, and waist circumference of participants. Questionnaires were provided to record the data on smoking (current or former smoker or not), drinking (former or every day, sometimes, rarely, or never), regular exercise (exercise to sweat lightly for over 30 min on each occasion, two times weekly), antihypertensive drug use, anti-hyperglycemic drug use, and lipid-lowering drug use. A participant was diagnosed with diabetes mellitus when the fasting plasma glucose level is ≥126 mg/dL, the HbA1c level is ≥6.5% (48 mmol/mol), or if the participant regularly uses anti-hyperglycemic drugs. A participant was diagnosed with hypertension if the systolic blood pressure was ≥140 mmHg, if the diastolic blood pressure was ≥90 mmHg, or if she/he regularly used antihypertensive drugs. A participant was diagnosed with dyslipidemia if the high-density lipoprotein (HDL) cholesterol level is <40 mg/dL (1.0 mmol/L), the low-density lipoprotein (LDL) cholesterol level is ≥140 mg/dL (3.6 mmol/L), the triglyceride level is ≥150 mg/dL (1.7 mmol/L), or if they regularly used lipid-lowering drugs. We calculated the estimated glomerular filtration rate (eGFR) using the Japanese formula (eGFR, mL/min/1.73 m²) = 194 × serum creatinine level (mg/dL) ^−1.094 × age (years) ^−0.287 (23).

Routine anthropometry and skeletal muscle mass, handgrip strength, walking speed, and body composition of the participants were assessed by trained staff, as previously reported (24). The waist circumference was measured at the level of the umbilicus (cm) in the standing position. Handgrip strength (kg) was measured using an isokinetic dynamometer (Smedley hand dynamometer) on both hands, and the values of the non-dominant arm were used. The fat and muscle composition in the whole body, trunk, arms, and legs were assessed using a body composition analyzer (InBody 770, InBody Japan Inc.) based on the segmental multifrequency bioelectrical impedance analysis (25, 26). The time required for walking 10 m was measured as described previously with slight modifications (27, 28). Fasting blood samples were collected after overnight fasting for ≥10 h and were assayed within 1 h using automatic clinical chemical analyzers. We excluded participants whose fasting blood samples could not be obtained. Nutritional intake indices were calculated using food frequency questionnaires as previously reported (24).

The definition and diagnosis of sarcopenia were based on the Asian Working Group for Sarcopenia (AWGS): 2019 Consensus Update on Sarcopenia Diagnosis and Treatment (2). In brief, “low muscle power” was defined as handgrip strength < 28 kg for men and < 18 kg for women; the criteria for “low physical performance” was walking speed < 1.0 m/s as evaluated by the time required for walking 10 m; and “low appendicular skeletal muscle mass (ASM)” was defined as a skeletal mass index (SMI) < 7.0 kg/m² in men and < 5.7 kg/m² in women. Sarcopenia is defined by low ASM and low muscle power or low physical performance.

The AUCs and the optimal cut-off values of albumin, GNRI, and CONUT for detecting sarcopenia are shown in Figure 2 and Table 3. In all participants, the AUCs were ordered from largest to smallest as follows: GNRI > albumin > CONUT, showing that the diagnostic power of GNRI was superior to albumin. The AUC of GNRI was also statistically significant and was superior to albumin in all, men, women, BMI ≥ 22, and diabetes duration ≥ 5 years subgroups (Figure 2; Table 3).

Based on the cut-off values of the nutritional indices, we calculated the ORs for the diagnosis of sarcopenia and its components. In all participants (Table 4), univariate and multiple logistic regression analysis showed that albumin was associated with low handgrip strength and sarcopenia but not with low SMI and low walking speed. The cut-off values of GNRI of 98 and 105 were associated with the diagnosis of low SMI, low handgrip strength, and sarcopenia in univariate and multiple logistic regression analysis (models 1 and 2). The cut-off value of CONUT was not associated with the diagnosis of sarcopenia and its components.

Multiple logistic regression analysis (Model 2 in Table 4) on the associations of nutritional indices with the diagnosis of sarcopenia and its components in the subclasses of current participants is shown in Table 5 (ORs in unadjusted and Model 1 not shown). The cut-off values of albumin were associated with a diagnosis of sarcopenia only in women but not in the men, BMI < 22 and BMI ≥ 22 subclasses. The cut-off values of GNRI were associated with a diagnosis of sarcopenia in the men (105), women (105), BMI < 22 (102), BMI ≥ 22 (112), and diabetes duration ≥ 5 years (105) subgroups, but not in the diabetes duration < 5 years (Table 5). When using an originally reported low nutrition-related cut-off at 98 (15), the GNRI was associated with sarcopenia in the men, women, and diabetes duration ≥ 5 years subgroups but not in the BMI <22, BMI ≥ 22, and diabetes duration < 5 years subgroups. The cut-off values of CONUT were not associated with sarcopenia and its components in these subclasses, except in patients with BMI < 22.

There are reports on the associations between the nutritional indicators such as SGA, MNA, MUST, and NRS2002 and diagnosis of sarcopenia (20, 30, 31). These reports repeatedly indicated that malnutrition determined by these indices and diagnosis of sarcopenia are closely linked (20, 30, 31). However, these indicators require interviews for history of body weight and dietary assessment, limiting clinical application. There are more simpler indices such as albumin and prealbumin. Xiu et al. reported that low prealbumin levels were associated with an increased risk for sarcopenia in older men with T2DM (32). However, prediction of undernutrition using these simple indices may be limited to some extent by potential confounding factors such as other clinical conditions (20, 30, 31). In our study, we adopted GNRI (15, 21, 22), CONUT (19), and albumin which are easily available in daily clinical practice for the assessment of undernutrition.

The cut-off values of GNRI were associated with a diagnosis of sarcopenia in multiple logistic regression analysis after correcting for potential cofounders. There are previous reports on the association between GNRI with the diagnosis of sarcopenia and its components. In Korean patients on hemodialysis, a GNRI of 97–101 (OR 0.064, 95% CI 0.005–0.883, compared to GNRI ≤ 96, p = 0.040) was associated with a lower sarcopenia risk (33). Xiang et al. reported that the overall diagnostic performance was the best for mid-arm circumference, followed by GNRI, calf circumference, BMI, and the worst for triceps skinfold thickness and albumin in detecting sarcopenia in community-dwelling Chinese adults aged 50 or older (34). The following two reports agreed with our findings, showing the association between low GNRI and the diagnosis of sarcopenia (13, 21). Takahashi et al. reported that a GNRI < 98 was related to the prevalence of sarcopenia [adjusted odds ratio, 4.88 (95%CI: 1.88–12.7), p = 0.001] in Japanese patients with T2DM (13). Matsuura et al. reported that a higher GNRI was associated with a lower risk of sarcopenia in older men and women with diabetes [multivariate-adjusted OR, 0.892; 95% CI, 0.839–0.948 for male; adjusted OR, 0.928; 0.876–0.982 for female] (21). However, the GNRI threshold and its relevance in subclasses for sarcopenia were not considered in the two studies. We further assessed the diagnostic utility of the cut-off values of GNRI to detect sarcopenia. The GNRI cut-off values of 105 and 98 were associated with a diagnosis of sarcopenia similarly in all participants and in the men and women subclasses. However, a GNRI of 102 in patients with BMI < 22 and a GNRI of 112 in patients with BMI ≥ 22, but not that of 98, were associated with a diagnosis of sarcopenia. Collectively, it is suggested that the optimal cut-off values of GNRI depend on the clinical characteristics of the target population.

There were reports indicating the association between low GNRI and low muscle power, and low muscle mass (22, 35). In Chinese elderly people, a low GNRI was associated with a higher incidence of low muscle mass (34). In Italian institutionalized elderly, GNRI was correlated with arm muscle area, handgrip strength, and handgrip strength/arm muscle area (22). Compared to other indices such as the ESPEN, GLIM, or SGA criteria, GNRI appears simple but still considers the serum albumin level in addition to current and ideal body weight (15).

The mechanisms by which low GNRI correlates with sarcopenia may include the lack of supply of muscle building blocks due to undernutrition and the involvement of chronic inflammation of muscle due to undernutrition (14, 36, 37). In Germany's older patients, a higher risk GNRI was associated with increased CRP levels (p < 0.05) and low lymphocyte counts (p < 0.05) after multivariable adjustment (36). Subclinical catabolic and inflammatory states, which are associated with chronic disease, led to increased production of catabolic cytokines, increased muscle catabolism, and decreased appetite with a negative effect on albumin levels (38–40). A reduction in serum albumin can therefore be a consequence of poor nutritional status or inflammation/disease (38–40).

Although low albumin has long been recognized as a crude indicator of undernutrition status (41), it is an unreliable indicator of nutritional status because it may be more related to inflammation or hydration status than to malnutrition (15, 42, 43). The GNRI was developed by Bouillanne et al. in 2005 to provide a prognostic nutritional index that enables quantitative determination of the risk of nutrition-related morbidity and mortality in elderly patients at admission into a geriatric hospital (15). They described that GNRI is not an index of malnutrition, but it is a “nutrition-related” risk index because GNRI scores are correlated to a severity score that considers nutritional status-related complications such as bedsores and infections (15). Importantly, GNRI is also based on measurements of weight loss, which are strong independent risk factors for comorbidities and mortality in older persons (44, 45). Applying the status of weight in the formula, GNRI can be a better predictor than serum albumin for low SMI and sarcopenia in the elderly with T2DM with a median age of 80 [IQR 74, 86]. The CONUT formula includes blood biomarkers such as serum albumin concentration, cholesterol level, and lymphocyte count but does not include body composition measures such as BMI (19). Because assessment of muscle mass is critical in considering the diagnosis of sarcopenia, the diagnostic power of CONUT can be low for detecting sarcopenia. The cut-off values of GNRI were rather different in the subclasses of T2DM participants. The GNRI cut-off values were associated with low SMI and diagnosis of sarcopenia in men and women (105), and BMI ≥ 22 (112) subclasses. While the GNRI cut-off value of 102 was associated with low handgrip strength and diagnosis of sarcopenia in patients with BMI < 22. When using an originally reported low nutrition-related cut-off of 98 (15), GNRI was associated with sarcopenia in men and women subclasses but not in the BMI < 22 and BMI ≥ 22 subclasses. The cut-off values of CONUT were not associated with a diagnosis of sarcopenia and its components in these subclasses, except in the diagnosis of sarcopenia in BMI < 22. The individual components of the CONUT score are shown in Supplementary Table 1. The distribution of scoring of lymphocyte count, total cholesterol, and albumin was not different between the sarcopenia– and sarcopenia+ groups. As discussed above, if the status of weight in the formula is applied, GNRI could be a better predictor for sarcopenia in the subclasses BMI < 22 and BMI ≥ 22.

Our study had some limitations. First, because this study was conducted at a single university hospital, there may be a bias toward patients with a high risk of developing sarcopenia. Second, the number of patients and duration of observation might have been insufficient to assess the development of sarcopenia. Third, this was a cross-sectional observational study, and low albumin followed by lower GNRI and the prevalence of low SMI and sarcopenia are mutually well-correlated. Therefore, we could not determine a cause-and-result relationship. Further large-scale longitudinal studies are needed to corroborate the results of this study.