Abstract
Background and aims
There are few reports on the association between malnutrition and hypoglycemia. The geriatric nutritional risk index (GNRI) allows risk classification by morbidity and mortality resulting from conditions often associated with malnutrition in older individuals. However, the association between GNRI and hypoglycemia is unclear. This study examined the associations between nutrition-related risk and hypoglycemia among older individuals with type 2 diabetes (T2D) using diabetes medication.
Methods
This single-center historical cohort study included hospitalized patients aged ≥ 65 years with T2D on medication. Nutrition-related risk was assessed using the GNRI and classified into four risk groups. Hypoglycemia and serious hypoglycemia were determined by oral or intravenous glucose intake and blood glucose < 3.9 mmol/L (70 mg/dL) as hypoglycemia, among them blood glucose < 3.0 mmol/L (54 mg/dL) as serious hypoglycemia. Data were recorded at least once during hospitalization.
Results
Patients who met the criteria (n = 1.754) were included in the study. The participants median age was 75.0 years. During the study, 81 patients (4.6%) experienced hypoglycemia and 7 patients (0.4%) experienced serious hypoglycemia. Hypoglycemia was observed in patients in the major risk (16.0%), moderate risk (9.7%), low risk (5.2%), and no risk (1.5%) groups (p for trend < 0.001). After adjusting for other risk factors, the hazard ratios of hypoglycemic among people with major, moderate, and low risk were 5.50, 3.86, and 2.55, respectively.
Conclusions
Hypoglycemia increased with increasing nutrition-related risk among older individuals with T2D using diabetes medication. The GNRI is a simple and useful assessment tool in the clinical setting.
Keywords: Hypoglycemia, Diabetes mellitus, Geriatric nutritional risk index, Malnutrition
Introduction
In the treatment of diabetes, hypoglycemia is a major obstacle to glycemic control. Hypoglycemia is a major risk factor for mortality, cardiovascular disease, and fatal arrhythmias [1, 2]. People who experience hypoglycemia during hospitalization are at higher risk for extended hospitalizations and death [3, 4]. Therefore, it is important to identify the risk factors for hypoglycemia.
According to previous studies, risk factors for hypoglycemia include age, diabetes medications, renal impairment, liver failure (LF), dementia, and strict glycemic control [5, 6]. However, the association between malnutrition and hypoglycemia is rarely reported. An association between body mass index (BMI) or serum albumin and hypoglycemia has been reported [7, 8]. In the hospital setting, these parameters are often used to screen for malnutrition. However, it has been suggested that BMI and serum albumin are not independently associated with malnutrition [9, 10]. Malnutrition is difficult to assess with a single indicator, hence multiple indicators should be used. A previous study on the association between malnutrition and hypoglycemia used the Nutritional Risk Screening 2002 (NRS-2002) [11]. However, in this study, an association between malnutrition and hypoglycemia was found only in non-diabetics, but not in diabetics. In addition, because the causal relationship was unclear as this study used a cross-sectional design,
We focused on older patients vulnerable to hypoglycemia and we used the geriatric nutritional risk index (GNRI) as a nutritional indicator. The GNRI allows risk classification by morbidity and mortality for medical conditions in older individuals associated with malnutrition [12]. Initially, the GNRI was developed for use in subacute care settings. Subsequently, its suitability in different settings, such as acute and long-term care, has become clearer [13–18]. For this reason, the use of GNRI is widespread. However, the association between GNRI and hypoglycemia is unclear.
Therefore, the present study aimed to examine the association between nutrition-related risk (assessed by GNRI on admission) and the incidence of hypoglycemia among hospitalized older patients with type 2 diabetes (T2D).
Materials and methods
Study design
This single-center historical cohort study was conducted at Soka Municipal Hospital—a secondary medical institution in Japan—and used electronic medical records.
Inclusion and exclusion criteria
The participants were hospitalized patients aged ≥ 65 years with T2D, who received diabetes medication between 1 September 2018, and 30 June 2020. The study period was defined as the date of admission until the date of discharge or 60 days from the date of admission. Patients who were either admitted to the intensive care unit or had incomplete records (missing data regarding height, weight, and serum albumin for calculating GNRI) were excluded. Hypoglycemia is associated with death for patients admitted to the intensive care unit [19]. Therefore, blood glucose targets in intensive care units are higher than those used in general wards, and care is taken to avoid hypoglycemia. To prevent this from creating selection bias, patients admitted to the intensive care unit were excluded from the present study. Among the subjects of this study, 22 died (17 without hypoglycemia and 5 with hypoglycemia). Those who died in the general ward were included in the study because we did not think it would cause selection bias. The extracted data included demographics, anthropometrics, laboratory values, date of admission, diagnosis, and medication.
Definitions of variables
The explanatory variable entailed nutrition-related risks were assessed using the GNRI screening tool and classified into four groups according to previous studies [12]: “major risk” (< 82), “moderate risk” (82 to < 92), “low risk” (92 to ≤ 98), and “no risk” (> 98). The GNRI was calculated using serum albumin, height, and current body weight with the following formula:
where the ideal body weight = (height [m])2 × 22. In cases where the current body weight exceeded the ideal body weight, the ratio of current body weight to ideal body weight was set to 1.
To calculate the GNRI as an indicator of nutrition-related risk on admission, the values that were used were height, current body weight, and serum albumin that were recorded between 60 days before and 7 days after the date of admission, before the onset of hypoglycemia, and close to the date of admission.
Hypoglycemia, which was the outcome variable, was defined as oral or intravenous glucose intake and a blood glucose level < 3.9 mmol/L (70 mg/dL). Among them, a blood glucose level < 3.0 mmol/L (54 mg/dL) was defined as serious hypoglycemia. These data were entered into the electronic medical record at least once during hospitalization. When hypoglycemia was observed more than once, only the first event was evaluated because patients who develop hypoglycemia are more carefully managed to prevent a recurrence. This management may reduce the risk of hypoglycemia development; therefore, the validity of the proportional hazard found during this analysis cannot be ensured. The time of occurrence of hypoglycemia was set from the date of admission to the first hypoglycemia episode. In addition, blood glucose levels were measured using a glucometer.
Diabetes medications included: insulins, sulfonylureas, glinides, dipeptidyl peptidase-4 inhibitors, α-glucosidase inhibitors, sodium-glucose transport protein 2 inhibitors, biguanides, thiazolidines, and glucagon-like peptide-1 receptor agonists. Because insulin and sulfonylureas are important factors in hypoglycemia, we conducted subgroup analyses stratified by patients using and patients not using this agent. The oral nutrition supplement comprised a high-calorie (1–1.6 kcal/mL), liquid nutrition product containing carbohydrates, protein, and fat, which were measured as ordered with meals.
High C-reactive protein (CRP) levels were defined as a recorded measurement of > 2 mg/dL by day 7 after admission. Previous studies have reported that inflammation with CRP levels > 2 mg/dL was strongly associated with hypoalbuminemia [20]. Therefore, high CRP was used to adjust for confounding. If CRP was not measured, it was considered as an absence of inflammation.
Chronic kidney disease (CKD) was defined as an estimated glomerular filtration rate (eGFR) < 60 mL/min/1.73 m2, while severe CKD was defined as an eGFR < 30 mL/min/1.73 m2. We used the Japanese eGFR estimation formula shown below:
LF was defined as the registered diagnosis of LF or cirrhosis (type B cirrhosis, type C cirrhosis, alcoholic cirrhosis, primary biliary cirrhosis, autoimmune cirrhosis, or other cirrhosis of the liver) or receiving medications for cirrhosis such as tolvaptan (use in cirrhosis), lactitol hydrate, or branched-chain amino acid preparations. Dementia was defined as the registered diagnosis of dementia or receiving medications for dementia such as donepezil hydrochloride, memantine hydrochloride, galantamine hydrobromide, or rivastigmine.
Statistical analyses
In the baseline population characteristics, the distributions of continuous variables were evaluated for normality using the Kolmogorov–Smirnov test. Continuous variables were described as medians (interquartile ranges), and categorical variables were presented as numbers and percentages of the total data. Baseline nutrition-related risk values were compared using the Kruskal–Wallis test for continuous variables and Fisher’s exact test for categorical variables.
The association between hypoglycemia and other variables was analyzed. During the univariate analysis, for the incidence of hypoglycemia, continuous variables were categorized and calculated using the median as the cutoff. Categorical variables were analyzed for associations using Fisher’s exact test. The Cochran-Armitage trend test was used to determine any associations of trends between groups. The cumulative incidence of hypoglycemia by GNRI class was estimated using the Kaplan–Meier method. In the multiple regression analysis, Cox regression analyses were used to estimate the hazard ratio (HR) and 95% confidence interval (CI). We entered the incidence of hypoglycemia as the outcome variable and nutrition-related risk classification, age, insulin use, severe CKD, dementia, and high-CRP as the explanatory variables. These variables were selected due to their association with hypoglycemia in previous studies [5, 6] and the large difference between the groups that did and did not experience hypoglycemia in this study. Finally, CRP was added as the explanatory variable because it affects serum albumin levels [20].
The GNRI formula included serum albumin and BMI. Therefore, we compared the association of these variables with hypoglycemia. The receiver operating characteristic (ROC) curves for each continuous variable were drawn, and the areas under the curve (AUCs) were compared. Furthermore, we have used Cox regression analysis to compare serum albumin (< 3.0, 3.0 to < 3.5, 3.5 to < 3.8, and ≥ 3.8 g/dL) and BMI (< 19, 19 to < 22, 22 to < 24, and ≥ 24 kg/m2) HRs. These were categorized according to previous studies [12]. The reason for the categorization and comparison is that categorization is easier to consider than continuous values in the assessment of nutrition-related risk and is more useful in clinical practice.
All statistical analyses were performed using EZR version 1.51 [21], an extended version of the R Commander designed to add statistical functions. p values < 0.05 were considered statistically significant.
Results
Characteristics of the study population
The analysis included 1,754 patients who met the inclusion criteria with a median age of 75.0 years (70.0–79.0). The population characteristics according to nutrition-related risk classification by GNRI are shown in Table 1. The variance between the date of measurement of height, weight, and serum albumin used to calculate GNRI and the date of admission was in the following quartiles: height: 0.0–0.5 days, weight: 0.0–0.6 days, and serum albumin: −4.0 to 1.0 days. The nutrition-related risk that was assessed by GNRI, revealed major risk for 94 (5.4%) patients, moderate risk for 289 (16.5%) patients, low risk for 481 (27.4%) patients, and no risk for 890 (50.7%) patients.
Table 1.
Baseline population characteristics and GNRI class
| Factor | Group | All | GNRI class | |||||
|---|---|---|---|---|---|---|---|---|
| Major risk | Moderate risk | Low risk | No risk | p value | ||||
| Total, n | 1754 | 94 | 289 | 481 | 890 | |||
| Demographics | Age, years (IQR) | 75.0 (70.0– 79.0) | 76.5 (72.3–82.0) | 76.0 (71.0–80.0) | 75.0 (71.0–79.0) | 74.0 (70.0–79.0) | < 0.001 | |
| Sex, n (%) | Male | 1231 (70.2) | 62 (66.0) | 204 (70.6) | 329 (68.4) | 636 (71.5) | 0.52 | |
| Length of stay, days (IQR) | 7.0 (2.0–13.0) | 15.5 (10.0–22.8) | 12.0 (6.0–19.0) | 7.0 (3.0–14.0) | 4.0 (2.0–10.0) | < 0.001 | ||
| Diabetes medications | Insulin, n (%) | Total | 533 (30.4) | 39 (41.5) | 113 (39.1) | 156 (32.4) | 225 (25.3) | < 0.001 |
| Rapid or short‑acting | 38 (2.2) | 0 (0.0) | 13 (4.5) | 12 (2.5) | 13 (1.5) | 0.008 | ||
| Long‑acting | 133 (7.6) | 14 (14.9) | 23 (8.0) | 34 (7.1) | 62 (7.0) | 0.048 | ||
| Rapid or short-acting + long‑acting | 206 (11.7) | 12 (12.8) | 36 (12.5) | 63 (13.1) | 95 (10.7) | 0.56 | ||
| Premixed | 54 (3.1) | 2 (2.1) | 10 (3.5) | 17 (3.5) | 25 (2.8) | 0.81 | ||
| Rapid or short‑acting + premixed | 7 (0.4) | 0 (0.0) | 0 (0.0) | 4 (0.8) | 3 (0.3) | 0.27 | ||
| Continuous infusion | 95 (5.4) | 11 (11.7) | 31 (10.7) | 26 (5.4) | 27 (3.0) | < 0.001 | ||
| Sulfonylureas, n (%) | 193 (11.0) | 5 (5.3) | 28 (9.7) | 39 ( 8.1) | 121 (13.6) | 0.003 | ||
| Glinides, n (%) | 257 (14.7) | 25 (26.6) | 51 (17.6) | 63 (13.1) | 118 (13.3) | 0.002 | ||
| DPP-4 inhibitors, n (%) | 1247 (71.1) | 69 (73.4) | 196 (67.8) | 333 (69.2) | 649 (72.9) | 0.26 | ||
| α-Glucosidase inhibitors, n (%) | 306 (17.4) | 18 (19.1) | 48 (16.6) | 90 (18.7) | 150 (16.9) | 0.78 | ||
| SGLT2 inhibitors, n (%) | 175 (10.0) | 5 (5.3) | 25 (8.7) | 36 (7.5) | 109 (12.2) | 0.010 | ||
| Biguanides, n (%) | 245 (14.0) | 9 (9.6) | 27 (9.3) | 43 (8.9) | 166 (18.7) | < 0.001 | ||
| Thiazolidines, n (%) | 77(4.4) | 5 (5.3) | 15 (5.2) | 18 (3.7) | 39 (4.4) | 0.77 | ||
| GLP-1 receptor agonists, n (%) | 62 (3.5) | 4 (4.3) | 12 (4.2) | 19 (4.0) | 27 (3.0) | 0.71 | ||
| Nutritional therapy | ONS, n (%) | 83 (4.7) | 10 (10.6) | 25 (8.7) | 29 (6.0) | 19 (2.1) | < 0.001 | |
| Laboratory data | Creatinine, mg/dL (IQR) | 0.92 (0.73–1.27) | 1.02 (0.71–1.82) | 1.02 (0.72–1.84) | 0.96 (0.73–1.46) | 0.89 (0.73–1.12) | < 0.001 | |
| eGFR, mL/min/1.73 m2 (IQR) | 55.3 (38.1–74.5) | 47.9 (26.4–74.9) | 48.4 (28.3–72.9) | 52.1 (33.4–75.8) | 57.7 (44.3–73.8) | < 0.001 | ||
| HbA1c, % (IQR) |
7.2 (6.6–8.1) (n = 1,465) |
6.9 (6.2–8.2) (n = 79) |
6.9 (6.3–7.8) (n = 237) |
7.3 (6.6–8.1) (n = 391) |
7.3 (6.7–8.1) (n = 758) |
< 0.001 | ||
| Albumin, g/dL (IQR) | 3.9 (3.6–4.2) | 2.6 (2.5–3.0) | 3.2 (3.1–3.4) | 3.7 (3.6– 3.8) | 4.2 (4.1–4.5) | < 0.001 | ||
| BMI, kg/m2 (IQR) | 23.5 (21.1–26.1) | 19.5 (17.8– 24.2) | 21.8 (19.5– 24.9) | 22.8 (20.7– 25.6) | 24.1 (22.3– 26.7) | < 0.001 | ||
| High CRP, n (%) | > 2 mg/dL | 369 (21.0) | 54 (57.4) | 130 (45.0) | 117 (24.3) | 68 (7.6) | < 0.001 | |
| Comorbidities | Severe CKD, n (%) | eGFR < 30 | 299 (17.0) | 25 (26.6) | 76 (26.3) | 105 (21.8) | 93 (10.4) | < 0.001 |
| CKD, n (%) | eGFR < 60 | 1008 (57.5) | 60 (63.8) | 180 (62.3) | 287 (59.7) | 481 (54.0) | 0.023 | |
| Dementia, n (%) | 118 (6.7) | 15 (16.0) | 20 (6.9) | 40 (8.3) | 43 (4.8) | < 0.001 | ||
| LF, n (%) | 54 (3.0) | 15 (16.0) | 15 (5.2) | 6 (1.2) | 18 (2.0) | < 0.001 | ||
BMI body mass index, CKD chronic kidney disease, CRP C-reactive protein, LF liver failure, GNRI geriatric nutritional risk index, IQR interquartile range, ONS oral nutritional supplements
Association between hypoglycemia and the GNRI according to the univariate analysis
During the study period, 81 patients (4.6%) experienced hypoglycemia; among them, 7 patients (0.4%) experienced serious hypoglycemia. The incidence of hypoglycemia and patient characteristics are shown in Table 2. Based on the GNRI classification, the risk of hypoglycemia was major for 15 of 94 (16.0%) patients, moderate for 28 of 289 (9.7%) patients, low for 25 of 481 (5.2%) patients, and none for 13 of 890 (1.5%) patients (p for trend < 0.001). Furthermore, the risk of serious hypoglycemia was major for 2 of 94 (2.1%) patients, moderate for 3 of 289 (1.0%), low for 0 of 481 (0.0%) patients, and none for 2 of 890 (0.2%) patients.
Table 2.
Incidence of hypoglycemia and population characteristics
| Factor | Group | Total, n | Patients with hypoglycemia (blood glucose < 3.9 mmol/L [70 mg/dL]), n (%) | p value | |
|---|---|---|---|---|---|
| All | 1754 | 81 (4.6) | |||
| Demographics | Age, years | ≥ Median | 914 | 45 (4.9) | 0.57 |
| Sex | Male | 1231 | 52 (4.2) | 0.26 | |
| Length of stay, days | ≥ Median | 904 | 72 (8.0) | < 0.001 | |
| Nutrition-related risk | GNRI class | No risk | 890 | 13 (1.5) | < 0.001 |
| Low risk | 481 | 25 (5.2) | |||
| Moderate risk | 289 | 28 (9.7) | |||
| Major risk | 94 | 15 (16.0) | |||
| Diabetes medications | Insulin | Total | 533 | 43 (8.1) | < 0.001 |
| Rapid or short‑acting | 38 | 3 (7.9) | 0.255 | ||
| Long‑acting | 133 | 11 (8.3) | 0.050 | ||
| Rapid or short-acting + long‑acting | 206 | 18 (8.7) | 0.007 | ||
| Premixed | 54 | 2 (3.7) | 1.000 | ||
| Rapid or short‑acting + premixed | 7 | 1 (14.3) | 0.282 | ||
| Continuous infusion | 95 | 8 (8.4) | 0.077 | ||
| Sulfonylureas | 193 | 4 (2.1) | 0.099 | ||
| Nutritional therapy | ONS | 83 | 9 (10.8) | 0.012 | |
| Laboratory data | Creatinine, mg/dL | ≥ Median | 890 | 41 (4.6) | 1.00 |
| eGFR, mL/min/1.73 m2 | ≤ Median | 878 | 43 (4.9) | 0.65 | |
| HbA1c, % | ≤ Median | 748 | 37 (4.9) | 0.72 | |
| Albumin, g/dL | ≤ Median | 903 | 65 (7.2) | < 0.001 | |
| BMI, kg/m2 | ≤ Median | 885 | 46 (5.2) | 0.26 | |
| High CRP | > 2 mg/dL | 369 | 32 (8.7) | < 0.001 | |
| Comorbidities | Severe CKD | eGFR < 30 | 299 | 23 (7.7) | 0.009 |
| CKD | eGFR < 60 | 1008 | 45 (4.5) | 0.73 | |
| Dementia | 118 | 13 (11.0) | 0.002 | ||
| LF | 54 | 6 (11.1) | 0.035 |
BMI body mass index, CKD chronic kidney disease, CRP C-reactive protein, GNRI geriatric nutritional risk index, LF liver failure, ONS oral nutritional supplements
We examined the interaction between the use of insulin and sulfonylurea drugs, which are known to be highly associated with hypoglycemia (Fig. 1). For patients using insulin or sulfonylurea, the risk was major for 8 of 43 (18.6%) patients, moderate for 20 of 138 (14.5%) patients, low for 15 of 192 (7.8%) patients, and none for 4 of 336 (1.2%) (p for trend < 0.001). For patients not using insulin or sulfonylurea, the risk was major for 7 of 51 (13.7%) patients, moderate for 8 of 151 (5.3%) patients, low for 10 of 289 (3.5%), patients, and none for 9 of 554 (1.6%) patients (p for trend < 0.001).
Fig. 1.
Subgroup analysis of the incidence of hypoglycemia with and without the use of insulin or sulfonylurea. GNRI, geriatric nutritional risk index
We examined the cumulative incidence of hypoglycemia based on the GNRI classification during the study period (Fig. 2). In this Kaplan–Meier curve, the cumulative incidence of hypoglycemia increased as the GNRI risk classification increased.
Fig. 2.
Cumulative incidence curve of hypoglycemia according to the geriatric nutritional risk index (GNRI) class
Association of GNRI, serum albumin, and BMI with the incidence of hypoglycemia.
Albumin and BMI are used to calculate the GNRI formula. Therefore, we investigated the relationships of GNRI, albumin, and BMI with hypoglycemia. Figure 3 shows an analysis using the ROC curve. The AUC of the ROC curves for GNRI, albumin, BMI, and albumin + BMI were predictive of hypoglycemia, with AUCs of 0.738, 0.722, 0.578, and 0.735, respectively, and GNRI showing the higher value.
Fig. 3.
Receiver operating characteristic curves of the geriatric nutritional risk index (GNRI), albumin, body mass index (BMI), albumin + BMI, the incidence of hypoglycemia, and the area under the curve (AUC)
Moreover, Cox regression analysis was performed to adjust for the incidence of hypoglycemia and confounding factors (Table 3). The nutrition-related risk class by GNRI increased the HR of incident hypoglycemia to a major risk by 5.50 times (95% CI, 2.43–12.45, p < 0.001), moderate risk by 3.86 times (95% CI, 1.91–7.80, p < 0.001), and low risk by 2.55 times (95% CI, 1.29–5.06, p = 0.007) compared with no risk as a reference. In addition, we created an albumin and BMI model that showed a significant difference only in the major risk for albumin class, while the BMI model showed no significant difference in the BMI class.
Table 3.
Adjusted HR (95% CI) of the incidence of hypoglycemia of GNRI, albumin, and BMI
| HR, 95% CI, p value | ||||
|---|---|---|---|---|
| GNRI model | Albumin model | BMI model | ||
| Risk class | No risk | Ref (GNRI > 98) | Ref (Albumin ≥ 3.8) | Ref (BMI ≥ 24) |
| Low risk | 2.55, 1.29–5.06, p = 0.007 (GNRI 92– ≤ 98) | 1.45, 0.76–2.78, p = 0.26 (Albumin 3.5– < 3.8) | 1.26, 0.70–2.29, p = 0.45 (BMI 22– < 24) | |
| Moderate risk | 3.86, 1.91–7.80, p < 0.001 (GNRI 82 < 92) | 1.79, 0.94–3.42, p = 0.076 (Albumin 3.0– < 3.5) | 1.23, 0.69–2.18, p = 0.48 (BMI 19– < 22) | |
| Major risk | 5.50, 2.43–12.45, p < 0.001 (GNRI < 82) | 3.86, 1.98–7.51, p < 0.001 (Albumin < 3.0) | 1.84, 0.96–3.54, p = 0.068 (BMI < 19) | |
| Age | 0.98, 0.94–1.02, p = 0.38 | 0.98, 0.94–1.02, p = 0.40 | 0.99, 0.95–1.03, p = 0.57 | |
| Diabetes medications | Insulin | 2.10, 1.34–3.29, p = 0.001 | 2.16, 1.37–3.38, p < 0.001 | 2.41, 1.54–3.77, p < 0.001 |
| Comorbidities | Severe CKD | 1.49, 0.91–2.46, p = 0.12 | 1.56, 0.94–2.58, p = 0.086 | 1.97, 1.20–3.22, p = 0.007 |
| Dementia | 2.10, 1.12–3.93, p = 0.020 | 2.34, 1.24–4.40, p = 0.009 | 2.17, 1.15–4.09, p = 0.017 | |
| Laboratory data | High CRP | 0.97, 0.60–1.58, p = 0.91 | 1.04, 0.63–1.72, p = 0.88 | 1.50, 0.95–2.38, p = 0.081 |
CKD chronic kidney disease, GNRI geriatric nutritional risk index, CRP C-reactive protein, LF liver failure, HR hazard ratio, CI confidence interval
aCox regression analysis was used to adjust for confounding factors and express the incidence of hypoglycemia. Models for GNRI, albumin, and BMI were created and compared
Discussion
This study aimed to examine the association between nutrition-related risk as assessed by the GNRI at the time of admission and the incidence of hypoglycemia during hospitalization in patients aged ≥ 65 years with T2D using diabetes medication. The results showed that the incidence of hypoglycemia increased as the nutrition-related risk increased. The incidence of hypoglycemia differed significantly by approximately 15% (major risk), 10% (moderate risk), 5% (low risk), and 2% (no risk). A similar trend occurred in the subgroups of patients who did and did not use insulin or sulfonylurea. Multivariate analysis was performed to adjust for other confounding factors. As a result, the HR was found to be approximately 2.5 (low risk) to 5.5 (major risk) times higher, when compared with the no risk as a reference.
The cause of hypoglycemia among the malnourished is unclear; however, it may be due to a depletion of glycogen stores because hypoglycemia is the main symptom of impaired glycogen storage [22, 23]. In glycogen storage diseases, glycogen accumulates in tissues, such as the liver and muscles, due to an inherent abnormality in the enzymes required for glycogenolysis. Hence, glycogen does not break down into glucose and results in hypoglycemia. People with malnutrition may be unable to synthesize glucose due to these conditions.
We included age, insulin, severe CKD, dementia, and high CRP to adjust for confounders in the multivariate analysis. Although sulfonylureas have been a risk factor for hypoglycemia in other studies, they were not a risk factor for hypoglycemia in this study. The incidence of hypoglycemia for sulfonylurea users (4/193; 2.1%) compared to the overall incidence of hypoglycemia (81/1754; 4.6%) tended to be lower, although not significantly. This may have occurred because approximately 60% of the patients using sulfonylurea drugs were in the no-risk group according to GNRI classification, which may have reduced the hypoglycemia occurrence. However, this is unclear because the number of patients using sulfonylureas was small.
In the present study, the incidence of hypoglycemia was proportional to the severity of nutrition-related risk. This suggests an association between hypoglycemia and nutritional status. In a previous study on the association between hypoglycemia and malnutrition, Leibovitz et al. [11] used the NRS-2002 to assess malnutrition and reported an association with the incidence of hypoglycemia during hospitalization. Specifically, malnutrition approximately doubled the OR of the incidence of hypoglycemia during hospitalization. Hence, these results were consistent with our study in which malnutrition increased the incidence of hypoglycemia.
However, in their study, an association between malnutrition and the incidence of hypoglycemia was observed among non-diabetic patients alone. In contrast, this study assessed patients with diabetes, and an association was observed. The use of NRS-2002 in previous studies might be responsible for the difference in results. The NRS-2002 has been shown to overestimate malnutrition when used for older people [24]. Their study population included a large number of older individuals. In other words, the malnourished group, which was prone to hypoglycemia, may have included false-positive malnourished who were less prone to hypoglycemia. Furthermore, the treatment with diabetic medications might have contributed to hypoglycemia and did not differ in the incidence of hypoglycemia between the two groups. The GNRI tool was a nutrition-related risk index for older individuals. Hence, it could accurately identify malnutrition in older individuals with diabetes, while an association between malnutrition and hypoglycemia was likely to be observed. Consequently, the assessment of GNRI is accurate for older individuals with diabetes.
The discriminatory performance of GNRI, serum albumin, and BMI in the development of hypoglycemia is illustrated by ROC curves. The results showed that GNRI had a good discriminatory performance. Furthermore, when comparing their respective HRs by Cox regression analysis, the results were also significant for all risk classes in the GNRI model, and the major-risk class in the serum albumin model. Therefore, GNRI is a better predictor than serum albumin and BMI in the development of hypoglycemia. Serum albumin was also affected by non-nutritional factors, such as fluid status, congestion, renal dysfunction, and inflammation [8]. Moreover, BMI was affected by fluid status and the body's ability to maintain homeostasis [25]. Increased volume of fluid (such as extracellular fluid) decreased serum albumin and increased the BMI. Because the GNRI assessed both serum albumin and BMI, it had the potential to offset the shortcomings of each measure. In fact, the ROC curve combining serum albumin and BMI improved discriminatory performance in the incidence of hypoglycemia to the same extent as GNRI. Therefore, GNRI may lead to an accurate identification of nutrition-related risk. It is easier to predict the incidence of hypoglycemia with GNRI than with a combination of serum albumin and BMI because the values are single and are classified by risk. Therefore, the GNRI is easier to apply in clinical practice.
In the present study, we suggested an association between nutrition-related risk and the incidence of hypoglycemia among hospitalized older individuals who received diabetes medications. We found that GNRI can be used to identify individuals who are likely to experience hypoglycemia. The GNRI’s strength is that it can facilitate nutritional assessment. This kind of convenience is not available with other nutritional screening tools. With the GNRI, assessing all hospitalized older diabetics and, if nutrition-related risks were identified, proactively implementing hypoglycemia prevention, including drug-prescription interventions (prescription changes to hypoglycemic low-risk drugs, glucose prescriptions), education regarding early symptoms of hypoglycemia, and immediate action during a hypoglycemic episode, and hypoglycemic-risk sharing with other professions (careful observation and prompt care by the nurse) were made possible. This would be a useful approach for all medical staff across professions.
This study had several limitations. First, this study retrospectively evaluated the evidence in a hospital database without subject randomization. The possibility of a selection bias must be considered. We performed multivariate analysis to balance groups for covariates that we thought would strongly influence the results. However, we cannot rule out the possibility of hidden covariates as this is a retrospective medical history review. Second, this was a single-center study, of which the external validity of the results was limited. However, the study facility, Soka Municipal Hospital, served as the flagship hospital for a city of 240,000 people. Therefore, the study sample was considered reasonably representative. Third, the GNRI assessment of nutrition-related risk indicators was obtained only at baseline. This index fluctuated with changes in a patient's medical condition, which may have affected the association between the index and the outcomes. Fourth, among the confounders, the diagnoses (LF, dementia) registered in the electronic medical records had limited reliability because they could have been classified according to insurance. Furthermore, there was a possibility that the diagnoses were not registered. However, the reliability was considered improved when information regarding medications used was available. Furthermore, the diagnosis of T2D is considered reliable because the design of our study was based on patients using diabetes medication. If the diagnosis was not registered, then we checked the description in the medical records. Therefore, due to these limitations, our results need to be validated by larger, prospective, multicenter studies.
Conclusion
This study suggested that the incidence of hypoglycemia increases with increasing nutrition-related risk among hospitalized older patients with T2D using diabetes medication. After adjusting for known risk factors, GNRI was found to be an independent predictor and showed a dose-dependent outcome. Therefore, assessment of nutritional status using the GNRI on admission and proactive intervention for hypoglycemia based on its severity may reduce the risk of incident hypoglycemia due to nutritional status.
Funding
No funding to declare.
Compliance with ethical standard
Conflict of interest
The authors declare that they have no conflict of interest.
Research involving human participants
All procedures followed were in accordance with the Helsinki Declaration of 1964 and later versions. The study protocol was approved by the Ethics Committees of Soka Municipal Hospital (approval no: 2020–06, approval date: 23 June 2020) and Meiji Pharmaceutical University (approval no: 202010, approval date: 6 July 2020).
Informed consent
Due to the retrospective observational design, written informed consent was not obtained from the participants, although it was obtained via the opt-out method through the website of the institution.
Footnotes
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
References
- 1.Zoungas S, Patel A, Chalmers J, De Galan BE, Li Q, Billot L, et al. Severe hypoglycemia and risks of vascular events and death. N Engl J Med. 2010;363:1410–1418. doi: 10.1056/NEJMoa1003795. [DOI] [PubMed] [Google Scholar]
- 2.Bonds DE, Miller ME, Bergenstal RM, Buse JB, Byington RP, Cutler JA, et al. The association between symptomatic, severe hypoglycaemia and mortality in type 2 diabetes: retrospective epidemiological analysis of the ACCORD study. BMJ. 2010;340:b4909. doi: 10.1136/bmj.b4909. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Brodovicz KG, Mehta V, Zhang Q, Zhao C, Davies MJ, Chen J, et al. Association between hypoglycemia and inpatient mortality and length of hospital stay in hospitalized, insulin-treated patients. Curr Med Res Opin. 2013;29:101–107. doi: 10.1185/03007995.2012.754744. [DOI] [PubMed] [Google Scholar]
- 4.Zapatero A, Gómez-Huelgas R, González N, Canora J, Asenjo Á, Hinojosa J, et al. Frequency of hypoglycemia and its impact on length of stay, mortality, and short-term readmission in patients with diabetes hospitalized in internal medicine wards. Endocr Pract. 2014;20:870–875. doi: 10.4158/EP14006.OR. [DOI] [PubMed] [Google Scholar]
- 5.Nicolucci A, Prosperini G, Buzzetti R, Decosmo S, Fanelli CG, Napoli R, et al. A multistep approach for the stratification of the risk of severe hypoglycemia in patients with type 2 diabetes. Minerva Endocrinol. 2018;43:501–510. doi: 10.23736/S0391-1977.18.02850-X. [DOI] [PubMed] [Google Scholar]
- 6.Yaffe K, Falvey CM, Hamilton N, Harris TB, Simonsick EM, Strotmeyer ES, et al. Association between hypoglycemia and dementia in a biracial cohort of older adults with diabetes mellitus. JAMA Intern Med. 2013;173:1300–1306. doi: 10.1001/jamainternmed.2013.6176. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Yun JS, Park YM, Han K, Cha SA, Ahn YB, Ko SH. Association between BMI and risk of severe hypoglycaemia in type 2 diabetes. Diabetes Metab. 2019;45:19–25. doi: 10.1016/j.diabet.2018.03.006. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Leibovitz E, Wainstein J, Boaz M. Association of albumin and cholesterol levels with incidence of hypoglycaemia in people admitted to general internal medicine units. Diabet Med. 2018;35:1735–1741. doi: 10.1111/dme.13792. [DOI] [PubMed] [Google Scholar]
- 9.Norman K, Pichard C, Lochs H, Pirlich M. Prognostic impact of disease-related malnutrition. Clin Nutr. 2008;27:5–15. doi: 10.1016/j.clnu.2007.10.007. [DOI] [PubMed] [Google Scholar]
- 10.Kuzuya M, Izawa S, Enoki H, Okada K, Iguchi A. Is serum albumin a good marker for malnutrition in the physically impaired elderly? Clin Nutr. 2007;26:84–90. doi: 10.1016/j.clnu.2006.07.009. [DOI] [PubMed] [Google Scholar]
- 11.Leibovitz E, Adler H, Giryes S, Ditch M, Burg NF, Boaz M. Malnutrition risk is associated with hypoglycemia among general population admitted to internal medicine units. Results from the MENU study. Eur J Clin Nutr. 2018;72:888–893. doi: 10.1038/s41430-018-0143-9. [DOI] [PubMed] [Google Scholar]
- 12.Bouillanne O, Morineau G, Dupant C, Coulombel I, Vincent JP, Nicolis I, et al. Geriatric nutritional risk index: a new index for evaluating at-risk elderly medical patients. Am J Clin Nutr. 2005;82:777–783. doi: 10.1093/ajcn/82.4.777. [DOI] [PubMed] [Google Scholar]
- 13.Abd-El-Gawad WM, Abou-Hashem RM, El Maraghy MO, Amin GE. The validity of geriatric nutrition risk index: simple tool for prediction of nutritional-related complication of hospitalized elderly patients. Comparison with mini nutritional assessment. Clin Nutr. 2014;33:1108–1116. doi: 10.1016/j.clnu.2013.12.005. [DOI] [PubMed] [Google Scholar]
- 14.Kinugasa Y, Kato M, Sugihara S, Hirai M, Yamada K, Yanagihara K, et al. Geriatric nutritional risk index predicts functional dependency and mortality in patients with heart failure with preserved ejection fraction. Circ J. 2013;77:705–711. doi: 10.1253/circj.CJ-12-1091. [DOI] [PubMed] [Google Scholar]
- 15.Cereda E, Limonta D, Pusani C, Vanotti A. Assessing elderly at risk of malnutrition: the new geriatric nutritional risk index versus nutritional risk index. Nutrition. 2006;22:680–682. doi: 10.1016/j.nut.2006.02.003. [DOI] [PubMed] [Google Scholar]
- 16.Cereda E, Zagami A, Vanotti A, Piffer S, Pedrolli C. Geriatric nutritional risk index and overall-cause mortality prediction in institutionalized elderly: a 3-year survival analysis. Clin Nutr. 2008;27:717–723. doi: 10.1016/j.clnu.2008.07.005. [DOI] [PubMed] [Google Scholar]
- 17.Cereda E, Limonta D, Pusani C, Vanotti A. Geriatric nutritional risk index: a possible indicator of short-term mortality in acutely hospitalized older people. J Am Geriatr Soc. 2006;54:1011–1012. doi: 10.1111/j.1532-5415.2006.00754.x. [DOI] [PubMed] [Google Scholar]
- 18.Yamada S, Yamamoto S, Fukuma S, Nakano T, Tsuruya K, Inaba M. Geriatric nutritional risk index (GNRI) and creatinine index equally predict the risk of mortality in hemodialysis patients: J-DOPPS. Sci Rep. 2020;10:5756. doi: 10.1038/s41598-020-62720-6. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.NICE-SUGAR Study Investigators. Finfer S, Liu B, Chittock DR, Norton R, Myburgh JA, et al. Hypoglycemia and risk of death in critically ill patients. N Engl J Med. 2012;367:1108–1118. doi: 10.1056/nejmoa1204942. [DOI] [PubMed] [Google Scholar]
- 20.Eckart A, Struja T, Kutz A, Baumgartner A, Baumgartner T, Zurfluh S, et al. Relationship of nutritional status, inflammation, and serum albumin levels during acute illness: a prospective study. Am J Med. 2020;133(6):713–22.e7. doi: 10.1016/j.amjmed.2019.10.031. [DOI] [PubMed] [Google Scholar]
- 21.Kanda Y. Investigation of the freely available easy-to-use software “EZR” for medical statistics. Bone Marrow Transplant. 2013;48:452–458. doi: 10.1038/bmt.2012.244. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Kanungo S, Wells K, Tribett T, El-Gharbawy A. Glycogen metabolism and glycogen storage disorders. Ann Transl Med. 2018;6:474. doi: 10.21037/atm.2018.10.59. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Ellingwood SS, Cheng A. Biochemical and clinical aspects of glycogen storage diseases. J Endocrinol. 2018;238:R131–R141. doi: 10.1530/JOE-18-0120. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Poulia KA, Yannakoulia M, Karageorgou D, Gamaletsou M, Panagiotakos DB, Sipsas NV, et al. Evaluation of the efficacy of six nutritional screening tools to predict malnutrition in the elderly. Clin Nutr. 2012;31:378–385. doi: 10.1016/j.clnu.2011.11.017. [DOI] [PubMed] [Google Scholar]
- 25.Kenchaiah S, Pocock SJ, Wang D, Finn PV, Zornoff LAM, Skali H, et al. Body mass index and prognosis in patients with chronic heart failure: insights from the candesartan in heart failure: assessment of reduction in mortality and morbidity (CHARM) program. Circulation. 2007;116:627–636. doi: 10.1161/CIRCULATIONAHA.106.679779. [DOI] [PubMed] [Google Scholar]