Multiple Antihyperglycemic Drug Use is Associated with Undernutrition Among Older Adults with Type 2 Diabetes Mellitus: A Cross-Sectional Study

We recruited community-dwelling individuals who were ≥ 65 years old through the geriatrics outpatient clinic as part of a cross-sectional survey on the associations between cardiovascular disease and common geriatric comorbidities. The study sample included subjects with a history of physician-diagnosed T2DM. Individuals who had a history of T2DM but no active treatment were also included when their glycated hemoglobin (HbA1c) value was > 7%. The control group included individuals who met the following criteria; no history of DM or hypoglycemic medications, no symptom of hyperglycemia, and normal fasting plasma glucose (< 100 mg/dl) and HbA1c (< 6.5%) levels. Subjects with dementia or cognitive deficits, delirium, or overt neuropsychiatric behavior, recent major surgery, limited mobility, or significant mobility impairment were excluded. The Institutional Review Board of Gulhane School of Medicine, Ankara, Turkey, approved the study; written informed consent was obtained from each participant, and the study procedures conform to the ethical guidelines of the 1975 Declaration of Helsinki.

Baseline characteristics of patients are listed in Table 1. Presence of coronary heart disease (CHD) was based on self-reported history of coronary revascularization, angiographic evidence of significant CHD, or a documented clinical history of myocardial infarction or unstable angina pectoris. Stroke diagnosis was based on self-reported history of a thrombotic event, cerebral hemorrhage, or small stroke/transient ischemic attack documented by hospital admission records. Peripheral arterial disease (PAD) was diagnosed by the ankle-brachial index measurement [18]. Cardiovascular disease (CVD) referred to the composite of CHD, stroke, and PAD. Due to lower LDL cholesterol goals, patients with T2DM were using more statins, and their mean LDL cholesterol level was lower than those without DM. Although a lower cholesterol level may be associated with malnutrition among the elderly [19], a recent study reported that statins also influence eating behavior in the long term [20]. Thus, the main dyslipidemia variable in our analyses was statin use. The definition of multimorbidity was consistent with ≥ 3 coexisting chronic conditions severe enough to require medications or lifestyle management [5, 21]. All participants had their weight, height, and body mass index (BMI) measured during the physical examination. Blood tests obtained during the outpatient geriatric evaluation were recorded for this research.

We evaluated the nutritional status using the Mini Nutritional Assessment short form (MNA-SF) [22]. The sum of questions on the MNA-SF can provide a maximum score of 14; a score ≥ 12 indicates normal nutritional status, a score between 8 and 11 indicates “at risk for malnutrition,” and a 0–7 score is indicative of “malnourishment.” Instead of the term “malnutrition,” “undernutrition” is used synonymously in the present report, which covers subjects who are classified as either malnourished or at risk of malnutrition by the MNA-SF [1].

We determined functional performance through the Barthel Index [23], which rates basic activities of daily living (BADL), and the Lawton-Brody instrumental activities of daily living (IADL) scale [24]. To assess cognitive functions, we used the Mini-Mental State Examination test (scores 0–30; higher scores show better cognitive functions) [25]. We evaluated depressive symptoms by the Geriatric Depression Scale (GDS) [26] (scores 0–15; scores > 5 show depression). Dementia was excluded using previously established criteria [27, 28].

Over- and undertreatment of DM was defined according to the health status-based HbA1c goals as recommended by the American Geriatrics Society (7.0–7.5%, 7.5–8.0%, and 8–9%) [29]. We defined the patients over- or undertreated when their HbA1c value remained below or above the recommended lower or upper limit for the corresponding health status category [5, 29]. Multiple antihyperglycemic drug use referred to the prescription of ≥ 2 antihyperglycemics (oral agents or oral agents plus insulin), and an insulin-based treatment indicated the use of insulin alone or insulin plus oral antihyperglycemics.

Continuous variables were presented as mean (standard deviation). Categorical data were described as an absolute number and a percentage of the total or either group. Nonnormal distribution was tested by the Shapiro-Wilk test. Differences between patients with and without diabetes were analyzed using the independent t test, χ² test, or Mann-Whitney U test. Spearman rank correlation coefficients were calculated to test linear correlations, and partial correlations were calculated by controlling for covariates. Uni- and multivariate logistic regressions were studied to identify independent variables associated with undernutrition. Associations at p  ≤ 0.25 in univariate analysis were entered in the full model. The Box-Tidwell test was used to check the linearity of the logit for the continuous independent variables in logistic regression analysis. No significant violations of the linearity in the logit assumption were observed in the two final models. Multicollinearity was studied in each model. Statistical significance was accepted at p < 0.05. All analyses were performed using SPSS (PASW) 23.0 software (SPSS Inc., Chicago, IL).

The total sample included 546 participants (mean age: 74.9 ± 6.3 years, female: 71.1%). T2DM (n = 215) and control (n = 331) groups were matched for age and gender. As shown in Table 1, hypertension, statin use, CHD, PAD, CVD, anemia, and multimorbidity were more prevalent in patients with T2DM. These patients also had lower scores on MMSE, BADLs, and IADLs and higher scores on GDS compared with controls. Medication count and polypharmacy were also significantly higher in the T2DM group. There was a marginally significant difference in mean BMI between the patients and controls, but the proportion of individuals with BMI < 25 was similar.

Mean MNA-SF score was significantly lower in patients with T2DM (Table 1). The proportion of subjects having undernutrition according to the MNA-SF score was 31.1% in the total sample. Most of the participants (90%) having undernutrition fell under the “at-risk” category. As shown in Table 1, patients with T2DM had a significantly higher prevalence of undernutrition compared with the controls. However, this difference was evident only in the “at-risk” category. The proportion of malnourished individuals was overall small in the two groups. There was no age-adjusted correlation between the MNA score and HbA1c in patients with T2DM (r = − 0.060, p = 0.423).

T2DM patients having undernutrition (n = 79) and normal nutritional status (n = 136) were similar for most parameters. However, those having undernutrition were more often female and had more CHD, PAD, composite CVD, anemia, and multimorbidity (Table 2). Compared with those with normal nutritional status, the mean BMI was slightly higher in patients with undernutrition, but the prevalence of BMI < 25 was similar. Patients with undernutrition also had lower scores on MNA-SF, MMSE, BADL, and IADL tests and higher scores on GDS. Interestingly, their fasting plasma glucose was higher than in those with normal nutritional status. There were age-adjusted partial correlations between the MMSE and MNA-SF scores in both the total sample (r = 0.306, p < 0.001) and T2DM group (r = 0.332, p < 0.001).

There were numeric but nonsignificant differences between the patients with T2DM with and without undernutrition in the length of diabetes diagnosis, HbA1c, and the rates of over- and undertreatment. More patients with undernutrition were using multiple antihyperglycemic agents (≥ 2) compared with those with normal nutritional status (Table 2). There was no patient with normal nutritional status taking triple antihyperglycemics (≥ 3 oral agents or insulin), but 8.9% of the patients with undernutrition were on such treatment. Use of insulin-based treatments (single insulin or insulin plus oral antihyperglycemics) was similar in patients with and without undernutrition. The length of diabetes duration (mean approximately 10 years) correlated positively with the number of antihyperglycemic drugs (r = 0.338, p < 0.001).

Uni- and multivariate associations of undernutrition with the study variables were tested in the total sample (Table 3). Logistic regression analysis showed that a lower BMI, lower scores on the Barthel Index, lower scores on MMSE, and presence of depressive symptoms were independently associated with undernutrition in the total sample.

In patients with T2DM, univariate associations were significant between undernutrition and female gender, lower education, CVD, lower scores on MMSE and BADL tests, and depressive symptoms (Table 4). Multivariable logistic regression analysis showed that, after adjustment for covariates, lower scores on the Barthel Index, depressive symptoms, and the use of multiple antihyperglycemic drugs were independently associated with undernutrition. Unlike the total sample, BMI < 25 was not associated with undernutrition in patients with T2DM. Also, there were no independent associations between undernutrition and insulin use, diabetes overtreatment or diabetes undertreatment in patients with T2DM.