Influence of Age, Gender, Frailty, and Body Mass Index on Serum IL-17A Levels in Mature Type 2 Diabetic Patients

Abstract

Background

The cytokine IL-17A is emerging as a marker of chronic inflammation in cardio-metabolic conditions. This study aimed to identify relevant factors that in older primary care patients with type 2 diabetes (T2D) could influence serum IL-17A concentrations. The results have a potential to improve risk stratification and therapy options for these patients.

Material/Methods

The study was conducted during a period of 4 months, in 2020, in the south-eastern region of Croatia. Patients from primary health care, diagnosed with T2D (N=170, M: F 75: 95, ≥50 years old), were recruited at their visits. Those with malignant diseases, on chemotherapy or biological therapy, with amputated legs, or at hemodialysis, were excluded. The multinomial regression models were used to determine independent associations of the groups of variables, indicating sociodemographic and clinical characteristics of these patients, with increasing values (quartiles) of serum IL-17A.

Results

The regression models indicated the frailty index and sex bias are the key modifying factors in associations of other variables with IL-17A serum values.

Conclusions

Sex bias and the existence of different frailty phenotypes could be the essential determining factors of the serum IL-17A levels in community-dwelling patients with T2D age 50 years and older. The results support the concept of T2D as a complex disorder.

Background

Chronic inflammation is considered the main driving force in accelerated aging; that is, aging burdened with cardio-metabolic conditions, such as type 2 diabetes (T2D), atherosclerosis, chronic heart disease (CHD) and chronic kidney disease (CKD), and other age-related chronic conditions [1]. The main sources of inflammation in older individuals are senescent cells and a chronically activated innate immune system. Various stimuli, like molecules that are released from damaged tissues and disturbed gut microbiota, can trigger receptors of the innate immune system, leading to increased production of pro-inflammatory cytokines, such as tumor necrosis factor-α (TNF-α) and interleukin-1β (IL-1β), IL-18, and IL-6 [2,3]. Obesity exacerbates age-related inflammation. This is mainly the effect of pro-inflammatory cytokines and other pro-inflammatory mediators produced by dysfunctional adipocytes and monocyte/macrophages that abundantly infiltrate adipose tissue of obese individuals [1,4]. In addition, metabolic intermediates, produced in excess in obesity and obesity-related conditions, including metabolic syndrome (MS) (a cluster of metabolic disorders and hypertension, associated with abdominal-type obesity) and T2D, such as free fatty acids, advanced glycation end-products, and oxidized lipoproteins, have been recognized as strong pro-inflammatory signals [2]. The pro-inflammatory cytokines induce insulin resistance (impaired insulin-mediated glucose utilization in metabolically active tissues), which further exacerbates metabolic disorders and inflammation, leading to the accelerated atherosclerosis and target organ damage [1].

In addition to secretory active macrophages, many other immune cells have been found in obesity-related adipose tissue [4,5]. Thus, the number of pro-inflammatory CD8+ T and CD4+ Th1 lymphocytes is increased, while the number of anti-inflammatory CD4+ Th2 lymphocytes is decreased, compared to adipose tissue of non-obese individuals. Recent evidence suggests that obesity-related pathologies, which are usually characterized by a high level of insulin resistance, including MS, hypertension, T2D, and their related vascular complications and target organ disease, and which play a major role in maintaining inflammation, involve pro-inflammatory CD4+ Th17 lymphocytes [6–10]. The role of Th17 lymphocytes in sustaining chronic inflammation has already been recognized in autoimmune and other inflammation-mediated diseases, such as inflammatory bowel disease, osteoarthritis, and periodontitis [6,11].

In the inflamed tissue microenvironment, the focus is turning from the predomination of anti-inflammatory T regulatory (Treg) lymphocytes toward a predomination of the pro-inflammatory Th1/Th17 pathway, which is associated with increased production of cytokines of the IL-17 family, known as a driver in maintaining tissue inflammation and damage [12,13]. Besides changes in the cytokine profile, changes in metabolic conditions can also shift the balance between Treg and Th17 cell lines. Both of these cell lines have a high level of adaptability to conditions in the microenvironment, which allows functional adaptation of the immune system to variations in physiological situations [14,15]. Th17/Treg polarization is imperfect, which makes inflammation and tissue remodeling/fibrosis stay even, with the balance oscillating between the predomination of either of these processes. For example, besides increased production of anti-inflammatory cytokines, an expansion of Treg is also associated with increased production of Transforming Growth Factor Beta (TGF-β), which is a major fibrotic factor [10].

The cytokine IL-17A is the most thoroughly investigated member of the IL-17 cytokine family, and its role in development of cardiovascular disease (CVD) and target organ damage has been demonstrated in both experimental and clinical conditions [13,16–18]. Some of the proposed mechanisms include increased mobilization of inflammatory and immune cells (eg, monocytes, neutrophils, and T lymphocytes) from circulation to tissues, increased production of pro-inflammatory molecules such as cytokines, chemokines, and adhesion molecules, and induction of extra-cellular matrix degradation and tissue fibrosis [19,20]. As in autoimmune diseases, neutrophils play a significant role in tissue damage caused by inflammation in cardio-metabolic disorders [21]. The therapeutic potential of anti-IL-17A antibodies for curing cardio-metabolic disorders has been demonstrated in animal models [19].

However, before routine implementation of IL-17A for diagnostic and therapeutic purposes in patients with T2D and CVD, it is necessary to understand factors that influence its serum level variations in these patients. This is important because, unlike in autoimmune diseases, where polymorphism of genes involved in immune reaction is the principal determinant of the clinical phenotype, cardio-metabolic disorders are complex, which means that share a common pathophysiology background, and that a variety of sociodemographic, lifestyle, and environmental factors contribute to inter-individual variations in disease expression [22,23]. In addition, evidence from experimental studies indicates involvement of this cytokine in various signaling cascades in both homeostatic (regulatory) and pathological pathways, so its regulation is highly context-dependent [24]. Besides the number of Th17 lymphocytes, local expression of IL-17A also depends on the size of the network of innate immune and residential tissue cells that in damaged tissue acquire the ability to produce IL-17A [25]. Ultimately, serum IL-17A concentrations depend on the net-effect of the synergistic and antagonistic signals from different organs and tissues [24]. For example, previous clinical studies investigating the classical inflammatory markers C-reactive protein (CRP) and neutrophil-to-lymphocyte ratio (NLR) have shown that particular sociodemographic and clinical factors affect their serum concentrations [26–28].

It is especially important to understand how particular factors influence serum IL-17A levels in older patients with T2D. The fact that T2D is mostly an age-related disease contributes to the complexity of these patients, by increasing their potential for development of multiple comorbidities and geriatric conditions such as sarcopenia, malnutrition, and frailty [29]. These conditions can change the body shape and thus also affect metabolic, inflammatory, and hormonal parameters, while increasing the inter-individual heterogeneity [30]. For example, in contrast to the conventional belief that only T2D duration and the quality of hyperglycemia control determine the outcomes in these patients, results of large epidemiologic studies suggest the importance of other factors such as current patient age and age at T2D onset [31]. In addition, frailty is emerging as being a major modulator of outcomes in diabetic patients [32]. Frailty is a progressive disorder (classified as pre-frailty and frailty), considered to be a result of exhaustion of homeostatic mechanisms, with nonspecific symptoms and signs such as sarcopenia (muscle loss), slow walking, weakness, and low activity levels [33]. The prevalence of frailty increases with age and the level of comorbidity and is higher in women than in men [34,35]. T2D and its related comorbidites – MS, CVD, and CKD – are strongly associated with frailty [36–38]. The pathogenesis of frailty is not well understood, and chronic inflammation is considered one of the several mechanisms involved [39]. There are no data showing how frailty relates to serum IL-17A levels in older diabetic patients.

Therefore, the aim of this study was to explore the relevance of sociodemographic and clinical factors in determining serum levels of IL-17A in patients with T2D who are age 50 years and older. The results are expected to inform future clinical studies and to accelerate implementation of the cytokine IL-17A in prediction of outcomes and treatment schemes in these patients. Of particular importance for clinical practice would be understanding the relationships between frailty and serum IL-17A levels in the context of other factors, especially when considering recent evidence on the existence of at least 2 well-defined metabolic frailty phenotypes – one associated with obesity and the other associated with weight loss [40].

Material and Methods

Ethics Statement

The study was conducted in accordance with the Declaration of Helsinki and was approved by the Expert and Ethics Council of the Health Centre Slavonski Brod, Croatia (ID: 1433-1/020).

Study Design and Participants

The study was conducted during a 4-month period in 2020 at the Health Centre of Slavonski Brod, a town in the south-eastern part of Croatia, with a population of about 50 000. We included only patients diagnosed with T2D and on therapy with antiglycemic agents, age 50 years and older, and of both sexes. They were recruited from 4 general practice offices out of a total of 20 that exist in this health center, as their general practitioners (GPs) agreed to participate in the study. This selection process did not, however, hamper the representativeness of the sample, because GPs in Croatia have a gatekeeping role, and almost all adults are registered with a GP. These 4 selected GP offices covered a total number of about 7000 adults. In addition, older people living in the area have similar living conditions and are generally of lower socioeconomic status.

The fact that participating GPs were all specialists, with at least of 7 years of working experience, secured the high level of accuracy of the collected data. The fact that the GPs all worked in the same health center means that they used similar professional vocabulary and content meaning of encoded terminology, which is important for data consistency. For diagnosis and treatment criteria for T2D, GPs in Croatia follow the guidelines of the European Society of Cardiology (ESC) and the European Association of the Study of Diabetes (EASD) [41]. Patient management is supported by the ICT system and electronic health records (eHRs). Although the general health check-ups for people aged 45–65 years include assessment of major cardiovascular (CV) risk factors, performed as a “fee-for-service reimbursement program” that has been in place for many years, screening for T2D is not fully successful, and the rate of undiagnosed diabetic patients in Croatia reaches 40% [42].

We used the age threshold of 50 years as a criterion for patient selection because the evidence suggests that the incidence rate of T2D is highest in people age 45–64 years [43]. Only patients who were able to come independently or with assistance to their GPs were included in the study. Exclusion criteria were acute health conditions, malignant diseases, autoimmune system diseases, and the use of corticosteroids or being on biological treatments. We also excluded patients with visible cognitive dysfunctions, leg amputation, had a transplanted kidney, and those who were on continuous renal replacement therapy.

The power analysis was performed to determine the study size. With the significance level of 0.05 and power of 0.8, and the moderate effect size of f2=0.15, the estimated sample size was 180 participants (G*Power, 3.1.9.4.). Patients were recruited during their regular GP visits. GPs explained the purpose of the study and the study protocol to all eligible patients, and asked for signed informed consent. The study lasted until the required number of subjects was reached. Some patients did not go to the laboratory to provide blood for testing, and we made a slight adjustment to keep participant numbers within the capacity of 1 cytokine kit, so that the final number of participants included in data analysis was 170 (M: F, 75: 95).

Dataset

Participants were thoroughly described by using 62 variables (Table 1A, 1B). Variables were selected to indicate: patient age and sex, smoking habits, anthropometric measures, MS, the number of comorbidities, diagnoses of CVD, the main groups of other chronic diseases, some common geriatric conditions such as malnutrition and frailty, the total number of prescribed drugs, the prescription rates of the main groups of antidiabetic and antihypertensive drugs, the hypolipidemic drug statins, and non-steroidal anti-inflammatory drugs (NSAIDs), as these are the most widely prescribed anti-inflammatory drugs [44].

Table 1A.

Participant characteristics (numerical variables).

Patient characteristics	Median (IQR)	Mean (SD)
Age (years)	66.00 (12.00)
Waist circumference (cm)	103.00 (13.00)
Midarm circumference (cm)	30.00 (4.00)
Erythrocyte number (×1012/L)		4.79 (0.44)
Glucose (mmol/L)	7.95 (3.18)
HbA1C (%)	6.90 (1.70)
Haemoglobin (g/l)	143.50 (17.00)
Total cholesterol (mmol/L)	5.20 (1.70)
Low-density lipoprotein (mmol/L)		3.21 (1.04)
High-density lipoprotein (mmol/L)	1.31 (0.40)
Triglycerides (mmol/L)	1.73 (0.90)
Estimated glomerular filtration rate (ml/min/1.73 m2)	83.00 (34.75)
Thyroid-stimulating hormone (mIU/L)	2.52 (1.70)

Table 1B.

Participant characteristics (categorical variables).

Patient characteristics	Absolute No.
Age (years)	50–65 – 79 66–75 – 63 >75 – 28
Sex (M, W)	M – 75 F – 95
Diabetes duration (years)	<5 – 74 <6, 10> – 38 >10 – 58
Hypertension duration (years)	<5 – 74 < 6, 10> – 49 >10 – 63
BMI (kg/m2)	<25 – 58 25–30 – 67 >30 – 83
Nutritional status (MNA-test screening score)	Normal nutrition – 144 At risk for malnutrition – 26
Smoking habit	Never – 68 Current – 31 Ex – 71
Frailty index (0 – robust; 1 – prefrail; 2 – frail)	0 – 97 1 – 42 2 – 31
Metabolic syndrome (M) = yes Metabolic syndrome (F) = yes	74.7% 96.8%
No. of comorbidities	≤3 – 8 >3 – 162
Hypertension = yes	89.4%
Cardiovascular dis. (one or more of) = yes	51.2%
Coronary artery dis. = yes	34.7%
Chronic heart dis. = yes	48.8%
Diabetic retinopathy = yes	28.2%
Chronic respiratory dis. (Chronic Obstructive Pulmonary Dis. or Asthma) = yes	8.2%
Gastrointestinal dis. = yes	42.4%
Osteoporosis = yes	45.9%
Osteoarthritis = yes	50.6%
Low back pain = yes	65.3%
Thyroid gland dis. = yes	19.4%
Urogenital dis. = yes	29.4%
Incontinentio urinae = yes	15.3%
Anxious disorders = yes	60.6%
Total No. of medications prescribed	≤3 – 9 >3 – 161
Metformin = yes	72.9%
Sulfonylureas = yes	22.4%
Pioglitazone = yes	5.9%
Metformin, Sulfonylureas, Pioglitazone – alltogether (Old fashioned oral antidiabetics) = yes	84.7%
dipeptidyl peptidase-4 inhibitor = yes	17.6%
glucagon-like peptide-1 receptor agonists = yes	8.8%
sodium-glucose cotransporter 2 inhibitors = yes	3.5%
DPP4inh, GLP1ra, SGLT2inh – all together (New fashioned oral antidiabetics) = yes	30.0%
Insulin therapy = yes	23.5%
Angiotensin converting enzyme inhibitors or Angiotensin receptor blockers = yes	78.2%
Calcium channel blockers = yes	42.4%
Beta-blockers = yes	45.3%
Diuretics = yes	66.5%
Statins = yes	84.1%
Non-steroidal anti-inflammatory drugs = yes	71.2%

Regarding selection of prescribed drugs, special focus was put on newly recommended cardio- and renal-protective antidiabetic drugs, glucagon-like peptide-1 receptor agonist (GLP1ra), and sodium-glucose cotransporter-2 inhibitors (SGLT2inh), for their potential anti-inflammatory effects [41,45]. An old-fashioned first-line antidiabetic drug, metformin, and statins are widely prescribed to diabetic patients, and for both drugs, besides their metabolic effects, direct involvement in IL-17A signaling has been identified [46,47]. Antidiabetic drugs of special interest due to their demonstrated effect in regulating the Th17/Treg balance also include dipeptidyl peptidase-4 inhibitors (DPP4inhs) [48,49]. Among antihypertensive drugs, the effect on serum IL-17A concentrations has been proposed for angiotensin-converting enzyme inhibitors (ACE-INHs) and angiotensin receptor blockers (ARBs) due to their effects on ameliorating tissue inflammation, which is mediated by cooperation of tissue-type angiotensin II and IL-17A [13].

Anthropometric measures included body mass index (BMI) (kg/m²), which is a measure of general obesity, waist circumference (wc), which is a measure of visceral obesity, and a marker of MS, and mid-arm circumference (mac), which is a measure of sarcopenia (muscle loss) [50]. Clinically relevant sarcopenia was defined as mac ≤22 cm [51].

To assess the nutritional status, we used the Mini Nutritional Assessment – Short Form (MNA-SF) test, a version with 18 questions [52]. This test examines patients’ eating habits, types of food consumed, number of long-term medications, and self-reported health conditions. The maximum number of points that can be scored is 30. A score ranging from 24 to 30 indicates good nutrition, 17 to 23 indicates a risk of malnutrition, and less than 17 points indicate malnutrition. The frailty status was assessed by using the Fried’s phenotype model [53], which considers 5 criteria – weight loss, feeling of exhaustion, low-level activity, slow walking, and hand grip strength measured using a hand dynamometer. Three to 5 positive criteria indicate frailty, 1 to 2 positive criteria indicate pre-frailty, and all negative criteria mean robustness. We also assessed participants on MS, following evidence indicating that diabetic patients with MS, compared to those without, have a higher predisposition for complications [54]. For this purpose, we used the modified definition of the National Cholesterol Education Program, Adult Treatment Panel III (NCP ATP III) [55]. The NCP ATP III definition of MS is at least 2 of the following: wc ≥102 cm (88 cm for F), triglycerides (TG) ≥1.7, HDL (high-density lipoprotein) cholesterol <1.0 (1.2 for F), and diagnosis of hypertension.

Some standard blood laboratory tests were performed to better characterize patient metabolic status, nutrition, and inflammation, and to determine the level of renal function decline. These tests included: erythrocyte number, leukocyte number, lymphocyte number and percentage, neutrophil number and percentage, hemoglobin (Hb), CRP, fasting glucose, glycosylated hemoglobin (HbA1c), total cholesterol, low-density lipoprotein (LDL) cholesterol, high-density lipoprotein (HDL) cholesterol, triglycerides, and serum creatinine, needed for estimation of glomerular filtration rate (eGFR), a measure of renal function decline [56]. HbA1c, a measure of average blood glucose in the last 2 or 3 months, indicates the quality of hyperglycemia control [57]. According to the EAS/EASD guidelines, the target HbA1c in diabetic patients should be, in general, <7%, but for older patients with comorbidities, less-stringent HbA1c goals, <8% or even ≤9%, may be adequate (26) [41]. The US National Kidney Foundation Guidelines recognize 4 stages of renal function decline (not counting for the terminal stage), where stage 2 (eGFR <90 >60 ml/min) indicates mild renal impairment, and stages 3 and 4 (eGFR of ≤60 ml/min) indicate moderately and severely decreased renal function [56]. In the laboratory panel test, we also included Thyroid-Stimulating Hormone (TSH), to detect latent hypothyroidism, a common disorder in older diabetic patients, and associated with metabolic disorders and increased risk of diabetic complications [58]. Latent hypothyroidism is indicated if TSH is ≥4 mU/L [59].

We used classical markers of inflammation, CRP, Hb, and neutrophil-lymphocyte ratio (NLR), to compare their levels of variation with that of IL-17A and their associations with IL-17A (Table 2) [60,61]. For variable Hb, evidence indicates that it can be considered as a marker of chronic inflammation associated with the presence of chronic health conditions, and as a predictor of frailty [62,63].

Table 2.

Markers of inflammation.

Markers of inflammation	Median (IQR)	Mean (SD)
Total No. of Leukocytes (×109/L)		7.58 (1.77)
No. of Lymphocytes (×103/mL)	2.45 (1.17)
No. of Neutrophils (×103/mL)	3.99 (1.43)
Lymphocytes%		34.26 (8.65)
Neutrophils%		53.06 (8.37)
Neutrophil-to-lymphocyte ratio	1.60 (0.90)
C-reactive protein (mg/L)	1.90 (2.20)
Haemoglobin (g/L)	143.00 (18.00)
Interleukin 17A (pg/mL)	1.71 (1.50)

The Data Collection Procedure

Data on age, sex, comorbidities, and treatment options were obtained from eHRs. In Croatia, the International Statistical Classification of Diseases and Related Health Problems, 10^th revision (ICD-10) is used for disease classification. Having ≥3 diagnoses of chronic diseases was considered as multimorbidity [35]. As CVD, we considered diagnoses of coronary artery disease (CAD), CHD, cerebrovascular disease, and periphery artery disease. Diabetic retinopathy was used as a microvascular complication, but it was not considered as CVD [57]. These diagnoses were recorded in eHRs if confirmed by the specialist examination. Information on grades of CHD is missing, because the heart ultrasound examination, necessary for grading, has not been systematicaly performed and recorded in eHRs [64]. From eHRs, we also used information on T2D and hypertension duration, as 2 frequently overlapping disorders, and checked it by talking with patients [65]. From T2D duration and patient age, we were able to conclude about the age at diagnosis of T2D, and this all information is prognostically important [31].

If not older than 6 months, data for anthropometric measurements was used from eHRs; otherwise, these measures were taken from participants during their visits. In addition, participants were asked about smoking habits and were evaluated for the presence of frailty, malnutrition, and MS. A detailed description of the screening procedure on frailty is provided in a previous publication [66].

Laboratory Testing

Participants were referred to the county hospital’s central laboratory for laboratory testing. For this purpose, 10 ml of venous blood (2 Vacutainer tubes) was obtained from participants by cubital venipuncture. The clot was removed by centrifugation, and serum was stored at −20°C until analyzed. A part of serum was separated and stored at −70°C for the purpose of measuring IL-17A. These specimens were transported in the transporter refrigerator to the Laboratory for Clinical Immunology and Allergology Diagnostics of the University Hospital Centre of Osijek, the capital of eastern Croatia.

Erythrocyte and leukocyte numbers and hemoglobin were assesssed by a routine analysis using a Sysmex XN 1000 (SYSMEX UK LTD, Wymbush, Milton Keynes, GB) hematology analyzer. For counting leukocyte subpopulations, including the number of lymphocytes and neutrophils, the Coulter impedance method was used. The NLR was estimated from the complete blood count with differential [61].

Biochemical analyses were performed in the automated process analyzer DxC 700AU (Beckman Coulter, Fullerton, SAD). Lipid parameters, including total cholesterol, LDL and HDL cholesterol, and triglycerides, were determined by enzymatic staining. The enzymatic method was also used to determine serum creatinine. For assessing fasting blood glucose, the enzymatic UV test (with hexokinase) was used, while HbA1c was analyzed by the turbidimetric method. CRP was analyzed by the immunoturbidimetric method. To determine TSH, the chemiluminescent microparticle immunoassay was performed in an Alinity, Architect analyzer (Abbot Laboratories, Abbot Park, IL 60064, SAD). From information on serum creatinine, sex, and age, we calculated eGFR according to the Modification of Diet in Renal Disease (MDRD) formula, using the online calculation system of the US National Kidney Fundation [67].

Determining Serum IL-17A

To analyze IL-17A, we used a high-sensitive IL-17A human ELISA kit (Invitrogen, ThermoFisher Scientific, SAD). Enzyme-linked immunosorbent assay (ELISA) was performed according to the manufacturer’s protocol. The company has stated the test’s sensitivity to be 4 pg/mL. Drawing from previous experiences in measuring IL-17A in human serum, we decided to add an additional calibration point, performed only with the puffer system and reagents, to allow determination of IL-17A at very low concentrations, starting from 0 pg/mL.

Statistical Analysis

Numerical data are presented as the mean±standard deviation (SD) or as the median and interquartile range (IQR), depending on the type of distribution (standard or not). Categorical data are presented as absolute and relative frequencies. To examine distributions of different variables according to the increasing values of IL-17A, we divided a range value of IL-17A into quartiles.

The collinearity and multicollinearity were investigated for numerical attributes using correlation analysis (Spearman’s correlation coefficient) and variance inflation factor (VIF). A value of VIF between 1 and 5 indicates a moderate correlation between the given predictor variable and other predictor variables in the model, but this is often not severe enough to require attention. A value greater than 5 indicates a potentially strong correlation between the given predictor variable and other predictor variables in the model. In this case, the coefficient estimates and P values in the regression outputs are likely unreliable. Next, we applied the Shapiro-Wilk’s test of normality with the null hypothesis of normal sample distribution. If this test is significant (P value less than 0.05), the distribution is non-normal. Based on this result, we considered parametric ANOVA test for variables with normal distribution and the Kruskal-Wallis rank-sum test as a non-parametric alternative. In the case of ANOVA (null hypothesis: no difference in means), we investigated the following assumptions: the records are collected independently and randomly from the population defined by the factor levels; the data are normally distributed; and these normal populations have a common variance (Bartlett’s test for testing homogeneity of variances). If the ANOVA is significant (P value less than 0.05), we applied the post hoc Tukey’s test to find which quartiles’ means are different from each other (P value less than 0.05). We used the same approach with the post hoc Dunn’s test with a Bonferroni correction (P value less than 0.05) with the Kruskal-Wallis rank-sum test results. Accessing the differences in categorical variables among quartiles of IL-17A, we performed Pearson’s chi-square test or Fisher’s exact test (if >20% of expected cell counts are less than 5). To assess how the examined variables were associated with quartiles of IL-17A, we used the multinomial logistic regression (MLR) model from R statistics. The Akaike Information Criterion (AIC) measured the model’s predictive performance quality.

Results

Participants were patients in primary health care, diagnosed with T2D, age 50 years and older, who were able to walk independently, and without severe conditions. They were selected by a consecutive sampling method. The sample was a good representation of older, community-dwelling, diabetic patients from the study area, since residents had good access to GPs, and GPs employ similar methods in screening and managing diabetic patients.

Most participants were age 50–75 years (median 66), with slightly more women than men. Almost equal proportions of participants had T2D of a short duration (0–5 years) and longer duration (>5 years). Most participants had hypertension (89.4%), mostly for a longer period (>5 years). Most participants were overweight/obese and most had abdominal (visceral) type of obesity. There was a high proportion of those with MS, especially among women. Many patients had multimorbidity. They were mostly in a good nutritional state, without signs of marked sarcopenia (none had mac ≤22 cm), and none were malnourished. About 43% of participants were pre-frail or frail, while fully frail individuals counted for about 20% of participants in the sample, with more women than men (Table 1A, 1B).

In most patients, HbA1c was <8.5% (median 6.9, ICR 1.7), indicating well-controlled hyperglycemia. Good metabolic control was confirmed also by LDL cholesterol, maintained in most patients within the recommended values of less than 5.0 mmol/L (mean 3.21, SD 1.04). Concerning CV complications, about one-third of patients were diagnosed with CAD, and almost a half of patients were diagnosed with CHD, although information on degree severity of CHD was missing. About two-thirds of participants had decreased renal function, but in most cases it was mild (eGFR 90–60 mL/min) or moderate (eGFR <60–45 mL/min). Of non-CV comorbidities, the most prevalent were musculoskeletal diseases and anxiety disorders (Table 1A, 1B).

Many patients had been prescribed ACE-INH/ARB antihypertensive drugs, the traditional antidiabetic drug metformin, and statins. Few participants were prescribed newer antidiabetic drugs such as GLP1ra and SGLT2inh. NSAID was prescribed to about two-thirds of participants (Table 1B).

Participants in the sample showed a narrow range of cytokine IL-17A (median 1.71, ICR 1.50 pg/mL), corresponding to its low variability. This was also a characteristic of the classical markers of inflammation, NLR and CRP (Table 2). The quartiles were calculated by the internal R function according to the standard rule: 1^st quartile (25^th percentile of the data under the produced value), 2^nd quartile (50^th percentile of the data under the produced value), and 3^rd quartile (75^th percentile of the data under the produced value). As shown in Figure 1, interquartile cut-offs of IL-17A were 1.42 pg/ml (between 1^st and 2^nd quartiles), 1.71 pg/ml (between 2^nd and 3^rd quartiles), and 2.92 pg/ml (between 3^rd and 4^th quartiles), indicating that the 2 middle quartiles (2^nd and 3^rd) were very close to each other, and that IL-17A was slightly higher in the upper quartile.

Figure 1.

The quartiles of IL-17A. R, 4.3.1, R Development Core Team.

Correlation analyses (Table 3) showed that several variables were strongly correlated with each other. It is necessary to consider this information during the regression models specification (1 of the paired variables should not be used in prediction models). By VIF, we detected strong multicollinearity (>5) within the following variables: LDL (8.55), total cholesterol (9.36), neutrophils% (11.32), neutrophils (71.6), lymphocytes (63.84), lymphocytes% (10.42), and NLR (8.08). These variables were excluded from the respective models.

Table 3.

The result of correlation analysis.

Variable	Variable	Correlation coefficient
Age	Hypertension duration	0.61
wc	mac	0.73
Erythrocyte	Hemoglobin	0.813
Glucose	HbA1C	0.68
Total cholesterol	LDL	0.91
Neutrophils (%)	Lymphocytes (%)	−0.87
Neutrophils (%)	NLR	0.86
Lymphocytes (%)	NLR	0.86
Neutrophils	Lymphocytes	0.90

Variables showing significant differences among quartiles of IL-17A were those indicating metabolic disorders (LDL cholesterol and HbA1c), chronic inflammation associated with the burden of comorbidities (hemoglobin), CVD (CAD and CHD), and some non-CV chronic conditions (eg, gastrointestinal disorders, low back pain, urogenital disorders, incontinent urinary, and anxiety disorders). In addition, some drugs, such as newer oral antidiabetics (including DPP4inh, GLP-1ra, and SGLT-2inh) and beta-blockers (Tables 4, 5) differed among quartiles of IL-17A.

Table 4.

Differences in distributions of examined variables among quartiles of IL-17A. Numerical variables.

Variable	1st quartile 39 records	2nd quartile 46 records	3rd quartile 40 records	4th quartile 45 records	p-value	Post hoc test (Quartile combinations and adjusted p-value)
No. of lymphocytes (St p-value <0.05, not a normal distribution) – Median (IQR)	2.43 (1.21)	2.51 (1.12)	2.40 (1.08)	2.58 (1.36)	0.80*
No. of neutrophils (St p-value <0.05, not a normal distribution) – Median (IQR)	4.10 (1.34)	3.97 (1.30)	3.97 (1.28)	3.95 (1.45)	0.80*
Lymphocytes% (St p-value >0.05, a normal distribution) – Mean (SD)	34.99 (8.19)	33.20 (9.49)	34.27 (7.52)	34.69 (9.23)	0.78**
Neutrophils% (St p-value >0.05, a normal distribution) – Mean (SD)	52.99 (7.60)	53.49 (8.04)	52.77 (8.17)	52.92 (9.69)	0.98**
Neutrophil-to-lymphocyte ratio (St p-value <0.05, not a normal distribution) – Median (IQR)	1.60 (0.90)	1.69 (1.10)	1.60 (0.62)	1.60 (1.05)	0.93*
C-reactive protein (mg/L) (St p-value <0.05, not a normal distribution) – Median (IQR)	2.00 (1.70)	1.85 (1.80)	1.80 (3.43)	1.90 (3.40)	0.30*
Estimated glomerular filtration rate (ml/min/1.73 m2) (St p-value <0.05, not a normal distribution) – Median (IQR)	80.00 (29.00)	86.00 (35.00)	80.50 (33.00)	82.00 (44.00)	0.54*
Age (years) (St p-value <0.05, not a normal distribution) – Median (IQR)	67.00 (10.00)	63.00 (10.75)	70.00 (9.00)	65.00 (12.00)	0.08* (0.02 – small effect)
Waist circumference (cm) (St p-value <0.05, not a normal distribution) – Median (IQR)	103.0 (11.50)	104.00 (14.00)	102.50 (11.25)	103.00 (11.00)	0.88*
Mac (cm) (St p-value <0.05, not a normal distribution) – Median (IQR)	30.00 (4.00)	30.00 (3.00)	29.00 (2.25)	31.00 (3.00)	0.22*
Diabetes duration (years) (St p-value <0.05, not a normal distribution) – Median (IQR)	9.00 (10.00)	9.00 (9.00)	6.50 (10.50)	8.00 (11.00)	0.96*
Hypertension duration (years) (St p-value <0.05, not a normal distribution) – Median (IQR)	8.00 (8.00)	8.00 (9.50)	8.50 (7.00)	8.00 (7.00)	0.66*
Total Leukocyte (×109/L) (St p-value >0.05, a normal distribution) – Mean (SD)	7.65 (1.63)	7.24 (1.47)	7.41 (1.74)	8.00 (2.14)	0.20**
Erythrocyte (×1012/L) (St p-value >0.05, a normal distribution) – Mean (SD)	4.80 (0.47)	4.78 (0.47)	4.89 (0.37)	4.69 (0.44)	0.20**
Low-density lipoprotein (mmol/L) (St p-value >0.05, a normal distribution) – Mean (SD)	3.51 (0.82)	3.05 (1.10)	2.91 (1.08)	3.39 (1.06)	0.03** (0.05 – medium effect)	Q1–Q3 (0.05)***
High-density lipoprotein (mmol/L) (St p-value <0.05, not a normal distribution) – Median (IQR)	1.29 (0.30)	1.33 (0.40)	1.32 (0.42)	1.37 (0.42)	0.45*
Haemoglobin (g/L) (St p-value <0.05, not a normal distribution) – Median (IQR)	146.00 (18.00)	143.50 (19.00)	147.00 (14.75)	137.00 (16.00)	0.02* (0.04 – small effect)	Q3–Q4 (0.02)****
Glucose (mmol/L) (St p-value <0.05, not a normal distribution) – Median (IQR)	7.00 (2.60)	8.45 (2.58)	8.95 (3.15)	7.70 (3.10)	0.06* (0.03 – small effect)
HbA1C (%) (St p-value <0.05, not a normal distribution) – Median (IQR)	6.60 (1.20)	7.00 (1.73)	7.60 (1.90)	6.80 (1.20)	0.02* (0.04 – small effect)	Q1–Q3 (0.05)****
Triglycerides (mmol/L) (St p-value <0.05, not a normal distribution) – Median (IQR)	1.80 (0.80)	1.75 (1.16)	1.60 (0.79)	1.77 (0.92)	0.46*
Total cholesterol (mmol/L) (St p-value <0.05, not a normal distribution) – Median (IQR)	5.40 (1.60)	5.20 (1.83)	4.75 (1.63)	5.20 (1.20)	0.35*
Thyroid-stimulating hormone (mIU/L) (St p-value <0.05, not a normal distribution) – Median (IQR)	2.40 (1.49)	2.42 (1.84)	3.10 (1.60)	2.39 (1.80)	0.30*

St – Shapiro test normality;

^*Kruskal-Wallis rank sum test;

^**ANOVA test (Bartlett’s test for testing homogeneity of variances);

^***ANOVA post hoc test (Tukey’s test);

^*****KW post hoc test (Dunn’s test with a Bonferroni correction).

Table 5.

Differences in examined variables among quartiles of IL-17A. Categorical variables.

Variable	1st quartile 39 records	2nd quartile 46 records	3rd quartile 40 records	4th quartile 45 records	Total 170 records	p-value
Estimated glomerular filtration rate (ml/min/1.73 m2)
<60	15 (38.5%)	11 (23.9%)	14 (35.0%)	13 (28.8%)	53 (31.2%)	0.75
60–90	15 (38.5%)	19 (41.3%)	13 (32.50%)	16 (35.6%)	63 (37.0%)
>90	9 (23.0%)	16 (34.8%)	13 (32.50%)	16 (35.6%)	54 (31.8)
Frailty index
0	21 (53.9%)	28 (60.9%)	19 (47.50%)	29 (64.4%)	97 (57,1%)	0.42*
1	12 (30.8%)	11 (23.9%)	13 (32.5%)	6 (13.3%)	42 (24.7%)
2	6 (15.3%)	7 (15.2%)	8 (20.0%)	10 (22.3%)	31 (18.2%)
Sex = Men	20 (51.3%)	19 (41.3%)	17 (42.5%)	19 (42.2%)	75 (44.1%)	0.78
Age (years)
50–65	16 (41.0%)	26 (56.5%)	13 (32.5%)	24 (53.3%)	79 (46.5%)	0.22
66–75	16 (41.0%)	12 (26.1%)	21 (52.5%)	14 (31.1%)	63 (37.1%)
>75	7 (41.0%)	8 (17.4%)	6 (15.0%)	7 (15.6%)	28 (16.5%)
Smoking
Never	17 (43.6%)	19 (41.3%)	16 (40.0%)	16 (35.6%)	68 (40%)	0.66
Current + Ex	22 (56.4%)	27 (58.7%)	24 (60.0%)	29 (64.4%)	102 (60%)
BMI (kg/m2)
<25	3 (7.7%)	8 (17.4%)	5 (12.5%)	4 (8.9%)	20 (11.8%)	0.70*
25–30	17 (43.6%)	19 (41.3%)	16 (40.0%)	15 (33.3%)	67 (39.4%)
>30	19 (48.7%)	19 (41.3%)	19 (47.5%)	26 (57.8%)	73 (42.9%)
Nutritional screening score normal nutritional state	34 (87.2%)	36 (78.3%)	35 (87.5%)	39 (86.7%)	144 (84.7%)	0.56
Non-steroidal anti-inflammatory drugs = Yes	30 (76.9%)	34 (73.9%)	24 (60.0%)	33 (73.3%)	121 (71.2%)	0.34
Chronic obstructive pulmonary dis. or asthma = Yes	6 (15.4%)	1 (2.2%)	4 (10.0%)	3 (6.7%)	14 (8.2%)	0.15*
Cardiovascular dis. and/or cerebrovascular dis. = Yes	2 (5.1%)	9 (19.6%)	9 (22.5%)	12 (26.7%)	32 (18.8%)	0.65
Coronary artery dis. = Yes	16 (41.0%)	10 (21.7%)	9 (22.5%)	24 (53.3%)	59 (34.7%)	<0.01 Q2–Q4 Q3–A4
Chronic heart dis. = Yes	21 (53.8%)	15 (32.6%)	15 (37.5%)	30 (66.7%)	81 (47.6%)	<0.01 Q2–Q4
Gastro-intestinal dis. = Yes	21 (53.8%)	10 (21.7%)	14 (35.0%)	28 (62.2%)	73 (42.9%)	<0.01 Q1–Q2 Q2–Q4
Osteoporosis = Yes	7 (17.9%)	8 (17.4%)	7 (17.5%)	16 (35.6%)	38 (22.4%)	0.10
Osteoarthritis = Yes	25 (64.1%)	22 (47.8%)	16 (40.0%)	23 (51.1%)	86 (50.6%)	0.19
Low back pain = Yes	30 (76.9%)	33 (71.7%)	18 (45.0%)	30 (66.7%)	111 (65.3%)	0.01 Q1–Q3
Thyroid gland dis. = Yes	4 (10.3%)	10 (21.7%)	10 (25.0%)	9 (20.0%)	33 (19.4%)	0.38
Incontinentio urinae = Yes	7 (17.9%)	4 (8.7%)	3 (7.5%)	12 (26.7%)	26 (15.3%)	0.04
Urogenital dis = Yes	17 (43.6%)	7 (15.2%)	8 (45.0%)	18 (40.0%)	50 (29.4%)	<0.01 Q1–Q2
Anxiety = Yes	26 (66.7%)	22 (47.8%)	19 (47.5%)	35 (77.8%)	102 (60.0%)	<0.01 Q2–Q4 Q3–Q4
Diabetic retinopathy = Yes	14 (35.9%)	14 (30.4%)	8 (20.0%)	12 (26.7%)	48 (28.2%)	0.45
Metformin = Yes	28 (71.8%)	33 (71.7%)	28 (70.0%)	35 (77.8%)	124 (72.9%)	0.86
Sulfonylureas = Yes	8 (20.50%)	7 (15.2%)	11 (27.5%)	12 (26.7%)	38 (22.4%)	0.47
Pioglitazone = Yes	4 (10.3%)	4 (8.7%)	1 (2.5%)	1 (2.2%)	10 (5.9%)	0.32*
Old fashioned oral antidiabetics = Yes	33 (84.6%)	36 (78.3%)	35 (87.5%)	40 (88.9%)	144 (84.7%)	0.51
Dipeptidyl peptidase-4 inhibitor = Yes	7 (17.9%)	12 (26.1%)	8 (20.0%)	3 (6.7%)	30 (17.6%)	0.10
Sodium-glucose cotransporter 2 inhibitors = Yes	3 (7.7%)	2 (4.3%)	1 (2.5%)	0 (0.0%)	6 (3.5%)	0.24*
Glucagon-like peptide-1 receptor agonists = Yes	2 (5.1%)	6 (13.0%)	3 (7.5%)	4 (8.9%)	15 (8.8%)	0.68*
New fashioned oral antidiabetics = Yes	12 (30.8%)	20 (43.5%)	12 (30.0%)	7 (15.6%)	51 (30.0%)	0.04 Q2–Q4
Insulin = Yes	7 (17.9%)	15 (32.6%)	8 (20.0%)	10 (22.2%)	40 (23.5%)	0.38
Angiotensin converting enzyme inhibitors or Angiotensin receptor blockers = Yes	31 (79.5%)	32 (69.6%)	32 (80.0%)	38 (84.4%)	133 (72.2%)	0.37
Calcium channel blockers = Yes	16 (41.0%)	17 (37.0%)	14 (35.0%)	25 (55.6%)	82 (48.2%)	0.20
Diuretics = Yes	25 (64.1%)	25 (54.3%)	31 (77.5%)	31 (68.9%)	112 (65.9%)	0.15
Beta-blockers = Yes	18 (46.2%)	15 (32.6%)	15 (37.5%)	28 (62.2%)	76 (44.7%)	0.03 Q2–Q4
Hyperlipidemia (statins therapy) = Yes	33 (84.6%)	37 (80.4%)	31 (77.5%)	42 (93.3%)	133 (78.2%)	0.20
Number of comorbidities = ≤3	35 (89.7%)	44 (95.7%)	39 (97.5%)	44 (97.8%)	162 (95.3%)	0.40*
Number of medications prescribed = ≤3	35 (89.7%)	43 (93.5%)	39 (97.5%)	44 (97.8%)	161 (94.7%)	0.38*
Metabolic syndrome (Male)	15 (38.5%)	13 (28.3%)	15 (37.5%)	13 (28.9%)	56 (32.9%)	0.48*
Metabolic syndrome (Female)	19 (48.7%)	26 (56.5%)	21 (52.5%)	26 (57.8%)	92 (54.1%)	0.38*

Pearson’s chi-squared test;

^*Fisher test;

Bolded – significant differences, pairwise proportions tests.

The values of these variables did not increase linearly, in parallel to the increasing quartile ranks of IL-17A. For example, participants in the lower and upper quartiles of IL-17A were more often diagnosed with CAD and CHD than those in the middle quartiles. There were more patients prescribed newer oral antidiabetics in lower quartiles of IL-17A than in the upper quartiles. Unexpectedly, some variables, including markers of inflammation, sex, age, BMI, the frailty index, and eGFR, as well as some of the drugs, did not show significant differences among quartiles of IL-17A. This could be also a result of collinearity or multicollinearity (Table 3), which in the context of patient complexity (great inter-individual variations and the existence of multiple patient subgroups) can differently, non-linearly, and unexpectedly affect serum IL-17A levels in particular patient subgroups. For this reason, we prepared several regression models to identify variables independently associated with increasing values (quartiles) of serum IL-17A levels.

As indicated by the regression models (Table 6), variables indicating comorbidity- and frailty-related inflammation (hemoglobin), glucose-related metabolic disorders (HbA1c), and some frailty-associated chronic health conditions such as osteoporosis, osteoarthritis, and urogenital diseases were independently associated with quartiles of IL-17A. Among geriatric syndromes, indicated by the model “Comorbidity level and functional disorders,” only frailty status was selected. Prescribed drugs selected by the regression models were antidiabetic drugs DPP4inh and pioglitazone. In models indicating comorbidities, the frailty index was the modifying factor.

Table 6.

Multinomial logistic regression models for quartiles of IL-17A. All models were adjusted for age, sex, T2D duration, and frailty index.

Model “Markers of inflammation”: variables in the input: Leukocyte number, Haemoglobin, and CRP; AIC: 488.37.
	Quartile 2		Quartile 3		Quartile 4
	z-value	OR (95% CI)	z-value	OR (95% CI)	z-value	OR (95% CI)
Haemoglobin (g/l)			2.06 (p-value 0.04)	1.05 (1.01–1.09)

Model “Laboratory tests”: variables in the input: eGFR, HbA1C, Triglycerides, HDL-cholesterol, TSH; AIC: 500.67.
	Quartile 2		Quartile 3		Quartile 4
	z-value	OR (95% CI)	z-value	OR (95% CI)	z-value	OR (95% CI)
HbA1C (%)			1.93 (p-value 0.05)	1.45 (1.00–2.00)

Model “Other comorbidities”: variables in the input: osteoporosis, osteoarthritis, low back pain, anxiety/depression, COPD/asthma, gastro-intestinal disorders, thyroid gland disorders, urogenital disease; AIC: 471.67.
	Quartile 2		Quartile 3		Quartile 4
	z-value	OR (95% CI)	z-value	OR (95% CI)	z-value	OR (95% CI)
Osteoporosis = yes					2.36 (p-value 0.02)	10.10 (2.01–50.66)
Osteoarthritis = yes	−2.17 (p-value 0.04)	0.21 (0.06–0.69)
Urogenital dis. = yes	−2.18 (p-value 0.02)	0.19 (0.06–0.66)
Frailty status = pre-frail					−2.25 (p-value 0.04)	0.16 (0.04–0.61)

Model “Comorbidity level and functional disorders”: variables in the input: frailty index (0,1,2), nutritional status, incontinentio urinae, number of comorbidities, and number of medications prescribed; AIC: 494.51.
	Quartile 2		Quartile 3		Quartile 4
	z-value	OR (95% CI)	z-value	OR (95% CI)	z-value	OR (95% CI)
Frailty status = pre-frail					−1.92 (p-value 0.05)	0.30 (0.11–0.84)

Model “Antidiabetic drugs”: variables in the input: metformin, sulfonylureas, pioglitazone, DPP4inh, GLP1ra, SGLT2inh, insulin therapy; AIC: 507.85.
	Quartile 2		Quartile 3		Quartile 4
	z-value	OR (95% CI)	z-value	OR (95% CI)	z-value	OR (95% CI)
Pioglitazone = yes					−1.98 (p-value 0.04)	0.06 (0.01–0.62)
DPP4inh = yes					−2.41 (p-value 0.04)	0.08 (0.02–0.46)

Model “Anthropometric measures”: variables in the input: BMI, wc; AIC: 505.80. These variables did not show a significant relation with the output variable (quartiles of IL-17A), all p-values higher than 0.05.

Model “Other prescribed drugs”: variables in the input: NSAID, ACE-INH/ARBs, Calcium channel blockers, Beta-blockers, Diuretics, Statins; AIC: 498.77. These variables did not show a significant relation with the output variable (quartiles of IL-17A), all p-values higher than 0.05.

Model “Cardio-metabolic comorbidities”: variables in the input: CAD, CHD, retinopathy, smoking habit, Metabolic Syndrome; AIC: 502.08. These variables did not show a significant relation with the output variable (quartiles of IL-17A), all p-values higher than 0.05.

Since regression models emphasized frailty as the key factor that modifies associations of other variables with variations in serum IL-17A levels, we performed several graphical presentations (Figures 2–5) to reveal associations of some factors, for which evidence indicates their associations with frailty, with the frailty status of patients in the sample. In addition, Figure 6A, 6B show how several factors considered together influence variations in serum IL-17A levels.

Figure 2.

Sex-dependent differences in frailty status: nonfrail, pre-frail and frail. R, 4.3.1, R Development Core Team.

Figure 3.

(A) Sex (men)-dependent distribution of older diabetic patients according to frailty status and BMI categories. (B) Sex (women)-dependent distribution of older diabetic patients according to frailty status and BMI categories. R, 4.3.1, R Development Core Team.

Figure 4.

Sex-dependent differences of diabetic patients according to the frailty status (yellow – men, violet – women). R, 4.3.1, R Development Core Team.

Figure 5.

Distribution of diabetic patients with/without CVD according to the frailty status. R, 4.3.1, R Development Core Team.

Figure 6.

(A) IL-17A in men and women <65 years old, distributed according to differences in the frailty status and BMI categories. (B) IL-17A in men and women ≥65 years old, distributed according to differences in the frailty status and BMI categories. R, 4.3.1, R Development Core Team.

Figure 2 shows that women predominate over men among frail patients. The proportion test across the frailty status categories showed a significant difference for men (P value 0.02) and for women (P value 0.02).

In men (Figure 3A), frailty status did not change according to changes in BMI; the proportion test (men) across the BMI categories showed no significant difference for “nonfrail” (P value 0.73), “pre-frail” (P value 0.75), and “frail” (P value 0.68) patients. The same proportion test (women) across the BMI categories showed no significant difference for “nonfrail” (P value 0.11), “pre-frail” (P value 0.07), and “frail” (P value 0.42) patients, although there was a tendency of overweight women (25–30 kg/m²) to be pre-frail (Figure 3B).

Figure 4 shows that there are no significant differences in age among patient subgroups defined by differences in the frailty status: nonfrail vs pre-frail (P value 0.22 by ANOVA), nonfrail vs frail (P value 0.20 by ANOVA), and pre-frail vs frail (P value 0.53 by ANOVA). Within the same subgroup, there were no significant differences from the sex perspective: nonfrail (P value 0.54 by ANOVA), pre-frail (P value 0.35 by ANOVA) and frail (P value 0.08 by Kruskal-Wallis rank-sum test). In fact, frail men (boxplots colored yellow) tended to be older than frail women (boxplots colored purple).

In Figure 5, the relationships between status CVD and the frailty index categories are detailed. P value 0.03 by Fisher test rejected the null hypothesis (no difference in CVD status within categories of the frailty index), and the post hoc test indicated significant differences for “light blue“ vs “dark blue” (P value 0.01, adjusted by Benjamini-Hochberg FDR method), “light blue” vs “green” (P value 0.01), and “light blue” vs “purple” (P value 0.01). The proportion test across the frailty index categories showed a significant difference for “light blue” (P value 0.003), “dark blue” (P value 0.02), “green” (P value 0.04), and no significant difference for “purple” (P value 0.61).

Figure 6 shows sex-dependent variations in serum IL-17A levels among patient subgroups defined by frailty status and BMI categories, depending on a difference in age. Figure 6A shows that in diabetic patients who are under 65 years of age, if they are not obese (BMI ≤30 kg/m²), frailty affects serum IL-17A levels, but this concerns only women, as they are tending to achieve frailty younger than men (in ages before 65 years) (P value 0.01 by ANOVA). Being obese (>30 kg/m²) (Figure 6A), justified serum IL-17A levels among nonfrail/pre-frail and frail patients, regardless of sex (no statistical difference by ANOVA). On the contrary, as seen in Figure 6B, in diabetic patients ≥65 years, there were no differences in serum IL-17A levels according to differences in frailty status, depending on whether patients are obese or not, and on sex (no statistical difference by ANOVA).

Discussion

This study aimed to identify relevant factors that in older primary care patients with T2D could influence serum IL-17A concentrations. The results have a potential to improve risk stratification and therapy options for these patients. The major finding was low IL-17A serum concentrations. Variables that showed significant variations across quartiles of IL-17A included HbA1c, LDL cholesterol, Hb, CVD (CAD and CHD), some non-CV comorbidities, and therapies with newer oral antidiabetics, taken all together (DPP4inh, GLP-1ra, and SGLT-2inh), and with beta-blockers. In regression models, independent associations with quartiles of IL-17A showed variables indicating comorbidity- and frailty-associated inflammation (hemoglobin), glucose-related metabolic disorders (HbA1c), frailty-associated non-CV conditions, notably including musculoskeletal diseases, and treatments with the antidiabetic drugs DPP4inh and pioglitazone. In models indicating comorbidities, the frailty index was a modifying factor.

The finding of low IL-17A serum concentrations is not surprising because this cytokine operates at the tissue level as a part of the cytokine network, so that small changes in its concentrations exert significant clinical effects [24]. The findings indicating low-level variability of markers of inflammation are in line with previous studies, where only classical markers of inflammation, such as IL-6 and CRP, have been explored, and where it was shown that these markers in cardio-metabolic conditions have only modest predictive value, or their validity as prediction factors have not been confirmed at all [68]. Specifically, the results of this study suggest a link between low IL-17A serum values and good metabolic control, as indicated by HbA1c <8.5% and LDL cholesterol <4.0 mmol/L, recorded in most participants, which are values within the recommended ranges for older diabetic patients.

Good metabolic control is largely the effect of treatment. In this regard, our results indicate that drugs with favourable metabolic effects, such as metformin, statins, and ACE-INH/ARBs, were widely prescribed to patients in the sample. The synergistic favourable effect of this combined therapy on metabolic disorders could have retarded the stimulatory effect of high insulin resistance on IL-17A serum values. The high insulin resistance state of patients in the sample was due to the fact that these patients had hypertension and MS as well as T2D [69,70], and all these drugs have recognized effects on ameliorating IL-17A-induced inflammation.

Diabetic patients had good metabolic control and low-level IL-17A-related system inflammation, so it is then reasonable that only a few of the examined factors were independently associated with variations in serum IL-17A concentrations. Another reason might be the complexity of patients in the sample, which implies complex interactions between connected factors, and can produce some hidden effects on the outcome variable [22]. The high level of collinearity and multicollinearity among the included variables confirms this.

Results of this study show that pre-frailty status is a key modifying factor of serum IL-17A concentrations, showing the negative effect (indicated by the negative sign of correlations), which is to be expected based on evidence indicating links among T2D, CVD, and IL-17A, and between these cardio-metabolic disorders and frailty. We also found that most frail diabetic patients had CVD (Figure 4), but evidence of an association between inflammation and frailty is not consistent [71,72]. When viewed from this angle, this result seems rather intriguing. Even more so, because pre-frailty, and not frailty, was shown to be a significant determinant of IL-17A serum levels, and the negative correlation is also curious.

These results may be explained by the emerging recognition of frailty as a multifaceted condition. There is an increasing awareness that to reveal variations in pathophysiological pathways and clinical phenotypes associated with frailty, new integrated analytical methods are needed [73]. At least 2 metabolic phenotypes of frailty in older diabetic patients are recognized – one associated with obesity and high insulin resistance (sarcopenic obese phenotype), and another associated with muscle and weight loss and low insulin resistance (anorexic malnourished phenotype). In addition, growing evidence indicates that the clinical expression of frailty is sex-dependent. Several pieces of evidence support this statement. Thus, it was shown that inflammatory markers better predict frailty in women than in men [74,75]. Women were more prone to frailty and frailty-related physical disability than men, while men can live longer with dysregulated physiological systems without becoming frail [76,77].

The sex bias in frailty expression might be especially strong in older diabetic patients because of their exceptional predisposition for CVD [78], and diabetic women were shown to have higher risk for CVD than diabetic men [79]. In addition, evidence suggests that women have predisposition for a certain type of CVD (subclinical CHD), and that women with CHD express higher levels of inflammation and higher predisposition for frailty than men with CHD [80,81]. These characteristics of women with CVD may be partly due to the tendency of women age 50 years and older (after menopause) to have abdominal obesity and MS-related T2D [82].

Our results need to be interpreted within the context of this described clinical framework, specific for older diabetic patients. In line with the facts stated above, diabetic women in the sample had a higher predisposition for frailty than diabetic men (Figure 2), as well as a tendency to develop frailty at a younger age (Figure 4). A very high proportion (>77%) of frail patients diagnosed with CVD (Figure 5) are older diabetic patients, in particular women, who have a high predisposition for CVD and for frailty. As shown in Figure 3A and 3B, although not statistically significant, overweight diabetic women are more likely to be pre-frail, as are postmenopausal diabetic women with abdominal-type obesity, which is associated with high level of insulin resistance, and consequently with increased IL-17A-related inflammation, which in turn predisposes them to development of CHD and frailty.

Whether progressive development of frailty (from pre-frailty to frailty), in line with gradually increasing severity of CHD, being more pronounced in diabetic women than men, can act in a way to revert the high level of insulin resistance/high level of inflammation to lower rates, or it is the effect of treatment, it cannot be accurately concluded from these results. Regression models suggested the negative correlations between pre-frailty status and IL-17A serum values in the upper quartile. The same correlation pattern applies to therapy with DPP4inh and pioglitazone. These drugs are used for treating overweight/obese diabetic patients, in particularly those with abdominal obesity, in whom these substances may exert an additional ameliorating effect on IL-17A-related inflammation. In line with this scenario, we found that more severe forms of CHD (such as CAD+CHD) characterize fully frail diabetic patients more than pre-frail ones (Figure 5). Also, evidence suggests that T2D usually appears in association with increased BMI and subclinical CVD and pre-frailty, but in later course of this disease progression, and, in parallel with accumulation of comorbidities, the link between severity of CVD and frailty grows stronger [83].

This dynamic association between CVD and frailty may be stronger in older diabetic patients, influenced by sex bias on CVD patterns and rates of progression, but also on clustering other, non-CV comorbidities [84,85]. Certain comorbidity patterns and frailty phenotpyes (obese vs non-obese) that may exist in a particular patient sample, depending on the age and sex structure of the sample, may variably affect IL-17A serum levels. The results of the regression models indicate musculoskeletal diseases, especially osteoporosis and osteoarthritis, as independent predictors of IL-17A serum levels. These disorders are commonly present in older, female, diabetic patients, and can contribute to frailty through the effect on sarcopenia development [86,87]. Particularly, the presence of osteoporosis, as inflammation-mediated disease, similar to autoimmune disease, can be a strong predictor of elevated IL-17A serum values [88].

The effect of certain comorbidities on IL-17A serum levels is mediated mostly through their association with frailty, as suggested by the selection of musculoskeletal diseases in regression models, and of urogenital disorders, which are increasingly considered part of frailty syndrome [73,89]. The selection into regression models of decreased hemoglobin (as indicated in Table 4), which is a marker of comorbidity-associated frailty, as an independent predictor of elevated IL-17A serum values, confirms this impression. Among different variables indicating geriatric conditions and the level of comorbidity, only variable indicating state pre-frailty were selected in the regression model. Participation of certain types of frailty (obese vs non-obese, including ratio pre-frailty/frailty) in the patient sample, which is age- and sex-dependent, might be important for predicting IL-17A serum values, as shown in Figures 6A, 6B, and 3A, 3B.

The results indicating the close association of CVD with frailty (Figure 5C), the sex bias in frailty (Figure 2), the distribution of pre-frailty and frailty across categories of BMI (Figure 3A, 3B), and the fact that particular frailty phenotypes can differentially affect IL-17A serum levels in diabetic men and women, and according to their age (Figure 6), can probably explain the lack of associations of variables CVD and BMI with IL-17A serum levels, although it was initially proposed. The link between BMI, CVD, and frailty, might be via MS (based on abdominal obesity). Arguments for this statement are given by the pieces of evidence, indicating, eg, that pre-frailty/frailty is linked to higher levels of adiposity, but only in the presence of abdominal obesity (a benchmark of MS) [90], and that in the elderly population (65+ years old), MS shows constant associations with the overall and CV mortality, across categories of BMI [91], and that obesity and underweight are associated with increased risk of frailty [92]. The high propensity of older diabetic patients for MS confirms by the fact that about three-quarters of men and almost all women had MS (Table 1B). The close relationship between MS and frailty realizes via sarcopenia, as indicated by the collinearity between variables MS and mac (Table 3).

A similar explanation may pertain to the lack of an association of eGFR with IL-17A serum levels, because of the cofounding effect of frailty, and the close association of MS and frailty, irrespective of BMI values [93]. In addition, evidence suggests that frailty does not develop until eGFR reaches some threshold of severity [94]. Because most participants in this study had a moderate decrease in renal function, they could not have developed frailty. Lastly, the mild-moderate stage of eGFR severity, along with the predisposition of women in the sample to have a subclinical form of CHD, might be associated with higher tendency of women in the sample to have pre-frailty (Figure 3A, 3B).

Taken together, the sex disparity in CV and non-CV comorbidity and frailty phenotypes in older diabetic patients, as indicated by the results of this study, must be considered when assessing the diagnostic and prognostic values of serum IL-17A levels in these patients. This further implies the possible different predictive and prognostic value of this inflammatory marker in diabetic women vs diabetic men. Longitudinal studies, conducted separately in women and men, would provide more details.

Limitations of the study

Limitations of this study include the relatively small number of participants for the number of variables used for analysis, which could have caused overestimation of the effect size in differences and regression analyses, and non-randomized participant selection, which could have impacted the variable selection in regression models. In addition, multivariable interactions and the presence of confounding factors could have masked or falsely demonstrated the actual associations. The mitigating circumstances are that residents in the study area have good access to GPs and that GPs employ similar methods in screening and managing diabetic patients, and that the applied regression models do not require many of the principles of linear regression models, and randomization does not justify the assumptions behind the model.

The recognized effect of sex bias and frailty phenotypes on IL-17A serum values should be addressed in future studies to separately assess several patient subgroups, defined at least by variables such as age, sex, frailty status, and BMI. The new methods for the complex data analysis from machine learning could provide benefits by separating diabetic patients into discrete subgroups, defined by a set of variables that tend to cluster together, or by phenotypes.

Other limitations are that individuals with subclinical atherosclerotic disease have not been identified, which could have influenced serum IL-17A levels in cases without a diagnosis of CVD, and there was no information on the grade severity of CHD.

Conclusions

This study is the first attempt to identify sociodemographic and clinical determinants of IL-17A serum levels in T2D patients age 50 years and older. The results have revealed the complexity of these patients, by means of the existence of multiple patient subgroups that can differentially impact IL-17A serum levels. Future studies should address the modifying role on IL-17A serum levels of sex, age, and frailty phenotypes.