Association Between Metabolic Syndrome Components, Clinical Characteristics, and Telomere Length: Factor Analysis of Mixed Data Based Cluster Analysis of LIPIDOGEN2015 Cross-Sectional Study

Abstract

Introduction

Telomere length is an acclaimed marker of aging, which has been previously shown to correlate with cardiovascular diseases and metabolic syndrome traits.

Aim

To identify the relationship between patient characteristics and telomere length.

Methods

The LIPIDOGEN was a random patient sample substudy of LIPIDOGRAM 2015 study (n = 13,724) conducted in primary care facilities in Poland. Data on risk factors, chronic diseases, treatment, and lifestyle were collected. Telomere length was determined with routine PCR from saliva. Factor Analysis for Mixed Data analysis was utilized to discern the principal components of patient clinical profiles. Furthermore, hierarchical clustering was used to obtain clusters of patients based on principal components.

Results

1556 patients (60% female, mean age 51 years) were included in the analysis after the exclusion of outliers and low DNA quality samples. Three clusters of patients were identified. Cluster 1 was characterized by low cardiovascular risk, without significant risk factors. Cluster 2 consisted of patients with a higher incidence of metabolic syndrome (MetS, 62%) and the highest smoking rate (22%). Cluster 3 had the highest incidence of MetS (94%), treatment with statin (62%), and diabetes mellitus (61%), and contained nearly all patients with myocardial infarction (17% of this cluster). Patients in Cluster 1 had significantly longer telomeres than patients in Cluster 2 and 3 (p = 0.01 and p < 0.001 respectively).

Conclusions

The pattern of clinical characteristics marked by classical cardiovascular risk factors including components of MetS, is inversely related to telomere length, underlining the potential role of metabolic disturbances in cellular aging.

Graphical Abstract

Supplementary Information

The online version contains supplementary material available at 10.1007/s40292-025-00753-6.

Introduction

Telomere length (TL) is a marker of cellular aging. Telomeres, located at the ends of chromosomes, contain sequence repeats (TTAGGG) and are necessary to maintain genome integrity [1]. Due to the fact that telomerase is not active in most somatic cells, telomeres shorten with each cell division due to the inability of DNA polymerase to replicate the end of a chromosome [2]. The shortening of telomers is a complex issue with numerous factors influencing this process, and genetics appears to be the most important one [3]. For example, it has been shown, that the dynamics of telomere length change over the course of life is largely heritable [4]. Regardless, acquired diseases may also affect the length of telomeres. A negative impact of various metabolic diseases, including obesity and diabetes mellitus (DM), on TL was found [5]. Studies using Mendelian randomization also indicated that shorter telomeres are causally associated with the risk of faster progression of type 2 DM and the need to include insulin in the treatment [6]. The diseases that influence TL rarely occur in isolation and complex interactions between them may determine TL. Exploratory dimension–reduction techniques enabling analysis of both categorical and continuous variables can identify complex patterns of associations between TL and sociodemographic and clinical variables.

Methods

In this study, we aimed to identify factors associated with relative telomere length (RTL) in a population of adult patients representative for the population under the care of family physicians in Poland.

LIPIDOGRAM2015/LIPIDOGEN2015 Study

The design of LIPIDOGRAM2015 study and LIPIDOGEN2015 substudy were described elsewhere in detail [7]. In brief, LIPIDOGRAM2015 study was a nationwide cohort study on cardiovascular diseases and cardiometabolic risk factors enrolling patients through primary care outpatient centers in Poland. In the LIPIDOGRAM2015 study, 438 primary care physicians from 398 cities in all of the 16 administrative regions in Poland recruited 13,724 patients. Patients recruited to LIPIDOGEN2015 substudy cohort (n = 1,788) were a random subset of patients of LIPIDOGRAM2015 cohort. Each patient was asked to complete a questionnaire that included questions related to the patient’s demographics, cardiovascular and cardiometabolic risk factors, chronic conditions, treatment, level of education, place of residence, and lifestyle. Patients from the LIPIDOGEN2015 substudy had saliva secured for DNA analyses (Fig. 1).

Fig. 1.

Study flow chart

Biochemical Analyses

Blood samples were obtained and sent in refrigerated containers to the Silesian Analytical Laboratories (SLA) in Katowice, Poland. Direct immunological techniques were used to assess lipid profiles. The analyses were performed using the Siemens Advia 1800 analyzer and Siemens reagents (Munich, Germany) within 12 hours of sample collection. Fasting blood glucose was measured utilizing Bionime glucometers and Rightest test strips, both sourced from Taichung City, Taiwan. Hemoglobin A1c (HbA1c) was assessed via high-performance liquid chromatography (HPLC) with the Variant II Turbo system (Bio-Rad, Hercules, California, USA).

Genetic Analyses

Two–milliliter saliva samples were collected according to the manufacturer’s instructions for the Oragene-DNA/OG-500 kit (DNA Genotek, Ottawa, Canada). Isolation of DNA from banked saliva was carried out using the PrepIT-L2P kit (DNA Genotek, Ottawa, Canada) according to the manufacturer’s protocol. DNA concentrations were determined using spectrophotometry by measuring the absorbance ratio at 260/280 nm using a BioSpectrometer (Eppendorf, Germany). The mean concentration of isolated DNA was 349.6 µg/ml (IQR—interquartile range 142.8 µg/ml–486.3 µg/ml). The multiplex quantitative polymerase chain reaction (MMQPCR) method, based on the modified Cawthon procedure [8, 9], was utilized to measure RTL. Standard curves were determined by serial three–fold dilutions of genomic DNA ranging from 60 ng to 0.74 ng. All experimental and control samples were conducted in triplicate. Each reaction well comprised 2 μl of DNA (10 ng), 2x SYBR Green PCR Master Mix (Bio-Rad, Hercules, California, USA), two sets of primers, and water to a final volume of 10 μl. Primer sequences are specified in Supplementary Table S1, Supplementary File 1.

Thermal cycling profile was as follows: initial denaturation by 15 min at 95 °C, 2 cycles of 15 s at 94 °C, 15 s at 49 °C; 35 cycles of 15 s at 94 °C, 10 s at 62 °C, 15 s at 74 °C with signal acquisition (signal for telomeres), 10 s at 84 °C, 15 s at 88 °C with signal acquisition (signal for albumin). Following thermal cycling and raw data acquisition, two standard curves were prepared for each plate: one for telomeres and another for the reference gene, albumin, using CFX Manager Software (Bio-Rad, USA). The efficiency of the telomere and albumin reactions was compared and was the same and not lower than 90%. MyiQ software (Bio-Rad, USA) was used to determine the T (telomere) and S (single copy gene) values. The ratio between telomere and albumin products (T/S) is proportional to the average TL per cell and represents the RTL. The average TL of the sample with a T/S ratio exceeding 1.0 is greater than that of standard DNA; conversely, the average TL of the sample with a T/S ratio of 1.0 is lesser than that of standard DNA.

Statistical Analyses

Dichotomous data was presented as percentages (%). Continuous variables were presented as means and standard deviations (SD) for normally distributed variables or medians with interquartile range (IQR) for variables whose distribution deviated from normal. To assess differences between LIPIDOGEN sample and LIPIDOGRAM2015 population, appropriate effect measures were used—standardized mean difference for normally distributed continuous variables, Cliff’s delta for non-normally distributed continuous variables and Cohen’s h for categorical variables. The normality assumption was tested with the visual inspection of QQ plots (See Supplementary Figure S1, Supplementary File 1). TL distribution was right-skewed as assessed by the histogram (See Supplementary Figure S2, Supplementary File 1). Outliers in the log-transformed TL data were identified and excluded based on the interquartile range (IQR) method. Specifically, outliers were defined as values below Q1 − 1.5 IQR or above Q3 + 1.5 IQR. Logging and exclusion of outliers led to the normalization of distribution (See Supplementary Figures S1B and S3, Supplementary File 1). To assess the relationship between clinical characteristics and log-transformed TL, univariate linear regression models with log-transformed TL as a dependent variable were fitted for each variable from clinical characteristics. It is widely known that several metabolic health parameters like BMI, WC, triglycerides, blood glucose, and Hb_A1c are highly correlated and colinear [10, 11]. Standard multivariable analysis may fail to properly reflect the structure of actual population characteristics in a given dataset and lead to erroneous results. To eliminate collinearities between the independent variables, the variance inflation coefficient (VIF) was calculated. A VIF of 5 threshold was assumed, above which variables were iteratively removed, starting with the one with the highest value. The process was repeated until the final model was obtained with all the variables meeting the VIF criterion ≤ 5. Such optimized multivariate regression models were also used in three age groups (< 45, 45–65, > 65 years).

To further diminish influence of collinearity on results we aimed to identify the latent structure of the data and the variables contributing to the main components of interpatient variation; Factor Analysis for Mixed Data (FAMD) was used, which handles both quantitative and qualitative variables. FAMD analysis allows for the resultant principal components to be represented in vector space. If two observations are close to each other in the principal component space, it means they are similar in terms of those clinical variables that contribute the most to the given principal component. Principal components are linear combinations of the original variables, with specific weights (referred to as loadings) assigned to each variable. These components are designed to capture the maximum amount of variation in the data by projecting it onto a new coordinate system. Variables with the highest contributions to a given principal component are visualized within the principal component space. Scores for each observation (i.e., the coordinates in the PCA space) were extracted and integrated back into the original dataset. The number of components to retain was determined based on Kaiser Criterion (eigenvalue >1), inflection points identified in the Scree plot, and their overall interpretability. To obtain clusters of patients with similar clinical profiles, hierarchical clustering from principal components was used to identify clusters of observations in the dataset. HCPC function from FactoMineR [12] was used. Clustering shows how clusters are formed at different levels of hierarchy, providing insight into the nested structure of the data. Rectangles were drawn around the clusters at the chosen cut–off level, with distinct colors for each cluster. Clusters were then assigned to the original dataset and treated as factors. Differences in clinical characteristics between patients in each cluster were assessed using the Jonckheree-Terpstra trend for test and testing for trends in proportions. Differences in RTL between clusters were assessed with the Kruskall-Wallis test and the post-hoc Tukey test. Analyses were performed using R studio [13]. Artificial intelligence [14], was used to assist when writing code. Figure 7 was drawn using Mind the Graph [15].

Fig. 7.

Environmental factors and their influence on telomere shortening. Exposure to environmental stressors increases the release of stress hormones (cortisol, epinephrine, norepinephrine), which are linked to the development of metabolic syndrome components. These include obesity, hypertension (associated with increased oxidative stress and inflammation), dyslipidemia (which promotes free radical production), and diabetes/prediabetes (linked to the formation of advanced glycation end products, AGEs). These processes cumulatively contribute to inflammation, oxidative stress, and ultimately, telomere shortening

Ethical Approval

The LIPIDOGRAM2015 study & LIPIDOGEN2015 substudy received approval from the Bioethical Committee of the District Medical Chamber in Częstochowa (approval no. K.B.Cz.–0018/2015). All studies adhered to the principles outlined in the Declaration of Helsinki. Written informed consent was obtained from all participants, and measures were taken to ensure both anonymity and confidentiality.

Results

Clinical Characteristics and Telomere Length

The LIPIDOGEN2015 substudy included 1785 randomly selected patients from the LIPIDOGRAM2015 study population (n = 13,700). The clinical characteristics of the patients were similar to those of the LIPIDOGRAM2015 study, except for the level of education, since there was a significantly higher percentage of patients with secondary or higher education in the LIPIDOGEN2015 substudy. High-quality DNA was isolated from 1628 patients, and reliable TL measurements were obtained from 1604 patients. After excluding outliers and patients with invalid data, 1556 patients were included in the analysis (Fig. 1). After exclusion of patients due to aforementioned technical and logistical reasons there were some differences between clinical characteristic sampled population and LIPIDOGRAM2015 population, meaningful for education and age (See Supplementary Table S2, Supplementary File 1). The mean age of the study group was 51 years (SD: 13 years). The majority (60%) were women. 74% of individuals reported having a secondary or higher education. Metabolic syndrome (MetS) could be diagnosed in 50% of the patients based on the Joint Interim Consensus criteria of 2009 [16]:

• Elevated waist circumference (≥ 102 cm in men; ≥ 88 cm in women);

• Elevated triglycerides (≥ 1.7 mmol/L or treatment of hypertriglyceridemia);

• Low HDL-C (< 1 mmol/l in men; < 1.3 mmol/l in women or treatment of low HDL-C);

• Elevated blood pressure (≥ 130 mmHg SBP or ≥ 85 mmHg DBP or hypertension treatment);

• Elevated fasting glucose (≥ 100 mg/dL or diabetes mellitus).

Detailed clinical characteristics of the patients included in the study are presented in Table 1.

Table 1.

Clinical characteristics of the study participants.

Characteristic	N = 15561
Age	51 (13)
Female sex	939 (60%)
Place of residence	858 (55%)
Secondary or higher education	1148 (74%)
Physical activity	706 (45%)
Smoking	252 (16%)
Alcohol	1131 (73%)
SBP (mmHg)	126 (120; 140)
DBP (mmHg)	80 (70; 85)
BMI (kg/m2)	28.1 (4.9)
Waist circumference (cm)	94 (14)
Metabolic syndrome	781 (50%)
Diabetes Mellitus	229 (15%)
Hypertension	650 (42%)
CKD	37 (2.4%)
Atrial fibrillation	40 (2.6%)
Previous myocardial infarction	59 (3.8%)
Family history of CV disease	706 (45%)
Dyslipidemia	1122 (72%)
Fibrate	83 (5.3%)
Statin	423 (27%)
Total cholesterol (mmol/l)	5.36 (1.18)
LDL-C (mmol/l)	3.22 (0.98)
HDL-C (mmol/l)	1.36 (1.14; 1.69)
Triglycerides (mmol/l)	1.43 (1.02; 2.00)
Non-HDL-C (mmol/l)	3.93 (1.13)
Remnants (mmol/l)	0.61 (0.49; 0.79)
Elevated remnant-C2	421 (27%)
Glucose (mg/dl)	98 (90;109)
HbA1c (%)	5.50 (5.20; 5.80)
RTL	0.82 (0.67; 0.99)

¹Mean (SD); n (%); Median (IQR)

²Elevated remnant-C: remnant cholesterol >0.77 mmol/l

The results of univariate analysis of factors affecting RTL are presented in Fig. 2. Clinical characteristics associated with insulin resistance and MetS that is increased waist circumference, obesity, HTN, elevated glucose concentration, increased HbA1c, increased remnant-C, and triglycerides were the strongest factors influencing RTL. Higher education and higher high-density lipoprotein cholesterol (HDL-C) levels were associated with increased RTL (Fig. 2). In the multivariate analysis HDL-C levels, higher education were associated with longer telomere length, whilst age and smoking were associated with shorter telomere length (Fig. 3). A multivariate analysis stratified by age groups showed that in the 50–64 age group, smoking and diabetes are negatively associated with telomere length, meaning that individuals with these conditions tend to have shorter telomeres. On the other hand, higher education appears to be positively associated, suggesting a potential protective effect. Other factors, such as LDL-C, hypertension, and female sex, exhibit weaker associations, with their impact being less pronounced (See Supplementary Figure S4, Supplementary File 1).

Fig. 2.

Results of univariate analysis. Correlation coefficients are ranked from highest positive to highest negative correlation

Fig. 3.

Results of multivariate analysis. Correlation coefficients are ranked from highest positive to highest negative correlation

Dimensions Identified with FAMD

Due to the collinearity of several metabolic health indices, multivariable analysis may produce flawed results. Instead of multivariable analysis, FAMD was used, which combines collinear variables into single ‘dimension’ variables that can be represented in a vector space. Ten dimensions with eigenvalues greater than one, explaining a total of 66.8% percent of the variance within the dataset were identified. Basing on the Scree plot and interpretability of dimensions we focused on the first four components explaining 18.2%, 12.7%, 6.9%, and 5.4% of the dataset variance (Supplementary Figure S5, Supplementary File 1). Variables associated with MetS that is hypertension (HTN) and blood pressure, obesity, triglyceride levels, diabetes mellitus (DM), glucose, and Hba1c levels contributed most and were positively correlated with Dimension 1—“Metabolic Syndrome and Cardiovascular Risk Factors” (Figs. 4 and 5, also see Supplementary Figure S6A, Supplementary File 1). Dimension 2, named “Lipid profile”, was most strongly correlated with lipid parameters—total cholesterol, low-density lipoprotein cholesterol (LDL-C), and non-high-density lipoprotein cholesterol (non-HDL-C) (Figs. 4 and 5, also see Supplementary Figure S6B, Supplementary File 1). Dimension 3—“Demographic and Lifestyle Factors” was mainly correlated with diet, alcohol consumption as well as age and sex (See Supplementary Figures S6D, S7A, S7B and S7C, Supplementary File 1). Dimension 4, explaining the least of the observed variance, was correlated with blood pressure and remnant cholesterol (remnant-C) and, similarly to Dimension 1, with glucose metabolism (See Supplementary Figures S6D, S7A, S7B and S7C, Supplementary File 1).

Fig. 4.

Contributions of variables to first and second dimensions. FAMD—factor analysis of mixed data

Fig. 5.

Correlations between variables and first and second dimensions. a The strength and direction of correlation of continuous variables and the first and the second dimension. b The strength and direction of correlation of categorical variables and the first and the second dimension

Cluster Analysis and Telomere Length

Hierarchical clustering based on principal components was used to identify clusters of observations in the dataset and determine their relationship with TL. Three clusters representing patients with different types of clinical characteristics were identified (Table 2, Fig. 6A). Clusters were most strongly differentiated by non-HDL-, LDL-C, TC, HbA1c, presence of diabetes mellitus, remnant-C, and triglyceride levels (all p < 0.001). The clinical characteristics of each subgroup are presented in Table 2.

Table 2.

Clinical characteristics of three identified clusters of patients.

Variable	Cluster 1 Low CV risk1 n = 660	Cluster 2 High CV risk, w/o prior MI1 n = 568	Cluster 3 Very high CV risk, MetS, w prior MI1 N = 328	p-value2
Age	45 (13)	52 (11)	60 (10)	< 0.001
Waist circumference (cm)	85 (11)	97 (12)	106 (12)	< 0.001
BMI (kg/m2)	25.3 (3.8)	29.1 (4.2)	32.0 (4.9)	< 0.001
Glucose (mg/dl)	94 (87; 100)	98 (91; 106)	117 (105; 148)	< 0.001
HbA1c (%)	5.30 (5.10; 5.50)	5.50 (5.20; 5.80)	6.20 (5.70; 7.20)	< 0.001
Total Cholesterol (mmol/l)	4.94 (0.84)	6.30 (0.98)	4.58 (1.00)	< 0.001
HDL-C (mmol/l)	1.58 (1.30; 1.85)	1.33 (1.10; 1.57)	1.16 (0.97; 1.35)	< 0.001
LDL-C (mmol/l)	2.83 (0.70)	4.01 (0.84)	2.62 (0.79)	< 0.001
Triglycerides (mmol/l)	1.01 (0.80; 1.29)	1.86 (1.45; 2.64)	1.67 (1.26; 2.28)	< 0.001
Non-HDL-C (mmol/l)	3.34 (0.73)	4.92 (0.89)	3.39 (0.98)	< 0.001
Remnants (mmol/l)	0.49 (0.41; 0.57)	0.78 (0.66; 1.00)	0.66 (0.53; 0.83)	< 0.001
SBP (mmHg)	120 (110; 130)	130 (120; 140)	140 (130; 150)	< 0.001
DBP (mmHg)	75 (70; 80)	80 (75; 90)	80 (80; 90)	< 0.001
Female sex	496 (75%)	287 (51%)	156 (48%)	< 0.001
Place of residence	377 (57%)	301 (53%)	180 (55%)	0.4
Secondary or higher education	576 (87%)	403 (71%)	169 (52%)	< 0.001
Previous MI	1 (0.2%)	1 (0.2%)	57 (17%)	< 0.001
Family history of CVD	312 (47%)	256 (45%)	138 (42%)	0.12
Smoking	76 (12%)	127 (22%)	49 (15%)	0.016
Alcohol	478 (72%)	426 (75%)	227 (69%)	0.5
Diabetes Mellitus	6 (0.9%)	22 (3.9%)	201 (61%)	< 0.001
Hypertension	113 (17%)	247 (43%)	290 (88%)	< 0.001
MetS	122 (18%)	352 (62%)	307 (94%)	< 0.001
Dyslipidemia	330 (50%)	540 (95%)	252 (77%)	< 0.001
Elevated remnant-C3	24 (3.6%)	297 (52%)	100 (30%)	< 0.001
Fibrate	14 (2.1%)	32 (5.6%)	37 (11%)	< 0.001
Statin	113 (17%)	107 (19%)	203 (62%)	< 0.001
CKD	12 (1.8%)	8 (1.4%)	17 (5.2%)	0.005
Physical activity	321 (49%)	262 (46%)	123 (38%)	0.002
Atrial fibrillation	8 (1.2%)	13 (2.3%)	19 (5.8%)	< 0.001
RTL	0.85 (0.71; 1.01)	0.80 (0.65; 0.97)	0.78 (0.63; 0.97)	< 0.001*

¹Mean (SD); Median (IQR); n (%)

²Jonckheere-Terpstra test; Chi-squared Test for Trend in Proportions

⁴Elevated remnant-C: remnant cholesterol > 0.77 mmol/l

^*Kruskal–Wallis rank sum test

Fig. 6.

Results of hierarchical clustering analysis and telomere length quintiles in each cluster. A The course of hierarchical clustering—identification of 3 clusters. B Scatter Plot with Telomere Length Quintiles: each point represents a patient, colored according to their assigned cluster, and the size of the point indicates the quintile of their log-transformed telomere length

Cluster 1 comprised 660 individuals, with females being most prevalent (75%). In general, Cluster 1 contained patients with low cardiovascular risk. On average, this subgroup comprised the youngest patients (mean 45 years, SD 13 years, range 19–89 years) with a favorable clinical risk profile, including the lowest BMI (25.3 kg/m², SD 3.8 kg/m²), waist circumference (85 cm; SD 11 cm), glucose (median 94 mg/dl, IQR 87; 100), and HbA1c levels (median 5.30%, IQR 5.10; 5.50). Patients in this group also had the lowest incidence of DM (0.9%), HTN (17%), and MetS (18%). HDL levels were the highest in this cluster (1.58 mmol/l IQR 1.30; 1.85) along with the percentage of patients with secondary or higher education (87%).

Cluster 2 included patients defined as primary prevention with high cardiovascular risk. This group included medium-aged patients (52 years, SD 11 years, range 20–82 years). The percentage of female patients was 51%. The incidence of previous myocardial infarction was very low (0.2%), similar to Cluster 1. Nearly two-thirds of patients had MetS (62%), and smoking was more prevalent in this group (22%) compared to Cluster 1 and 3 patients.

Cluster 3 had clinical characteristics that opposed those of Cluster 1. This subgroup comprised older patients (median 60 years, SD 10 years, range 26–84 years). Females constituted 48% of patients in this group, similar to Cluster 2. Cluster 3 was characterized by the highest BMI (32.0 kg/m², SD 4.9 kg/m²), waist circumference (106 cm, SD 12 cm), glucose (median 117 mg/dl, IQR 105; 148), and HbA1c levels (median 6.20%, IQR 5.70; 7.20), which corresponded to the highest incidence of DM (61%) in this group. Nearly all (94%) patients in this group had MetS. Patients in this subgroup had lower total cholesterol (4.58 mmol/l, SD 1.00), LDL-C (2.62 mmol/l, SD 0.79), and triglycerides (median 1.67 mmol/l, IQR 1.28, 2.28) compared to Clusters 1 and 2—possibly due to treatment, as 62% of patients were prescribed statins compared to 17% in Cluster 1 and 19% in Cluster 2. Additionally, this cluster included almost all of the patients with previous myocardial infarction (17%). TL differed significantly between the three identified clusters (Table 2, Fig. 6B). In the post-hoc analysis, Cluster 1 had significantly higher RTL than either Cluster 2 or Cluster 3 (p = 0.01 and p < 0.001 respectively), while the difference between Cluster 2 and 3 was not significant (p = 0.19).

Discussion

Our study aimed to determine the factors associated with TL. Because TL reflects both genetic propensity and the cumulative impact of environmental influences on the organism, it is widely accepted as an aging marker [17]. We found that shorter RTL apart from age and smoking were associated directly with unfavorable clinical and metabolic profiles including HTN, DM, elevated glucose, and Hba1c, increased remnant-C, non-HDL-C, and triglycerides. Older age is associated with telomere shortening as shown by multiple studies beginning from the second half of the previous century [18]. In our population smoking was also associated with TL shortening which was also similar to the results of other studies [19], although this effect could be even more pronounced in our population as we used oral epithelium for DNA isolation [7]. Notably, all the other factors were associated with insulin resistance: increased waist circumference, glucose, HbA1c levels, and DM. Increased remnant-C and triglycerides were also significantly associated with reduced TL and both of those lipid abnormalities are frequent in patients with insulin resistance [20].

To simultaneously examine the complex relationships between multiple variables, a multivariate analysis was conducted, which revealed significant associations between clinical and demographic factors, particularly telomere length and cardiovascular risk factors (diabetes mellitus, HDL-C and LDL levels, age, sex, smoking status, level of education, hypertension) across different age groups. Increased age and smoking are linked to shorter telomere length [18, 19], while higher HDL-C levels and education appear protective, as also demonstrated by other studies conducted on the U.S. population [21]. Surprisingly, the analysis revealed that hypertension in the 50–64 age group may be positively associated with telomere length. This result may stem from the use of certain medication groups, such as ACE inhibitors, which may have a protective effect on telomeres [22]. The impact of these factors is most pronounced in individuals aged 50-64, highlighting the need for targeted interventions. Nevertheless, the factors comprising MetS are highly intertwined, resulting in them being correlated and therefore collinear [10, 11]. As a result, standard multivariable regression may produce results with wider and somewhat unstable error estimates and p-values for some of the factors [23]. To strengthen our analysis and better understand the underlying correlations and patterns behind clinical characteristics, we utilized PCA. This method allows us to automatically select the principal components, which best explain the variability of data and complex interactions between the individual factors. Instead of treating the variables individually, PCA assigns weights to several variables at once and lumps them into “dimensions”. Importantly, PCA attempts to produce non-collinear dimensions which explain the variability of the dataset without redundancy [24]. In further analyses, we used clustering analysis to find groups of patients with similar clinical profiles.

Adjusting the clustering algorithm to extract three patient subgroups revealed an interesting pattern: Cluster 1 with low cardiovascular risk, Cluster 2 included patients with high cardiovascular risk but no prior myocardial infarction, and Cluster 3 consisted almost entirely of patients with MetS, as defined by the Joint Interim Statement criteria [16], along with nearly all patients with a history of myocardial infarction recruited in the LIPIDOGEN2015 substudy. Therefore, we suggest that the underlying pathology that leads to telomere shortening in this subgroup of patients, apart from age–related shortening, is mainly MetS–related.

MetS is a central obesity-related constellation of factors such as elevated blood pressure and dyslipidemia – which encompasses both increased triglycerides and low HDL-C. Obesity is not synonymous with MetS, as some obese patients do not have significant metabolic disorders [25]. Adipose tissue acts as a storage reservoir for excess calories from overfeeding; however, when ectopic fat accumulates within muscles, the abdominal cavity (visceral adipose tissue), and the liver, metabolic disorders arise. The common denominator of these disorders is insulin resistance [26]. There are genetic predispositions to the development of MetS, meaning that obesity is not a sine qua non-condition for its onset [27]. An extreme example is lipodystrophy—a genetic disorder in which the amount of subcutaneous fat tissue and the ability to store fat is minimal, relying primarily on ectopic fat tissue, mainly associated with the storage of excess fatty acids in the liver [28]. In this syndrome, there are severe metabolic disorders despite a normal BMI [28]. Additionally, the quality of the diet, not just calorie intake, which affects BMI, plays a significant role in the development of metabolic disorders [25].

At the molecular level, environmental factors such as heavy metals, chronic stress, lack of physical activity or improper diet influence TL by increasing the levels of so-called stress hormones, promoting inflammation, and oxidative stress (Fig. 7) [29–34]. It has been shown that elevated levels of epinephrine, norepinephrine, and cortisol are associated with shorter TL [29, 35]. At the same time, stress hormones, particularly cortisol, promote the development of the main component of MetS—obesity (Fig. 7) [36]. HTN, another component of MetS, also has a documented link with chronic stress [37]. At the same time, HTN, through increasing oxidative stress, inducing inflammation, and causing mitochondrial dysfunction (especially in arterial endothelial cells), contributes to telomere shortening [38]. Similarly, hyperglycemia and diabetes, including even gestational diabetes, also increase oxidative stress, which can have further impact on telomeres. Dyslipidemia, particularly triglyceride levels, is strongly associated with telomere shortening through the induction of inflammation and free radicals [39]. In our previous work, we demonstrated that oxidative stress is particularly elevated in lipid disorders associated with atherogenic dyslipidemia—low HDL-C and high triglycerides [40], and to a lesser extent LDL-C levels. Finally, the last component of MetS, carbohydrate metabolism disorders, including both DM and prediabetes, contributes to the induction of inflammation and the generation of reactive oxygen species and the increased formation of advanced glycation end products (AGEs), ultimately leading to telomere shortening (Fig. 7). The impact of type 1 diabetes and the impact of gestational diabetes on the unborn appears broadly similar to the impact of T2D, highlighting the direct effects of hyperglycemia independent of broader metabolic dysfunction [41–45].

This study is a sub-analysis of the LIPIDOGRAM2015 study. Although patients were selected randomly, the clinical characteristics of the LIPIDOGRAM2015 and LIPIDOGEN2015 populations differ in terms of the percentage of individuals with secondary or higher education (which is higher in the LIPIDOGEN2015 substudy). However, patients were similar regarding other clinical characteristics. Another limitation stems from the nature of the study itself—namely, in a cross-sectional study, we are unable to assess causal relationships, only correlations. Moreover, telomere length was the only measured telomere-related parameter and telomerase activity was not assayed, which could have provided additional insights considering its implications in atherosclerosis and T2D [46]. Additionally, the exclusion of outliers may have introduced an inclusion bias, with an unclear effect on the results. Our survey methods for physical activity did not allow for an in-depth differentiation of different activity types and levels. Finally, the confounding impact of unsurveyed lifestyle or dietary factors cannot be excluded. For example, omega-3 fatty acids may influence both telomere length [47, 48] and lipid profile [49].

Conclusions

The pattern of clinical characteristics marked by classical cardiovascular risk factors including components of MetS, is inversely related to TL, underlining the potential role of metabolic disturbances in cellular aging.

Supplementary Information

Below is the link to the electronic supplementary material.

Author Contributions

Tadeusz Osadnik , Jacek Jóźwiak, Maciej Banach, Kamila Osadnik, Joanna Katarzyna Strzelczyk, and Natalia Pawlas designed the study. Jacek Jóźwiak, Maciej Banach, Marek Gierlotka, Anna Goc, Ewa Boniewska-Bernacka, and Anna Pańczyszyn contributed to the acquisition and interpretation of data. Tadeusz Osadnik performed the statistical analyses. Nikodem Baron and Martyna Fronczek wrote the manuscript. Tadeusz Osadnik, Nikodem Baron, Martyna Fronczek, and Marcin Goławski prepared figures. All authors reviewed the manuscript. All authors were involved in the interpretation of results and accept responsibility to submit for publication.

Funding

Analyzes of telomere length measurement were funded by the statutory agreements of the Medical University of Silesia in Katowice: PCN-1-113/N/1I. The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: The Polish Lipid Association (PoLA) and the College of Family Physicians in Poland (CFPIP) provided non-material support by endorsing the project. The present study was funded by an unrestricted educational grant from Valeant (Warsaw, Poland). Valeant had no role in the study design, data analysis, data interpretation, or writing of the report.

Data Availability

The data presented in this study are available on request from the corresponding author. The data are not publicly available due to the planned preparation of subsequent publications based on the collected dataset (data may be publicly available after the end of the project, currently only upon reasonable request).

Declarations

Ethics Approval and Consent to Participate

The LIPIDOGRAM2015 study & LIPIDOGEN2015 substudy received approval from the Bioethical Committee of the District Medical Chamber in Częstochowa (approval no. K.B.Cz.–0018/2015). All studies adhered to the principles outlined in the Declaration of Helsinki. Written informed consent was obtained from all participants, and measures were taken to ensure both anonymity and confidentiality.

Consent for Publication

Not applicable.