Targeting biological age with bioactive, microbiota-accessible nutritional complexes: a pilot study on healthspan extension in medically healthy adults Authors: Andrei Biţă, Adina Turcu-Ştiolică, Ion Romulus Scorei, Cătălina Gabriela Pisoschi, Cristina Elena Biţă, Gabriela Rău, Maria Viorica Ciocîlteu, Simona Ştefănescu, Laura Dincă, Johny Neamţu, Ion Rogoveanu, George Dan Mogoşanu, Dan Ionuţ Gheonea Published: 2025 ABSTRACT BackgroundMicrobiota-accessible nutritional complexes (MAC), a formulation comprising prebiotics, postbiotics, autophagy stimulators, senolytic activators, and natural probiotics, may influence systemic biomarkers and biological aging in healthy individuals. This pilot interventional study aimed to evaluate the effects of a 60-day MAC supplementation on serum biomarkers and biological age (BioAge) in medically healthy adults. Methods: Of 13 screened, 12 enrolled; 3 were excluded from the final analysis. Nine participants (five females, four males; mean age 61 ± 9.29 years) completed 60 days of daily MAC supplementation and were included in the analyses. Serum biomarkers were measured at baseline and post-intervention. BioAge was estimated using three machine-learning regressors: Support Vector Regression (SVR), Random Forest (RF), and eXtreme Gradient Boosting (XGBoost). Feature importance analysis was conducted to identify key predictors of BioAge. Results: No adverse events occurred. A significant reduction in high-sensitivity C-reactive protein (hs-CRP) levels was observed from 2.66 ± 4.65 to 0.84 ± 0.54 mg/L (-69%; p = 0.009; Cohen’s d ≈ 0.55; post-mean 95% CI: 1.44 to 5.10), indicating decreased systemic inflammation. Lactate dehydrogenase (LDH) also declined significantly from 171.11 ± 21.32 to 159.44 ± 26.86 U/L (-6.8%; p = 0.038; Cohen’s d ≈ 0.22; post-mean 95% CI: 0.97 to 22.37). Other biomarkers, including gamma-glutamyl transferase (GGT), alkaline phosphatase (ALP), and glucose, showed trends toward improvement without reaching statistical significance. Stratified analysis revealed that females experienced a significant reduction in hs-CRP (p = 0.043) and a mild increase in creatinine (p = 0.042), whereas males exhibited non-significant trends toward improved inflammatory and metabolic markers. AI modeling indicated a reduction in BioAge for several participants. The XGBoost model consistently captured moderate improvements (e.g., Participant 7: 3.3 years), while the RF model showed more variability. SVR did not detect significant changes. An independent empirical model confirmed a statistically significant reduction in BioAge post-intervention (p < 0.0001). Top predictors were low-density lipoprotein cholesterol (LDL-C), glucose, and total cholesterol (TC) as key predictors in the RF and SVR models, while ferritin and hs-CRP ranked highest in the XGBoost model. Conclusions: Sixty days of MAC was safe and associated with clinically relevant hs-CRP reductions and small LDH decreases, alongside AI-inferred BioAge improvements most stably detected by XGBoost. Limitations include small sample size (n = 9), single-arm design, 60-day duration, non-fasting sampling, and a multicomponent intervention that precludes mechanistic attribution, with no microbiome/postbiotic readouts. Larger randomized trials with microbiome/metabolomic profiling and pre-registered, externally validated AI pipelines are required to confirm causality. Trial registration: ISRCTN, ISRCTN85957759. Registered 04 February 2025, https://www.isrctn.com/ISRCTN85957759.Supplementary InformationThe online version contains supplementary material available at 10.1038/s41598-025-31590-1.Keywords: Biological age, Targeting, Microbiota-accessible nutritional complexes, Pilot study, Healthspan extension, Healthy adultsSubject terms: Biomarkers, Diseases, Medical research Background Healthy life expectancy denotes the period lived free of major morbidity, emphasizing functional longevity rather than survival alone; the World Health Organization’s HALE (“health-adjusted life expectancy at birth”) metric operationalizes this perspective1. Extending healthspan has the potential to significantly reduce healthcare costs for both individuals and society. This can be achieved through a combination of lifestyle modifications, such as optimized diet, physical activity, sleep hygiene, and stress management, as well as through targeted therapies and nutritional supplementation. Before the age of 75, the rate of biological aging accelerates, after which it tends to stabilize. Women generally live longer than men2. Compared to younger adults, the elderly experience a biological aging rate that is approximately 1.8 times faster3,4. Chronological age is the time elapsed since birth5, whereas biological age (BioAge) captures organismal functional decline and more closely predicts health and longevity6,7. Although chronological age is irreversible, BioAge can potentially be reversed through informed lifestyle and dietary choices8. Age acceleration—the deviation of BioAge from chronological age—has emerged as a clinically informative construct, yet its mechanistic drivers and modifiable levers remain incompletely defined. A broad toolkit has been used to estimate BioAge—telomere length, DNA methylation, proteo-/metabolomics, glycomics, wearable-derived metrics, and clinical blood biomarkers9. Among these, standard serum biomarkers can detect BioAge differences even in ostensibly healthy individuals and predict healthspan/lifespan via glycemic, lipid, hepatic, renal, and inflammatory axes10,11. However, key gaps persist: (i) whether a short-term, safe, nutrition-based intervention can shift BioAge in healthy adults; (ii) which biological pathways are most responsive; and (iii) how robust such inferences are across machine-learning (ML) models applied to clinical biomarkers. To address these gaps, we investigate a microbiota-accessible nutritional complex (MAC) - a multi-component formulation comprising prebiotics, postbiotics, autophagy stimulators, senolytic activators, and natural probiotics—selected for complementary mechanisms relevant to aging biology. Prebiotics foster short-chain-fatty-acid (SCFA) production and enrich beneficial taxa (e.g., Bifidobacterium, butyrogenic guilds), thereby lowering systemic inflammatory tone and supporting epithelial barrier integrity. Postbiotics (defined microbial metabolites/structural components) exert immunomodulatory and antioxidant effects that stabilize mucosal defense. Autophagy stimulators (e.g., polyamine-linked nutrients) promote proteostasis and mitochondrial turnover, processes tied to longevity signaling. Senolytic activators (e.g., selected polyphenols) target senescent-cell burden to mitigate pro-inflammatory secretory phenotypes. Probiotics contribute ecological stabilization and cross-kingdom signaling that support resilience of the gut ecosystem. Together, these domains converge on hallmarks of aging—chronic low-grade inflammation (“inflammaging”), impaired mitochondrial quality control, cellular senescence, and barrier fragility—providing a mechanistic rationale for testing MAC in healthy adults. This pilot study evaluates 60-day MAC supplementation in medically healthy adults, quantifying paired pre–post changes in standard serum biomarkers and estimating BioAge using three ML regressors (Support Vector Regression, Random Forest, eXtreme Gradient Boosting). By integrating inflammatory and metabolic endpoints with ML-based BioAge, the study tests the hypothesis that a microbiota-accessible, multi-pathway nutritional strategy can reduce systemic inflammation and produce measurable, model-consistent shifts in biological aging over a short timeframe, while explicitly acknowledging the pilot design and the need for larger randomized trials for confirmation and external validation. Methods Study design and conduct The trial design was approved by the Bioethics Board of the Bioboron Research Institute, S.C. Natural Research S.R.L., Podari, Romania and subsequently reviewed and authorized by the Institutional Ethics Committee of the University of Medicine and Pharmacy of Craiova, Romania (Approval No. 7/10.01.2025). This pilot study was registered in the UK Clinical Study Registry (identifier: ISRCTN85957759)12. All study procedures—including visits, informed consent and follow-up assessments—were conducted at the University of Medicine and Pharmacy of Craiova. The study adhered to the ethical principles outlined in the 1975 Declaration of Helsinki, as revised in 2008. It was additionally reviewed by the study sponsors: S.C. Natural Research S.R.L. and the University of Medicine and Pharmacy of Craiova. Participants were recruited on a voluntary basis from the general community in Craiova, with an equal sex ratio (1:1) among participants. The eligibility was assessed via telephone pre-screening, followed by an on-site screening visit. This single-center study was conducted at the Bioboron Research Institute, according to TREND checklist [Supplement 1]. In the study protocol, and provision of informed consent, full inclusion and exclusion criteria are available at: https://www.isrctn.com/ISRCTN8595775912. Thirteen volunteers, aged 45–72, were screened for eligibility and recruited. Enrollment, which included obtaining consent, took place from December 2024 to January 2025, with the study concluding in March 2025. Twelve participants initiated supplementation (six adult males and six adult females); however, one individual withdrew prior to baseline blood sampling and was not formally enrolled. Three participants were excluded from the final analysis due to pre-specified criteria or intercurrent illness: one male (multiple sclerosis), one female (rheumatoid arthritis), and one male (acute SARS-CoV-2 infection during intervention). Ultimately, nine participants completed all visits and were thus included in the per-protocol analysis. Volunteers were screened by medical history, physical examination, and baseline laboratory testing (CBC, CMP, hs-CRP). Inclusion required absence of acute or chronic illness and no chronic medication use. Exclusion criteria comprised cardiovascular disease, hypertension requiring treatment, diabetes or other metabolic disorders, hepatic or renal impairment, autoimmune/inflammatory diseases, active infection, pregnancy/lactation, psychiatric instability requiring medication, and use of prescription drugs, NSAIDs/corticosteroids, or dietary supplements within the preceding 14 days (washout). Participants committed to maintaining stable lifestyle (diet, activity) and to avoiding new medications/supplements during the study. Concomitant medication use was assessed at screening and at each visit by structured questionnaire and AE log. Prohibited during the 60-day intervention: prescription drugs, NSAIDs, corticosteroids, antihistamines, anticoagulants/antiplatelets, and any dietary supplements. Initiation of short-course therapy for intercurrent illness would have led to temporary suspension and documentation; no such cases occurred among analyzed participants. Among screened/enrolled individuals, no concomitant medication was reported during the intervention in the per-protocol cohort. At enrollment, participants received non-mandatory lifestyle recommendations to minimize extraneous variability during the 60-day intervention: a plant-forward diet emphasizing fruits/vegetables, seeds/cereals, fermented dairy, and marine fish/seafood; recommended beverages (water, unsweetened coffee/tea; optional ~ 150 mL red wine/week); and limits on red meat, sweets, and bread (with polenta/corn cakes as alternatives). Participants were encouraged to achieve ≥ 4,000 steps/day and approximately 150 Google Fit “cardio points”/week, and to use Google Fit to track steps and cardio points. They were asked to report step counts and body weight at day 60. As this was a pilot, adherence reporting was optional and incomplete; consequently, diet/activity data were not incorporated into inferential analyses. The participant notice is provided in Supplementary File S1. A Transparent Reporting of Evaluations with Nonrandomized Designs (TREND) flow diagram is shown in Fig. 1 and baseline characteristics are shown in Table 1. The age range of 45–72 was selected as a time when age-related vulnerabilities typically manifest (over 45 years)13,14. The analyzed cohort (n = 9) comprised healthy adults without chronic medication use; baseline CBC/CMP were within reference ranges and no pathological conditions were present. TREND flow-diagram of pilot study. TREND: transparent reporting of evaluations with nonrandomized designs. Table 1Baseline characteristics of study participants.No.Subject codeAge (years)Height (cm)Body weight (kg)BMI (kg/m2)1.M1601808225.312.M27118010632.723.F3541666723.674.F4551707224.915.F6581656624.246.M8681748327.417.F9671615822.388.F11431686723.749.M127218510630.97BMI: Body mass index. Baseline characteristics of study participants. BMI: Body mass index. The present clinical study involved the oral administration of bioactive MACs dietary supplement15, which included: organic boron from a boron-enriched green coffee extract (BE-GCE); 12 mg elemental B/day, calcium butyrate 200 mg, curcumin 50 mg, EGCG 50 mg, spermidine 5 mg, fisetin 10 mg, quercetin 50 mg, glutamine 200 mg, and a probiotic blend (10⁹ CFU/day: Lactobacillus plantarum, Streptococcus thermophilus, Bifidobacterium animalis, B. adolescentis, B. longum, Bacillus subtilis, and Bacillus coagulans). Elemental boron content in BE-GCE was quantified by ICP-based assays; no structural assignment to individual boron esters is claimed. We hypothesize that boron–polyphenol complexes (from BE-GCE), together with butyrate and autophagy/senolytic modulators, underlie the observed reductions in hs-CRP and estimated BioAge. Each supplement was taken once daily, either after breakfast or lunch, for 60 days. All blood samples were collected in the non-fasting state in red capped vacutainers without any anticoagulant and processed for serum separation. Samples were centrifuged at 3000 rpm for 10 min at room temperature, and serum aliquots were immediately frozen at − 80 °C until batch analysis. To minimize pre-analytical variation, all paired samples (baseline and day 60) for a given subject were analyzed in the same analytical run using calibrated reagents. Primary outcome measures We selected 12 blood-based biomarkers shown in scientific literature16–19 to be associated with healthy aging and longevity (Table 2). Those assessing the outcomes were blinded to study condition assignment. Laboratory analysts performing biochemical assays and the independent data scientist conducting machine-learning analyses were blinded to participant identity and timepoint (baseline vs. post-intervention). All datasets were pseudonymized before analysis, and model training/testing were performed on coded identifiers only. Table 2Blood biomarkers related to healthy aging and longevity.BiomarkerObservationsUAthe latest studies show that UA should not exceed 4.9 mg/dL (292 µmol/L) and not be less than 4.0 mg/dL (238 µmol/L) to ensure exceptional longevity over 100 years; a significant increase in the risk of cardiovascular and all-cause mortality was observed in men with serum UA levels ≥ 7.0 and < 3.0 mg/dL (≥ 416 and < 178 µmol/L, respectively) and in women with levels ≥ 5.0 and < 3.0 mg/dL (≥ 297 and < 178 µmol/L, respectively)20hs-CRPcompared with those with hs-CRP < 1.0 mg/L, individuals (men and women) with hs-CRP > 3.0 mg/L had a significantly higher risk of all-cause mortality21cholesterolboth in men and women, TC between 200 and 240 mg/dL (5.2 and 6.2 mmol/L, respectively), is a serious biomarker to consider for decreased risk of mortality after 65 years for healthy longevity; in addition, a LDL-C between 82 and 140 mg/dL (2.1 and 3.6 mmol/L, respectively) indicates a stable lipid profile for exceptional longevity22,23glucosea comparison between centenarians and non-centenarians for a long and healthy life showed that a blood glucose level between 80 and 110 mg/dL (4.4 and 6.1 mmol/L, respectively) is needed to reach over 100 years of age24biochemical liver testslow levels of GGT in the blood (< 30 U/L) were associated with a higher likelihood of becoming a centenarian; low serum levels of ALP activity (< 150 U/L) were associated with a higher likelihood of becoming a centenarian; low serum levels of LDH activity (< 200 U/L) have been shown to increase the odds of older people becoming centenarians25creatininecentenarians generally had low serum creatinine values, between 0.47 and 1.24 mg/dL (41.5 and 109.6 µmol/L, respectively)26total iron,ferritin,TIBC, and UIBCto increase the chances of being a centenarian, total blood iron reference range for exceptional longevity for men and women is 55–125 µg/dL (10–22.4 µmol/L); ferritin reference range for exceptional longevity for men and women is 20–100 ng/mL (4.2–21 µmol/L) and coupled with a decrease in TIBC level to range 50–65 µmol/L18,27ALP: Alkaline phosphatase; GGT: Gamma-glutamyl transferase; hs-CRP: High-sensitivity C-reactive protein; LDH: Lactate dehydrogenase; LDL-C: Low-density lipoprotein cholesterol; TC: Total cholesterol; TIBC: Total iron-binding capacity; UA: Uric acid; UIBC: Unsaturated iron-binding capacity. Blood biomarkers related to healthy aging and longevity. total iron, ferritin, TIBC, and UIBC ALP: Alkaline phosphatase; GGT: Gamma-glutamyl transferase; hs-CRP: High-sensitivity C-reactive protein; LDH: Lactate dehydrogenase; LDL-C: Low-density lipoprotein cholesterol; TC: Total cholesterol; TIBC: Total iron-binding capacity; UA: Uric acid; UIBC: Unsaturated iron-binding capacity. Serum levels of uric acid (UA), creatinine, high-sensitivity C-reactive protein (hs-CRP), lipid profiles (high-density lipoprotein cholesterol (HDL-C), low-density lipoprotein cholesterol (LDL-C), total cholesterol (TC), and triglycerides (TGs)), liver enzymes (gamma-glutamyl transferase (GGT), alkaline phosphatase (ALP), and lactate dehydrogenase (LDH)), glucose, total iron and total iron-binding capacity (TIBC) measured using blood tests at baseline and after 60 days. Biochemical parameters were analyzed using an Abbott Alinity C chemistry analyzer (Diamond Diagnostics Inc., Holliston, MA, USA) provided with standardized enzymatic methods dedicated to the quantitative detection of each analyte in biological fluids (serum, plasma, urine), the results being compared to the reference values considered as follows: UA reference values, 2.5–6.2 mg/dL for women and 3.7–7.7 mg/dL for men; creatinine reference values, 0.55–1.02 mg/dL for women and 0.73–1.18 mg/dL for men; TC reference values, 70–240 mg/dL; TGs reference values, < 150 mg/dL; GGT reference values, 0–38 U/L for women and 0–55 U/L for men; ALP reference values, 46–122 U/L for women and 50–116 U/L for men; LDH reference values, 125–220 U/L; glucose reference values for the glucose oxidase–peroxidase (GOD–POD) method, 70–110 mg/dL. hs-CRP, an inflammatory biomarker, was assessed using CRP Vario Reagent kit (Sentinel Diagnostics, Milano, Italy) based on a quantitative immunoturbidimetric determination of CRP with variable assay ranges – CRP48 and CRP16 (hs-CRP) – and reference values between 0 and 5 mg/L for CRP and < 1 mg/L for hs-CRP. Included in the lipid profile, HDL-C was assessed using Ultra HDL assay, a homogeneous method for directly measuring of HDL-C in serum or plasma without the need for off-line pretreatment or centrifugation steps, accomplished using finally a spectrophotometric method with reference values between 55 and 100 mg/dL for women and 50–100 mg/dL for men. LDL-C was calculated using the Eq. (1) (Friedewald formula): Reference values fall between 60 and 100 mg/dL, very-low-density lipoprotein cholesterol (VLDL-C) being calculated from the TGs level divided by 5, if TGs are less than 400 mg/dL. Serum iron level was determined using a colorimetric method based on complexing iron with ferrozine and dedicated to the quantitation of iron in serum or plasma, reference values being 50–170 µg/dL for women and 65–175 µg/dL for men. Ferritin was quantified in human serum and plasma using electrochemiluminescence immunoassay performed on a Cobas 6000 platform (Roche Diagnostics, Rotkreuz, Switzerland); reference values: 13–150 ng/mL for women and 30–400 ng/mL for men. TIBC was calculated using the Eq. (2): In order to apply this formula, transferrin was measured at baseline and after 60 days using an immunoturbidimetric method dedicated to the quantitation in serum or plasma, with reference values 50–65 µmol/L. Unsaturated iron-binding capacity (UIBC), with reference values 21–84 µmoles/L, representing the amount of transferrin available to bind more iron, was calculated with the following Eq. (3): Secondary outcome measures Adverse events, as secondary outcome measures, were assessed during the 60 days. Adverse events were defined using the Romanian National Agency for Medicines and Medical Devices (ANMDM) Guidelines and measured using data collected from the patient card for spontaneous reporting of adverse drug reactions28. Statistical analysis Continuous variables were presented as mean ± standard deviation (SD) and median with interquartile range (IQR), as appropriate. For all statistical comparisons, the results convey the p-value and the direction of the effect of MACs dietary supplement after 60 days of treatment. We reported effect sizes (Cohen’s d and % change) for key biomarkers to give practical meaning beyond p-values. Machine learning models for biological age Estimation To estimate BioAge from clinical biomarkers, we developed a predictive model using different machine learning (ML) algorithms: Random Forest (RF), eXtreme Gradient Boosting (XGBoost), and Support Vector Regression (SVR)29. The underlying hypothesis was that individuals whose biomarker profiles closely resembled those of centenarians would exhibit a lower BA, indicating more favorable aging trajectories. A higher BioAge reflects a higher risk of illness, incapacity, and death, while a smaller value reflects the opposite. A total of 12 serum biomarkers were selected based on their established relevance to aging and systemic health, including markers of liver function (e.g., ALP, GGT), inflammation (hs-CRP), metabolism (glucose, cholesterol, LDL-C), and iron homeostasis (iron, ferritin, TIBC, UIBC). Centenarian reference ranges were obtained from published literature and used as normative aging targets. Data were stratified by sex due to sex-specific biomarker distributions, and standardization (z-score normalization) was applied within each subgroup. For each participant, a Mahalanobis distance was calculated relative to the centenarian profile, accounting for inter-variable correlations via the covariance matrix30,31. To invert the scale (i.e., lower distances indicating better biological similarity), we multiplied the Mahalanobis distance by − 1, generating an “inverted score” where higher values indicate greater proximity to centenarian health profiles. For a more precise distance from centenarian health status, we calculated the Mahalanobis distance instead of simple normalized deviations. We did not use linear regression because biomarkers are likely non-linear with age, and we used RF regressor from sklearn.ensemble. We employed an RF regressor (100 estimators, random_state = 42) using the inverted score as the primary predictor of chronological age. Model validation followed a leave-one-out cross-validation (LOOCV) scheme appropriate for small sample size, whereas, for robustness checks, fivefold cross-validation was also performed. The model was trained to learn the mapping from biomarker similarity to centenarian profiles to age, then used to generate predicted BioAge values for everyone. The difference Δ between baseline BioAge and BioAge after 60 days of treatment. This estimate reflects the age that best corresponds to their biomarker profile under the learned model. To explore which biomarkers contributed most to the BioAge prediction, we retrained the RF model using all standardized biomarkers as input features. Besides RF, we explored more artificial intelligence (AI) models in order to compare their performance and robustness: XGBoost, and SVR. The strength of SVR is that it is good for small datasets like ours. Feature importance scores were extracted based on the mean decrease in impurity across all trees. These scores quantify the relative contribution of each biomarker to the model’s accuracy and are presented as a normalized importance scale. To compare the biomarkers, BioAge baseline vs. after 60 days of treatment, we performed Wilcoxon Signed-Rank test (small sample size and/or non-normal distribution). We used Python 3.11, and the libraries scikit-learn 1.2, scipy.spatial, pandas, numpy, xgboost, sklearn.preprocessing, sklearn.svm, matplotlib. The level of p-value of 0.05 was considered statistically significant. Secondary method for biological age estimation: longevity index model A complementary method was implemented to provide a simple, interpretable estimation of the BioAge. This model was designed for practical clinical use and is implemented in a standalone Excel tool. Each of the 12 selected biomarkers received a Longevity Score ranging from 0 to 1: (i) score 1 – biomarker within the optimal reference range; (ii) score 0 – biomarker outside the reference range; (iii) intermediate values were assigned proportionally based on deviation from the optimal range. The Longevity Index is the average of all Longevity Scores (biomarker scores). This index ranged from 0 (poor alignment with centenarian profiles) to 1 (perfect alignment). BioAge was then estimated using the linear Eq. (4): It ranges from 0 (poor alignment with longevity profiles) to 1 (perfect alignment). For example: (i) score 1.0 → BioAge = 50 (excellent profile); (ii) score 0.5 → BioAge = 75 (intermediate profile); (iii) score 0.0 → BioAge = 100 (highest risk). The equation was empirically derived to maintain a 1:1 mapping between normalized biomarker alignment and a 50-year biological aging span (50–100 years range). This simple deterministic model provides transparency complementary to the AI-based estimators. The tool is implemented entirely in Excel using built-in formulas, requiring no programming or internet access. Users can input participant biomarker values to instantly receive a Longevity Index and corresponding BioAge estimate. This model assumes as limitations equal weight for each biomarker and a linear relationship between biomarker alignment and BioAge. It does not incorporate genetics, lifestyle, or clinical history data. Nevertheless, it serves as a transparent and practical framework for evaluating BioAge in clinical settings. This pilot study was designed to generate preliminary effect-size estimates rather than to test specific hypotheses. The sample size aligns with accepted pilot thresholds (8–12 completers) for feasibility and variance estimation, as recommended32. These data support planning of future randomized controlled trials with powered group comparisons. Results Biomarkers We included nine participants (five females and four males), the mean age ± SD was 61 ± 9.29 years. The effect of the 60-day MACs administration on serum biomarker levels was evaluated by comparing baseline and post-treatment values after 60 days. No adverse reactions were observed. UA levels showed an increase from 4.48 ± 1.05 mg/dL at baseline to 4.57 ± 0.76 mg/dL after treatment (p = 0.678), as shown in Supplementary File S2a. Although not statistically significant, the clinical interpretation of this change may require context-dependent evaluation, as UA is influenced by both oxidative defense and metabolic imbalance. hs-CRP demonstrated a marked and statistically significant reduction, from 2.66 ± 4.65 to 0.84 ± 0.54 mg/L (p = 0.009), representing a 69% decrease from baseline (Cohen’s d = 0.55, a medium effect, post-mean 95% CI: 1.44 to 5.10). This suggests a notable decrease in systemic inflammation, which is consistent with a favorable biological aging profile. The violin plots from Fig. 2a visually support this pronounced decrease, with most post-treatment values clustered near the lower end of the scale. The mean hs-CRP levels dropped substantially, representing approx. 68% reduction. Comparison of biomarkers between baseline and after 60 days of treatment: (a) hs-CRP; (b) LDH. hs-CRP: High-sensitivity C-reactive protein; LDH: Lactate dehydrogenase. TC levels decreased from 205.44 ± 60 to 193.33 ± 49.02 mg/dL (p = 0.81), as shown in Supplementary File S2b, indicating a potential improvement in lipid metabolism and cardiovascular risk, even if non-significant. No statistically significant changes were observed in LDL-C (125.78 ± 54.23 vs. 120.43 ± 52.34 mg/dL; p = 0.477), glucose (98.22 ± 21.69 vs. 96.67 ± 11.01 mg/dL; p = 0.722), GGT (25.4 ± 7.08 vs. 24.32 ± 7.18 U/L; p = 0.287), ALP (81.26 ± 36.17 vs. 76.26 ± 36.6 U/L; p = 0.138), creatinine (0.81 ± 0.24 vs. 0.83 ± 0.22 mg/dL; p = 0.258), ferritin (131.2 ± 74.98 vs. 116.04 ± 63.29 ng/mL; p = 0.066), TIBC (43.96 ± 6.17 vs. 46.58 ± 16.27 µmol/L; p = 0.72), UIBC (43.96 ± 7.85 vs. 46.58 ± 7.94 µmol/L; p = 0.678). The violin plots (Supplementary File S2, c–j) show considerable overlap in distributions and no clear trend in individual subject trajectories (lines connect paired samples). Although these biomarkers did not reach statistical significance, some showed trends toward improvement (e.g., reductions in GGT, ALP, and glucose), suggesting a potential biological response that may not have been detectable in this small sample size. Statistically significant changes were observed in LDH (171.11 ± 21.32 vs. 159.44 ± 26.86 U/L; p = 0.038), representing a 9% decrease from baseline (Cohen’s d = 0.22, a small effect), as shown in Fig. 2b. Results differences among females and males are summarized in Table 3 and presented as mean ± SD and median (IQR), with corresponding p-values for within-subject sex comparisons. We have split the table into "Inflammation/Metabolism" and "Iron/Liver" groups. Table 3Characteristics of the participants and biomarker changes following 60 days of treatment.BiomarkerSexBaselineAfter 60 days of treatmentp-value*Inflammation/Metabolism UA (mg/dL)M5.21 ± 1.155.12 (4.18–6.34)5.19 ± 0.735.30 (4.45–5.83)0.854F3.90 ± 0.523.90 (3.45–4.35)4.08 ± 0.284.20 (3.85–4.25)0.588 hs-CRP (mg/L)M4.86 ± 6.761.93 (0.79–11.86)1.09 ± 0.631.07 (0.50–1.71)0.068F0.91 ± 0.50.78 (0.45–1.43)0.63 ± 0.400.74 (0.22–0.99) 0.043 TC (mg/dL)M168.50 ± 31.67175.50 (136.75–193.25)172.00 ± 48.15171.50 (125.75–218.75)0.715F235.00 ± 63.19241.00 (178.50–288.50)210.40 ± 47.41210.00 (168.50–252.50)0.345 LDL-C (mg/dL)M95.85 ± 30.8196.50 (65.85–125.20)98.93 ± 41.2992.80 (62.73–141.25)1.000F148.64 ± 57.53140.00 (96.90–204.70)126.96 ± 48.21129.00 (86.10–166.80)0.279 Glucose (mg/dL)M112.25 ± 19.69104.50 (99.25–133.00)103.75 ± 9.11104.50 (95.00–111.75)0.273F87.00 ± 17.2081.00 (77.00–100.00)91.00 ± 9.49129.00 (86.10–166.80)0.588 LDH (U/L)M163.25 ± 26.25162.00 (138.50–189.25)148.25 ± 26.55139.00 (129.75–176.00)0.068F177.40 ± 16.77180.00 (161.50–192.00)168.40 ± 26.25164.00 (146.00–193.00)0.225 Creatinine (mg/dL)M1.02 ± 0.131.04 (0.89–1.13)1.01 ± 0.140.99 (0.88–1.15)0.461F0.64 ± 0.150.66 (0.50–0.76)0.68 ± 0.160.74 (0.54–0.81) 0.042 Iron/Liver ALP (U/L)M61.83 ± 14.9567.65 (46.08–71.75)53.58 ± 12.2956.15 (40.75–63.83)0.068F96.80 ± 42.06101.00 (57.50–134.00)94.40 ± 40.5078.00 (63.50–133.50)0.893 GGT (U/L)M26.90 ± 7.2625.30 (21.00–34.40)25.48 ± 9.0521.45 (20.23–34.75)0.461F24.20 ± 7.5323.00 (17.50–31.50)23.40 ± 6.2722.00 (18.00–29.50)0.194 Ferritin (ng/mL)M180.03 ± 76.43166.50 (114.63–258.95)153.00 ± 72.11151.20 (85.06–222.75)0.144F92.14 ± 50.7167.65 (50.68–145.85)86.47 ± 40.6588.33 (48.65–123.35)0.345 TIBC (µmol/L)M65.57 ± 3.9566.25 (61.49–68.98)64.68 ± 4.9965.25 (59.66–69.13)1.000F63.24 ± 4.8161.74 (59.72–67.50)63.45 ± 4.7860.73 (60.09–68.16)0.893 UIBC (µmol/L)M46.46 ± 6.9743.92 (41.65–53.80)43.81 ± 7.3246.89 (36.27–48.28)0.465F41.96 ± 8.6947.82 (32.48–48.52)48.80 ± 8.4950.76 (40.57–56.05)0.225*Wilcoxon Signed Rank test; bold values of p-value are statistically significant. ALP: Alkaline phosphatase; F: Female; GGT: Gamma-glutamyl transferase; hs-CRP: High-sensitivity C-reactive protein; LDH: Lactate dehydrogenase; LDL-C: Low-density lipoprotein cholesterol; M: Male; TC: Total cholesterol; TIBC: Total iron-binding capacity; UA: Uric acid; UIBC: Unsaturated iron-binding capacity. Characteristics of the participants and biomarker changes following 60 days of treatment. 5.21 ± 1.15 5.12 (4.18–6.34) 5.19 ± 0.73 5.30 (4.45–5.83) 3.90 ± 0.52 3.90 (3.45–4.35) 4.08 ± 0.28 4.20 (3.85–4.25) 4.86 ± 6.76 1.93 (0.79–11.86) 1.09 ± 0.63 1.07 (0.50–1.71) 0.91 ± 0.5 0.78 (0.45–1.43) 0.63 ± 0.40 0.74 (0.22–0.99) 168.50 ± 31.67 175.50 (136.75–193.25) 172.00 ± 48.15 171.50 (125.75–218.75) 235.00 ± 63.19 241.00 (178.50–288.50) 210.40 ± 47.41 210.00 (168.50–252.50) 95.85 ± 30.81 96.50 (65.85–125.20) 98.93 ± 41.29 92.80 (62.73–141.25) 148.64 ± 57.53 140.00 (96.90–204.70) 126.96 ± 48.21 129.00 (86.10–166.80) 112.25 ± 19.69 104.50 (99.25–133.00) 103.75 ± 9.11 104.50 (95.00–111.75) 87.00 ± 17.20 81.00 (77.00–100.00) 91.00 ± 9.49 129.00 (86.10–166.80) 163.25 ± 26.25 162.00 (138.50–189.25) 148.25 ± 26.55 139.00 (129.75–176.00) 177.40 ± 16.77 180.00 (161.50–192.00) 168.40 ± 26.25 164.00 (146.00–193.00) 1.02 ± 0.13 1.04 (0.89–1.13) 1.01 ± 0.14 0.99 (0.88–1.15) 0.64 ± 0.15 0.66 (0.50–0.76) 0.68 ± 0.16 0.74 (0.54–0.81) 61.83 ± 14.95 67.65 (46.08–71.75) 53.58 ± 12.29 56.15 (40.75–63.83) 96.80 ± 42.06 101.00 (57.50–134.00) 94.40 ± 40.50 78.00 (63.50–133.50) 26.90 ± 7.26 25.30 (21.00–34.40) 25.48 ± 9.05 21.45 (20.23–34.75) 24.20 ± 7.53 23.00 (17.50–31.50) 23.40 ± 6.27 22.00 (18.00–29.50) 180.03 ± 76.43 166.50 (114.63–258.95) 153.00 ± 72.11 151.20 (85.06–222.75) 92.14 ± 50.71 67.65 (50.68–145.85) 86.47 ± 40.65 88.33 (48.65–123.35) 65.57 ± 3.95 66.25 (61.49–68.98) 64.68 ± 4.99 65.25 (59.66–69.13) 63.24 ± 4.81 61.74 (59.72–67.50) 63.45 ± 4.78 60.73 (60.09–68.16) 46.46 ± 6.97 43.92 (41.65–53.80) 43.81 ± 7.32 46.89 (36.27–48.28) 41.96 ± 8.69 47.82 (32.48–48.52) 48.80 ± 8.49 50.76 (40.57–56.05) *Wilcoxon Signed Rank test; bold values of p-value are statistically significant. ALP: Alkaline phosphatase; F: Female; GGT: Gamma-glutamyl transferase; hs-CRP: High-sensitivity C-reactive protein; LDH: Lactate dehydrogenase; LDL-C: Low-density lipoprotein cholesterol; M: Male; TC: Total cholesterol; TIBC: Total iron-binding capacity; UA: Uric acid; UIBC: Unsaturated iron-binding capacity. While most biomarkers remained stable in males, females exhibited a significant decrease in hs-CRP (baseline: 0.91 ± 0.5 mg/L vs. post-treatment: 0.63 ± 0.4 mg/L; p = 0.043) and an increase in creatinine (baseline: 0.64 ± 0.15 mg/dL vs. post-treatment: 0.68 ± 0.16 mg/dL; p = 0.042). Males showed a non-significant trend toward decreased hs-CRP (p = 0.068), LDH (p = 0.068), and ALP (p = 0.068). Notably, baseline UA levels were higher in males (5.21 ± 1.15 mg/dL) than females (3.9 ± 0.52 mg/dL), but neither sex experienced significant changes post-treatment (p = 0.854 and p = 0.588, respectively). Lipid profiles (TC, LDL-C) and other metabolic markers (glucose, GGT, ferritin, UIBC) did not differ significantly within either sex over time (all p > 0.05). Biological age measured by AI modeling We used three different AI algorithms to predict the BioAge. Figure 3(a–c) exhibits BioAge changes from baseline to 60 days using SVR model, RF model and XGBoost model. The biological age changes from baseline to 60 days using (a) SVR, (b) RF and (c) XGBoost models. Dotted line represents the chronological age reference. RF: Random Forest; SVR: Support Vector Regression; XGBoost: eXtreme Gradient Boosting. The SVR model predicted no change in BioAge after 60 days: the chronological age reference lines are present, but the model doesn’t differentiate between baseline and follow-up. The SVR model likely underfit the data or was not sensitive enough to detect subtle shifts between timepoints. The RF model observed significant variability in change in BioAge across participants: some show marked improvements (e.g., Participant 4: Δ=−5.1 years, Participant 7: Δ=−4.2 years), while others show deterioration (e.g., Participant 1: Δ=+9.5 years). We observed good alignment with chronological age for some participants. RF captures non-linear patterns and interactions, making it responsive to change. However, the large positive Δ in some participants might suggest overfitting or lack of regularization. Regarding XGBoost model, we observed more balanced and moderate changes in BioAge, several participants improved (e.g., Participant 7: Δ=−3.3 years, Participant 9: Δ=−3.1 years), while others had minor increases. Δ values are generally smaller than RF, suggesting more conservative estimates. XGBoost tends to offer better generalization and control over model complexity. Its predictions are more robust and less volatile than RF, but more responsive than SVR. Biological age measured by second method In most cases, the BioAge at baseline is significantly higher than actual chronological age (p < 0.05). After 60 days, all participants show a statistically significant reduction in BioAge (p < 0.0001), indicating a potential rejuvenating or health-optimizing effect of the treatment, as shown in Fig. 4a. The BioAge after treatment was closer to actual age (p = 0.063), as in Fig. 4b. The chronological age (CAge) vs. biological age (BioAge) changes from baseline to 60 days (a). The BioAge changes from baseline to 60 days using an empirical formula (b). While both sexes experienced improvement, females may exhibit a greater reduction of BioAge — though this observation needs to be verified in a larger, sex-balanced sample for statistical validity. Biomarkers’ impact on biological aging We assessed the biomarker importance for predicting BioAge using the three AI models and the results were presented in Fig. 5(a–c). A comparative analysis of the biomarker importance across the three models (SVR, RF, and XGBoost) for predicting BioAge was done. We observed LDL-C appears highly important in all three models (top 4 in each), glucose ranks in the top 5 across all models, and TC is top 3 in both RF and XGBoost. The model-specific patterns were observed for SVR (linear kernel): it showed that UA as most important (2.5 score), ALP and glucose next most important, very high relative important scores (0–2.5 range), and likely capturing linear relationships only. The model-specific patterns were observed for RF: TC dominates (0.356 score), LDL-C second most important (0.198 score), more balanced distribution of importance, capturing non-linear interactions well. The model-specific patterns were observed for XGBoost: ferritin emerges as most important (0.30 score), LDL-C and hs-CRP next, includes “sex” as a feature (other models didn’t), showing finer gradations in importance scores. Biomarker importance for predicting biological age using (a) SVR, (b) RF and (c) XGBoost models. ALP: Alkaline phosphatase; GGT: Gamma-glutamyl transferase; hs-CRP: High-sensitivity C-reactive protein; LDH: Lactate dehydrogenase; LDL-C: Low-density lipoprotein cholesterol; RF: Random Forest; TC: Total cholesterol; TIBC: Total iron-binding capacity; SVR: Support Vector Regression; UA: Uric acid; UIBC: Unsaturated iron-binding capacity; XGBoost: eXtreme Gradient Boosting. We observed some notable differences: UA was top in SVR but bottom in XGBoost, ferritin was only top ranked in XGBoost, hs-CRP was much more important in XGBoost vs. others, and ALP was important in SRV, but less in others. Discussion Our pilot study evaluated the effects of a 60-day MACs treatment on biomarkers of aging, inflammation, metabolism, and AI-modeled BioAge in nine participants. The results highlight three major themes: (i) a significant anti-inflammatory effect (reduced hs-CRP), (ii) neutral or modest metabolic impacts, and (iii) divergent BioAge estimates across AI models, with evidence of potential rejuvenation. The most striking result was the 68% reduction in hs-CRP (p = 0.009), particularly in females (p = 0.043). This aligns with evidence that hs-CRP, a marker of systemic inflammation, correlates with biological aging and age-related diseases (e.g., cardiovascular disease)33. The magnitude of reduction (~ 1.82 mg/L) is clinically meaningful, as some studies demonstrated reduction in hs-CRP is associated with decreased incidence of cardiovascular events risk34. The sex-specific response (stronger in females) may reflect hormonal influences on inflammation35, though larger studies are needed to confirm this. Despite favorable trends (e.g., TC decreased by ~ 12 mg/dL), no significant changes were observed in lipids (TC, LDL-C), glucose, or liver enzymes (GGT, ALP). This suggests the treatment’s primary mechanism may not involve metabolic modulation, at least within this short timeframe. However, the trend toward reduced ferritin (p = 0.066) hints at improved iron metabolism, potentially linked to lower inflammation36. LDH decrease (p = 0.038) could reflect attenuated cellular stress, though its clinical relevance requires further study. Biomarker importance analysis underscored LDL-C, glucose, and hs-CRP as key predictors, aligning with known aging biomarkers29. Ferritin’s prominence in XGBoost (but not other models) highlights the value of multi-model approaches. Focusing on biomarkers that appear important across multiple models, our results recommend LDL-C and TC consistently important, suggesting that lipid metabolism is strongly associated with BioAge. Glucose importance aligns with known aging pathways (insulin resistance). hs-CRP’s prominence in XGBoost suggests non-linear inflammatory effects that tree models capture better. Ferritin’s top position in XGBoost may indicate iron homeostasis is important in aging. It may indicate threshold effects because it shows up strongly only in XGBoost. Next steps include validating top biomarkers in independent cohorts, examining interaction effects between top biomarkers and consider combining models in an ensemble approach. The differences highlight how model choice affects biological interpretation – linear models (SVR) emphasize different aspects than tree-based methods. The consistency in lipid markers across models suggests they’re robust predictors of BioAge. The AI models yielded divergent but complementary insights regarding the BioAge estimation. XGBoost and RF suggested BioAge reduction in some participants (e.g., Δ=−3 to − 5 years), consistent with hs-CRP improvements. Variability (e.g., Participant 1: Δ=+9.5 years in RF) may reflect overfitting or heterogeneous responses. The empirical formula showed uniform reductions (p < 0.0001), supporting a systemic anti-aging effect. Although this pilot was not designed for causal microbiome inference, the MAC formulation was intended to shift the intestinal ecosystem toward an anti-inflammatory, barrier-supportive, and metabolically efficient state—conditions predicted to lower hs-CRP and improve composite biological-age estimates. The MAC components collectively target these axes: BE-GCE (polyphenols) as microbiota-accessible substrates, calcium butyrate as a direct SCFA source, direct supporting of epithelial energy needs and anti-inflammatory signaling, and probiotics with anti-inflammatory/metabolic resilience profiles. Polyphenol/senolytic/ autophagy modulators (curcumin, EGCG, fisetin, quercetin, spermidine) are involved in modulation of oxidative stress and clearance of senescent cells—lowering the inflammatory “noise” that feeds into BioAge and reducing oxidative stress and senescent-cell burden. Interindividual variation in metabolic rate and substrate utilization can modulate AI-estimated biological age by altering biomarker turnover, mitochondrial redox balance, and inflammatory tone. Inefficient energy metabolism produces greater oxidative by-products and low-grade inflammation, which can bias composite aging scores upward, whereas efficient oxidative metabolism tends to reduce inflammatory load. Within this framework, MACs may influence BioAge indirectly. Butyrate supports β-oxidation and improves redox/epigenetic signaling (HDAC inhibition; GPR41/43/109A), polyphenols (curcumin, EGCG, fisetin, quercetin) modulate mitochondrial biogenesis/efficiency and oxidative stress, spermidine promotes organelle quality control, and microbiota-dependent bile-acid signaling (FXR/TGR5) enhances metabolic homeostasis. Collectively, these pathways could lower hs-CRP and other downstream biomarkers that contribute to BioAge. Notably, the present pilot did not measure metabolic rate; therefore, these mechanisms remain hypotheses to be evaluated in larger, mechanism-focused trials that include indirect calorimetry and mitochondrial proxies. In this pilot of healthy adults, we interpret “healthspan improvement” at the biomarker level: a shift toward lower systemic inflammatory load and higher physiological resilience, operationalized by reductions in hs-CRP and LDH, together with a decrease in AI-estimated BioAge. These markers correlate with morbidity, mortality, and healthspan in longitudinal cohorts, but in a short, single-arm study they remain surrogate indicators. The pre–post design, stable lifestyle instructions, and absence of concomitant medications/supplements limit confounding and support attribution of the observed within-subject changes to the MAC intervention. Nevertheless, we refrain from claiming clinical extension of healthspan; rather, we report a biomarker profile consistent with improved healthspan that warrants confirmation in randomized trials with functional endpoints (e.g., PROMs, physical performance, resilience to standardized perturbations), microbiome/postbiotic assays, and metabolic readouts. We identified several limitations in this study. One is related to the small sample size (n = 9) and the other to the short duration of our pilot study. The small sample size limits statistical power and generalizability, whereas longer treatment might reveal delayed metabolic effects. The study did not include direct assessments of energy metabolism (e.g., indirect calorimetry, respiratory quotient) or mitochondrial function; thus, any effect of MACs on metabolic rate remains inferential. Future trials will incorporate resting metabolic rate, substrate oxidation indices, and mitochondrial proxies to test whether metabolic changes mediate BioAge reductions. The study did not include stool microbiome profiling or postbiotic measurements (SCFAs, bile acids), limiting causal inference for the microbiota pathway. Future randomized trials will incorporate metagenomics and targeted metabolomics to test whether MAC-induced microbiome shifts mediate changes in hs-CRP and BioAge. Also, discrepancies in BioAge estimates (e.g., SVR vs. RF) emphasize the need for standardized AI aging metrics. The study did not include functional healthspan endpoints (e.g., PROMs, physical performance, resilience tests) and was not powered to assess clinical events; thus, healthspan inference is restricted to biomarker-level signals (hs-CRP, LDH, BioAge). The single-arm design precludes definitive causal attribution. Future randomized trials will incorporate functional and patient-centered outcomes, microbiome/postbiotic profiling, and metabolic assessments to test whether biomarker changes translate into clinically meaningful healthspan. Our results have clinical and research implications. It could be suggested that it is a targeted anti-inflammatory therapy: the hs-CRP reduction supports further study in conditions driven by inflammation (e.g., sarcopenia, atherosclerosis). On the other hand, personalized aging interventions could be recommended due to the sex differences (e.g., female hs-CRP response) and warrant exploration in larger cohorts. Also, while promising, BioAge models require validation against longitudinal outcomes. In this small pilot, among all three AI models, XGBoost yielded the most stable estimates, but this requires confirmation in larger datasets. SVR failed to capture any meaningful biological change, and RF captured significant changes but may require better tuning or feature selection to reduce potential overfitting. Conclusions This pilot study provides preliminary evidence that microbiota-accessible nutritional complexes (MACs) may reduce systemic inflammation and modestly influence biological age trajectories in healthy adults. The consistent decrease in hs-CRP and the AI-inferred reductions in BioAge suggest that targeted nutritional interventions acting on the gut–metabolic axis could modulate early biomarkers of aging. From a translational standpoint, these findings highlight the potential of MAC formulations as candidates for next-generation, microbiome-informed nutritional therapeutics aimed at promoting healthspan. However, such applications remain exploratory at this stage. Clinical translation will require large-scale randomized trials to replicate the effects observed here, establish dose–response relationships, and verify safety and efficacy under real-world conditions. To ensure responsible progression toward standardized anti-aging interventions, it is crucial to develop consensus-based frameworks integrating nutritional science, microbiome profiling, and systems biology modeling. Regulatory recognition of aging as a modifiable physiological process may further accelerate the transition from experimental formulations to validated clinical applications. Our results should therefore be interpreted as hypothesis-generating rather than prescriptive. No actionable clinical recommendations can be drawn from this small pilot cohort. Nevertheless, the multi-component, microbiota-accessible design of the MAC intervention offers a promising direction for nutritional geroscience. Finally, this study underscores the emerging role of AI in modeling biological aging. While AI models such as XGBoost and RF provided valuable interpretive power for biomarker integration, challenges remain regarding sample size requirements, feature stability, and external validation. As datasets expand and standardized pipelines emerge, AI-driven age modeling will increasingly complement traditional clinical and molecular metrics, refining our understanding of how interventions can truly modulate biological aging. Supplementary Information Below is the link to the electronic supplementary material. Supplementary Material 1 Supplementary Material 2 Supplementary Material 3 Supplementary Material 4 Acknowledgements AI tools were used to improve the English of the text. The authors have reviewed and take full responsibility for the content of this publication. Author contributions Conceptualization, A.B., A.T.-Ş. and I.R.S.; methodology, I.R.S., C.G.P. and D.I.G.; software, A.B., A.T.-Ş. and G.D.M.; validation, I.R.S., C.E.B. and D.I.G.; formal analysis, A.B., A.T.-Ş., I.R.S. and I.R.; investigation, C.G.P., G.R., M.V.C., S.Ş. and L.D.; resources, I.R.S., J.N. and D.I.G.; data curation, C.E.B., S.Ş., L.D. and G.D.M.; writing—original draft preparation, A.B., A.T.-Ş., I.R.S. and G.D.M.; writing—review and editing, A.B., I.R.S. and G.D.M.; visualization, C.G.P., G.R. and M.V.C.; supervision, I.R.S., A.T.-Ş., J.N. and I.R. All authors have read and agreed to the published version of the manuscript. Funding The Article Processing Charges were funded by the University of Medicine and Pharmacy of Craiova, Romania. Data availability The original contributions presented in this study are included in the article. Further inquiries can be directed to the corresponding author. Declarations Ethics approval and consent to participate The study was conducted in accordance with the Declaration of Helsinki and approved by the Ethics Committee of the University of Medicine and Pharmacy of Craiova, Romania (protocol code No. 7 from 10/01/2025). Informed consent was obtained from all subjects involved in the study. Consent for publication Not applicable. Competing interests The authors declare no competing interests. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. References Associated Data Supplementary Materials Supplementary Material 1 Supplementary Material 2 Supplementary Material 3 Supplementary Material 4 Data Availability Statement The original contributions presented in this study are included in the article. Further inquiries can be directed to the corresponding author.