Modelling of biological age in stable and acute exacerbations of chronic obstructive pulmonary disease

Abstract

Background

Aging has been established as an independent risk factor for chronic obstructive pulmonary disease (COPD). Biological age (BA), a novel metric for gauging the extent of aging, has rarely been investigated in the context of acute exacerbation of COPD (AECOPD). Our study aimed to elucidate the association between BA and AECOPD, thereby highlighting the potential of BA as a predictive tool in clinical practice.

Methods

The dataset encompasses patients hospitalized at Chengdu Third People's Hospital between 2018 and 2022. The AECOPD patients enrolled in this study were hospitalized due to rapidly worsening symptoms, including cough, sputum production, and dyspnea, whereas the COPD patients were clinically stable. BA and biological age acceleration were ascertained through the Klemera-Doubale method (KDM). A multivariable logistic regression analysis was conducted to evaluate the correlation between BA, biological age acceleration, and the incidence of AECOPD, complemented by subgroup analyses to explore the dose‒response dynamics between biological age acceleration and the risk of AECOPD. The dataset was partitioned into training and validation sets at a 7:3 ratio, and LASSO regression was applied to refine the model's variable composition. To assess the ability of different variables to discriminate current disease status, we developed the initial model and three subsequent models, with the following variables added in the new model: Chronological age (CA), BA, and biological age acceleration. The models were subsequently evaluated within both datasets.

Results

The study cohort comprised 2,511 patients, through an analysis of the transect data, with 59.1% experiencing acute exacerbations. Both BA (79.14 ± 9.49 years) and biological age acceleration (1.04 ± 2.82 years) emerged as independent risk factors for AECOPD (P < 0.001). In Model 3, each year increment in BA and biological age acceleration corresponded to a 1.04-fold (95% CI = 1.027–1.048, P < 0.001) and 1.18-fold (95% CI = 1.14–1.224, P < 0.001) increase in exacerbation risk, respectively. The biological age of patients with stable COPD was significantly lower than the actual age (-0.36 ± 2.56 years), which suggests a significant inter-individual heterogeneity in the biological aging process of COPD patients. Subgroup analysis confirmed a pronounced dose‒response relationship between biological age acceleration and AECOPD risk(Q4 vs. Q1: OR = 2.7, 95% CI = 2.172–3.518). LASSO regression pinpointed BMI, Diabetes, Hypertensive heart disease, Cor pulmonale, Stroke, and Hyperlipidemia as critical variables within the model. The internal validation process revealed AUC values of 0.735 (95% CI = 0.7–0.77), 0.742 (95% CI = 0.707–0.777), 0.753 (95% CI = 0.719–0.787), and 0.766 (95% CI = 0.733–0.8) for the respective models. The HL test confirmed the models' good fit (P = 0.128, P = 0.121, P = 0.272, P = 0.795), with Model 4 exhibiting the most precise calibration against the diagonal reference. Decision curve analysis (DCA) indicated that all the models provided a net benefit in disease outcome discrimination, with Model 4 yielding the most significant advantage.

Conclusions

The acceleration of aging portends an increased propensity for acute exacerbations, and a distinct dose–response relationship is observable between biological age acceleration and exacerbation events. BA and biological age acceleration outperform chronological age in discerning the likelihood of acute exacerbations, underscoring their enhanced ability to predict this critical health outcome.

Introduction

Chronic obstructive pulmonary disease (COPD) is a chronic respiratory disease (CRD) characterized by irreversible airway limitation and persistent chronic airway inflammation, accounting for approximately 55% of global CRD cases. COPD is the third leading cause of death worldwide, with acute exacerbations of COPD (AECOPD) being one of the main contributors to this statistic [1]. Exacerbations in patients are caused primarily by infections, exposure to irritants such as smoke, or poor adherence to treatment [2].

Aging is a major risk factor for a variety of age-related diseases [3]. Research has also identified age as an independent risk factor for the occurrence of AECOPD [4]. For every decade increase in age, the risk of AECOPD increases by 21% [5]. However, previous studies have not considered that COPD likely accelerate the aging process, and few have discussed the relationship between this accelerated aging and AECOPD.

The process of aging is influenced by genetics, the environment, lifestyle habits, and other factors, making the aging process unique to each individual. Relying solely on chronological age (CA) to define and study aging is not rigorous enough [6]. To overcome the shortcomings of chronological age, the concept of biological age (BA), which better represents the degree of biological aging, has been proposed [7]. With respect to the calculation of biological age, some studies have used the methylation status of GpC sites in the human genome to construct epigenetic clocks for calculating biological age, but obtaining individual genomic methylation information is difficult in clinical practice [8]. The KDM, on the other hand, can be used to calculate an individual's biological age using a variety of biological markers [9]. When comparing the predictive power of biological age calculated by five different methods for disease, the KDM method was found to be the most convincing [6].

Therefore, this study employed the KDM method to determine the true extent of patients'aging, identify individuals experiencing accelerated aging, explore the relationship between accelerated aging and AECOPD, and quantify the dose‒response relationship between aging and disease. Discrimination models were constructed to verify the clinical application value of biological age-related variables. By investigating the relationship between aging and AECOPD through clinical data, this study aimed to provide clinical data to support future research.

Methods

Study population

The study population consisted of patients hospitalized at Chengdu Third People's Hospital from January 2018 to December 2022. All patients initially collected were required to fulfill the GOLD diagnostic criteria for COPD, defined as a post-bronchodilator FEV1/FVC ratio < 0.7 [10].

Study design

We first excluded patients who might be disturbed by other diseases; A total of 11,918 patients were excluded due to missing data for at least one of the following variables: (1) General patient characteristics: including patient hospital admission ID, age, sex, blood pressure upon admission, height, weight, smoking history, drinking history, and past medical history; or (2) Bio-age calculation data (all laboratory parameters were obtained from the first post-admission blood tests).The detailed exclusion criteria are depicted in a flowchart (Fig. 1). This study was a retrospective analysis, approved by the Ethics Committee of Chengdu Third People's Hospital, and the need to obtain informed consent from patients was waived.

Fig. 1.

The Inclusion and Exclusion Process of Study Samples

Definitions

AECOPD: determined by a respiratory specialist based on acute worsening of clinical symptoms (significant worsening of cough, sputum, dyspnea, and other symptoms within 48 h) and meeting the criteria for requiring hospitalization for respiratory support [11]. COPD: clinically stable (defined as no acute exacerbation events occurring) at the time of inclusion. Biological age acceleration: Since healthy aging was not included as a control group in this study, we defined"accelerated aging"as a state of aging relative to that of patients without premature aging during the stabilization period.

BA Calculation

(1) Method Selection: The KDM algorithm was originally developed using NHANES III cohort data. In recent years, this method has been extensively applied [12, 13]and validated [14] in studies utilizing Chinese population data. Based on this established reliability, we selected KDM as the biological age calculation method for the current study. The biological age acceleration is defined as the difference between biological age and chronological age. A positive value indicates accelerated aging in the individual [15].(2) Data preprocessing: Eleven biomarkers were selected based on their high correlation (r > 0.8) with chronological age, as validated in previous studies [15]. The required data includes alkaline phosphatase、mean corpuscular volume、C-reactive protein、glycated hemoglobin、white blood cell count、creatinine、albumin、triglycerides、total cholesterol、red blood cell count and systolic blood pressure [16].Samples with missing data for one or more biomarkers were excluded, resulting in a final cohort of 2,511 samples with complete biomarker data. No additional preprocessing (e.g., missing value imputation or outlier removal) was performed. Unit standardization was applied to all clinical biomarkers based on NHANES III cohort data, and logarithmic transformation was conducted for non-normally distributed variables (e.g., ln-CRP). (3)Biological age calculation: The calculation process employs the open-source BioAge software package developed collaboratively by Dayoon Kwon and Daniel W. Belsky (GitHub: http://github.com/dayoonkwon/BioAge), in which built-in parameters and standardized procedures already existed in the pre-training function, and the specific KDM-BA calculation formula was provided in the original study [17]. The linear relationship between biomarkers and chronological age was first analyzed to obtain the regression slope (k), intercept (q), root mean square error (s), and explained variance (r2). Biological age correlates with actual age in this study (r = 0.952) (Table 1).

Table 1.

11 biomarkers and dosage units

1): an estimate of the age of the preliminary organism in the linear relationship();

2): calculate the feature correlation coefficient();

3): Calculate the scaling factor()

4): Calculate the final biological age();

Covariates

Previous studies have demonstrated that gender [18], BMI [19], smoking status [20], and alcohol consumption [21] significantly influence both the incidence and prognosis of AECOPD to varying degrees. In line with existing literature, we additionally adjusted for the prevalence of multiple comorbidities potentially associated with AECOPD development as potential confounders [22]. This study incorporated several covariates [23], including sex, height (meters), weight (kilograms), smoking status (never smoked, former smoker, current smoker), drinking status (never drank, former drinker, current drinker), and the presence of comorbid conditions such as Hypertension, Diabetes, Atherosclerosis, Atrial fibrillation, Heart failure, Coronary heart disease, Hypertensive heart disease, Cor pulmonale, Hyperlipidemia, Stroke and CKD-5. Variables showing clinically meaningful between-group differences are incorporated into subsequent multivariate analyses.

Statistical analysis

For the cross-sectional analysis component, patients were stratified into COPD and AECOPD groups according to disease status. Continuous data are presented as the means ± standard deviations, and categorical data are expressed as percentages (%). Independent sample t-tests and chi-square tests were used for intergroup comparisons, with P < 0.05 indicating a significant difference between groups [24]. Multivariate logistic regression was employed to investigate the relationships of biological age and biological age acceleration with AECOPD [25]. Model 1 was an unadjusted crude model, Model 2 was adjusted for sex, BMI, and smoking variables, and Model 3 incorporated comorbidity-related covariates based on Model 2 to calculate the odds ratios (ORs) and 95% confidence intervals (95% CIs). Subgroup analysis was conducted to observe the dose‒response relationship between biological age acceleration and AECOPD within Model 3, by categorizing biological age acceleration into quartiles (Q1: top 25%; Q2: 25%−50%; Q3: 50%−75%; Q4: 75%−100%), to analyzing the linear trend across quartiles, and examining different dose‒response relationships between males and females [26]. To evaluate the discrimination value of biological age for COPD acute exacerbations, we developed a logistic regression model. Receiver operating characteristic (ROC) curves were plotted for both the training and validation sets, and the area under the curve (AUC), sensitivity and specificity were calculated to compare the predictive abilities of different models [27]. Decision curve analysis (DCA) was used to assess the clinical net benefit of each model at a 50% prevalence threshold [28], and calibration curves were plotted with the Hosmer‒Lemeshow (HL) test to evaluate model calibration. The original model was considered the old model, and the net reclassification index (NRI) and integrated discrimination improvement (IDI) were used to test the improvement in predictive performance when new variables were introduced.

Results

Basic characteristics of patients

As shown in Table 2, a total of 2,511 participants were included in the study, with an average age of 77.42 ± 9.58 years. The incidence of AECOPD was 59% (n = 1,483). Intergroup comparisons revealed that patients who experienced acute exacerbations were predominantly male (67.4%), had a lower BMI (22.92 ± 3.81 kg/m²), and had higher proportions of those who had quit smoking (30.4%) and were still smoking (21.6%). Consistent with the expected outcomes, patients in the AECOPD group had higher chronological ages (78.09 ± 9.46 years) and biological ages (79.14 ± 9.49 years), with a positive mean biological age acceleration (1.04 ± 2.82 years), and a greater proportion of patients with higher biological ages, indicating accelerated aging (61.8%). Overall, most patients had a combination of hypertension and diabetes mellitus, and AECOPD patients in particular had a higher prevalence of diabetes mellitus. All these differences were statistically significant.

Table 2.

The baseline characteristics of the participants included in the study (N = 2511)

P-values for parametric (T-test) and non-parametric tests (Chi-squared test); COPD: chronic obstructive pulmonary disease; AECOPD: acute exacerbation of COPD; Never: the patient has never smoked or drank alcohol. Former: the patient has quit smoking or drinking. Mild: the patient has not quit smoking or drinking.CKD-5: Stage 5 chronic kidney disease; Statistically significant results (P < 0.05) are highlighted in bold throughout the table

n:sample number; Continuous variables: Mean ± SD, Categorical variables: %; BMI: body massindex

Relationship between Biological Age and Risk of AECOPD

Table 3 presents the detailed results of the multivariate logistic regression, indicating that biological age and biological age acceleration are independent risk factors for the occurrence of AECOPD. The crude model demonstrated that for each additional year of biological age and biological age acceleration, the risk of AECOPD increased by 3.3% (95% CI = 1.024–1.041) and 21.5% (95% CI = 1.176–1.254), respectively. In Model 3, increase of one year in biological age and biological age acceleration were associated with a 3.7% (95% CI = 1.027–1.048) and 18.1% (95% CI = 1.14–1.224) increase in the risk of AECOPD, respectively.

Table 3.

Multifactorial logistic regression analysis of Age, KDM age and KDM-Age advance in relation to AECOPD

Model 1: crude model with no adjustment for covariates. Model 2 was based on gender, BMI, and smoking status. Model 3 was additionally adjusted for chronic disease history based on model 2. Ref: Reference Group (Patient characteristics: Biological age < Chronological age). Statistically significant results (P < 0.05) are highlighted in bold throughout the table. Biological acceleration: the difference between biological age and chronological age; Biological age acceleration no: negative biological age acceleration

n:sample number; Continuous variables: Mean ± SD, Categorical variables: %; BMI: body mass index

Patients in Model 3 who experienced accelerated aging had an increased risk of acute exacerbation that was 1.85-fold (95% CI = 1.544–2.217, P < 0.001) increased risk of acute exacerbation. Subgroup analyses revealed significant dose–response relationships across different populations (P for trend < 0.001). In the non-sex-stratified models, the risk of AECOPD in Groups Q2, Q3, and Q4, compared with that in the baseline group (Q1) increased by 1.4 times (95% CI = 1.007–2.064), 1.5 times (95% CI = 1.204–1.87), and 2.8 times (95% CI = 2.172–3.518), respectively. For males, the risk of acute exacerbation for Groups Q2, Q3, and Q4 was greater than that for the baseline Group Q1. Only the female sex in Group Q4 was significantly associated with the risk of AECOPD. These findings suggest that the impact of biological age acceleration on AECOPD varies by sex, highlighting the necessity of considering sex specificity in the management and prevention of AECOPD, as detailed in Table 4.

Table 4.

Subgroup: Relationship between the degree of accelerated aging and AECOPD

Ref: Reference Group (Patient characteristics: Q1: top 25%). Statistically significant results (P < 0.05) are highlighted in bold throughout the table

Constructing discrimination models and comparing the disease discrimination abilities of different models

The cohort was randomly split into training (n = 1,758) and validation (n = 753) sets at a 7:3 ratio. Least Absolute Shrinkage and Selection Operator (LASSO) regression was used as a regularized modeling approach [29]. Owing to the multitude of variables and significant collinearity among some variables, LASSO regression was applied to select variables for model inclusion, employing L1 regularization to reduce model complexity, and cross-validation to determine the optimal lambda value, thereby balancing model fitness and simplicity. LASSO regression ultimately led to the selection of six variables for constructing the disease discrimination model: BMI, Diabetes, Hypertensive Heart Disease, Cor pulmonale, Hyperlipidemia and Stroke. Model 1 was the baseline model, whereas Models 2, 3, and 4 enhanced the baseline model by incorporating chronological age, biological age, and biological age acceleration variables, respectively, to improve its discrimination performance. ROC curves were plotted for the four models in both the training and validation sets (Figs. 2-A and B), and the area under the curve (AUC), sensitivity and specificity of the models were calculated. The specific results are presented in Table 3. Model 4 showed similar discrimination in both datasets, with AUC values of 0.751 and 0.766, respectively. The calibration curves of the four models (Figs. 3-C and D) indicate that Model 4 more closely fit the diagonal line in both the training and validation sets, especially in the high-risk area, suggesting that Model 4 performed better in distinguishing whether patients will experience an acute exacerbation. The HL test results revealed that all four models fit well in the internal validation set (P = 0.128, P = 0.121, P = 0.272, P = 0.795). The DCA curves of different models (Figs. 24-E and F) suggested that all four predictive models have a net benefit in disease discrimination at the same level of diagnostic performance, but that Model 4 has better prospects for clinical application (Fig. 4 and Table 5).

Fig. 2.

Receiver Operating Characteristic Curve (ROC) for discrimination model one, discrimination model two, discrimination model three, discrimination model four. A: ROCs of discrimination models in training cohort; B: ROCs of discrimination models in testing cohort

Fig. 3.

Calibration curves for discrimination model one, discrimination model two, discrimination model three, discrimination model four. C: Calibration curves of discrimination models in training cohort; D: Calibration curves of discrimination models in testing cohort

Fig. 4.

Decision Curve Analysis (DCA) for discrimination model one, discrimination model two, discrimination model three, discrimination model four. E: DCAs of discrimination models in training cohort; F: DCAs of discrimination models in testing cohort

Table 5.

The area under the curve (AUC) values, sensitivity, and specificity of the discrimination models in the two data sets

To assess the improvement in the ability of the three predictive models introducing the new variables to discriminate disease composition on the basis of the original models, NRI and IDI were introduced for model comparison. The results showed that the overall disease discrimination abilities of both Model 3 and Model 4 were enhanced compared with those of the original model, and Model 4, which incorporated the biological age acceleration variable, showed particularly prominent improvement in combined disease discrimination ability. The details are given in Table 6.

Table 6.

NRI and IDI analysis of different Subsequent models compared to the Model one

NRI: Net Reclassification Improvement for Classification; IDI: Integrated Discriminant Improvement; Significant results are marked in bold (p < 0.05)

Discussion

This study included a total of 2,511 patients, with biological age being calculated using the KDM method. Research has revealed that patients experiencing acute exacerbations have higher biological ages. Multivariable logistic regression analysis adjusted for potential confounders demonstrated that both biological age (OR = 1.037, 95% CI: 1.027–1.048) and accelerated biological aging (OR = 1.181, 95% CI: 1.140–1.224) were independently associated with AECOPD (P < 0.001). Additionally, a significant dose–response relationship was observed between biological age acceleration and the development of AECOPD. The discrimination models incorporating biological age and biological age acceleration demonstrated superior discrimination and calibration, as well as better clinical applicability, and showed an enhanced overall composite discrimination capability compared with the previous models.

Prior studies have reported that COPD accelerates the aging process [30].AECOPD represents a critical sentinel event in COPD progression. The study found that approximately 50% of patients experience at least one acute exacerbation annually [31]. However, few studies have specifically addressed the subset of patients with acute exacerbations. Our study of clinical data provides a detailed elucidation of the relationship between biological age/biological age acceleration and AECOPD, and thereby fills a gap in the literature.

Similar to transcriptomic studies of COPD patients that identified multiple disease-related differentially expressed genes [32], gene sequencing has also revealed numerous differentially expressed genes in AECOPD patients at the molecular level. Through transcriptomic analysis and pathway enrichment of blood from patients with deterioration by different pathogens, IFITM3, ISG15, DEFA3 and CD47 were considered as potential biomarkers of acute exacerbation of COPD, and the mitochondrial damage pathway, the pathway of positive regulation of NF-κB transcription factor activity, and the pathway of positive regulation of ROS metabolic processes were significantly up-regulated [33], of which interferon-stimulated gene 15 (ISG15), a small ubiquitin-like protein whose main function is to participate in the regulation of viral infections, has now been shown to be a central substance in the association of aging and age-related diseases [34]. The WNT/β-catenin signaling pathway has been found to be down-regulated during aging [35], and this specific change also occurred in an AECOPD patient model with a significant up-regulation of phosphorylated β-catenin levels. The study also found that the PI3K/Akt signaling pathway was associated with the deterioration of COPD [36], and it has been demonstrated that PI3K has an important regulatory role in a wide range of aging characteristics [37].

Patients with acute exacerbations experience physiological lung aging and accelerated aging due to the combined effects of various mechanisms. Studies have shown that healthy lungs begin to age early in life [38]. During the aging process, lung tissue stiffness and decreased compliance [39], with weakening of respiratory muscles resulting in reduced exercise tolerance in patients [40], enlargement of alveolar spaces with a relatively constant total lung capacity leading to increased residual volume, and a progressive decrease in FEV1/FVC, which results in elderly individuals exhibiting pulmonary functional changes similar to those observed in COPD patients [41].

At the cellular level, aging is reflected by irreversible cell cycle arrest due to various types of damage [42]. Senescent cells accumulate and promote chronic inflammation in the lungs by expressing the senescence-associated secretory phenotype (SASP) and releasing a plethora of proinflammatory factors [43]. Cellular senescence is considered a fundamental mechanism in the progression of AECOPD. Connective tissue growth factor (CTGF) induces cellular senescence by inhibiting the renewal capacity of alveolar epithelial cells, exacerbating the degree of airway obstruction and driving the occurrence of AECOPD [44]. Varying degrees of hypoxia in most patients during AECOPD may promote immune senescence [45]. Studies have identified significant upregulation of hypoxia-related genes CXCL9 and CXCL12 in COPD-PH patients [46], CXCL12 has been investigated as a therapeutic target for pulmonary arterial hypertension [47]. Targeting hypoxia-related genes represents a potential new direction for AECOPD treatment. Furthermore, owing to the"end replication problem,"telomere length decreases with increasing age [48], and increased oxidative stress in the lungs is one of the reasons for accelerated telomere shortening [49]. Shortened telomeres may be the basis for the decline in lung function associated with both aging and COPD [50]. telomere shortening increases the risk of acute exacerbation, thereby also serving as a biomarker for identifying various adverse outcomes, including acute exacerbations in COPD patients [51, 52]. Oxidative stress is a failed defense mechanism against oxidative effects, and enhanced smoke-induced irritation or airway inflammation in COPD patients leads to increased oxidative stress levels, accelerating lung aging [53].

Several meaningful findings have been used in clinical settings. Effective anti-aging treatments have been shown to improve Acute Exacerbations of Chronic Obstructive Pulmonary Disease (AECOPD). Traditional Chinese medicine, such as Tong Sai Granules, can modulate cell aging through the MAPK-SIRT1-NF-κB pathway, thereby reducing the levels of C-reactive protein (CRP) and serum amyloid A (SAA) during inflammatory episodes. This helps control inflammation and mitigate the severity of AECOPD [54]. Additionally, inhibiting PM2.5-induced NOX4/Nrf2 redox imbalance and reducing mitochondrial autophagy can also alleviate the symptoms of AECOPD [55].

Moreover, recent research has identified COPD as a"multimorbidity"condition, which is consistent with the findings of this study, where almost all patients had chronic diseases affecting other systems. This is due to the shared risk factor of age across various diseases, making COPD patients prone to chronic conditions such as cardiovascular and cerebrovascular diseases, chronic kidney disease, and diabetes [56]. Emerging evidence demonstrates that hypertension not only elevates cardiovascular risk but also accelerates biological aging [57]. Accelerated biological aging has been identified as an independent predictor of adverse clinical outcomes in hypertensive patients [58]. Notably, the prevalence of cor pulmonale was significantly higher in AECOPD patients than in stable COPD controls (45.6% vs. 13%, P < 0.001). Mechanistically, pulmonary hypertension exhibits a robust senescent cell phenotype, and early-phase studies demonstrate the therapeutic potential of senolytic interventions in this context [59]. Type 2 diabetes mellitus (T2DM) similarly exhibits an accelerated aging phenotype [60]. In our cohort, 52% of AECOPD patients had current or prior smoking history. Tobacco exposure independently accelerates biological aging, with studies demonstrating smokers develop age-related comorbidities up to a decade earlier than non-smokers [61].

Our study identified premature aging (biological age > chronological age) in 42.5% of COPD patients. Strikingly, stable-phase patients exhibited younger biological ages than chronological ages (− 0.36 ± 2.56 years), while AECOPD patients showed significant acceleration (1.04 ± 2.82 years, P < 0.001). This disparity not only substantiates the association between COPD and premature aging, but also demonstrates that patients during acute exacerbation exhibit significantly higher biological age compared to those in stable phase (P < 0.001). While most studies report positive associations between COPD and biological age acceleration [62, 63], emerging evidence intriguingly suggests potential negative correlations in specific subgroups [64].This apparent contradiction may reflect hitherto unrecognized heterogeneity in COPD-related aging trajectories.

We hypothesize that methodological variations, particularly in patient selection criteria and biological age calculation approaches, may account for these discrepant findings. Unlike previous studies that defined COPD solely by FEV1/FVC < 0.7 without explicit AECOPD exclusion [62, 65], our study rigorously enrolled only stable-phase patients, thereby minimizing exacerbation-induced confounding in aging assessments.

Furthermore, in this study, stable-phase patients showed significantly lower levels of inflammatory markers than those with acute exacerbations (CRP: 15.67 ± 32.3 vs. 37.9 ± 49.32 U/L, p < 0.001; white blood cell count: 6.52 vs. 7.42 × 10⁹/L, p < 0.001), the observed attenuation of aging progression in stable COPD patients may be attributed to either their low-grade systemic inflammation or the anti-aging effects of effective anti-inflammatory therapies maintaining disease stability. Compelling evidence demonstrates that anti-inflammatory therapy serves as a biologically plausible anti-aging intervention [66], this is supported by observations of DNA methylation age (DNAmAge) deceleration in COPD patients receiving corticosteroid treatment [64].

The KDM exhibits heterogeneity in calculating biological age, where inter-individual variability may introduce measurement errors [67], and current biological age algorithms fail to account for key disease-independent aging modifiers—including genetic predisposition, environmental exposures, and dietary factors—representing a critical limitation for future methodological refinement. Moreover, recent studies have provided mechanistic insights into anti-aging pathways in COPD, including, Lipid reprogramming that generates anti-inflammatory metabolites in COPD patients [68], and innovative senotherapeutic approaches targeting immune pathways of senescent cells for prevention and reversal of aging [69]. Therefore, the mechanisms underlying this inverse association require further in-depth investigation. It needs to be emphasized that COPD patients did not exhibit a significant increase in biological age compared to AECOPD patients, but this does not imply an absolute reduction in biological age among COPD patients.

In conclusion, the AECOPD discrimination model developed in this study demonstrated robust clinical applicability for biological age, with strong discriminative performance (Model 3: AUC = 0.754; Model 4: AUC = 0.767). While previous studies have developed similar discrimination models [70], our study introduces a critical innovation by incorporating accelerated biological age—a composite metric of systemic aging—as a key variables. The model demonstrated robust clinical discrimination performance across both training and validation cohorts. As shown in Table 6, the NRI of Model 4 was 0.115 (P = 0.003), indicating that compared with Model 1, Model 4 with the biological age acceleration variable correctly reclassified 11.5% of AECOPD patients. The significant improvement in IDI (0.044, P < 0.001) further confirmed the enhanced discriminatory ability of Model 4 [71], suggesting its potential clinical utility in reducing underdiagnosis.

Limitations: Despite the valuable insights provided by this study, several limitations should be acknowledged. First, the retrospective nature of the study precluded the inclusion of a healthy control group, which may limit the robustness of the analysis regarding the relationship between COPD patients and biological age. Second, the study employed a single method for calculating biological age, which does not permit a comprehensive validation of its feasibility, suggesting a need for further refinement in future research. Third, the biological age calculation model was trained on a European population, which may introduce biases due to racial differences, thus affecting the generalizability of the findings. Fourthly, in addition to longitudinal validation of distinct biological age algorithms to assess their reliability and robustness, further research is needed to explore whether additional protective factors against aging exist in stable COPD patients, for example, in prospective studies to explore whether inflammation control is associated with specific aging manifestations in stable patients. Fifth, this cross-sectional analysis cannot establish causal relationships between premature aging and AECOPD. We are currently designing a prospective cohort study to further investigate biological age dynamics, which will facilitate clinical translation of these findings. Finally, the study utilized traditional statistical methods for data analysis, and there is a significant opportunity to enhance the predictive power and accuracy by incorporating machine learning techniques to build and compare disease discrimination models.

Conclusion

In summary, our findings suggests a significant correlation between accelerated aging and an elevated risk of acute exacerbations of chronic obstructive pulmonary disease (AECOPD). The magnitude of increase in biological age, as determined by biomarker calculations, is directly proportional to the risk of acute exacerbations, may hinting the clinical utility of biological age and its acceleration in predicting such events. Considering the escalating detrimental impacts of acute exacerbations due to environmental and demographic shifts, delaying the aging process emerges as a promising therapeutic and preventive strategy. However, this approach necessitates further research to identify intervenable pathways and targets within the aging process, coupled with the conduct of well-designed randomized controlled trials.

Authors’ contributions

Wang Yujiao and Mu Ting contributed equally to this study, participating in all aspects including study design, data cleaning, data analysis, figure preparation, manuscript writing, and manuscript revision. Fu Yufen was involved in data cleaning, data analysis, and manuscript revision. Wang Yuxin contributed to data collection and data analysis. Li Guoping participated in study design and manuscript revision.

Funding

Development center for medical science and technology national health commission of the people’s republic of China (WKZX2023HK0123).

Data availability

The datasets analyzed during the current study are available from the corresponding author upon reasonable request.

Declarations

Ethics approval and consent to participate

This study was conducted in accordance with the ethical standards of the Declaration of Helsinki and its later amendments. The study protocol was approved by the Ethics Committee of Chengdu Third People's Hospital (approval number: Chengdu Third Hospital Ethics [2024] -S-401). All patient information was collected anonymously to ensure the confidentiality and privacy of the participants.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.