Dynamic frailty trajectories and risk of chronic obstructive pulmonary disease: a large-scale prospective cohort study

Abstract

Background

Frailty progression may represent a dynamic clinical marker associated with future chronic obstructive pulmonary disease (COPD) diagnosis, but longitudinal evidence is limited. This study examined the longitudinal association between frailty trajectories and incident COPD.

Methods

This prospective cohort study analyzed UK Biobank data. Participants (n = 30,557) aged 40–69 with ≥ 2 valid frailty assessments and no baseline COPD were included. Frailty was assessed longitudinally using a modified Fried phenotype (5 components: weight loss, exhaustion, slow walking, low activity, weak grip). A joint longitudinal-survival model simultaneously analyzed frailty trajectories (linear mixed-effects) and COPD incidence (Cox proportional hazards), linked by shared random effects. Models adjusted for covariates and established COPD risk factors. Sensitivity and subgroup analyses assessed robustness.

Results

Among 30,557 UK Biobank participants, baseline frailty classification revealed 60.8% non-frail, 38.2% pre-frail, and 1.0% frail. Frailty scores increased significantly over time (β = 0.020, P < 0.001). An increasing frailty trajectory was positively associated with the instantaneous hazard of incident COPD (association parameter α = 2.33, 95% CI: 1.64–3.02, P < 0.001). This detrimental association between worsening frailty and COPD incidence remained consistent across all pre-specified subgroups (all P < 0.05) and was robust to multiple sensitivity analyses addressing potential biases, including excluding early events or high-risk groups (all P < 0.001). The joint model showed moderate-to-acceptable discriminative performance (time-dependent AUC 0.63–0.77) for individualized risk assessment.

Conclusions

Our study demonstrates that progressive worsening of frailty is dynamically associated with incident COPD, supporting the potential value of repeated frailty assessment for COPD risk stratification.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12890-026-04380-2.

Introduction

Frailty, a multidimensional geriatric syndrome characterized by diminished physiological reserve and increased vulnerability to stressors, has emerged as a critical prognostic indicator of adverse health outcomes in aging populations [1]. Frailty encompasses physical, cognitive, and social domains, with physical frailty characterized by weight loss, fatigue, low activity, slow gait, and weak grip strength [2]. As populations age globally, understanding the relationship between frailty and chronic disease development has become increasingly important for public health planning and clinical decision-making [3]. Chronic obstructive pulmonary disease (COPD), affecting over 400 million individuals worldwide and ranking as the third leading cause of death, represents a particularly relevant chronic condition given its strong associations with aging and functional decline [4, 5].

The intersection of frailty and respiratory health has garnered substantial research attention over the past decade [6]. Studies have consistently demonstrated that frail older adults exhibit lower forced expiratory volume in one second (FEV1), reduced forced vital capacity, and increased susceptibility to respiratory infections compared to their robust counterparts [7, 8]. Longitudinal investigations have further revealed that baseline frailty status is associated with future accelerated lung function decline, increased COPD exacerbation rates, and higher respiratory-related mortality [9]. For instance, Hanlon et al. demonstrated that frailty was significantly associated with increased mortality and major adverse cardiovascular events [10]. Similarly, Wang et al. conducted a comprehensive systematic review and meta-analysis showing that frailty in COPD patients was associated with significantly increased all-cause mortality and hospitalization risk [7].

However, these studies have employed traditional analytical approaches that treat frailty as a static baseline characteristic, failing to capture its dynamic nature over time [11]. Recent longitudinal research has increasingly recognized frailty as a dynamic and potentially reversible condition that can improve or worsen throughout the aging process, influenced by interventions, lifestyle changes, and disease progression [12]. Furthermore, the temporal relationship between frailty development and COPD onset remains poorly understood, with recent studies beginning to explore the bidirectional associations between physical frailty and multiple health outcomes, including respiratory conditions [13]. Emerging evidence suggests that frailty trajectories in the final years of life follow distinctive patterns, with rapid initial progression followed by plateaus, highlighting the importance of longitudinal assessment rather than single time-point measurements [14].

Despite growing evidence linking frailty and respiratory health, significant methodological limitations persist in the existing literature. Most studies have relied on single baseline frailty assessments, potentially missing important changes in frailty status that may influence disease risk [15]. Additionally, traditional survival analysis methods cannot adequately model the bidirectional relationship between evolving frailty trajectories and time-to-event outcomes. Joint modeling approaches, which simultaneously analyze longitudinal biomarker trajectories and survival outcomes, offer a sophisticated analytical framework to address these limitations [16], but have been underutilized in frailty-respiratory health research despite their proven effectiveness in avoiding biased estimates when informative dropout occurs [17]. This represents a critical gap in our understanding of how dynamic frailty progression is related to COPD development over time.

To address these limitations, this study aimed to investigate the dynamic longitudinal association between frailty trajectories and incident COPD using a joint longitudinal-survival modeling approach. We hypothesized that steeper increases in frailty over time would be associated with a higher instantaneous hazard of subsequent COPD diagnosis. Joint models were employed to analyze data from the UK Biobank cohort, with repeated frailty assessments over multiple visits and comprehensive covariate adjustment incorporated to account for potential confounders [13]. The analytical approach enables simultaneous modeling of individual frailty trajectories and their association with COPD risk, providing novel insights into the temporal dynamics of frailty-respiratory health relationships and identifying high-risk populations for targeted preventive interventions [18].

Methods

Study design and study population

This prospective cohort study utilized data from the UK Biobank, a large-scale population-based cohort that recruited approximately 500,000 participants aged 40–69 years between 2006 and 2010. Data from the baseline visit (2006–2010) and three follow-up visits (2012–2013, 2014, and 2019–2021) were analyzed. Participants were included if they had a baseline assessment and at least one follow-up visit with complete frailty assessment data. Exclusion criteria were as follows: (1) Insufficient frailty data: participants were required to have at least two valid measurements for each of the five frailty components (weight loss, fatigue, walking speed, physical activity, grip strength) across four visits; (2) Pre-existing COPD diagnosis at baseline; (3) Severe comorbidities: participants with baseline asthma (ICD-10: J45), interstitial lung disease (J84), chronic kidney disease (N18), cirrhosis (K74), or heart failure (I50) were excluded to minimize confounding; (4) Final quality control: this included excluding participants with missing baseline frailty scores or inconsistent temporal records (i.e., longitudinal assessments occurring after COPD diagnosis or death) to ensure model integrity. After screening, 30,557 participants were included in the final analysis (Fig. 1). This research was conducted in accordance with the ethical principles of the Declaration of Helsinki, approved by the Northwest Ethics Committee (REC 21/NW/0157). All participants provided informed consent.

Fig. 1.

workflow of study

Frailty assessment

Frailty was assessed using a modified Fried frailty phenotype comprising five components: (1) unintentional weight loss based on self-reported weight change, defined as answering “Yes - lost weight” to the weight change question; (2) exhaustion assessed through standardized questionnaires, defined as experiencing tiredness or lethargy “more than half the days” or “nearly every day”; (3) slow walking speed determined by self-reported usual walking pace, with “slow pace” classified as abnormal; (4) low physical activity evaluated using a composite assessment of activity types and frequency, where participants reporting “none of the above” activities or only light DIY activities performed once weekly or less frequently were classified as having low activity; and (5) weak grip strength measured using a dynamometer with sex- and BMI-specific thresholds, where the minimum value of left and right hand grip strength was compared against established cutoffs [2]. Each component was scored as 0 (normal) or 1 (abnormal), yielding a total frailty score of 0–5. Participants were categorized as non-frail (score 0), pre-frail (score 1–2), or frail (score ≥ 3). Frailty assessment was performed at baseline and each follow-up visit to capture longitudinal changes in frailty status.

Outcome definition

COPD status was defined using the UK Biobank’s pre-validated algorithmic classification (Category 46: COPD outcomes). This classification combines baseline self-reported medical conditions (Field 20002), hospital admission diagnoses/procedures (Data-Coding 2000), and death registry ICD-10 codes (Fields 40001, 40002) to maximize the accuracy of case ascertainment.

Covariates

Covariates were selected based on established risk factors for COPD and potential confounders of the frailty-COPD relationship. Covariates included age (continuous), sex, ethnicity (White/non-White), education (university degree/non-degree), smoking status (never/previous/current), BMI categories (normal 18.5–24.9 kg/m²; overweight 25.0–29.9 kg/m²; obese ≥ 30.0 kg/m²), self-reported diabetes (yes/no), self-reported cardiovascular disease (yes/no), and Townsend deprivation index quartiles (Q1-Q4, Q4 indicating highest deprivation).

Statistical analysis

Descriptive statistics were presented as means ± standard deviations for continuous variables and frequencies (percentages) for categorical variables. Group differences were assessed using t-tests or chi-square tests as appropriate. To address missing data in baseline covariates, we utilized the multivariate imputation by chained equations (MICE) algorithm, extracting a complete imputed dataset for the primary analysis.

A joint longitudinal-survival modeling approach was implemented to simultaneously analyze frailty trajectory evolution and COPD incidence risk. The analytical framework comprised two interconnected sub-models linked through shared random effects. Specifically, a linear mixed-effects model was employed to characterize individual frailty score trajectories over time, accounting for individual baseline heterogeneity. Concurrently, a Cox proportional hazards model was utilized to describe COPD incidence risk, incorporating the latent underlying frailty process derived from the longitudinal sub-model. Detailed mathematical formulations, parameter specifications, and estimation algorithms for the joint model are provided in Supplementary Methods S1.

Before fitting the joint model, the proportional hazards assumption of the Cox model was assessed using Schoenfeld residuals. Predictive discrimination over follow-up was further assessed using dynamic time-dependent areas under the receiver operating characteristic curve (time-dependent AUC).

Subgroup analyses were performed based on prespecified stratification factors: age (< 56 vs. ≥ 56 years), sex (male/female), smoking status (never vs. ever/current smokers), cardiovascular disease (CVD) status (yes/no), education (university vs. non-university), and BMI categories (normal/overweight/obese). Six sensitivity analyses were conducted to verify result robustness: (1) excluding participants aged ≥ 65 years; (2) restricting to never-smokers; (3) removing obese individuals (BMI ≥ 30 kg/m²); (4) excluding COPD cases within 2 years; (5) requiring ≥ 5 years follow-up; and (6) excluding extreme frailty (top 5%). All analyses maintained the primary joint modeling approach, with consistent effect directions and significance confirming robustness.

All analyses were performed using R software (version 4.3.0). All statistical tests were two-sided with P < 0.05 considered statistically significant. Effect sizes were reported as hazard ratios (HR) with 95% confidence intervals.

Results

Baseline characteristics

Of the 30,557 participants included in the analysis, 18,586 (60.8%) were classified as non-frail, 11,672 (38.2%) as pre-frail, and 299 (1.0%) as frail based on the modified Fried frailty phenotype. Baseline characteristics varied significantly across frailty groups (Table 1). Frail participants had a higher proportion of women (70.9% vs. 51.4% in non-frail; P < 0.001) and were more likely to be obese (42.8% vs. 12.6%; P < 0.001). Educational attainment was inversely associated with frailty status, with university education being less common among frail participants (29.8% vs. 52.8% in non-frail; P < 0.001). The prevalence of diabetes (7.7% vs. 1.4%; P < 0.001) and cardiovascular disease (35.1% vs. 17.8%; P < 0.001) increased progressively with frailty severity.

Table 1.

Baseline characteristics by frailty status

Characteristic	Pre-frail (n = 11,672)	Frail (n = 299)	Non-frail (n = 18,586)	P-value
Age, years
Median (IQR)	57.0 (50.0, 62.0)	57.0 (51.0, 62.0)	55.0 (49.0, 61.0)	< 0.001
Sex, n (%)				< 0.001
Female	6,706 (57.5)	212 (70.9)	9,546 (51.4)
Male	4,966 (42.5)	87 (29.1)	9,040 (48.6)
BMI category, n (%)				< 0.001
Normal	4,332 (37.1)	65 (21.7)	8,355 (45.0)
Overweight	4,805 (41.2)	106 (35.5)	7,883 (42.4)
Obese	2,535 (21.7)	128 (42.8)	2,348 (12.6)
Smoking status, n (%)				< 0.001
Never	6,993 (60.0)	176 (59.3)	11,716 (63.1)
Previous	3,916 (33.6)	96 (32.3)	5,744 (30.9)
Current	738 (6.3)	25 (8.4)	1,105 (6.0)
Ethnicity, n (%)				< 0.001
White	11,343 (97.2)	289 (96.7)	18,282 (98.4)
Non-white	329 (2.8)	10 (3.3)	304 (1.6)
Education, n (%)				< 0.001
University	5,449 (46.7)	89 (29.8)	9,813 (52.8)
Non-university	6,223 (53.3)	210 (70.2)	8,773 (47.2)
Diabetes, n (%)				< 0.001
No	11,240 (96.5)	276 (92.3)	18,307 (98.6)
Yes	411 (3.5)	23 (7.7)	258 (1.4)
Cardiovascular disease, n (%)				< 0.001
No	8,788 (75.4)	194 (64.9)	15,263 (82.2)
Yes	2,867 (24.6)	105 (35.1)	3,311 (17.8)
Townsend deprivation index quartile, n (%)				< 0.001
1 (least deprived)	2,768 (23.7)	51 (17.1)	4,816 (25.9)
2	2,836 (24.3)	64 (21.4)	4,735 (25.5)
3	2,989 (25.6)	73 (24.4)	4,572 (24.6)
4 (most deprived)	3,072 (26.3)	111 (37.1)	4,451 (24.0)

Data are presented as median (interquartile range) for continuous variables and n (%) for categorical variables. P-values were calculated using Kruskal-Wallis test for continuous variables and chi-square test for categorical variables. BMI body mass index, IQR interquartile range. Frailty status was defined according to the modified Fried frailty phenotype: non-frail (0 points), pre-frail (1–2 points), frail (≥ 3 points)

Joint model results

The longitudinal sub-model demonstrated a significant increase in frailty scores over time (β = 0.020, 95%CI: 0.019–0.021, P < 0.001). Age showed a positive association with frailty score (β = 0.012, P < 0.001), followed by obesity (β = 0.383, P < 0.001) and overweight status (β = 0.097, P < 0.001). Male sex (β=-0.110, P < 0.001) and university education (β=-0.072, P < 0.001) were associated with lower frailty scores (Table 2).

Table 2.

Longitudinal sub-model results: linear mixed-effects model for frailty score trajectory

Variable	Estimate (95% CI)	P-value
Intercept	-0.202 (-0.245, -0.159)	< 0.001
Time	0.020 (0.019, 0.021)	< 0.001
Age	0.012 (0.012, 0.013)	< 0.001
Sex (Male vs. Female)	-0.110 (-0.121, -0.098)	< 0.001
Education (University vs. Others)	-0.072 (-0.083, -0.061)	< 0.001
BMI (Overweight vs. Normal)	0.097 (0.085, 0.110)	< 0.001
BMI (Obese vs. Normal)	0.383 (0.366, 0.399)	< 0.001

CI Confidence Interval. Estimates represent the change in frailty score per unit change in covariate. Positive estimates indicate increased frailty

The survival sub-model (Table 3) identified smoking as the predominant risk factor for COPD incidence. Current smokers had an 8.82-fold increased risk (HR = 8.816, 95%CI: 6.590-11.795, P < 0.001), while former smokers showed a 2.69-fold increased risk (HR = 2.692, 95%CI: 2.092–3.464, P < 0.001) compared to never smokers. Other significant risk factors included male sex (HR = 1.463, P = 0.001), advancing age (HR = 1.054, P < 0.001), and cardiovascular disease history (HR = 1.619, P < 0.001). Paradoxically, overweight (HR = 0.646, P < 0.001) and obesity (HR = 0.420, P < 0.001) were associated with reduced COPD risk, as was university education (HR = 0.724, P = 0.006). During the follow-up period, significant differences in COPD incidence were observed across frailty groups (Fig. 2). The Kaplan-Meier survival analysis revealed that individuals with baseline frailty had significantly lower COPD-free survival compared to non-frail and pre-frail participants (log-rank test, P < 0.0001).

Table 3.

Survival sub-model results: cox regression model for COPD incidence

Variable	HR (95% CI)	P-value
Age	1.054 (1.035–1.074)	< 0.001
Sex (Male vs. Female)	1.463 (1.160–1.845)	0.001
Smoking (Previous vs. Never)	2.692 (2.092–3.464)	< 0.001
Smoking (Current vs. Never)	8.816 (6.590-11.795)	< 0.001
BMI (Overweight vs. Normal)	0.646 (0.501–0.832)	< 0.001
BMI (Obese vs. Normal)	0.420 (0.282–0.628)	< 0.001
Education (University vs. Others)	0.724 (0.575–0.910)	0.006
Ethnicity (White vs. Others)	0.618 (0.335–1.138)	0.122
Townsend Deprivation Index	1.047 (0.954–1.150)	0.334
Diabetes (Yes vs. No)	1.244 (0.759–2.038)	0.386
CVD (Yes vs. No)	1.619 (1.286–2.038)	< 0.001
Association Parameter (α)*	2.333 (1.642–3.025)	< 0.001

HR Hazard Ratio, CI Confidence Interval. HR > 1 indicates increased risk of COPD incidence. The association parameter quantifies the strength of association between frailty trajectory and COPD risk

* α = association parameter linking frailty trajectory change to COPD risk. Value represents the association parameter estimate (α) and its 95% CI from the joint model

Forest plot showing the association parameter (α) and 95% confidence intervals for the relationship between frailty score and COPD risk across different subgroups. The horizontal lines represent 95% confidence intervals; an interval that does not cross the red dashed line (α = 0) indicates a statistically significant association. Subgroups are categorized by age (< 56 years vs. ≥56 years), sex (female vs. male), body mass index (normal weight, overweight, obese), smoking status (never smoker vs. ever smoker), education level (non-university vs. university), and cardiovascular disease status (no CVD vs. CVD). Point estimates are represented by circles, with circle size proportional to sample size. The vertical dashed red line indicates the null effect (α = 0). Different colors represent different subgroup categories. All analyses were performed using joint modeling of longitudinal frailty trajectories and time-to-COPD incidence, adjusting for age, sex, and relevant covariates. COPD chronic obstructive pulmonary disease, CVD cardiovascular disease, BMI body mass index

Fig. 2.

Kaplan-Meier Survival Curves for COPD-Free Survival by Baseline Frailty Status. Kaplan-Meier curves showing COPD-free survival probability over follow-up time stratified by baseline frailty status (non-frail, pre-frail, and frail). The log-rank test showed significant differences between groups (P < 0.0001). The frail group demonstrated the lowest COPD-free survival probability, particularly after 12 years of follow-up

The joint longitudinal-survival model revealed a positive association between the underlying frailty trajectory and the hazard of incident COPD (Association parameter α = 2.33, 95% CI: 1.64–3.02, P < 0.001), indicating that worsening frailty over time was dynamically coupled with a higher instantaneous COPD hazard (Table 3).

Model performance

Using longitudinal data from the first 7 years of follow-up, the joint model demonstrated moderate-to-acceptable discriminative performance, with time-dependent AUC values ranging from 0.63 to 0.77 for estimating COPD incidence risk across different future time horizons (Fig. 3A). The model’s capability for individualized risk assessment is illustrated through representative subject trajectories (Fig. 3B), demonstrating how early frailty trajectory patterns can inform long-term COPD risk evolution.

Fig. 3.

Joint Model Performance and Individual Risk Assessment. A Time-dependent AUC values. Higher AUC values indicate a stronger ability of the model to distinguish future COPD cases from non-cases across different follow-up years. B Individual COPD risk trajectories for five representative subjects. These plots illustrate how the model provides real-time updates of a participant’s future COPD risk as new frailty measurements are recorded. Shaded areas represent 95% confidence intervals

Subgroup analysis

Subgroup analyses revealed consistent associations between frailty score and COPD risk across different population strata (Fig. 4). The association parameter (α) ranged from 1.84 (95% CI: 0.98–2.69) in participants aged ≥ 56 years to 2.77 (95% CI: 1.24–4.30) in university-educated participants. Notably, numerically larger association estimates were observed in female participants (α = 2.72, 95% CI: 1.41–4.03, P < 0.001) compared to male participants (α = 2.33, 95% CI: 1.51–3.16, P < 0.001), and in participants with cardiovascular disease (α = 2.73, 95% CI: 1.52–3.94, P < 0.001) compared to those without (α = 2.37, 95% CI: 1.34–3.40, P < 0.001). Among BMI categories, the association was consistent across normal weight (α = 2.18, 95% CI: 0.68–3.67, P = 0.004), overweight (α = 2.36, 95% CI: 1.10–3.62, P < 0.001), and obese participants (α = 2.31, 95% CI: 1.04–3.58, P < 0.001). The association was stronger in ever smokers (α = 2.49, 95% CI: 1.51–3.47, P < 0.001) compared to never smokers (α = 2.38, 95% CI: 0.98–3.77, P < 0.001). Educational differences showed university-educated participants had a stronger association (α = 2.77, 95% CI: 1.24–4.30, P < 0.001) compared to non-university educated participants (α = 2.40, 95% CI: 1.49–3.30, P < 0.001). All subgroup analyses demonstrated statistically significant positive associations, with generally consistent effect directions across strata.

Fig. 4.

Subgroup analysis of the association between frailty score and COPD risk

Sensitivity analyses

Sensitivity analyses demonstrated the robustness of the association between frailty score and COPD risk across different analytical scenarios (Supplementary Figure S1 and Supplementary Table S1). When excluding participants aged ≥ 65 years, the association remained significant (α = 2.10, 95% CI: 1.21–2.99, P < 0.001). Excluding current smokers yielded similar results (α = 2.38, 95% CI: 0.98–3.77, P < 0.001), as did excluding obese participants with BMI ≥ 30 kg/m² (α = 2.33, 95% CI: 1.64–3.02, P < 0.001). To address potential reverse causation, we excluded events occurring within the first 2 years of follow-up, which strengthened the association (α = 2.52, 95% CI: 1.80–3.24, P < 0.001). Restricting the analysis to participants with ≥ 5 years of follow-up maintained the significant association (α = 2.10, 95% CI: 1.37–2.83, P < 0.001). Finally, excluding participants with extreme frailty scores confirmed the robustness of our findings (α = 2.33, 95% CI: 1.64–3.02, P < 0.001). All sensitivity analyses yielded statistically significant associations, confirming the consistency and reliability of the observed relationship between frailty and COPD risk.

Discussion

This large-scale prospective cohort study provides longitudinal evidence that frailty progression is dynamically associated with incident COPD in middle-aged and older adults. Through joint modeling to simultaneously analyze frailty trajectories and risk of COPD, it was demonstrated that an increasing frailty trajectory is strongly associated with incident COPD. This association remained robust across multiple sensitivity analyses and consistent across diverse population subgroups, suggesting that frailty progression may serve as an early dynamic marker of respiratory vulnerability in aging populations.

Our findings align with and extend previous studies that have identified associations between frailty and respiratory diseases [19, 20]. However, this study represents a large-scale longitudinal investigation showing that frailty progression is associated with incident COPD diagnosis. The observed association suggests that frailty trajectories may provide additional dynamic information beyond conventional baseline risk factors, highlighting the clinical relevance of frailty as a risk-stratification marker. It is important to note that the association parameter derived from our joint model should not be interpreted as a conventional static hazard ratio. Instead, it serves as a mathematical link quantifying the strength of the dynamic relationship between the latent underlying frailty process and the instantaneous hazard of incident COPD. The strong positive association parameter observed in our study clinically indicates that an upward continuous shift in a participant’s frailty trajectory is significantly and continuously coupled with an elevated instantaneous hazard of incident COPD over time. This metric emphasizes the dynamic progression of physiological decline rather than comparing fixed baseline risk multipliers. Recent evidence from the English Longitudinal Study of Ageing has similarly demonstrated that preserved ratio impaired spirometry and COPD are associated with accelerated frailty progression, supporting the bidirectional nature of this relationship [21].

Previous studies have primarily focused on frailty because of established COPD. Our longitudinal design with careful exclusion of prevalent COPD cases and early events (sensitivity analysis excluding events within 2 years) supports a temporal association between frailty progression and subsequent COPD diagnosis, although the possibility that subclinical respiratory decline contributed to worsening frailty cannot be fully excluded. The strengthened association after excluding early events further supports this temporal relationship. This approach is consistent with recent methodological advances in frailty research that emphasize the importance of longitudinal assessment to capture the dynamic nature of frailty transitions [22].

Several plausible biological mechanisms may partly underlie the observed association between frailty progression and incident COPD, although these pathways were not directly tested in the present study. First, frailty is characterized by chronic low-grade inflammation, with elevated levels of inflammatory markers such as C-reactive protein, interleukin-6, and tumor necrosis factor-α [23]. While not directly measured in our study, this systemic inflammatory state is hypothesized to play a role in airway inflammation and remodeling, key processes in COPD pathogenesis [24]. Second, frailty-associated muscle weakness and reduced physical activity may be linked to respiratory muscle deconditioning, impaired lung function, and increased susceptibility to respiratory infections. Third, frailty often coexists with malnutrition and oxidative stress, which are associated with compromised immune function and antioxidant defenses in the respiratory system. Recent systematic reviews have identified these inflammatory biomarkers as promising targets for early frailty detection and intervention [25]. It should be noted that these proposed mechanisms remain speculative in the context of the present study, as no direct biological validation was performed.

Our findings have important implications for COPD prevention and early intervention strategies. The identification of frailty as a potential risk marker of COPD risk suggests that frailty assessment could be incorporated into routine clinical practice for risk stratification. Validated frailty screening tools, such as the Fried phenotype or Clinical Frailty Scale, could help identify high-risk individuals who would benefit from targeted interventions [26]. However, recent evidence suggests that screening tools may need to be tailored for specific populations, as the Hospital Frailty Risk Score showed limited sensitivity in detecting frailty among hospitalized COPD patients compared to bedside clinical assessments [27].

Moreover, frailty is potentially modifiable through multicomponent interventions including exercise training, nutritional supplementation, and comprehensive geriatric assessment [28]. Our results suggest that such interventions might not only improve functional outcomes but also help identify or modify trajectories of respiratory vulnerability, although interventional studies are needed to determine whether improving frailty can reduce COPD incidence. This represents a paradigm shift from traditional COPD prevention strategies that focus primarily on smoking cessation and environmental exposure reduction. Recent advances in frailty intervention research have demonstrated that individually tailored multifactorial interventions delivered by multidisciplinary teams can effectively improve frailty status and maintain physical function [26].

The subgroup analyses revealed interesting patterns that warrant discussion. The stronger association observed in women compared to men may reflect sex-specific differences in frailty presentation, hormonal influences, or differential exposure to risk factors. Recent multimorbidity network analyses have shown that frail females have higher prevalence of obesity, hypertension, COPD, and diabetes compared with frail males, suggesting distinct pathophysiological pathways [29]. The enhanced association in participants with cardiovascular disease supports the concept of multimorbidity clustering and shared pathophysiological pathways between frailty, cardiovascular disease, and COPD. Studies in populations with HIV have demonstrated that cerebrovascular disease, diabetes, COPD, or liver disease co-occurring with multimorbidity substantially is associated with substantially increased frailty hazards [30].

Interestingly, the survival sub-model revealed that overweight and obesity were associated with a reduced risk of incident COPD. While seemingly counterintuitive, this finding aligns with the well-documented obesity paradox in respiratory medicine [22]. This inverse association may be driven by several methodological and biological factors. Methodologically, this result is highly susceptible to survival bias, as susceptible individuals are often depleted from the cohort early on. Obese individuals, particularly smokers, face a substantially higher risk of premature cardiovascular mortality. Therefore, obese participants who survived long enough to be included in our longitudinal analysis might represent a robust subpopulation, potentially lowering their observed COPD incidence. Furthermore, low BMI strongly overlaps with the weight loss component of our frailty score. This overlap creates a complex interaction where low weight may reflect cachexia as an early systemic manifestation of subclinical COPD, rather than merely acting as a baseline risk factor [31]. Biologically, overweight status may indicate a better nutritional reserve that protects against the severe muscle wasting commonly seen in the emphysematous phenotype of COPD [32]. Despite this paradox, the detrimental association between frailty progression and COPD risk remained consistent across all BMI strata, highlighting that frailty involves distinct pathophysiological pathways independent of obesity-related mechanisms [33].

This study has several strengths, including a large sample size, an extended follow-up period, and comprehensive sensitivity analyses. Furthermore, the use of joint modeling represents a methodological advance that captures the dynamic nature of frailty, allowing the dynamic association between repeated frailty measurements and incident COPD to be modeled more appropriately than with baseline-only survival analyses.

However, several limitations should be acknowledged. First, COPD diagnosis relied on algorithmic definitions from hospital records and death certificates without spirometric confirmation. Consequently, we likely missed undiagnosed or mild cases, restricting our findings primarily to clinically symptomatic or exacerbation-prone COPD, though this algorithmic approach likely improves specificity for clinically recognized COPD cases. Second, frailty was assessed using available UK Biobank variables rather than standardized clinical instruments. Notably, walking speed was determined by self-reported pace. While this composite score has been previously validated, the lack of objective physical performance quantification may limit assessment precision. Third, despite extensive covariate adjustment, residual confounding cannot be entirely excluded. Specifically, our adjustment for smoking relied on categorical status rather than cumulative intensity, meaning frailty might partially proxy for unmeasured smoking damage. Additionally, comprehensive environmental/occupational exposures, detailed medication histories, and nutritional status were not fully accounted for. Fourth, the UK Biobank’s “healthy volunteer” bias and lack of ethnic diversity limit generalizability [13]. The extremely low baseline frailty prevalence restricts statistical power for extensive subgroup analyses among the severely frail, though it implies our findings might underestimate the true frailty-driven COPD incidence in the broader population. Finally, the observational nature of this study precludes definitive causal inferences. Despite our robust longitudinal modeling, we cannot entirely rule out reverse causation—where subclinical respiratory decline accelerates frailty progression prior to formal COPD diagnosis.

Conclusions

This large-scale longitudinal study shows that progressive frailty is dynamically associated with incident COPD. These findings support the potential value of repeated frailty assessment for COPD risk stratification and early identification of respiratory vulnerability, while further studies are needed to determine whether interventions targeting frailty can reduce COPD incidence.

Abbreviations

AUC

Area Under the Curve

BMI

Body Mass Index

CI

Confidence Interval

COPD

Chronic Obstructive Pulmonary Disease

CVD

Cardiovascular Disease

DIY

Do It Yourself

EM

Expectation-Maximization

FEV1

Forced Expiratory Volume in one second

HR

Hazard Ratio

ICD-10

International Classification of Diseases, 10th Revision

SHS

Suboptimal Health Status

Authors’ contributions

Qi Sun : Conceptualization; Methodology; Writing-original draft; Writing-review editing. Jinxuan Li: Data curation; Writing-original draft; Writing-review editing. Li Yu: Data curation; Methodology; Writing-review editing. Han Lv: Project administration; Conceptualization; Writing-review editing.

Funding

This work was supported by the National Natural Science Foundation of China (No: 52500256) and Program of the National Research Institute for Family Planning (No: 2024GJPY03).

Data availability

The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request.

Declarations

Ethics approval and consent to participate

This study was approved by the Northwest Ethics Committee with approval number 21/NW/0157. All participants provided written informed consent. The study was conducted in accordance with the ethical principles of the Declaration of Helsinki. All participants were informed about the purpose of the study, assured of confidentiality, and provided written consent prior to participation. Participation was voluntary, and respondents could withdraw at any time without consequence.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.

Competing interests

The authors declare that they have no competing interests.