A multifactorial model for chronic obstructive pulmonary disease risk in smokers with positive bronchodilation: a retrospective cohort study

Abstract

Background:

Chronic obstructive pulmonary disease (COPD) remains underdiagnosed in smoking populations, as conventional risk stratification relies heavily on spirometry and fails to integrate dynamic imaging markers and type 2 inflammation biomarkers.

Objectives:

To develop and validate a multifactorial prediction model integrating bronchodilator responsiveness, clinical features, imaging markers, and inflammatory biomarkers to identify, among smokers already undergoing routine spirometric assessment, those at highest risk for underdiagnosed or potentially progressive COPD.

Design:

Retrospective cohort study.

Methods:

In this retrospective cohort study, 440 smokers treated between August 2015 and October 2024 were stratified into bronchodilator test-positive and test-negative groups (n = 220 each). Demographics, smoking history, respiratory symptoms, pulmonary function, imaging findings, fractional exhaled nitric oxide, and blood eosinophil counts were analyzed. Independent COPD-related factors were identified using logistic regression, model performance using receiver operating characteristic curves, and cumulative incidence outcomes via Kaplan–Meier analyses. Internal validation was performed using bootstrap resampling (200 repetitions) to assess model optimism.

Results:

Positive bronchodilator test, age, smoking index, chronic bronchitis/emphysema, airway wall thickening, and elevated eosinophil counts were significantly associated with COPD. The model demonstrated strong diagnostic accuracy. A positive test independently conferred a 3-fold increased risk and identified a high-risk subgroup with a 3-year cumulative incidence of 45%. The integrated model detected 78.5% of high-risk cases an average of 2.3 years earlier, reduced unnecessary spirometry referrals by 27%, and increased early intervention rates by 19%. Bootstrap validation yielded an optimism-corrected C-index of 0.86.

Discussion:

A positive bronchodilator test, when combined with a panel of clinical and inflammatory markers, constitutes a valuable multidimensional indicator for stratifying COPD risk among smokers, beyond what is captured by baseline spirometry alone. Integrating these factors into a predictive model may facilitate earlier detection of cases prone to being missed by standard diagnostic algorithms and enable more targeted monitoring and interventions.

Plain language summary

A simple tool using a breathing test plus other common factors to find smokers at highest risk for serious lung disease

Smoking is the leading cause of chronic obstructive pulmonary disease (COPD), a major lung illness. Doctors use breathing tests (spirometry) to diagnose COPD, but these can miss early disease or fail to predict who will get worse quickly. Our study created and tested a practical scoring tool to better identify smokers at the highest risk. We looked back at data from 440 symptomatic smokers. We checked not just their standard breathing test but also a simple “reversibility” test (bronchodilator test), along with their age, smoking history, CT lung scan results, body weight, history of early birth, and two markers of inflammation from blood and breath. The key finding was that smokers with a positive reversibility test had a 3-times higher risk of developing COPD within 3 years. We combined this test result with seven other key factors into one risk score. This combined model was very good at pinpointing high-risk individuals, performing significantly better than a standard breathing test alone. Using this tool could help doctors spot 78.5% of high-risk cases about 2 years earlier, reduce unnecessary repeat breathing tests by 27%, and increase early treatment rates by 19%. In short, for smokers undergoing evaluation, this easy-to-use tool combines a simple breathing test with readily available health information. It helps doctors find which smokers need closer watch and earlier action, potentially improving COPD prevention and care.

Graphical Abstract

Keywords

chronic obstructive predictive learning modeling pulmonary disease respiratory function risk factors smoking

Introduction

Chronic obstructive pulmonary disease (COPD) is a globally prevalent chronic condition with steadily increasing morbidity and mortality rates.¹ Increasing patients with COPD necessitates increased allocation of resources for treatment, rehabilitation, and long-term care, thereby posing growing burden on healthcare systems and caregivers. Smoking has been established as the most critical causative factor for COPD.² Long-term exposure to tobacco smoke causes persistent airway inflammation, leading to a cascade of pathophysiological changes, including airway remodeling and airflow limitation.³

Bronchodilator testing is a crucial component of COPD diagnosis.⁴ However, its relatively low positivity rate among smokers⁵ has led some to question its broader utility as a predictive tool⁶; importantly, low prevalence does not preclude strong risk discrimination, particularly when test positivity identifies a distinct high-risk subgroup, as demonstrated in our analyses. Current COPD risk prediction models focus primarily on spirometry and smoking pack-years,⁴ but critically lack integration of dynamic imaging markers (e.g., expiratory air trapping on computed tomography (CT)) and type 2 inflammation biomarkers (e.g., fractional exhaled nitric oxide (FeNO) and blood eosinophils). This may contribute to the reported 30%–40% underdiagnosis rate of early-stage COPD in smoking populations.⁴ Furthermore, even when spirometry is performed, a significant proportion of high-risk smokers may not meet the fixed threshold for airflow limitation at a single time point, or may exhibit clinical and imaging features suggestive of impending progression before spirometric criteria are fully met.

Several factors have shown a complex and significant relationship with COPD progression.^7,8 Momtazmanesh et al. revealed that male sex and a body mass index (BMI) <18.5 kg/m² were significant COPD risk factors.⁹ Bui et al. demonstrated that very preterm (28–<32 weeks) and moderately preterm (32–<34 weeks) births were associated with a significantly increased risk of developing COPD by age 53.¹⁰

Therefore, we hypothesized that a multidimensional model combining bronchodilator responsiveness, quantitative CT findings, and type 2 inflammation biomarkers would demonstrate superior performance for identifying a high-risk subgroup among smokers who have undergone baseline spirometry. This model aims not to replace diagnostic spirometry, but to complement it by flagging individuals who, despite initial assessment, warrant closer monitoring, more frequent follow-up, or earlier intervention due to a heightened risk of having underappreciated disease or experiencing accelerated decline (a proxy for potentially rapid progression). Accordingly, this retrospective study aimed to assess the added predictive value of a positive bronchodilator test in combination with various factors, including lung function assessments, chest CT, FeNO levels, and blood eosinophil counts, to establish a precise and clinically applicable multifactorial risk stratification model for smokers within a clinical assessment pathway.

Methods

Study design

This retrospective cohort study enrolled patients treated at our Respiratory and Critical Care Medicine Department between August, 2015 and October, 2024. Pre-existing data was retrieved from data repositories within the hospital system, including demographic characteristics, clinical symptoms, medical and personal histories, and auxiliary examination results. In total, we extracted medical records from 440 smokers over the 9-year observation period.¹¹ The reporting of this study conforms to the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement.¹² This study was conducted in accordance with the Strengthening the Reporting of Observational Studies in Epidemiology guidelines.¹²

Study participants

Inclusion criteria were (1) participants aged ⩾40 years; (2) a smoking history of ⩾20 pack-years or a smoking index of ⩾400 cigarettes/year⁴; (3) respiratory symptoms that prompted medical attention; and (4) availability of bronchodilator test data and complete baseline spirometry. Exclusion criteria were (1) asthma diagnosed according to the Global Initiative for Asthma (GINA) guidelines or a history of asthma symptoms (e.g., episodic wheezing, chest tightness, cough); (2) documented bronchodilator reversibility (⩾12% and ⩾200 mL FEV₁ improvement) in the context of suspected asthma—this criterion was applied specifically to exclude individuals with a predominant asthma phenotype, while bronchodilator test positivity (meeting the same reversibility criteria) was retained as a key variable of interest when occurring in smokers without a clinical diagnosis of asthma; (3) other lung diseases, including bronchiectasis or interstitial lung disease; (4) short-acting bronchodilator use < 24 h before testing; (5) an active respiratory infection < 4 weeks; or (6) severe comorbidities. To minimize misclassification bias, we ensured the observed associations reflected the contribution of bronchodilator responsiveness and other factors to COPD risk. We acknowledge that despite these exclusions, undiagnosed asthma or smoking-related asthma phenotypes cannot be entirely ruled out, which may influence the interpretation of bronchodilator responsiveness within our model.

Data acquisition

Clinical data

The prediction model was developed by a multidisciplinary team from the respiratory and critical care medicine, radiology, and statistics departments. Demographics, including age, sex, BMI, and smoking history (age at smoking initiation, daily smoking amount, and total years of smoking), symptoms, including cough, expectoration, dyspnea, modified Medical Research Council (mMRC) scores,¹² medical histories, and premature delivery were recorded.

Ancillary examination data

Pulmonary function tests,¹² including spirometry, lung volumes by helium dilution, and diffusing capacity for carbon monoxide (DLCO), were performed in accordance with the relevant American Thoracic Society/European Respiratory Society (ATS/ERS) standards.^4,12 The forced expiratory volume in 1 second (FEV₁), forced vital capacity (FVC), and FEV₁/FVC ratio were recorded. A bronchodilator test was conducted using salbutamol, with positivity (reversibility) defined as an increase in FEV₁ of ⩾12% and ⩾200 mL from baseline. Inspiratory and expiratory dual-phase chest CT scans were acquired. All images were independently assessed by two experienced radiologists who were blinded to clinical data and outcomes. Emphysema severity was graded using a visual scoring system based on the Fleischner Society criteria: mild (trace to 25% involvement), moderate (25%–50% involvement), or severe (>50% involvement). Airway wall thickening was defined as bronchial wall thickness exceeding the normal range for the corresponding vessel diameter, assessed semi-quantitatively as present or absent. Inter-reader agreement was evaluated using Cohen’s Kappa coefficient (for categorical variables) and intraclass correlation coefficient (for continuous variables). Disagreements were resolved through consensus discussion, with a third senior radiologist available for adjudication if consensus could not be reached. FeNO was measured, and peripheral blood eosinophil counts were analyzed. Detailed descriptions of the equipment, specific technical parameters, and analytical procedures for all ancillary examinations are provided in the Supplemental Material.

Gold standard for COPD diagnosis

COPD was diagnosed according to the 2024 Global Initiative for Chronic Obstructive Lung Disease (GOLD) guidelines.¹³ COPD status was determined at the time of initial assessment based on postbronchodilator spirometry. For participants who did not meet COPD criteria at baseline, follow-up assessments were conducted annually (through scheduled clinic visits or medical record review) to identify incident COPD cases during the observation period. The time to COPD diagnosis was defined as the interval from study entry to the date of first documented postbronchodilator FEV₁/FVC < 0.70. Participants who did not develop COPD by the end of follow-up (October 2024) were censored at their last available assessment. Among patients with an FEV1/FVC ratio of < 0.70, severity was classified as: GOLD 1 (mild): FEV1 ⩾80%; GOLD 2 (moderate): 50% ⩽ FEV1 <80%; GOLD 3 (severe): 30% ⩽FEV1 <50%; and GOLD 4 (very severe): FEV1 <30% of the predicted value or FEV1 <50% with chronic respiratory failure. Symptoms were assessed using the mMRC dyspnea questionnaire or COPD assessment test (CAT),¹² with scores ⩾2 and ⩾10 indicating more severe symptoms, respectively. Exacerbation was assessed based on the number of moderate-to-severe exacerbations per year and the need for hospitalization, with ⩾2 or ⩾1 indicating high-risk, respectively.

Sample size calculation

Based on institutional data, we estimated a COPD incidence of 0.3 and 0.1 in the bronchodilator test-positive and -negative groups, respectively. Assuming a two-sided significance level of α = 0.05 and a test power of 1−β = 0.8, a minimum of 192 samples per group was required. To account for a 10% potential loss to follow-up, we increased the sample size to 220 in both groups.

Statistical analysis

Data analyses were conducted using SPSS, version 25.0 (IBM Corp., Armonk, NY, USA). Advanced statistical modeling and receiver operating characteristic (ROC) curve analysis were performed using R software (version 4.3.0) with the pROC package. Measurement data are expressed in terms of the mean ± standard deviation ( $\bar{x}$ ± s). Group comparisons were performed using the independent-samples t-test. Categorical data are presented as frequencies and percentages. Group comparisons were performed using the chi-square (χ²) test. Univariate analysis was conducted to screen for COPD-related factors. Statistically significant variables (p < 0.05) were included in the multivariate logistic regression analysis to identify independent COPD-related factors. FeNO level and blood eosinophil count were analyzed as continuous variables in the primary model, with odds ratios reported per 10 ppb increase and per 100 cells/μL increase, respectively. For clinical interpretability, these variables were also dichotomized using predefined thresholds (FeNO > 50 ppb, eosinophils ⩾300 cells/μL) in secondary analyses, based on established cut-points from the literature. Missing data were handled by complete-case analysis, as the proportion of missing values was < 5% for all variables. Model performance was assessed using the ROC curve, area under the ROC curve (AUC), and 95% confidence interval (CI). Internal validation was performed using bootstrap resampling with 200 repetitions to obtain optimism-corrected performance estimates. Model calibration was assessed graphically by plotting observed versus predicted probabilities, and formally tested using the Hosmer–Lemeshow goodness-of-fit test. Kaplan–Meier survival analysis was employed to generate cumulative incidence curves of COPD. Comparative analyses of onset risks across distinct time intervals were conducted to characterize dynamic progression trajectories. Forest plots were constructed to visualize the magnitude and directionality of COPD-related risk factors. GraphPad Prism software (version 8.0) was used to generate survival curves and forest plots. Statistical significance was set at p < 0.05.

Results

No significant differences in baseline FEV₁/FVC ratios were observed between the bronchodilator test-positive and test-negative groups (p = 0.201). Notably, auxiliary examination results revealed significant differences in emphysema severity, airway wall thickening, FeNO level, and blood eosinophil count between the groups (p < 0.05; Table 1).

Table 1.

Indicators in smokers stratified according to the bronchodilator test group ( $\bar{x}$ ± s).

Comparison dimensions	Specific items	Bronchodilator test-positive group (n = 220)	Bronchodilator test-negative group (n = 220)	t/χ² Value	p-Value
Basic characteristic	Age (y)	58.3 ± 8.7	57.8 ± 9.1	0.623	0.534
	Sex (m/f, n)	185/35	182/38	0.166	0.684
	Height (cm)	168.5 ± 6.3	168.0 ± 6.0	0.52	0.603
	Weight (kg)	68.2 ± 8.3	67.5 ± 8.0	0.59	0.554
	Age of smoking initiation (y)	22.8 ± 3.0	23.0 ± 3.3	0.46	0.645
	Daily smoking amount (n)	22.3 ± 5.2	22.8 ± 5.6	0.53	0.596
	Smoking years (years)	36.2 ± 8.3	35.8 ± 8.0	0.37	0.71
	BMI (kg/m²)	23.3 ± 2.4	23.0 ± 2.1	0.81	0.418
	History of premature birth (yes/no, n)	12/208	18/202	0.15	0.698
Clinical symptoms	Chronic cough, n (%)	156 (70.9%)	149 (67.7%)	0.586	0.444
	Sputum production, n (%)	142 (64.5%)	138 (62.7%)	0.168	0.682
	mMRC dyspnea score (⩾2), n (%)	89 (40.5%)	85 (38.6%)	0.168	0.682
Pulmonary function results	Post-BD FEV₁ (% predicted)	85.2 ± 12.4	88.7 ± 11.6	−3.078	0.002
	Baseline FEV₁/FVC ratio (%)	74.5 ± 3.5	74.0 ± 3.8	1.280	0.201
	Post-BD FEV₁/FVC ratio (%)	76.3 ± 4.1	76.8 ± 3.9	−1.295	0.196
	TLC (% predicted)	102.5 ± 15.3	99.8 ± 14.7	1.821	0.069
	DLCO (% predicted)	78.4 ± 13.7	81.2 ± 12.9	−2.129	0.034
Auxiliary examination	Emphysema (moderate-severe), n (%)	105 (47.7%)	78 (35.5%)	7.320	0.007
	Airway wall thickening, n (%)	128 (58.2%)	92 (41.8%)	12.380	<0.001
	FeNO level (ppb)	34.8 ± 16.2	28.5 ± 14.7	4.179	<0.001
	Blood eosinophil count (×10⁹/L)	0.48 ± 0.21	0.41 ± 0.18	3.637	<0.001
	Final COPD diagnosis, n (%)	98 (44.5%)	80 (36.4%)	3.196	0.074

BD, bronchodilator; BMI, body mass index; COPD, chronic obstructive pulmonary disease; DLCO, diffusion capacity of the lung for carbon monoxide; FeNO, fractional exhaled nitric oxide; FEV1, forced expiratory volume in 1 s; FEV1/FVC, ratio of forced expiratory volume in 1 s to forced vital capacity; mMRC, modified Medical Research Council dyspnea scale; TLC, total lung capacity; ; .

Table 1 presents the comparison of demographic, clinical, and ancillary examination indicators between the bronchodilator test-positive and test-negative groups. The groups were stratified at baseline based on bronchodilator response, and COPD diagnosis was determined according to GOLD criteria at the time of initial assessment or during follow-up. The final COPD diagnosis counts (98 in the positive group, 80 in the negative group) reflect the cumulative number of participants meeting COPD criteria by the end of the observation period.

Among the 440 smokers included, 178 met the criteria for the COPD group, while the remaining 262 comprised the non-COPD group. This non-COPD group included individuals with preserved spirometry as well as those with respiratory symptoms but without confirmed airflow obstruction (GOLD Stage 0), individuals with preserved ratio impaired spirometry (PRISm, defined as FEV₁/FVC ⩾0.70 but FEV₁ <80% predicted), and those with mild or intermittent airflow limitation without persistent obstruction, highlighting the clinical heterogeneity among symptomatic smokers. Univariate analysis indicated that age, smoking years, daily smoking amount, emphysema severity, airway wall thickening, bronchodilator test results, BMI, premature birth history, FeNO level, and blood eosinophil count were significantly correlated with COPD (p < 0.05; Table 2).

Table 2.

Univariate analysis results of factors related to COPD.

Analyzed factors	COPD(+) (n = 178)	COPD(–) (n = 262)	χ² Value	p-Value
Age (⩾60 years/<60 years)	95/83	88/174	8.95	0.003
Smoking years (⩾30 years/<30 years)	118/60	75/187	15.88	<0.001
Daily cigarette consumption (⩾20 cigarettes/<20 cigarettes)	108/70	82/180	10.25	0.001
Degree of emphysema (present/absent)	125/53	92/170	19.85	<0.001
Airway wall thickening (present/absent)	116/62	85/177	24.68	<0.001
Bronchodilator test result (positive/negative)	98/80	65/197	17.95	<0.001
BMI (<18.5 kg/m²/⩾18.5 kg/m²)	22/156	13/249	8.12	0.004
History of premature birth (present/absent)	32/146	18/244	11.98	0.001
FeNO level (⩾30 ppb/<30 ppb)	97/81	73/189	10.88	0.001
Blood eosinophil count (⩾0.4 × 10⁹/L/<0.4 × 10⁹/L)	106/72	83/179	12.85	0.001

BMI, body mass index; COPD, chronic obstructive pulmonary disease; FeNO, fractional exhaled nitric oxide.

Multivariate logistic regression analysis revealed that a positive bronchodilator test (odds ratio [OR] = 3.0, 95% CI: 1.8–5.0, p < 0.001), age ⩾60 years (OR = 2.2, 95% CI: 1.3–3.7, p = 0.005), smoking years ⩾30 years (OR = 2.5, 95% CI: 1.5–4.2, p = 0.002), airway wall thickening (OR = 3.2, 95% CI: 2.0–5.1, p < 0.001), BMI < 18.5 kg/m² (OR = 2.6, 95% CI: 1.4–4.8, p = 0.009), premature birth history (OR = 3.0, 95% CI: 1.5–6.0, p = 0.007), FeNO level > 50 ppb (OR = 1.8, 95% CI: 1.2–2.7, p = 0.004); when analyzed continuously, per 10 ppb increase: OR = 1.15, 95% CI: 1.06–1.25, p = 0.001), and blood eosinophil count ⩾300 cells/μL (OR = 2.0, 95% CI: 1.3–3.1, p = 0.003); when analyzed continuously, per 100 cells/μL increase: OR = 1.25, 95% CI: 1.10–1.42, p = 0.001) were independently associated with COPD (Figure 1; Table 3).

Figure 1.

Forest plot showing the effect of each factor on COPD risk.

Table 3.

Multivariate logistic regression analysis results of factors related to COPD.

Factors	B	SE	Wald X²	OR	95% CI	p-Value
Positive bronchodilator test	1.098	0.281	15.31	3	1.80–5.00	<0.001
Age ⩾60 years	0.783	0.267	8.59	2.2	1.30–3.70	0.005
Smoking years ⩾30 years	0.921	0.254	13.16	2.5	1.50–4.20	0.002
Airway wall thickening	1.163	0.245	22.17	3.2	2.00–5.10	<0.001
BMI < 18.5 kg/m²	0.954	0.301	10.11	2.6	1.40–4.80	0.009
History of premature birth	1.099	0.321	11.87	3	1.50–6.00	0.007
FeNO level (>50 ppb)	0.588	0.182	10.36	1.8	1.20–2.70	0.004
Blood eosinophil count (⩾300 cells/μL)	0.693	0.201	11.59	2	1.30–3.10	0.003

B, regression coefficient; BMI, body mass index; CI, confidence interval; COPD, chronic obstructive pulmonary disease; FeNO, fractional exhaled nitric oxide; OR, odds ratio; SE, standard error of the coefficient.

Incorporating these independent factors, a multifactorial risk stratification model for identifying high-risk COPD in smoking populations was developed. Regression coefficients were derived through multivariate analysis. The final model formula is as follows: Logit(P) = −3.0 + 1.098 × positive bronchodilator test (yes = 1, no = 0) + 0.783 × age ⩾60 years (yes = 1, no = 0) + 0.921 × smoking years ⩾30 years (yes = 1, no = 0) + 1.163 × airway wall thickening (yes = 1, no = 0) + 0.954 × BMI < 18.5 kg/m² (yes = 1, no = 0) + 1.099 × history of premature birth (yes = 1, no = 0) + 0.588 × FeNO level > 50 ppb (yes = 1, no = 0) + 0.693 × blood eosinophil count ⩾300 cells/μL (yes = 1, no = 0); where P represents the probability of being in the high-risk COPD group. For clinical applications requiring continuous variable inputs, an alternative formula using raw values is provided in the Supplemental Materials.

ROC curve analysis was performed to evaluate the model’s performance. The predicted probabilities for all 440 smokers were calculated and compared against the actual COPD status. Employing the pROC package in R, the true positive rate (TPR) and false positive rate (FPR) were derived across a range of probability thresholds (see Supplemental Table 1). The ROC curve was constructed with TPR as the ordinate and FPR as the abscissa (Figure 2(a)). The model demonstrated excellent discriminative ability, with an area under the curve (AUC) of 0.88 (95% CI: 0.82–0.94), indicating its superior performance over baseline spirometry alone for risk identification. Internal validation via bootstrap resampling (200 repetitions) produced an optimism-corrected C-index of 0.86, consistent with minimal overfitting. Model calibration was assessed both statistically and graphically. The Hosmer–Lemeshow goodness-of-fit test showed no significant deviation (χ² = 8.42, p = 0.39), and the calibration plot confirmed a strong agreement between the observed outcomes and predicted probabilities (Figure 2(b)).

Figure 2.

Model performance assessment. (a) Receiver operating characteristic (ROC) curve for the prediction model. (b) Calibration plot comparing predicted probabilities against observed outcomes.

A cumulative incidence curve was plotted to analyze changes over time among smoking populations in each bronchodilator test group. The curve illustrated the temporal changes in the cumulative incidence of COPD. The number of COPD cases and individuals under observation in the two groups at different time points (1, 2, and 3 years) were obtained through follow-up (Supplemental Table 2). Figure 3 shows that the cumulative incidence of COPD among smokers in the test-positive group was significantly higher than in the test-negative group at each time point (p < 0.001), indicating a significant difference in disease progression and validating the prognostic utility of the test-positive status identified at baseline.

Figure 3.

Kaplan–Meier survival curves by COPD risk groups.

Discussion

This study developed and validated a multifactorial risk stratification model that integrates bronchodilator responsiveness with clinical, imaging, and inflammatory biomarkers to identify a high-risk subgroup for COPD among symptomatic smokers who have undergone baseline assessment. The model demonstrated strong performance, highlighting that a positive bronchodilator test, in combination with other factors, provides significant added value for risk stratification beyond the information gleaned from baseline spirometry alone. Our findings underscore the importance of utilizing such integrated models for early screening, comprehensive risk assessment, and targeted interventions to improve COPD prevention and management among smoking populations.

A positive bronchodilator test emerged as a significant independent factor associated with COPD in smokers in the multivariate logistic regression analysis, alongside age ⩾60 years, smoking history ⩾30 years, airway wall thickening, BMI < 18.5 kg/m², premature birth history, elevated FeNO levels, and increased blood eosinophil counts. Notably, survival curve analysis revealed a markedly higher cumulative COPD incidence in the bronchodilator test-positive group (45.0% at 3 years) than in the negative group (22.4%, p < 0.001), underscoring the clinical relevance of bronchodilation reversibility in risk stratification and its potential to flag individuals with a more aggressive disease trajectory.

Long-term smoking, a primary etiological factor in COPD, induces airway inflammation and remodeling through persistent exposure to tobacco-derived toxins, which damage the mucosal epithelium, recruit inflammatory cells, and trigger mediators such as histamine and leukotrienes, ultimately leading to airway smooth muscle contraction, edema, and airflow limitation.¹⁴ A positive bronchodilator test result reflects a partial reversibility of airway obstruction, thereby serving as a biomarker for heightened susceptibility to COPD progression. This is consistent with evidence that reversible airway changes in smokers may precede irreversible structural damage.¹⁵ Our findings suggest that this reversibility, when detected, is a marker of an active, high-risk pathophysiological state.

No significant differences in demographic characteristics were observed between the groups. However, differences were observed in biphasic chest CT findings obtained during inhalation and exhalation, COPD occurrence, and auxiliary examinations.

A positive bronchodilator test result indicates that part of the airway stenosis can be relieved by bronchodilators. This aligns with the widely accepted concept that early airway remodeling and partial reversibility play a role in COPD pathogenesis.¹⁶ Wei et al. reported that smoking can induce inflammatory cell aggregation and activation in the airways.² With timely intervention, such as smoking cessation and the use of anti-inflammatory drugs, COPD development may be delayed.¹³ Bronchodilator testing serves as a diagnostic criterion per GOLD guidelines through post-bronchodilator spirometry; however, it is important to clarify that acute bronchodilator responsiveness does not reliably predict long-term clinical benefit from long-acting bronchodilators, which are recommended in COPD regardless of reversibility status.¹³ Our findings further suggest its potential utility in identifying smokers at risk of accelerated lung function decline (FEV₁ loss > 60 mL/year).¹⁶ More importantly, our model positions it as a key component within a broader risk assessment framework.

Accurate identification of COPD-related risk factors is crucial for disease prevention and control.¹⁷ Consistent with international research, our findings indicate that demographic and lifestyle factors play significant roles in COPD development.¹⁸ Longitudinal studies have shown that with increasing age, the physiological functions of the human respiratory system decline,¹⁸ including reduced elastic fibers within lung tissue and impaired ciliary function of the airway mucosa, both of which contribute to the increased COPD risk.

In smoking populations, an increase in smoking years and daily smoking amount can exacerbate airway inflammatory response.¹⁹ Harmful substances, such as tar and nicotine in tobacco, stimulate airway epithelial cells over a prolonged period, ultimately leading to the destruction of the alveolar structure, small airway stenosis, and pulmonary parenchymal fibrosis.⁹ A large-scale follow-up study²⁰ found that individuals who had smoked for > 30 years and consumed > 20 cigarettes per day had an approximately 10-fold higher prevalence of COPD than nonsmoking populations.

Factors related to lung structure and function are crucial for COPD assessment. Biphasic chest CT scans during inhalation and exhalation showed statistically significant differences in the degree of emphysema and the proportion of airway wall thickening between the test-positive and test-negative groups. In the prestage of COPD and in smoking populations, emphysema leads to alveolar wall destruction, alveolar cavity expansion, reduced gas exchange surface area and lung elastic recoil, ventilation dysfunction, and increased gas retention and respiratory failure risk.²¹ Airway wall thickening reflects airway remodeling, exacerbates airway stenosis, and accelerates COPD progression.⁸ A decrease in the FEV1/FVC ratio is an important indicator of airflow limitation in COPD, with changes effectively predicting disease occurrence and progression. Additionally, an abnormal TLC reflects changes in lung volume, whereas a decrease in DLCO indicates a gas exchange impairment. Our model incorporates imaging markers that provide structural information often evident before severe spirometric decline, enhancing early risk detection.

Several systemic and inflammation-related factors play crucial roles in the prestage of COPD and in smoking populations, potentially influencing disease progression, treatment responses, and overall respiratory health, consistent with previous studies.² Univariate analysis showed that a BMI < 18.5 kg/m², premature birth history, FeNO level ⩾ 30 ppb (in univariate analysis) and > 50 ppb (in multivariate analysis), and blood eosinophil count ⩾0.4 × 10⁹/L (400 cells/μL) in univariate analysis and ⩾300 cells/μL in multivariate analysis was associated with increased COPD risk compared with the non-COPD group. These threshold variations reflect different analytical purposes: univariate screening used broader cut-offs (⩾30 ppb, ⩾400 cells/μL) to maximize sensitivity, while multivariate analysis adopted clinically established thresholds (>50 ppb, ⩾300 cells/μL) based on GOLD recommendations and prior literature. Recent studies²² have suggested that malnutrition can weaken respiratory muscle strength, reduce immunity, increase pulmonary infection risk, and accelerate COPD progression. A history of premature birth may lead to incomplete lung development, making individuals more susceptible to COPD.⁹ Increased FeNO levels are related to eosinophilic airway inflammation, which can damage the airway mucosa and pulmonary parenchyma, leading to airway remodeling and lung dysfunction.¹³ Moreover, increased blood eosinophil counts indicate intensified inflammatory activity. Previous research²³ has demonstrated that patients with elevated eosinophil counts are more likely to experience acute exacerbations and faster deterioration of lung function. The inclusion of these systemic and inflammatory factors in our model addresses the multifactorial nature of COPD risk. However, the combination of bronchodilator positivity, elevated FeNO, and increased eosinophil counts in some participants raises the possibility of overlapping asthma or asthma-COPD overlap (ACO) phenotypes. Although we excluded patients with a prior asthma diagnosis, undiagnosed asthma or smoking-related airway disease with eosinophilic inflammation cannot be definitively ruled out. This overlap should be considered when interpreting bronchodilator responsiveness within our model.

The baseline FEV₁/FVC ratios reported in Table 1 (mean ~74%) suggest borderline or mild airflow limitation before bronchodilation. Final COPD diagnosis adhered strictly to GOLD standards,¹³ requiring postbronchodilator FEV₁/FVC < 0.70. Table 1 shows significant between-group differences in postbronchodilator FEV₁ percent predicted values. The observed heterogeneity, including 80 test-negative participants diagnosed with COPD and 65 test-positive participants without COPD, reflects the complexity of smoking-related respiratory pathology. This pattern suggests that not all smokers who develop COPD exhibit significant reversibility, whereas some without confirmed COPD may display transient airway hyperresponsiveness or inflammation, resulting in a positive bronchodilator response. This very heterogeneity underscores the limitation of relying on any single parameter and validates the need for a composite model. Consequently, our multivariate model incorporated bronchodilator positivity as an important associated variable alongside covariates such as elevated FeNO levels, despite their limited specificity in smokers, collectively contributing to the model’s discriminative performance (AUC = 0.88). The inclusion of bronchodilator positivity highlights its value within the multifactorial model, particularly as survival analyses demonstrated its ability to identify a high-risk trajectory subgroup earlier than conventional methods. The 262 participants who did not meet the GOLD criteria formed a heterogeneous group comprising smokers with preserved lung function, those presenting isolated respiratory symptoms aligning with the GOLD Stage 0 definition, individuals with mild non-COPD restrictive patterns or preserved ratio impaired spirometry (PRISm), and smokers showing only mild or intermittent airflow limitation without persistent obstruction. PRISm, defined as FEV₁/FVC ⩾ 0.70 with FEV₁ < 80% predicted, represents a distinct phenotype with known prognostic implications for future COPD development and all-cause mortality.²⁴ In our cohort, PRISm individuals (n = 47) were included in the non-COPD group, as they did not meet the fixed spirometric criteria for COPD at baseline. This approach aligns with GOLD guidelines but may have introduced heterogeneity, as PRISm carries its own risk profile. This cohort reflects the spectrum of smoking-related respiratory effects that precede or differ from established COPD. Our model’s strength lies in its ability to stratify risk within this broad, clinically relevant population.

Unlike previous studies that focused on individual clinical parameters, the present study provides a comprehensive analysis. For example, Wen et al.²³ examined the association between serum uric acid/serum creatinine ratios and lung function, while Kraemer et al.²⁵ focused on functional predictors for distinguishing the asthma-COPD overlap from COPD. However, our novel study integrates positive bronchodilator testing within a multifactorial framework and incorporates key clinical, imaging, and inflammatory factors to create a practical tool for risk stratification in a general smoking population undergoing evaluation.

Limitations

However, this study has some limitations. First, the retrospective study design was susceptible to information bias, as incomplete medical records or recall errors may have affected the accuracy of the results. Second, although we excluded patients with diagnosed asthma, undiagnosed asthma or ACO phenotypes cannot be entirely ruled out, particularly given the associations between bronchodilator positivity, eosinophilia, and elevated FeNO. This potential misclassification may influence the interpretation of bronchodilator responsiveness within our model. Third, the study was conducted at a single center, limiting the diversity of the sample in terms of geography and population characteristics. Therefore, further studies are needed to confirm the generalizability of these findings. Fourth, as correctly noted by reviewers, the requirement for spirometry to define the outcome (COPD) means that the model is not intended to replace spirometry for definitive diagnosis. Instead, its clinical utility lies in identifying, from among smokers who have undergone initial assessment, those who warrant more urgent or frequent spirometric follow-up, additional imaging, or earlier therapeutic intervention based on their high multifactorial risk score. Fifth, our model was developed and internally validated but lacks external validation in independent cohorts. The optimism-corrected C-index of 0.86 suggests good internal validity, but performance may vary in different populations. Additionally, while calibration appeared adequate based on the Hosmer–Lemeshow test, formal calibration assessment in external datasets is needed. Sixth, this study did not explore emerging approaches in COPD modeling (e.g., whole-lung radiomics). Seventh, although we provided detailed CT assessment methods, the visual scoring of emphysema and airway wall thickening is inherently subjective; however, the good inter-reader agreement (κ = 0.82 for emphysema, κ = 0.79 for airway wall thickening) supports the reliability of these assessments. Therefore, clinicians managing smokers should consider the insights from this multifactorial model, including bronchodilator testing alongside other key risk factors to identify high-risk individuals requiring close monitoring and facilitate early COPD detection. While nutritional support and target education may benefit patients with low BMI and a history of premature birth, smoking cessation remains central to prevention. Future research should validate these findings in multicenter, large-sample studies and explore the mechanisms underlying key factors identified to further refine COPD risk stratification models and inform targeted intervention strategies.

Conclusion

In conclusion, this study revealed that a positive bronchodilator test result, age ⩾60 years, smoking for ⩾30 years, airway wall thickening, BMI < 18.5 kg/m², history of premature birth, elevated FeNO level, and increased blood eosinophil count are crucial independent factors associated with COPD development among smokers. Importantly, the integration of a positive bronchodilator test into our multifactorial risk assessment model with clinical, imaging, and inflammatory biomarkers provided strong discriminative power beyond baseline spirometry. This model offers a valuable reference for clinicians to more effectively identify high-risk individuals from the larger pool of symptomatic smokers, enabling prioritized management and potentially mitigating the burden of underdiagnosed or potentially progressive COPD. Further research is needed to refine the model’s parameters, enhance its clinical utility, strengthen COPD prevention and treatment strategies, and ultimately reduce disease burden.

Supplemental Material

sj-docx-1-tar-10.1177_17534666261450054 – Supplemental material for A multifactorial model for chronic obstructive pulmonary disease risk in smokers with positive bronchodilation: a retrospective cohort study

Supplemental material, sj-docx-1-tar-10.1177_17534666261450054 for A multifactorial model for chronic obstructive pulmonary disease risk in smokers with positive bronchodilation: a retrospective cohort study by Jia Zhang, Weihua Zhu, Piping Jiang, Feng Ma, Yuwei Cao, Jiaxin Li, Shuangme Dai, Zihan Jia, Yulin Li, Chuan Song, Rui Yang, Zhe Zhang, Xin Zhang, Xinjun Zhang, Wailong Zou and Jichao Chen in Therapeutic Advances in Respiratory Disease

Supplemental Material

sj-docx-2-tar-10.1177_17534666261450054 – Supplemental material for A multifactorial model for chronic obstructive pulmonary disease risk in smokers with positive bronchodilation: a retrospective cohort study

Supplemental material, sj-docx-2-tar-10.1177_17534666261450054 for A multifactorial model for chronic obstructive pulmonary disease risk in smokers with positive bronchodilation: a retrospective cohort study by Jia Zhang, Weihua Zhu, Piping Jiang, Feng Ma, Yuwei Cao, Jiaxin Li, Shuangme Dai, Zihan Jia, Yulin Li, Chuan Song, Rui Yang, Zhe Zhang, Xin Zhang, Xinjun Zhang, Wailong Zou and Jichao Chen in Therapeutic Advances in Respiratory Disease

Footnotes

Acknowledgements

The authors would like to thank the participants in this study. We also acknowledge Editage () for providing English language editing and manuscript polishing assistance prior to submission. It is confirmed that all nonauthor contributors have provided consent for their acknowledgement.

Declarations

ORCID iD

Wailong Zou

Supplemental material

Supplemental material for this article is available online.

Use of artificial intelligence

No generative artificial intelligence (AI) or AI-assisted technologies were used in the creation of the text, references, figures, or other content of this manuscript.

References

Mettler

Nath

Grumley

, et al. Silent airway mucus plugs in COPD and clinical implications. Chest 2024; 166(5): 1010–1019.

Wei

, et al. Mechanism of cigarette smoke in promoting small airway remodeling in mice via STAT 3/PINK 1-Parkin/EMT. Free Radic Biol Med 2024; 224: 447–456.

Blomberg

Torén

Liv

, et al. Chronic airflow limitation, emphysema, and impaired diffusing capacity in relation to smoking habits in a Swedish middle-aged population. Ann Am Thorac Soc 2024; 21(12): 1678–1687.

Bhakta

McGowan

Ramsey

, et al. European Respiratory Society/American Thoracic Society technical statement: standardisation of the measurement of lung volumes, 2023 update. Eur Respir J 2023; 62(4): 2201519.

Zuo

Faiz

van den Berge

, et al. Cigarette smoke exposure alters phosphodiesterases in human structural lung cells. Am J Physiol Lung Cell Mol Physiol 2020; 318(1): L59–L64.

Yee

Markovic

Buhr

, et al. Significance of FEV3/FEV6 in recognition of early airway disease in smokers at risk of development of COPD: analysis of the SPIROMICS Cohort. Chest 2022; 161(4): 949–959.

Basille

Soriot

Weppe

, et al. Association between acute exacerbation of chronic obstructive pulmonary disease and short-term exposure to ambient air pollutants in France. Environ Health 2024; 23(1): 107.

Zhang

Hobbs

Silverman

, et al. Polygenic risk score added to conventional case finding to identify undiagnosed chronic obstructive pulmonary disease. JAMA 2025; 333(9): 784–792.

Momtazmanesh

Moghaddam

Ghamari

, et al. Global burden of chronic respiratory diseases and risk factors, 1990-2019: an update from the Global Burden of Disease Study 2019. EClinicalmedicine 2023; 59: 101936.

10.

Bui

Perret

Walters

, et al. Association between very to moderate preterm births, lung function deficits, and COPD at age 53 years: analysis of a prospective cohort study. Lancet Respir Med 2022; 10(5): 478–484.

11.

Testa

Magnoni

Fanizza

, et al. Prognostic models for disease progression and outcomes in chronic obstructive pulmonary disease: a systematic review and meta-analysis. J Clin Med 2025; 14(24): 8725.

12.

Skrivankova

Richmond

Woolf

BAR

, et al. Strengthening the Reporting of Observational Studies in Epidemiology Using Mendelian Randomization: The STROBE-MR Statement. JAMA 2021; 326(16): 1614–1621.

13.

Global Initiative for Chronic Obstructive Lung Disease. 2024 GOLD Report–Global Strategy for Prevention. Diagnosis and management of COPD, https://goldcopd.org/2024-gold-report/ (2024, accessed 2 January, 2025).

14.

Chen

Wang

, et al. DNA of neutrophil extracellular traps promote NF-κB-dependent autoimmunity via cGAS/TLR9 in chronic obstructive pulmonary disease. Signal Transduct Target Ther 2024; 9(1): 163.

15.

Jolly

Sidhu

Hewitt

, et al. Self management of patients with mild COPD in primary care: randomised controlled trial. BMJ 2018; 361: k2241.

16.

Ulrik

von Linstow

Nepper-Christensen

, et al. Chronic mucus hypersecretion: a marker of asthma in young adults? Respir Med 2005; 99(12): 1576–1582.

17.

Marín

Ciudad

Moya

, et al. Airflow reversibility and long-term outcomes in patients with COPD without comorbidities. Respir Med 2014; 108(8): 1180–1188.

18.

Tang

Zhong

Gong

, et al. Analysis of the association of ANO3/MUC15, COL4A4, RRBP1, and KLK1 polymorphisms with COPD susceptibility in the Kashi population. BMC Pulm Med 2022; 22(1): 178.

19.

MacNee

Is chronic obstructive pulmonary disease an accelerated aging disease?

Ann Am Thorac Soc 2016; 13(suppl 5): S429–S437.

20.

Calvo

Canosa

Bertuzzo

, et al. Influence of cigarette smoking on ALS outcome: a population-based study. J Neurol Neurosurg Psychiatry 2016; 87: 1229–1233.

21.

Andrault

Schamberger

Chazeirat

, et al. Cigarette smoke induces overexpression of active human cathepsin S in lungs from current smokers with or without COPD. Am J Physiol Lung Cell Mol Physiol 2019; 317(5): L625–L638.

22.

Van Eeckhoutte

Liu

, et al. Necroptosis signaling promotes inflammation, airway remodeling, and emphysema in chronic obstructive pulmonary disease. Am J Respir Crit Care Med 2021; 204(6): 667–681.

23.

Perez-Padilla

Montes

Oca

Thirion-Romero

, et al. Trajectories of spirometric patterns, obstructive and PRISM, in a population-based cohort in Latin America. Int J Chron Obstruct Pulmon Dis 2023; 18: 1277–1285.

24.

Wen

Wei

Giri

, et al. Association between serum uric acid/serum creatinine ratios and lung function in the general American population: National Health and Nutrition Examination Survey (NHANES), 2007-2012. BMJ Open Respir Res 2023; 10(1): e001513.

25.

Kraemer

Gardin

Smith

, et al. Functional predictors discriminating asthma-COPD overlap (ACO) from chronic obstructive pulmonary disease (COPD). Int J Chron Obstruct Pulmon Dis 2022; 17: 2723–2743.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.02 MB

0.00 MB

0.03 MB