Repeatability and replicability of the Respiratory Symptom Experience Scale (RSES): internal and external validation of smoking-status differences in respiratory symptoms

Abstract

Background

Respiratory symptoms are common among smokers without COPD, yet symptom instruments are often optimized for established disease and may under-capture early changes relevant to cessation and tobacco harm reduction. We conducted a repeatability and replicability study of the Respiratory Symptom Experience Scale (RSES) in a rigorously phenotyped cohort to confirm both measurement stability and reproducible discrimination by smoking status.

Methods

Healthy adults were recruited at a UK hospital and classified as current, former, or never smokers using self-report, exhaled CO, and a comprehensive panel of biomarkers of exposure. RSES was administered at Visit 1 and repeated after 7 ± 3 days at Visit 2 to assess test–retest repeatability. Known-group validity (discriminability by smoking status) was evaluated by comparing RSES scores across biochemically verified groups using non-parametric tests, with ANCOVA for exploratory covariate adjustment. True replicability across independent datasets was not formally tested in the present study.

Results

Of 180 enrolled participants, 170 were included (55 current smokers, 60 former smokers, 55 never smokers) after excluding clinical ineligibility and biomarker-inconsistent smoking status. RSES demonstrated high repeatability (Pearson r = 0.87; Spearman ρ = 0.85) with minimal bias (− 0.07) and good agreement. RSES also showed consistent known-group discrimination by smoking status, differing between current and never smokers (p < 0.0001) and across all three groups (p < 0.0001). Current smokers reported higher scores than never (median 2.0 vs. 1.0) and former smokers (2.0 vs. 1.2), whereas former and never smokers did not differ (p = 0.686). Covariate adjustment did not materially alter results.

Conclusions

RSES shows strong repeatability and consistent known-group discrimination by smoking status, supporting its use as a patient-relevant endpoint for cessation, and harm-reduction trials, as well as for clinical respiratory and regulatory research.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s12955-026-02591-z.

Introduction

Tobacco smoking remains highly prevalent worldwide and is responsible for about eight million deaths each year [1, 2]. Prolonged exposure to the toxic constituents of cigarette smoke leads to progressive inflammatory injury throughout the respiratory tract. This persistent inflammation, together with subsequent airway remodelling, contributes to the development of chronic respiratory diseases such as chronic obstructive pulmonary disease (COPD) and asthma [3–5].

Importantly, many smokers experience respiratory symptoms such as cough, wheeze, and sputum production even before meeting diagnostic criteria for COPD or other recognised respiratory disorders [6–8]. These symptoms may represent early markers of smoking-related respiratory damage, offering a window to characterise functional impairment before overt disease develops. They are also among the earliest outcomes expected to improve after smoking cessation [9] and after complete switching to non-combustible nicotine products [10, 11].

Reducing the respiratory harm caused by tobacco smoking depends on eliminating chronic exposure to combustion-derived toxicants by stopping smoking. Both smoking cessation and switching to the exclusive use of non-combustible nicotine products are associated with improvements in respiratory health outcomes [12, 13].

Most of the evidence linking smoking abstinence to improvements in respiratory symptoms comes from prospective longitudinal studies in patients with established COPD [14, 15]. In contrast, relatively few data are available for smokers without COPD. Existing questionnaires developed and validated for assessing respiratory symptoms in COPD [16, 17] are unlikely to be optimal in this population, as they are calibrated to reflect the greater symptom burden typical of COPD rather than the milder, earlier manifestations seen in otherwise healthy smokers. For example, questions focusing on severe breathlessness or marked limitations in daily activities may not be relevant for smokers with modest symptoms and may therefore underestimate the true impact of smoking-related respiratory complaints.

The need for a symptom scale tailored specifically to non-COPD smokers and former smokers is further underscored by the growing interest in combustion-free alternative nicotine products, such as e-cigarettes and heated tobacco products (HTPs), as tools for tobacco harm reduction. Cigarette smoke exposes users to a complex mixture of combustion-derived toxic chemicals that contribute directly to respiratory symptoms. By contrast, aerosols generated by non-combustible products do not contain these combustion products [18–20], suggesting that complete switching from cigarettes may lead to improvements in respiratory symptoms. Nevertheless, because these products are inhaled, it remains essential to employ appropriate and sensitive instruments capable of detecting even subtle changes in respiratory symptoms among smokers who switch, as well as among those who quit entirely.

A major limitation in evaluating changes of respiratory health after smoking cessation and tobacco harm reduction strategies lies in the lack of sensitive and clinically meaningful respiratory endpoints [21]. This gap has slowed progress in understanding the full health impact of cessation therapies and combustion-free nicotine products. Advancing regulatory science and clinical research therefore requires early indicators of respiratory health recovery to capture the earliest signs of clinical improvement. At the University of Catania’s Center of Excellence for the Acceleration of Harm Reduction (CoEHAR), ongoing work is assessing early-effect indicators to detect physiological/biological changes after stopping smoking, including reported respiratory symptoms in non-COPD smokers [10, 11].

Research in this field would be greatly strengthened by the availability of a clinical assessment tool specifically designed and validated to quantify respiratory symptoms in adult current and former smokers. The Respiratory Symptom Experience Scale (RSES) [22] was developed for this purpose and has shown good psychometric properties: it distinguishes individuals with and without respiratory disease, differentiates smokers from former smokers, and shows lower symptom scores among smokers who switch to e-cigarettes compared with those who continue smoking.

Because RSES appears to respond early to changes in airway toxicant burden it is a promising respiratory health indicator for cessation studies and switching trials involving alternative nicotine products (e-cigarettes, heated tobacco products, and oral nicotine pouches). However, evidence of internal and external validation remains limited.

Repeatability (the ability of a measure to yield consistent results when repeated under the same conditions) and replicability (the ability to obtain consistent results across independent studies addressing the same question using different datasets) [23] are essential to establish measurement reliability and to support its use as an endpoint. In medicine, the “replication crisis” has further highlighted the importance of replicability for maintaining confidence in the evidence base [24].

We therefore conducted a study assessing repeatability (test-retest reliability) and known-group validity (discriminability across smoking-status groups) of the RSES in a rigorously phenotyped cohort. It should be noted that, in this context, the cross-group comparison constitutes an assessment of construct validity rather than replicability in the strict sense (i.e., reproduction of findings in a fully independent dataset), which remains to be established in future work. To enhance classification accuracy, self-reported smoking behaviour was validated using a panel of biomarkers of exposure (BoEs), ensuring precise characterisation of smoking status and exposure intensity [25].

Establishing strong internal and external validity is essential for supporting RSES as a robust, clinically meaningful measure in smoking-related research (including cessation and switching studies) and as a patient-centred outcome that can complement biomarker data in clinical respiratory research and regulatory review (e.g., FDA submissions) In this work, internal validation refers to the assessment of construct validity within the study dataset, operationalised through (i) test-retest repeatability and (ii) known-group discrimination across biochemically verified smoking categories. External validation refers to the cross-cohort generalisability of findings, evaluated by comparing observed patterns with those previously reported in independent U.S. and Italian cohorts using the same instrument.

Methods

Study population

The study enrolled healthy adults who were allocated to three groups. Individuals who had been smoking at least 10 cigarettes per day for ≥ 15 years and who, at screening, showed an exhaled carbon monoxide (eCO) concentration ≥ 7 ppm were classified as current smokers. Individuals who had abstained from smoking for a minimum of 6 months before screening and who had an eCO value < 7 ppm were classified as former smokers. Self-reported non-smoker with a lifetime consumption of fewer than 100 cigarettes and presenting an eCO < 7 ppm, to minimise the likelihood of substantial passive smoke exposure or environmental carbon monoxide contamination were classified as never smokers.

To maximise the accuracy of group assignment, smoking status was corroborated by an extensive panel of biomarkers of exposure (BoEs). This panel included nicotine and its five main metabolites [26]; tobacco-specific nitrosamines (TSNAs) [27]; 2-cyanoethylmercapturic acid (CEMA, indicative of recent exposure) [28]; cyanoethylvaline (CeVal, reflecting longer-term exposure) [29]; anatabine and anabasine (markers specific to tobacco use) [27]; propylene glycol (a marker of e-cigarette use) [30]; and carboxyhaemoglobin (COHb) [25, 31] The cut-off levels applied to these biomarkers for the verification of smoking status were consistent with those described by Sajid et al. [25].

Participants (current, former, and never smokers) were not eligible for inclusion if any of the following criteria were met:

• Presence of conditions that might cause or confound assessment of respiratory symptoms, including upper respiratory tract infection within the preceding 14 days, asthma, COPD, bronchiectasis, or cystic fibrosis.

• Ongoing use of e-cigarettes or heated tobacco products.

• Relevant exposure to secondhand smoke (for former and never smokers).

• History of environmental or occupational exposure to pollutants (e.g. living in close proximity to major roads or industrial sites, or working in chemical or metallurgical industries).

• Pregnancy.

Study design

This cross-sectional study compared respiratory symptoms scores across three groups: current, former, and never smokers. The research was conducted at Ashford and St Peter’s Hospitals NHS Foundation Trust (Chertsey, England)) and involved three morning visits: an initial screening visit, a baseline visit (V1), and a follow-up visit at day 7 ± 3 days (V2) to assess measurements repeatability. To limit acute effects of smoking on study outcomes, participants who smoked were asked to refrain from smoking for at least 1 h before each visit.

During the screening visit, potential participants were provided with verbal and written information outlining the study rationale and objectives. Investigators then evaluated eligibility and recorded basic sociodemographic characteristics (e.g. sex, age, smoking status). Individuals meeting the eligibility criteria were invited to return for the baseline visit (V1), at which point inclusion criteria were rechecked, written informed consent was obtained, biospecimens (spot urine, blood, and exhaled breath) were collected, and respiratory symptoms questionnaire administered. Participants returned for a follow-up visit (V2), scheduled 7 ± 3 days after V1, during which eligibility was reverified and questionnaire repeated to evaluate its repeatability.

Respiratory symptoms were assessed using the Respiratory Symptoms Experience Scale (RSES), a brief, self-administered instrument developed and validated for adults who smoke cigarettes but may not have a diagnosed respiratory disease [22]. Respondents rated the frequency of five symptoms experienced in the past 30 days on a 5-point response scale (ranging 1 to 5), with higher values indicating worse respiratory symptoms. The composite RSES score, calculated as the unweighted mean of the scores for each of the five questions answered, served as the primary outcome of the study. The individual symptoms and scoring details of the RSES are reported in the Supplemental Table.

Statistical analysis

The RSES score was an exploratory outcome within a broader study investigating the capacity of a range of biomarkers (biochemical, functional, and imaging-based) to distinguish healthy smokers from non-smokers and former smokers. As such, the study was not specifically powered for the RSES outcome.

The analysis population included all participants who met the inclusion criteria after applying the following post-recruitment exclusions: (1) participants in whom study assessments revealed an unsuspected or undeclared condition that would have led to exclusion at baseline; and (2) participants whose biochemical assay results were inconsistent with their self-reported smoking status, as determined by exposure biomarkers described by Sajid et al. [25]. Briefly, a multi-biomarker approach was used to assess exposure to nicotine, tobacco-specific nitrosamines, acrylonitrile, and propylene glycol (PG). All biomarkers, except PG, were significantly elevated in the current smoker group, leading to the identification of six non-compliant participants (four current smokers, one former smoker, and one non-smoker). The biomarker panel outperformed eCO and was also able to distinguish distinct use patterns such as vaping versus smoking, thereby enhancing data accuracy. It should be noted that eCO was measured during morning visits; as eCO typically accumulates throughout the day, morning measurements may underestimate the true CO burden in active smokers, potentially limiting the sensitivity of eCO as a standalone classification tool.

Baseline characteristics and RSES values were described by smoking status subgroup, with continuous variables presented as median ± IQR and mean ± SD, and categorical variables as proportions. Although RSES overall score is technically an ordinal variable, given that there are 21 potential values for the overall score, for descriptive purposes it can reasonably be expected to behave in a broadly continuous fashion and therefore median/mean and IQR/sd values are presented.

A stepwise approach to analysis was adopted.

Test–retest reliability for RSES was next assessed using the Pearson product–moment correlation coefficient, with results validated by Spearman’s rank correlation coefficient to account for non-normal distributions. A Bland-Altman plot was generated to explore any evidence for systematic bias.

Given the discontinuous nature of RSES scores discussed above, the underlying assumptions for standard parametric comparison are unlikely to be met. Consequently, the association between the RSES score and smoking status was tested using the non-parametric Kruskal–Wallis H test, an extension of the Mann-Whitney U-test that allows comparison of ordinal outcomes. The initial comparison focused on current versus never smokers.

Only if RSES showed significant discrimination between current and never smokers was reanalysis to be preformed in a second step including all three groups (current, former, and never smokers). This approach is based on the premise that a metric that is unable to distinguish smokers from non-smokers were unlikely to be informative when including former smokers, whose biomarker and physiological values were expected to fall between the two extremes. Where three-way analysis was carried out, the Kruskal–Wallis H-test was once again used followed by pairwise comparisons.

In the absence of a covariate-corrected version of the Kruskal-Wallis test, a conventional analysis of covariance (ANCOVA) model was fitted to explore whether individual characteristics (age, sex, ethnicity, body mass index [BMI]) exerted an independent effect on RSES values.

All tests were two-sided, and p < 0.05 was considered the nominal threshold for statistical significance. Statistical analyses and graph generation were performed using MedCalc v23.3.7.

Results

A total of 180 subjects were enrolled: 60 current smokers, 59 never smokers, and 61 former smokers. Ten subjects were excluded from the analysis: one with a BMI < 17 kg/m² (undernutrition), one with lung cancer documented on CT, six whose biomarker-of-exposure profiles were inconsistent with their self-reported smoking category [25], and two whose spirometry suggested restrictive lung disease. The final analysis population comprised 170 participants (55 current smokers, 55 never smokers, and 60 former smokers), with demographic and anthropometric characteristics reported in Table 1. Most former smokers (n = 45) had been abstinent for more than 10 years, and the cohort therefore largely reflects long-term cessation (mean time since quitting 17.2 ± 10.1 years).

Table 1.

Demographic data of the study population (n = 170)

Sex, % males	Never smokers (n = 55)	Current smokers (n = 55)	Former smokers (n = 60)	P value Never smokers Vs. Current smokers	P value Former smokers Vs. Current smokers	P value Former smokers Vs. Never smokers
Sex, % males	29.1	38.2	41.7	0.315	0.704	0.161
Ethnicity: (%)
Caucasian	45.5	74.5	88.3	0.002	0.136	< 0.0001
Asian	45.5	21.8	8.3
Black African/Caribbean	9.1	-	1.7
Mixed race	-	3.6	1.7
Age (years)	45.4 ± 8.8	47.8 ± 8.4	51.5 ± 8.4	0.006	0.020	0.0002
Body Mass Index (Kg/m2)	27.4 ± 5.7	27.6 ± 6.7	28.3 ± 5.3	0.910	0.546	0.382
Smoking (Pack/years)	-	20.4 ± 11.5	12.5 ± 9.6	n/a	0.0001	n/a
Time since quitting (years)	-	-	17.2 ± 10.1	n/a	n/a	n/a

Values are expressed as frequencies or mean±standard deviation. P-values are estimated using chi-squared test for categorical variables and ANOVA for continuous variables

Test-retest analysis to assess measurement repeatability between Visit 1 and 2 was carried out using data from all 170 subjects. The correlation between Visit 1 and Visit 2 measurements was high (r = 0.87; Fig. 1). Consistent results were obtained using a non-parametric approach (Spearman’s rho = 0.85) (Table 2).

Fig. 1.

Test–retest repeatability of RSES between Visit 1 and Visit 2 (n = 170) Scatter plot showing individual participant values at Visit 1 (x-axis) versus Visit 2 (y-axis). The association indicates high repeatability (Pearson’s r = 0.87)

Table 2.

– Test–retest repeatability of RSES between Visit 1 and Visit 2 (n = 170). Parametric and non-parametric correlation coefficients by smoking status

	Pearson’s r	Spearman’s rho
All subjects	0.869	0.847
Current smokers	0.774	0.763
Never smokers	0.757	0.762
Former smokers	0.916	0.811

The Bland–Altman plot with mean bias and 95% limits of agreement are shown in Fig. 2. The plot shows a small negative mean bias (− 0.07) with an even dispersion of differences across the measurement range, and few outliers (6.5% of observations outside ± 1.96 SD), indicating a good level of agreement between the two measurements (Table 3).

Fig. 2.

Bland–Altman plot for test–retest agreement of RSES between Visit 1 and Visit 2 (n = 170)

Difference between measurements (Visit 2 − Visit 1) plotted against the mean of the two visits for each participant. The solid horizontal line indicates the mean bias (− 0.07), and the dashed lines represent the 95% limits of agreement (mean bias ± 1.96 SD). The even dispersion of points and the small proportion of observations outside the limits (6.5%) indicate good agreement between visits

Table 3.

– Test–retest repeatability of RSES between Visit 1 and Visit 2 (n = 170). Summary of results of Bland-Altman plot by smoking status

	Mean bias	95% CI	% >1.96 sd
All subjects	0.07	-0.55; 0.68	6.47%
Current smokers	0.11	-0.80; 1.01	5.45%
Never smokers	0.09	-0.40; 0.57	9.09%
Former smokers	0.01	-0.32; 0.34	8.33%

The primary analysis of quantitative between-group differences was conducted in 110 participants (55 current smokers and 55 never smokers). The two groups were comparable in sex distribution, age, and BMI. However, the proportion of Asian participants was significantly higher in the never-smoker group. Smoking history in the smoker group was consistent with heavy exposure (mean 20.4 pack-years). For the primary analysis, RSES scores at Visit 1 were compared between current smokers and never smokers and showed a statistically significant between-group difference (p < 0.0001).

Given the clear discriminatory performance of the RSES between smokers and never smokers, we repeated the analysis including former smokers. This also showed a significant overall difference across the three groups (p < 0.0001). Pairwise comparisons confirmed significant differences between current smokers and never smokers (median RSES 2.0 vs. 1.0) and between current smokers and former smokers (median RSES 2.0 vs. 1.2). No significant difference was observed between never smokers and former smokers (median RSES 1.0 vs. 1.2) (Fig. 3).

Fig. 3.

Respiratory Symptoms Experience Scale (RSES) scores by smoking status.

Distribution of Visit 1 RSES scores (Median ± 95% CI) in current smokers, former smokers, and never smokers

Adjustment for baseline covariates did not materially change the overall findings (Table 4). Table 5 presents an ANCOVA including age, BMI, height, sex, ethnicity (Caucasian), and smoking category. The model was significant (F = 16.178, p < 0.001) and explained a moderate proportion of variance (R²=0.411; adjusted R²=0.386). BMI was the only covariate independently associated with the outcome (F = 8.097, p = 0.005), whereas age, height, sex, and ethnicity were not (all p > 0.10). Smoking category remained strongly associated (F = 51.935, p < 0.001), confirming that between-group differences were robust after adjustment for covariates.

Table 4.

Respiratory Symptoms Experience Scale (RSES) scores by smoking status. Adjusted mean results after correction for potential confounders (ANCOVA)

Category	RSES score (Original analysis)		RSES score (Adjusted analysis)
	Median	Mean	Estimated marginal mean
Current smoker	2.00	2.09	2.11
Never smoker	1.00	1.28	1.27
Former smoker	1.20	1.31	1.30

Table 5.

ANCOVA of Visit 1 RSES adjusting for baseline covariates (age, BMI, height, sex, ethnicity [Caucasian]) and smoking category (current, former, never)

Tests of Between-Subjects Effects
Source	Sum of Squares	DF	Mean Square	F	P
Corrected Model	26.563	7	3.795	16.178	< 0.001
Intercept	0.0481	1	0.0481	0.205	0.651
Age	0.0251	1	0.0251	0.107	0.744
BMI	1.899	1	1.899	8.097	0.005
Height	0.642	1	0.642	2.739	0.100
Male	0.211	1	0.211	0.901	0.344
Caucasian	0.227	1	0.227	0.966	0.327
Category	24.364	2	12.182	51.935	< 0.001
Residual	37.999	162	0.235
Total	475.160	170
Corrected Total	64.562	169
Coefficient of determination R²			0.4114
R²-adjusted			0.3860

Discussion

A key objective of this work was to evaluate the test-retest repeatability of the Respiratory Symptoms Experience Scale (RSES) and its known-group validity, i.e., its ability to discriminate symptom burden across biochemically verified smoking categories. The cross-group comparison addresses construct validity rather than replicability in the strict methodological sense. We therefore compared RSES scores in biochemically verified current, former, and never smokers. Establishing strong internal and external validity is essential to support RSES as a robust, clinically meaningful endpoint for smoking-related research (including cessation and switching studies) and as a patient-centred outcome that complements biomarker data in clinical respiratory research and regulatory review (e.g., FDA submissions).

In this cohort of healthy adults with biochemically confirmed smoking status, RSES demonstrated strong construct validity (a component of internal validity), evidenced by clear cross-sectional discrimination by smoking category. RSES scores differed between current and never smokers and across current, former, and never smokers. Current smokers reported higher symptom scores than never smokers and former smokers, whereas never and former smokers had similar scores. Although the clinical interpretation of RSES score differences requires further anchoring to patient-relevant thresholds, the observed between-group difference of 1.0 median unit (current vs. never smokers: 2.0 vs. 1.0) represents a near-doubling of the symptom frequency rating and aligns with the distinctions between “rarely” and “occasionally” symptomatic patients. Establishing the Minimal Clinically Important Difference (MCID) for the RSES will be important for interpreting the magnitude of future intervention-related changes.

The absence of a difference between never and former smokers likely reflects the long time since quitting in this cohort, as most former smokers had been abstinent for more than 10 years, suggesting near-complete normalization of respiratory symptoms after cessation. Covariate adjustment did not materially alter these findings, consistent with symptom improvement after sustained abstinence. Our results in English participants align with prior findings in U.S [22]. and Italian cohorts [32], both using independent datasets with different sampling frames, supporting the external validity of the RSES across distinct cultural and linguistic populations. This cross-cohort consistency constitutes the external validation evidence.

The internal and external validity of the RSES are essential prerequisites for its use as an endpoint in clinical and regulatory research, and our findings support this role in a “non-COPD smoker” population. Symptom tools developed for established COPD may under-capture milder burdens and early disease because they are tuned to more severe breathlessness and activity limitation. In contrast, the present data indicate that the RSES retains sufficient sensitivity to detect smoking-category differences even in a carefully screened cohort that excluded conditions likely to confound symptom reporting (e.g., asthma, COPD, recent respiratory infection). Overall, these results support RSES as a practical measure for detecting early respiratory changes and recovery in cessation and tobacco harm reduction studies, where validated short-horizon respiratory endpoints remain a recognized gap [21].

A notable strength of this paper is the rigorous verification of smoking status. Classification into current, former, and never smokers did not rely solely on self-report, but was corroborated using an extensive panel of biomarkers of exposure [25]. Participants with biomarker profiles inconsistent with their self-reported category were excluded, reducing the risk of smoking misclassification (a recurrent problem in tobacco research) that can dilute true between-group differences. This strengthens confidence that observed symptom differences reflect genuine exposure categories rather than reporting artefacts, and is particularly valuable in regulatory settings (e.g., switching or abstinence evaluations) where classification accuracy is critical for interpretability.

This work has clear relevance to both regulatory science and clinical respiratory research. For tobacco/nicotine product evaluation, regulators (e.g. FDA) increasingly look for clinically meaningful, patient-relevant outcomes (PROs) to complement chemistry and biomarker evidence [33, 34], especially in otherwise “healthy” smokers without established COPD, in whom spirometric changes are often undetectable over typical trial time horizons and may take years to become clearly manifest. In this context, RSES helps fill a key gap: it is patient-centered, repeatable, and discriminates by smoking status in a rigorously phenotyped cohort, supporting its use as a Clinical Outcome Assessment (COA) alongside biomarkers in regulatory submissions, pending longitudinal validation of responsiveness to intervention. Clinically, because many symptom tools were developed for COPD and may under-capture milder burdens, RSES offers a pragmatic endpoint for studies of early respiratory effects and recovery after cessation or complete switching, and can be integrated with spirometry, imaging, and biomarkers for a more complete picture of respiratory health.

Several limitations should be acknowledged. First, RSES was an exploratory outcome within a broader biomarker programme and the study was not powered specifically for this endpoint. Nevertheless, we observed statistically significant, clinically meaningful difference between smokers vs. never and former smokers despite the modest sample size, which can inform future study design. Establishing the MCID and responsiveness to interventions will be important for interpreting prospective trials. Second, participants were asked to abstain from smoking for at least one hour before visits to limit acute effects. While pragmatic, this does not fully standardize recent exposure and may not eliminate short-lived symptom fluctuations. Additionally, exhaled CO was measured during morning visits; since eCO levels typically accumulate over the course of the day rather than being stable across hours, morning measurements may not reflect peak daily exposure and may explain in part why the biomarker panel outperformed eCO alone for classifying smoking status. Third, the cross-sectional design cannot capture within-person change after quitting. Longitudinal studies should assess RSES responsiveness to cessation and complete switching away from combustible cigarettes. Fourth, although covariate adjustment did not materially change findings, residual confounding remains possible in these observational comparisons. Fifth, it should be acknowledged that external validation in the present work is based on cross-cohort consistency rather than formal pre-registered replication in an independent dataset; prospective independent replication remains an important next step.

Despite these limitations, our findings support RSES as a practical, repeatable, and discriminative symptom endpoint in healthy smokers and comparators, with potential value for detecting early respiratory changes after cessation or switching. This is particularly relevant to cessation and harm-reduction research, where symptom improvement may precede measurable spirometric change and represents a core patient-relevant outcome.

In conclusion, RSES demonstrated high repeatability and robust discrimination between current smokers and both former and never smokers in a rigorously phenotyped cohort. These properties support its further development and prospective validation as an endpoint for detecting early, clinically meaningful respiratory symptom differences in smoking-related research, though the present cross-sectional design should be noted as a limitation when interpreting the strength of this evidence.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

P.C. conceived the study and designed the research protocol. L.S., D.C. and M.N. coordinated participant recruitment and data collection. J.B. performed the statistical analyses and prepared the main tables and figures. D.Cam., I.J., E.A., and G.G. contributed to data curation and quality control. N.P. and M.S. conducted biomarker analyses and contributed to methodological validation. D.Ca. and J.B. drafted the initial manuscript text. P.C., L.S., D.Ca., J.B., D.Ca., M.N., I.J., E.A., G.G., N.P., M.S., and R.P. critically reviewed and revised the manuscript. R.P. supervised the project and secured funding. All authors approved the final version of the manuscript.

Funding

This investigator-initiated research is sponsored by Altria Client Services LLC (ALCS). Sponsor played no role in designing, implementing, data analysis, or interpretation of the study results. The contents, selection, and presentation of facts, as well as any opinions expressed, are the sole responsibility of the authors and should not be regarded as reflecting the positions of the sponsor.

Data availability

No datasets were generated or analysed during the current study.

Declarations

Human ethics and consent to participate declarations

This study was conducted in accordance with the Declaration of Helsinki. Ethical approval was granted by the Health Research Authority (HRA) and Health and Care Research Wales (HCRW), the designated Institutional Review Board (IRB/Ethics Committee) for this research (IRAS project ID: 322282; approval date: 14 March 2023). Written informed consent was obtained from all participants prior to enrolment.

Competing interests

The authors declare no competing interests.