History of Smoking and Prognosis of Patients With Non-small Cell Lung Cancer

Abstract

Background/Aim

Smoking is a known risk factor for the development of non-small-cell lung cancer (NSCLC). However, conflicting results are reported in the literature regarding the risk of smoking on the prognosis of patients with NSCLC. Moreover, these studies were often conducted before the era of personalized targeted therapy and immunotherapy, which can significantly improve the prognosis of patients. The aim of the present study was to evaluate the prognosis of patients with NSCLC in relation to their smoking history.

Patients and Methods

This retrospective multicentre study included all patients from the Czech LUCAS registry with baseline tumour characteristics. Overall survival was assessed using Kaplan-Meier estimates, with differences between groups evaluated by the log-rank test and multivariate Cox regression model, adjusting for covariates and stratifying by disease stage when proportional hazards assumptions were violated.

Results

A total of 3,372 patients were included. In the univariate model, non-smokers had a significantly better prognosis than former and current heavy smokers (with a history of ≥40 pack-years), hazard ratio=1.19 (95% confidence interval=1.01-1.42), p=0.043; and hazard ratio 1.19 (95% confidence interval=0.03-1.37), p=0.018, respectively. This relationship was not significant in the multivariate model. The absence of a prognostic significance of smoking at the time of NSCLC diagnosis was also demonstrated in a multivariate model both in patients with advanced disease treated with anticancer drugs and in patients who underwent surgery.

Conclusion

Although smoking/ex-smoking at the time of NSCLC diagnosis had a negative prognostic impact in univariate analysis, multivariate analysis did not demonstrate smoking itself as an independent prognostic factor for NSCLC. Therefore, other factors, including poorer performance status, likely contributed to the poorer outcome for smokers.

Introduction

Non-small-cell lung cancer (NSCLC) is strongly associated with smoking, which is the most significant risk factor for its development (1). However, there are also cases of NSCLC in non-smokers (15-25% cases), which are usually associated with different characteristics compared to NSCLC in smokers (2). NSCLC in non-smokers occurs more frequently in women and is more commonly represented by adenocarcinoma histologically (2, 3). Non-smokers also have a higher incidence of some driver mutations (especially for the epidermal growth factor receptor, EGFR) and some other biological characteristics (2, 4, 5). On the other hand, NSCLC in non-smokers is usually diagnosed at a higher stage, although according to some authors, the primary tumour itself tends to be larger in smokers at diagnosis (3, 6). The effect of smoking on the effectiveness of some oncology drugs is also described (7, 8). Therefore, the question arises whether these groups also have a different prognosis for overall survival (OS).

Studies (usually older publications) on this topic have inconsistent conclusions and often included only a group of non-smokers and smokers, ignoring former smokers, as well as the different numbers of cigarettes smoked by former and current smokers (9-36). Moreover, both surgical treatment (increased use of minimally invasive methods) and oncological therapy (advancement of personalized targeted therapy and immunotherapy) have significantly changed the treatment approaches for NSCLC in recent years and thus improved prognostic outcomes of patients (37). Therefore, conclusions regarding the effect of smoking on OS in more recent studies may differ from those of older studies.

The aim of this study was to compare OS between non-smokers, former smokers and current smokers, taking into account the total number of cigarettes smoked, both in the total population of patients with NSCLC and in subgroups of patients undergoing radical surgery and patients with stage IV disease treated with anticancer drug therapy (including personalized targeted therapy and immunotherapy).

Patients and Methods

The LUCAS registry and patient cohort. TheLUCAS (Lung Cancer Focus) registry is a prospective registry, established in June 2018, in which consecutive patients with lung cancer have been enrolled across 11 centres in the Czech Republic. The registry is a project of the Czech Pneumological and Phthisiological Society of the Czech Medical Association of Jan Evangelista Purkyně and is registered at ClinicalTrials.gov under registration number NCT04228237. In the LUCAS registry, basic demographic and clinical characteristics, as well as data on pharmacotherapy (including treatment types, combinations, and sequences) and surgery, are recorded.

All patients enrolled in the LUCAS registry with complete basic patient characteristics were included in the study (inclusion criteria: included in the LUCAS registry, diagnosis of NSCLC after June 2018; exclusion criteria: unavailability of current survival data or incomplete data in the LUCAS registry). In total, 11 centres from the Czech Republic participated. Information on the diagnosis was extracted from patients’ medical records.

A non-smoker was defined as one who smoked fewer than 200 cigarettes in their lifetime, a former smoker as one who smoked more than 200 cigarettes in their lifetime and did not smoke for more than 1 year before the diagnosis of NSCLC, and a smoker as one who smoked more than 200 cigarettes in their lifetime and continued to smoke or quit smoking within 1 year before the diagnosis of NSCLC. A pack-year (PY) was defined as smoking 20 cigarettes per day for 1 year or its equivalent (e.g., 10 cigarettes per day for 2 years, etc.). Patients were divided into the following groups: non-smokers, former smokers and smokers. Former smokers and current smokers were further divided according to their PY history into groups: <20 PY, 20-<40 PY and ≥40 PY.

All participants were required to give their written informed consent prior to inclusion in the registry. This study was approved by the Ethics Committee of University Hospital Olomouc, Faculty of Medicine and Dentistry, Palacky University, on June 11, 2018 (reference number 63/18 MEK 13) and was conducted in accordance with the Declaration of Helsinki (as revised in 2013).

The Institute of Health Information and Statistics of the Czech Republic, in cooperation with the Czech Statistical Office, was requested to perform patient vital status verification (ascertaining whether enrolled patients were alive or deceased). This verification was completed at all participating centres, with a reference date of June 30, 2023 to December 31, 2024, depending on the specific centre. Data collection in the LUCAS registry was completed in May 2025. Consequently, a total of 3,372 patients were available for the current analysis.

Statistical methods. OS, defined as the time from NSCLC diagnosis to death from any cause, was evaluated using the Kaplan-Meier method. Patients who were alive at study end were censored at the last date on which their vital status was reliably known (e.g., based on treatment administered or their last clinical visit until May 2025).

Survival differences between study groups were assessed using the log-rank (Mantel-Cox) test and the Cox regression model. The Cox regression analysis was performed as a multivariate model, with statistically non-significant covariates removed using backward elimination, resulting in the final model applied in this analysis. Hazard ratios for smoking history and sex were derived from a multivariate Cox proportional hazards model, fully adjusted for age and tumour histology, and stratified by clinical stage (according to eighth TNM classification) (38), Eastern Cooperative Oncology Group performance status (ECOG PS), and treatment modality. Compared with the model including stage as a covariate, this approach had only a minimal impact on the overall results and was considered a more appropriate statistical method for evaluating the effects of individual covariates. The median follow-up was estimated using the reverse Kaplan-Meier method.

Differences between study groups were analysed using the Mann-Whitney U-test for continuous variables and the Pearson chi-square test or Fisher’s exact test for categorical variables. Hypothesis testing was performed at a significance level of 5%, with values of p<0.05 considered statistically significant. All analyses were conducted using R (Version 4.5.1, 13-06-2025) (39).

Results

Patient characteristics. A total of 3,372 patients were enrolled. A STROBE flow diagram illustrating the participant selection process from the total LUCAS registry population is shown in Figure 1. The median follow-up was 38.2 (95% confidence interval=36.9-39.3) months. The median age of the patients was 70 years. Other patient characteristics are shown in Table I.

Figure 1.

STROBE flow diagram illustrating the process of participant selection from the total LUCAS registry population. NSCLC: Non-small cell lung cancer.

Table I. Baseline patients’ characteristics.

ECOG PS: Eastern Cooperative Oncology Group performance status; LC: large-cell; NEC: neuroendocrine carcinoma; NSCLC NOS: non-small-cell lung cancer not otherwise specified; PY: pack-years.

Radical surgery was performed in 432 (12.8%) of patients overall: in 132 patients with stage I disease (75% of stage), 157 with stage II (63.1% of stage) and 143 with stage IIIA (34.5% of stage).

Pharmacological oncological treatment was administered to 1,939 (57.4%) patients with stage IV disease. Of these, a total of 296 (15.3%) patients received targeted therapy, 727 (37.5%) patients received immunotherapy with/without chemotherapy, and 916 (47.2%) patients received chemotherapy alone.

Patient characteristics according to smoking history. Non-smokers were more often women, histologically more often had adenocarcinoma and more of them were diagnosed in stage IV. Non-smokers had more often ECOG PS 0. Non-smokers were more often treated in stage IV with targeted therapy and less with immunotherapy. When comparing former smokers and current smokers, it is evident that former smokers were more often lighter smokers (less PY in smoking history) than smokers. The differences between the individual groups are summarized in detail in Table II.

Table II. Differences between groups of non-smokers, former smokers and smokers with non-small-cell lung cancer (NSCLC).

CHT: Chemotherapy; ECOG PS: Eastern Cooperative Oncology Group performance status; LC: large-cell; NEC: neuroendocrine carcinoma; NOS: not otherwise specified; PY: pack-years. Statistically significant p-values are shown in bold.

Prognosis by smoking status – univariate model. Current and former heavy smokers had significantly shorter OS than non-smokers. Lower OS was also observed in men, with increasing tumour stage and ECOG PS, and for patients with NSCLC not otherwise specified. Factors related to survival in the univariate model are summarised in Table III.

Table III. Univariate Cox regression model of prognostic factors for overall survival of patients with non-small-cell lung cancer (NSCLC).

CI: Confidence interval; ECOG PS: Eastern Cooperative Oncology Group performance status; HR: hazard ratio; LC: large-cell; NEC: neuroendocrine carcinoma; NOS: not otherwise specified; PY: packyears. *N=3,364, except for ECOG PS where n=3,261. Statistically significant p-values are shown in bold.

Kaplan-Meier curves of the individual groups according to smoking history are shown in Figure 2 and OS values from these curves are given in Table IV. Overall, the log-rank test value was p=0.2. In the log-rank test for comparison of non-smokersandformer smokers, the p-value was 0.2. In the log-rank test for comparison of non-smokers andcurrent smokers, the p-value was 0.08.

Figure 2.

Kaplan-Meier curves according to smoking history. PY: Pack-years.

Table IV. Overall survival (OS) durations estimated from Kaplan–Meier curves.

CI: Confidence interval; PY: pack-years.

Prognosis by smoking status – multivariate model. In the stratified multivariate Cox model, smoking history had no relation to OS (p=0.9). Likewise, no relationship to OS according to smoking history was found in subgroups of patients who had undergone radical resections (p>0.9) or in patients with stage IV disease with pharmacological oncology treatment (p=0.8). The model details for the whole cohort, as well as for the two aforementioned subgroups are shown in Table V.

Table V. Multivariate Cox proportional hazards model stratified by Eastern Cooperative Oncology Group performance status, stage at diagnosis and treatment.

CI: Confidence interval; HR: hazard ratio; PY: pack-years HRs are shown for smoking history and sex; the model was additionally adjusted for histological type of cancer and age at enrolment (estimates not shown). Statistically significant p-values are shown in bold.

Discussion

The key finding of this study is that the adverse outcomes of heavy smoking on prognosis were not associated with smoking itself but with other factors. This was also observed for surgically operated patients and patients treated with pharmacological oncological therapy (including targeted therapy and immunotherapy). Moreover, these results were found for both former and current smokers, regardless of the total number of cigarettes smoked.

Our study confirmed characteristics of non-smokers and smokers with NSCLC as being different, with a particularly higher prevalence of adenocarcinoma and women among non-smokers with NSCLC. This is also described in other studies (1-3, 13). We also showed differences in oncological treatment of advanced stages, with non-smokers being treated more frequently with targeted therapy, while smokers were treated more frequently with immunotherapy. This corresponds to a higher prevalence of targetable mutations [especially EGFR and anaplastic lymphoma kinase (ALK)] in non-smokers (8, 40).

The effect of smoking on the prognosis of patients with NSCLC has been investigated by a number of studies. Some of them, which with our results agree with, did not show significant effects on OS in multivariate models analysing smoking history at diagnosis (9-15). On the contrary, a number of other studies have shown better OS in non-smokers (or former smokers – however, only a minority of studies have examined this group of patients) compared to current smokers (15-35).

In older studies, the worse OS in smokers may be explained due to the fact that EGFR inhibitors were used without knowledge of EGFR mutations, which are more common in non-smokers (8). Moreover, in older studies, where the use of first-generation EGFR inhibitors was expected, it is also necessary to mention the reduced efficacy of erlotinib in the presence of smoking due to a decrease in its plasma level (7). Overall, treatment with EGFR inhibitors (in relapsed or metastatic disease) in smokers with a low probability of EGFR mutations and possibly lower efficacy of erlotinib did not achieve the same effect as in non-smokers (with higher prevalence of EGFR mutation), which may have significantly differently affected the OS of these groups. This is particularly true in Asian studies, where the incidence of EGFR mutations is much more common than in Europe (41). On the contrary, in our study, some patients (more frequently among smokers) were already treated with immunotherapy, which may have partially bridged this gap in treatment effectiveness. Some studies even show results of immunotherapy to be better in smokers compared to non-smokers (42, 43) while others have not confirmed this (44). This may be also due to a difference between mono-immunotherapy (probably with better results in smokers) compared with chemoimmunotherapy (probably no difference between non-smokers and smokers) (45, 46). Other factors that in some studies might have played a role in achieving a result different from that of our study were: the inclusion of patients with small-cell lung carcinomas, the absence of ECOG PS as an important prognostic factor, selection of only a limited group of patients that our study did not address in detail (only irradiated patients, only patients with a limited age range, only women or only patients with stage I disease), or a different division of the studied groups (above/below a certain number of PY regardless of smoking/non-smoking/ex-smoking history) (11, 17, 20, 23-26, 28, 30, 31, 33-35, 47, 48).

Only a very limited number of studies on this topic published after 2020 are available; since 2020, the previously widespread use of erlotinib in higher lines of therapy for patients with wild-type EGFR is no longer expected to such an extent, and immunotherapy has also entered into use. Zhu et al. did not demonstrate a difference in OS between smokers and former smokers (9), which our data are consistent with. Werner et al. then described a significant difference in OS between patients treated with radiotherapy according to smoking history, reporting that patients with history of smoking <40 PY had better survival than those with ≥40 years (47). However, our study did not specifically use these subgroups, moreover we stratified smoking history in a different way. The study by Takamori et al. is interesting, which did not demonstrate an influence of smoking history in the form of PYs on OS, which our data for operated patients are in agreement with, but did point out the significance of the duration of smoking (<40 vs. ≥40 years) (18). We do not have duration data available in our study. Wang et al. demonstrated worse OS in current and former smokers with early-stage NSCLC (28). However, they did not collect data on ECOG PS, which might have played a significant role. Eng et al. published a large study (n=48,531) of patients who participated in a TNM staging project (27). Current smokers had worse OS than former smokers, and both groups had worse OS than never smokers, even after including other covariates (including ECOG PS) in the multivariate model. However, possible influence of comorbidities cannot be ruled out.

Overall, the non-significance of the difference in our study in the multivariate model for smokers was most likely caused by other factors which accounted for the significantly worse OS for heavier smokers in the univariate model. Firstly, non-smokers had better ECOG PS than smokers. Worse OS as well as ECOG PS of heavy smokers may be related to the numerous comorbidities that are usually associated with smoking (chronic obstructive pulmonary disease, cardiovascular disease, and others) (49, 50). This idea was also addressed by Tammemagi et al. but in a multivariate model, they did not demonstrate a significant effect of the most common comorbidities on OS (31). On the other hand, this study by Tammemagi et al. assessed comorbidities individually and not as a whole, as smokers usually have multiple comorbidities at the same time, which may ultimately have led to a lower baseline ECOG PS compared to non-smokers (31).The partial influence of comorbidities on higher mortality in smokers is indicated by the study by Hanagiri et al., who described high pulmonary and cardiovascular complication-related mortality as one of the causes of poor prognosis of smokers with lung cancer (among those with lower stage of disease and female sex) (15). The difference may also be due to the greater representation of women in the non-smoker cohort, as sex was an independent prognostic factor in our study and many other studies on this topic (9, 10, 16, 17, 19, 27, 28, 30, 32, 35). A possible explanation for this might be the different biological behaviour of NSCLC in men and women (51).

Among the factors that may play a role in prognosis is the number of patients who quit smoking after their diagnosis of NSCLC, as this is likely to be associated with a better prognosis (2, 52-56). However, information on smoking cessation after diagnosis was not available in our registry, nor did the other included studies investigate the effect of smoking cessation on OS at diagnosis.

Study limitiations. A limitation of this study is its retrospective nature and the associated possible bias. Firstly, smoking status was recorded only based on patient history taking as part of routine diagnostics, which may have led to inaccuracies for some patients. Furthermore, the type of pharmacological oncological treatment for some patients was unknown, as were the criteria for selecting the treatment strategy. Moreover, patient comorbidities that might have influenced the results were not systematically collected. On the other hand, due to the total number of patients, this is a robust analysis that partially eliminates some of these possible biases.

Conclusion

In our study, a history of heavy smoking was a negative prognostic factor in univariate analysis, however, smoking history did not prove to be an independent prognostic factor considering the entire patient population, patients treated with surgery, and patients treated for metastatic disease. Other factors therefore played a role in the unfavourable prognosis of patients with a smoking history, such as performance status. The influence of comorbidities (often associated with smoking) and possible cessation of smoking after the diagnosis of NSCLC cannot be ruled out.

Conflicts of Interest

The Authors declare that they have no conflicts of interest related to this article.

Authors’ Contributions

MS and PD designed the study; PD performed statistical analyses; all authors collected and interpreted the data; MS, MM and PD wrote the manuscript with support from all other authors. MM supervised the project.

Funding

This study was supported by the Cooperation Programme, research area SURG.

Artificial Intelligence (AI) Disclosure

No artificial intelligence (AI) tools, including large language models or machine learning software, were used in the preparation, analysis, or presentation of this manuscript.