Impact of socioeconomic determinants and clinical characteristics on non-small cell lung cancer survival: a real-world cohort study

Introduction

Lung cancer remains the leading cause of cancer-related mortality worldwide, accounting for approximately 18% of all cancer deaths (1). In 2020, over 2 million new cases and 1.8 million deaths were reported globally (1). In Brazil, lung cancer is the third most common cancer in men and the fourth in women (2), underscoring its significant public health impact (2).

According to the American Cancer Society (ACS) 2026 report, the 5-year relative survival rate for lung cancer has improved to 28% for those diagnosed between 2015 and 2021, up from 15% in the mid-90s. Following the advances in screening and immunotherapy the 5-year survival rates for lung cancer has improved and is now tabulated as local, regional, distant in the Surveillance, Epidemiology, and End Results (SEER) registry and all-combined survival (67%, 40%, 12% and 32%, respectively) but remains among the lowest and some of the top cause of cancer death (2-4).

Key determinants of survival include cancer stage at diagnosis, comorbidities, and histological subtype (5). Recent attention has increasingly focused on the role of socioeconomic factors in shaping patient care and clinical outcomes. Socioeconomic disparities strongly influence access to healthcare, early detection, and timely treatment. Patients from lower socioeconomic backgrounds often face barriers to advanced diagnostics and treatments, which adversely affect their outcomes (6). These challenges are particularly pronounced in middle-income countries, where significant economic inequalities persist. This study aims to evaluate long-term survival and its potential predictors within a large cohort of patients with non-small cell lung cancer (NSCLC) treated in the real-world public healthcare system of Brazil, a middle-income country. Specifically, the research assesses how clinical and socioeconomic determinants—particularly variations in the Human Development Index (HDI)—impact both overall survival (OS) and cancer-specific survival (CSS). To achieve this, we analyzed comprehensive data from the Hospital Cancer Registry (HCR) of the State of São Paulo, a geographical area with a population of 40 million inhabitants. We present this article in accordance with the STROBE reporting checklist (available at https://jtd.amegroups.com/article/view/10.21037/jtd-2026-1-0334/rc).

Methods

Patients, clinical stages, and cancer registry

A hospital-based retrospective cohort study including 16,038 patients with all complete data from diagnosed with NSCLC (5) [International Classification of Diseases for Oncology (ICD-O 3rd ed. 8140/3)], admitted with lung cancer [tumor-node-metastasis (TNM) 6^th, 7^th and 8^th edition] between January 2000 and December 2016, with follow-up through December 09, 2022, sourced from the Oncocenter Foundation of São Paulo/Brazil. All patients’ information was extracted from the Hospital’s Cancer Registry (HCR), coordinated by the Information and Epidemiology Directorate of the Sao Paulo Oncocentro Foundation (FOSP), responsible for the registry of cancer of the State of Sao Paulo (SISRHC) in 76 HCRs, and available on the FOSP website (https://www.fosp.saude.sp.gov.br/publicacoes/downloadarquivos) (7).

Patients with lung cancer, older than 18 years, with confirmed histologic NSCLC were included. It excluded the patients who had undergone previous treatment to any other neoplasm and patients with small cell lung cancer. Patients diagnosed after 2016, and those with treatment dates later than their diagnosis, and whose last information date exceeded their diagnosis date were excluded. All patients received care within the Brazilian public health system in the same geographical region (Southeast). The detailed process of study sample selection, including the specific inclusion and exclusion criteria and the final participant count, is illustrated in the study flowchart (Figure 1).

Figure 1 Flowchart of the study participant selection process, illustrating the inclusion and exclusion criteria and the final cohort of 16,038 patients with NSCLC diagnosed between 2000 and 2016 in the State of São Paulo, Brazil. NSCLC, non-small cell lung cancer.

Analyzed variables included gender, age, education level, treatment modality, clinical stage at diagnosis, and time to treatment, as well as final vital status and tumor laterality. Furthermore, we assessed healthcare resource density, including the rates of oncologists, thoracic surgeons, and their combined per capita availability. Diagnostic infrastructure was evaluated based on the number of computed tomography (CT) scanners available since 2005 and positron emission tomography (PET)/CT devices incorporated into the public healthcare system starting in 2014. Additionally, the HDI was utilized as a primary socioeconomic indicator. Regarding imaging resources, a composite variable (PET/CT + CT) was created to analyze the total volume of imaging modalities utilized in patient evaluation. We based HDI on 2010 per capita income data for each municipality in the cohort, reflecting conditions for patients treated between 2000 and 2019.

To evaluate the impact of the socioeconomic environment on 5-year NSCLC survival, we utilized the HDI as refined in 2010. While the Socio-demographic Index (SDI) emphasizes fertility and the Area Deprivation Index (ADI) is geographically limited to neighborhood-level data, the HDI is the global gold standard for standardized, peer-reviewed comparisons (8). The 2010 HDI captures three critical dimensions—longevity, knowledge, and standard of living—using a geometric mean to ensure a balanced reflection of human capability (9). This prevents high economic status from masking deficiencies in healthcare or education (10,11). Such multidimensionality is vital for NSCLC research, where long-term survival depends not only on treatment availability but also on the educational levels necessary for early symptom recognition and the robust health infrastructure required for complex 5-year follow-up (3,12,13).

Ethical statement

The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. This study was approved by the Research Ethics Committee of Hospital Alemão Oswaldo Cruz (protocol 49258615.4). Informed consent was waived in this retrospective study.

Statistical analysis

Clinical and sociodemographic characteristics of the sample (including missing category where appropriate) were described by median (interquartile range) or mean (standard deviation) for continuous variables, and absolute and relative frequencies for categorical variables.

Survival outcomes were calculated from the date of diagnosis to the date of the event of interest. OS was defined as the time to death from any cause; patients alive at the last follow-up were censored.

CSS was analyzed using a competing risks framework. The primary event of interest was death from cancer. Deaths from causes other than cancer were treated as competing events rather than simple censoring. For both outcomes, patients alive at the last contact date were censored.

Survival curves were estimated using the Kaplan-Meier method and compared via the log-rank test (14). For CSS, cumulative incidence curves were calculated to account for competing risks, with differences between groups evaluated using Gray’s test (15).

The association between covariates and survival outcomes was initially assessed using univariable Cox proportional hazards and Fine-Gray’s (16) regression models for OS and CSS, respectively. The assumption of linearity between a continuous covariate and the log-hazard in Cox regression was examined graphically with martingale residuals (17-19).

Where linearity not supported, continuous covariates were dichotomized according to optimal cut-off points determined by the Contal and O’Quigley method (20).

Covariates with univariate P<0.20, no considerable multicollinearity, and <10% missing data were included in multivariable analyses. Standard Cox regression was used for OS, whereas Fine-Gray regression (26) was applied for CSS to address competing events. A backward selection (BS) procedure (removal alpha =0.05) defined the final independent predictors for both outcomes (21,22). Effect sizes are presented as hazard ratios (HRs) for OS and sub-distribution HRs (sHRs) for CSS, with 95% confidence intervals (CIs) (23,24).

Covariates exhibiting multicollinearity, defined by a variance inflation factor (VIF) exceeding 2.5, were excluded from the multivariable Cox models during BS selection (17,18,25,26). To maintain parsimony and prevent over-parameterization, systematic testing for two-way interactions was not performed, avoiding the detection of spurious, clinically meaningless associations in this highly powered cohort (27,28).

Given the substantial proportion of missing data for education level and clinical stage, sensitivity analyses were performed using multivariable backward Cox regression following missing data imputation. Imputation was conducted using a machine learning-based approach, specifically a 1-nearest neighbor (1-NN) learner, to enhance predictive accuracy by capturing complex patterns within the covariates (29,30). The HRs and significance levels from this imputed model were then evaluated alongside those of the primary complete-case model to qualitatively assess the robustness of the identified independent predictors.

The proportional hazards assumption was verified using weighted Schoenfeld residuals plotted against time, with penalized B-spline curves (and 95% CIs) fitted to the points (13,20-22). No violations of the proportional hazards assumption were observed (31-34).

Before BS, VIFs values were calculated to assess multicollinearity among the candidate predictors. To obtain the VIFs, a multiple linear regression model was fitted with time to event as the dependent variable and the candidate predictors as covariates (25).

A VIF greater than 2.5 was considered to indicate considerable multicollinearity; therefore, variables exceeding this threshold were excluded from the multiple Cox regression analysis during BS (26).

Proportional hazards assumption was assessed by using weighted Schoenfeld residuals plotted against time with a penalized B-spline curve (with 95% CI) fitted to the points (23,32-35). There was no violation in proportional hazards.

Uno’s C-statistic (36) were used to assess the discrimination of the multiple cox regression models. C-stats values were interpreted as: <0.60, poor discrimination; 0.60 to 0.75, possibly helpful discrimination; and >0.75, clearly useful discrimination (37,38).

Except for multivariable regression models, which underwent sensitivity analysis using multiple imputation, all other analyses followed the complete-case principle, excluding individuals with missing data (25). Median follow-up and its 95% CI were calculated using the reverse Kaplan-Meier approach based on OS (35,39).

Statistical significance was defined by a two-sided P value <0.05. Analyses were conducted using SAS 9.4 (SAS Institute Inc., Cary, NC, USA), while missing data imputation was performed separately with Orange Data Mining (version 3.38).

Results

Patient demographics, clinical stages, specialist access, and HDI trends

Patients were mostly male (62.8%), and only 10.5% having completed high school. Surgical intervention was performed in 25.9% patients, while 78.9% were diagnosed at stage III and IV. Patients mostly resided in regions with an aggregate of 22.97 oncologists and thoracic surgeons per million people, including 17.6 oncologists per million people and 5.4 thoracic surgeons per million people. Overall medical density was 2.9/1,000 people (Table 1).

Univariate predictors for all-cause mortality

Univariate Cox regression showed that gender, education level, clinical stage, PET/CT + CT devices [2005–2019], PET/CT devices [2014–2019], age, income, HDI 2010, and medical rate were associated with OS.

Males had a 28% higher risk of all-cause mortality than females (HR 1.28, 95% CI: 1.24–1.32; P<0.001). Advanced clinical stages increased mortality risks: stage IV (HR 4.50, 95% CI: 4.22–4.80; P<0.001), stage III (HR 2.95, 95% CI: 2.76–3.15; P<0.001), and stage II (HR 1.73, 95% CI: 1.59–1.89; P<0.001) compared to stage I.

Each 10-year increase in age raised the risk of all-cause mortality by 4% (HR 1.04, 95% CI: 1.03–1.06; P<0.001). Each 0.1-unit increase in HDI 2010 reduced the risk of all-cause mortality by 7% (HR 0.93, 95% CI: 0.90–0.96; P<0.001). Higher medical rates improved survival (HR 0.95, 95% CI: 0.92–0.97; P<0.001) (Table 2).

Predictors of all-cause mortality from multiple Cox regression

In the multivariate analysis, gender, clinical stage, age, and HDI 2010 independently predicted all-cause mortality. Male gender (HR 1.27, 95% CI: 1.22–1.31; P<0.001), advanced clinical stages (stage IV: HR 4.58, 95% CI: 4.29–4.88; P<0.001; stage III: HR 2.90, 95% CI: 2.71–3.09; P<0.001; and stage II: HR 1.69, 95% CI: 1.55–1.85; P<0.001), and age significantly increased mortality risk. Each 10-year increase in age raised the risk of all-cause mortality by 7% (HR 1.07, 95% CI: 1.06–1.09; P<0.001). Each 0.1-unit increase in HDI 2010 reduced overall mortality risk by 2% (HR 0.98, 95% CI: 0.94–1.01; P<0.001) (Table 3).

Predictors of all-cause mortality from multiple Cox regression after missing imputation are shown in Appendix 1.

Univariate predictors for cancer-specific mortality

The results of the univariate regression analyses for CSS are given in Table 4 and significant associations were observed for the same covariates as for OS (Table 4).

Predictors of cancer-specific mortality from multiple Cox regression

The multivariable Fine-Gray model identified three independent predictors of cancer-specific mortality after adjusting for deaths from other causes as competing risks (Table 5).

Male gender was associated with an increased risk of cancer-specific death (sHR 1.21; 95% CI: 1.16–1.25; P<0.001) compared to females. Clinical stage at diagnosis was the most significant predictor of outcome. Relative to stage I, patients diagnosed at stage II (sHR 2.20; 95% CI: 1.99–2.44), stage III (sHR 3.70; 95% CI: 3.42–4.00), and stage IV (sHR 5.73; 95% CI: 5.31–6.18) demonstrated progressively higher risks of mortality (all P<0.001 compared to stage I). Additionally, the socioeconomic proxy PET/CT + CT devices (2005–2019) was significantly associated with survival. Patients in the highest group (≥34) demonstrated improved outcomes, with a 7% reduction in the hazard of cancer-specific death (sHR 0.93; 95% CI: 0.90–0.97; P<0.001) compared to those below this threshold (Table 5).

Predictors of cancer-specific mortality from multiple Fine-Gray regression after missing imputation are shown in Appendix 1.

Discussion

This study involving 16,038 patients with lung cancer underscores the critical factors influencing survival outcomes, including the role of socioeconomic conditions, healthcare infrastructure, and technology availability. Disparities in education, access to diagnostic tools, and human development emerge as key areas requiring attention to improve survival rates in Brazil.

While higher HDI-GNI was associated with reduced all-cause mortality, cancer-specific mortality was more closely linked to healthcare infrastructure, specifically PET/CT availability, rather than macro-level development indices.

Typical demographic factors, such as male gender and advanced age, were associated with poorer outcomes; however, while each 10-year increase in age raised all-cause mortality by 7%, it was not an independent predictor of cancer-specific mortality (43-47).

The HDI serves as a robust macro-level predictor of lung cancer outcomes, reflecting a complex interplay between socioeconomic infrastructure and clinical metrics (44,48-52).

According to the Global Cancer Statistics 2024 provided by Sung and Ferlay [2024], while high-HDI regions often report higher crude incidence rates due to historical exposure factors, they paradoxically maintain superior survival outcomes compared to developing areas (53).

In a systematic evaluation of global disparities, Onwuka et al. [2025] have demonstrated that the era of immunotherapy has widened the survival gap, as regions with lower socioeconomic status exhibit higher mortality-to-incidence ratios and limited access to precision medicine (53).

Furthermore, in a comprehensive global ecological study, Soheylizad et al. [2016] evaluated the correlation between the HDI and lung cancer metrics across 172 countries (54). The study reported that in 2012, lung cancer accounted for 13% of all cancer diagnoses, with a global crude incidence rate of 25.9 per 100,000 and a mortality rate of 22.5 per 100,000. Their statistical analysis revealed a strong positive correlation between HDI and both lung cancer incidence (R=0.79, P<0.05) and mortality (R=0.77, P<0.05).

Linear regression models further demonstrated that higher life expectancy, increased mean years of schooling, and greater Gross National Income (GNI) per capita—alongside urbanization and obesity prevalence—significantly contribute to the rising burden of lung cancer in high and very high HDI regions. The authors concluded that while the epidemiological transition increases lung cancer prevalence in developed settings, the mitigation of this burden necessitates a dual approach: robust primary prevention (tobacco and alcohol control) and enhanced equitable access to curative treatments to lower mortality rates (43).

Limited access to advanced diagnostic technologies, such as PET/CT and endobronchial ultrasound (EBUS) that is not provided by Brazilian’s Public Health System, and reduced treatment adherence due to geographic barriers in rural areas significantly hinder outcomes in socioeconomically disadvantaged regions.

The survival gap between socioeconomic strata is further exacerbated by the restricted availability of precision medicine within the public sector. Currently, molecular testing for actionable mutations, such as epidermal growth factor receptor (EGFR), and the subsequent provision of targeted therapies or immunotherapy are not routinely funded by the SUS. Consequently, patients in lower HDI regions are not only diagnosed at more advanced stages but are also deprived of the therapeutic advances that have significantly improved survival rates in high-income countries over the last decade. This underscores the role of socioeconomic status as a primary determinant of biological and treatment equity (6,8,10,48).

Innovative treatments, including targeted therapies and immunotherapy, remain largely unavailable within the Brazilian public health system (SUS), further exacerbating disparities in lung cancer survival. Our finding that male gender is associated with poorer lung cancer outcomes aligns with emerging evidence on sex-specific differences; however, our dataset lacked information on actionable gene alterations—primarily EGFR mutations. These alterations are more prevalent in women and may contribute to superior survival outcomes; nonetheless, as previously noted, such molecular testing and subsequent therapies are not routinely provided in the Brazilian Public Health System.

Concerning the gender variable, Florez et al. [2024] highlight that women face unique risk factors, including greater exposure to second-hand smoke and indoor pollution, alongside biological distinctions such as hormonal influences and enhanced immune responses, which may improve treatment tolerability and efficacy. Additionally, societal factors, including gender roles that delay diagnosis and underrepresentation of women in clinical trials (<40% enrollment), may exacerbate disparities by relying on male-dominated treatment protocols. These multifaceted factors suggest that while actionable mutations play a role, environmental exposures and systemic inequities likely contribute to the observed gender differences, warranting further studies integrating genomic and social determinants (55).

Comparatively, in the United States, sex, education, and socioeconomic status are also associated with lung cancer outcomes, with better survival rates observed among women, individuals with higher income, and those with higher education levels. However, Brazil faces additional challenges due to its socioeconomic heterogeneity, which amplify barriers related to healthcare infrastructure and technology access (55).

Technology access inequality

CT and PET/CT are vital for lung cancer management, aiding in staging, treatment planning, and monitoring Brazil’s PET-CT density is just 0.53 per million people, requiring a doubling of equipment to meet recommendations (56). However, our findings did not establish a clear correlation between this deficiency and mortality outcomes in lung cancer patients (57).

There are significant disparities in PET-CT availability between low- and high-income countries, with more than 30 scanners per million people in high-income countries compared to less than one per million—or none at all—in low-income countries. This disparity leads to late diagnostic and inadequate cancer staging, contributing to increased cancer-related mortality and morbidity (47). Closing the global diagnostic gap requires a minimum investment of US$229 million in PET-CT infrastructure, with 62% of these resources—US$142 million—critically needed to address the severe technological deficit in 61 low-income countries (58).

Simulation models suggest that integrating imaging and treatment could increase survival rates tenfold, potentially saving 9.5 million lives by 2030 for six more prevalent cancers, including lung-cancer (59). The synergy between advanced diagnostic tools and effective treatment not only improves health outcomes but also provides sustainable economic benefits (46).

In Brazil, with fewer than 478 nuclear medicine specialists and only 23 training centers, most concentrated in the Southeast, limiting the access to early diagnosis and specialized care, directly compromising survival rates (56).

In this study, 79% of participants were diagnosed in stages III and IV. Limited access to low-dose CT screening for at-risk populations in Brazil leads to a predominance of advanced-stage diagnoses, with approximately 85% of patients with lung cancer diagnosed at stages III and IV (45,46,60). In contrast, countries with widespread screening programs, like the United States and the United Kingdom, detect lung cancer earlier, with 47% and around 29% of cases, respectively, diagnosed at stages I and II (45,46,60).

Several limitations inherent to the retrospective nature of this registry-based study must be acknowledged. Firstly, internal differences in therapeutic modalities between hospitals were not available, nor were the specific patient characteristics across different hospital types. Secondly, our database does not capture the methodology utilised for clinical or pathological staging, which may affect the precision of our stage-specific analysis. Furthermore, data regarding surgical quality, including the status of surgical margins and the extent of mediastinal lymphadenectomy—are not recorded in the registry. Selection bias may also be present, as the database consists of patients already integrated into the foundation’s system; this potentially excludes those who succumbed to the disease before accessing specialized care, thereby skewing survival estimates toward a more stable patient population. Additionally, the lack of data on the Charlson Comorbidity Index (CCI) and specific surgical metrics—such as the radicality of mediastinal lymphadenectomy—poses a threat to the internal validity of our survival estimates. These unmeasured confounders may influence mortality outcomes independently of the socioeconomic variables analysed, and their absence should be considered when interpreting the strength of the associations reported in this study.

Conclusions

This study highlights the significant impact of socioeconomic and healthcare determinants on lung cancer survival within a large cohort from a middle-income country. Our findings demonstrate that, among patients with NSCLC, favorable regional socioeconomic determinants—specifically higher HDI and greater PET/CT availability—are independently associated with improved overall and CSS, respectively. Consequently, policymakers should prioritize equitable resource allocation, investing in oncology centers and diagnostic facilities in underserved areas to mitigate diagnostic delays and improve clinical outcomes.

Acknowledgments

None.

Footnote

Reporting Checklist: The authors have completed the STROBE reporting checklist. Available at https://jtd.amegroups.com/article/view/10.21037/jtd-2026-1-0334/rc

Peer Review File: Available at https://jtd.amegroups.com/article/view/10.21037/jtd-2026-1-0334/prf

Funding: None.

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://jtd.amegroups.com/article/view/10.21037/jtd-2026-1-0334/coif). The authors have no conflicts of interest to declare.

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. This study was approved by the Research Ethics Committee of Hospital Alemão Oswaldo Cruz (protocol 49258615.4). Informed consent was waived in this retrospective study.

Open Access Statement: This is an Open Access article distributed in accordance with the Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International License (CC BY-NC-ND 4.0), which permits the non-commercial replication and distribution of the article with the strict proviso that no changes or edits are made and the original work is properly cited (including links to both the formal publication through the relevant DOI and the license). See: https://creativecommons.org/licenses/by-nc-nd/4.0/.

References

1. Sung H, Ferlay J, Siegel RL, et al. Global Cancer Statistics 2020: GLOBOCAN Estimates of Incidence and Mortality Worldwide for 36 Cancers in 185 Countries. CA Cancer J Clin 2021;71:209-49. [Crossref] [PubMed]
2. Cancer Today [cited 2026 Apr 26]. Available online: https://gco.iarc.who.int/today/en/dataviz
3. Siegel RL, Kratzer TB, Wagle NS, et al. Cancer statistics, 2026. CA Cancer J Clin 2026;76:e70043. [Crossref] [PubMed]
4. International Classification of Diseases for Oncology, 3rd Edition (ICD-O-3) [cited 2026 Apr 26]. Available online: https://www.who.int/standards/classifications/other-classifications/international-classification-of-diseases-for-oncology
5. Siegel RL, Miller KD, Fuchs HE, et al. Cancer statistics, 2022. CA Cancer J Clin 2022;72:7-33. [Crossref] [PubMed]
6. McMaughan DJ, Oloruntoba O, Smith ML. Socioeconomic Status and Access to Healthcare: Interrelated Drivers for Healthy Aging. Front Public Health 2020;8:231. [Crossref] [PubMed]
7. Liu X. Survival analysis: Models and applications. Wiley; 2012:1-446.
8. Kind AJH, Buckingham WR. Making Neighborhood-Disadvantage Metrics Accessible - The Neighborhood Atlas. N Engl J Med 2018;378:2456-8. [Crossref] [PubMed]
9. United Nations Development Programme. Human Development Report 2010: The Real Wealth of Nations: Pathways to Human Development. New York, NY: United Nations Development Programme (UNDP); 2010.
10. Yang X, Deng L, Li M, et al. Impact of socioeconomic status on cancer staging, survival in non-small cell lung cancer. Front Public Health 2022;10:992944. [Crossref] [PubMed]
11. Van der Heyden JH, Schaap MM, Kunst AE, et al. Socioeconomic inequalities in lung cancer mortality in 16 European populations. Lung Cancer 2009;63:322-30. [Crossref] [PubMed]
12. GBD Results | Institute for Health Metrics and Evaluation [cited 2026 Apr 26]. Available online: https://www.healthdata.org/data-tools-practices/interactive-visuals/gbd-results
13. Tan HB, Law SH. Nonlinear dynamics of the finance-inequality nexus in developing countries. J Econ Inequal 2012;10:551-63.
14. Borucka J. Assessment of Cox Proportional Hazard Model Adequacy Using PROC PHREG and PROC GPLOT. PhUSE 2010. Paper SP05.
15. Gray RJ. A Class of K-Sample Tests for Comparing the Cumulative Incidence of a Competing Risk. The Annals of Statistics 1988;16:1141-54.
16. Zhang X, Zhang MJ, Fine J. A proportional hazards regression model for the subdistribution with right-censored and left-truncated competing risks data. Stat Med 2011;30:1933-51. [Crossref] [PubMed]
17. Data T. SURVIVAL ANALYSIS Techniques for Censored and Second Edition [cited 2026 Apr 26]. Available online: https://books.google.com/books/about/Survival_Analysis.html?hl=pt-BR&id=jS2Cy0lezJIC
18. Brembilla A, Olland A, Puyraveau M, et al. Use of the Cox regression analysis in thoracic surgical research. J Thorac Dis 2018;10:3891-6. [Crossref] [PubMed]
19. Contal C, O’Quigley J. An application of changepoint methods in studying the effect of age on survival in breast cancer. Computational Statistics & Data Analysis 1999;30:253-70.
20. Liu Y, Nickleach DC, Zhang C, et al. Carrying out streamlined routine data analyses with reports for observational studies: introduction to a series of generic SAS (®) macros. F1000Res 2018;7:1955. [Crossref] [PubMed]
21. Zhang C, Liu Y, Jia Y. SAS® Macros on Performing Competing Risks Survival Data: CIF Plot, Backward Elimination Fine & Gray’s Model and Cause-Specific Hazard Model. J Med Stat Inform 2019;7:2.
22. Therneau TM, Grambsch PM. Modeling Survival Data: Extending the Cox Model 2000. Statistics for Biology and Health. Springer; 2000. doi:10.1007/978-1-4757-3294-8.
23. Kleinbaum DG, Klein M. Survival Analysis: A Self-Learning Text, Third Edition. Statistics for Biology and Health. Springer; 2012. doi:10.1007/978-1-4419-6646-9.
24. Demšar J, Erjavec A, Hočevar T, et al. Orange: Data Mining Toolbox in Python. Journal of Machine Learning Research. 2013;14:2349-53.
25. Introduction to Regression Methods for Public Health Using R - 1st Edi [cited 2026 Apr 26]. Available online: https://www.routledge.com/Introduction-to-Regression-Methods-for-Public-Health-Using-R/Nahhas/p/book/9781032203072
26. Dormann CF, Elith J, Bacher S, et al. Collinearity: A review of methods to deal with it and a simulation study evaluating their performance. Ecography 2013;36:27-46.
27. Harrell FE Jr. Regression Modeling Strategies: With Applications to Linear Models, Logistic and Ordinal Regression, and Survival Analysis 2015. 2nd edition. Springer Series in Statistics; 2015. doi:10.1007/978-3-319-19425-7.
28. Royston P, Sauerbrei W. Multivariable Model-Building: A Pragmatic Approach to Regression Analysis based on Fractional Polynomials for Modelling Continuous Variables. Multivariable Model-Building: A Pragmatic Approach to Regression Analysis based on Fractional Polynomials for Modelling Continuous Variables. 2008:1-303. doi: 10.1002/9780470770771.
29. Park SY, Park JE, Kim H, et al. Review of Statistical Methods for Evaluating the Performance of Survival or Other Time-to-Event Prediction Models (from Conventional to Deep Learning Approaches). Korean J Radiol 2021;22:1697-707. [Crossref] [PubMed]
30. Gardiner JC. Evaluating the accuracy of clinical prediction models for binary and survival outcomes.
31. Jeon Y, Lee WK. Competing Risk Model in Survival Analysis. Cardiovascular Prevention and Pharmacotherapy 2020;2:77.
32. Grambsch PM, Therneau TM. Proportional hazards tests and diagnostics based on weighted residuals. Biometrika 1994;81:515-26.
33. SAS Documentation | SAS Support [cited 2026 Apr 28]. Available online: https://support.sas.com/en/documentation.html
34. Lu J, Shen D. Application of Survival Analysis in Multiple Events Using SAS. PharmaSUG 2018. 2018.
35. Schemper M, Smith TL. A note on quantifying follow-up in studies of failure time. Control Clin Trials 1996;17:343-6. [Crossref] [PubMed]
36. Uno H, Cai T, Pencina MJ, et al. On the C-statistics for evaluating overall adequacy of risk prediction procedures with censored survival data. Stat Med 2011;30:1105-17. [Crossref] [PubMed]
37. Alba AC, Agoritsas T, Walsh M, et al. Discrimination and Calibration of Clinical Prediction Models: Users’ Guides to the Medical Literature. JAMA 2017;318:1377-84. [Crossref] [PubMed]
38. D’Agostino RB, Nam BH. Evaluation of the Performance of Survival Analysis Models: Discrimination and Calibration Measures. Handbook of Statistics 2003;23:1-25.
39. Sathish N, Wu CLA. Let’s Flip: An Approach to Understand Median Follow-up by the Reverse Kaplan-Meier Estimator from a Statistical Programmer’s Perspective.
40. Harrell FE Jr, Califf RM, Pryor DB, et al. Evaluating the yield of medical tests. JAMA 1982;247:2543-6.
41. Guo C, So Y, Jang W. Evaluating Predictive Accuracy of Survival Models with PROC PHREG.
42. SAS Help Center: Overview: PHREG Procedure [cited 2026 Apr 28]. Available online: https://documentation.sas.com/doc/en/pgmsascdc/v_074/statug/statug_phreg_overview.htm
43. da Silva Brandão Rodrigues M, Magne TM, Santos-Oliveira R. Inside Brazilian nuclear medicine: numbers, projections and behaviors. Eur J Nucl Med Mol Imaging 2023;50:3809-16. [Crossref] [PubMed]
44. Câncer de pulmão — Instituto Nacional de Câncer - INCA [cited 2026 Apr 28]. Available online: https://www.gov.br/inca/pt-br/assuntos/cancer/tipos/pulmao
45. Instituto Nacional de Câncer - INCA. Distribuição de Oncologistas no Brasil. 2021 2021 [cited 2026 May 10]. Available online: https://www.gov.br/inca/pt-br
46. Oliveira NPD, de , Mendes T de MC, Vasconcelos HS, de , Castro JL, de , Souza DLB, de . Rede de Atenção Oncológica: Distribuição Espacial dos Serviços e da Força de Trabalho em Saúde no Brasil. Revista Brasileira de Cancerologia. 2025;71:e-145282.
47. Gallach M, Mikhail Lette M, Abdel-Wahab M, et al. Addressing Global Inequities in Positron Emission Tomography-Computed Tomography (PET-CT) for Cancer Management: A Statistical Model to Guide Strategic Planning. Med Sci Monit 2020;26:e926544. [Crossref] [PubMed]
48. Hochhegger B, Camargo S, da Silva Teles GB, et al. Challenges of Implementing Lung Cancer Screening in a Developing Country: Results of the Second Brazilian Early Lung Cancer Screening Trial (BRELT2). JCO Glob Oncol 2022;8:e2100257. [Crossref] [PubMed]
49. Lung Association A. State of Lung Cancer 2023 Report.
50. U.S. Cancer Statistics Lung Cancer Stat Bite | U.S. Cancer Statistics | CDC [cited 2026 Apr 28]. Available online: https://www.cdc.gov/united-states-cancer-statistics/publications/lung-cancer-stat-bite.html
51. Lung cancer statistics | Cancer Research UK [cited 2026 Apr 28]. Available online: https://www.cancerresearchuk.org/health-professional/cancer-statistics/statistics-by-cancer-type/lung-cancer
52. Bray F, Laversanne M, Sung H, et al. Global cancer statistics 2022: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries. CA Cancer J Clin 2024;74:229-63. [Crossref] [PubMed]
53. Onwuka JU, Zahed H, Feng X, et al. Association between socioeconomic position and lung cancer incidence in 16 countries: a prospective cohort consortium study. EClinicalMedicine 2025;82:103152. [Crossref] [PubMed]
54. Soheylizad M, Khazaei S, Khazaei S, et al. Relation Between Lung Cancer Incidence and Mortality Rates with Human Development Index and Its Components: A Global Ecological Study. Iranian Journal of Cancer Prevention 2016;9:e5310.
55. Florez N, Kiel L, Riano I, et al. Lung Cancer in Women: The Past, Present, and Future. Clin Lung Cancer 2024;25:1-8. [Crossref] [PubMed]
56. Ministério da Saúde - Brasil. Relatório de Equipamentos de Diagnóstico por Imagem no Brasil 2022 [cited 2026 May 10]. Available online: https://www.gov.br/saude/pt-br/composicao/sectics/desid
57. Hricak H, Abdel-Wahab M, Atun R, et al. Medical imaging and nuclear medicine: a Lancet Oncology Commission. Lancet Oncol 2021;22:e136-72. [Crossref] [PubMed]
58. Stewart BW, Wild CP. World Cancer Report 2014. Lyon: IARC Press; 2014.
59. dos Santos RS, Franceschini JP, Chate RC, et al. Do Current Lung Cancer Screening Guidelines Apply for Populations With High Prevalence of Granulomatous Disease? Results From the First Brazilian Lung Cancer Screening Trial (BRELT1). Ann Thorac Surg 2016;101:481-6; discussion 487-8.
60. Holden CE, Wheelwright S, Harle A, et al. The role of health literacy in cancer care: A mixed studies systematic review. PLoS One 2021;16:e0259815. [Crossref] [PubMed]