Impact of the immuno-oncology era on cerebrovascular disease and cancer-specific mortality in older adults with advanced non-small cell lung cancer: a competing-risks cohort study

Introduction

Non-small cell lung cancer (NSCLC) remains a leading cause of cancer-related mortality worldwide, and outcomes are particularly poor among older adults with advanced-stage disease (1,2). Since 2015, the clinical introduction of immune checkpoint inhibitors (ICIs) has fundamentally altered the treatment landscape for advanced NSCLC (3). In the past decade, ICIs have fundamentally reshaped the treatment landscape of advanced NSCLC, improving survival in multiple clinical settings (4). Moreover, clinical trials have demonstrated the efficacy of these agents in older adults; however, real-world outcomes in the “oldest-old” population remain complex due to the presence of significant comorbidities (5). The survival gains observed in routine practice may be accompanied by changes in non-cancer mortality patterns, especially among older populations who carry a higher baseline burden of vascular disease and frailty.

Cerebrovascular disease (CVD), including ischemic stroke and intracranial hemorrhage, is a leading cause of death and disability in the elderly (6,7). In addition to traditional vascular risk factors, contemporary cancer therapies may influence thrombo-inflammation, endothelial function, and immune-mediated vascular injury (8,9). In patients with advanced cancer, CVD serves as a formidable “competing risk”—a phenomenon where a patient may die from a non-cancer cause before the cancer-specific outcome occurs. The prolongation of life facilitated by ICIs may inadvertently increase the cumulative incidence of cerebrovascular events by extending the “at-risk” window or through potential immune-related inflammatory effects on the vasculature (10,11). As a result, an “improved cancer survival but increased competing mortality” scenario is plausible in the immuno-oncology (I-O) era: patients may live longer and thus accumulate more time at risk for non-cancer events, while ICI-related inflammatory or prothrombotic mechanisms may further increase vascular risk.

Despite increasing attention to cardiotoxicity in oncology, population-level evidence describing how the introduction of ICIs has reshaped cause-specific mortality—particularly cerebrovascular mortality—among older adults with advanced NSCLC remains limited. Importantly, standard survival models may be inadequate when competing events (e.g., NSCLC death) preclude observation of CVD death. Therefore, competing-risks methods are essential to quantify the real-world cumulative burden of CVD death since the I-O era.

In this study, we used the Surveillance, Epidemiology, and End Results (SEER) database to evaluate whether the introduction of ICI therapy corresponds to changes in CVD mortality and NSCLC-specific mortality among adults aged ≥70 years diagnosed with advanced NSCLC. We compared outcomes between the pre-immuno-oncology era (Pre-I-O; 2007–2014) and a post-immuno-oncology era (Post-I-O; 2015–2022) using propensity score matching (PSM), Fine-Gray competing-risks models, cause-specific Cox regression, and multistate modeling, with subgroup analyses to assess consistency across clinically relevant strata. We present this article in accordance with the STROBE reporting checklist (available at https://tcr.amegroups.com/article/view/10.21037/tcr-2026-1-0169/rc).

Methods

Data source and study design

We conducted a retrospective cohort study using the SEER-Medicare linked database. SEER provides population-based cancer registry data, including tumor characteristics and cause of death, and Medicare claims provide information on demographics and treatments (12). The study was designed to compare mortality outcomes before and after the introduction and widespread adoption of I-O therapies for advanced NSCLC in routine care. The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments.

Study population

We identified patients who: (I) were aged ≥70 years at diagnosis; (II) had advanced NSCLC at diagnosis (as defined in your analytic datasets using SEER staging variables); (III) were diagnosed between January 1, 2007 and December 31, 2022.

Patients were categorized by diagnosis period into Pre-I-O (January 2007 to December 2014) and Post-I-O (January 2015 to December 2022).

The index date was the date of NSCLC diagnosis. Follow-up started at diagnosis and continued until death or end of follow-up.

Exposure definition

The primary exposure was the I-O era, operationalized as diagnosis period: Pre-I-O vs. Post-I-O. This approach captures system-level changes in treatment availability and clinical adoption associated with ICIs, rather than requiring confirmed ICI receipt in claims (which can be incomplete or influenced by coding changes over time).

Outcomes

The primary mortality outcomes were CVD death, NSCLC-specific death and other-cause death (as a competing event category). Cause of death was determined using SEER cause-of-death fields and categorized into CVD, NSCLC, or other causes according to your predefined coding rules.

Covariates

We adjusted for the following covariates: (I) demographics and social factors: age, sex, race, marital status, rural urban distribution, median household income; (II) tumor characteristics: tumor grade, tumor (T) stage, node (N) stage, metastasis (M) stage (all strictly analyzed as categorical variables); and (III) treatment variables: surgery recode, radiotherapy recode, chemotherapy recode.

PSM

To reduce confounding between eras, we estimated each patient’s propensity to be diagnosed in the Post-I-O using the covariates above. We then performed 1:1 PSM to generate balanced cohorts for the primary analyses. Covariate balance was assessed using standardized differences and/or distributional checks (as implemented in your analysis).

Statistical analysis

Cumulative incidence (absolute risk)

We estimated and visualized the cumulative incidence functions (CIFs) for CVD death and NSCLC death, treating other causes of death as competing events. Differences in CIF patterns were interpreted as changes in the population-level absolute burden of each cause of death across eras.

Primary model: Fine-Gray competing-risks regression

We used Fine-Gray sub-distribution hazard models to estimate the association between I-O era and CVD death [sub-distribution hazard ratio (sHR)] as well as NSCLC-specific death (sHR). This approach directly models the CIF and is appropriate when the clinical question concerns absolute risk accumulation over time under competing risks. To aid clinical interpretation, the sHR estimates the relative probability of a specific event occurring, while appropriately accounting for the fact that some patients will experience a competing event (e.g., dying from cancer) and are thus removed from the “at-risk” pool for the primary event

Sensitivity analysis: cause-specific Cox regression

As sensitivity analysis, we fitted cause-specific Cox proportional hazards models for: CVD death (treating non-CVD deaths as competing censoring) and NSCLC death (treating non-NSCLC deaths as competing censoring). Cause-specific hazard ratios (HRs) provide complementary inference regarding instantaneous hazard among those still event-free.

Additional sensitivity/subgroup framework: multistate model

To further evaluate robustness and to support subgroup inference within a coherent event-process framework, we implemented a multistate model representing transitions from “Alive after diagnosis” to each absorbing death state (NSCLC death, CVD death, other death). Era effects were estimated as hazard ratios for transitions under the multistate formulation.

Subgroup analyses

We performed subgroup analyses (pre-specified in your results) across clinically relevant strata, including age groups, sex, race categories, and treatment-related variables, reporting hazard ratios and confidence intervals (CIs). Consistency of effect direction and statistical interaction patterns were assessed descriptively and/or by interaction terms as implemented. All tests were two-sided, with significance set at P<0.05.

Results

Cohort characteristics and matching

The screening process for NSCLC patients is shown in Figure 1. Before matching, the study included 40,585 patients in the Post-I-O and 41,772 patients in the Pre-I-O. As shown in Table 1, baseline characteristics differed between eras across several domains, including sociodemographic variables (e.g., income distribution), tumor features (grade and stage distributions), and treatment patterns (surgery, radiotherapy, chemotherapy), supporting the need for confounding control.

Figure 1 The screening process for NSCLC patients. NSCLC, non-small cell lung cancer; Post-I-O, post-immuno-oncology era; Pre-I-O, pre-immuno-oncology era; SEER, Surveillance, Epidemiology, and End Results.

After 1:1 PSM, 37,169 patients were retained in each era cohort (Table 2). Matching improved balance across key demographic, tumor, and treatment variables. Some variables (notably income distribution and certain staging or race categories) showed residual differences, which is expected in very large samples and was further addressed through multivariable adjustment in regression models.

Cumulative incidence patterns

Cumulative incidence analyses demonstrated a divergent pattern after the introduction of I-O: the cumulative incidence of NSCLC-specific death was lower in the Post-I-O than in the Pre-I-O, consistent with improved cancer outcomes in routine practice (Figure 2A). In contrast, the cumulative incidence of CVD death was higher in the Post-I-O, indicating an increased absolute burden of cerebrovascular mortality despite improved cancer survival (Figure 2B). These CIF patterns motivated formal competing-risks regression to quantify era-associated differences while appropriately accounting for competing causes of death.

Figure 2 Cumulative incidence curve. Cumulative incidence rate of NSCLC-related death (A) and CVD-related death (B). CVD, cerebrovascular disease; NSCLC, non-small cell lung cancer; Post-I-O, post-immuno-oncology era; Pre-I-O, pre-immuno-oncology era.

Primary competing-risks analysis

The result of multivariable Fine-Gray competing-risks regression is shown in Table 3. Diagnosis in the Post-I-O was associated with a substantially lower sub-distribution hazard of NSCLC-specific death with a sHR of 0.624 (95% CI: 0.613–0.635; P<0.001). This corresponds to an approximate 38% relative reduction in the sub-distribution hazard of NSCLC death in the Post-I-O, consistent with lower CIF observed in Figure 3. Conversely, diagnosis in the Post-I-O was associated with a higher sub-distribution hazard of CVD death with a sHR of 1.424 (95% CI: 1.293–1.657, P<0.001). This indicates an approximate 42% relative increase in the sub-distribution hazard of CVD death, aligning with the higher CVD CIF in Figure 2.

Figure 3 Subgroup analysis of all caused death. CI, confidence interval; G, grade; HR, hazard ratio; M, metastasis; Post-I/O, post-immuno-oncology era; Pre-I/O, pre-immuno-oncology era; T, tumor.

Sensitivity analysis

Cause-specific Cox models yielded consistent directions of association. Diagnosis in the Post-I-O was still significantly associated with a substantially lower risk of NSCLC-specific death (HR =0.739; 95% CI: 0.727–0.752, P<0.001) after adjusting for other confounding factors (Table 4). After adjusting for other confounding factors, patients diagnosed in the Post-I-O correlated with a substantially higher risk of CVD death (HR =1.425; 95% CI: 1.217–1.699; P<0.001). Moreover, female, married cases, metropolitan area, higher median household income, lower tumor grade, without distant disease, surgery of primary tumor and chemotherapy were independent protective factors for both NSCLC-specific death and CVD death. Thus, both the CIF-focused Fine-Gray approach and the instantaneous-risk cause-specific approach supported the conclusion that the Post-I-O is associated with lower NSCLC mortality but higher cerebrovascular mortality.

Multistate model analysis

In multistate modeling of the death process after diagnosis, Post-I-O diagnosis was associated with a lower overall hazard of all caused death in the modeled transition framework (HR =0.746; 95% CI: 0.735–0.758; P<0.001) (Table 5). This result is consistent with improved overall survival dynamics in the I-O era while still allowing for cause-specific increases in CVD mortality observed in competing-risks models.

Subgroup analyses

Across most subgroups, the era effect on mortality demonstrated consistent directionality, with hazard ratios generally favoring the Post-I-O for the all caused mortality pattern in the subgroup framework (HRs approximately ranging from ~0.71 to ~0.87). Effects were observed across age strata (70–74 years, 75–79 years, 80–84 years, and ≥85 years), across sex, and across most race categories. For smaller race subgroups such as American Indian/Alaska Native, CIs were wider and statistical significance was not consistently achieved, likely reflecting limited sample size and reduced precision. Collectively, these subgroup results suggest that the observed era-associated changes are broadly consistent across clinically relevant patient groups.

Discussion

In this large SEER cohort of adults aged ≥70 years with advanced NSCLC, the introduction of I-O (post-2015) was associated with a dual mortality shift. (I) Lower NSCLC-specific mortality in the Post-I-O, supported by both competing-risks regression (Fine-Gray sHR =0.624) and cause-specific Cox models (HR =0.739), and reflected in lower cumulative incidence curves. Importantly, while standard P values indicate statistical significance, evaluating the minimal clinically important difference (MCID) provides necessary context for clinical relevance. Based on the recently proposed Horita’s MCID framework for evaluating effect sizes other than mean differences, an sHR of 0.624 represents a robust and clinically meaningful reduction in cancer-specific mortality risk, confirming the profound real-world impact of the I-O era (13). (II) Higher CVD mortality in the Post-I-O, again consistent across Fine-Gray (sHR =1.424) and cause-specific Cox (HR =1.425), with higher cumulative incidence of CVD death. These findings indicate that while cancer outcomes improved in the I-O era for older adults with advanced NSCLC, the population-level burden of cerebrovascular mortality increased, highlighting the importance of integrated oncologic and vascular risk management in this vulnerable group.

Several non-mutually exclusive mechanisms may explain the observed rise in CVD death. Improved cancer control reduces early NSCLC deaths, leading to longer survival time and increased exposure window for non-cancer events such as CVD. In competing-risk terms, when the dominant competing event (NSCLC death) becomes less frequent or occurs later, the CIF for other causes (including CVD) can rise because more patients remain alive long enough to experience those events. The concordance between lower NSCLC CIF and higher CVD CIF in Figure 2 supports this explanation.

Immunotherapy can trigger systemic immune activation, endothelial inflammation, and prothrombotic pathways (14-16). Even if ICI-related vascular events are uncommon on an individual basis, small increases in risk may translate to meaningful population-level changes in older adults—particularly those with baseline vascular vulnerability. The observed increase in CVD death in both Fine-Gray and cause-specific Cox frameworks is compatible with either true risk elevation or increased detection/attribution of CVD as cause of death in longer-term survivors.

The post-2015 era includes broader changes: improved imaging, evolving coding, modifications in chemotherapy use, radiotherapy techniques, and supportive care practices. Some of these changes could influence vascular outcomes indirectly. While matching and multivariable adjustment mitigate measured confounding, unmeasured secular factors may contribute to the observed pattern.

Our results emphasize a critical “success paradox” in modern oncology: as cancer mortality declines, non-cancer causes of death become increasingly visible and clinically relevant. For older adults with advanced NSCLC in the I-O era, cerebrovascular mortality appears to represent a growing competing risk. These findings support routine vascular risk assessment at diagnosis and during systemic therapy planning for older NSCLC patients and coordination between oncology and primary care/cardiology/neurology to optimize blood pressure control, diabetes management, lipid therapy, smoking cessation, and antithrombotic strategies when indicated.

This study has several strengths. This study was a large, population-based cohort focused on older adults (≥70 years), who are underrepresented in clinical trials. Moreover, competing-risks methodology (Fine-Gray) aligned with the clinical question of cumulative cause-specific mortality. We also performed robustness checks via cause-specific Cox and multistate models, plus subgroup evaluation of effect consistency.

Several limitations merit consideration: era-based exposure definition does not confirm individual ICI receipt. The Post-I-O reflects availability and adoption but includes treatment heterogeneity. Residual confounding is possible due to unmeasured factors (e.g., smoking history, performance status, frailty, comorbidities not captured or incompletely modeled, detailed regimen lines, and supportive care). Cause-of-death misclassification may occur, particularly between CVD and other causes in older adults. Changes in clinical practice over time (diagnostics, coding, supportive care) may contribute to observed differences independent of immunotherapy. Despite these limitations, the consistent direction of associations across analytic frameworks and alignment between CIF curves and model estimates strengthen confidence in the overall conclusion.

Conclusions

Among Medicare-linked older adults with advanced NSCLC, the Post-I-O [2015–2022] was associated with substantially lower NSCLC-specific mortality but higher CVD mortality compared with 2007–2014. These results underscore the need for integrated cancer and vascular risk management to maximize long-term benefits of I-O in older patients. Specifically, these findings highlight the increasingly vital role of the cardio-oncology discipline in providing comprehensive, interdisciplinary survivorship care for the aging cancer population.

Acknowledgments

None.

Footnote

Reporting Checklist: The authors have completed the STROBE reporting checklist. Available at https://tcr.amegroups.com/article/view/10.21037/tcr-2026-1-0169/rc

Peer Review File: Available at https://tcr.amegroups.com/article/view/10.21037/tcr-2026-1-0169/prf

Funding: This study was funded by Hunan Provincial Natural Science Foundation of China (No. 2020JJ8068).

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://tcr.amegroups.com/article/view/10.21037/tcr-2026-1-0169/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.

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. Liu L, Soler J, Reckamp KL, et al. Emerging Targets in Non-Small Cell Lung Cancer. Int J Mol Sci 2024;25:10046. [Crossref] [PubMed]
2. Xue M, Ma L, Zhang P, et al. New insights into non-small cell lung cancer bone metastasis: mechanisms and therapies. Int J Biol Sci 2024;20:5747-63. [Crossref] [PubMed]
3. Desai A, Peters S. Immunotherapy-based combinations in metastatic NSCLC. Cancer Treat Rev 2023;116:102545. [Crossref] [PubMed]
4. Hendriks LEL, Remon J, Faivre-Finn C, et al. Non-small-cell lung cancer. Nat Rev Dis Primers 2024;10:71. [Crossref] [PubMed]
5. Tsukita Y, Tozuka T, Kushiro K, et al. Immunotherapy or Chemoimmunotherapy in Older Adults With Advanced Non-Small Cell Lung Cancer. JAMA Oncol 2024;10:439-47. [Crossref] [PubMed]
6. Adams HP Jr. Cancer and Cerebrovascular Disease. Curr Neurol Neurosci Rep 2019;19:73. [Crossref] [PubMed]
7. Goldstein LB. Introduction for Focused Updates in Cerebrovascular Disease. Stroke 2020;51:708-10. [Crossref] [PubMed]
8. Kulkarni P, Do D, Shrestha S, et al. Cancer Immunotherapy-An Overview. Cancer Treat Res 2025;129:1-16. [Crossref] [PubMed]
9. Michot JM, Bigenwald C, Champiat S, et al. Immune-related adverse events with immune checkpoint blockade: a comprehensive review. Eur J Cancer 2016;54:139-48. [Crossref] [PubMed]
10. Taylor X, Noristani HN, Fitzgerald GJ, et al. Amyloid-β (Aβ) immunotherapy induced microhemorrhages are linked to vascular inflammation and cerebrovascular damage in a mouse model of Alzheimer's disease. Mol Neurodegener 2024;19:77. [Crossref] [PubMed]
11. Xiong M, Jiang H, Serrano JR, et al. APOE immunotherapy reduces cerebral amyloid angiopathy and amyloid plaques while improving cerebrovascular function. Sci Transl Med 2021;13:eabd7522. [Crossref] [PubMed]
12. Che WQ, Li YJ, Tsang CK, et al. How to use the Surveillance, Epidemiology, and End Results (SEER) data: research design and methodology. Mil Med Res 2023;10:50. [Crossref] [PubMed]
13. Horita N, Yamamoto S, Mizuki Y, et al. Minimal Clinically Important Difference (MCID) of Effect Sizes other than Mean Difference. Journal of Clinical Question 2024;1:116-27.
14. Martins F, Sofiya L, Sykiotis GP, et al. Adverse effects of immune-checkpoint inhibitors: epidemiology, management and surveillance. Nat Rev Clin Oncol 2019;16:563-80. [Crossref] [PubMed]
15. Mahalingam P, Newsom-Davis T. Cancer immunotherapy and the management of side effects. Clin Med (Lond) 2023;23:56-60. [Crossref] [PubMed]
16. Brown TJ, Mamtani R, Bange EM. Immunotherapy Adverse Effects. JAMA Oncol 2021;7:1908. [Crossref] [PubMed]