Follow-up–dependent calendar-period associations with cardiac mortality after lung cancer radiotherapy: a competing-risks SEER analysis

Figures

• 

  Figure 1:. Study cohort composition and propensity score matching. The original unmatched cohort included 56,241 patients: 27,995 in the surgery cohort and 28,246 in the beam radiotherapy cohort. After 1:1 propensity score matching, 12,204 patients remained in each treatment cohort, resulting in a total matched cohort of 24,408 patients.

• 

  Figure 2:. Cumulative incidence of cardiac death by treatment cohort. (A) Original unmatched cohort. (B) Propensity score-matched cohort. Death from lung and bronchus cancer was the competing event; Gray’s test yielded p<0.001 in both panels. Numbers below each panel show those at risk. Comparisons are descriptive because treatment-selection differences and residual imbalance remained.

• 

  Figure 3:. Full-follow-up cumulative incidence of cardiac death by diagnosis period. (A) Surgery cohort. (B) Beam radiotherapy cohort. Curves compare 2009–2014 with 2015–2021; death from lung and bronchus cancer was the competing event. Gray’s test yielded p=0.003 and p=0.097, respectively. Numbers below each panel show those at risk. The comparisons are descriptive because potential follow-up differed by diagnosis period.

• 

  Figure 4:. Cumulative incidence of cardiac death by diagnosis period within a fixed 36-month window. (A) Surgery cohort. (B) Beam radiotherapy cohort. Patients diagnosed in 2009–2014 were compared with those diagnosed in 2015–2018, and follow-up for patients in both diagnosis-period groups was administratively censored at 36 months after diagnosis. Death from lung and bronchus cancer was treated as the prespecified competing event. Gray’s test yielded p=0.001 in the surgery cohort and p=0.333 in the beam radiotherapy cohort. Numbers below each panel show the numbers at risk.

• 

  Figure 5:. Adjusted Fine–Gray estimates of diagnosis-period associations with cardiac death in the beam radiotherapy cohort. Models compared 2015–2021 with 2009–2014 under full follow-up and 2015–2018 with 2009–2014 in the diagnosis-restricted full-follow-up and fixed 36-month analyses. The latter two models used the same 2009–2018 cohort; only the fixed-window model censored follow-up at 36 months. Models adjusted for age group, sex, race, summary stage, histological subtype, and primary tumor site. Squares show sHRs, horizontal lines show 95 % CIs, and the dashed line indicates an sHR of 1.0.

• 

  Figure 6:. Diagnosis-period subdistribution hazard ratios across fixed observation windows. Points show adjusted sHRs and vertical lines show 95 % CIs for the beam radiotherapy and surgery cohorts. The dashed horizontal line indicates an sHR of 1.0. Eligibility was restricted separately for each window to diagnosis years with the corresponding potential follow-up; the populations therefore differed across window lengths.