Association of age and primary treatment with risk of non-lymphoma-related death and long-term survival outcomes in adult patients with early-stage follicular lymphoma: a population-based analysis Authors: Chang Xu, Yunpeng Wu, Xin Liu, Qiu-Zi Zhong, Yong Yang, Tao Wu, Si-Ye Chen, Bo Chen, Yong-Wen Song, Hui Fang, Jing Jin, Yue-Ping Liu, Hao Jing, Yuan Tang, Ning Li, Ning-Ning Lu, Wen-Wen Zhang, Shu-Lian Wang, Jingru Zhu, Shu-Nan Qi, Ye-Xiong Li Published: 2026 ABSTRACT AbstractIn adult patients with early-stage follicular lymphoma (FL), age-related differences in the risk of non-lymphoma-related death (LRD) during the follow-up period after treatment are unclear. We aimed to investigate the interactive effects of age and primary treatment on the risk of non-LRD and survival. We retrospectively analyzed the data of 6030 adult patients with early-stage FL from the Surveillance, Epidemiology, and End Results (SEER) database (2000–2020). Primary therapy involved observation (n = 2509), chemotherapy (CT) alone (n = 1844), radiotherapy alone (RT, n = 1191) or RT and CT (RT + CT, n = 486). The groups were balanced by inverse probability of treatment weighting (IPTW). Comparisons of the cumulative incidence of non-LRD were performed using competing-risk analysis and transformed Cox regression analysis. The median patient age was 64 years (RT, 62 years; RT + CT, 60 years; CT, 63 years; observation, 66 years). Over a median follow-up period of 132 months, 830 patients (13.8%) died of FL, while non-LRD accounted for 1279 patients (21.2%). Age significantly influenced the non-LRD risk, especially in patients aged > 60 years. After IPTW adjustment, RT alone, RT + CT or CT alone was not associated with an increased risk of the cumulative incidences of non-LRD or its specific causes in comparison with observation. Age-dependent analyses showed that the adjusted hazard ratios (HRs) for RT alone, RT + CT or CT versus observation remained close to 1.0 across the age spectrum, indicating no significant difference in non-LRD risk between treatment groups. RT alone and RT + CT conferred sustained overall survival (OS), disease specific survival (DSS) and relative survival (RS) advantages over CT or observation, with RS consistently exceeding 90% over the extended follow-up period. The survival benefit of RT versus non-RT persisted across age groups, with the adjusted HRs for OS and RS remaining below 1.0 for most ages. Older age, not primary treatment, was associated with an increased risk of non-LRD. The results indicated the safety and long-term survival benefit of RT in adult patients with early-stage FL.Supplementary InformationThe online version contains supplementary material available at https://doi.org/10.1007/s00277-026-07144-4.Keywords: Follicular lymphoma, Non-lymphoma-related death, Competing-risk analysis, Radiotherapy, Chemotherapy, Observation Introduction Follicular lymphoma (FL) is the most common form of indolent non-Hodgkin lymphoma [1–3]. This disease is characterized by a predominance in older adults (median age, > 60 years), commonly presents with advanced-stage disease (80%−90%), and has a favorable prognosis (median survival, 10 years) [4]. Given the improved overall survival (OS) and relative survival (RS) in large population-based comparative studies [5–7], radiotherapy (RT) has been recommended as the mainstay treatment for early-stage FL [8, 9]. However, primary treatments for early-stage FL include RT, chemotherapy (CT), immunotherapy, combined modality therapy, or observation, and are tailored according to patient age, performance status, extent of disease, primary site, and other clinical factors [2, 8, 10–13]. The application of different first-line treatments for early-stage FL may be associated with concerns regarding their net survival benefit, long-term toxicities, and overtreatment in elder patients at a high risk of non-lymphoma-related death (LRD) or by successful salvage treatment when the disease progresses. Given the long life expectancy and high competing risks of mortality in adult patients with early-stage FL, understanding the interactive effects of age and treatments on non-LRD and survival is crucial. Although RT, especially extended-field irradiation for Hodgkin’s lymphoma, is associated with increased risks of secondary malignancies and cardiovascular toxicity [14, 15], long-term treatment-related mortality in patients with early-stage FL remains unclear. Previous studies have also demonstrated that exposure to alkylating agents and anthracycline-based CT is associated with an increased risk of specific secondary malignant neoplasms (SMNs), including breast cancer, soft-tissue sarcoma, thyroid cancer, and melanoma [16, 17]. Moreover, breast cancer survivors who received anthracycline-based CT face a significantly elevated long-term risk of cardiomyopathy, heart failure, and ischemic heart disease [18]. However, whether these treatment-related toxicities translate into increased risk of non-LRD in adult FL patients treated with involved-field or involved-site RT and/or anthracycline-based CT remains uncertain. Moreover, the influence of the age-dependent increased risk of non-LRD on the long-term survival benefit of primary treatment for early-stage FL is also unclear. In this regard, only a few of studies have analyzed non-LRD as a distinct endpoint in adult patients with early-stage indolent lymphomas [19]; the relative contributions of RT, CT, or their combination to non-LRD have not been systematically assessed. Furthermore, although FL predominantly affects older adults, most studies did not evaluate the effect of age as a continuous variable on the risk of non-LRD or assess its interaction effect with treatment strategies. The vulnerability of elderly patients, who have a higher baseline risk for comorbidities such as cardiovascular disease (CVD), and the long-term toxic effects of treatment necessitate a closer examination of how age modifies the risk of non-LRD. Understanding the intersection of treatment-related late toxicities, particularly SMN and CVD, with age-associated vulnerability could provide critical insights into the long-term survival outcomes of different treatment strategies for patients with FL. Recognizing the importance of non-LRD in long-term survivors, we conducted a large-scale population-based study using the Surveillance, Epidemiology, and End Results (SEER) database to assess whether age and primary treatments are associated with an increased risk of non-LRD, mainly SMN and CVD. We also analyzed the cumulative incidence of non-LRD across different treatment and age groups and assessed the hazard ratios (HRs) for non-LRD over the follow-up period to provide a comprehensive evaluation of the long-term impact of RT and/or CT on patient outcomes. Methods Data source and study population The study population consisted of 6030 adult patients with newly diagnosed grade I-II FL identified from the SEER-17 registries between 2000 and 2020. The inclusion criteria were as follows: (1) histologically confirmed grade I-II FL; (2) Ann Arbor stage I and II disease; (3) adult patients aged ≥ 20 years at the time of diagnosis; and (4) data available on treatment and survival outcomes. The exclusion criteria were as follows: (1) Ann Arbor stage III-IV disease or unknown stage disease; (2) unknown cause of death and unknown treatment; and (3) no follow-up data after diagnosis. A flowchart showing the participant-selection process is presented in Supplemental Fig. 1. Definitions and treatment In this study, non-LRD was categorized into deaths attributable to SMN, CVD, and other causes on the basis of the data for the variables “SEER cause-specific death classification” and “SEER other cause of death classification” in the SEER database. The cumulative incidences of LRD and non-LRD were investigated via competing risk analysis using Gray’s test. The competing event for the cumulative incidence was death without the event of interest. For instance, in non-LRD analyses, non-LRD was treated as the event of interest, with LRD as the competing event. Moreover, in LRD analyses, LRD was treated as the event of interest, with non-LRD as the competing event. Patient age was analyzed as both continuous and categorical variables (20–50, 51–60, 61–70, 71–80, and > 80 years). Race was categorized as white, black, and other/unknown. The primary sites were categorized as supradiaphragmatic lymph nodes (ICD-O-3 770, 771, 773), subdiaphragmatic lymph nodes (ICD-O-3 772, 774, 775), extranodal, and other sites. Year of diagnosis was classified as 2000–2004, 2005–2010, 2011–2014, and 2015–2020. Primary treatment was defined as RT alone (n = 1191, 19.8%), RT with CT (RT + CT; n = 486, 8.0%), CT alone (n = 1844, 30.6%), and observation (n = 2509, 41.6%). Endpoints and statistical analysis The primary endpoint was the cumulative incidence of non-LRD. The secondary endpoints were the OS and RS. LRD was defined as death from FL, while non-LRD was defined as death from treatment-related and other causes but not FL. OS was defined as the time from diagnosis to death from any cause. Disease-specific survival (DSS) was estimated using the Kaplan–Meier method, treating LRD as an event. RS was estimated using the Ederer II method, which compares the observed survival of patients with FL to the expected survival of the age-, sex-, race-, and calendar year-matched general US population. The cumulative incidence of mortality was evaluated by competing-risk analysis using Fine-Gray’s test [20]. The sub-distribution hazard ratio (SHR) was used to compare the risk of non-LRD between different treatments. The SHR provides a measure of the effect of covariates on the cumulative incidence function in the presence of competing risks, allowing for comparison of the cumulative incidence of non-LRD while accounting for the competing risk of LRD [21]. To evaluate the proportion and HR for non-LRD by treatment, splines were fitted using age as a continuous variable [22]. The spline knots were optimized on the basis of the C-index and Akaike information criterion. A linear regression model was applied to estimate the cumulative incidence trends of non-LRD, and corresponding equations with R² values were used to assess the goodness-of-fit. A multivariate Cox proportional-hazard model was used to estimate the HR for non-LRD while adjusting for potential confounders. To further evaluate age-dependent treatment effects, we plotted adjusted HR curves for treatment comparisons using patient age as a continuous variable. Penalized spline functions were applied to flexibly characterize how the association between treatment and non-LRD and its specific causes varied across the age spectrum. The propensity score-based inverse probability of treatment weighting (IPTW) approach was used to balance the distribution of covariates between treatment arms [23]. The propensity-score model included age, sex, race, marital status, Ann Arbor stage, tumor grade, B symptoms, primary site, and year of diagnosis. Missing or unrecorded categories were retained as separate categories when applicable. Because IPTW is a weighting approach rather than a matching procedure, patients were reweighted rather than excluded; therefore, all eligible patients were retained in the IPTW-adjusted analyses. Covariate balance was assessed using the standardized mean difference, with values less than 0.1 indicating acceptable balance. Differences among categorical variables were assessed using the chi-square test. Statistical analyses were performed using the tidycmprsk, survival, dplyr, mstate, cmprsk, survminer, riskRegression, cmprskcoxmsm, splines packages in R, version 4.3.3 (R Foundation for Statistical Computing, Vienna, Austria). The SEER*Stat software version 8.4.2 was used to query data from the SEER database (SEER Research Data, 17 Registries, Nov 2022 Sub [2000–2020]). Results Patient characteristics and patterns of deaths The baseline characteristics of the study population are summarized in Table 1. The median age was 64 years (range, 20–90 years), and most patients (61.8%) were older than 60 years. The male-to-female ratio was approximately 1.02:1. Most patients (89.3%) were white, and 60.1% were married. Stage I disease was observed in 65.7% of the patients. Table 1Baseline characteristics of early-stage FL patients stratified by primary treatment from the SEER database (2000–2020)CharacteristicAll patientsTreatmentsp*No. (%)RT aloneNo. (%)RT + CTNo. (%)CT aloneNo. (%)ObservationNo. (%)Total6030 (100)1191 (100)486 (100)1844 (100)2509 (100)Age, years< 0.001 Median age, year (range)64 (20–90)62 (20–90)60 (20–90)63 (20–90)66 (20–90) 20–602302 (38.2)493 (41.4)232 (47.7)728 (39.5)849 (33.8) > 603728 (61.8)698 (58.6)254 (52.3)1116 (60.5)1660 (66.2)Gender< 0.001 Male3046 (50.5)635 (53.3)274 (56.4)938 (50.9)1199 (47.8) Female2984 (49.5)556 (46.7)212 (43.6)906 (49.1)1310 (52.2) Race0.188 White5387 (89.2)1059 (89.0)437 (89.9)1669 (90.6)2222 (88.6) Black301 (5.0)60 (5.0)18 (3.7)80 (4.3)143 (5.7) Other293 (5.0)67 (5.6)30 (6.2)82 (4.4)114 (4.5) Unrecorded49 (0.8)5 (0.4)1 (0.2)13 (0.7)30 (1.2)Ann Arbor stage< 0.001 Stage I3960 (65.7)995 (83.5)313 (64.4)883 (47.9)1769 (70.5) Stage II2070 (34.3)196 (16.5)173 (35.6)961 (52.1)740 (29.5)Grade< 0.001 I2859 (47.4)512 (43.0)202 (41.6)844 (45.8)1301 (51.8) II3171 (52.6)679 (57.0)284 (58.4)1000 (54.2)1208 (48.2)B symptoms< 0.001 Present273 (4.5)36 (3.0)17 (3.5)131 (7.1)89 (3.5) Absent2039 (33.8)523 (43.9)123 (25.3)454 (24.6)939 (37.4) Unrecorded3718 (61.7)632 (53.1)346 (71.2)1259 (68.3)1481 (59.1)Primary site< 0.001 Supradiaphragmatic lymph nodes1361 (22.6)279 (23.4)124 (25.5)339 (18.4)619 (24.7) Subdiaphragmatic lymph nodes1686 (27.9)248 (20.8)122 (25.1)621 (33.7)695 (27.7) Extra nodal1931 (32.0)587 (49.3)160 (32.9)408 (22.1)776 (30.9) Other/unknown1052 (17.5)77 (6.5)80 (16.5)476 (25.8)419 (16.7)Year of diagnosis< 0.001 2000–20051769 (29.3)328 (27.5)192 (39.5)598 (32.4)651 (26.0) 2005–20101733 (28.7)316 (26.5)164 (33.7)622 (33.8)631 (25.1) 2010–20151469 (24.4)325 (27.4)82 (16.9)377 (20.4)685 (27.3) 2015–20201059 (17.6)222 (18.6)48 (9.9)247 (13.4)542 (21.6)Marital status0.681 Married3626 (60.1)787 (66.1)304 (62.6)1112 (60.3)1423 (56.7) Divorced/separated/widowed1242 (20.6)220 (18.5)89 (18.3)374 (20.3)559 (22.3) Single/unmarried/domestic partner769 (12.8)135 (11.3)75 (15.4)252 (13.7)307 (12.2) Unrecorded393 (6.5)49 (4.1)18 (3.7)106 (5.7)220 (8.8)End point status< 0.001 Alive3921 (65.0)895 (75.2)315 (64.8)1067 (57.8)1644 (65.5) Lymphoma-related death (LRD)830 (13.8)87 (7.3)64 (13.2)374 (20.3)305 (12.2) Non-LRD1279 (21.2)209 (17.5)107 (22.0)403 (21.9)560 (22.3)FL follicular lymphoma, RT ± CT radiotherapy with or without chemotherapy, CT chemotherapy, LRD lymphoma-related death, Non-LRD non-lymphoma-related death Baseline characteristics of early-stage FL patients stratified by primary treatment from the SEER database (2000–2020) FL follicular lymphoma, RT ± CT radiotherapy with or without chemotherapy, CT chemotherapy, LRD lymphoma-related death, Non-LRD non-lymphoma-related death Patients managed with observation (median age, 66 years) were older than those who underwent RT (62 years), RT + CT (60 years) or CT (63 years; Table 1, P < 0.001). Patients managed with observation included a higher proportion of those aged ≥ 60 years (66.2%), in comparison with those who underwent RT, RT + CT and CT (58.6%, 52.3% and 60.5%, respectively). The clinical use of observation rapidly increased from 37.5% in 2009 to 55.2% in 2020 (Supplemental Fig. 2 A). Among patients aged ≥ 50 years, the clinical use of observation increased from 33.3% in those aged 50 years to 56.4% in those aged 90 years (Supplemental Fig. 2B); in contrast, the clinical use of RT + CT and CT continuously decreased over time or with age beyond 50 years. Linear-dependent effect of increasing age as a continuous variable on non-LRD. The estimated logarithm HRs (solid line) with 95% CIs (shading) for the association of patient age with non-LRD in patients with early-stage FL based on the dfmacox in a smoothHR-the optimal extended Cox-type additive hazard regression model adjusted for covariables. The effect of age on the risk of non-LRD (A), SMN-related deaths (B), CVD-related deaths (C), and deaths attributable to other causes (D) was modeled using a penalized spline (P-spline) expansion, with patient age as a continuous covariable. An age cutoff of 60 years (indicated by the vertical line) was used as the reference value for calculating the HR. HR, hazard ratio; CI, confidence interval; dfmacox, degrees of freedom in multivariate additive Cox models; Non-LRD, non-lymphoma-related death; CVD, cardiovascular disease; SMN, secondary malignant neoplasm Linear-dependent effect of patient age on the cumulative incidence of non-LRD Over a median follow-up period of 132 months, 2109 patients died (non-LRD, 1279 patients [21.2%]; FL, 830 patients [13.8%]). Non-LRDs included 267 SMN-related deaths (20.9%), 217 CVD-related deaths (17.0%), and 795 deaths attributable to other causes (62.1%). The risks of LRD and non-LRD varied with extended follow-up time. Within the first five years after diagnosis, patients were at a slightly higher risk of dying from FL than non-LRD, and the risk of dying of causes other than FL increased thereafter (Fig. 1A). The cumulative incidence of LRD slightly increased from 10.1% at 5 years to 14.5% and 16.9% at 10 and 15 years, respectively. However, the cumulative incidence of non-LRD rapidly increased from 10.0% at 5 years to 19.8% at 10 years and 30.0% at 15 years. Notably, beyond 10 years, non-LRD continued to accumulate more substantially than LRD, indicating that non-LRD became increasingly prominent during extended follow-up. Similarly, the cumulative incidences of SMN-related, CVD-related, and other deaths at 5, 10, and 15 years were, respectively, 2.0%, 4.3%, and 6.3%; 1.8%, 3.4%, and 5.2%; and 6.2%, 12.2%, and 18.8% (Fig. 1B). Cumulative incidence of non-LRD over the follow-up period in all patients. (A) Cumulative incidence of LRD and non-LRD in the entire study population over time since diagnosis. (B) The linear regression analysis reveals a strong correlation between the follow-up time and the cumulative incidences of non-LRD and SMN-related deaths, CVD-related deaths, and deaths attributable to other causes (R² = 0.990–0.999). (C-F) The linear regression analysis reveals a strong correlation between the follow-up time and the cumulative incidences of non-LRD (C), SMN-related death (D), CVD-related deaths (E), and deaths attributable to other causes (F) after stratification by five age groups (20–50, 51–60, 61–70, 71–80, and > 80 years). In this analysis of the cumulative incidences of non-LRD, non-LRD was treated as the event of interest, with LRD as the competing event. Non-LRD, non-lymphoma-related death; LRD, lymphoma-related death; CVD, cardiovascular disease; SMN, secondary malignant neoplasm; R², the coefficient of determination Linear regression analysis revealed that the cumulative incidence of non-LRD and SMN-related, CVD-related, and other deaths increased over the follow-up period (Fig. 1B; non-LRD, 2.0% per year [determination coefficient, R2 = 0.999, non-LRD% = 2 × time (years) + 0.1]; SMN-related deaths, 0.6% per year [R2 = 0.989, SMN% = 0.6 × time (years)−0.4]; CVD-related deaths, 0.5% per year [R2 = 0.990, CVD% = 0.5 × time (years)-0.2]; and other deaths, 1.6% per year [R2 = 0.996, other% = 0.5 × time (years)-1.1]). After age stratification, linear regression analysis showed a strong relationship of the cumulative incidence of non-LRD (R2 = 0.957–0.995; Fig. 1C) and SMN-related (R2 = 0.868–0.976; Fig. 1D), CVD-related (R2 = 0.870–0.986; Fig. 1E), and other deaths (R2 = 0.933–0.984; Fig. 1F) with age across the five groups. This finding indicated a sustained and predictable rise in the cumulative incidence of non-LRD, depending on the patient’s age at initial diagnosis and follow-up time. Effect of age as a continuous variable on the risk of non-LRD Patients’ age as a continuous variable was entered into the Cox proportional-hazard regression using P-splines in smoothHR to allow for a nonlinear relationship between age and non-LRD. After adjusting for covariates (sex, stage, tumor grade, marital status, race, B symptoms, primary site, and year of diagnosis), the risk (lnHR) of non-LRD, including SMN-related, CVD-related, and other deaths, increased steadily with increasing age (Fig. 2A). When 60 years of age was set as the reference, the HRs of non-LRD continuously increased for older patients, being 0.85 (95% CI: 0.75–0.97) at 50 years, 1.24 (95% CI: 1.12–1.37) at 70 years, and 1.82 (95% CI: 1.56–2.12) at 80 years. Similar HR trends with increasing age were also observed for SMN-related deaths (Fig. 2B), CVD-related deaths (Fig. 2C), and other deaths (Fig. 2D). These findings indicated a positive linear relationship between increasing age and non-LRD without a cutoff point. Effect of primary treatments on the cumulative incidence of non-LRD We further evaluated the effect of primary treatments on the risk of non-LRD. The cumulative incidence of non-LRD in elder patients with observation (median age, 66 years) was higher than that in younger patients who underwent RT (median age, 62 years), RT + CT (median age, 60 years) or CT (median age, 63 years). Before IPTW adjustment, the 5-, 10-, and 15-year cumulative incidences of non-LRD were, respectively, 7.9%, 16.3%, and 25.7% for RT (SHR, 0.71; 95% CI: 0.60–0.83; P < 0.001), 7.8%, 17.0%, and 26.9% for RT + CT (SHR, 0.76; 95% CI: 0.62–0.93; P = 0.008), and 9.4%, 18.3%, and 29.5% for CT (SHR, 0.83; 95% CI: 0.73–0.94; P = 0.003), which were significantly lower than the corresponding values for observation (12.0%, 23.3%, and 33.8%; Fig. 3A). After IPTW adjustment (Fig. 3B), there was no significant difference in cumulative incidence of non-LRD between treatment groups. The 5-, 10-, and 15-year cumulative incidences of non-LRD were 10.0%, 16.8%, and 20.6% for observation, compared with 8.4%, 14.7%, and 17.9% (SHR, 0.85; 95% CI: 0.70–1.02, P = 0.084) for RT, 7.4%, 14.5%, and 19.3% (SHR, 0.91; 95% CI: 0.73–1.13, P = 0.390) for RT + CT, and 8.3%, 15.1%, and 19.8% (SHR, 0.95; 95% CI: 0.82–1.80, P = 0.421) for CT. Comparison of cumulative incidence of non-LRD between treatment arms. The cumulative incidences of non-LRD, SMN-related deaths, CVD-related deaths, and deaths attributable to other causes between treatment arms were compared before (A, C, E, G) or after (B, D, F, H) balancing with IPTW and adjusting for age, sex, race, diagnosis year, primary site, Ann Arbor stage, and marital status based on the Cox multivariable regression. In this analysis of the cumulative incidences of non-LRD, non-LRD was treated as the event of interest, with LRD as the competing event. Non-LRD, non-lymphoma-related death; SMN, secondary malignant neoplasm; CVD, cardiovascular disease; IPTW, inverse probability of treatment weighting Additionally, RT, RT + CT and CT versus observation did not show a significant difference in the cumulative incidences of SMN-related deaths. The unadjusted 5-, 10-, and 15-year cumulative incidences of SMN were, respectively, 2.3%, 5.0%, and 7.1% with observation (Fig. 3C), compared with 1.3%, 3.0%, and 4.9% with RT (SHR, 0.64; 95% CI: 0.45–0.93; P = 0.018), 2.1%, 4.7%, and 6.5% with RT + CT (SHR, 0.98; 95% CI: 0.64–1.48; P = 0.900), and 2.1%, 4.1%, and 6.2% with CT (SHR, 0.87; 95% CI: 0.66–1.15; P = 0.340). The adjusted 5-, 10-, and 15-year cumulative incidences of SMN-related deaths were, respectively, 2.1%, 3.8%, and 4.5% with observation (Fig. 3D), 1.3%, 2.7%, and 3.4% with RT (SHR, 0.71; 95% CI: 0.46–1.09; P = 0.120), 1.7%, 3.5%, and 4.4% with RT + CT (SHR, 0.94; 95% CI: 0.61–1.46; P = 0.788), and 1.6%, 3.3%, and 4.1% with CT (SHR, 0.88; 95% CI: 0.66–1.17; P = 0.389). Similarly, RT, RT + CT and CT versus observation showed no significant difference in the cumulative incidence of CVD-related deaths. The unadjusted 5-, 10-, and 15-year cumulative incidences of CVD-related deaths were, respectively, 2.0%, 3.9%, and 5.9% with observation (Fig. 3E), 1.6%, 2.8%, and 4.7% with RT (SHR, 0.88; 95% CI: 0.61–1.27; P = 0.490), 1.7%, 3.0%, and 4.7% with RT + CT (SHR, 0.92; 95% CI: 0.57–1.48; P = 0.720), and 1.5%, 3.1%, and 4.7% with CT (SHR, 0.82; 95% CI: 0.59–1.12; P = 0.210). The adjusted 5-, 10-, and 15-year cumulative incidences of CVD-related deaths were, respectively, 1.8%, 2.8%, and 3.5% with observation (Fig. 3F), 2.2%, 3.0%, and 3.7% with RT (SHR, 1.13; 95% CI: 0.73–1.74; P = 0.575), 1.8%, 2.7%, and 3.6% with RT + CT (SHR, 1.12; 95% CI: 0.68–1.87; P = 0.638), and 1.4%, 2.4%, and 3.2% with CT (SHR, 0.90; 95% CI: 0.64–1.26; P = 0.549). The unadjusted 5-, 10-, and 15- cumulative incidences of other non-LRD were, respectively, 7.6%, 14.4%, and 20.8% with observation (Fig. 3G), 5.1%, 10.4%, and 16.1% with RT (SHR, 0.72; 95% CI: 0.59–0.88; P = 0.001), 3.9%, 9.3%, and 15.7% with RT + CT (SHR, 0.70; 95% CI: 0.53–0.91; P = 0.008), and 5.8%, 11.1%, and 18.5% with CT (SHR, 0.85; 95% CI: 0.73–1.00; P = 0.051). The adjusted 5-, 10- and 15-year cumulative incidences for other deaths were, respectively, 6.2%, 10.2%, and 12.6% with observation (Fig. 3H), 4.9%, 9.0%, and 10.7% with RT (SHR, 0.84; 95% CI: 0.67–1.07; P = 0.156), 3.9%, 8.3%, and 11.2% with RT + CT (SHR, 0.86; 95% CI: 0.64–1.15; P = 0.303), and 5.3%, 9.4%, and 12.6% with CT (SHR, 0.99; 95% CI: 0.84–1.19; P = 0.983). Association of age as a continuous variable and treatments with the risk of non-LRD Because the cumulative incidence of non-LRD varied according to age and treatment (Figs. 1, 2 and 3), the influence of age as a continuous variable on the risk of non-LRD between treatment groups was investigated (Fig. 4). Using IPTW-adjusted Cox models with age as a continuous variable, RT alone, RT + CT or CT alone versus observation was not associated with an increased risk of non-LRD and SMN-related, CVD-related, and other deaths, with adjusted HRs close to 1.0 across the entire age spectrum and all P values > 0.05 (range, 0.081–0.95). The adjusted HRs for RT alone versus observation were 0.49 (95% CI: 0.17–1.41, P = 0.188; Fig. 4A) for all non-LRDs, 0.13 (95% CI: 0.02–1.21, P = 0.143; Fig. 4B) for SMN-related deaths, 0.16 (95% CI: 0.01–2.72, P = 0.207; Fig. 4C) for CVD-related deaths, and 0.96 (95% CI: 0.23–3.96, P = 0.950; Fig. 4D) for other deaths. The adjusted HRs for RT + CT versus observation were 1.05 (95% CI: 0.32–3.48, P = 0.932; Fig. 4E) for all non-LRDs, 0.14 (95% CI: 0.01–1.42, P = 0.097; Fig. 4F) for SMN-related deaths, 0.30 (95% CI: 0.01–6.64, P = 0.447; Fig. 4G) for CVD-related deaths, and 4.12 (95% CI: 0.83–10.41, P = 0.081; Fig. 4H) for other deaths. Similarly, CT alone versus observation showed no increased risk of non-LRD, with adjusted HRs of 1.09 (95% CI: 0.45–2.62, P = 0.854; Fig. 4I) for all non-LRDs, 0.32 (95% CI: 0.06–1.62, P = 0.169; Fig. 4J) for SMN-related deaths, 0.49 (95% CI: 0.05–5.13, P = 0.552; Fig. 4K) for CVD-related deaths, and 2.43 (95% CI: 0.75–7.91, P = 0.141; Fig. 4L) for other-related deaths. Age-dependent adjusted HRs for non-LRD between treatment groups. The effect of patient age as a continuous covariable on the risk of non-LRD (A), SMN (B), CVD (C), and other causes (D) between treatment groups was modeled using a penalized spline (P-spline) expansion. Adjusted HRs for non-LRD, SMN, CVD, and other causes were presented by RT alone versus observation (A-D); RT + CT versus observation (E-H); or by CT versus observation (I-L). The solid line represents the HR estimate, and the shadow represents the 95% CI. HR, hazard ratio; CI, confidence interval; non-LRD, non-lymphoma-related death; SMN, secondary malignant neoplasm; CVD, cardiovascular disease Long-term OS, DSS and RS benefits of RT versus CT or observation Since the incidence of both LRD and non-LRD increased with age and follow-up time, we evaluated the effect of the primary treatments on long-term OS and RS. The median follow-up time was 132 months for the entire cohort (RT, 125 months; RT + CT, 161 months; CT, 147 months; and observation, 114 months). Although the OS rates continuously decreased over follow-up time, RT and RT + CT versus CT or observation were associated with improved OS. The unadjusted 5-, 10-, and 15-year OS rates were, respectively, 87.3%, 76.0%, and 65.4% for RT, 84.2%, 70.6%, and 61.2% for RT + CT (HR, 1.27; 95% CI: 1.05–1.53; P = 0.013), 74.4%, 61.2%, and 46.9% for CT (HR, 1.82; 95% CI: 1.59–2.08; P < 0.001), and 79.5%, 62.9%, and 50.0% for observation (HR, 1.62; 95% CI: 1.42–1.85; P < 0.001; Fig. 5A). Patients receiving RT alone were more likely to have stage I disease, whereas patients with observation were more likely to have grade I disease (Table 1). After IPTW adjustment, prognostic factors and marital status were well balanced between treatment arms (Supplemental Fig. 1B). After IPTW adjustment, RT alone maintained a survival advantage over CT or observation. The IPTW-adjusted 5-, 10-, and 15-year OS rates were, respectively, 84.1%, 71.8%, and 62.6% for RT, significantly higher than the corresponding values for RT + CT (82.6%, 67.5%, and 54.0; HR, 1.24; 95% CI: 1.00–1.54; P = 0.044), CT (75.9%, 61.7%, and 46.8%; HR, 1.55; 95% CI: 1.32–1.81; P < 0.001), and observation (80.8%, 65.5%, and 53.4%; HR, 1.28; 95% CI: 1.09–1.49; P = 0.002). Comparison of OS, DSS and RS between treatment arms in early-stage FL. (A) Kaplan–Meier survival analysis of OS among patients who underwent RT alone, RT + CT, CT alone, or observation. (B) RS differences between treatment arms. (C) Unadjusted Kaplan–Meier curves of DSS before IPTW. (D) IPTW-adjusted DSS curves after balancing baseline covariates across treatment groups. Adjusted HRs for OS (E) and RS (F) are presented by RT (with or without CT) versus non-RT (CT and observation), with age as the continuous variable. The number-at-risk table shown in panel D reflects weighted pseudo-population counts and does not represent the actual number of patients at risk. The solid line in panel E and F represents the HR estimate, and the shadow with dashed lines represents the 95% CI. The simulated plot of the interaction between age and RS efficacy of RT, depicting the HRs for OS and RS among patients treated with RT vs. non-RT. This shows the results of the transformed Cox regression testing for the interaction of age with the OS and RS benefits of RT versus non-RT, after balancing for IPTW and controlling for age, sex, race, diagnosis year, primary site, Ann Arbor stage, and marital status. OS, overall survival; DSS, disease specific survival; RS, relative survival; FL, follicular lymphoma; HR, hazard ratio; RT: radiotherapy; non-RT, non-radiotherapy; CT, chemotherapy; IPTW: inverse probability of treatment weighting We further evaluated the net survival benefit of RT relative to the age-, sex-, race, and calendar year-matched general US population. Patients who received RT alone maintained the highest RS rate with an extended follow-up time over 10 years. However, patients who received CT alone had significantly higher risks of excess mortality, and exhibited a continuous decrease in RS over time (Fig. 5B). The 5-year, 10-year, and 15-year RS rates were, respectively, 99.9%, 101.0%, and 104.0% (approximately 100%) for RT, 93.2%, 91.1%, and 89.6% for RT + CT (HR, 1.27; 95% CI: 1.05–1.53; P = 0.013), 85.2%, 82.5%, and 76.4% for CT (HR, 1.82; 95% CI: 1.59–2.08; P < 0.001), and 93.6%, 89.8%, and 87.4% for observation (HR, 1.62; 95% CI: 1.42–1.85; P < 0.001). This finding indicated that RT based approaches achieved very favorable net survival after correcting the risk of non-LRD with increasing age. In addition, DSS was analyzed by using LRD as the event of interest. Before IPTW adjustment, RT alone versus RT + CT, CT or observation was associated with the higher DSS over long-term follow-up. The unadjusted 5-, 10-, and 15-year DSS rates were 95.1%, 91.7%, and 90.3% for RT alone, compared with 91.7%, 86.6%, and 83.7% for RT + CT (HR, 1.68; 95% CI: 1.22–2.32; P = 0.002), 83.2%, 78.2%, and 73.7% for CT (HR, 2.96; 95% CI: 2.35–3.74; P < 0.001), and 91.0%, 84.6%, and 81.1% for observation (HR, 1.87; 95% CI: 1.47–2.37; P < 0.001; Fig. 5C). After IPTW adjustment, the adjusted 5-, 10-, and 15-year DSS rates were 93.2%, 89.1%, and 87.8% for RT alone, compared with 90.5%, 84.7%, and 82.0% for RT + CT (HR, 1.43; 95% CI: 0.98–2.08; P = 0.065), 84.4%, 79.3%, and 74.8% for CT (HR, 2.10; 95% CI: 1.59–2.78; P < 0.001), and 91.4%, 85.3%, and 82.1% for observation (HR, 1.34; 95% CI: 1.01–1.77; P = 0.043; Fig. 5D). Based on the treatment-stratified DSS curves, no clear lymphoma-specific survival plateau was observed. Although the decline in DSS appeared to attenuate after 10 years, DSS continued to decrease up to 15 years, indicating persistent but slower accumulation of LRD. This pattern was consistent with the cumulative incidence of LRD in the overall cohort, which increased from 10.1% at 5 years to 14.5% at 10 years and 16.9% at 15 years. Interactive effect of age and primary treatment on long-term survival benefit The influence of age as a continuous variable on OS and RS was examined. After adjusting for all covariates and treatments using IPTW and multivariable analyses, RT (with or without CT) versus non-RT (CT plus observation) was associated with improved OS (HR, 0.40; 95% CI: 0.21–0.78; P = 0.006; Fig. 5E) and RS (HR, 0.39; 95% CI: 0.20–0.76; P = 0.005; Fig. 5F). However, no significant interaction between age and RT was identified for OS (HR, 1.01; 95% CI: 1.00–1.02; P = 0.061) and RS (HR, 1.01; 95% CI, 0.99–1.02; P = 0.058), indicating similar survival benefits for RT across all-ages patients. Discussion To our knowledge, this is the largest population-based study to quantify the age-dependent risk of non-LRD after initial therapy for early-stage FL, with a long-term follow-up period of more than 10 years. Additionally, this is the first study to evaluate the potential influence of non-LRD on the survival benefit of RT in relation to the patient’s age and follow-up period. Increasing age was associated with a significantly increased risk of non-LRD in a linear regression pattern, independent of treatment strategy. The cumulative incidence of non-LRD increased with age at initial diagnosis over the follow-up period. Importantly, RT ± CT or CT was not associated with an increased risk of non-LRD in comparison with observation, underscoring the favorable long-term safety profile for these treatments. The OS and RS benefits of RT vs. non-RT for early-stage FL persisted across all age groups, indicating that optimal treatment should depend on disease-related risk factors rather than chronological age. Taken together, our results underscore the importance of considering age-related competing risk of mortality in survivorship counseling, and support the continued use of first-line RT with or without CT in appropriately selected early-stage FL patients. In this population-based study with long-term follow-up data, age emerged as the most dominant predictor of non-LRD in patients with early-stage FL. A consistent and progressive increase in the cumulative incidence of non-LRD was observed with advancing age, particularly beyond 60 years, consistent with prior reports [7, 24]. Age-dependent cumulative incidence and log-HR curves further revealed a linear association between age and non-LRD risk. Similarly, cause-specific analysis also showed that the risk of CVD-related, SMN-related, and other non-cancer deaths increased steadily with age. Importantly, RT alone, RT + CT or CT alone was not associated with an increased risk of non-LRD, including SMN- and CVD-related deaths, in comparison with observation even over a prolonged follow-up period. Moreover, this result remained consistent across multiple analytic approaches, including cumulative incidence curves, multivariable Cox regression, and age-adjusted interaction modeling. Although the risks of late treatment-induced morbidity and mortality associated with CT or RT for lymphomas have been established, particularly in childhood cancer survivors [24–26], we observed no significant excess risk of non-LRD or CVD- or SMN-related deaths in adult patients undergoing RT or CT for early-stage FL in this study. The lack of an increased risk of non-LRD with RT could be partially due to the application of low-dose involved-site RT techniques in indolent lymphomas [27–30], and the potential influence of improved disease control in reducing treatment-related complications, relapse-associated mortality, or systemic inflammatory burden, thereby mitigating long-term non-cancer risks [31]. These findings may challenge the long-standing assumption that definitive treatment inevitably leads to late adverse outcomes, particularly in adult patients with early-stage FL [25, 32, 33]. Instead, they provide reassuring evidence that RT ± CT and CT, when appropriately applied, are effective and safe in the long run. This may help address concerns regarding the historical association between curative-intent therapy and long-term toxicity [15, 18, 26], especially in light of recent trends toward treatment de-intensification in early-stage FL [27–29, 34]. Moreover, the age-stratified modeling demonstrated that the absolute cumulative incidence of non-LRD continued to increase with longer follow-up in older patients, highlighting the importance of age-based survivorship stratification. These findings suggest that age-driven vulnerability, rather than late treatment-related toxicity, may be the predominant contributor to non-LRD in this population. However, the results should be interpreted with caution, as the cumulative incidences of non-LRD may be influenced by unmeasured baseline factors, such as comorbidities and functional status, which are not available in the SEER database. Two features of contemporary practice provide a plausible mechanistic basis for these findings. The first feature is field design: Early-stage indolent lymphoma is routinely treated with involved-site RT (ISRT) instead of extended-field RT [9]; because mediastinal lymph nodes are rarely involved in FL, most patients do not receive mediastinal irradiation. This factor limits incidental cardiac and pulmonary exposure in comparison with that historically observed during extended-field RT, reducing the long-term risk of CVD. The second factor is dose de-escalation: Modern FL protocols employ lower curative doses or de-intensified schedules. Dose-randomization trials and recent reports (including a report by our group) have shown that dose reduction preserves disease control while lowering acute toxicity [27–29, 34]. The large population covered by the registry of the SEER database along with the use of follow-up data spanning more than 10 years enabled long-term detailed evaluation of the patterns of non-LRD risk and survival benefits after initial treatments. We observed a higher risk of non-LRD with increasing age in patients with early-stage FL. However, after controlling for covariables of background mortality and treatment, older age was not an unfavorable prognostic factor for RS. This study specifically examined how age or background mortality influence the treatment benefits of RT. RT based treatments was associated with improved OS, DSS and RS in comparison with CT and observation. Notably, RT alone and RT + CT achieved very favorable RS regardless of age. Patients treated with RT alone exhibited the lowest excess mortality over time, and their 10-year and 15-year RS rates of approximately 100% were comparable to those in the matched general US population. In contrast, patients in the CT or observation groups experienced continuous declines in RS with extended follow-up time, clearly indicating a higher risk of LRD or disease-related excess mortality. This study had several notable strengths. First, given the median survival time of > 10 years in FL, large sample sizes and long follow-up periods are essential to demonstrate statistically significant survival differences between treatments. The use of a large population-based dataset in this study yielded robust comparisons and generalizable conclusions. The extended median follow-up of more than 10 years allowed for thorough dynamic evaluation of late treatment-related deaths, which is particularly relevant in cases of early-stage FL treated with RT and anthracycline-based CT, since SMN and CVD often emerge 10–15 years post-treatment in such cases. On the other hand, we had previously demonstrated the time-dependent effect of RT versus non-RT on OS and RS benefits, indicating the need for at least 10 years of follow-up to define greater treatment benefits. Importantly, more than 250 patients remained at risk at 15 years, ensuring sufficient statistical power to assess long-term non-LRDs and survival outcomes. Moreover, the inclusion of age-stratified competing-risk analyses provided critical insights into age-specific patterns of non-LRDs, highlighting age as a key modifier of late treatment effects. This study had several limitations. Data for several treatment-related variables, such as RT dose, target volume, chemotherapy type and cycles, and use of immunotherapy, were unavailable in the SEER database, precluding evaluation of dose–response effects and modern strategies such as ISRT, PET staging, and anti-CD20–based therapies [35, 36] Furthermore, as data on disease progression and relapse are not available in the SEER database, progression-free survival and cumulative incidence of lymphoma progression and relapse could not be analyzed in this study. The non-LRD estimates are restricted to the period before lymphoma progression, non-LRD events occurring after progression and subsequent therapies are not assessed, and the counterintuitive finding of lower non-LRD in RT/CT versus observation is likely influenced by residual confounding and by the handling of competing events. Additionally, the inability to distinguish involved-field RT from ISRT limited the assessment of contemporary volume-sparing techniques [30]. As with all retrospective registry analyses, the possibility of residual confounding and treatment selection bias could not be ruled out despite IPTW adjustment and multivariable modeling. Data for many unknown confounders, such as number of chemotherapy cycle and second line treatment after observation, and other unknown confounders that could potentially influence treatment bias and outcomes in comparisons between treatments were not collected; therefore, these confounders were not accounted for in the analysis. Moreover, the SEER database does not provide patients’ socioeconomic status, comorbidities and performance status, insurance coverage, rural and urban residence, and other confounding factors that may influence access to radiotherapy and treatment selection in real-world practice. Patients with high tumor burden, such as bulky, stage II, or non-contiguous disease, tended to receive chemotherapy, immunotherapy or combined modality therapy; while patients who are older or in poor performance status were more likely to receive radiotherapy alone or observation. The absence of these variables limited our ability to adjust for potential disparities in treatment availability related to healthcare access. Lastly, the cohort included only adult patients aged ≥ 20 years, limiting the generalizability of these findings for pediatric or adolescent patients. In addition, modern treatment approaches may modify the interpretation of our findings. PET-based staging could reduce occult stage migration and better define truly localized disease, while contemporary low dose ISRT may lower normal-tissue exposure and late SMN- or CVD-related mortality. Anti-CD20–based therapy and modern salvage treatments may improve lymphoma control and alter subsequent treatment exposure, thereby influencing DSS, OS, RS, and cause-specific mortality. Because these details are unavailable in SEER, our findings should be interpreted with caution when extrapolated to fully contemporary practice. In conclusion, non-LRDs increased with advancing age and accumulated over the follow-up period. RT alone, RT + CT or CT was not associated with excess long-term non-LRDs, including SMN- and CVD-related deaths. The survival benefit of RT versus non-RT persisted regardless of age, indicating the importance of definitive therapy with survivorship care age-tailored to competing mortality. Supplementary Information Below is the link to the electronic supplementary material. Supplementary Material 1 (DOCX 1.98 MB) Abbreviations Akaike information criterion Confidence interval Chemotherapy Cardiovascular disease Follicular lymphoma Hazard ratio Inverse probability of treatment weighting Involved-site radiotherapy Lymphoma-related death Overall survival Relative survival Radiotherapy Surveillance, Epidemiology, and End Results Sub-distribution hazard ratio Secondary malignant neoplasm Author contributions C.X., Y.W., and X.L. performed data extraction, statistical analysis, and drafted the manuscript. Q.-Z.Z., Y.Y., T.W., S.-Y.C., B.C., Y.-W.S., and H.F. contributed to data collection and validation. J.J., Y.-P.L., H.J., Y.T., N.L., N.-N.L., W.-W.Z., J.Z., and S.-N.Q. assisted in data interpretation and manuscript revision. S.-L.W. and Y.-X.L. conceived and supervised the study and critically revised the manuscript. All authors reviewed and approved the final version of the manuscript. Funding This work was supported by grants from the Chinese Academy of Medical Sciences Innovation Fund for Medical Sciences (2023-I2M-1-001), and the Clinical Research Fund for the “Rotational Intensity-Modulated Radiation Therapy Clinical Application Study” (Project No. A01113, Clinical Discipline Key Program A01). Data availability The datasets generated and/or analysed during the current study are available in the SEER repository, https://seer.cancer.gov/. Declarations Ethics approval and consent to participate This study is based on data from the Surveillance, Epidemiology, and End Results (SEER) public database. Since the SEER database contains de-identified patient information, no ethical approval or informed consent was required for this study. Consent for publication This study used de-identified data from the Surveillance, Epidemiology, and End Results (SEER) public database. As the SEER database contains anonymized patient information, consent for publication was not required. Clinical trail number Not applicable. Competing interests The authors declare no competing interests. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. References Associated Data Supplementary Materials Supplementary Material 1 (DOCX 1.98 MB) Data Availability Statement The datasets generated and/or analysed during the current study are available in the SEER repository, https://seer.cancer.gov/.