Risk stratification using a nomogram model for postoperative cancer-related fatigue in elderly survivors following early-stage non-small cell lung cancer resection.

A total of 649 eligible participants with complete and valid responses were enrolled in this study. The median age was 74 years, and the cohort consisted of 427 (65.8%) males and 222 (34.2%) females. The prevalence of mild, moderate, and severe CRF was 19.6% (127/649), 47.3% (307/649), and 32.7% (212/649), respectively (Fig.  2 ). Univariate analysis revealed significant differences between the severe (212 cases) and non-severe CRF (437 cases) groups with respect to variables such as age ( P  = 0.020), ADL ( P  < 0.001), FoP ( P  < 0.001), resilience ( P  = 0.011), social support ( P  < 0.001) and coping styles ( P  = 0.015). Detailed comparisons are provided in Table  1 . Among these participants, 455 (70%) were randomly assigned to the training set, while 194 (30%) were assigned to the validation set. The baseline characteristics of participants in the training and validation sets are presented in Supplementary Table S3. No statistically significant differences were observed in any parameters between the training and validation sets (all p  > 0.05, Supplementary Table S3). Fig. 2 Violin plots depicting postoperative CRF scores among 649 participants stratified by CRF severity Violin plots depicting postoperative CRF scores among 649 participants stratified by CRF severity Table 1 Comparison of baseline characteristics between severe and non-severe CRF participants ( n  = 649) Variables a Total ( n  = 649) Non-severe CRF ( n  = 437) Severe CRF ( n  = 212) Statistics b P -value Age (years) 74.0 (72.0, 77.0) 75.0 (72.0, 78.0) 74.0 (72.0, 77.0) −2.319 0.020* BMI (kg/m 2 )  <24 447 (68.9) 297 (68.0) 150 (70.8) 0.519 0.471  ≥24 202 (31.1) 140 (32.0) 62 (29.2) Gender  Male 427 (65.8) 304 (69.6) 123 (58.0) 8.456 0.004**  Female 222 (34.2) 133 (30.4) 89 (42.0) Marital status  Single/divorced/windowed 237 (36.5) 172 (39.4) 65 (30.7) 4.660 0.031*  Married 412 (63.5) 265 (60.6) 147 (69.3) Education level  Below High school 370 (57.0) 244 (55.8) 126 (59.4) 0.754 0.385  High school or above 279 (43.0) 193 (44.2) 86 (40.6) Caregivers  Yes 80 (12.3) 27 (6.2) 53 (25.0) 46.792 < 0.001***  No 569 (87.7) 410 (93.8) 159 (75.0) Residence  Rural 277 (42.7) 186 (42.6) 91 (42.9) 0.008 0.930  Urban 372 (57.3) 251 (57.4) 121 (57.1) Smoking status  Never 325 (50.1) 222 (50.8) 103 (48.6) 2.328 0.312  Previous 308 (47.5) 207 (47.4) 101 (47.6)  Current 16 (2.5) 8 (1.8) 8 (3.8) Drinking status  Never 245 (37.8) 164 (37.5) 81 (38.2) 2.517 0.284  Previous 289 (44.5) 202 (46.2) 87 (41.0)  Current 115 (17.7) 71 (16.2) 44 (20.8) Monthly income (Yuan)  <3,000 222 (34.2) 144 (33.0) 78 (36.8) 5.345 0.069  3,000–5,000 327 (50.4) 233 (53.3) 94 (44.3)  ≥5,000 100 (15.4) 60 (13.7) 40 (18.9)  Medical payment method  Self-pay 243 (37.4) 156 (35.7) 87 (41.0) 1.738 0.187  Medical Insurance 406 (62.6) 281 (64.3) 125 (59.0) Employment status  Work 123 (19.0) 85 (19.5) 38 (17.9) 0.216 0.642  Unemployment/retired 526 (81.0) 352 (80.5) 174 (82.1) Comorbidities  Hypertension 254 (39.1) 167 (38.2) 87 (41.0) 0.477 0.490  Diabetes 276 (42.5) 184 (42.1) 92 (43.4) 0.097 0.755  Cardiovascular diseases 525 (80.9) 347 (79.4) 178 (84.0) 1.918 0.166  COPD 198 (30.5) 122 (27.9) 76 (35.8) 4.236 0.040*  Osteoarthrosis 242 (37.3) 164 (37.5) 78 (36.8) 0.033 0.856  Asthma 359 (55.3) 252 (57.7) 107 (50.5) 2.989 0.084 TNM stage  I 451 (69.5) 313 (71.6) 138 (65.1) 2.871 0.090  II 198 (30.5) 124 (28.4) 74 (34.9) Surgical approach  Lobectomy 471 (72.6) 324 (74.1) 147 (69.3) 3.589 0.166  Wedge resection 131 (20.2) 87 (19.9) 44 (20.8)  Segmentectomy 47 (7.2) 26 (5.9) 21 (9.9) Tumor type  Squamous carcinoma 33 (5.1) 23 (5.3) 10 (4.7) 0.088 0.766  Adenocarcinoma 616 (94.9) 414 (94.7) 202 (95.3) Lesion location  Left 295 (45.5) 186 (42.6) 109 (51.4) 4.512 0.034*  Right 354 (54.5) 251 (57.4) 103 (48.6) Tumor size (cm)  <3 445 (68.6) 299 (68.4) 146 (68.9) 0.013 0.908  ≥3 204 (31.4) 138 (31.6) 66 (31.1)  Postoperative ICU admission 47 (7.2) 24 (5.5) 23 (10.8) 6.098 0.014*  Postoperative pulmonary complications c 82 (12.6) 49 (11.2) 33 (15.6) 2.451 0.117  Adjuvant therapy 237 (36.5) 148 (33.9) 89 (42.0) 4.054 0.044  Time since surgery (years) 2.0 (1.0, 3.0) 2.0 (1.0, 3.0) 2.0 (1.0, 3.0) −2.096 0.036* Baseline PFTs  FVC% predicted 81.0 (63.0, 98.0) 79.0 (58.3, 95.8) 83.0 (64.0, 98.5) −1.337 0.181  FEV1% predicted 74.0 (55.0, 93.0) 73.0 (54.0, 91.0) 76.0 (55.0, 93.0) −1.055 0.292  FEV1/FVC (%) 69.0 (62.0, 79.0) 72.0 (63.0, 79.0) 69.0 (61.5, 77.0) −2.022 0.043* Nutrition status  Malnutrition 118 (18.2) 36 (8.2) 82 (38.7) 109.722 < 0.001***  Malnutrition risk 364 (56.1) 252 (57.7) 112 (52.8)  Normal 167 (25.7) 149 (34.1) 18 (8.5) ADL  Impaired 449 (69.2) 264 (60.4) 185 (87.3) 48.278 < 0.001***  Normal 200 (30.8) 173 (39.6) 27 (12.7) mMRC (score)  <2 266 (41.0) 187 (42.8) 79 (37.3) 1.803 0.179  ≥2 383 (59.0) 250 (57.2) 133 (62.7)  Poor sleep quality 521 (80.3) 350 (80.1) 171 (80.7) 0.029 0.864  Anxiety 335 (51.6) 225 (51.5) 110 (51.9) 0.009 0.924  Depression 369 (56.9) 248 (56.8) 121 (57.1) 0.006 0.938 FoP  Low 268 (41.3) 202 (46.2) 66 (31.1) 13.412 < 0.001***  High 381 (58.7) 235 (53.8) 146 (68.9) Resilience  Low 486 (74.9) 314 (71.9) 172 (81.1) 6.534 0.011*  High 163 (25.1) 123 (28.1) 40 (18.9) Social support  Low 212 (32.7) 90 (20.6) 122 (57.5) 92.926 < 0.001***  Medium 208 (32.0) 155 (35.5) 53 (25.0)  High 229 (35.3) 192 (43.9) 37 (17.5) Coping styles  Negative coping 323 (49.8) 203 (46.5) 120 (56.6) 5.883 0.015*  Positive coping 326 (50.2) 234 (53.5) 92 (43.4) Abbreviations ADL activities of daily living, BMI body mass index COPD chronic obstructive pulmonary disease CRF cancer-related fatigue, FEV1 forced expiratory volume in 1 s, FVC forced vital capacity, FEV1/FVC FEV1 to FVC ratio, FoP fear of disease progression, ICU Intensive Care Unit, mMRC modified Medical Research Council, PFTs pulmonary function tests, TNM Tumor-Node-Metastasis * P  < 0.05; ** P  < 0.01; *** P  < 0.001 a Continuous variables are expressed as median (interquartile range) and categorical variables are presented as frequencies (%) b Mann-Whitney U test, Chi-square test, or Fisher’s exact test were conducted for all continuous and categorical variables c Postoperative pulmonary complications include prolonged air leak, pulmonary embolism, atelectasis, pneumonia, atrial arrhythmia, empyema, chylothorax, acute respiratory distress syndrome, bronchopleural fistula, or respiratory failure. Comparison of baseline characteristics between severe and non-severe CRF participants ( n  = 649) Abbreviations ADL activities of daily living, BMI body mass index COPD chronic obstructive pulmonary disease CRF cancer-related fatigue, FEV1 forced expiratory volume in 1 s, FVC forced vital capacity, FEV1/FVC FEV1 to FVC ratio, FoP fear of disease progression, ICU Intensive Care Unit, mMRC modified Medical Research Council, PFTs pulmonary function tests, TNM Tumor-Node-Metastasis * P  < 0.05; ** P  < 0.01; *** P  < 0.001 a Continuous variables are expressed as median (interquartile range) and categorical variables are presented as frequencies (%) b Mann-Whitney U test, Chi-square test, or Fisher’s exact test were conducted for all continuous and categorical variables c Postoperative pulmonary complications include prolonged air leak, pulmonary embolism, atelectasis, pneumonia, atrial arrhythmia, empyema, chylothorax, acute respiratory distress syndrome, bronchopleural fistula, or respiratory failure. LASSO regression analysis was performed to identify potential variables influencing severe CRF. As shown in Fig.  3 A, as the λ value increased, the absolute values of the coefficients for each factor decreased. In this study, five variables with non-zero coefficients were selected at the λ.1 standard error (λ.1se), which was 0.0581 (Fig.  3 B). These variables included FoP, caregivers, social support, ADL, and nutrition status (Fig.  3 C). The multicollinearity analysis showed that the VIF values for all variables ranged from 1.006 to 1.083, which were below the threshold of 5, suggesting no substantial multicollinearity among these risk factors. This result confirmed the reliability of the regression analysis, as detailed in Supplementary Table S4. Subsequently, these candidate variables were further validated through multivariable logistic regression (all P  < 0.05; Fig.  3 D). Fig. 3 Feature selection for postoperative severe CRF. A  Profiles of LASSO coefficients. B  LASSO regression analysis utilizing ten-fold cross-validation. C  Final variables selected by LASSO regression with standardized coefficients. D  Multivariable logistic regression of the five retained risk factors Feature selection for postoperative severe CRF. A  Profiles of LASSO coefficients. B  LASSO regression analysis utilizing ten-fold cross-validation. C  Final variables selected by LASSO regression with standardized coefficients. D  Multivariable logistic regression of the five retained risk factors A nomogram was developed to stratify the risk of severe CRF in elderly patients after early-stage NSCLC resection. As shown in Fig.  4 A, each variable was assigned a specific “Point” value on the top axis. The “Total points” were calculated by summing the points for all variables. These “Total points” were converted into the predicted probability of severe CRF through the logistic transformation. A higher total point value indicated a higher probability of developing severe CRF. For example, a 74-year-old patient with caregivers, impaired ADL, low social support, risk of malnutrition, and no FoP accumulated “Total points” of 323, corresponding to a predicted probability of 49.1%. Fig. 4 Nomogram for predicting postoperative severe CRF. A  Nomogram incorporating five variables. B  An example illustrating the application of the nomogram model to predict the likelihood of severe CRF Nomogram for predicting postoperative severe CRF. A  Nomogram incorporating five variables. B  An example illustrating the application of the nomogram model to predict the likelihood of severe CRF The discriminative ability of the nomogram was evaluated using the AUC. For the training set, the AUC was 0.864 (95% CI: 0.828, 0.900). The optimal cutoff value was 0.247, which resulted in a specificity of 0.734 and a sensitivity of 0.871 (Fig.  5 A). These results underscored the nomogram’s strong discriminative ability. In the validation set, the AUC was slightly lower at 0.845 (95% CI: 0.786, 0.903), showcasing the model’s consistent performance across different datasets. The best cutoff value was identified as 0.404, with a specificity of 0.884 and a sensitivity of 0.662 (Fig.  5 A). Supplementary Table S5 provided a comprehensive overview of the performance metrics across various probability thresholds. At a threshold of 0.825, the PPV reached 93.8%, but sensitivity decreased to 23.1%, reflecting the trade-off between high specificity and reduced case detection (Supplementary Table S5). The calibration curves demonstrated satisfactory agreement between the predicted probabilities and the observed outcomes in both the training and validation sets (Fig.  5 B). The HL test further confirmed that the nomogram was well calibrated, with χ2 = 9.999 ( P  = 0.351) for the training set and χ2 = 12.133 ( P = 0.206) for the validation set. Additionally, DCA revealed that the threshold probability ranged from 6% to 98% in the training set (Fig.  5 C). Similarly, the DCA for the validation set demonstrated high clinical utility of the nomogram, with a threshold probability ranging from 6.8% to 99.9% (Fig.  5 C). Moreover, the CIC illustrated the projected number of individuals classified as high risk for severe CRF at various potential risk thresholds and visually depicted the proportion of these patients who were true positive cases. For example, in the training set (Fig.  5 D), applying a 60% risk threshold to a cohort of 1,000 screened patients would identify approximately 200 individuals as high risk, among whom an estimated 180 would be confirmed as having severe CRF. Similar results were observed in the validation set (Fig.  5 D). Fig. 5 Assessment of the nomogram’s predictive performance. A  ROC curve demonstrating discriminative ability. B  Calibration curve assessing the accuracy of predicted probabilities. C  DCA evaluating clinical utility. D  CIC demonstrating the practical implications within a clinical context Assessment of the nomogram’s predictive performance. A  ROC curve demonstrating discriminative ability. B  Calibration curve assessing the accuracy of predicted probabilities. C  DCA evaluating clinical utility. D  CIC demonstrating the practical implications within a clinical context

This study employed a cross-sectional design utilizing convenience sampling. Data were collected from eligible participants undergoing routine postoperative follow-up at the First Affiliated Hospital of China Medical University in Shenyang between March 2022 and December 2024. Inclusion criteria were: (1) diagnosis of primary NSCLC confirmed by pathological examination, with patients having undergone curative lung resection; (2) age ≥ 70 years; (3) Tumor-Node-Metastasis (TNM) staging IA-IIB; (4) voluntary participation with written informed consent; (5) ability to communicate in Chinese; (6) completion of surgery and/or adjuvant therapy at least three months prior. Exclusion criteria included: (1) psychiatric disorders; (2) metastases or tumor recurrence; (3) incomplete or missing data > 20%; (4) life-threatening diseases or other malignancies; (5) expected survival time of less than one year; (6) history of COVID-19 pneumonia. This study was approved by the Ethics Committee of the First Affiliated Hospital of China Medical University (No. 2022-124-2) and was conducted in accordance with the Declaration of Helsinki. Written informed consent was obtained from all participants before the study. To ensure the reliability of the predictive model and minimize the risk of overfitting, we determined the sample size based on the Event Per Variable (EPV) guideline [ 19 ]. Anticipating 10 risk variables and an incidence of severe CRF of 30%, as estimated from a preliminary evaluation of 100 samples, we applied the EPV formula: sample size = (EPV × number of variables)/event rate. Adhering to the EPV = 10 rule, we calculated a minimum required sample size of 372, accounting for a 10% dropout rate. Ultimately, 649 eligible patients were recruited, as shown in Fig. 1 . Fig. 1 Flow diagram illustrating the study design Flow diagram illustrating the study design The Cancer Fatigue Scale (CFS), developed by Okuyama et al. in 2000 [ 20 ], was employed to evaluate the severity of CRF in participants. The CFS comprises 15 items, each rated on a 5-point Likert scale ranging from 0 (“Not at all”) to 4 (“Very much”). The total score ranges from 0 to 60, with higher scores indicating more severe CRF: ≤5, no fatigue; 6–15, mild fatigue; 16–30, moderate fatigue; and 31–60, severe fatigue. In this study, the Cronbach’s α coefficient for the CFS was 0.941. In this study, we initially identified 40 candidate risk factors potentially associated with severe CRF. These included 18 demographic parameters (e.g., age, gender, education), 12 clinical variables (e.g., surgical approach, TNM stage, tumor size), and 10 multidimensional physical and psychological factors—including activities of daily living (ADL), breathlessness, nutritional status, sleep quality, anxiety, depression, psychological resilience, fear of progression (FoP), social support, and coping styles—evaluated using validated measurement instruments (see Supplementary Table S1 for details). All candidate variables and their corresponding assignments are presented in Supplementary Table S2. All statistical analyses were performed using R software (version 4.3.0), with a two-tailed P -value < 0.05 deemed statistically significant. The dataset was randomly divided into a training cohort (70%) and a validation cohort (30%). Continuous variables were presented as median (interquartile range) and were analyzed using the Mann-Whitney U test. Categorical variables were expressed as frequencies (%) and compared using the Chi-square test or Fisher’s exact test. Features with a missing ratio > 20% were excluded due to their limited information value. For the remaining data, multiple imputation was employed to minimize bias and ensure robustness. This approach accounts for the uncertainty associated with missing values and helps maintain the integrity of the dataset. The least absolute shrinkage and selection operator (LASSO) regression was applied to handle overfitting and identify significant risk factors for severe postoperative CRF. Ten-fold cross-validation was performed, with a random seed set at 123 to ensure reproducibility. The optimal penalty parameter (λ) was selected using the λ at 1 standard error (λ.1se) rule, providing a balance between model simplicity and predictive performance. Independent risk factors from the LASSO model were incorporated into a multivariable logistic regression analysis to evaluate their associations with CRF. Results were reported as adjusted odds ratios (ORs) with 95% confidence intervals (CIs). Multicollinearity among the independent variables included in the final multivariable logistic regression was assessed using the Variance Inflation Factor (VIF), with a threshold of VIF < 5 confirming no significant collinearity issues. Subsequently, a nomogram was constructed using the coefficients derived from the multivariable logistic regression analysis to provide a visual representation of the model. The total points derived from the nomogram were converted to the linear predictor (LP) using the model coefficients and intercept. The predicted probability of severe CRF was then calculated as p  = 1/(1 + exp(− LP)). Model discrimination was evaluated using the receiver operating characteristic (ROC) curve for both training and validation sets. Performance metrics, including Youden’s Index, specificity, sensitivity, positive predictive value (PPV), and negative predictive value (NPV), were calculated at various thresholds. Calibration plots and the Hosmer-Lemeshow (HL) test were utilized to assess goodness-of-fit. The clinical utility of the nomogram was evaluated using decision curve analysis (DCA) and clinical impact curve (CIC). Key R packages included glmnet (LASSO), lrm (nomograms), rms (calibration), pROC (ROC analysis), and rmda (DCA and CIC).

In conclusion, the long-term persistence of severe CRF was observed among elderly survivors with early-stage NSCLC resection. A LASSO-based nomogram was developed and validated using both training and validation sets to stratify the risk of severe CRF. The model identified five key risk factors—FoP, caregiver support, social support, ADL, and nutritional status—highlighting the multifactorial nature of CRF. By integrating behavioral variables (ADL and nutritional status) with psychosocial factors (FoP, caregiver support, and social support), this nomogram facilitates the targeted risk stratification of high-risk patients, thereby enabling the timely implementation of evidence-based interventions to mitigate severe postoperative CRF in this vulnerable population.

In this study, the prevalence of severe CRF among elderly NSCLC survivors was found to be 32.7%, slightly higher than the range of 17–31% reported in prior studies [ 10 , 21 , 22 ], which underscores the clinical importance of CRF in this demographic. The LASSO-based nomogram identified five key indicators—FoP, caregivers, social support, ADL, and nutrition status—as significant factors associated with the severity of CRF in these patients. The nomogram demonstrated strong accuracy, discrimination, and clinical utility in both the training and validation sets, highlighting its potential as a robust tool for stratifying the risk of severe CRF following surgery. To the best of our knowledge, this is the first model specifically developed to assess the long-term risk of severe CRF in elderly NSCLC patients following early-stage curative resection. By identifying critical risk factors, our nomogram provides an enhanced framework for understanding the etiology of CRF and offers actionable insights to aid clinicians in prioritizing interventions for those most at risk. Our findings identified that ADL impairment served as a pivotal factor in stratifying the risk of severe CRF in post-surgical elderly NSCLC patients. Postoperative declines in ADL reflect diminished physiological reserves, which undermine recovery from surgical stress and exacerbate cancer-related inflammatory and metabolic dysregulation [ 23 ]. This aligns with evidence from colorectal cancer cohorts, where ADL dysfunction predicted moderate-to-severe CRF following chemotherapy [ 17 ]. In our cohort, elderly patients with impaired ADL exhibited accelerated deterioration in physiological function and musculoskeletal health, potentially increasing their susceptibility to severe CRF. Consequently, preserving proficiency in ADL may be as crucial as tumor-specific management in optimizing postoperative quality of life for elderly NSCLC patients [ 24 ]. Routine assessment of ADL during follow-up visits is strongly recommended. Targeted interventions, such as personalized exercise programs, energy conservation strategies, and interdisciplinary rehabilitation support, may mitigate the severity of CRF by enhancing functional capacity in this vulnerable population. The identification of malnutrition as an independent risk factor for severe CRF is consistent with established frameworks that link nutritional deficits to the persistence of fatigue in oncology populations [ 25 ]. Malnutrition constitutes a persistent biological stressor that directly impairs mitochondrial bioenergetics, redox homeostasis, and neurotransmitter regulation—core mechanisms in the pathogenesis of CRF [ 26 ]. Furthermore, malnutrition exacerbates the progression of sarcopenia and the severity of anemia, compounding physical debilitation and subjective perception of fatigue in elderly patients following resection [ 27 ]. This relationship is bidirectional, as CRF may reciprocally exacerbate malnutrition by reducing appetite and nutrient intake [ 28 ]. Concordant findings in LC cohorts [ 29 ] and the predictive model by Hang et al. for colorectal cancer patients [ 17 ] reinforce nutritional status as a transdiagnostic determinant of CRF. Consequently, targeted nutritional interventions warrant integration into multidisciplinary care protocols to disrupt this vicious cycle among high-risk elderly NSCLC survivors. The significant association between FoP and severe CRF underscores a bidirectional psychosomatic relationship among elderly NSCLC survivors. FoP, characterized by persistent worry about cancer recurrence, leads to chronic stress and energy depletion, which hinder recovery and result in prolonged fatigue [ 30 ]. CRF symptoms, such as sleep disturbances and emotional exhaustion, further intensify FoP, thereby exacerbating fears of recurrence [ 31 ]. This reciprocity is supported by findings in cohorts of patients with endometriosis, where elevated FoP predicts the severity of fatigue [ 32 ]. Our findings confirmed heightened FoP as a critical risk factor for severe CRF in elderly NSCLC patients, a cohort particularly vulnerable to post-surgical existential distress. Notably, LC-specific symptoms, such as CRF, frequently persist long-term, demonstrating significant associations with FoP and diminished quality of life among survivors [ 33 ]. Therefore, routine screening for FoP (e.g., using FoP assessment tools) during postoperative surveillance is recommended to facilitate timely psychological interventions. The absence of caregiver support emerged as a significant risk factor for severe CRF in elderly NSCLC survivors. Individuals who lack caregiver support face increased logistical burdens and unmet needs following surgery, exacerbating physical and psychological distress [ 34 ]. Insufficient social networks have been well-documented to correlate with the persistence of chronic fatigue [ 35 ]. Patients without caregivers experience heightened psychological distress due to the lack of encouragement and supervision from family and friends [ 36 ]. This finding aligns with previous research linking lack of caregiver support to postoperative fatigue in gastrointestinal cancer patients [ 37 ]. For elderly NSCLC survivors, whose functional abilities are severely disrupted by surgery and adjuvant therapies, adequate social support can mitigate the negative effects of symptoms, motivate patients to overcome challenging conditions, and provide essential practical assistance [ 38 ]. Clinicians are urged to screen high-risk patients—particularly those who are isolated—and offer interventions such as caregiver counseling and peer support groups. Notably, previous studies have identified a critical role of psychological factors—including poor sleep quality [ 39 ], anxiety [ 40 ], depression [ 41 ], resilience [ 42 ], and coping styles [ 43 ]—in the development of CRF across diverse patient populations. However, in contrast to these findings, our analysis demonstrated no statistically significant associations between these psychological variables and the incidence of severe CRF among NSCLC survivors aged ≥ 70 years following surgery. This discrepancy, observed within our elderly cohort, suggests that the CRF etiology in older adults may diverge from established mechanisms found in younger or mixed-age populations. Contributing factors may include age-related differences in fatigue perception, reduced statistical power due to inherent sample size limitations in geriatric oncology studies, and complex interactions between physiological senescence processes and diminished psychological adaptive capacity in advanced age. Further investigation is warranted to elucidate the interplay between biological aging and psychosocial determinants in the pathogenesis of CRF among elderly cancer survivors. This study presents several limitations. First, the nomogram was developed based on a single-city cohort, which may limit its generalizability across different populations and settings. Second, the relatively modest sample size—particularly among frail elderly individuals with comorbidities—increases the likelihood of selection bias and limits the ability to detect rare or common risk factors. Third, reliance on self-reported measures for assessing CRF introduces potential recall bias and subjective interpretation, which could affect the accuracy of findings. Fourth, previously recognized risk factors for CRF, such as sleep quality, anxiety, depression, resilience, and coping styles were not included in the model, highlighting a need for further exploration to provide a more comprehensive assessment. Fifth, although internal validation has confirmed the model’s reliability, external validation with diverse cohorts is essential to ensure broader applicability. Finally, a key limitation of this study is the concurrent measurement of risk factors and CRF outcomes during follow-up. While the nomogram effectively classifies patients into different risk categories based on the data available at the time of assessment, it does not imply temporal causality or predict the future onset of CRF. Future longitudinal studies with repeated assessments of predictors before CRF manifestation are recommended. Such an approach could help establish the temporal sequence between potential risk factors and CRF, allowing for the development of models with true predictive utility.

Non-small cell lung cancer (NSCLC) accounts for approximately 85% of lung cancer (LC) cases worldwide and remains a leading cause of cancer-related mortality [ 1 ]. In China, the incidence and mortality of NSCLC increase substantially among individuals over 60 years of age [ 2 ], with rising life expectancy and an aging population underscoring a growing NSCLC burden in the elderly population [ 3 ]. Surgical resection with curative intent remains the cornerstone of early-stage NSCLC management [ 4 ]. Although advances in early detection and minimally invasive surgical techniques have improved survival outcomes [ 5 ], prolonged survivorship has revealed critical challenges in managing long-term treatment sequelae. Cancer-related fatigue (CRF), a common complication, is frequently underrecognized despite its disproportionate impact on elderly survivors [ 6 ]. This underscores the need for targeted interventions to manage and mitigate its effects during survivorship. Postoperative CRF, characterized by persistent physical, emotional, and cognitive exhaustion unrelated to exertion, affects 57% to 100% of patients from diagnosis to end-of-life and exerts a detrimental influence on prognosis [ 7 , 8 ]. In elderly populations, CRF is exacerbated by age-related physiological decline, multimorbidity, and diminished psychosocial resilience [ 9 ]. In NSCLC patients, a cohort study reported a postoperative CRF prevalence of 57%, with 17% experiencing moderate-to-severe fatigue lasting 1–6 years after resection [ 10 ]. CRF impedes functional recovery, daily activities, psychological well-being, and long-term quality of life [ 11 ]. Notably, severe CRF correlates with poor treatment adherence, elevated recurrence risk, and increased mortality rates [ 12 ]. Despite its clinical importance, CRF is often underrecognized or misattributed to normal aging in clinical practice. Timely screening and early interventions for severe postoperative CRF in elderly NSCLC survivors are essential to mitigate adverse outcomes and improve survivorship care. Current methods for assessing CRF typically depend on generalized assessment tools that lack specificity for elderly surgical populations and may overlook the multifaceted risk factors affecting these patients [ 13 ]. Risk stratification models have become increasingly valuable for identifying high-risk patients and guiding personalized interventions. Notably, nomograms are graphical calculators that combine multiple risk factors to estimate individualized risk levels, offering promising solutions for clinical application [ 14 ]. While models for CRF have demonstrated utility across various cancer types, including breast, cervical, and colorectal cancers [ 15 – 17 ], they rarely focus on elderly patients with early-stage NSCLC following surgical treatment. Prior studies have predominantly addressed perioperative CRF in LC patients [ 18 ], underscoring a need for long-term postoperative CRF assessment within this demographic. This study aims to develop and validate a nomogram specifically designed to stratify the risk of severe CRF in elderly survivors (aged ≥ 70 years) with early-stage NSCLC following curative surgical resection. Utilizing a multivariable framework, the study identifies key risk factors from demographic, clinical, psychosocial, and treatment-related domains, integrating them into a user-friendly visual tool. The model’s performance is rigorously evaluated through internal validation and calibration to ensure its applicability in clinical settings, thereby addressing the critical need for targeted risk assessment in this vulnerable population.

Supplementary Material 1. Supplementary Material 1.