Development and validation of a multi-variable prediction model for major postoperative complications after lung resection in patients aged ≥70 years with non-small-cell lung cancer

Introduction

Lung cancer is the leading cause of cancer deaths in China (1), with non-small-cell lung cancer (NSCLC) accounting for approximately 80–87%. NSCLC is a malignancy of the elderly with the median age of diagnosis being 70 years (2). Amidst the gloom of population aging, the prospect of healthy aging is not optimistic. The elderly will continue to account for a larger proportion of China’s population, and the number of oncogeriatric patients with NSCLC, subsequently, is expected to increase in the coming decades.

Surgical resection is the primary curative therapeutic option for NSCLC. Previous studies have shown that age is not a contraindication to surgical intervention (3,4). Surgical treatment in elderly patients can provide equivalent long-term survival compared to their younger counterparts (5,6). However, given the high incidence of postoperative complications in elderly patients ranging from 15% to 67% (7-9), surgeons must carefully assess the risk of major postoperative complications (MPC) and death before deciding to perform thoracic surgeries on elderly patients. Conversely, elderly patients would miss curative treatment if they refused surgical treatment because of the presumed high surgical risk (10). Thus, there is an urgent need to identify and predict the risk of MPC in elderly patients undergoing surgery for implementing rational treatment decisions.

Currently, few prediction models related to MPC have been proposed (8,10), however, small sample size or non-exclusive preoperative indicators hinder their clinical practice. Reported prediction model with an area under the receiver operating characteristic curve (AUC) of 0.70 was typically single-center study with small sample size (10). Another reported prediction model with an AUC of 0.61 based on the public database included 11 predictor variables and was inconvenient to use. Therefore, it is warranted to assess MPC comprehensively and identify the risk factors associated with MPC occurrence, and to construct a model for predicting the risk of MPC in elderly patients with NSCLC based on preoperative clinical variables.

To elucidate the risk factors and develop a credible prediction model of MPC in older patients, logistic regression analysis was used for better clinical practice. Preoperative variables were analyzed among the 63 variables from 989 elderly patients, specific relationship among the variables was explored and the new prediction model with bootstrapping AUC of 0.779 was constructed. We present this article in accordance with the TRIPOD reporting checklist (available at https://jtd.amegroups.com/article/view/10.21037/jtd-2025-1636/rc).

Methods

Data sources and study population

This was a single-center retrospective cohort study. All 989 NSCLC patients older than 70 years who underwent radical pulmonary resection of thoracotomy or video-assisted thoracoscopic surgery (VATS) approach at the Department of Thoracic Surgery of Peking University Cancer Hospital in China from October 1995 to June 2019 were screened. NSCLC cases were identified from medical records according to the International Classification of Diseases of Oncology, Third Edition (ICD-O-3) codes C34.0-C34.3 and C34.8-C34.9 (11). The NSCLC ICD-O-3 morphology codes included adenocarcinoma, bronchoalveolar carcinoma, squamous cell carcinoma, carcinoma not otherwise specified, and other variants. Clinical and follow-up data of the patients were obtained from medical records and collected through the hospital database by trained coordinators. The project was approved by the Ethics Committee of Peking University Cancer Hospital (No. 2020KT76). All participants provided written informed consent. The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments.

Clinical variables and outcomes

Five dimensions of preoperative variables were collected from the medical record: demographic characteristics, tumor characteristics, surgery-related characteristics, pulmonary function (echocardiographic results were not available in >70% of records and were therefore not considered as candidate predictors), and laboratory test results. Composite variables were merged to compensate weakness of the single variables, consisting of neutrophil to lymphocyte ratio (NLR), platelet to lymphocyte ratio (PLR), lymphocyte to monocyte ratio (LMR), prognostic nutritional index (PNI), systemic immunoinflammatory index (SII), and lymphocyte to white blood cell ratio (LWR). These variables and their coding formats have been described in the Appendix 1. MPC was defined as major complications within 30 days of surgery. Complications occurring within 30 days of surgery were meticulously documented and defined as MPC according to the Clavien-Dindo grading system, grades IIIb–V (12).

Statistical analysis

Continuous variables were presented as medians [interquartile range (IQR)] and compared by using Wilcoxon rank sum test. Categorical variables were summarized as frequency and proportion in each category and compared through Fisher’s exact test or Pearson’s Chi-squared test. Patients with missing data were excluded from the analysis.

Risk factors analysis

Logistic regression was used in the univariate analysis to identify risk factors associated with MPC. Factors that were significant in univariate analysis (P<0.05), coupled with age and gender, comprised the variable set and were included in multivariate logistic regression analysis. To evaluate the presence of interactions in the variable set, pairwise interaction analysis using the likelihood ratio test was performed to denote the interdependence between the effects of factors within the confines of given model of risk (13). Considering multiple comparisons, we performed Bonferroni correction for the P value of the likelihood ratio test, correlation heatmap of 38 clinical variables was shown in Figure S1. Additionally, RCS model was used to visualize the non-linear relationship between significant continuous variables in the variable set and MPC (14). Analyses were adjusted for the remaining variables in the variable set. Restricted cubic spline models with knots at the 5th, 50th, and 95th percentiles were fitted for each continuous variable measure to assess nonlinear associations. We tested for nonlinearity by comparing nested models with and without the spline terms using likelihood ratio tests.

Model construction

Two-step approach was used for dimensionality reduction and model construction. Variables were selected by the traditional multivariable logistic regression with backward stepwise regression and Akaike information criterion (AIC) minimization. The process we adhered to the BMJ TRIPOD guidance for prediction model development (15). Firstly, preoperative variables were categorized into five dimensions of demographic characteristics, tumor characteristics, surgery-related characteristics, cardiopulmonary function indicators, and laboratory test results. A stepwise backward regression with AIC criterion was performed to screen variables in each dimension. Secondly, the candidate variables screened from the five dimensions were recombined and the final model variables were again determined using multivariate stepwise regression. The structure of the model was visualized by a Nomogram.

Model performance

The fitness of prediction model was assessed by the Hosmer-Lemeshow goodness-of-fit Chi-squared test. Prediction performance was evaluated using discrimination and calibration. In which model discrimination was measured using AUC. The maximum AUC is 1.0 indicating a perfect prediction model. Conversely, a value of 0.5 indicates the outcome is correctly predicted by chance (16). The bootstrapping technique was used for internal validation with 1,000 re-samplings to estimate the impact of overfitting of the model (17). The calibration capability of our prediction model was visually evaluated with calibration plots. Perfect calibration is characterized by a line with an intercept of 0 and slope of 1. Decision curve analysis (DCA) was used to evaluate the clinical benefit of our model (18). By using DCA, we can make a clinical judgment about the relative value of benefits (true positive) and harms (false positive) associated with a clinical prediction tool.

To evaluate the effectiveness of the model for risk stratification and clinical usefulness, we divided patients based on their risk probabilities into 3 groups: low-risk, moderate-risk, and high-risk groups. The cutoff value that would achieve the desired partitioning was defined as the highest probability that maintained sensitivity of 98% and 80%, respectively. The number of patients, identified rate, and identified rate ratio in each risk group were calculated for the development sets.

All tests were two-tailed and P values less than 0.05 were considered statistically significant. The R Foundation for Statistical Computing software version 4.3.1 was used for all statistical analyses. The P values for the univariate statistical tests were not corrected for multiple testing, because those tests were taken as exploratory. Other statistical results relating to further refinement of the model were secondary and taken as descriptive only and did not require correction of their P values for multiple testing.

Results

Characteristics of patients

A total of 989 elderly patients were finally included, with about 69.4% aged between 70 and 74 years, about 26.3% aged between 75 and 79 years, the male to female ratio was comparable (59.5% vs. 40.5%). Body mass index (BMI) was also analyzed, about 41.7% were categorized as normal (BMI: 18.5–24.0 kg/m²), about 3.4% were grouped as lean (<18.5 kg/m²), about 54.9% were categorized as fat (>24.0 kg/m²). In terms of clinical staging, 686 (69.4%) were stage I, 118 (11.9%) were stage II, 185 (18.7%) were stage III or IV. Lobectomy was the most common surgery procedure, comprising 74.9% of the surgeries. In which 60.9% of the surgeries were performed minimally through video-assisted thoracotomy and 39.1% received thoracotomy. The characteristics of patients are summarized in Table 1. As patients were divided into two groups by MPC, significant difference can be observed in forced expiratory volume in the first second (FEV₁)% prediction (P=0.04), tumor diameter (P=0.003), staging (P=0.04), surgery method (P=0.009), surgery approach (P<0.001), low density lipoprotein cholesterol (P=0.005), red blood cell count (P<0.001), NLR (P=0.002), PLR (P=0.02), LMR (P=0.004), PNI (P=0.001), SII (P=0.02) and LWR (P=0.02).

Characteristics of predictor variables

Sixty-three preoperative variables consisting of 32 clinical variables and 24 laboratory test-related results were collected. Composite variables were also calculated according to previously published articles (19,20), which were proven to have a significant correlation with prognosis. Trends were also analyzed by the case year, female ratio, obesity ratio and VATS ratio showed slow but significant increasing trend in the 25 years (Figure 1). Ratio of lobectomy, tumor diameter and clinical staging showed a stable tendency (Figure 1). That is, female elderly patients have been increasing to the same level as males, and obesity has gradually increased, with the proportion close to 50%. VATS occupied the position of the main surgery strategy after 2011, which was now the preferred surgery regime, while lobectomy was still the standard surgery procedure in the past 2 decades.

Figure 1 Trends of predictor variables over 25 years. (A) Trend of gender proportion. (B) Trend of different BMI proportion. (C) Trend of tumor diameter proportion. (D) Trend of clinical staging proportion. (E) Trend of surgery approach proportion. (F) Trend of surgery method proportion. BMI, body mass index; VATS, video-assisted thoracoscopic surgery.

MPCs

There are 66 patients (6.67%) suffering from MPC (Table 2).

Among these 66 patients with MPC, 5 tops ranked were displayed here as severe hemoptysis (24.0%), respiratory failure (16.4%), severe postoperative bleeding (15.2%), pulmonary embolism (15.2%) and heart failure (6.3%) (Figure 2). Of note, the cohort included 4 patients (5.1%) who died in hospital and were diagnosed with respiratory failure caused by pulmonary infection. All of them have ever received sputum suction by fiberoptic bronchoscopy, indicating the significance of sputum elimination, especially for elderly patients, who would be hindered by pain and weakness.

Figure 2 Proportion of different MPC of all the MPC patients. MPC, major postoperative complication.

Analysis of risk factors

To explore the risk factors of MPC, all the preoperative variables were analyzed according to the logistic regression analysis. On univariate analysis using the predictors listed in Table 3, longer tumor diameter (P<0.001), lung cancer clinical staging III–IV (P=0.02), multi-lobectomy (P=0.004), thoracotomy (P<0.001), higher levels of low-density lipoprotein cholesterol (LDL-C) (P=0.001), higher levels of total protein (P=0.01), higher levels of creatinine (P=0.005), higher hematocrit level (P=0.003), higher levels of red blood cell count (P=0.001), higher levels of neutrophil count (P=0.001), higher levels of hemoglobin (P=0.003), NLR (P=0.002), PLR (P=0.02), PNI (P=0.002), and SII (P=0.02) were significantly increased the risk of MPC occurrence. While LMR >4.5 and >0.3 were protective against the development of MPC. After adjustment for multivariate logistic regression, five independent risk factors for MPC were identified: operation age ≥80 years [odds ratio (OR) =3.52; 95% confidence interval (CI): 1.11–11.16], thoracotomy (OR =4.84; 95% CI: 2.53–9.27), LDL-C (OR =1.53; 95% CI: 1.11–2.11), PLR >140.9 (OR =2.25; 95% CI: 1.06–4.78), and PNI >57.9 (OR =3.50; 95% CI: 1.62–7.56). Notably, operation age was not significant in the univariate analyses but was significant in the multivariate adjustments.

Intriguingly, we observed a significant interaction association between surgical methods and LDL-C (P_Bonferroni for interaction =0.02). Risk for MPC was significantly increased by the LDL-C in lobectomy group (OR =1.56; 95% CI: 1.06–2.29) and multi-lobectomy group (OR =3.06; 95% CI: 1.19–7.82) (Figure 3). No significant evidence was found for a non-linear association between LDL-C levels and MPC risk (Figure S2).

Figure 3 Interaction analysis. (A) displays the predicted probability of MPC on the y-axis against LDL-C levels on the x-axis. The three curves represent different surgical methods, illustrating how the predicted probability of MPC varies with increasing LDL-C across surgical subgroups; (B) shows the adjusted effect estimates derived from a multivariable logistic regression model that included multiple covariates and the interaction term between LDL-C and surgical method. CI, confidence interval; LDL-C, low density lipoprotein cholesterol; LMR, lymphocyte to monocyte ratio; LWR, lymphocyte to white blood cell ratio; MPC, major postoperative complication; NLR, neutrophil-to-lymphocyte ratio; OR, odds ratio; PLR, lymphocyte ratio; PNI, prognostic nutritional index; RBC, red blood cell count; SII, systemic immunoinflammatory index.

Prediction model construction and validation

The final model included 9 predictors consisting of sex (OR =1.29; 95% CI: 0.69–2.44), tumor diameter (OR =1.17; 95% CI: 1.00–1.37), thoracotomy (OR =4.88; 95% CI: 2.64–9.00), FEV1% prediction (OR =0.44; 95% CI: 0.03–6.29), LDL-C (OR =1.48; 95% CI: 1.09–2.02), red blood count (OR =2.97; 95% CI: 1.49–5.91), NLR (OR =2.25; 95% CI: 1.16–4.39), PLR (OR =1.90; 95% CI: 1.02–3.57) and PNI (OR =3.49; 95% CI: 1.76–6.93) (Figure 4A). The nomogram for this model is presented in Figure 4B. Figure S3 shows the variable importance ranking results.

Figure 4 Prediction model. (A) Multi-variable logistic analysis. (B) Nomograph for the prediction model: practical use—(I) locate the patient’s value on each axis; (II) drop a vertical line to the points row and record the score; (III) sum the nine scores to obtain total points; (IV) drop a vertical line from total points to the bottom “predicted probability” axis to read the risk. For example: 76-year-old man, LDL 7.3 mmol/L, platelet count 100×10⁹/L, absolute neutrophil count 6×10⁹/L, absolute lymphocyte count 2×10⁹/L, absolute monocyte count 0.5×10⁹/L, sex male, tumor diameter 3 cm, surgical approach open thoracotomy → predicted probability =17.7%. High-risk threshold (>16.47%) is exceeded. CI, confidence interval; FEV₁, forced expiratory volume in 1 second; LDL-C, low density lipoprotein cholesterol; NLR, neutrophil-to-lymphocyte ratio; OR, odds ratio; PLR, lymphocyte ratio; PNI, prognostic nutritional index; RBC, red blood cell count; VATS, video-assisted thoracoscopic surgery.

The AUC was 0.815 (95% CI: 0.759–0.871) for the development set (Figure 5A). Internal validation using bootstrapping resulted in a slightly lower AUC of 0.779 (95% CI: 0.723–0.836) after correction (Figure 5B). The Hosmer-Lemeshow test indicated good model fit with a P value of 0.67. The calibration plot showed excellent agreement between predicted and observed risk, which runs very close to the diagonal, showing excellent calibration (Figure 5C). Based on clinical experience suggesting 10–15% as an acceptable threshold for MPC risk, DCA demonstrated that the prediction model had higher net benefit compared to the “treat all” and “treat none” approaches across this risk threshold (Figure 5D), indicating good clinical utility.

Figure 5 Prediction model construction and validation. (A) Area under curve of the development set. (B) 1,000 Bootstrapping boxplot. (C) Calibration plot. (D) Decision curve analysis for MPC. AUC, area under the curve; CI, confidence interval; MPC, major postoperative complication; ROC, receiver operating characteristic.

Risk stratification

To facilitate clinical application, patients were stratified into low-, moderate- and high-risk groups based on predicted risk probabilities (Table 4). The risk cutoffs were defined as the highest predicted probabilities that maintained 98% and 80% sensitivity in the development set, which were 0.0375 and 0.1647, respectively. This divided patients into low- (50.15%), moderate- (39.84%) and high-risk (10.01%) groups. With this stratification, 31.82% of MPC cases were in the high-risk group while only 15.15% were in the low-risk group. The MPC identified rate was 10.5 times higher in the high- versus low-risk group, indicating enrichment of patients likely to develop MPC.

Sensitivity analysis

To further evaluate the robustness and validity of the prediction model, we performed the sensitivity analysis by validating it in elderly patients with different case years and in different age groups. The model showed stable discrimination with AUC of 0.784 in the before 2011 group and 0.689 in the after 2011 group (Figure S4). Elderly patients were also stratified by age (70–74 years, 75–79 years, ≥80 years). The model had consistent AUC across these groups: 0.802 for 70–74 years, 0.806 for 75–79 years, and 0.879 for ≥80 years (Figure S4). We also re-classified the cohort into two age groups: 70–80 years (n=946, major complication rate 6.4%), >80 years (n=43, major complication rate 7.6%). The prediction model’s discrimination was similar in both subgroups (70–80 years—AUC 0.80, 95% CI: 0.74–0.86; >80 years—AUC 0.86, 95% CI: 0.67–0.99), and the DeLong test indicated no significant difference between them (P=0.58).

Discussion

Populations are ageing at a faster pace than in the past and this demographic transition will have effects on all aspects of society. According to WHO, healthy ageing is more than the absence of disease; it is the process of developing and maintaining the functional ability that enables wellbeing in older age (21). It has been well documented that the elderly carry a higher risk for lung cancer as well as multimorbidity and frailty (22). Patients were less likely to undergo surgical resection for early-stage lung cancer if they perceived that their overall quality of life would be worse at 1 year postoperatively (23), thus, they required a careful selection for proper therapeutic strategy. The “age” may play an important role in predicting postoperative short-term outcomes; however, “elderly” could not directly deny the possibility of thoracic surgery in clinical work. Our study, among the population-based cohort of older NSCLC patients, aims to elucidate the characteristics of elderly lung cancer patients and construct the prediction model for MPC.

The variation of preoperative characteristics in all these in-hospital patients over the past 25 years exhibits a certain trend. Until the beginning of 1995, males represented the majority among elderly lung cancer patients while afterwards the proportion of female was continuously increasing. The trend of female ratio was consistent with the published survey report in Cancer Statistics, which indicated a global trend of decline in tumor incidence at 2.6% per year in men and 1.1% per year in women. However, female population had a lower rate of decline than men in tumor incidence (24). Another more pronounced trend was observed in VATS and thoracotomy; VATS has nearly fully replaced thoracotomy in the last 10 years from 7.5% to 96.9%. Previous articles revealed that postoperative hospitalization was shorter in the VTAS groups compared to thoracotomy (25). Furthermore, surgery method has remained relatively stable over the past 25 years for lung cancer; lobectomy has remained the accepted standard of care for early-stage lung cancer. Moreover, LDL-C level was identified as the independent risk factor of MPC, which was previously proven to increase the risk of lung cancer for elderly women (26,27) and high-level LDL-C is often accompanied by high metastatic potential (28). Notably, surgery method can be the potential variable affecting LDL in resulting in MPC. LDL-C produced increases in MPC that varied with the extent of surgical resection, that is, multi-lobectomy would make the elderly more prone to MPC caused by the high-level LDL-C than sub-lobectomy or lobectomy. Further investigation is needed to clarify whether low LDL-C is a causal risk factor or simply a marker of frailty/inflammation in the elderly surgical population.

Some associations between these single preoperative variables with MPC could be owing to chance alone; results should be interpreted accordingly. To build and improve prediction model, reliable metrics to evaluate the accuracy and reliability are essential. The prediction model based on preoperative predictor variables of the elderly patients finally reached an AUC of 0.815, both sensitivity and specificity demonstrated model’s favorable discrimination ability. From a more practical point of view, the prediction model performed better than the previously reported models in AUC. Single parameter failed to provide a comprehensive assessment for the prediction model. During the final four years of the study period [2016–2019], all patients were managed within a standardized ERAS pathway, suggesting that the derived risk estimates are applicable regardless of ERAS implementation. Nevertheless, we now emphasise that adoption of ERAS should be considered standard of care for high-risk elderly patients, and future external validation studies should explicitly report whether an ERAS protocol was in place.

DCA was furtherly applied to evaluate the multivariable prediction models, which displayed the probability of a positive event and standard diagnostic tests that produce a simple binary result (18). The elderly lung cancer patients were evaluated by the threshold probability of postoperative complications, then theoretical association between the threshold probability and the relative value of false-positive and negative results was ascertained to determine the value of the prediction model. Given the low baseline rate of major complications (6.7%), the model’s primary clinical value lies in safe exclusion of very low-risk patients rather than in precise identification of high-risk individuals. More broadly, the model could help both doctors and patients obtain the greatest benefit, indicating its good clinical practice.

Compared with the previous five-variable model (AUC 0.75), current prediction model gains discriminative ability while remaining entirely pre-operative. An open-access web calculator (https://PKUCH.org/MPC-nomogram) has also been developed to facilitate real-time bedside use, rendering it more suitable for pre-admission risk counselling and resource allocation. Risk of MPC for elderly patients can infer a probability, yet it fails to provide intuitive conclusion. To facilitate the application of the model in clinical practice, predictive values were stratified into low-, moderate-, and high-risk. The detection rate in high-risk group was 17.9 times higher than that in low-risk group, which reflects a considerable enrichment for potential patients with MPC. The model can segregate elderly patients into various risk groups with distinct MPC outcomes, showing good discrimination and clinical availability.

There are several limitations in this study. Although the prediction model reached a higher AUC value than the previous models, there was a paucity of validation data from other medical centers. Then, patients were retrospectively included from 1995 to 2019, heterogeneity of preoperative variables was likely to influence the statistical outcome, and lack of systematic cardiac-function data (ejection fraction, valve disease, or cardio-pulmonary exercise testing) may have omitted an important domain of risk in elderly patients. Additionally, this was a single-center study consisting of older lung cancer patients; medical information was collected according to the medical history documents, some information may be lost. Lack of geriatric frailty indices or performance-status scores [e.g., Karnofsky or Eastern Cooperative Oncology Group Performance Status (ECOG PS)] limits our ability to quantify baseline vulnerability and may confound the association between age and post-treatment complications. Besides, without external validation, the clinical utility remains limited. Finally, the incidence of MPCs was low; thus, our prediction model can accurately predict the negative outcome, while its positive predictive value is limited. The final model comprises nine predictors, including several laboratory variables, which may limit its immediate usability in daily practice.

Conclusions

Minimally invasive VATS occupied a dominant position in lung cancer surgery and overwhelmingly reduced the risk of MPC. LDL-C can be negatively affected by the extent of surgery in resulting MPC. The excellent prediction model helps facilitate surgical decision-making and encourages the implementation of more effective perioperative care guidelines for elderly patients. Further validation with other datasets is needed for a stable MPC prediction model.

Acknowledgments

We acknowledge Elsevier Academic Editing for editing this manuscript.

Footnote

Reporting Checklist: The authors have completed the TRIPOD reporting checklist. Available at https://jtd.amegroups.com/article/view/10.21037/jtd-2025-1636/rc

Data Sharing Statement: Available at https://jtd.amegroups.com/article/view/10.21037/jtd-2025-1636/dss

Peer Review File: Available at https://jtd.amegroups.com/article/view/10.21037/jtd-2025-1636/prf

Funding: This study was supported by the Noncommunicable Chronic Diseases–National Science and Technology Major Project (Nos. 2023ZD0501700 and 2023ZD0512400), and the Science Foundation of Peking University Cancer Hospital (No. BJCH2024CZ03).

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://jtd.amegroups.com/article/view/10.21037/jtd-2025-1636/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. The project was approved by the Ethics Committee of Peking University Cancer Hospital (No. 2020KT76). All participants provided written informed consent.

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. Siegel RL, Miller KD, Fuchs HE, et al. Cancer Statistics, 2021. CA Cancer J Clin 2021;71:7-33. [Crossref] [PubMed]
2. Lung Cancer Specialty Committee of Chinese Elderly Health Care Association. Enhanced Recovery after Surgery Specialty Committee of Tianjin Medical and Health Association. Consensus of Chinese Experts on Surgical Treatment of Lung Cancer in the Elderly (2022 Edition). Zhongguo Fei Ai Za Zhi 2023;26:83-92. [Crossref] [PubMed]
3. Thomas P, Sielezneff I, Ragni J, et al. Is lung cancer resection justified in patients aged over 70 years? Eur J Cardiothorac Surg 1993;7:246-50; discussion 250-1. [Crossref] [PubMed]
4. Jack CI, Lye M, Lesley F, et al. Surgery for lung cancer: age alone is not a contraindication. Int J Clin Pract 1997;51:423-6.
5. Higton AM, Monach J, Congleton J. Investigation and management of lung cancer in older adults. Lung Cancer 2010;69:209-12. [Crossref] [PubMed]
6. Cerfolio RJ, Bryant AS. Survival and outcomes of pulmonary resection for non-small cell lung cancer in the elderly: a nested case-control study. Ann Thorac Surg 2006;82:424-9; discussion 429-30. [Crossref] [PubMed]
7. Spaggiari L, Scanagatta P. Surgery of non-small cell lung cancer in the elderly. Curr Opin Oncol 2007;19:84-91. [Crossref] [PubMed]
8. Kates M, Perez X, Gribetz J, et al. Validation of a model to predict perioperative mortality from lung cancer resection in the elderly. Am J Respir Crit Care Med 2009;179:390-5. [Crossref] [PubMed]
9. Castillo MD, Heerdt PM. Pulmonary resection in the elderly. Curr Opin Anaesthesiol 2007;20:4-9. [Crossref] [PubMed]
10. Wang Y, Wu N, Zheng Q, et al. Prediction of Surgical Outcome by Modeling Based on Risk Factors of Morbidity After Pulmonary Resection for Lung Cancer in Older Adults. Ann Thorac Surg 2016;102:971-8. [Crossref] [PubMed]
11. World Health Organization. International classification of diseases for oncology (ICD-O). 3rd ed., 1st revision. 2013. Available online: https://iris.who.int/handle/10665/96612
12. Dindo D, Demartines N, Clavien PA. Classification of surgical complications: a new proposal with evaluation in a cohort of 6336 patients and results of a survey. Ann Surg 2004;240:205-13. [Crossref] [PubMed]
13. Rothman KJ, Greenland S, Walker AM. Concepts of interaction. Am J Epidemiol 1980;112:467-70. [Crossref] [PubMed]
14. Durrleman S, Simon R. Flexible regression models with cubic splines. Stat Med 1989;8:551-61. [Crossref] [PubMed]
15. Efthimiou O, Seo M, Chalkou K, et al. Developing clinical prediction models: a step-by-step guide. BMJ 2024;386:e078276. [Crossref] [PubMed]
16. Mandrekar JN. Receiver operating characteristic curve in diagnostic test assessment. J Thorac Oncol 2010;5:1315-6. [Crossref] [PubMed]
17. Steyerberg EW, Harrell FE Jr, Borsboom GJ, et al. Internal validation of predictive models: efficiency of some procedures for logistic regression analysis. J Clin Epidemiol 2001;54:774-81. [Crossref] [PubMed]
18. Vickers AJ, Elkin EB. Decision curve analysis: a novel method for evaluating prediction models. Med Decis Making 2006;26:565-74. [Crossref] [PubMed]
19. Wang C, Jin S, Xu S, et al. High Systemic Immune-Inflammation Index (SII) Represents an Unfavorable Prognostic Factor for Small Cell Lung Cancer Treated with Etoposide and Platinum-Based Chemotherapy. Lung 2020;198:405-14. [Crossref] [PubMed]
20. Yamamoto T, Kawada K, Obama K. Inflammation-Related Biomarkers for the Prediction of Prognosis in Colorectal Cancer Patients. Int J Mol Sci 2021;22:8002. [Crossref] [PubMed]
21. WHO. World report on ageing and health. 2015. [Accessed June 10, 2022]. Available online: https://apps.who.int/iris/handle/10665/186463
22. CDC. Older adults. 2021. [Accessed 19 February 2021]. Available online: https://www.cdc.gov/coronavirus/2019-ncov/need-extra-precautions/older-adults.html
23. Cykert S, Dilworth-Anderson P, Monroe MH, et al. Factors associated with decisions to undergo surgery among patients with newly diagnosed early-stage lung cancer. JAMA 2010;303:2368-76. [Crossref] [PubMed]
24. Siegel RL, Miller KD, Wagle NS, et al. Cancer statistics, 2023. CA Cancer J Clin 2023;73:17-48. [Crossref] [PubMed]
25. Yang HX, Woo KM, Sima CS, et al. Long-term Survival Based on the Surgical Approach to Lobectomy For Clinical Stage I Nonsmall Cell Lung Cancer: Comparison of Robotic, Video-assisted Thoracic Surgery, and Thoracotomy Lobectomy. Ann Surg 2017;265:431-7. [Crossref] [PubMed]
26. Chang AK, Barrett-Connor E, Edelstein S. Low plasma cholesterol predicts an increased risk of lung cancer in elderly women. Prev Med 1995;24:557-62. [Crossref] [PubMed]
27. Benn M, Tybjærg-Hansen A, Stender S, et al. Low-density lipoprotein cholesterol and the risk of cancer: a mendelian randomization study. J Natl Cancer Inst 2011;103:508-19. [Crossref] [PubMed]
28. Li R, Liu B, Liu Y, et al. Elevated serum lipid level can serve as early signal for metastasis for Non-Small Cell Lung Cancer patients: A retrospective nested case-control study. J Cancer 2020;11:7023-31. [Crossref] [PubMed]