We retrospectively examined 3,125 consecutive patients with NSCLC who underwent surgical treatments by a single surgical team at the National Taiwan University Hospital from January 2011 to June 2018. The inclusion criteria were pStage I NSCLC, age ≥75 years, and patients undergoing thoracoscopic LR or SR. In total, 258 patients were included in the study ( Figure 1 ). Our hospital research ethics committee approved this retrospective study (202005006RINC) and waived the requirement for informed consent.

Preoperative clinical data was obtained from a prospectively collected database at our institute. Preoperative comorbidities were evaluated according to the Charlson Comorbidity Index (15). All preoperative studies were performed at our institute within two months before the operation, including pulmonary function test, echocardiography, test for preoperative serum carcinoembryonic antigen (CEA) levels, chest radiography, contrast-enhanced chest/abdomen CT, brain CT or magnetic resonance imaging (MRI), bone scanning, and positron emission tomography (PET). Tumor depth was defined as the shortest distance from the tumor to the pleura. The consolidation-to-tumor ratio (C/T ratio) was the ratio of the maximum diameter of consolidation to the maximum tumor diameter, measured by using CT according to the Japan Clinical Oncology Group 0201 definition (16). All imaging studies were independently reviewed by two thoracic surgeons in our surgical team using standard pictures with a commercially available viewer (IMPAX 5.2; Agfa HealthCare N.V., Mortsel, Belgium).

At our institute, SR was indicated if the tumor diameter was ≤2 cm and the C/T ratio was ≤50% in patients with NSCLC without lymph node or distant metastases. SR was considered if the resection margin was at >2 cm from than the tumor diameter. Otherwise, LR was performed to achieve the appropriate resection margin, which was defined based on gross measurements during surgery. Each surgeon made the decision for which surgical method (i.e., LR or SR) to use, which was then approved during a weekly multidisciplinary lung cancer meeting attended by surgeons, radiologists, pathologists, and oncologists before the surgery. After the SR was completed, mediastinal lymph nodes were sampled, depending on the location of the lung nodule, for lung cancer staging. The severity of postoperative complications was graded according to the Clavien-Dindo classification, which classifies postoperative complications as requiring pharmacological treatment (Grade II), surgical or radiological intervention (Grade III), or life-threatening complication (Grade IV) (17).

The histological diagnosis and pathological features were retrospectively collected from pathological analysis documents at the National Taiwan University Hospital. The pathological staging of lung tumors was determined according to the 8th Edition of the American Joint Committee on Cancer TNM staging system (18). The histopathology pattern was classified according to the 2015 World Health Organization criteria (19). Tumor spread through air spaces (STAS) was defined as tumor cells within air spaces in the lung parenchyma at a distance of at least one alveolus away from the main tumor (20).

Patients were monitored postoperatively in the outpatient clinic with physical examinations, serum CEA measurements, and chest CT examinations every 6 months for the first 2 years and every 6–12 months thereafter. Lung cancer recurrence was initially assessed using CT, PET, and MRI. It was further confirmed using CT-guided, endobronchial ultrasound-guided, and transthoracic echogram-guided biopsies or pleural effusion tapping.

For patients who underwent sublobar resection with an involved margin, the sequential treatments are completion lobectomy or radiotherapy within eight weeks. Adjuvant cisplatin-based chemotherapy is indicated for patients with nodal metastasis. Radiation therapy dose of 45-60 Gy as the highest cumulative radiation dose is indicated for lymph node metastases with extranodal extension.

The classification of the cause of death were lung cancer specific and non-lung cancer specific. Lung cancer-specific mortality was defined as death owing to recurrent disease associated with resected NSCLCs. Non-lung cancer-specific mortality was defined as death owing to specific causes other than NSCLCs. Death as a result of other malignancies was also regarded as non-lung cancer-specific death.

A propensity score matching (PSM) analysis was adopted to select comparable cases in two groups for further analyses and to remove the potential impact of confounding factors. Briefly, the propensity score was calculated using a multivariate logistic regression model. Age, performance status, total tumor size, solid tumor size, and C/T ratio were included as covariates. Patients undergoing LR were matched to those undergoing SR in a 1:1 ratio, with a standardized mean difference (SMD) of <0.2 of the logit propensity score (21). The PSM algorithm was executed with PSM for the Statistical Package for the Social Sciences (SPSS) (add-on for SPSS, version 3.04) and the underlying R packages (version 3.3.0).

Descriptive statistics are reported as mean ± standard deviation for continuous data and as percentages (%) for categorical data. Before PSM, chi-squared tests were performed for categorical variables, and Student’s t-test was performed for continuous variables. These values were also estimated after PSM using chi-squared tests for categorical variables and paired t-tests for continuous data. Overall survival (OS) and disease-free survival (DFS) rates were calculated using the Kaplan-Meier method and analyzed using log-rank tests. Multivariate analyses were performed using the Cox regression model after adjusting for significant confounding factors in the univariate model. Statistical analyses were performed using SPSS version 25 (IBM, Armonk, New York, USA). All tests were two sided, and p-values <0.05 were considered significant.

The study cohort included 84 (32.6%) patients who underwent LR and 174 (67.4%) patients (including 128 wedge resections and 46 segmentectomies) who underwent SR. The mean follow-up period was 39.8 ± 24.6 months. The mean age of all 258 patients was 78.6 ± 3.3 years (range: 75-90). The majority of patients were female (57.0%) and non-smoker (79.5%). The comorbidity index of the study group was 1.5 ± 1.4. The details of patient demographics and clinical characteristics of the study cohort are listed in Supplementary Table 1 .

Before PSM, patients in the LR group were younger (p = 0.013) and had a better Eastern Cooperative Oncology Group performance status (p = 0.007) than those in the SR group. The total tumor diameter, solid component diameter, and tumor depth were significantly larger in the LR group than in the SR group ( Supplementary Table 1 ). The pathological features showed that the pathological tumor diameter and pStage were significantly different between the two groups. The LR group had a larger pathological tumor diameter (p < 0.001) and a higher presence of IB pStage (p < 0.001) than the SR group ( Supplementary Table 2 ). There was no significant difference in positive STAS between lobectomy and sublobar resection group in this study cohort (42.1% vs. 31.5%, p=0.577).

PSM was used to identify 60 well-balanced patients in each group for a survival comparison. After PSM, only the tumor depth was greater in the LR group than in the SR group (p = 0.001). No other differences in demographic, clinical, histopathologic features, and pStage were observed between the two groups ( Tables 1 , 2 ). The distribution of the propensity scores and SMDs for both groups before and after PSM is shown in Supplementary Figure 1 .

There was no 30-day mortality in the study cohort, and only one patient in the LR group underwent conversion to thoracotomy. After matching, SR was associated with better perioperative outcomes, including lower blood loss and shorter operative times, hospital stays, and chest tube duration ( Table 3 and Supplementary Table 3 ). Postoperative complications were less common in the SR group than in the LR group (10.0% vs. 28.3%, p = 0.011). Severe postoperative complications with a Clavien-Dindo classification (17) greater than 3a were also less common in the SR group than in the LR group (10.0% vs. 25.0%, p = 0.031). The details of postoperative complications are listed in Supplementary Table 4 .

Our results showed favorable clinical outcomes for both thoracoscopic LR and SR in geriatric patients with pStage I NSCLC. The Kaplan-Meier OS analysis showed no difference between the LR and SR groups with respect to lung cancer-specific OS (5-year lung cancer-specific OS: 100% vs. 98.2%, p = 0.081), non-lung cancer-specific OS (5-year non-lung cancer-specific OS: 83.9% vs. 90.7%, p = 0.853), and DFS (5-year DFS: 78.8% vs. 82.4%; p = 0.643). Within the follow-up period, 33 patients (12.8%) developed recurrence, among whom, 14 underwent LR and 19 underwent SR ( Supplementary Table 5 ).

In this study cohort, 11 patients received SR with microscopic resection margin involvement, and only one patient (9.1%, 1/11) received reresection. The rest of the patients received radiotherapy for positive resection margin. In total, 11 patients developed local recurrence, 27.3% (3/11) patients received reresection, and the remaining received adjuvant therapy (tyrosine kinase inhibitor for mutant epidermal growth factor receptor [EGFR], or chemotherapy for wild-type EGFR mutation) plus radiotherapy.

After matching, the Kaplan-Meier OS analysis showed no difference between the two groups with respect to both lung cancer-specific OS (LR vs. SR; 5-year lung cancer-specific OS: 100% vs. 94.6%, p = 0.116) and non-lung cancer-specific OS (LR vs. SR; 5-year non-lung cancer-specific OS: 83.9% vs. 90.7%, p = 0.408) ( Figure 2 ). The Kaplan-Meier DFS analysis also showed no significant difference between the LR and SR groups (5-year DFS: 80.3% vs. 81.3%; p = 0.597) ( Figure 2 ). The causes of mortality stratified according to the lung cancer specific and non-lung cancer specific variables are shown in Supplementary Table 6 . There was no significant difference in the cause of mortality between the LR and SR groups (p = 0.356).

The results of the univariate analysis, which examined the correlation between clinicopathological features and OS, are listed in Supplementary Tables 7–9 . We evaluated both lung cancer-specific and non-lung cancer-specific survival. Because of the limited number of events, no multivariate analysis was performed for lung cancer-specific OS. The multivariate analysis results are listed in Supplementary Table 9 . Female sex (p = 0.049) was the only independent factor of non-cancer-specific death and was associated with increased survival. The resection method was not a prognostic factor for lung cancer-specific and non-lung cancer-specific OS.