Predictors of Recurrence After Surgery in Patients with Stage I Non-Small Cell Lung Cancer.

1. Introduction
Lung cancer remains one of the most common malignancies worldwide, with approximately 1.8 million new diagnoses, accounting for about 13% of all cancer cases diagnosed [1]. Surgical resection represents the cornerstone of the treatment of locoregionally confined non-small cell lung cancer (NSCLC; clinical stages I to III) [2]. Despite curative intentions, approximately 40–55% of lung cancer patients experience recurrence, depending on the stage of the disease [3,4,5,6]. Several studies have documented recurrence in patients who underwent surgery for stage I NSCLC. Martini et al. [7] investigated 598 patients diagnosed with stage I NSCLC and found a recurrence rate of 27%, with 60% of these recurrences occurring within two years post-surgery. Maeda et al. [8] examined 483 patients who remained recurrence-free for five years following surgery, discovering that 4.8% experienced a delayed recurrence more than 5 years following their procedure. An improved risk stratification for these patients could justify the intensification of adjuvant therapies for a subgroup of early-stage lung cancer patients deemed to be at a high risk for recurrence [9]. Nevertheless, none of these studies have identified the specific factors associated with early recurrence or identified reliable prognostic indicators following recurrence.
This study aims to identify the prognostic factors associated with early and late recurrence in patients who have undergone radical surgery for stage I NSCLC.

2. Materials and Methods
We reviewed the medical records of patients from four major lung cancer centers who underwent surgical treatment for clinical stage I NSCLC from 2013 to 2021.
Complete resection was defined as a radical procedure without any macroscopic residual lesions and with a microscopically clear margin, associated with a systematic lymph node dissection including a minimum of six nodes/stations, three of which had to be mediastinal, always including the sub-carinal station. Wedge resections with an adequate margin greater than 10 mm were also considered radical in those patients with limited residual pulmonary function and for lesions less than 15 mm in diameter. Only patients with confirmed pN0 status were eligible for analysis to ensure accuracy of the final pathological stage. Patients with multiple lung cancers (either synchronous or metachronous) and those who underwent incomplete resection were excluded from this analysis.
All patients had a histological diagnosis of NSCLC before or during surgery. Surgical approaches included open thoracotomy, video-assisted thoracic surgery (VATS), and robotic-assisted thoracic surgery (RATS). All resected specimens underwent pathological examination. The histological classification included a subdivision of adenocarcinoma histotypes, and the pathologic T factor was reclassified according to the eighth edition of the Tumors, Node and Metastasis (TNM) classification system described by the International Association for the Study of Lung Cancer (IASLC) [10,11].
Collected variables included: age, gender, body mass index (BMI), comorbidities (yes/no), smoking history (ever or never), imaging appearance of the tumor on computed tomography (CT) scans (solid, ground-glass opacity, mixed), resection volume (wedge, segmentectomy, lobectomy or bilobectomy), surgical approach (open, VATS or RATS), and pT status. Early recurrence was defined as any recurrence occurring within 12 months after surgery. Local recurrence was defined as tumor relapse in the same chest cavity (including mediastinal structures and chest wall), while distant recurrence referred to tumor appearance in the other organs.
Data collected and reported in this study were retrieved from patient’s medical records and pathological and surgical registries. Follow-up data were obtained through outpatient visits, imaging studies (total body CT scan or FDG-PET scan) and phone interviews. To account for potential inter-institutional variability and temporal changes in surgical practice, additional sensitivity analyses were performed including center of treatment and calendar year of surgery as covariates in multivariable models. The data collection for this retrospective analysis was approved by the institutional review board (IRB) of Humanitas Research Hospital, and individual patient consent was obtained to authorize the use of their clinical data for research purposes and retrospective analysis in an anonymous form. The STROBE reporting recommendations were adopted in the reporting of this study [12].

Statistical Analysis
Descriptive statistics were used to summarize the demographic, clinical, surgical, and pathological characteristics of the cohort. Continuous variables were reported as mean ± standard deviation (SD) and compared using the Wilcoxon rank-sum test. Categorical variables were presented as counts and percentages, and differences between groups were assessed using the chi-square test or Fisher’s exact test, as appropriate. Firstly, the association of major variables with the time (early or late) and site (local or distant) of the recurrence was investigated. Then, disease-free survival (DFS) and overall survival (OS) were estimated using the Kaplan–Meier method. DFS was defined as the time from surgery to the first recurrence (local, distant, or both) or last follow-up. DFS in the presence/absence of different factors was compared with the log-rank test. The Cox proportional hazards model was used to determine independent predictors of recurrence. Univariate analyses were performed to identify variables potentially associated with recurrence. Variables with a p < 0.10 in the univariate analysis were included in the multivariate analysis. Hazard ratios (HRs) with 95% confidence intervals (CIs) were calculated. BMI was treated as a purely descriptive variable for the available subset and was not included in any further statistical testing.
All tests were two-sided, and a p-value < 0.05 was considered statistically significant. Statistical analyses were conducted using STATA v17.0 (StataCorp, College Station, TX, USA).

3. Results
We retrieved data from a total of 1132 patients; the main demographic, surgical and pathological features are summarized in Table 1.
A lobectomy was performed in 995 (88%) patients, an anatomic segmentectomy was performed in 56 (4%) and a wedge procedure was performed in 81 (7%). The majority of the operations (57.1%) were carried out through a minimally invasive approach: 552 VATSs and 106 RATSs, respectively. The open approach consisted of a standard lateral thoracotomy.
The histological classification revealed adenocarcinoma in 920 (81.3%) patients and squamous cell carcinoma in 141 (12.5%) patients. According to eighth edition of the TNM classification, 212 (18.7%) patients had stage pT1a, 402 (35.5%) had stage pT1b, 223 (19.7%) had stage pT1c and 295 (26.1%) had stage pT2a.
All the patients were followed over time with a mean follow-up of 57 ± 37 months.

3.1. Recurrences
Recurrence were observed in 224 patients (19.8%), of which 86 were local, 106 were distant-only and 32 involved both local and distant sites. Patients who experienced both local and distant recurrences simultaneously were classified and counted only within the ‘distant recurrence’ group, as this reflects the most clinically significant event.
In 72 (32.1%) patients, the recurrence occurred within one year after surgery. Patients who developed early recurrence were evaluated through a multidisciplinary approach. In cases where the recurrence was isolated, a local treatment with radical intent was considered.
Several factors were associated with early recurrence (Table 2). However, a multivariable logistic regression analysis was performed to identify independent predictors of early recurrence (Table 3). After adjustments for age, sex, the pT status, histology, and the year of surgery, the surgical approach was not independently associated with early recurrence. In contrast, later pT stages and squamous histology were significantly associated with an increased risk of early recurrence, while RATS showed a statistically significant association with early recurrence. This finding should be interpreted cautiously due to the small sample size and large confidence intervals.
Regarding the site of recurrence, distant relapses were more frequent in patients undergoing lobar resection (p = 0.0001) or receiving an open approach (p = 0.007) or with a pT2a tumor (p = 0.008) (Table 4).

3.2. Survival Analysis
Overall survival rates at 12, 36, and 60 months were 98.4%, 92.1% and 89.3% (Figure 1A), respectively. Disease-free survival rates for the same intervals were 93.5%, 83.3% and 76.5%, respectively (Figure 1B).
The disease-free survival rate was significantly higher for pT1a (p = 0.002, Figure 2A) and for the lepidic/papillar adenocarcinoma histology (p = 0.0015, Figure 2B). VATS/RATS resections had better outcomes compared to the open approach, although these were not significant (p = 0.170, Figure 2C).

3.3. Cox Regression
In the univariable analysis, patients who developed recurrence were more commonly operated on via an open approach (p < 0.001), with a pT stage above 1a (p < 0.001) and non-lepidic adenocarcinoma subtypes (p < 0.001) (Table 5).
The multivariable analysis showed that an open surgery (p = 0.032), pT stage > 1a (pT1b p = 0.036; pT1c p = 0.052; pT2a p = 0.003) and non-lepidic subtype among adenocarcinomas (acinar: p = 0.012; solid: p = 0.016, other: p = 0.008) were independent predictors of recurrence (Table 5). In the primary multivariable model, the surgical approach appeared to be associated with the recurrence risk. In this adjusted model, the association between the surgical approach and recurrence was attenuated and no longer statistically significant (VATS vs open: HR 1.37, 95% CI 0.94–2.01, p = 0.102; RATS vs open: HR 1.97, 95% CI 0.91–4.27, p = 0.087). The calendar year showed a modest association with the recurrence risk (HR 1.09 per year, 95% CI 1.01–1.18, p = 0.025). Full results are reported in Supplementary Table S1.

4. Discussion
Despite having radical surgery, patients with completely resected node-negative NSCLC remain at risk for recurrence. Many retrospective studies have analyzed the risk factors for relapse in stage I NSCLC following radical resection and have attempted to develop a risk prediction model based on clinical factors and biomarkers [13,14,15,16,17,18,19,20,21,22].
In our cohort of 1132 of patients, the 5-year recurrence rate was 19.8%, consistent with previous reports [23,24]. We found that the pathological stage, particularly stages greater than IA, significantly contributed to the risk of recurrence. This aligns with findings from Zhao et al. [25], who reported similar conclusions in their study of stage I lung adenocarcinoma. We also identified strong associations between other clinic-pathological features and recurrence, specifically adenocarcinoma subtypes. Our analysis also showed that the non-lepidic subtype of adenocarcinoma was linked to a higher risk of recurrence within five years, and the multivariate analysis confirmed this finding. A recent validation study of the IASLC histologic grading system also supported these findings [16]. Woo et al. [26] demonstrated that grade 3 tumors and non-lepidic subtypes were associated with worse recurrence-free survival and overall survival [20,26].
Our analysis suggested that patients undergoing open thoracotomy may have a higher recurrence risk. However, after additional adjustments for the center and calendar year, this association is no longer significant, and the direction changes. This finding is not widely corroborated in the current literature [27]. At least several studies have suggested that the open surgery is the safer operation, and the minimally invasive approach is not inferior in terms of overall survival or the risk of recurrence in early-stage lung cancer [28,29,30]. The apparent association between the surgical approach and recurrence should therefore be interpreted cautiously. In retrospective surgical series, an open thoracotomy is often selected for patients with more complex or radiologically suspicious tumors, which may inherently carry a higher risk of recurrence. In addition, given the retrospective design and the long inclusion period (2013–2021), open surgery was more frequently performed in the earlier years of the study period, when minimally invasive approaches were less widely adopted and surgeon expertise with VATS or robotic techniques was still evolving across centers.
Our findings indicate that the association between the surgical approach and outcomes is model-dependent and loses statistical significance after adjustments for the center and calendar year, suggesting that institutional and temporal factors may partly explain the observed differences.
As far as the resection volume, we found that sublobar resection was not associated with an increased rate of local or distant recurrence. Similarly, several studies suggest that a sublobar resection may be equivalent to a lobectomy in a selected subset of patients, such as ground-glass opacities and well-differentiated tumors less than 2 cm [31,32,33]. On the other hand, Varlotto et al. reported an increase in local recurrence rates among patients who underwent a sublobar resection for moderately differentiated NSCLC tumors larger than 2 cm [34], while El-Sherif et al. found that patients with stage IB NSCLC had worse disease-free survival rates when treated with sublobar resection compared to the standard lobectomy [35]. The low percentage of sublobar resections (12.1%) in our series may suggest a strict selection of patients that justifies the results being comparable to lobar resection.
Based on the survival analysis presented in our study, the possibility of offering adjuvant therapy to selected patients should be considered, even when pathological staging indicates early-stage oncologic disease.

Limitations
The main limitation of our study is its retrospective design, as selection bias can play a role that cannot be reliably controlled. Furthermore, due to the multicenter nature of this study, the choice of surgical approach or the extent of the pulmonary resection may be influenced by the center of origin; in particular, the variable availability of robotic and minimally invasive platforms across centers may have resulted in a heterogeneous distribution of cases and should therefore be considered a potential confounding factor. We developed our multivariable analysis by selecting variables that were reported in the literature and supported by our clinical experience as potential predictors of recurrence. However, given the sample size, we were limited to including only five variables in the multivariable model.
The histopathologic classification of lung tumors has undergone major changes over the past decade, and certain features (such as spread through airspaces, STAS) were not systematically recorded at the start of our study. Given the prognostic relevance of STAS [36], further research is warranted as its reporting becomes increasingly standardized.

5. Conclusions
In our analysis, the 5-year recurrence rate is comparable to that reported in the literature, and we demonstrated that the disease-free survival rate was significantly worse for patients at stages greater than pT1a and in those with non-lepidic adenocarcinoma.