Primary outcomes

Overall incidence of N1 pathological lymph nodes (Hilar 10/11, Lobar 12, Sublobar 13/14)

Incidence of N1 pathological lymph nodes (Hilar 10/11, Lobar 12, Sublobar 13/14) in patients with pathological evidence of visceral pleural invasion.

Secondary outcomes

Overall survival (1-3-5 years)

Disease free-survival (1-3-5 years)

Tumor recurrence (pattern: local, regional, distance)

Inclusion criteria

Anatomical resection for NSCLC <3 cm (lobectomy, bilobectomy, segmentectomy) with radical mediastinal lymphadenectomy (at least 6 lymph nodes from 3 mediastinal nodal stations)

Samples from the intrapulmonary stations 12, 13, and 14 lymph nodes

Resection of lymph nodes station 10 and 11 during hilar separation

R0 resection

Exclusion criteria

Prior or synchronous lung cancer

pN2

Pneumonectomy

R1/R2 resection

M1

Neo-adjuvant treatment

Variables to be collected

Age

Sex (M/F/Other)

Cardiac comorbidity (Y/N)

Pulmonary comorbidity (Y/N)

Pre-op FEV1 (L,%)

Pre-op DLCO (%)

Pre-op PET (Y/N)

Pre-op mediastinal Staging (Y/N) if Y (EBUS/EUS, Mediastinoscopy, others + biopsied stations level)

Date of Surgery

Surgical access (Open, VATS, Robotic)

Site (RUL, MLL, RLL, LUL, LLL)

Type of resection (Wedge, Segmentectomy, Lobectomy, Bilobectomy, Pneumonectomy)

Length of stay

Tumor Size

Histology

Extent of lymphadenectomy

cTNM

pTNM

PL status

Metastatic Lymph nodes N1 (10, 11, 12, 13/14)

PDL1 (%)

EGFR Mutation (Y/N/U)

ALK Mutation (Y/N/U)

Post-op complication (Y/N)

Re-op (Y/N) if Y Reason

Adjuvant treatment (Y/N) (Chemotherapy, Radiation, Chemo-Radiation, Immunotherapy, Chemo Immunotherapy, Chemo-Radio-Immunotherapy, Radio-Immunotherapy)

- If Chemotherapy: type of Chemo

- If Radiotherapy: regimen

Recurrence (Y/N) if Y: date of recurrence

Death (Y/N) if Y: date of death

Data management

Data for this study will be collected, recorded and stored using REDCap (Research Electronic Data Capture). REDCap is a secure, web application designed to support data capture for research studies. It includes features for HIPAA (Health Insurance Portability and Accountability Act) compliance including real-time data entry validation (e.g. for data types and range checks), a full audit trail, user-based privileges, de-identified data export mechanism to statistical packages (SPSS), and integration with the institutional Active Directory. Access to study data in REDCap will be restricted to the members of the study team with authentication through hospitals credentials. The REDCap database and web server are housed on secure platforms that are backed up daily. The data collected for the review will be stored at Lucerne Cantonal Hospital under the direction of the Principal Investigator. In accordance with the Medical Professional Code, all important study documents (e.g., CRFs) will be archived for at least ten years after the end of the study.

The results of the study will be published in peer-reviewed journals and presented to the scientific community at national and international conferences.

Sample size calculation

The minimum sample size to estimate the incidence was calculated using the method described by Humphrey et al. [9]:

Setting the expected incidence (of patients with an N1 lymph node metastasis to 0.2, and the allowable error of the incidence (L) to 0.2 the minimum required sample size is 385 patients. The allowable error was chosen because it represents a reasonable balance between precision and feasibility. After doubling the number to account for loss to follow-up and a lower true incidence, the final sample size is 770 patients.

The minimum sample size for the survival analysis was calculated using the method described by Schoenfeld [10]:

We set the type I error rate to 0.05 (), and the type II error rate to 0.2 (). Based on previous studies the expected percentage of patients with N1 lymph node metastasis was set to 20% (q1) and the relative hazard (RH) was set to 1.5 [11–13]. The number of events needed to achieve a power of 80% in the cohort was 298. To account for this we increased the final sample size by 10% to 857.

Statistical analysis

The incidence of N1 lymph node metastasis overall and the incidence of metastasis to the different lymph node stations (Hilar 10/11, Lobar 12, Sublobar 13/14) will be calculated by dividing the number of the respective events by the patient years separately. Ninety-five percent confidence intervals will be calculated using the method described by Ulm [14].

To investigate the association between visceral pleural invasion and the presence of metastatic lymph nodes univariable and multivariable logistic regression models will be fitted to the data. The multivariable logistic regression will be adjusted for age, sex, and tumor characteristics (size, histology, and mutation status).

For the secondary outcomes, Kaplan-Meier curves for the overall cohort and the subgroups (presence of N1 metastasis (Yes/No), presence of metastasis in intrapulmonary lymph nodes (Yes/No), presence of pleural invasion (Yes/No), sex (male/female), and age (≥ median, < median) will be created, for the disease-free survival and overall survival separately. Invariable Cox proportional-hazards models will be fitted for each of the predictors of interest. Hazard ratios and corresponding 95% confidence intervals will be calculated from the coefficients. Furthermore, we will also create multivariable Cox models adjusting for age, sex, and relevant comorbidities. The assumption of constant hazards will be checked by calculating the scaled Schoenfeld residuals and inspecting the correlation with time both statistically and visually [10]. The assumption of non-informative censoring will be tested by creating optimistic models in which all censoring will be set to the longest survival time, and pessimistic models, in which all censored time points will be regarded as events. If the results point in the same direction, we will regard the censoring as non-informative. In case of a violation of the assumption of non-informative censoring we will conduct a thorough examination of key variables, such as disease status, surgery success, and follow-up length, to identify any relationship between these variables and the censoring event as well as the outcome of interest. If we find any evidence of such a relation, we will then adjust for the relevant variables and perform sensitivity analyses to evaluate the robustness of our findings to potential violations of the non-informative censoring assumption.

If the number of missing data of a variable does not exceed 5%, it will be considered for impuatation. The relationship of missingness between different variables will be graphically explored using the “naniar” package for R. Heatmaps of missing data and scatter plots of numerical data indicating missigness will be created. If the data is considered to be at “missing at random” or “missing completely at random” we will impute it using a random-forest based imputation algorithm as implemented in the “missForest” package for R. This algorithm has been shown to perform as good as other state-of-the-art imputation methods and can handle a wide array of data types [15]. All analysis will be performed using the statistical software “R” (Version 4.1.2).

Ethical considerations

Individual patient consent is not required as data are collected under existing permissions (general consent) cancer registration in each jurisdiction.

Only members of the study team will have access to the information. The purpose of sharing data is to increase the ability of the investigators to analyze and translate data into a meaningful report validating their individual findings. A Data Transfer Agreement (namely a formal contract that documents which data are being transferred, the format of the data collection file and how the data can be used) will be signed between the principal investigator and the local investigators. The agreement protects all the centers who will provide the data, ensuring that data will not be misused. Each center is represented by one local investigator who will participate to regular meetings, held quarterly throughout the year, where the trend of the data collection along with critical points will be discussed in order to improve the understanding of the project and protocol.

Data will be de-identified, coded securely, and stored on a password-protected, virus protected, firewalled and encrypted server at Cantonal Hospital of Lucerne. The data from the other centers will be entered in REDCAP by the local investigator through a link provided by the principal investigator. Since this protocol describes a no-intervention study, the only risk to the patient would involve a breach of privacy. This risk is minimized, however, since only indirect identifiers will be recorded at all points throughout the course of the study, ensuring that patient confidentiality is safely kept and the patients will only be recorded in the database by their participant identification code. The link between this code and the patient identity will be in hard copy only and will be stored in a locked location separate from the data collections forms. This code document will be destroyed upon completion of the study data extraction. The members of the study team will maintain professionalism at all times, ensuring the full protection of each patient’s privacy.