Sentinel lymph node biopsy vs observation in high-risk cutaneous squamous cell carcinoma in immunosuppressed
Description
Supplementary material of observational study of the survival benefit of sentinel lymph node biopsy (SLNB) in immunocompetent and immunosuppressed patients with high-risk cutaneous squamous cell carcinoma (cSCC)
Files
Steps to reproduce
We have analysed two group SLNB and observation in high-risk cutaneous squamous cell carcinoma by the inverse probability of treatment weighting, which adjusts for confounding and allows analysis of all patients included, a propensity score method. First, it was calculated by performing logistic regression analysis of predictors of SLNB, tumor progression, and mortality (sex, age, year of diagnosis, and American Joint Committee on Cancer 8th Edition [AJCC8] and Brigham Women’s Hospital [BWH] tumor stages). Two weighting approaches were used: one targeting ATE (average treatment effect) and the other targeting ATT (average treatment effect among the treated population). The former assumes that all patients undergo SLNB, while the latter estimates the effect for all patients actually treated. Stabilized weights were used to prevent extreme weights from affecting certain variables. IPTW and all additional analyses were carried out separately for immunocompetent and immunosuppressed patients. We also adjusted for confounding effects using other well-established methods, such as propensity score matching (with different matching ratios), full optimal matching, doubly robust propensity score adjustment methods, and classic multivariate covariate adjustment. To test the robustness of our findings, we built additional propensity score models using other prognostic factors, such as perineural and lymphovascular invasion, and found no significant differences. Statistical analysis To compare survival between the treatment groups, IPTW-adjusted Kaplan-Meier curves with ATE weights were computed for each survival outcome of interest.We compared survival at 5 years after diagnosis and computed the following measures to assess the absolute magnitude of the effect of SLNB on clinical outcomes: absolute risk reduction (ARR), number needed to treat (NNT), and differences in restricted mean survival time (RMST). RMST is a measure of the average time free from an event up to a given time point. RMST differences, in turn, provide clinicians with an intuitive absolute measure of how long a given treatment delays an event in those treated. To assess the relative effect of SLNB in treated patients, we built an IPTW-weighted Cox regression model with robust standard errors and ATT weights. We also explored the effect of SLN positivity on nodal recurrence and disease-specific mortality using a multivariate Cox regression model including both immunocompetent and immunosuppressed patients. All tests were two-tailed and statistical significance was set at P < .05. All analyses were performed using R software and the additional packages “tidyverse”, “survival”, “MatchIt”, “IPWsurvival”, “WeighIt”, “cobalt”, and “arsenal”. The STROBE (Strengthening the Reporting of Observational Studies in Epidemiology) guidelines were applied to report our findings.