br Statistical analysis br Mean and median values of continu
2.3. Statistical analysis
Mean and median values of continuous variables were calculated for the population description, while categorical variables were described with frequencies and percentages. For baseline categorical variables, the two groups of NSCLC patients based on TP53 mutation status were compared using a Chi2 test or a Fisher’s exact test for smaller sample sizes (≤ 5). Univariate analysis by logistic regression was used to de-termine the variables associated with ORR. Unadjusted Odds Ratio (OR)
with 95% Confidence Interval (CI) were reported. Multivariate analysis was performed using logistic regression. The variables associated with ORR (p value ≤ 0.20) in the univariate analysis were included in multivariate model. Kaplan-Meier analysis and log-rank test were re-spectively used to estimate the median survival times and their 95% CI, and statistical significance for the two subgroups: TP53-mutated and TP53-wild-type. Hazard Ratios (HR) and their 95% CI were calculated from univariate analyses with Cox proportional hazards models. Multivariate Cox analyses were performed using stepwise regression after checking the proportional hazard hypothesis graphically for all variables, and the absence of multicolinearity between variables by calculating inflation variance factors. Variables with more than 5% missing data were not introduced in the multivariate analysis. The variables tested in the univariate analyses of OS and PFS were gender, age, smoking status, performance status (PS), histological subtype, number of lines of treatment, PD-L1 Puromycin dihydrochloride level, KRAS and TP53 mutational status. Only variables that showed association with survival (p-value set at < 0.20) were introduced in the multivariate models of OS and PFS, with three additional clinical variables (gender, age and smoking status) forced in the OS multivariate model because of their clinical interest. Findings were considered statistically significant if two-sided p values were < 0.05.
All statistical analyses were performed using SAS™ version 9.4 software, University Edition for Windows™. Median follow-up was evaluated by the reverse Kaplan-Meier method. We censored follow-up on February 26th, 2018.
status available for analysis. Of all the patients with no NGS available (n = 50), 32 (64.0%) presented with squamous NSCLC (understandably so given that NGS is not reimbursed for squamous cell carcinoma [SCC] patients in France), two (4.0%) with large-cell neuroendocrine carci-noma, one (2.0%) with large-cell carcinoma, one (2.0%) with sarco-matoid carcinoma, one (2.0%) with undiﬀerentiated carcinoma, four (8.0%) only had liquid biopsy, four (8.0%) had uninterpretable NGS due to lack of DNA amplification, and finally, adequate tumor material was exhausted for five (10.0%) patients. Comparison of the clinical and pathological features of included patients and patients with no avail-able NGS, is presented in Supplementary Table 1. There were no sig-nificantly diﬀerent characteristics between patients with or without NGS according to age, gender, smoking, performance status. However, SCC histology and nivolumab monotherapy were more frequent and the number of lines more important in the non NGS group, reflecting the specific immunotherapy registrations and indications in the SCC sub-type. Table 2 reports the baseline characteristics of patients in the TP53-mutated and TP53-wild-type groups. Overall, 57 (79.2%) patients were younger than 70 years of age at ICI initiation, and median age at ICI initiation was 61 (range: 33–83) years; 52 (72.2%) patients were male, and 53 (73.6%) were current smokers or recent quitters (< 10 years). There were 53 (73.6%) patients with adenocarcinoma and 19 (26.4%) with other histological subtypes, 11 (15.3%) of which were SCC, three (4.2%) sarcomatoid carcinomas, one (1.4%) large-cell neu-roendocrine carcinoma features, one (1.4%) large-cell carcinoma, and three (4.2%) very undiﬀerentiated carcinomas not otherwise specified (NOS). NGS was performed in these patients because they were light or never-smokers, suggesting the possibility of potentially targetable ad-dictive oncogenic alterations. On ICI initiation, 67 (93.1%) patients were metastatic; five (6.9%) presented with Stage IIIA, four of whom were refused surgery and radiotherapy due to unresectable tumors or associated comorbidities, and one suﬀered early localized relapse fol-lowing platinum-based chemotherapy and thoracic irradiation. ICI was administered to 20 (27.8%) in first line and to 52 (72.2%) in second line or after. Most patients (n = 59, 81.9%) received nivolumab alone. There were 12 (16.7%) participating in clinical trials assessing ICI in combination with another ICI (n = 8) or with platinum-based che-motherapy (n = 4) in a first-line setting. On ICI initiation, 39 (54.2%) patients had a performance status (PS) of 1 or less. A total of 25 (34.8%) patients harbored KRAS mutations, two (2.8%) harbored EGFR acti-vating mutations, but none had ALK or ROS1 rearrangement. There were 41 (56.9%) harboring TP53 mutations and 31 (43.1%) were TP53 wild-type. Of the 65 (90.3%) patients with available tumor PD-L1 im-munohistochemical data with E1L3N monoclonal antibody, 36 (55.4%) expressed PD-L1 ≥50%, 20 (30.8%) from 1 to 49%, and nine (13.8%) produced negative results.