Prognostic Image Biomarkers in the Treatment of Patients with Locally Advanced NSCLC

Abstract

<h3>Purpose/Objective(s)</h3> Image biomarkers on planning PET/CT-scans are potential prognostic tools in patient-specific predisposition to first site of failure in the treatment of locally advanced NSCLC-patients (LA-NSCLC) with curative intended chemo-/radiotherapy (cCRT). <h3>Materials/Methods</h3> Patients treated with cCRT for LA-NSCLC at a single institution from 2012-2018 were retrospectively included (N=197). Planning PET/CT scans were analyzed (commercially available software) with respect to selected image biomarkers (volume and sphericity of GTV-T on CT and SUV_peak and percentage of GTV-T volume with PET-signal above 50% of SUV_peak on PET). Clinical baseline characteristics (gender, stage, histology, performance status (PS)) were collected. Progression was scored as either loco-regional (LR), distant metastasis (M), simultaneous loco-regional and distant (LR+M) or death with no evidence of disease (DNED). Data were analyzed with a Fine and Gray competing risk analysis and variables included are shown in table 1. For each failure mode, subdistributed hazard ratios (sHR) with 95%-confidence intervals are reported. <h3>Results</h3> Median follow-up-time was 24 months. Significant predictors of LR failure were histology (squamous cell carcinoma (SCC) vs. adenocarcinoma (AC), sHR=2.05 [1.02-4.13], p=0.045) and SUV_peak (sHR=1.079 pr. SUV-increase [1.01-1.16], p=0.03). Histology remained a significant predictor for M failure (sHR=0.195 [0.07-0.52], p<0.01). No significant predictive parameters for LR+M-failure or DNED were found. Comparing cumulative incidences between the four groups (divided by histology and SUV_peak above/below median) using Fine-Gray test showed significant differences between the four groups in terms of LR-failure (P<0.01), M-failure (p<0.01) and LR+M-failure (p<0.01). LR-failures 2 years after treatment were similar for SCC regardless of SUV_peak (37% vs. 32%). Meanwhile, AC with SUV_peak above median had a significantly higher risk of LR-failure after 2 years than AC with SUV_peak below median (27% vs. 8%). M-failure after 2-years varied with histology but not with SUV_peak (AC: 35% vs. 28%, SCC: 9% vs. 5%). <h3>Conclusion</h3> In a competing risk analysis of 197 LA-NSCLC patients treated with cCRT, histology and SUV_peak prior to cCRT were identified as significant predictors of LR-failure. Patients with AC displayed a lower risk of LR-failure if SUV_peak was below median, which separated this group from the remaining patients. In general, AC were significantly more prone to M-failure than SCC.