Session Item

Combined chemo-radiotherapy (cRT) is standard of care for patients with locally advanced non-small cell lung cancer (LA-NSCLC). This study evaluates the early radiologic and metabolic tumour response after chemotherapy and its prognostic value in predicting pattern of failure for LA-NSCLC-patients.

Patients with LA-NSCLC treated with curative intended cRT (2012-2019) at a single institution were retrospectively reviewed (n=188). All patients had diagnostic PET/CT-(dPC) and planning PET/CT (pPC)-scans, between which they received platinum-based chemotherapy. Volume, sphericity and SUV_peak for the gross tumour volume (GTV-T) were investigated on dPC and pPC. First failure was characterized as either loco-regional (LR), distant metastasis (M), simultaneous (LR+M) or death with no evidence of disease. Two multivariate competing risk analyses (Fine-Gray model) were performed, including patient characteristics only (Model 1) and patient characteristics, baseline and changes of image features (Model 2). The variables included are given in Fig. 1. The models are described with sub-distribution hazard ratios (SHR) with 95%-Confidence intervals for each failure mode. Cumulative incidences were plotted for discrete subgroups with the strongest effects in multivariable modelling and a Fine and Gray test was used to test for difference in each of the outcomes. 

Median follow-up for patients alive was 33 months. Baseline image features on dPC as well as changes between dPC and pPC in response to chemotherapy varied greatly. Median decrease in GTV-T volume and SUV_peak were 19.1% and 32.7%, respectively (fig. 1). In Model 1, histology was the only significant prognostic factor. Squamous cell carcinomas (SCC) presented a significantly lower risk of M failure (SHR=0.246 [0.0887-0.684], p<0.01), and higher risk of LR failure (SHR=2.15 [1.09-4.21], p=0.026) compared to adenocarcinomas (AC). In Model 2, SUV_peak at diagnosis was the only significant predictor of LR failure (SHR=1.07 [1.02-1.13], p<0.01), while histology was still the significant predictor of M failure (SHR=0.259 [0.0970-0.694], p<0.01, SCC). Analysis on four subgroups (histology, SUV_peak above and below median) by Fine and Gray tests showed significant difference for LR (p<0.01), M (p<0.01) and LR+M failure (p=0.048) between the four groups. Cumulative incidence plots for the four subgroups (fig. 2) illustrate that the incidence of M failure varies with histology, but not with SUV_peak. For LR failure, ACs with high SUV_peak are notably more prone to failure than ACs with low SUV_peak. For SCC, the incidence of LR failure is independent of SUV_peak.

Histology and tumour SUV_peak at diagnosis were significant predictors of LR failure in a multivariate model based on 188 LA-NSCLC patients treated with cRT. AC patients with high SUV_peak had a higher risk of LR failure than those with low SUV_peak, while the risk of M failure depended solely on histology.