Tumour cell budding and spread through air spaces in squamous cell carcinoma of the lung – Determination and validation of optimal prognostic cut-offs

Prognostic stratification of patients with squamous cell carcinomas of the lung (SCC-L) is challenging. Therefore, we investigated several histomorphological parameters (tumour cell budding (TCB), spread through air spaces (STAS), tumour-stroma-ratio, immune cell infiltration) which could potentially serve as prognostic parameters in SCC-L. We aimed to systematically determine optimal cut-off-values and assess the prognostic capability of these patterns. We furthermore assessed interobserver variability (IOV) for prognostically significant patterns TCB and STAS.The Cancer Genome Atlas (TCGA) study cohort consisted of 335 patients with SCC-L. Histomorphological parameters analysed comprised TCB, minimal cell nest size (MCNS), STAS, stroma content and immune cell infiltration. The most significant cut-off-values were determined and univariate and multivariate survival outcomes were estimated. The identified cut-off-points were validated in an independent SCC-L cohort (n = 346 patients). Two experienced pathologists probed IOV in the validation cohort.In the TCGA study cohort, TCB, STAS and immune cell infiltration were identified as significant prognostic parameters. TCB-high tumours, a high number of STAS foci, extensive STAS for distance of STAS in alveoli and a low immune cell infiltration remained as independent prognostic factors in multivariate Cox proportional hazard analyses for overall survival (OS). The significance of TCB, number of STAS foci and distance of STAS in alveoli for OS could be validated in the validation cohort. IOV reached a Kappa ≥ 0.89 for prognostic parameters.We determined optimal cut-offs and identified TCB and STAS (number of STAS foci, distance of STAS in alveoli) as independent and uncorrelated prognostic factors for patients with SCC-L. The significance was validated in a large independent cohort. IOV was almost perfect for prognostic parameters. We propose the application of TCB- and STAS-based grading in SCC-L as prognostic morphological classifiers.