Reliable entity subtyping is paramount for therapy stratification in lung cancer. Morphological evaluation remains the basis for entity subtyping and directs the application of additional methods such as immunohistochemistry (IHC). The decision of whether to perform IHC for subtyping is subjective, and access to IHC is not available worldwide. Thus, the application of additional methods to support morphological entity subtyping is desirable. Therefore, the ability of convolutional neuronal networks (CNNs) to classify the most common lung cancer subtypes, pulmonary adenocarcinoma (ADC), pulmonary squamous cell carcinoma (SqCC), and small-cell lung cancer (SCLC), was evaluated. A cohort of 80 ADC, 80 SqCC, 80 SCLC, and 30 skeletal muscle specimens was assembled; slides were scanned; tumor areas were annotated; image patches were extracted; and cases were randomly assigned to a training, validation or test set. Multiple CNN architectures (VGG16, InceptionV3, and InceptionResNetV2) were trained and optimized to classify the four entities. A quality control (QC) metric was established. An optimized InceptionV3 CNN architecture yielded the highest classification accuracy and was used for the classification of the test set. Image patch and patient-based CNN classification results were 95% and 100% in the test set after the application of strict QC. Misclassified cases mainly included ADC and SqCC. The QC metric identified cases that needed further IHC for definite entity subtyping. The study highlights the potential and limitations of CNN image classification models for tumor differentiation.

An increasing number of treatment-determining biomarkers has been identified in non-small cell lung cancer (NSCLC) and molecular testing is recommended to enable optimal individualized treatment. However, data on implementation of these recommendations in the "real-world" setting are scarce. This study presents comprehensive details on the frequency, methodology and results of biomarker testing of advanced NSCLC in Germany. This analysis included 3,717 patients with advanced NSCLC (2,921 non-squamous; 796 squamous), recruited into the CRISP registry at start of systemic therapy by 150 German sites between December 2015 and June 2019. Evaluated were the molecular biomarkers EGFR, ALK, ROS1, BRAF, KRAS, MET, TP53, RET, HER2, as well as expression of PD-L1. In total, 90.5 % of the patients were tested for biomarkers. Testing rates were 92.2 % (non-squamous), 70.7 % (squamous) and increased from 83.2 % in 2015/16 to 94.2% in 2019. Overall testing rates for EGFR, ALK, ROS1, and BRAF were 72.5 %, 74.5 %, 66.1 %, and 53.0 %, respectively (non-squamous). Testing rates for PD-L1 expression were 64.5 % (non-squamous), and 58.5 % (squamous). The most common testing methods were immunohistochemistry (68.5 % non-squamous, 58.3 % squamous), and next-generation sequencing (38.7 % non-squamous, 14.4 % squamous). Reasons for not testing were insufficient tumor material or lack of guideline recommendations (squamous). No alteration was found in 37.8 % (non-squamous), and 57.9 % (squamous), respectively. Most common alterations in non-squamous tumors (all patients/all patients tested for the respective biomarker): KRAS (17.3 %/39.2 %), TP53 (14.1 %/51.4 %), and EGFR (11.0 %/15.1 %); in squamous tumors: TP53 (7.0 %/69.1 %), MET (1.5 %/11.1 %), and EGFR (1.1 %/4.4 %). Median PFS (non-squamous) was 8.7 months (95 % CI 7.4-10.4) with druggable EGFR mutation, and 8.0 months (95 % CI 3.9-9.2) with druggable ALK alterations. Testing rates in Germany are high nationwide and acceptable in international comparison, but still leave out a significant portion of patients, who could potentially benefit. Thus, specific measures are needed to increase implementation.