Clinical validation and implementation of droplet digital PCR for the detection of BRAF mutations from cell-free DNA.

Droplet digital PCR (ddPCR) has been demonstrated in many research studies to be a sensitive method in the analysis of circulating tumour DNA (ctDNA) for identifying mutations and tracking disease. The transition of ddPCR into the diagnostic setting requires a number of critical steps including the assessment of accuracy and precision and ultimately implementation into clinical use. Here we present the clinical validation of ddPCR for the detection of BRAF mutations (V600E and V600K) from plasma. We describe the performance characteristics assessed including the limit of blank, limit of detection, ruggedness, accuracy, precision and the effect of the matrix. Overall, each assay could achieve a limit of detection of 0.5% variant allele fraction and was highly accurate, with 100% concordance of results obtained from routine diagnostic testing of formalin fixed tumour samples or reference controls (n=36 for BRAF V600E and n=30 for BRAF V600K). Inter-laboratory reproducibility across 12 plasma samples for each assay was also assessed and results were 100% concordant. Overall, we report the successful validation and translation of a ddPCR assay into clinical routine practice.

ROS1 rearrangements in non-small cell lung cancer: screening by immunohistochemistry using proportion of cells staining without intensity and excluding cases with MAPK pathway drivers improves test performance.

Therapeutically actionable ROS1 rearrangements have been described in 1-3% of non-small cell lung cancer (NSCLC). Screening for ROS1 rearrangements is recommended to be by immunohistochemistry (IHC), followed by confirmation with fluorescence in situ hybridisation (FISH) or sequencing. However, in practise ROS1 IHC presents difficulties due to conflicting scoring systems, multiple clones and expression in tumours that are wild-type for ROS1. We assessed ROS1 IHC in 285 consecutive cases of NSCLC with non-squamous histology over a nearly 2-year period. IHC was scored with ROS1 clone D4D6 (n=270), clone SP384 (n=275) or both clones (n=260). Results were correlated with ROS1 break-apart FISH (n=67), ALK status (n=194), and sequence data of EGFR (n=178) and other drivers, where possible. ROS1 expression was detected in 161/285 cases (56.5%), including 13/14 ROS1 FISH-positive cases. There was no ROS1 expression in one ROS1 FISH-positive case in which sequencing detected an ALK-EML4 fusion, but not a ROS1 fusion. The other 13 ROS1 FISH-positive cases showed moderate to strong staining with both IHC clones. However, one case with a TPM3-ROS1 fusion would have been scored as negative with SP384 and D4D6 clones by some previous criteria. ROS1 expression was also detected in 58/285 cases (20.4%) that had driver mutations in genes other than ROS1. A sensitivity of 100% for detecting a ROS1 rearrangement by FISH was achieved by omitting intensity from the IHC scoring criteria and expression in >0% cells with D4D6 or in ≥50% cells with SP384. Excluding cases with driver events in any MAPK pathway gene (e.g., in ALK, EGFR, KRAS, BRAF, ERBB2 and MET) substantially reduced the number of cases proceeding to ROS1 FISH. Only 15.9% of MAPK-negative NSCLC would proceed to FISH for an IHC threshold of >0% cells with D4D6, with a specificity of 42.4%. For a threshold of ≥50% cells with SP384, only 18.5% of MAPK-negative cases would proceed to FISH, with a specificity of 31.4%. Based on our data we suggest an algorithm for screening for ROS1 rearrangements in NSCLC in which ROS1 FISH is only performed in cases that have been demonstrated to lack activating mutations in any MAPK pathway gene by comprehensive sequencing and ALK IHC, and show staining at any intensity in ≥50% of cells with clone SP384, or >0% cells with D4D6.