104PFirst national external quality assessment for the interpretation of somatic variants: Assessment of 25 variants in colorectal, lung, ovarian cancers and melanoma in France

Abstract Background The interpretation of testing results in molecular pathology is an essential phase. It is difficult to assesss this in external quality assessment on real samples. Here we propose a national measure of the concordance in the variant interpretation between laboratories participating in the national French quality control program Gen&tiss. Methods Laboratories were asked to interpret 5 variants each for melanoma, lung and colorectal cancers and 10 for ovarian cancer. These variants in KRAS, NRAS, BRAF, PIK3CA, PTEN, AKT1, TP53, EGFR, RAC1, KIT, BRCA1 and BRCA2 were not classical hotspots of mutation and 15 of these were annotated in OncoKB. A maximum of 25 variants was assessed per laboratories. Nine possible interpretations were proposed: from neutral to oncogenic and from no therapeutic impact to targeted by a drug. For each variant, a consensus score was determined based on public database as OncoKb and validated by 4 molecular experts with literature review. Three scores were possible: right answer (2 points), acceptable answer without clinical impact (1 point), wrong answer with clinical impact (0 points). Results 60 laboratories participated to the program for variant interpretation. The average scores on 10 were 5.9 for colorectal (N=52), 7.0 for lung (N=49), 6.0 for melanoma (N=49) and 7.0 for ovarian cancer (N=25). For BRCA genes in ovarian cancer, there was a pilot project in 2017, in which 5 identical variants were proposed to interpretation. 18 laboratories participated both in 2017 and 2018: 50% performed better, 28% performed worse, 17% had similar results. Conclusions As test results have will have a direct impact on patient management, ensuring a correct interpretation becomes more and more important. This EQA could help to validate the interpretation process used in the laboratories (SOP and databases) and could be extend to the individual expertise.The average score is acceptable, yet improvable and it really depends on the variants proposed. International consensus guidelines equivalent to the ACMG classification of constitutional variants would be helpful to improve reproducibility and inter-laboratory homogeneity. Legal entity responsible for the study Gen&tiss - GFCO and AFAQAP. Funding AstraZeneca. Disclosure All authors have declared no conflicts of interest.