Concordance of ATM (ataxia telangiectasia mutated) immunohistochemistry between biopsy or metastatic tumor samples and primary tumors in gastric cancer patients.

ATM (ataxia telangiectasia mutated) is one of several DNA repair proteins that are suggested to sensitize tumor cells to the poly(ADP-ribose) polymerase inhibitor olaparib when deficient. The aim of this study was to assess the spatiotemporal concordance of ATM immunohistochemistry (IHC) in gastric cancer in order to determine if measurements made at the level of various sample types and times could be inferred as having the potential to be relevant to treatment decisions made at the patient level. Two independent cohorts composed of 591 gastric cancer patients divided into a gastrectomy cohort (n = 450) and a metastasis cohort (n = 141) were used in this study. A total of 2,705 ATM IHC samples were examined, including 450 whole tissue, 3 sets of 450 tissue microarray (TMA), 301 biopsy, 222 metastatic tumor and 2 additional whole tissue samples of 50 cases from the gastrectomy cohort, and 141 pairs of primary and metastatic tumors from the metastasis cohort. The prevalence of ATM negativity was 13.1% in biopsies, 13.9, 15.1, and 16.0% in TMAs and 15.9% in whole tissue samples of the gastrectomy cohort, and 21.4% in primary tumor and 21.5% in metastatic tumor samples of the metastasis cohort. coefficients were 0.341 for biopsy, 0.572 as the average of 3 TMAs and 0.415 for the largely synchronous metastatic tumors of the gastrectomy cohort, and 0.153 for the largely asynchronous metastatic tumors of the metastasis cohort. Using whole tissue sections from tumor resections or primary tumor, respectively, as the reference standards, specificity and sensitivity were 91.6 and 41.0% for biopsy, 93.9 and 61.9% as the average of 3 TMAs, and 86.6 and 58.8% for metastatic tumors of the gastrectomy cohort and 81.7 and 33.3% for metastatic tumors of the metastasis cohort, respectively. Although we have demonstrated that the IHC assay for ATM was robust and reproducible in gastric tumor samples, we have also found that measurements were subject to significant discordance across multiple sample types from the same patient. Further work will be necessary to determine if classification may be made more consistent by multiple sampling. However, the lack of agreement between primary and asynchronous metastatic samples suggests that such sampling would need to be performed at the time of any treatment decision.