Breast, Colorectal, and Pancreatic Cancer Mortality With Pathogenic Variants in <i>ATM</i> , <i>CHEK2</i> , or <i>PALB2</i>
Christine M. Veenstra, Paul Abrahamse, Ann S. Hamilton, Kevin C. Ward, Scarlett Lin Gomez, Lihua Liu, Steven J. Katz, Timothy P. Hofer, Allison W. Kurian
Abstract
PURPOSE Oncologists encounter patients with pathogenic variants (PVs) in ATM , CHEK2 , or PALB2 , but little is known about their cancer mortality. METHODS Patients who were 20 years or older, diagnosed in 2013-2019 with breast, colorectal, or pancreatic cancer, and reported to SEER registries in California and Georgia were linked to germline genetic testing results from four clinical laboratories and followed through 2021. Multivariable models of cancer mortality were fit; for each cancer, the reference group was the average hazard across all genetically tested patients with that diagnosis. Each cancer was modeled separately, followed by a single model that interacted the cancer type with all covariates. In addition to fixed effects models, random effects models were used as a regularization approach to reduce overfitting. RESULTS A total of 70,272 tested patients with breast (48,473 estrogen receptor–/progesterone receptor–positive, human epidermal growth factor receptor 2 (HER2)–negative; 9,957 HER2-positive; 11,842 triple-negative) cancer, 5,822 with colorectal cancer, and 1,861 with pancreatic cancer were analyzed; the mean follow-up was 3.9 years. Patients with ATM , CHEK2 , or PALB2 PVs had no differences in breast, colorectal, or pancreatic cancer mortality. Patients with ATM PVs in triple-negative breast cancer appeared to have higher mortality in fixed effects models (hazard ratio [HR], 3.7 [95% CI, 1.8 to 7.8]), but not in random effects models (HR, 1.2 [95% CI, 0.8 to 1.6]) that reduce overfitting. Patients with BRCA1 / 2 PVs had lower triple-negative breast cancer mortality in both models (fixed HR, 0.6 [95% CI, 0.5 to 0.9], random HR, 0.7 [95% CI, 0.6 to 0.8]). Patients with Lynch syndrome gene PVs had lower colorectal cancer mortality in both models (fixed HR, 0.5 [95% CI, 0.4 to 0.8], random HR, 0.7 [95% CI, 0.5 to 0.9]). CONCLUSION Patients with ATM , CHEK2 , or PALB2 PVs had similar breast, colorectal, and pancreatic cancer mortality to the average genetically tested patient with their cancer type.