CH Mutations in 16,812 Advanced Cancer Patients–Response

We appreciate the thoughtful analysis included in the letter to the editor from Antonarakis and colleagues about the purported overestimation of BRCA1/2 clonal hematopoiesis (CH) variants and the concern that this inflates the risk of inappropriate administration of PARP inhibitors (PARPi), but we respectfully disagree. In fact, new evidence just published in April 2025 further elucidates the importance of understanding the prevalence of CH in cell-free DNA and its presence in tumor tissue itself. Pich and colleagues report in The New England Journal of Medicine that 42% of patients with NSCLC with CH also had CH present in their tumor tissue, with an increased risk of death or recurrence compared with those patients without CH in the tumor. Across all solid tumors, 26% of those with CH had CH in their tumor, and the tumor CH conferred a risk of death from any cause that was 1.17 times (95% confidence interval, 1.06–1.29) as high as the risk with CHIP in the absence of TI-CH (1). Also in Clinical Cancer Research, Bernard and Micol point out, specifically in reference to our article, that CH alterations in blood not just pose the danger of being false positives for cancer treatment selection but also represent an opportunity to diagnose occult hematologic malignancies and, in the long term, better understand CH-associated risk for cardiovascular events (2).There is significant off-label use of PARPis that Antonarakis and colleagues may not have considered when making their conclusions. The FDA approval for PARPis in breast and pancreatic cancers is for deleterious germline mutations only, and administration to patients with somatic mutations is off-label, though it is on guidelines (3). The letter authors may not be aware of the extent of PARPi prescribing in patients with no pathogenic/likely pathogenic variant. In our database of nearly 500,000 patients with a tissue profile, we observed that 21.7% patients with cancer who received a PARPi only harbor a benign or VUS mutation and do not have a co-occurring pathogenic mutation (ovarian: 23.8%, pancreatic: 19.6%, and breast: 10.5%; manuscript in preparation). Thus, there is a large population of patients who are receiving PARPis who probably should not be, and in this aspect we agree with the letter authors.A second consideration is that our study population largely consists of patients with late-stage metastatic disease where ∼65% of them had received treatment prior to blood being collected and where the landscape of CH variants has not been previously established on a large scale. There is a known increase in CH variants due to treatment with chemotherapy (4), which is consistent with the increased rate we observed in this study. Chemotherapy (including platinum) causes CH variants, and we also know that PARPi responses are higher in platinum-sensitive patients. So, this highlights the importance for this research reporting all CH variants to better understand how patterns of CH variants could potentially impact PARPi sensitivity.As noted in Bernard and colleagues (2), there is no standardly accepted method for CH identification. Sequencing depth influences CH detection, and different studies sequence at different depths. Our very deep sequencing is likely to detect more CH variants than testing at lower depths. In their NEJM study, Pich and colleagues (1) use a variant allele frequency cutoff of 2% in 77 genes for calling CH. We agree with Bernard and Michol that the field would benefit by a standardized CH calling approach.Antonarakis and colleagues correctly stated that certain BRCA1/2 variants included in the study are benign, intronic, and of unknown significance, which we recognize could be further clarified. The publication states in the methods that “Only the pathogenic and likely pathogenic variants were reported,” but this was intended to refer to the physician-facing final report we deliver, not the global makeup of all CH variants that the study was intended to address. These additional variants were retained for a simple reason: A benign, intronic CH variant is still a somatic variant, and therefore the prevalences should not be limited to pathogenic and likely pathogenic alterations. It is of significant importance that these variants are not interpreted in a manner that would lead to inappropriate therapy administration.Table 1 in Magee and colleagues reports that 2.6% of patients harbored a BRCA1 germline mutation compared with 1.1% with a somatic mutation, whereas 5.1% and 1.9% of patients harbored germline and somatic variants in BRCA2, respectively. The supposed overestimation of CH BRCA1/2 variants of 58.6% and 66.5%, respectively, indicates the breakdown of variants of the <2% of patients harboring a somatic mutation at all. We note that although Antonarakis and colleagues focus on BRCA1/2 variants, PARPi selection in prostate cancer is driven by additional genes and the rate of CH-mutated patients with prostate cancer published in Figure 4 of Magee and colleagues (5) is under 6%, compared with 10% established in the prior work, which contradicts claims of overestimated CH rates. Furthermore, the findings of the TRITON2 trial (6) suggest BRCA CH as a possible reason for the difference in response rates observed among BRCA1/2-positive patients selected via tissue or plasma, although they suggested that this was not likely based on the understanding of CH association of BRCA at the time of the publication. Of note, one third (2 of 6) of the patients enrolled via liquid BRCA mutations that also had tissue testing were negative for BRCA in the tissue. Although not statistically significant, the precise mechanism by which objective response rates could diverge starkly between patients enrolled via tissue- versus plasma-only sequencing highlights unresolved questions about BRCA positivity and PARPi response.As far as we are aware, our study represents the largest collection of paired cell-free DNA and gDNA variant data ever published, enabling researchers to more definitively answer such questions. Our view is that dismissal of these results would not be beneficial to the late-stage cancer population or other researchers seeking to establish mechanisms for lack of response to PARPi or other therapies. Although we focus on tumor-derived pathogenic and likely pathogenic mutations in our therapy recommendations, the fact remains that not all labs do that, and many plasma-only liquid profiling labs may be reporting somatic CH mutations, although they are coming from the tumor. Antonarakis and colleagues raise the concern that many of the BRCA CH variants detected also had mutations in the tissue or germ line and thus would not result in inappropriate treatment. We disagree that this is a valid criticism of the work, as we included in the article a thorough analysis of the concurrence of mutations in tissue, germ line, and CH. Figure 4 specifically elucidates this. We also disagree that a clinician would always have tissue or germline results available when considering the results of a liquid biopsy test. The real-world potential for mistreating false positives due to CH is considerable. Based on our work and the subsequently published work on the importance of CH, we believe directly testing for and reporting CH variants is in the best interest of patients.D. Magee reports other support from Caris Life Sciences during the conduct of the study, as well as other support from Caris Life Sciences outside the submitted work. R. Hahn-Lowry reports employment and equity ownership with Caris Life Sciences. G.W. Sledge reports other support from Caris Life Sciences during the conduct of the study. M.J. Oberley reports other support from Caris Life Sciences during the conduct of the study. M. Radovich reports other support from Caris Life Sciences during the conduct of the study. D.B. Spetzler reports employment and equity ownership with Caris Life Sciences.This work was funded by Caris Life Sciences.