Personalized/Tumor-Informed ctDNA and Patient Survival in Pancreatic Cancer

Gregory P. Botta, MD, PhD: In our abstract, “Association of personalized and tumor-informed ctDNA [circulating tumor DNA] with patient survival outcomes in pancreatic adenocarcinoma,” we were trying to identify biomarkers in pancreatic cancer that are predictive of recurrence. We were also trying to characterize an association between circulating tumor DNA positivity before or after surgery as well as with adjuvant therapy, and if there’s an association with recurrence-free survival. Lastly, we wanted to see how that stacked up against the standard of care marker CA 19-9 and see if we have an improved sensitivity or improved prognostic information based on the ctDNA.

Pancreatic cancer CA 19-9 levels can be elevated for a variety of reasons, including biliary obstruction or liver failure. In those cases, the CA 19-9 doesn’t reflect the actual tumor burden in the body. We also know that about 5% to 10% of our patients with pancreatic cancer fail to make CA 19-9. In those cases, we’re left with either radiographic evidence of their recurrence or clinical symptoms. At the end of the day, we wanted to have another tool in our toolbox to help guide decision-making as well as when to decide on whether a treatment was working.

We evaluated 116 patients, and of those 116 patients, we had 450 plasma time points in which we were able to evaluate ctDNA. This was done in a tumor-informed approach. We took the patient’s pancreatic cancer, ran whole-exome sequencing on it, removed the germline mutations, and were left with the pathologic mutations in which we formed a gene signature. That gene signature was followed in the blood over time. The median time from follow-up from the surgery was about a year in all of these patients. We evaluated the recurrence rates for all stages in stage I through stage III, but also in some stage IV metastatic patients. The division between male and female was nonsignificant, at about 50% on each side. Some patients were given neoadjuvant therapy prior to their resection.

Ultimately, we were able to find that in patients with stage I, stage II, and stage III disease, the ctDNA increase—not just that was positive or negative—the amount of mean tumor molecules or ctDNA circulating was correlated with the stage of disease. If we saw that the ctDNA was highly elevated, that usually meant that the patient had a higher stage of disease. In addition, we were able to determine if a patient was either going to relapse before radiographic prediction or if they had a response to adjuvant therapy.

In these cases, we were able to find that after surgery, some patients actually become ctDNA negative. In addition, those patients who didn’t become ctDNA negative after surgery were able to become ctDNA negative after being given adjuvant therapy. Furthermore, the patients who remained ctDNA negative over a longer period of time had an improved relapse-free survival. We found that the ctDNA was able to be more prognostic of survival compared with CA 19-9 with a hazard ratio of about 8.2, meaning that patients whose ctDNA remained negative over a longer period of time had a much improved recurrence-free survival of up to about 60 months.

In addition, as I discussed earlier, can ctDNA predict recurrence significantly earlier than radiographic recurrence? That would make sense. When you’re looking at mean tumor molecules, the sensitivity is very high compared with radiographic evidence of approximately a half of a cm. We found that that was the case, that many times with our patients, we were able to detect that their tumor had recurred. Their ctDNA went from negative to positive before imaging was able to do so.

In those cases, the question that’s ultimately left is, what do we do in those situations? Do we continue to monitor the ctDNA as it rises, preparing the patient for either additional therapy or possibly a clinical trial? Do we initiate therapy immediately, knowing that there’s molecular residual disease and that we will possibly be able to change the patient’s outcome by starting our treatment early? Or is it lead time bias and that it doesn’t matter that the patient has recurred and it doesn’t matter when we determine the patient recurs, because their overall survival is going to remain the same? These are questions that are left to be determined in future prospective studies.

Transcript edited for clarity.