Skip to main content
Premium Trial:

Request an Annual Quote

Model Predicts Blood Cancer Progression Risk for Clonal Hematopoiesis Patients


NEW YORK – A model derived from genetic, laboratory, and outcomes data on 438,890 UK Biobank participants can distinguish patients with clonal hematopoiesis at high risk of progressing to leukemia from those at low risk, according to a recently published study.

Clonal hematopoiesis is an expansion of blood cells derived from a single hematopoietic stem cell typically caused by mutations in leukemia driver genes that can precede malignancy. It's a common occurrence with aging, first described in a 2015 article in Blood by a research group headed by Dana-Farber Cancer Institute hematologist Benjamin Ebert. More than 10 percent of people over age 60 have clonal hematopoiesis, but only about 1 percent convert to malignancy each year. Although some biomarkers have been associated with progression to malignant neoplasm, including some high-risk genes and patterns of co-mutation, to date, effective tools for predicting cancer risk have not been available.

Ebert, Lachelle Weeks, a medical oncologist at Dana-Farber, and other researchers from the cancer institute, Harvard University, and Massachusetts Institute of Technology collaborated on a new study, published Tuesday in NEJM Evidence, detailing their attempt to calculate that risk for patients with clonal hematopoiesis.

"Some of our greatest inroads into decreasing the burden of cancer have been from early diagnosis and interventions that prevent the development of advanced malignancies," Ebert said. "Although we are making progress, the treatment of advanced malignancies is always a major challenge."

Weeks explained that comparable to a colon polyp that develops into colon cancer or a skin lesion that later becomes melanoma, clonal hematopoiesis is essentially a precancerous lesion. "There are individuals walking around who are completely asymptomatic, who have mutations in these genes that cause leukemia, but they don't have any of the bone marrow findings or the blood count findings that would diagnose them with overt leukemia," said Weeks.

However, unlike a mole or a colon polyp, clonal hematopoiesis cannot be surgically removed. That raises the question of whether there's a way to stratify those patients by risk levels for progressing to blood cancer. Weeks' team set out to develop a clinical tool that could be used at the bedside to quantify a patient's risk. Using data from the UK Biobank, they identified more than 11,000 people with clonal hematopoiesis and found genes that could confer a higher risk. Out of that analysis they were able to also determine that a specific mutation in the DNMT3A gene conferred a low risk of cancer progression.

"With many projects of this type, a central challenge is having datasets that are large and well annotated enough to do very robust modeling," said Ebert. "The UK Biobank has been a tremendous resource for this field."

The UK Biobank holds data from 500,000 people, including from whole-exome and whole-genome sequencing, data from electronic health records, and participants' blood samples. Weeks' team developed the underlying algorithm that produces a clonal hematopoiesis risk score (CHRS) for patients using data from 193,743 UK Biobank participants and validated it using data from 245,147 participants. The algorithm is available online.

The researchers found a wide range among UK Biobank participants in terms of their probability of developing cancer over a 10-year period, ranging from 85 percent for one group to less than 1 percent for another.

The researchers also validated their risk model in two independent populations of patients whose blood samples had been sequenced. One was a group of patients diagnosed in hematology clinics at Dana-Farber between 2014 and 2019, who had one of two types of clonal hematopoiesis: clonal hematopoiesis of indeterminate potential (CHIP) or clonal cytopenia of undetermined significance (CCUS). Those patients were followed through December 2021. Another group of 99 patients had bone-marrow biopsy-confirmed CCUS between 2003 and 2019 at the University of Pavia in Italy.

The model predicted the risk of myeloid neoplasm well with a concordance of 0.788 for the Dana-Farber group and 0.727 for the Pavia group where a value of 1 indicates a perfect model and below 0.5 a very poor model.

"We're hoping clinicians will begin to utilize this [score] to calculate the risk of blood cancer for their patients," said Weeks, adding that for over 90 percent of the population the risk of developing a blood cancer is very low, while only about 1.1 percent of patients are at high risk. "That high-risk population warrants being seen in specialized hematology clinics like the Center for Prevention of Progression [at Dana-Farber], whereas the people who are at low risk for developing a blood cancer don't necessarily need that same degree of surveillance."

Weeks said another application of the CHRS is in selecting patients for clinical trials. "This is a completely asymptomatic [condition] normally," said Weeks. "You don't have a lot of room to tolerate risk in somebody who is pretty healthy." But using the CHRS, investigators could identify high-risk patients who could potentially benefit from preventive treatment, since any added risk of adverse events would be justified if that treatment prevented cancer.

Weeks' team is currently developing a clinical trial with funding from the foundation Break Through Cancer to assess whether a hypermethylating agent commonly used to treat myelodysplastic syndromes can be used preventively in a population at high risk for developing myelodysplastic syndromes and acute myeloid leukemia. Although the researchers began designing the trial years ago, before the CHRS tool had been developed, they have revised the study to incorporate the calculator for patient selection.

Weeks' group is interested in exploring ways to predict and prevent other health conditions in patients with clonal hematopoiesis. In addition to having a risk for developing blood cancer, these patients also have nearly double the risk of coronary heart disease. "That's a work in progress," Weeks said, noting that her team is "thinking about how [to] develop other calculators" and design interventions for patients with increased cardiovascular risk, including more aggressive cholesterol control, blood pressure medications, or lifestyle interventions.

For their next steps, Ebert and Weeks plan to validate their model in more diverse patient cohorts. "It is important to note that the population that we used to develop the CHRS was not a population of people who are actively receiving chemotherapy, and that is a large fraction of the patients with clonal hematopoiesis that we see in the clinic," said Weeks. "Making sure the model holds in this setting is essential."

Weeks and her colleagues have already begun using the score at Dana-Farber to guide conversations with patients. She anticipates that as the model is validated in additional cohorts and proves to be generalizable, there will be wider clinical uptake and adoption of CHRS into clinical management guidelines. "For companies that are creating commercial sequencing assays and routinely detecting clonal hematopoiesis, I hope the CHRS can be useful in tailoring reporting to include commentary on the risk of blood cancer progression," Weeks said.