NEW YORK – Researchers are developing a polygenic score to identify children with acute myeloid leukemia who are less likely to respond to cytarabine chemotherapy and could benefit from a higher dose or additional therapies.
The score needs to be prospectively studied before real-world implementation, but efforts are already underway to refine an early version of the score in patients from Africa and other countries.
In a study published recently in the Journal of Clinical Oncology, a team of researchers led by University of Florida's Jatinder Lamba described a genomic score, dubbed the "10-SNP ARA-C_SNP," or ACS10, associated with pediatric AML patients' outcomes on the standard-of-care chemotherapy agent cytarabine.
While newer therapies are now available for pediatric AML, cytarabine remains a backbone treatment that most patients receive before or with these other drugs. But between 10 percent and 15 percent of pediatric AML patients don't achieve complete remission on cytarabine, also called ARA-C, and around 40 percent of those who do go on to experience disease relapse.
Lamba and colleagues attempted to identify this subset of unlikely responders and determine whether they might benefit from additional, or augmented, therapy. On the flip side, a response score that identifies poor responders could also identify those likely to achieve durable remissions and may do well with a lower cytarabine dose.
"I know we need to find new drugs and new targets, but we also need to use the older drugs more smartly and in better combinations," Lamba said, pointing out that cytarabine has been used for nearly 50 years, and while individual dosing varies, those decisions don't have a genetic rationale.
Discovering, validating a score
The recent JCO study is the culmination of nearly two decades of work for Lamba, beginning with research sequencing individual genes associated with the metabolic pathway activating cytarabine. For the drug to have an enduring benefit, she explained, the metabolic pathway genes need to be intact to collectively activate the drug. Lamba sequenced these genes to identify SNPs associated with cytarabine activating pathways and supplemented her findings with information from other databases.
"In a patient, one gene or one SNP is not going to have an impact by itself … it's the whole combination," she said. As such, Lamba and colleagues set out to create a polygenic score using SNPs within 16 genes, which they selected based on Lamba's previous research, published research, and whether each SNP was located in a region important for gene function.
To develop the score, the researchers used data from a St. Jude Children's Research Hospital and National Cancer Institute-sponsored trial of newly diagnosed pediatric AML patients, in which samples were taken 24 hours after starting cytarabine. Patients received cytarabine at various doses across multiple study arms. Lamba's team used univariate analyses to identify the top SNPs associated with patients' treatment outcomes.
They then tested different combinations of SNPs three at a time to see how well they predicted patients' outcomes. "We ran that thousands of times to make sure that particular combinations were picked up every time," she explained. "It was a pretty robust approach."
From there, the researchers picked the top SNPs associated with specific endpoints of interest, including patients' cytarabine triphosphate levels, and their minimal residual disease after chemo and event-free survival outcomes. The researchers then created a genotype score for each SNP based on its mode of inheritance and if it was associated with better or worse outcomes. By adding the score for each SNP, the researchers were able to generate a composite polygenic score for each patient in the study.
If a patient scored less than zero, that indicated they wouldn't be able to efficiently convert cytarabine into its active form and might benefit from a higher dose or additional drugs that work via different pathways, such as the CD33-targeted gemtuzumab ozogamicin (Pfizer's Mylotarg). High scores, on the other hand, indicated patients who could lower toxicity risks with less drug.
Next, researchers set out to retrospectively validate the ARC10 score using data from a Phase III Children's Oncology Group and NCI-sponsored clinical trial of pediatric AML patients treated with cytarabine alone or combined with gemtuzumab ozogamicin.
They found that a greater percentage of patients with ACS10 low scores were alive at five years after receiving cytarabine and gemtuzumab ozogamicin than after treatment with just cytarabine. The same was true in terms of patients with low scores who were alive without disease progression. The researchers saw the same trend for high- versus low-dose cytarabine arms in the ARC10-low-scoring group, indicating that patients with low polygenic scores benefited from combination therapy and from higher-dose chemo.
Patients with high polygenic scores, on the other hand, had better outcomes with just chemo, particularly when given at a lower dose. While "0" worked well as a cutoff for delineating high and low scoring patients and stratifying those likely to benefit from low-dose and high-dose cytarabine, Lamba and colleagues acknowledged that a more refined cutoff may be needed to determine the benefit of adding gemtuzumab ozogamicin to chemo.
Considering race disparities
Among myriad demographic factors that Lamba and her team considered in analyzing pediatric AML patients' characteristics across the low- and high- polygenic score groups, race was the only factor corresponding to a stark difference between the groups.
Specifically, among Black pediatric AML patients, roughly 70 percent had low scores with Lamba's ACS10 polygenic score, whereas, among white patients, the same was only true of 30 percent. "That was really eye-opening," Lamba said. "Because there are racial disparities in AML outcomes … and we don't exactly know how to explain that."
Prior research has shown that Black AML patients have poorer outcomes on treatment, she explained, and these findings with the polygenic score could provide a genetic explanation. Unfortunately, there weren't enough Black patients in this study for this finding to reach statistical significance, but it did get Lamba thinking about how the polygenic score might be used to identify the best treatment strategies for Black patients and narrow race-based outcomes disparities.
To further probe this finding, Lamba is conducting a project with researchers in Africa, including with the consortium H3Africa, African Genomic Studies, and the University of Nigeria. The researchers are creating a consortium, which as of now involves 20 different centers in nine different countries, within which they are working to establish an institutional review board that can oversee the conduct of multiple studies to test out the polygenic score in African patient populations.
Lamba has ambitions to test the score in other populations as well and is in early talks to initiate similar research efforts in Central America and Asia.
Ultimately, she said, her goal is to validate the score globally to determine how and whether it can be used for all AML patient populations. For example, if there is a population of patients with high polygenic scores in a developing country with limited resources, "you could probably treat those patients at a lower cost, and maybe treat more patients, by reducing their [dose of] intensive chemotherapy." The example is hypothetical, but it offers a glimpse of the potential utility of such a polygenic score.
Lamba's commitment to evaluating the polygenic score across global populations is particularly valuable since most polygenic risk scores developed to date have been based on data from genome-wide association studies with a pervasive European ancestry bias that dates back decades.
While Lamba said that she couldn't discount these types of deep-rooted biases playing into the race-based findings she saw with the ACS10 score, she said that much of the early research conducted years ago informing her work today drew on HapMap [the international haplotype project], which includes diverse populations from Nigeria, Beijing, China, Tokyo, Japan, Utah, and the Yoruba in Ibadan, and has since included other populations. Information gleaned from the project's cell lines as opposed to patient DNA. She hopes the HapMap origins make her team's findings less vulnerable to bias, but knows it is far from bias-free.
"We will have to see what this score looks like when we start the research in these African countries," she acknowledged, adding that there may be nuanced differences in West and East African patient populations, too.
Prospective validation, next steps
While Lamba is currently focused on expanding her research to these diverse populations, she and her team also want to evaluate the polygenic score in the context of newer drugs added to cytarabine.
"I'm really excited to see how other agents that have been approved for AML do in patients with low scores for cytarabine," she said, noting the example of venetoclax (AbbVie/Genentech's Venclexta). Of course, for newer agents approved in recent years, it may take time to generate the patient numbers and datasets needed for this type of research.
Additionally, Lamba said that the fact that her findings deal with the germline polymorphisms suggest they could be relevant for adult AML, too, not just pediatric populations, but more research is needed.
The JCO study describes early-stage research on a promising polygenic score, and for this to ultimately guide treatments for AML patients in the real world, the ACS10 score will need to be prospectively studied. Lamba envisions a trial where in one arm patients receive a genotype-guided induction chemo regimen and in another arm patients receive standard of care. A comparison of arms in a study like this could "push this [score] very fast toward standard of care," she said.
From a feasibility standpoint, AML patients already have their bone marrow samples drawn for diagnosis before treatment, and therefore could have those samples tested to determine a polygenic score for guiding treatment.
That said, prospective validation could take years, as could the process of bringing drug companies on board, getting regulatory bodies to update cytarabine package inserts with genotyping information, and encouraging guideline bodies to amend their recommendations. "The biggest challenge is showing [these findings] in multiple independent cohorts," Lamba said. But down the line, if the score winds up being consistently useful and validated across the board, she said moving it forward "will be a no-brainer."