Algorithm may be ‘transformative’ for predicting GVHD after stem cell transplant
Researchers at Mount Sinai Health System developed a two-biomarker model that can predict graft-versus-host disease before signs and symptoms appear in individuals who underwent allogeneic hematopoietic stem cell transplant, according to study results published in The Journal of Clinical Investigation.
Approximately half of transplant recipients develop severe GVHD, and nearly 40% of that subset will die of the condition.
The Mount Sinai Acute GVHD International Consortium (MAGIC) algorithm has the potential to save many lives through early detection and intervention, according to James L.M. Ferrara, MD, chair of cancer medicine and professor in the departments of oncological sciences, medicine and pediatrics at The Tisch Cancer Institute of Icahn School of Medicine at Mount Sinai, and colleagues.
Ferrara and colleagues analyzed blood samples from nearly 1,300 allogeneic HSCT recipients treated at 11 cancer centers. The researchers identified two proteins, ST2 and REG3a, that can predict GVHD mortality weeks before disease onset.
“The [MAGIC] algorithm provides doctors a roadmap to save many lives in the future,” Ferrara said in a press release. “This simple blood test can determine which bone marrow transplant patients are at high risk for a lethal complication before it occurs.”
HemOnc Today spoke with Ferrara about the study results, their potential implications, and what additional research is necessary before the MAGIC algorithm is ready for widespread clinical use.
Question: Can you describe the need for your study?
Answer: The idea for this study was in gestation for quite a while. GVHD is the major complication of allogeneic HSCT, yet there has been no progress on this issue during the past 40 years. All trials have failed, which is hugely frustrating for patients and clinicians. I have worked on this issue for more than 30 years, and I knew a long time ago that we were going to need objective laboratory tests to help define patients at high risk before they became refractory to therapy. Working with animal models on GVHD, I knew donor lymphocytes started reacting to host tissues and were proliferating within hours of the transplant, even though clinical symptoms may not be apparent for weeks. Analyzing the cells in the blood was not going to help because, after transplant, all the white blood cells disappear from peripheral circulation. The lymphocytes that cause GVHD are already buried, trapped and activated in the tissues; they are not circulating. I also knew from animal studies that there were soluble proteins that would be released from these activated lymphocytes, and that they may be able to indicate where the GVHD was developing.
Q : How did your research come about?
A: Fifteen years ago, I received $1 million from the Doris Duke Foundation. With this funding, we set up a database and biorepository through which we collected blood samples from patients. We made detailed notes in our database about their clinical status at the time of the blood sample. We looked for biomarkers and found several. About 5 years ago, the data on these biomarkers were very promising, but all the data were from one center. A big problem in transplantation is that there are ‘strong center effects,’ where what happens in one transplant center is not necessarily what happens in another. There is not a lot of uniformity between centers in terms of how certain problems are being managed. We knew that if these biomarkers were going to be useful, we would need to transition to a multicenter biorepository and database, but this is difficult and expensive. I came to Mount Sinai Hospital in 2014 because they were willing to spend millions of dollars to help us go global and essentially help us create this multicenter database consortium.
Q: How did you conduct the study?
A: We collected data from 20 centers across the United States and Europe. We started working closely with colleagues in Germany who had very similar ideas about what was needed, and we aligned our practices so we could essentially combine our data into a joint database and repository, with the ultimate goal of conducting clinical trials. We now have more than 1,000 patients from 11 centers where both blood samples and clinical outcomes data have been gathered.
Our initial studies, published more than 2 years ago, showed biomarkers could be combined in a model or in an equation where we used the plasma concentrations of biomarkers when GVHD started. We could tell who was at low risk, intermediate risk or high risk for death, who was going to respond to therapy and who was not going to respond. This is the next step: We want to know if we can find patients before GVHD starts.
Our inspiration for this was the work that had been done decades ago in cytomegalovirus. People who acquired this disease would die because they could not predict who was going to get the disease and who was not. Along came the polymerase chain reaction test, which could tell us when the disease was going to multiply and divide before the symptoms of the disease occurred. We wanted to see if our biomarkers could do the same thing for GVHD. We did not use a polymerase chain reaction test, but we had a very sensitive test for some of the biomarkers that could tell us nomogram quantities.
Q: What did you find?
A: On day 7, we were able to identify about 20% of the 1,300 patients who were at very high risk for lethal GVHD. This is the ‘nugget’ of this paper. When we conducted further analysis among the 200 patients who sent us additional samples at the onset of GVHD, we saw that use of two biomarkers with this new test was even more accurate than the three-biomarker algorithm. We now know we can use two biomarkers — not only before GVHD but also at the onset — to predict who is at high risk for severe or lethal GVHD and who is at low risk.
We also found that, when the GVHD is most severe, it is located in the gastrointestinal tract. There are three target organs in the disease: the skin, the gastrointestinal tract and the liver. Our findings suggest we can see subclinical disease in the gastrointestinal tract before clinical symptoms become apparent.
Q: Can you describe the clinical implications of these data?
A: We think our results are very good news, because we are now able to start to develop clinical trials to test new drugs that can stop this reaction. This biomarker is ready for clinical research, but it is not ready for widespread clinical use.
Q: What will future research entail?
A: We are conducting additional research to intervene in patients who are at high risk with a pre-emptive strategy that hopefully will prevent disease — particularly in the gastrointestinal tract, where lethal GVHD eventually happens. We were amazed that this one test could give us a huge difference in long-term outcomes on day 7. We believe we should be able to look weekly for the first 2 or 3 weeks to see whether the GVHD is developing in the gastrointestinal tract — as we do for cytomegalovirus testing — and that we will be able to improve the sensitivity of the test as we move forward. These studies are ongoing.
Another area of research will be looking at how well this test works within specific groups, such as minorities and children. Finally, although this two-biomarker algorithm is excellent, there may be other biomarkers. As they are discovered, we can compare them with those we already have to see if they represent an advance.
Q: Is there anything else you would like to mention?
A: We believe this algorithm can be transformative. These biomarkers represent precision medicine for transplant patients, and we can now target our therapies to those patients who need them the most. – by Jennifer Southall
Ferrara JLM, et al. J Clin Invest. 2017;doi:10.1172/jci.insight.89798.
Levine JE, et al. Lancet Haematol. 2015;doi:10.1016/S2352-3026(14)00035-0.
For more information:
James L.M. Ferrara, MD, can be reached at Icahn School of Medicine at Mount Sinai, 1468 Madison Ave., New York, NY 10029; email: firstname.lastname@example.org.
Disclosure: Ferrara reports grant funding from an American Cancer Society clinical research fellowship and a Doris Duke Charitable Foundation clinical research mentorship. Grants from the NCI supported this study.