Researchers take aim at cancer drugs’ toxic side effects

The patient was a success story, his advanced melanoma erased by a popular new cancer treatment. Known as immune checkpoint inhibitors, the drugs coax the immune system to seek and destroy cancer cells—and in this case, they “worked beautifully,” says Kerry Reynolds, an oncologist at Massachusetts General Hospital (MGH) who helped care for the man.

But about a month after an infusion, without a melanoma cell detectable in his body, the 64-year-old was admitted to the hospital, gravely ill. The drugs were sending his immune system into overdrive, wreaking havoc on his colon and nervous system. Doctors struggled for more than 3 weeks to save him, but “he died of overwhelming toxicity,” Reynolds says. She was haunted by his story. “We felt so hopeless.”

Before he died the man implored Reynolds to learn from his experience, and she promised she would. Soon after, in 2017, Reynolds founded the Severe Immunotherapy Complications Service at MGH, where immunologist and genomicist Alexandra-Chloé Villani took on a parallel research effort; together they aim to treat and study people with immune complications from these breakthrough cancer drugs. The program is now expanding—part of a larger push by scientists around the world. They are launching clinical trials to test treatments for the side effects, turning to computer algorithms to try to predict who’s at risk, and analyzing single cells to parse the biology of these vexing assaults.

Villani, who came to the field after her mother was saved by checkpoint inhibitors but left with arthritis as a consequence, says wider use of checkpoint treatments is making the research more urgent. “We’re curing patients that were incurable a decade ago,” but side effects limit how the drugs can be used.

About 10% of those who get checkpoint inhibitors are hospitalized with immune toxicities. As many as 1% die. A 2021 study suggested that, like Villani’s mother, about 40% of those taking checkpoint drugs develop chronic complications, often arthritis or endocrine dysfunction. “When people have 4 months to live, the risk makes sense,” Reynolds says. For less advanced cancers, “the risk profile changes” and doctors crave more information about who stands to benefit.

Today that tension between the drugs’ risks and benefits is especially acute, because 11 years after the first checkpoint inhibitor was approved in the United States for metastatic melanoma they’re being cleared for earlier stages of several cancers, including melanoma, lung cancer, and breast cancer. More than 40% of cancer patients in the United States are eligible to take the drugs, and they constitute a $30 billion—and growing—market.

The complications superficially resemble known autoimmune diseases, such as hepatitis or colitis, but “the abrupt development is very different,” says Afreen Shariff, an endocrinologist at Duke University. Colon biopsies of patients with drug-induced colitis suggest a mix of overlapping and distinct features compared with biopsies from patients with ulcerative colitis and Crohn’s disease, says Villani, whose lab is studying this biology.

Some side effects are chronic but manageable. Dysfunction in the adrenal gland or thyroid, for example, may be controlled by medicine once a day, says Douglas Johnson, a melanoma oncologist at Vanderbilt University. Others are devastating: The drugs can cause a myocarditis in which the immune system obliterates heart muscle, for example. Although far rarer than many other immune complications, it’s fatal between one-quarter and half the time.

Most patients with immune complications currently receive steroids, a blunt tool that risks interfering with the cancer-directed attack the checkpoint inhibitors are meant to spur—and that don’t always help patients. So, researchers are seeking better countermeasures. In Paris, Sorbonne University cardiologist Joe-Elie Salem has been investigating an arthritis drug called abatacept, which disrupts the activity of T cells, to treat checkpoint-induced myocarditis. Researchers are still trying to determine whether abatacept interferes with checkpoint therapy’s benefits, but in 2019, Salem and colleagues reported in The New England Journal of Medicine that a woman with lung cancer and myocarditis was successfully treated with the arthritis drug without suffering tumor progression. Soon after, Salem was a co-author on another paper that described success with abatacept in a mouse model of checkpoint-induced myocarditis. (The senior author on that paper, published in Cancer Discovery, was James Allison of the MD Anderson Cancer Center, one of two scientists awarded a Nobel Prize in 2018 for cancer immunotherapy, including checkpoint inhibitors.)

Salem has launched a clinical trial to test abatacept in more patients. Another trial, funded by Bristol Myers Squibb, the maker of several checkpoint drugs (and abatacept), this month began to enroll myocarditis patients. The preliminary abatacept results are raising hopes for more targeted treatment, and the possibility that the biology driving immune side effects can be “decoupled” from the drugs’ ability to combat cancer. If so, managing their hazards without blunting their effectiveness would be far easier.

This is a major focus of the work at MGH, where Villani and her lab members are running a slew of tests on blood and tissue samples from more than 300 (and counting) patients affected by immune side effects. The team hopes to learn which cell populations and signaling pathways are behind the complications in different patients. “We do have some early results suggesting we can decouple” the good and bad sides of checkpoint drugs, Villani says, but the picture is complex. “It’s not the same immune component that’s upregulated in every patient, not even every patient with the same toxicity.”

Such immune signatures might offer an early warning of looming problems before the patient’s health spirals downward. Immune complications can take weeks or even months after treatment to manifest, and symptoms alone aren’t always a good indicator: Early signs can be vague and common among people with cancer, such as fatigue, weight loss, and loss of appetite, Shariff says.

To refine predictions of who’s careening toward serious illness, Shariff has developed an algorithm based on electronic health records data from 5000 patients treated at Duke for checkpoint inhibitor complications. People on the drugs get blood tests every 3 weeks, and Shariff has noted patterns that seem to anticipate toxicities, such as abrupt changes in lab results like liver function. The algorithm also accounts for risk factors such as taking a combination of checkpoint inhibitors, a popular strategy that is often more effective against cancer. Lesser risks may include a history of autoimmune disease.

In the next month, Shariff hopes to put the algorithm to its first real-world test in some of Duke’s cancer clinics. She wants to see whether it correctly predicts brewing toxicities and influences how doctors care for patients, perhaps enabling them to prevent hospitalizations by starting immune suppression and other treatments sooner. Initially the algorithm will draw on details from a patient’s medical history and lab results, but she hopes with time to incorporate the kind of molecular detail Villani’s lab is studying.

Pinpointing patients at sky-high risk of immune side effects before they get checkpoint drugs is another frontier. In January researchers reported in Nature Medicine that in blood taken from advanced melanoma patients before treatment, high numbers of a certain type of memory T cell, among other features, can signal an up to eightfold increase in the risk of severe immune-related complications. The team now plans to enroll 75 cancer patients getting checkpoint inhibitors and follow them, to validate this crystal ball. Combined with emerging data on who is most likely to benefit from checkpoint drugs, especially in an earlier stage of disease, the information could guide treatment decisions. But Johnson, who is studying various other markers of immune function to see whether they might anticipate side effects, is cautious. “I’m not so convinced we’re going to find a really good predictive biomarker” that makes forgoing the therapy worthwhile.

Still, with progress on several fronts, “I think that in the next 4 or 5 years, we will have good answers” on how to counsel people about checkpoint therapy, says Jon McDunn, a biomedical engineer in Cary, North Carolina. McDunn is the executive director of Project Data Sphere, a nonprofit that recently worked with MGH and others to help develop definitions of neurologic side effects, and funded a registry to identify affected patients.

In Boston, meanwhile, Reynolds’s program has expanded to 73 doctors and scientists across specialties who meet regularly. Every morning, a smaller subset that includes Villani and some lab members is notified of potential immune complications in MGH’s cancer patients who have agreed to participate in research studies. Funding has been scarce, Reynolds says: “We have done this by Band-Aid and bootstrap.”

About once a month, with permission from the patient before their death and from the family, an autopsy is performed and tissues collected for Villani’s lab. The man with melanoma was the program’s first autopsy, and Reynolds’s promise to him remains fresh in her mind. “We have to get to the bottom of this,” she says.