While immunotherapy has revolutionized the way in which oncologists treat cancer, its use comes with a variety of side effects that are very different from those seen with other systemic therapies.
These treatments are "revving up the immune system," said Bryan J. Schneider, MD, of the University of Michigan in Ann Arbor, who co-chaired the group that updated the American Society of Clinical Oncology (ASCO) guidelines for the management of immune-related adverse events (irAEs). "Sometimes the immune system becomes so hyperactive that it starts attacking healthy organs it shouldn't. If it attacks the colon you get colitis, if it attacks the liver you get hepatitis, if it attacks the lungs you get pneumonitis."
Most patients don't get them, he added, but for the minority of those who have severe side effects, they can be "potentially life-altering and even fatal."
Different Rules
These irAEs are different from AEs seen with other systemic therapies and that "has really been a struggle for us as medical oncologists," Schneider said.
For example, with cytotoxic chemotherapy, oncologists know what drugs will cause patients to lose their hair, have a higher risk of nausea and vomiting, or experience drops in blood counts.
"Almost uniformly, stopping the drug will give patients time for those toxicities to resolve," he said. "But those rules just don't apply to checkpoint inhibitors because this is driven by the patient's immune system. So, unfortunately, particularly for higher toxicities, just stopping the checkpoint inhibitor is unlikely to reverse the toxicity."
As reported in the ASCO guidelines, irAEs can involve any organ or system of the body, with dermatologic, gastrointestinal, hepatic, endocrine, and pulmonary toxicities being the most prevalent.
For example, cutaneous toxicities -- including rash, pruritus, and vitiligo -- have been reported in up to 70% of patients treated with immune checkpoint inhibitors. Cardiovascular complications are rare, but can result in devastating clinical consequences, with mortality rates ranging from 35% to 50%.
Who Is at Risk?
According to Schneider, patients who receive dual immune checkpoint therapy have a much higher risk of immune checkpoint toxicity, "so they really need to be monitored much more carefully."
He said there are also higher risks for certain subpopulations, including patients with underlying lung disease who may be at an incrementally higher risk for pneumonitis, and patients receiving CTLA-4 inhibitors, who may have a higher risk of hypophysitis compared with those receiving PD-1/PD-L1 inhibitors.
"And then anyone with pre-existing autoimmune disease like rheumatic syndromes or inflammatory bowel disease always has a significant risk of flare," Schneider said. "So, although those populations aren't excluded from getting treatment, clinicians are going to have to be very mindful and make sure there is open communication with the treating subspecialists because of that risk of flare."
Finally, Schneider noted that oncologists worry about triple-M -- myocarditis, myositis, myasthenia gravis -- overlap syndrome, which is an extremely rare and dangerous complication of immune checkpoint inhibitors.
"If any patient presents with one of those, they are at very high risk of having the other ones, and those should be ruled out as well, because those toxicities can oftentimes be life-threatening and potentially catastrophic," he said.
These irAEs can vary in terms of onset. "Most of these toxicities occur within the first 3 months of initiating therapy, which is again why we like to keep a close eye on patients," he added, pointing out that it's possible that these toxicities can develop much later than that.
"We've seen pneumonitis occur 18 months after starting treatment, and there are even reports of these toxicities occurring 6 months after stopping immunotherapy," he said.
Managing irAEs
Oncologists will typically be the first physicians to identify these toxicities, "assuming the patient doesn't present to the emergency department," Schneider said.
For most non-endocrine toxicities, stopping immune checkpoint inhibitor therapy and treating the AEs with steroids will be successful. "I would say 90% of these cases will resolve with a steroid, and oncologists are comfortable with steroid management," he said. "That typically doesn't apply to endocrine toxicities -- including hypophysitis, or hypothyroidism. Typically, steroids don't work and hormone replacement is required."
If the toxicity doesn't respond to steroids, subspecialists should be involved in order to guide alternative treatments that may include non-steroid immunosuppressants, which these specialists may be more familiar with, Schneider said. "Our gastroenterologists may recommend vedolizumab [Entyvio], our neurologists may use plasmapheresis or IVIG [intravenous immunoglobulin], and for our rheumatology colleagues, there are many DMARDs [disease-modifying anti-rheumatic drugs] to manage inflammatory arthritis that oncologists may not have experience or comfort using."
"But the toxicities need to be managed quickly," he said. "My personal bias is to start steroids in a reasonable time and then get subspecialists involved early if there isn't significant improvement."
At the same time, Schneider said that research efforts are needed to move the field forward with treatments that can eliminate, or at least reduce, the use of steroids.
"Although they typically work, they do come at the cost of other side effects," he said. For instance, Schneider pointed out that in the case of refractory pneumonitis cases, prolonged steroid use can lead to infectious complications or accelerated osteoporosis -- "things that can be life-threatening or debilitating. Research that could help manage these with steroid-sparing agents would, I think, bring a lot of value."
Learning About irAEs
Questions still remain about the characteristics and outcomes of severe irAEs, and how they may affect the overall risk-benefit relationship of immunotherapy, said Kerry Reynolds, MD, of Massachusetts General Hospital and Harvard Medical School in Boston.
Using myocarditis as an example, Reynolds pointed out that it has a high mortality rate, and that "we have a lot to learn about this condition."
"At our own institution, we've had a couple of hundred cases, but when we think about what we're really going to learn from all case presentations, we have to pull cases together," she said. "We really don't have enough cases to glean what we need to on how these people present, on different trends relating to the prognosis of the condition, or how treatments work."
Thus, Reynolds said, it can be challenging to understand and manage these irAEs in real-world settings.
In a , Reynolds and colleagues explained how Mass General Brigham set up a severe immunotherapy complication service to track all patients admitted with suspected immunotoxicities. However, according to Reynolds, that kind of operation, which is highly resource-intensive, "is not sustainable and not scaleable."
Thus, she and her colleagues incorporated the use of a pre-built large language model (LLM) to identify information relating to instances of irAEs in the electronic health record, and tested it against the use of ICD codes in a dataset of hospital admissions individually reviewed and manually adjudicated for the presence of irAEs.
The LLM demonstrated higher sensitivity than ICD codes (94.7% vs 68.7%); achieved significance for immune checkpoint inhibitor hepatitis (P<0.001), myocarditis (P <0.001), and pneumonitis (P=0.003); and picked up additional cases of irAEs not picked up by manual adjudication, and all at only 9.53 seconds per chart.
"You can see the sensitivity is amazing," Reynolds said. "It was tested on a gold standard. And the tool performs very well at kicking out cases [that aren't irAEs] and, if it says yes, you have the human review, which is critically important."
As a free and open-source model, Reynolds and her colleagues said the LLM could be used by other institutions to enhance data sharing and collaborative discovery across institutions.
"We could link real-world datasets so that we could get what we need to inform guidelines and how we should treat patients," she explained.
Disclosures
Schneider reported receiving research funding from Merck.
Reynolds reported relationships with Project Data Sphere, Biogen, CME Outfitters, Medscape, SAGA Diagnostics, and Bristol Myers Squibb.