How often it happens depends on how hard you look
Most published estimates come from patients who were tested after something had already gone wrong. Test everyone, and the number settles down.
At Michigan Medicine, 2,606 patients receiving checkpoint therapy underwent systematic serial blood testing, and 27 of them, 1.0%, were diagnosed with myocarditis (Vasbinder, JACC CardioOncol 2022). Estimates drawn from trial safety reporting run several times lower, because trials of that era were not screening for it and caught mainly the patients who became severely ill.
The same study found the disease leaving traces in tests nobody was reading for it. Creatine phosphokinase, a cheap and routine measure of muscle injury, was raised in 89% of the patients with myocarditis, was independently associated with developing it and with dying, and was 99% sensitive at a specificity of 23% (Vasbinder, JACC CardioOncol 2022). On its own that is a poor test. As a prompt to look properly, in a disease this easy to miss, it is a useful one, and it is already being drawn. It works because this is not purely a disease of the heart: the same process attacks skeletal muscle, which is where the enzyme comes from. The most useful pattern, though, was not any single result but how many were abnormal at once: 95% of the patients with myocarditis had at least three routine biomarkers elevated, against 5% of those without (Vasbinder, JACC CardioOncol 2022).
Which troponin your hospital runs changes the answer
Cardiac troponin is the test of record for injury to heart muscle. It comes in two forms that are treated as interchangeable, and in this disease they are not.
In 60 patients managed at cardio-oncology units in Paris and Heidelberg, with more than 1,700 troponin T and 900 troponin I measurements between them, troponin T was raised on admission in 93% and troponin I in 64%; the same gap appeared again in 87 independent cases from an international registry (Lehmann, Circulation 2023).
The gap matters most at the moment it would do harm. Within 72 hours of a patient's first major cardiac event, troponin T was raised in all 23 of them. Troponin I was still inside its normal range in 11%, and creatine kinase in 27% (Lehmann, Circulation 2023). A patient with a reassuring troponin I can be having the disease.
Troponin T also grades it. The highest ratio of troponin T to its upper reference limit in the first 72 hours predicted a major event within 90 days with an area under the curve of 0.84, and a ratio at or above 32 carried an eleven-fold hazard. Below 32 identifies a group at genuinely low risk (Lehmann, Circulation 2023).
It can look exactly like a heart attack
The commonest way to miss this diagnosis is not to overlook it. It is to make a different one that fits the presentation just as well.
Chest pain, a rising troponin and electrocardiographic changes in a patient in their seventies describe an acute coronary syndrome, and most of the time that is what they are. Endomyocardial biopsy remains the reference test for myocarditis, but the diagnosis in practice is assembled from clinical suspicion, biomarkers and imaging together, and suspicion is the part that has to come first, because the treatment that follows is difficult enough without also being late (Lehmann, JAMA Cardiol 2021).
The cost of getting it wrong is measurable. Among 261 patients with suspected checkpoint myocarditis who underwent coronary angiography, 22.6% turned out to have significant coronary disease, and a third of those were revascularized. Those patients were less likely to have received steroids within 24 hours of admission, and their myocarditis-related 90-day mortality was 52.7%, against about 16% both in those with coronary disease who were not revascularized and in those with none (Nowatzke, Eur J Cancer 2022).
That comparison cannot separate the delay from the fact that patients with two diseases are sicker than patients with one. The practical point stands: finding a stenosis does not exclude myocarditis, and in a patient on checkpoint therapy the two should be considered together rather than in sequence.
Who is susceptible, before the first dose
Most risk in oncology is discovered after treatment starts. This one can be seen beforehand, and it points at a mechanism rather than merely marking a group.
The thymus is where developing T cells are taught not to attack the body's own tissue, so a diseased thymus is the first place to look for a failure of that tolerance. Patients with thymic epithelial tumors, thymoma in particular, developed checkpoint-related heart and muscle toxicity more often than patients with other cancers, developed it sooner, and did worse when they did (Fenioux, Nat Med 2023).
The effect is large enough to survive a multivariable model in the largest series available: among 748 patients admitted with checkpoint myocarditis across 17 countries, active thymoma carried a hazard ratio of 3.6 for a severe cardiac or muscular event (Power, Eur Heart J 2026). Combining two checkpoint drugs raises risk as well, which is the other thing known before the first dose is given.
Predicting the course once it is found
The treatment question is not whether to suppress the immune response but how hard. That turns on being able to tell, on the day of admission, who is going to deteriorate.
In those 748 patients, five things available at presentation predicted a severe event: active thymoma, cardiac or muscular symptoms, low voltage on the electrocardiogram, an ejection fraction under 50%, and the size of the troponin rise, which carried a hazard of 1.8 at twenty times the upper reference limit and 4.6 at two thousand times (Power, Eur Heart J 2026). The electrocardiogram earns its place here: conduction disease, a widened QRS, a prolonged QTc and a fall in voltage against a patient's own earlier tracing all mark the illness (Power, Circulation 2021).
Ejection fraction is worth a closer look, because a third of these patients arrive with it already reduced. Of 757 patients in the international registry, 35% had an ejection fraction below 50% on admission, and those patients were more likely to have had chest radiation in the past, 24.2% against 13.5% (Chen, JACC CardioOncol 2025). A heart that has been irradiated is not a fresh starting point for an autoimmune attack on the myocardium.
Combined into a score, those five separate a 30-day event rate of 4% at a score of zero from 81% at four or more (Power, Eur Heart J 2026). It has been validated in independent French and United States cohorts, and used prospectively in the French one to identify low-risk patients who were then managed without immunosuppression, none of whom had a cardiac or muscular event. A score that only sorts patients is an academic exercise; one that lets a physician safely withhold high-dose steroids from someone who does not need them changes what happens to that patient.
Treating it is the hard part
Finding it early matters because what follows is difficult. The regimen is assembled from experience rather than from evidence, every option costs something, and the disease is not confined to the heart.
Corticosteroids are the first move, and for a substantial minority they are not enough. What follows has been assembled from experience: abatacept, which blocks the same molecule one class of these drugs was designed to block, with its dose titrated against how much of that receptor it actually occupies, and ruxolitinib alongside it (Salem, Cancer Discov 2023). None of it has been through a randomized trial; the first is under way, a dose-finding study with seven patients per arm (Salem, Arch Cardiovasc Dis 2025).
The other half of the treatment is not a drug. The same autoimmune process attacks skeletal muscle, and when it reaches the muscles of breathing a patient can die of respiratory failure with a heart that is recovering. Screening for that deliberately, and ventilating early when it is found, was half of the strategy that brought fatality down (Salem, Cancer Discov 2023). It is why the field increasingly calls this cardiomyotoxicity rather than myocarditis.
The dilemma underneath all of it is that every one of these drugs suppresses the immune response that is treating the cancer. The checkpoint inhibitor was given because the tumor needed it. Turning that response off to save the heart is a trade with no fixed exchange rate, and it has to be made within days.
Knowing when to stop is as unsettled as knowing what to start. Withdrawing immunosuppression too quickly risks the myocarditis returning; too slowly risks infection and leaves the cancer unopposed. Tracking resolution with troponin and the non-cardiac biomarkers rather than by the calendar is the part of this the lab has worked on (Vasbinder, Curr Cardiol Rep 2023). It is the same argument as the risk score, applied in the other direction: match the intensity of treatment to the severity of the disease, on the way down as well as on the way up.
What it costs, and why that is changing
The disease earned a reputation for killing about half the people who developed it. It kills far fewer now, and the reason is not that it became milder.
Among the 748 registry patients, 33% had a severe cardiac or muscular event within 30 days, 13% died of that toxicity and 17% died of any cause (Power, Eur Heart J 2026). That is already well below the figures still widely quoted, which come from the first case series and pharmacovigilance reports and by construction collected the patients sick enough to be noticed and reported (Mahmood, J Am Coll Cardiol 2018; Salem, Lancet Oncol 2018).
The clearest evidence that recognition and treatment are doing the work comes from a single center that changed its practice partway through a series. Of 40 patients with definite myocarditis, the first 10 were managed by the guidance of the day, and 60% died of the myotoxicity, in line with historical controls. The next 30 were screened systematically for respiratory muscle involvement, ventilated early where it was found, and treated with abatacept and ruxolitinib. One of the 30 died (Salem, Cancer Discov 2023).
A before-and-after comparison at one center is not a trial, and the later patients may also have been caught earlier as awareness grew. But a difference of that size is not explained by drift in who gets counted, and every part of it was something a clinician decided to do. That is the case for the assay, the routine biomarkers and the score above: none of them treat anybody, and all of them buy the time in which treatment still works.
Where this goes next: the arteries
Myocarditis is the dramatic toxicity and the rare one. The larger cost may be slower, quieter and vascular, and that is the question the program is taking up next.
We followed 30,720 propensity-matched patients starting either checkpoint therapy or chemotherapy in a federated health record network. Over two years, checkpoint therapy carried a 40% higher risk of admission to hospital for a major cardiovascular event, consistent across heart failure, myocardial infarction and stroke, with the two-year cumulative incidence reaching 16.2% on one agent (Vosooghi, JCO Oncol Pract 2026). The control cohort was treated in an earlier era, so background cardiovascular care differs between the groups and the excess cannot be attributed to the drugs alone. We report it as a signal worth a prospective study.
There is a mechanism waiting for it. The checkpoints these drugs block are themselves brakes on the inflammation that builds an atherosclerotic plaque, so releasing them systemically has an obvious place to act. Others have already found a three-fold higher rate of atherosclerotic events after starting a checkpoint inhibitor, with plaque progressing faster on imaging over the same period (Drobni, Circulation 2020).
Whether that is happening in the patients we follow, how early it can be detected, and whether it can be modified, is where this work goes next. It is also where this program meets the lab's other one, since what would have to be measured is inflammation.
Read next: