Intended for healthcare professionals

Clinical Review State of the Art Review

Autoimmune complications of immunotherapy: pathophysiology and management

BMJ 2020; 369 doi: (Published 06 April 2020) Cite this as: BMJ 2020;369:m736
  1. Karmela K Chan, assistant professor of medicine,
  2. Anne R Bass, professor of clinical medicine
  1. Weill Cornell Medicine, Hospital for Special Surgery, New York, USA
  1. Correspondence to A Bass bassa{at}


Immune checkpoint inhibitors (ICIs) are monoclonal antibodies that target inhibitory molecules, such as cytotoxic T-lymphocyte-associated protein 4 (CTLA-4), programmed cell death protein 1 (PD-1), or its ligand, programmed cell death protein ligand 1 (PD-L1), and lead to immune activation in the tumor micro-environment. ICIs can induce durable treatment responses in patients with advanced cancers, but they are commonly associated with immune related adverse events (irAEs) such as rash, colitis, hepatitis, pneumonitis, and endocrine and musculoskeletal disorders. Almost all patients experience some form of irAE, but high grade irAEs occur in approximately half of those on combination therapy (eg, anti-CTLA-4 plus anti-PD-1), and up to one quarter receiving ICI monotherapy. Fatal irAEs occur in approximately 1.2% of patients on CTLA-4 blockade and 0.4% of patients receiving PD-1 or PD-L1 blockade, and case fatality rates are highest for myocarditis and myositis. IrAEs typically occur in the first three months after ICI initiation, but can occur as early as one day after the first dose to years after ICI initiation. The mainstay of treatment is with corticosteroids, but tumor necrosis factor inhibitors are commonly used for refractory irAEs. Although ICIs are generally discontinued when high grade irAEs occur, ICI discontinuation alone is rarely adequate to resolve irAEs. Consensus guidelines have been published to help guide management, but will likely be modified as our understanding of irAEs grows.


Use of immune checkpoint inhibitors (ICIs) has revolutionized the management of many advanced cancers (table 1) and results in durable responses in some patients with metastatic disease (∼10-60% depending on the cancer and ICI regimen).123 ICIs are monoclonal antibodies that block inhibitory molecules, such as cytotoxic T-lymphocyte-associated protein 4 (CTLA-4, eg, ipilimumab), programmed cell death protein 1 (PD-1, eg, nivolumab, pembrolizumab, cemiplimab), or its ligand programmed cell death protein ligand 1 (PD-L1, eg, atezolizumab, avelumab, durvalumab), which leads to immune cell activation in the tumor microenvironment (fig 1).4 However, because ICIs activate T immune cells in a variety of tissues, they are often associated with autoimmune side effects, termed “immune related adverse events” (irAEs). These irAEs commonly affect the skin, colon, liver, lungs, endocrine organs, and joints, but can affect almost any organ. IrAEs can mimic autoimmune diseases seen in other settings, but often differ at the level of tissue pathology.56 Unlike the treatment of autoimmune diseases in other settings, irAE treatment paradigms must take into account the potential impact of immunosuppression on cancer outcomes. The supporting evidence for the incidence of irAEs is relatively strong, based on multiple systematic literature reviews and meta-analyses of randomized controlled trials. However, it is also limited because of misclassification of irAEs in trials, such as ICI associated colitis characterized as “diarrhea” because endoscopy was not done, or inflammatory arthritis or polymyalgia rheumatica characterized as “arthralgia” or “myalgia.” The evidence describing clinical manifestations of irAEs and their pathobiology is weak, based on case series and observational cohorts, and evidence supporting their treatment is even weaker, driven largely by expert consensus.

Table 1

Checkpoint inhibitor approved indications

View this table:
Fig 1
Fig 1

Checkpoint inhibitor mechanism of action. Panel A: after an antigen, presented by MHC, binds to a naive T cell’s receptor, CD28 on the T cell binds to B7 on the antigen presenting cell and activates the T cell. CTLA-4 is then expressed by the T cell and competes for binding to B7. Unlike CD28, CTLA-4 sends an inhibitory signal to the T cell. Anti-CTLA-4 antibodies block that inhibitory signal. CD4+ regulatory T cells are important targets of anti-CTLA-4 antibodies. This process takes place in the lymph node. Panel B: activated T cells, particularly cytotoxic CD8+T cells, search for their tumor target in the tissues and bind to surface antigens, presented by MHC, via their T cell receptor. Later, PD-1 is expressed on the surface of the T cell. When PD-1 binds to PD-L1 on the target cell, it sends an inhibitory signal to the T cell, turning off cytotoxicity. Antibodies against either PD-1 on the T cell, or PD-L1 on the tumor cell, block this inhibitory pathway. This figure is adapted from © Terese Winslow LLC, US Govt. has certain rights

Because indications for ICIs are rapidly expanding, clinicians need to be able to recognize and treat irAEs. In this article, directed at internal medicine subspecialists, hospitalists, intensivists, and emergency department physicians, we describe the clinical characteristics of irAEs, their pathophysiology, relationship to cancer survival, and general approaches to management.

Sources and selection criteria

We searched Medline and Cochrane Reviews from 1 January 2000 through 15 July 2019, using terms for specific ICIs plus drug related side effects/adverse events/toxicities, patient reported outcomes, biologic disease modifying anti-rheumatic drugs, and the microbiome. We included English language systematic literature reviews and studies that provided primary data about irAEs, such as high quality observational cohorts or ancillary studies from clinical trials. We also included small case series and translational or non-human studies if they provided important clinical or mechanistic insight.

Grading of immune related adverse events

Immune related adverse events are characterized by grade, defined as

Grade 1: mild

Grade 2: moderate

Grade 3: severe or requiring hospitalization, but not life threatening

Grade 4: life threatening

Grade 5: death.

“Common Terminology Criteria for Adverse Events (CTCAE) 5.0”7 provides detailed characterizations of each grade for all adverse events.


The incidence of irAEs varies depending on the ICI regimen used. The best estimates come from large systematic reviews and meta-analyses. The incidence of any irAEs with anti-CTLA-4 antibodies is estimated to be 72% (95% confidence interval, 65 to 79), and of high grade irAEs 24% (18 to 30).8 Patients treated with anti-PD-1 or anti-PD-L1 (“anti-PD-(L)1”) antibodies have a similar total incidence of irAEs, 74% (69 to 79), but fewer high-grade irAEs, 14% (12 to 16).9 Combination therapy with anti-CTLA-4 plus anti-PD-1 antibodies is associated with the highest incidence of irAEs overall, 88% (84 to 92), and of high grade irAEs, 41% (35 to 47).10 A three arm clinical trial in 945 patients with advanced melanoma that directly compared anti-CTLA-4 monotherapy with anti-PD-1 monotherapy with the combination of anti-CTLA-4 monotherapy and anti-PD-1 monotherapy, showed treatment related adverse events in 85%, 82%, and 96%, and high grade adverse events in 27%, 16%, and 55%, respectively.11

The frequency of organ specific irAEs was the subject of a large systematic review and meta-analysis of 35 randomized controlled trials including 16 485 patients, and is summarized in table 2.9 Hypothyroidism and pneumonitis are more frequent with anti-PD-(L)1 antibodies, while pruritis, rash, colitis, and hypophysitis are more frequent with anti-CTLA-4 antibodies. Colitis and hypothyroidism are more common with combination therapy than with either anti-CTLA-4 or anti-PD-(L)1 monotherapy.9 Colitis, hepatitis, pancreatitis, and type 1 diabetes are more likely to be high grade when they occur.13 In practice, the incidence of irAEs may vary depending on whether combination ICIs are given concurrently or sequentially, and whether ICIs are given together with chemotherapy or with targeted agents such at tyrosine kinase inhibitors.

Table 2

Incidence of organ specific autoimmune side effects from checkpoint inhibitors (Adapted with permission)12

View this table:

Incidence of irAEs differs slightly based on the underlying cancer being treated, although the number of trials using identical regimens across tumor types is too small to make truly accurate comparisons. A systematic review and meta-analysis of 106 anti-PD-(L)1 clinical trials showed that patients with melanoma had a mean irAE incidence rate 1.72% (1.45 to 2.27) compared with lung cancer patients, who had the lowest mean incidence rate, 1.55% (1.23 to 1.81).13 In this same systematic review, high grade irAEs occurred more frequently with anti-PD-1 than with anti-PD-(L)1 antibodies (odds ratio 1.58; 95% confidence interval 1.00 to 2.54), which could be due to the presence of a second PD-1 ligand, PD-L2, which is spared by antibodies targeting PD-L1. However, PD-1 and PD-L1 inhibitors have never been directly compared in a clinical trial.


A pooled analysis of 325 patients with melanoma treated with anti-CTLA-4 antibodies assessed the kinetics of tumor responses and irAEs. Skin related irAEs typically started two to three weeks after initiation of ICIs, gastrointestinal and hepatic irAEs after six to seven weeks, and endocrine irAEs after a mean of nine weeks.14 A similar analysis used pooled data from three trials of combination anti-CTLA-4/PD-1 antibodies for melanoma. Median (interquartile range) time of onset of irAEs involving the skin was 3 (1-8) weeks, gastrointestinal tract 7.1 (4-11) weeks, liver 8.4 (5-12) weeks, endocrine organs 11.4 (7-14), lungs 9.4 (4-20), and kidneys 16.3 (4-24) weeks.15 Most high grade irAEs resolve in two to five weeks with immunosuppression but some, such as inflammatory arthritis, tend to persist16 and endocrine events usually require prolonged hormone replacement.15

Pathophysiology of immune related adverse events

Cellular autoimmunity

Many irAEs are characterized by cellular autoimmunity, with T cells and histiocytes seen in target tissues. The exact mechanism of most irAEs is unknown, but cross-reactive T cells (T cells that bind to both tumor and irAE target tissues) may play a role. An autopsy study in two patients with melanoma who had died from fatal myocarditis and myositis after treatment with combination anti-CTLA-4/PD-1 antibodies showed shared T cell clones in the tumor, heart, and skeletal muscle, but not in smooth muscle.17 A study in patients with non-small cell lung cancer who had been treated with anti-PD-1 identified shared T cell antigens in the lung tumor, and in the skin in 25 patients with ICI associated skin toxicity.18 Although this suggests that shared antigenic targets may drive both antitumor responses and organ specific autoimmunity, no study has shown a T cell clone capable of both tumor lysis and irAE target tissue damage.

Humoral immunity

Humoral immunity plays an important role in some irAEs. For example, in nine patients with anti-PD-(L)1-induced bullous pemphigoid, seven had antibodies to BP180, a basement membrane protein that is targeted in de novo bullous pemphigoid.19 The study did not document whether the autoantibodies were present before ICI initiation, and the absence of BP180 antibodies in two of nine patients points to other antigenic targets in some individuals. Antibody mediated cytotoxicity caused by anti-CTLA-4 antibodies may play a role in anti-CTLA-4 associated hypophysitis. This was shown in an animal model in which pituitary expression of CTLA-4 was shown at both the RNA and protein levels, and which showed that injection of anti-CTLA-4 antibodies induces lymphocytic infiltration and complement activation in the pituitary gland.20 In a human autopsy study, a deceased patient who had had severe anti-CTLA-4 associated hypophysitis had high levels of pituitary CTLA-4 expression, whereas four similarly treated patients without hypophysitis did not.21

Cytokines and chemokines

Cytokines and chemokines are also important intermediaries in the pathogenesis of irAEs. For example, irAEs are associated with high levels of IFN-gamma inducible chemokines, such as CXCL9 and CXCL10, which are chemotactic for T cells.22 Unlike chimeric antigen receptor (CAR)-T cells, ICIs do not typically induce cytokine storm but instead induce toxicity through cellular intermediaries. Although it is possible that fatigue, a very common irAE in patients treated with ICIs,13 is cytokine driven, this is conjecture.


Genetic susceptibility to irAEs may play an important role in pathogenesis but has been poorly studied because most observational cohorts for specific irAEs are too small to power a genetic analysis. In one study of 26 patients with ICI associated inflammatory arthritis, 62% had at least one copy of the rheumatoid arthritis associated HLA-DR susceptibility allele (the “shared epitope”), similar to rheumatoid arthritis but higher than in healthy controls.23 However, shared epitope homozygosity was lower in the ICI arthritis patients compared with rheumatoid arthritis—8% versus 24%.23 A systematic review that described 71 cases of ICI associated type 1 diabetes similarly noted that 27 of 32 patients tested had at least one type 1 diabetes HLA-DR risk allele.24

The role of the microbiome

The development of irAEs may be influenced by the microbiome, although a mechanistic understanding of the association remains obscure. In a study of 34 patients treated for metastatic melanoma with anti-CTLA-4, increased Bacteroidetes in the gut microbiome correlated with a lower risk of ICI induced colitis.25 Similarly, in another study of 26 patients with melanoma treated with anti-CTLA-4, Firmicutes but not Bacteroidetes were associated with ICI associated colitis, and patients with Faecalibacterium and other Firmicutes also had longer progression free and overall survival.26 In a study of 112 patients with melanoma treated with anti-PD-1, greater gut microbiome diversity was also associated with antitumor response.27

Manipulation of the gut microbiome can influence tumor responses and irAEs. In a study of 249 patients treated with anti-PD-(L)1 for non-small cell lung cancer, renal cell carcinoma, or urothelial carcinoma, early use of antibiotics (two months before to one month after ICI initiation) shortened progression free and overall survival.28 In animal models, fecal microbiota transplantation from ICI responders into germ free or antibiotic treated mice ameliorates the antitumor effects of PD-1 blockade, whereas fecal microbiota transplantation from non-responding patients does not.28 Two cases of refractory ICI associated colitis were successfully treated with fecal microbiota transplantation, and clinical improvement was associated with an increased proportion of T regulatory cells in the colonic mucosa.29

Biomarkers of immune related adverse events

Table 3 lists cellular, soluble, and genetic biomarkers that have been linked to irAEs. These studies are limited by their small size and the preponderance of gastrointestinal toxicity (the most common grade ≥3 irAE). None of these biomarkers has been incorporated into clinical practice, but they may provide insight into irAE pathogenesis. For example, the association of neutrophil markers (CD177),35 neutrophil growth factors (G-CSF, GM-CSF),34 a high neutrophil:lymophocyte ratio,31 and high IL-6 and IL-17 levels3233 with high grade irAEs, particularly colitis, suggests there is an important role for Th17 adaptive immune responses and innate immunity, at least in ICI associated colitis. The association of B cell changes30 and baseline autoantibodies with irAE3637 suggests that humoral immunity may also play a role in some organ specific irAEs.

Table 3

Biomarkers associated with immune related adverse events after treatment with a checkpoint inhibitor

View this table:

The relationship between immune related adverse events and survival

A systematic review of 137 immunotherapy studies encompassing 5737 patients with melanoma (who were treated with anti-CTLA-4, cancer vaccines, and other immunological therapies) showed that the development of vitiligo was associated with longer progression free (hazard ratio 0.51; 95% confidence interval 0.32 to 0.82) and overall survival (0.25; 0.10 to 0.61).42 Apart from vitiligo, other irAEs have also been shown to provide a survival advantage. For example, an analysis of seven prospective trials in patients with urothelial cancer treated with anti-PD(L)1 found that having any irAE was associated with increased overall survival compared with not having an irAE (0.53; 0.43 to 0.66).43 An important caveat in interpreting these studies is that patients who live longer on treatment have more of an opportunity to experience an irAE, although the bulk of high grade irAEs tend to occur early after treatment initiation.1415 Although irAEs can sometimes lead to discontinuation of ICIs, a retrospective analysis of pooled data from 409 patients in phase II and III trials of combination anti-CTLA-4/PD-1×4 doses followed by anti-PD-1 monotherapy found that patients who discontinued ICIs during the induction phase (n=96) had similar median progression free survival compared with those who never discontinued ICIs due to adverse events (n=231), 8.4 months versus 10.8 months (P=0.97),44 suggesting that discontinuing ICIs for toxicity may not abrogate cancer responses.

Immunotherapy re-challenge after an adverse event

Patients who discontinue ICIs due to irAEs may or may not have another irAE upon ICI re-challenge. In a group of 40 patients who discontinued anti-PD-(L)1 antibodies because of grade ≥2 irAEs who were re-challenged with the same agent, irAEs (the same or de novo) occurred in only 22 (55%).45 Similarly, in a retrospective study of 482 patients with non-small cell lung cancer who were treated with anti-PD-(L)1, of 38 who discontinued treatment because of irAEs and were re-challenged, only 20 (53%) had an irAE, although there were two treatment related deaths.46 Finally, in a study of 80 patients with melanoma who discontinued combination anti-CTLA-4/PD-1 for irAEs and were then given an anti-PD-1 antibody, only 14 (18%) had a recurrent irAE, although seven were grade 3-4 and there was one death.47 This suggests that ICI re-challenge can be considered, but that it does carry a higher risk of a fatal irAE than de novo ICI treatment.

Patients with underlying autoimmune diseases

Patients with underlying autoimmune diseases have been systematically excluded from ICI clinical trials although they represent a sizeable subpopulation of the cancer population. In a study of 210 509 patients with lung cancer in the Surveillance, Epidemiology, and End Results (SEER) database, 5.9% had rheumatoid arthritis, 2.8% psoriasis, and 1.8% polymyalgia rheumatica.48 Underlying autoimmunity may improve cancer patients’ responses to ICI. In a retrospective study of 137 patients with advanced non-small cell lung cancer treated with anti-PD-1, the presence of autoantibodies before ICI treatment was associated with prolonged progression free survival compared with patients without autoantibodies, 6.5 months (95% confidence interval, 4.4 to 12.9) versus 3.5 months (2.4 to 4.1).37

Patients with underlying autoimmunity may be at somewhat higher risk of irAEs, however. In a prospective registry study, patients with an underlying autoimmune disease (n=45) had a shorter time to irAE onset than those without (n=359), median 5.4 months versus 13 months (P=2.1×10−4).49 A prospective observational study of 751 cancer patients (85 with an autoimmune disease) treated with anti-PD1 showed that irAEs were more common in patients with an underlying autoimmune disease than in those without, 66%, (50 to 86) versus 40% (35 to 45) but that high grade irAEs occurred with similar frequencies, 9.4% (4.1 to 19) versus 8.8% (6.7 to 11.4)50 (in that study, no relationship was seen between the presence of an autoimmune disease and progression free or overall survival). In a retrospective cohort study of 102 patients with pre-existing inflammatory bowel disease treated with ICIs, 41% experienced gastrointestinal adverse events compared with 11% in patients without inflammatory bowel disease (odds ratio, 3.61; 95% confidence interval 0.85 to 15.27; P=0.081), but the incidence of de novo irAEs was not reported.51 A systematic review of case reports, case series, and observational studies detailed clinical features in 123 patients with autoimmune disease who were treated with ICIs. Fifty per cent had a flare of their autoimmune disease and 34% had a de novo irAE (75% either or both).52 Flares of the underlying autoimmune disease were more common with anti-PD-(L)1 therapy than with anti-CTLA-4 (36% anti-CTLA-4 versus 62% with anti-PD-(L)1), whereas de novo irAEs were more common with anti-CTLA-4 (42% with anti-CTLA-4 versus 26% with anti-PD-(L)1).52 There were no differences in the frequency of irAE in patients with active compared with inactive autoimmune disease at baseline, but patients on immunosuppressive therapy at the time of ICI initiation had fewer irAEs. Reporting bias is an inherent limitation of case reports, therefore this review may have overestimated the true incidence of flare and irAEs in this population.52

Clinical presentation of checkpoint inhibitor associated autoimmunity

Clinical information about organ specific irAEs is derived from observational cohorts and case series and is especially limited in the case of rare irAEs. Table 2 lists the frequency of organ specific irAEs.9 The data in that table are derived from a systematic review of randomized controlled trials, including from 19 trial arms (4762 patients) treated with anti-CTLA-4 alone, 24 trial arms (5879 patients) treated with anti-PD(L)-1 alone, and four trial arms (545 patients) treated with the combination.9

Rash In a single center study of 98 patients with ICI associated rashes over a two year period, the most common were lichenoid, maculopapular, psoriasiform, eczematous, and bullous eruptions and most were accompanied by pruritus.53 Vitiligo resulting from anti-PD-1 antibodies differs from de novo vitiligo in that it has a patchy rather than confluent distribution, it occurs in sun exposed areas, and lacks the Koebner phenomenon.5 Bullous skin disease has been reported primarily in patients receiving anti-PD-(L)1 rather than anti-CTLA-4 antibodies.54 Low grade skin reactions can be treated with topical corticosteroids but systemic treatment is indicated for high grade events.

Diarrhea and colitis ICI associated colitis occurs more commonly with anti-CTLA-4 and with combination therapy than with anti-PD-(L)1 antibodies alone (table 2) and untreated can lead to intestinal perforation.55 Histologically, patients have infiltration of the lamina propria with lymphocytes, plasma cells, neutrophils, and eosinophils, and crypt abscesses can sometimes be seen.65657 Colitis associated autoantibodies, such as anti-Saccharomyces cerevisiae antibodies and perinuclear anti-neutrophilic cytoplasmic antibodies are rarely present.58 In a single center study of 117 patients treated with ICIs who had diarrhea, about one third required no treatment, corticosteroids, or corticosteroids plus infliximab (a tumor necrosis factor (TNF) inhibitor).59 A randomized controlled trial showed that prophylactic budesonide was not effective against anti-CTLA-4 induced colitis,60 but it is commonly used to treat ICI-associated microscopic colitis.61 Infliximab has been associated with a shorter time to symptom resolution than corticosteroids alone,62 and vedolizumab (an antibody to α4β7 integrin), has also been used as a steroid sparing agent.63 Retrospective data suggest that early institution of immunosuppressive therapy with either infliximab or vedolizumab shortens symptom duration and reduces steroid exposure.64 However, physicians may be more likely to rapidly taper steroids in a patient they know has received biologics, and corticosteroids have never been prospectively compared with biologics.

Hepatitis and pancreatitis ICI associated hepatitis can be asymptomatic or can present with fever, fatigue, nausea, and jaundice. In one study, time to onset was shorter with anti-CTLA-4 than with anti-PD-(L)1 antibodies, three weeks (range 1 to 7) versus 14 weeks (range 2 to 49) (P=0.02).65 In this study, half of the patients were positive for antinuclear antibodies and 19% had anti-smooth muscle antibodies; patients treated with anti-CTLA-4 tended to have a granulomatous hepatitis whereas those treated with anti-PD-(L)1 more commonly had a lobular hepatitis.65 Pancreatitis can also be seen in response to ICIs. In a retrospective review of 2279 patients treated with ICIs, 82 (4%) had elevated pancreatic enzymes but only 32 (1.4%) had symptoms. In that study, 12 (0.5%) patients developed long term adverse sequelae of pancreatitis including type 1 diabetes in six.66

Thyroiditis ICI associated hypothyroidism is more common after anti-PD-(L)1 or combination therapy than with anti-CTLA-4 antibodies alone (table 2) and most patients receiving ICIs undergo regular thyroid testing (with both thyroid stimulating hormone and free T4) to screen for the condition. Patients usually experience a subclinical or clinical hyperthyroidism, which is followed, in most cases, by hypothyroidism requiring thyroid hormone replacement.67 Having thyroid autoantibodies before ICI treatment increases the risk of ICI associated thyroiditis.39 ICIs can be continued in the context of thyroiditis, and most patients are simply treated with thyroid replacement. Treatment with corticosteroids has not been shown to affect the duration or degree of thyroid dysfunction.68

Hypophysitis Hypophysitis is a complication of anti-CTLA-4 treatment but is only rarely seen in anti-PD-L(1)-treated individuals (table 2). The anterior pituitary gland is generally affected (most commonly the thyroid and adrenal axes) and patients often present with headache and/or fatigue.6970 In most cases, the pituitary gland and/or infundibulum is enlarged on magnetic resonance imaging (MRI) of the brain.697071 Treatment with high dose corticosteroids does not shorten the time to resolution of hypophysitis compared with replacement dosing.71 Adrenal insufficiency rarely recovers, although other hormone deficiencies can.71

Diabetes Anti-PD-(L)1 antibodies can be associated with autoimmune type 1 diabetes, which usually presents as diabetic ketoacidosis, although this is rare.72 C-peptide levels are very low in these individuals as a result of destruction of the insulin secreting β cells of the pancreas. About half of affected patients have type 1 diabetes associated antibodies (anti-glutamic acid decarboxylase), and many have type 1 diabetes associated human leukocyte antigen-risk alleles (DR4–DQ8 and DR3–DQ2).24

Pneumonitis Patients with ICI pneumonitis commonly present with dyspnea and/or cough. Fever and chest pain are uncommon, and one third of patients with pneumonitis are asymptomatic.73 There are no characteristic radiographic findings on chest computed tomography (CT); radiographic findings can include ground glass opacities, cryptogenic organizing pneumonia, or interstitial changes.7374 Lung biopsies have been reported to show organizing pneumonia, interstitial pneumonitis, or in some cases, non-necrotizing granulomas.7374 The differential diagnoses often include infection or metastatic cancer, therefore diagnosis and management of these patients is complex. Risk factors for anti-PD-1 associated pneumonitis include previous thoracic radiotherapy, previous lung disease, and the presence of fibrosis on pre-treatment chest CT.7576

Arthritis Inflammatory arthritis occurs in approximately 4% of patients treated with ICIs and most commonly presents as a symmetrical polyarthritis resembling rheumatoid arthritis.77 Other presentations include polymyalgia rheumatica and oligoarthritis with or without tenosynovitis.77787980 Rheumatoid factor and anti-citrullinated peptide antibodies occur in a minority of patients.7981 For some patients this can be managed with non-steroidal anti-inflammatories or intra-articular steroid injections, but most require systemic glucocorticoids, and many require steroid sparing agents, especially those who received combination ICIs.81 Unlike other irAEs, inflammatory arthritis often persists after discontinuing ICIs and can require prolonged immunosuppression.16

Myositis Patients with ICI associated myositis can present with acute or subacute myalgia, or weakness in a limb girdle, axial, or oculomotor distribution.82 Concomitant myocarditis and myasthenia gravis-like symptoms such as ptosis and oculomotor weakness are common, and case fatality rates are higher than with other irAEs.83 Muscle biopsy typically shows endomysial inflammation with T cells and histiocytes, as well as myonecrosis.82 In one study, eight of 24 patients with ICI myositis had one myositis associated antibody (anti-TIF1-y, SRP, Ro52; PL-7, PL-12, or SRP).84 Interestingly, anti-striated muscle antibodies are frequently found in these patients, even in the absence of clinical myasthenia gravis.8586 Patients with ICI myositis are generally treated with high dose corticosteroids, but intravenous immunoglobulin and plasmapheresis are also used in refractory cases and when concomitant neuromuscular symptoms are present.8284

Myocarditis In an observational cohort study of 35 patients with ICI associated myocarditis, most had an elevated troponin level and abnormal results on electrocardiogram, but half had preserved ejection fraction.87 Patients in this study typically had late gadolinium enhancement on cardiac MRI, and endomyocardial biopsies showed a T cell infiltrate in the myocardium. Major adverse cardiac events occurred in half of these patients, including those with preserved ejection fraction. A troponin T level ≥1.5 ng/mL was associated with a fourfold increase in major adverse cardiac events during follow-up (hazard ratio 4.0; 95% confidence interval 1.5 to 10.9).87 Case fatality rates are high in patients with ICI myocarditis, so early and aggressive treatment with immunosuppressive therapy (including very high dose corticosteroids) is critical.878889

Neurological Neurological sequelae of ICIs include headache, myasthenia gravis, peripheral neuropathy, meningitis, and encephalitis.9091 Myasthenia gravis and encephalitis are reported more commonly with anti-PD-1 antibodies, whereas Guillain Barré and meningitis are reported more commonly with anti-CTLA-4. ICI associated myasthenia gravis commonly presents with concomitant myositis and myocarditis and case fatality rates are high.9192 Half of ICI associated myasthenia gravis cases have anti-acetylcholine receptor antibodies.92 Treatments include high dose corticosteroids and, in the case of neuromuscular disorders, intravenous immunoglobulin and plasma exchange.92

Renal ICI associated renal injury is usually due to interstitial nephritis,93 but glomerulonephritis has also been described.94 Interstitial infiltrates consist of T cells, plasma cells, and eosinophils, similar to other forms of drug induced interstitial nephritis.939495

Ocular ICI treatment can be complicated by uveitis, which can involve any segment.9697 Rare cases of Vogt-Koyanagi-Harada disease have also been described in patients with melanoma, suggesting that there may be cross reactivity between T cells targeting melanoma cells and healthy ocular melanocytes in these patients.98 ICI associated sicca syndrome affects the mouth more than the eyes, and can be quite severe.99 Sjogren’s syndrome antibodies are typically negative and salivary gland biopsies reveal a T cell infiltrate and a notable absence of B cells, unlike de novo Sjogren’s syndrome, where B cells predominate.99

Hematologic Neutropenia, autoimmune hemolytic anemia, and immune thrombocytopenia occur rarely after treatment with ICIs.100 Approximately 60% of patients with ICI associated autoimmune hemolytic anemia have a positive direct antiglobulin test.101 Steroids are the mainstay of treatment, but some patients require intravenous immunoglobulin or rituximab and patients with neutropenia may require G-CSF.100101

Fatal immune related adverse events

The incidence of fatal irAEs has been assessed by the same authors using three approaches.102 The first used the VigiLyze vigibase, a voluntary international database of adverse event reports, and identified 31 059 unique reports associated with ICIs, including 613 (2%) that were fatal. The most common causes of fatal irAE with anti-CTLA-4 was colitis (70% of fatalities), while the most common cause of anti-PD-(L)1 associated fatality was pneumonitis (35%). With combination therapy, 37% of deaths were owing to colitis, and 25% to myocarditis. Case fatality rates were highest for myocarditis, myositis, hematologic toxicity, and nephritis, and the occurrence of multiple concurrent irAEs was common. The second approach assessed seven academic ICI treatment databases that included 3545 patients. Twenty one patients (0.6%) experienced a fatal irAE a median of 15 days (range 3 to 543) after ICI initiation. The onset of fatal irAE was earlier with combination ICIs than with monotherapy (14.5 days versus 40 days, P<0.001). Finally, the authors performed a systematic review of 112 randomized controlled trials involving 19 217 patients and found that the incidence of fatal irAEs was 0.38% with anti-PD-(L)1, 1.08% with anti-CTLA-4, and 1.23% with combination therapy (χ2=58.8; P<0.001 for PD-(L)1 versus anti-CTLA-4 or combination).102

Patient reported health related quality of life

A systematic review of health related quality of life in patients treated with ICIs suggested that the treatment is tolerated at least as well as traditional therapies, but the study was limited by the fact that few trials actually included patient reported outcomes.103 To quote a patient, himself a physician, who experienced a high grade irAE, “the massive fatigue and relentless sensation of flu-like symptoms are hard to glean from an [adverse event] checklist and significantly alter quality of life. Furthermore the … side effects of checkpoint inhibitors cannot be viewed in a vacuum, because prolonged corticosteroid use [to treat adverse events] is associated with high grade adverse events. For example, in my specific situation it led to fractures, weight gain, and diabetes.”

As with chemotherapy, ICI treatment can produce long lasting effects in survivors. For example, in a study of 84 patients with metastatic melanoma who achieved a durable response to an ICI and had been followed for ≥12 months, patients had a statistically significantly lower health related quality of life with regard to physical, social, and physical role functioning and general health compared with population norms. Of these patients, 43% had experienced high grade irAEs and 14% had required treatment with an immunosuppressive agent in addition to corticosteroids.104 For those patients with durable remission from their cancer but left with a new chronic illness, including type 1 diabetes24 or chronic inflammatory arthritis,16 their quality of life is affected in the long term.


Most treatment strategies for irAEs are empirical, supported only by evidence from retrospective cohort studies and expert opinion. General recommendations are to treat grade 1 irAEs with symptomatic management and close observation. Grade 2 irAEs may require low to moderate dose corticosteroids and temporary discontinuation of immunotherapy. High grade (grade 3-4) irAEs generally require high dose corticosteroids, consideration of a synthetic or biologic immunomodulatory drug, immunotherapy discontinuation, and consultation with a subspecialist. It is always important to consider other causes of the patient’s symptoms, such as infection. Close collaboration between members of the treatment team, and shared decision making with the patient are pillars of management.

Whether treatment of irAEs with immunosuppression counteracts the beneficial anti-cancer effects of ICIs has been difficult to determine because of confounding by indication, since the irAEs themselves are associated with improved survival. A retrospective study of 78 patients with multiple cancers treated with multiple ICIs showed that corticosteroid use before ICI initiation reduced the risk of irAE (adjusted odds ratio 0.14; 95% confidence interval 0.04 to 0.56; P=0.005).105 However, a retrospective study of 640 patients with non-small cell lung cancer treated with anti-PD-(L)1 monotherapy showed that patients on prednisone ≥10 mg/day at treatment initiation had shorter overall survival (hazard ratio 1.7; 1.3 to 2.2, P<0.001).106 Similarly, in a retrospective study of 98 melanoma patients with anti-CTLA-4-induced hypophysitis, patients treated with low dose corticosteroids (cut-off ≤7.5 mg prednisone/day) had significantly longer overall survival than those treated with high dose corticosteroids (0.24; 0.07 to 0.62).107

Synthetic immunomodulatory agents are often used as steroid sparing agents in patients with high grade irAEs, but targeted biologic agents are increasingly chosen in this setting, in part because of their more rapid onset of action (use of synthetic and biologic immunomodulatory agents is off-label in the setting of irAEs). Tumor necrosis factor inhibitors (TNFi) are frequently used to treat steroid refractory diarrhea, colitis, and other irAEs, and pre-clinical data suggest they may be safer than corticosteroids. For example, one pre-clinical study compared the effect of corticosteroids with TNFi on the ability to activate melanoma-specific tumor infiltrating lymphocytes in vitro and showed that dexamethasone (a long-acting corticosteroid) at low or intermediate or high doses impaired tumor infiltrating lymphocyte activation by 46% and 62%, respectively. By contrast, infliximab, a TNFi, reduced their activation by only 20%,108 although caution must be taken in extrapolating this to the in vivo setting where multiple factors may be at play. In a mouse melanoma model, investigators showed that mice deficient in TNF have prolonged survival when treated with anti-PD-1 antibodies compared with mice that are TNF sufficient.109 In addition, mice with cancer (xeno transplanted) and colitis (dextran sulfate induced) that are given TNFi when they are administered anti-PD-1 therapy have improvement in their colitis without an abrogation of antitumor responses.110 In a study of 75 patients with ICI colitis, treatment with infliximab was associated with more rapid resolution of diarrhea, more rapid corticosteroid tapering, and no difference in time to treatment failure or overall survival compared with corticosteroids alone.111 Early initiation of targeted therapy with a biologic may also improve outcomes, as shown in a study of 84 patients with ICI colitis who received targeted treatment with either infliximab or vedolizumab. In this study, initiation of the targeted therapy less than 10 days after symptom onset was associated with fewer hospitalizations, a shorter duration of symptoms, and a shorter duration of steroid therapy.64 These studies are limited by their retrospective design, however.

The IL-6/JAK/STAT3 pathway is upregulated in many cancers and is associated with a poor prognosis.112 STAT3 signaling negatively regulates NK cells, effector T cells, and dendritic cells, and positively regulates regulatory T cells.112 However, intact JAK/STAT3 signaling is needed to respond to PD-1 blockade, and JAK 1/2 mutations have been linked to anti-PD-1 resistance.113114 For this reason, JAK/STAT inhibitors are generally avoided in managing irAEs, especially those resulting from anti-PD-1 treatment. Interleukin-6 receptor (IL-6R) blockade does not directly target intracellular JAK/STAT signaling and therefore does not raise similar concerns. One report described the successful use of tocilizumab (anti-IL-6R) in 34 patients with corticosteroid refractory irAEs without a significant difference in survival compared with patients treated with corticosteroids alone.115 Rituximab (anti-CD20) is commonly used to treat steroid refractory bullous skin disease,116 and IVIG and plasmapheresis are commonly used to treat ICI associated neurological syndromes such as Guillain-Barré and myasthenia gravis.92 The use of secukinumab (anti-IL-17A) has been reported in two patients with a psoriasiform reaction to anti-PD1 therapy, but one was temporally associated with cancer progression.117118 Use of abatacept (CTLA-4-Ig) and alemtuzumab (anti-CD52) is supported only by rare case reports119120 and, based on their mechanism of action, both these agents might be expected to interfere with antitumor immunity.

Emerging treatments

No registered trials are aimed specifically at irAE treatment; however, in a small number of studies immunomodulatory therapies are being given together with ICIs to prevent toxicity and/or to improve survival. NCT04163289—In this phase I study, 20 patients with renal cell carcinoma will be given fecal microbiota capsules ≥7 days before initiation of combination ipilimumab/nivolumab (anti-CTLA-4/PD-1) and 1-3 days before the next two treatments, and the frequency of high grade colitis will be measured. NCT03293784—This 12 week open label phase Ib trial will assess the incidence of dose limiting toxicity in 30 patients with advanced melanoma treated with a TNFi (certolizumab or infliximab) concomitantly with combination ipilimumab/nivolumab. NCT03999749—This phase II, single arm open label trial will assess the safety and tolerability of tocilizumab (interleukin-6 receptor inhibitor) every six weeks for five doses, in combination with ipilimumab/nivolumab induction followed by nivolumab monotherapy, in 67 patients with advanced melanoma. Although the primary outcome is tumor response, the incidence of irAE will also be measured. NCT03719131—This single center phase II trial will compare rates of high grade irAEs in 44 patients with stage III-IV melanoma treated with ipilimumab/nivolumab, with or without four weekly doses of rituximab and hyaluronidase human. Tumor response will also be measured. NCT03190174—This single center phase Ib study is assessing the maximum tolerated dose of an mTOR inhibitor (nab-rapamycin) together with nivolumab in 40 patients with advanced sarcoma, and should provide interesting information on irAE incidence. NCT03816332—This phase I trial will examine cancer outcomes in 16 renal transplant recipients with advanced cancer treated with combination ipilimumab/nivolumab plus tacrolimus and prednisone, but will also provide important information about irAE incidence and the frequency of transplant rejection in these patients.


Several organizations have published guidelines for the management of irAEs, including the Society for Immunotherapy of Cancer,121 the American Society of Clinical Oncology,122 and the National Comprehensive Cancer Network,123 however their recommendations are based on expert consensus. All three guidelines recommend holding ICIs for ≥grade 3 irAEs and starting systemic corticosteroids (except for endocrine irAEs, which are usually treated with hormone replacement). The guidelines provide clinical information on irAEs in addition to detailed recommendations for organ specific irAE management. These guidelines will require frequent updating as our knowledge grows.


ICIs provide durable remissions in a substantial percentage of patients with metastatic cancer, including in up to 58% with metastatic melanoma.2 However, ICIs are commonly associated with autoimmune toxicities, including high grade toxicity in up to a half of patients and fatal toxicities in 0.4%-1.2%, depending on the ICI regimen given.102 Although patients with pre-existing autoimmune diseases were excluded from all ICI clinical trials, they do not appear to have higher rates of high grade irAEs than other patients. However, approximately half of these patients will have a flare of their underlying autoimmune condition with ICI treatment.52 Most adverse events occur 2-9 weeks after ICI initiation, but adverse events can be seen as late as two years after.15102 Commonly affected organ systems include the skin, gastrointestinal tract, lungs, and endocrine system. Colitis and hypophysitis are more frequent with anti-CTLA-4 treatment, and thyroiditis and pneumonitis are more common with anti-PD-(L)1 treatment. Patients who experience an adverse event are more likely to survive their cancer, suggesting shared mechanisms driving efficacy and toxicity. The microbiome may be an important predictor of both efficacy and toxicity of ICIs, particularly ICI associated colitis, but otherwise we lack pre-treatment biomarkers of specific ICI toxicities to guide treatment and monitoring. We also have little understanding of the genetics of ICI response and toxicity.

Treatment of high grade ICI related toxicity with high dose corticosteroids may prevent substantial morbidity and mortality, however high dose corticosteroids may be associated with reduced cancer survival.107 In retrospective studies, early intervention with a targeted biologic, such as infliximab or vedolizumab, was associated with better outcomes than corticosteroids alone in patients with colitis, but this has not been studied in the case of other irAEs. Given current knowledge, the key to management is early recognition. Whether in the inpatient or outpatient setting, it is critical to consider autoimmune toxicity in the differential diagnosis for any ICI treated patient presenting with new symptoms. The management of these complex patients is best accomplished with a team based approach that allows for close communication between oncologists and medical subspecialists, and incorporates the values and preferences of the patient.

Research questions

  • How can we predict who will experience high grade adverse events from immunotherapy?

  • What is the safest approach to treating immune related adverse events without compromising survival?

  • Can manipulation of the microbiome be used to reduce immune related adverse events other than colitis?

Patient involvement

In drafting this review we involved patients who had experienced irAEs. Patients were sent an outline of the review for comment early in the writing process. They were also given an opportunity to review and comment on the manuscript before it was finalized. One patient emphasized the importance of irAE biomarkers to help predict who is at risk, and commented on the underappreciated impact of multiple, albeit “low grade,” irAEs on quality of life. Another mentioned that we should comment on toxicities that occur when ICIs are combined with other therapies. One patient noted that, although side effects from the immunotherapy mimic other autoimmune illnesses, he was frustrated that they cannot always be treated with traditional autoimmune disease medicines that might compromise his survival.


  • Series explanation: State of the Art Reviews are commissioned on the basis of their relevance to academics and specialists in the US and internationally. For this reason they are written predominantly by US authors

  • Competing interests: The BMJ has judged that there are no disqualifying financial ties to commercial companies. The authors declare the following other interests: none.

  • Further details of The BMJ policy on financial interests are here:

  • Contributorship statement and guarantor: Drs Bass and Chan contributed equally to drafting the first draft of this manuscript. Dr Bass was responsible for overall content and final edits, and serves as the guarantor.

  • Provenance and peer review: commissioned; externally peer reviewed.


View Abstract