The PRISMA flowchart is shown in Fig. 1 for identification, screening, and selection of study records, including reason for exclusion.  A total of 5885 studies were identified in the three databases. After excluding duplicates, 3595 studies remained. After reviewing title and abstract screening and manual search of references, 2230 studies met the PICOS criteria and were eligible for full text screening. Following this, 357 studies proceeded to data extraction where additional studies were excluded. Finally, 293 studies were included for analysis. Details of search strategy are available in Supplementary Table S1.

Study design and total patient numbers of the included studies are shown in Tables 1 and 2. In 293 included studies, irAE were reported in 58,291 patients among 305,879 total patients who were reported to have received ICI treatment. There were 24 studies that did not report on the number of patients with irAE and/or the study population size. There were 221 retrospective studies including 187 cohort or observational, 14 case-control, nine pharmacovigilance or surveillance, eight clinical trial subsets and three experimental/exploratory/pilot studies. Retrospective studies accounted for the majority of ICI patients included: 240,683 patients (78.7%). There were 55 prospective studies including 31 cohort or observational, 18 experimental/exploratory/pilot studies and six clinical trial subsets accounting for 4462 patients (1.5%). Additionally, there were 17 systematic reviews or meta-analyses containing 60,734 patients (19.9%). Full details of study characteristics are shown in Supplementary Table S2.

Of all the included studies, 170 (58.0%) performed a cancer-specific approach and 123 (42.0%) performed a pan-cancer approach to data analysis. Among all studies that performed a cancer specific approach, 81 focused on lung cancer (74 NSCLC, three NSCLC or SCLC, four other lung cancers), 62 studies focused on skin cancer (61 melanoma, one other skin cancers), three on renal cell carcinoma, two studies each on gastric cancer, head and neck cancer and urothelial carcinoma, one study each on liver, upper gastrointestinal cancer, Hodgkin’s lymphoma, and fifteen studies that included multiple specific cancer types.

Among all studies, 193 (65.9%) focused on monotherapy across any class of ICI, 92 (31.4%) studies included both monotherapy and combination ICI therapies while eight (2.7%) focused solely on combination therapies.

Among all studies, 162 (55.3%) reported on general irAE occurrence, 124 (42.3%) reported on specific irAE type(s), while seven studies (2.3%) reported on both specific irAE type(s) and general irAE occurrence. Overall, 46 (15.5%) studies reported on high grade irAE; 43 studies reported general high grade irAE and three studies reported on specific high grade irAE types. Endocrine irAE and pulmonary irAE were reported in 31 studies, gastrointestinal in 26, renal in 13, cardiac in 11, skin in 7, musculoskeletal in three, neurological in one study. Definition of irAE varied across included studies. Common Terminology Criteria for Adverse Events (versions 2, 3, 4 and 5) were the most widely used resource to define irAE among included studies.

Cardiac irAE included myocarditis, acute vascular events, pericardial disease, atrial fibrillation, cardiac failure, pericardial effusion, vasculitis, and dyspnoea. Endocrine irAE included hypo- and hyperthyroidism, adrenal insufficiency, hypophysitis, thyroiditis, isolated adrenocorticotropic hormone deficiency, new onset and worsening type 2 diabetes. Gastrointestinal irAE predominately included diarrhoea and colitis, and enteritis, pancreatitis, hepatitis. Musculoskeletal irAE included myositis alone or overlap manifestation (myositis and myocarditis and/or myasthenia gravis), inflammatory arthritis, arthralgia, and polymyalgia rheumatica. Neurological irAE included myasthenia gravis, neuropathy, Guillain-Barre syndrome, meningitis, encephalitis, myelitis, and demyelinating disorders. Pulmonary irAE predominately included pneumonitis and interstitial lung disease. Renal irAE included acute kidney injury, kidney failure and nephritis. Skin irAE included pruritus, rash, erythema, vitiligo, and skin eruption (macular-papular or eczematous). Full details of irAE types is provided in Supplementary Table S3.

Event rates were calculated for 269 studies representing 37,442 patients with irAE among 278,772 total patients who received ICI. Event rate could not be calculated for 24 studies due to insufficient reporting on number of patients with irAE (14 studies), study population on ICI (8 studies), or neither number of patients with irAE nor study population on ICI (3 studies).

Event rate was calculated for 132 (n = 75,988) and 43 (n = 25,607) studies that reported either general and high grade irAE occurrence, respectively. The mean event rate for general irAE occurrence across any grade was 40.0% (37.3–42.7%; 95% CI) and high grade was 19.7% (15.8–23.7%; 95% CI). Figure 2 shows the distribution of general irAE occurrence for any grade and high grade among studies.

n = 175 studies. Any grade: minimum 3.5%; first quartile 29.3%; median 40.0%; third quartile 49.4%; maximum 84.0%; mean 40.0%; interquartile range 20.1%. High grade: minimum 1.7%; first quartile 9.6%; median 16.9%; third quartile 25.8%; maximum 56.0%; mean 19.7%; interquartile range 16.2%.

Overall, event rates were calculated for 121 studies (n = 188,506) that reported specific irAE occurrence. The mean event rates for each type of specific irAE occurrence were as follows: cardiac irAE 18.0% (5.9–30.2%; 95% CI), endocrine irAE 23.9% (18.8–29.1%; 95% CI), gastrointestinal irAE 19.4% (14.1–24.6%; 95% CI), pulmonary irAE 18.9% (15.1–22.7%; 95% CI), renal irAE 15.5% (7.3–23.7%; 95% CI), and skin irAE 28.7% (22.9–34.6%; 95% CI). Figure 3 shows the distribution of specific irAE occurrence among studies. Due to low number of studies, event rates are not presented for musculoskeletal and neurological irAE.

n = 118 studies. Cardiac: minimum 0.8%; first quartile 2.2%; median 5.9%; third quartile 37.6%; maximum 52.3%; mean 18.0%; interquartile range 35.4%. Endocrine: minimum 1.5%; first quartile 12.4%; median 22.8%; third quartile 33.5%; maximum 65.6%; mean 23.9%; interquartile range 21.0%. Gastrointestinal: minimum 1.3%; first quartile 5.7%; median 12.5%; third quartile 33.3%; maximum 52.5%; mean 19.4%; interquartile range 27.6%. Pulmonary: minimum 2.8%; first quartile 11.7%; median 15.0%; third quartile 24.7%; maximum 45.4%; mean 18.9%; interquartile range 13.0%. Renal: minimum 1.4%; first quartile 4.5%; median 14.9%; third quartile 17.3%; maximum 50.0%; mean 15.5%; interquartile range 12.8%. Skin: minimum 16.7%; first quartile 18.2%; median 25.1%; third quartile 40%; maximum 47.7%; mean 28.7%; interquartile range 21.8%. Note: Due to low number of studies, event rates are not presented for musculoskeletal and neurologic irAE.

Event rates varied across different ICI treatment types among 183 studies (n = 168,132) that focused on monotherapy, six (n = 339) combination therapy, and 80 studies (n = 110,301) that included both monotherapy and combination therapy. The mean event rate was expectedly lowest in monotherapy, 30.5% (28.1–32.9%; 95% CI), and highest with combination therapy, 45.7% (29.6–61.7%). Studies that had both monotherapy and combination therapy cohorts had a mean event rate of 30.0% (25.3–34.6%; 95% CI). Figure 4 shows the distribution of overall irAE occurrence across different ICI treatments among included studies.

n = 269 studies. Monotherapy: minimum 0.1%; first quartile 17.1%; median 29.7%; third quartile 43.5%; maximum 84.0%; mean 30.5%; interquartile range 26.4%. Combination: minimum 12.8%; first quartile 24.9%; median 48.0%; third quartile 66.3%; maximum 71.7%; mean 45.7%; interquartile range 41.4%. Combination & monotherapy: minimum 0.4%; first quartile 9.8%; median 32.6%; third quartile 46.3%; maximum 84.0%; mean 30.0%; interquartile range 36.5%.

Of the 293 studies analysed, 225 (76.8%) reported a difference, association, or correlation between the presence or absence of a specific factor, characteristic, or measure and the development of irAE at any stage prior to or during treatment. In contrast, 68 studies (23.2%) found no such relation. Further breakdown of the studies revealed that 158 (53.9%) specifically aimed to investigate the relation between a risk factor or predictor and the occurrence or severity. Meanwhile, 135 studies (46.1%) did not have this aim but contained data comparing the occurrence or severity of irAE between patients who did and did not experience irAE. It is worth noting that the statistical methods employed to assess differences, associations, or correlations varied considerably across these studies.