A Systematic Review and Meta-analysis of Observational Studies and Uncontrolled Trials Reporting on the Use of Checkpoint Blockers in Patients with Cancer and Pre-existing Autoimmune Disease

Abstract

Background:

Cancer patients with autoimmune disease have been excluded from randomized trials of immune checkpoint blockers (ICBs). We conducted a systematic review of observational studies and uncontrolled trials including cancer patients with pre-existing autoimmune disease who received ICIs.

Methods:

We searched 5 electronic databases through November 2023. Study selection, data collection, and quality assessment were performed independently by 2 investigators. We performed a meta-analysis to pool incidence of immune-related adverse events (irAEs), including de novo events and flares of existing autoimmune disease, hospitalizations due to irAEs, as well as deaths.

Results:

A total of 95 studies were included (23,897 patients with cancer and preexisting autoimmune disease). The most common cancer evaluated was lung cancer (30.7%) followed by skin cancer (15.7%). Patients with autoimmune disease were more likely to report irAEs compared to patients without autoimmune disease (relative risk 1.3, 95% CI 1.0 to 1.6). The pooled occurrence rate of any irAEs (flares or de novo) was 61% (95% CI 54% to 68%); that of flares was 36% (95% CI 30% to 43%), and that of de novo irAEs was 23% (95% CI 16% to 30%). Flares were mild (grade <3) in half of cases and more commonly reported in patients with psoriasis/psoriatic arthritis (39%), inflammatory bowel disease (37%), and rheumatoid arthritis (36%). 32% of the patients with irAEs required hospitalization and treatment of irAEs included corticosteroids in 72% of the cases. The irAEs mortality rate was 0.07%. There were no statistically significant differences in cancer response to ICIs between patients with and without autoimmune disease.

Conclusions:

Although more patients with pre-existing autoimmune disease had irAEs, these were mild and managed with corticosteroids in most cases, with no impact on cancer response. These results suggest that ICIs can be used in these patients, but careful monitoring is required, as over a third of the patients will experience a flare of their autoimmune disease and/or require hospitalization. These findings provide a crucial foundation for oncologists to refine their monitoring and management strategies, ensuring that the benefits of ICI therapy are maximized while minimizing its risks.

Immune checkpoint blockers (ICBs) have revolutionized the treatment and prognosis of many cancers, including those in advanced stages. Three major classes of ICIs are currently approved by the U.S. Food and Drug Administration (FDA): anti–cytotoxic T-lymphocyte-associated protein 4 (CTLA-4) antibodies, anti–programmed cell death protein and anti–programmed death ligand 1 (PD-1/PD-L1) antibodies, and lymphocyte activation gene 3 (LAG-3) (Supplementary Table 1).

When compared to alternative treatment options, ICI therapy, whether as monotherapy or in combination, resulted in considerable improvements in survival rates for various malignancies.^1–6 Despite their benefits, ICIs frequently generate immune-related adverse events (irAEs), which can impact practically any system or organ and can be severe.^7,8

The precise pathophysiology of irAEs is unknown and may vary depending on the toxicity phenotype, however irAEs are commonly related to the widespread activation of immune pathways, culminating in inflammatory and autoimmune symptoms. As a result, cancer patients with autoimmune conditions have been excluded from participating in ICI trials due to concerns about increasing their risk for immune toxicities and/or flares of their concomitant autoimmune disease. Limited data exist regarding the safety of ICIs in patients with cancer and autoimmune disease.^7,9,10 About 15% of the U.S. population suffers from an autoimmune disorder (approximately 50 million people).^11–13 Furthermore, autoimmune and chronic inflammatory diseases have been significantly associated with an increased risk for cancer.¹⁴ It is thus critical to quantify the potential risk for immune toxicity and for increased disease activity in patients with cancer who could benefit from ICI therapy but have a concomitant autoimmune disorder.

To the best of our knowledge, no controlled clinical trials have been reported to date evaluating the use of ICIs in patients with autoimmune disease. We systematically reviewed observational studies and uncontrolled trials reporting on outcomes of cancer patients with pre-existing autoimmune disease who received ICIs to determine the frequency of flares of autoimmune disease, de novo irAEs, hospitalizations, and deaths. Previous reviews have summarized irAE rates in patients with an underlying autoimmune disease (Supplementary Table 2), but the evolving nature of the topic, new ICI approvals, and the heterogeneity of autoimmune diseases (e.g., a psoriasis flare may not have the same outcome as a lupus flare) points to the need for a more comprehensive and systematic synthesis that estimates the incidence of irAEs, overall and according to specific condition among patients with cancer and pre-existing autoimmune disease.

Methods

For this systematic review, we followed the Cochrane Collaboration methodology and the 2020 PRISMA statement for reporting our results.^15,16

Eligibility criteria

We included prospective or retrospective studies reporting the effects of ICIs in cancer patients with pre-existing autoimmune disorders. We excluded animal studies, clinical trials not reporting on autoimmune disease, and review articles. We also excluded case reports or series, studies where it was unclear whether all patients had an autoimmune disease (e.g., nonspecific interstitial lung disease, thyroid dysfunction, autoantibodies, paraneoplastic syndromes), studies exploring association between comorbidities and irAEs, and studies that did not provide separate data or outcome data (e.g., proportion of irAEs, flares, or deaths) for the patients with autoimmune disease.

Information sources

The databases searched were Medline (through Ovid), EMBASE (through Ovid), The Cochrane Library, Web of Science, and PubMed (non-Medline), from the start of the database through November 2023. In addition, the reference lists of included articles and prior comprehensive reviews were hand-searched. EndNote software (Clarivate Analytics) was used to store all citations and for duplicate checking.

Search strategy

The first search strategy was developed and conducted by a librarian (G.P.) with expertise in systematic reviews in the medical field to encompass all ICIs approved by the FDA. As opposed to previous reviews, ours was an exceptionally comprehensive search aimed to find all published studies on the topic. We included broad terms such as “cancer”, “tumor”, and “neoplasms” and drug-related terms such as “ipilimumab” and “nivolumab”, among others. The second search was run by another expert librarian (Y.G.) and incorporated terms for ICIs approved from 2019 onward. The final search was done by the first author of this study using the same strategy as the second but incorporating the FDA-approved ICIs in 2023 with the help of the librarian (Y.G.) with an end date of November 2023 and restricting results to new citations and human studies. The Medline search strategies are shown in Supplementary Table 3.

Selection process

The eligibility of the identified citations was assessed independently by pairs of reviewers each time the search was updated (N.A.; M.A.L.-O.; J.D.A.; M.K.; X.P.; J.J.K.; and P.C.) in a 2-step approach. Titles and abstracts of all retrieved citations were screened first, and the full text of those deemed potentially relevant was retrieved for further screening. Disagreements at either stage were resolved through discussion and, if needed, third-party adjudication (M.A.L.-O. or M.S.A.). We used DistillerSR (Evidence Partners Incorporated; Ottawa, Canada) to manage selection. For the last update (years 2022 and 2023), we used a modified screening approach using DistillerSR’s machine learning tool to estimate the likelihood of relevance based on a previously validated approach.¹⁷

Data collection process

One reviewer extracted data from individual studies, and another reviewer cross-checked the information (N.A.; M.A.L.-O.; J.D.A.; M.K.; X.P.; J.J.K.; and P.C.). Disagreements were resolved through discussion. We used Microsoft Excel to extract and tabulate the data.

Data items

We collected the following information: i) general information including title, authors, contact address, publication source, publication year, country, study sponsor; ii) study characteristics including type of study design, study period, sample size, exclusion criteria, and number of patients with and without pre-existing autoimmune disease; iii) participants’ baseline characteristics (age, gender), type of autoimmune disease, time since autoimmune disease diagnosis, autoimmune disease treatments, number of patients discontinuing their autoimmune disease treatment before the initiation of ICI, number of patients with autoimmune disease activity (i.e., symptoms from the autoimmune disease at the time of ICI treatment initiation); and iv) cancer characteristics including type of cancer, cancer stage, number of patients with previous use of ICI, current ICI treatment, and ICI doses. Outcome data included number of patients experiencing an irAE, categorized as a flare of their concomitant autoimmune disease or an unrelated de novo irAE. We also recorded irAE grade (based on the Common Terminology Criteria for Adverse Events), median time to flares after ICI initiation, treatment used for management of irAEs (i.e., corticosteroids, immunosuppressants), hospitalizations due to irAEs, permanent discontinuation of ICIs after irAEs, and deaths due to irAEs. The cancer outcomes collected included complete responder, partial responders and progressive disease.

Study risk of bias assessment

One reviewer assessed the risk of bias of the included studies, and another reviewer cross-checked the information (N.H.A.-W.; M.A.L.-O.; J.J.K.; and P.C.). Studies were appraised using The Newcastle Ottawa Scale (NOS).¹⁸ This instrument assesses the potential for bias due to confounding, in selection of participants, in measurement of interventions, due to departures from intended interventions, due to missing data, and in measurement of outcomes. Studies can be awarded a maximum of 1 star for each domain (2 if studies controlled for additional confounders). The maximum score allocated is 4 in the selection domain, 2 in the comparability domain, and 3 in the outcome (or exposure, for case-control studies) domain, with a total maximum score of 9. A higher score indicates a lower risk of bias.¹⁹

Effect measures

Our primary outcomes measures were flare and any irAEs occurrence. Secondary outcomes included de novo irAEs, median time to flare after ICI initiation, use of corticosteroids and/or immunosuppressants to manage irAEs, deaths related to irAEs, hospitalizations, permanent ICI discontinuation after an irAE, and cancer outcomes (i.e., complete responders, partial responders, and progressive disease). For the primary outcome, our numerator was the number of cancer patients with autoimmune disease and flares in each study. For the denominator, we considered all patients with cancer and autoimmune disease. We excluded from the denominator non-autoimmune diseases included in the individual analysis as pre-existing autoimmune conditions, such as gout, asthma, atopic dermatitis, Bell’s palsy, eczema, fibromyalgia, idiopathic pulmonary fibrosis, lipoid nephrosis, mucositis, reactive arthritis, and sicca syndrome. We pooled the proportions and calculated relative risks and their 95% CIs. For comparisons of dichotomous variables, we estimated the relative risk and 95% CIs.

Synthesis methods

Analyses were conducted using Stata statistical software version 18 (StataCorp). Descriptive statistics were used to report frequencies of autoimmune disease and cancer types. Pooled frequencies were calculated using the Freeman-Tukey arcsine transformation to stabilize variances and conducted a meta-analysis using inverse variance weights. Then, estimates and their 95% CI boundaries were back-transformed into proportions. We used random effects models to calculate more conservative pooled estimates (incidences/proportions and RR). Study heterogeneity was assessed by using the I² statistic and subsequent χ2 test. We performed exploratory subgroup analyses to evaluate: 1) the impact of the treatment strategy (combination therapy (2 ICIs or an ICI combined with another treatment) vs monotherapy); and 2) study design (prospective vs retrospective studies). A meta-regression was performed to investigate the extent to which statistical heterogeneity was explained by study sample size, year in which the study was conducted, type of ICI agent received, and grade of irAE reported.

Reporting bias assessment

The risk of publication bias was assessed through funnel plots. When the regression slope significantly deviated from the vertical slope, this was considered to indicate significant potential for bias.

Certainty assessment

The strength of the evidence was evaluated using the GRADE system.²⁰ It divides findings into four categories: i) high quality, which indicates that there is little chance that future research will alter our conclusions; ii) moderate quality, which indicates that future research will probably have an impact on our conclusions; iii) low quality, which indicates that future research will likely have a significant impact on our conclusions; and iv) very low quality, which indicates that there is uncertainty in our conclusions. Findings based on observational studies are rated as low or very low quality of evidence with the GRADE system; however, they can lead to strong recommendations in clinical settings where randomized trials are not available.

Results

Study selection

Flow of studies.

Our searches found 67,006 citations; after exclusion of duplicates, we screened titles and abstracts of 37,163 citations. Of these, we excluded 29,649 citations. The reasons for exclusion at each screening step are shown in Figure 1. We screened the full text of the remaining 7,514 citations, and we found 120 publications representing 95 studies meeting our eligibility criteria.^9,21–139

Figure 1.

PRISMA flow diagram: Study selection.

Excluded studies.

There were 7,301 citations excluded because they were case reports or series (with or without pre-existing autoimmune disease) or observational studies without patients with pre-existing autoimmune disease. The remaining 93 excluded were studies with an unclear number of patients with pre-existing autoimmune disease, studies reporting on patients with pre-existing autoantibodies (e.g., rheumatoid factor, anti-nuclear antibodies), studies exploring association between comorbidities and irAEs, or studies with insufficient data provided for analysis (i.e., outcome data not reported or preliminary data from studies included in the analysis). These references are listed in Supplementary Table 4.

Study characteristics

Fifty-two studies (54.1%) also reported data in cancer patients without a pre-existing autoimmune disease; 11 (11.5%) were results from national registries, regulatory agencies, or claims data; 4 (4.2%) were data from uncontrolled trials (single arm or phase IV). In total, the included studies evaluated 23,897 patients with cancer and autoimmune diseases and 51,676 cancer patients without autoimmune disease. Thirty-three studies (34.4%) were multicenter. The studies were conducted in Australia, Canada, China, Europe, Israel, Japan, Turkey, and the United States. The most commonly reported autoimmune diseases were inflammatory bowel diseases (34.7%), type 1 diabetes (13.1%), rheumatoid arthritis (11.6%), psoriasis (7.5%), and Addison disease (3.0%) (Table 1).

Table 1.

Characteristics of the studies included in our analysis (n=95 studies).

Characteristic	Value
Design
Uncontrolled cohort	28 (29.2%)
Controlled cohort	51 (54.1%)
Registries	11 (11.5%)
Uncontrolled trials	4 (4.2%)
Individual patient data from multiple controlled trials	1 (1.0%)
Centers
Single-center	63 (65.6%)
Multi-center	32 (34.4%)
Countries
Australia	4 (4.2%)
Canada	3 (3.1%)
China	2 (2.1%)
Europe	30 (31.3%)
Israel	1 (1.0%)
Japan	6 (6.3%)
United States	40 (42.7%)
Global collaboration	9 (9.4%)
Time period (range)	1999 to 2022
Number of patients without pre-existing autoimmune disease	51,676
Number of patients with pre-existing autoimmune disease	23,897
Number of autoimmune diseases reported (n= 30,493a)
Inflammatory bowel disease	10,585 (34.7%)
Type 1 diabetes	4,007 (13.1%)
Rheumatoid arthritis	3,542 (11.6%)
Unspecified	2,436 (8.0%)
Psoriasis	2,296 (7.5%)
Addison’s disease	922 (3.0%)
Hashimoto’s thyroiditis	758 (2.5%)
Systemic lupus erythematosus	595 (2.0%)
Dermatologic (unspecified)	525 (1.7%)
Vasculitis	515 (1.7%)
Autoimmune thyroiditis	467 (1.5%)
Graves’ disease	458 (1.5%)
Multiple sclerosis	339 (1.1%)
Rheumatic (unspecified)	325 (1.1%)
Lichen	300 (1.0%)
Otherb	2,423 (7.9%)

^aSome patients had overlapping disease, therefore the number of autoimmune diseases is larger than the number of patients with pre-existing autoimmune diseases reported.

^bOther autoimmune disease with less than 1% individually presence were: celiac disease, vitiligo, CREST syndrome (calcinosis, Raynaud phenomenon, esophageal dysmotility, sclerodactyly, and telangiectasia), spondyloarthropathies, endocrine (unspecified), systemic sclerosis, myasthenia gravis, ulcerative colitis, autoimmune hepatitis, Crohn’s disease, alopecia areata, polymyalgia rheumatica, dermatomyositis, sarcoidosis, bullous pemphigus, psoriasis with psoriatic arthritis, psoriatic arthritis, neurologic (unspecified), pernicious anemia, immune thrombocytopenic purpura, gastrointestinal (unspecified), inflammatory arthritis, Sjogren, Guillain-Barré syndrome, Raynaud’s phenomenon, uveitis, discoid lupus, IgA nephropathy, mixed or undifferentiated connective tissue disorder, cutaneous lupus, myositis, hemolytic anemia, polymyositis, granulomatosis with polyangiitis, autoimmune hypophysitis, Behçet’s disease, primary biliary cholangitis, chronic inflammatory demyelinating neuropathy, giant cell arteritis, hidradenitis suppurativa, rheumatic fever, autoimmune urticaria, anti-phospholipid syndrome, Churg-Strauss syndrome, erythema nodosum, IgG4-sclerosing, juvenile idiopathic arthritis, glomerulonephritis, myelitis, autoimmune pancreatitis, polyarteritis nodosa, autoimmune retinopathy.

Participants’ characteristics

Table 2 shows the baseline demographic characteristics of participants (n=23,897) in the included studies. The mean age ranged from 54 to 76 years (pooled mean 66.1, 95% CI 65.3 to 67.0). The proportion of male participants ranged between 19% and 100%. Lung cancer (30.7%) was the most commonly reported cancer, followed by skin cancer (15.7%). Studies included patients with resectable cancer (i.e., melanoma) and advanced cancer stages (i.e., II to IV). Four studies reported use of previous ICIs in the included patients (CTLA-4 in 3 and PD-1/PD-L1 in 2).^{9,40,62,89,140} Nivolumab, pembrolizumab, and ipilimumab were the ICIs most commonly evaluated. Duration of current ICI treatment ranged from 120 to 196 days, and the number of cycles ranged between 1 and 20. The percentage of patients on treatment for autoimmune disease in each study ranged from 0% to 87.5%; this treatment varied and included nonsteroidal anti-inflammatory drugs, corticosteroids (topical, oral, or systemic), and conventional and targeted disease-modifying anti-rheumatic drugs (e.g., hydroxychloroquine, methotrexate, sulfasalazine, infliximab, apremilast). At the time of initiating ICI, the percentage of patients with active autoimmune disease ranged from 0% to 67% across studies.

Table 2.

Baseline demographic characteristics of participants with autoimmune disease in the included studies.

Characteristic	Value
Age, pooled mean (95% CI)	66.1 (65.3 to 67.0)
Age, range of means	54 to 76 years
Male sex, range	19% to 100%
Cancer type, no. (%) (n=23,897 participants with autoimmune disease)
Lung (non–small cell carcinoma, small cell carcinoma, mesothelioma, unspecified)	7,336 (30.7%)
Skin (melanoma, squamous cell carcinoma, Merkel cell carcinoma)	3,752 (15.7%)
Genitourinary (renal cell, urothelial, bladder, prostate, penile, unspecified)	2,127 (8.9%)
Gastrointestinal (colorectal, hepatocellular, esophageal, pancreatic, unspecified)	1,936 (8.1%)
Head and neck (sinonasal, unspecified)	263 (1.1%)
Gynecological (breast, ovarian, endometrial, epithelial trophoblastic, unspecified)	134 (0.56%)
Hematological (Hodgkin’s lymphoma, B-cell lymphoma, unspecified)	127 (0.53%)
Unspecified	8,149 (34.1%)
Othersa	74 (0.31%)
Distribution of ICIs, no. (%) (n=23,897 participants with autoimmune disease)
Anti–PD-1	5,807 (24.3%)
Nivolumab	4,134 (17.3%)
Pembrolizumab	2,557 (10.7%)
Ipilimumab	2,246 (9.4%)
ICI monotherapy	1,768 (7.4%)
Anti–PD-1/PD-L1	1,147 (4.8%)
ICI combination	550 (2.3%)
Combination	502 (2.1%)
Durvalumab	406 (1.7%)
Anti-CTLA–4	263 (1.1%)
Atezolizumab	48 (0.2%)
Avelumab	24 (0.2%)
ICI + non-ICI	27 (0.1%)
Unspecified ICI	4397 (18.4%)
Duration of current ICI treatment, range	120 to 196 days
Number of cycles, range	1 to 20
Patients on treatment for autoimmune disease, range	0% to 87.5%
Active autoimmune disease, range	0% to 67%

^aOthers included neuroendocrine, brain unspecified, sarcoma, thymoma, thyroid.

Risk of bias within studies

Supplementary Figure 1 shows the risk of bias assessment of the included studies in an aggregate manner. NOS score ranged from 4 to 7 stars. For 28 studies (29.2%), there was no control group (i.e., patients without autoimmune disease receiving ICIs). Few studies adjusted their analysis for potential confounders. In 1 study, the follow-up time was not reported.¹²⁵ Four studies reported no source of funding, 1 cited the U.S. National Institutes of Health,⁶² and 1 reported 7 funding sources.⁸⁰

Safety of ICIs in patients with pre-existing autoimmune diseases

Table 3 shows the reported number of patients not experiencing irAEs and those experiencing de novo irAEs, flares, and deaths. Patients with autoimmune disease were 30% more likely to report an irAE compared to patients without autoimmune disease (relative risk 1.3, 95% CI 1.0 to 1.6). The pooled frequency of any irAEs (flares or de novo) was 61% (95% CI 54%-68%); the pooled frequency of flares was 36% (95% CI 29% to 43%), and that of de novo irAEs was 23% (95% CI 16% to 30%). Out of 1,827 de novo irAEs reported, the most common were colitis (17%), thyroid dysfunction (11%), skin rash (6%), and arthritis/arthralgia (5%). When categorized by type of ICI, 25% (95% CI 15% to 36%) of the patients receiving anti–PD-1/PD-L1 agents had irAEs, as did 38% (95% CI 22% to 54%) of those who received anti–CTLA-4 agents. Although most irAEs were mild, 30% of the irAEs were grade 3–4, and 32% of all irAEs in patients with pre-existing autoimmune disease required hospitalization. Pooled mortality related to ICIs was less than 0.1%; however, the pooled frequency of permanent discontinuation of the ICI due to irAEs was 35% (95% CI 24% to 46%). Most irAEs were managed with local or systemic corticosteroids (72%), and less than a third of patients in the pooled analysis required immunosuppression (27%; 95% CI 15% to 40%). The irAEs rates were similar when patients with combination therapy were removed from the analysis (Supplementary Table 5).

Table 3.

Effects associated with ICIs in patients with pre-existing autoimmune diseases in the included studies.

	Number of studies with these data	Number of patientsa	Pooled incidence or relative risk (95% CI)
Patients not experiencing irAEs	68	23430	35.7% (28.5% to 43.3%)
Risk of irAEs (autoimmune vs no autoimmune disease), relative risk	21	2399 vs 20348	1.3 (1.0 to 1.6)
Risk of irAEs in patients receiving treatment for autoimmune disease vs no treatment), relative risk	4	183	1.1 (CI 0.72–1.6)
irAEs (any)	68	4350	61.0% (54.3% to 67.5%)
Grade of irAEs
1–2	29	966	50.1% (39.7% to 60.5%)
3–4	25	713	29.6% (20.6% to 39.4%)
Management of irAEs
Corticosteroids	35	478	71.6% (59.2% to 82.9%)
Immunosuppressants/biologics	26	570	26.5% (14.7% to 39.7%)
irAEs per type of ICI agent
Anti–PD-1/PD-L1	3	74	25.2% (15.4% to 36.3%)
Anti–CTLA-4	4	98	37.7% (22.2% to 54.4%)
De novo irAEs (not flares)	59	1827	22.9% (16.3% to 30.0%)
Patients with irAEs requiring hospitalization	6	103	31.7% (13.8% to 52.0%)
Permanent ICI discontinuation after irAEs	48	968	34.9% (24.4% to 46.0%)
Deaths related to irAEs	43	943	0.07% (<0.01% to 3.9%)
Flares (any autoimmune disease)	74	2753	36.3% (29.5% to 43.3%)
Median time to flare after ICI initiation, range	25	687	7 to 470 days
Flares per type of autoimmune disease (most commonly reported)
Inflammatory bowel disease	10	480	36.9% (28.5% to 45.7%)
Rheumatoid arthritis	13	789	35.9% (22.2% to 50.4%)
Psoriasis/psoriatic arthritis	13	560	38.9% (22.9% to 55.8%)
Autoimmune thyroiditis (any)	10	458	15.4% (8.2% to 23.8%)
Multiple sclerosis	8	291	18.0% (<1% to 49.3%)
Flares per use of DMARDs/corticosteroids at ICI start	1	47	29.4% (9.7% to 53.6%)
Complete responders (autoimmune vs no autoimmune disease), relative risk	3	94 vs 780	1.2 (0.47 to 2.9)
Partial responders (autoimmune vs no autoimmune disease), relative risk	5	172 vs 1179	1.1 (0.83 to 1.3)
Progressive disease (autoimmune vs no autoimmune disease), relative risk	6	129 vs 629	1.4 (0.79 to 2.3)

DMARD, disease-modifying antirheumatic drugs

^aNumber of patients in studies reporting the data.

Flares

The median time to flare after ICI initiation ranged from 7 to 470 days. Flares were most commonly reported in patients with psoriasis/psoriatic arthritis (39%), inflammatory bowel disease (37%), rheumatoid arthritis (36%), multiple sclerosis (18%), and any type of autoimmune thyroiditis (15%). The pooled incidence of flares for people using immunosuppressive agents or corticosteroids at the start of ICI therapy was 29% (95% CI 10% to 54%).

Cancer response to ICI in patients with cancer and pre-existing autoimmune diseases

Rates of complete response after treatment with any ICI did not significantly differ between patients with autoimmune disease and patients without autoimmune disease (relative risk 1.2, 95% CI 0.47 to 2.9). Similarly, no significant differences between those groups were observed in the number of partial responders or patients with progressive disease (Table 3).

Risk of bias across studies

Supplementary Figure 2 shows the funnel plot of the occurrence of flares in the reported studies. The plot looks asymmetrical, with greater numbers of studies with less variability reporting lower occurrence rates and studies with higher variability reporting higher occurrence rates. The Egger test for small-study effects was significant (p=0.006), indicating smaller studies reporting larger effect sizes than the larger studies. However, when studies with sample sizes less than 5 patients (33%, 95% CI 27% to 40%; 58 studies; Egger test p=0.11) and less than 10 patients (29%, 95% CI 24% to 36%; 41 studies, Egger test p=0.66) were removed, the pooled occurrence rates remained similar.

DISCUSSION

We synthetized current published observational studies reporting on the use of ICI in patients with cancer and pre-existing autoimmune disease. We found that irAEs of any type were commonly reported in more than two thirds of the patients, with autoimmune disease flares reported in over a third of the patients and de novo irAEs in over 20% of the patients. The frequency of flares appeared to differ according to the type of autoimmune disease and the ICI received. Most studies reported flares or de novo irAEs that were mild to moderate with low rates of discontinuations and a rate of less than 0.1% for deaths related to irAEs. However, our findings also suggest that these patients required careful monitoring given that a third required hospitalization due to irAEs. We also found that responses to the ICI treatment were similar between cancer patients with and without pre-existing autoimmune disease.

Our systematic review differs from other comprehensive reviews published in the literature (Supplementary Table 2).^7,141–162 We found 23 reviews with information on this topic, with the number of included studies ranging from 5 to 53. Eleven were narrative, 12 were systematic reviews and of these, 5 included meta-analyses (2 out of the 5 included non-autoimmune comorbid conditions). None of the search strategies reported were built as a broad search to cover any type of publication reporting on adverse events with ICIs like ours. Although our method increased the number of citations to screen, which we leveraged using machine learning screening methods, our search proved more sensitive and retrieve 95 relevant citations. This facilitated the exclusion of conditions that are not autoimmune (e.g., gout). We excluded studies in which it was unclear if all patients had an autoimmune condition such as interstitial lung disease and thyroid dysfunction; inclusion of this type of disorders could have overestimated the rates of flares or de novo irAEs, given the frequency of these disorders (such as subclinical hypothyroidism), which are also found in the general population, and in fact only few in this category can be labeled as autoimmune disease.¹⁴³ Nonetheless, our pooled estimate for incidence of flares was similar to those reported in prior meta-analyses (range 23% to 37%).^143,161,162 Previously, we reported a review of 123 patients in 49 publications with 92 (75%) reporting an exacerbation of autoimmune disease, irAEs, or both.⁷ In our previous report, more than 50% of the total number of cases had disease exacerbation, more than a third had de novo irAEs, and 9% irAEs were fatal. In our current review, the frequency of irAEs in patients with autoimmune disease was lower. The differences could be due to increased reporting of poor outcomes in case reports, versus more unbiased reporting in series that systematically identified patients with autoimmune disease receiving ICIs, which is the study design for publications included in the current review.

Our study also reports findings according to different autoimmune diseases, which is more clinically relevant, given the different presentations of each autoimmune condition. A broad spectrum of autoimmune diseases was reported, including psoriasis and/or psoriatic arthritis, rheumatoid arthritis, autoimmune thyroid disease, ulcerative colitis, Crohn disease, multiple sclerosis, myasthenia gravis, and sarcoidosis. Previous comprehensive reviews included meta-analysis according to organ system rather than disease.^143,160 Another meta-analysis found that patients undergoing immunosuppression had a similar risk of flare compared to those without immunosuppression.¹⁶¹ When the authors compared the risk of developing flares between rheumatoid arthritis and other autoimmune diseases, the only significant difference observed was between rheumatoid arthritis and autoimmune thyroiditis after removal of the study with a higher proportion of flares reported for autoimmune thyroiditis.¹⁶¹ A prior systematic review including 1 observational study and 3 case reports also found that the frequency of irAEs differed by type of ICI. In our study, we found that flares were more commonly reported with anti–CTLA-4.¹⁴⁵

Previous studies have also reported no differences in cancer outcomes between patients with and without autoimmune disease,¹⁴⁴ except when the data were analyzed by type of cancer in patients with interstitial lung disease.¹⁴³ Patients with non–small cell lung cancer and pre-existing interstitial lung disease were more likely to achieve complete response than those without interstitial lung disease.¹⁴³ Another study pooled the objective response rate (combining complete, partial, and stable response and progressive disease) and found that patients with flares had better objective response rates than those without flares.¹⁶⁰

The results of our systematic review and meta-analysis were limited by the available data. Although our search was run recently, we acknowledge the potential of missing newly indexed articles. This is because of the nature of the research question, with multiple studies being published rapidly and continuously. There is also the possibility of newer studies being more methodologically rigorous, which could impact the confidence in our findings. The current state of the evidence is derived from observational/pharmacovigilance studies. However, we are confident that our estimates are robust, and new evidence from observational studies could only change the magnitude of the occurrence.

While some of our results corroborate existing knowledge, the methodological rigor, information on hospitalization rates due to irAEs, inclusion of the most recent ICI approved agents, and the detailed breakdown by type of autoimmune disease add a layer of depth to the current understanding. These enhancements help refine clinical approaches and inform future research directions. We believe these aspects underscore the value of our study and hope that this clarification helps emphasize how our findings contribute to the broader scientific and clinical discourse on the treatment of cancer patients with preexisting autoimmune diseases using ICIs.

Given the high occurrence rate of irAEs, oncologists should implement enhanced monitoring protocols for patients with pre-existing autoimmune diseases undergoing ICI therapy. 1) A thorough baseline assessment of the patient’s autoimmune condition before ICI initiation, could help anticipate the likelihood and severity of flares and potentially guide the choice and timing of ICI therapy, balancing cancer treatment efficacy with potential autoimmune exacerbations. 2) An integrated care approach during ICI treatment would help promptly address irAEs, optimize immunosuppressive therapy, and ensure continuity of cancer care. 3) The risk-benefit ratios of immunotherapy in patients with pre-existing autoimmune diseases needs to be carefully discussed with those patients. Patients should be thoroughly informed about the risks of irAEs, potential symptoms, and the importance of prompt reporting of any changes in their health, and the potential for modifications to their immunosuppressive or cancer treatments. Educated patients are better prepared to manage their conditions in partnership with their healthcare providers. 4) Oncology facilities should be prepared for the potential need for a rapid response to severe irAEs, including immediate access to autoimmune disease treatment interventions and hospital care.

Our results can assist in providing evidence-based data to patients with cancer and pre-existing autoimmune diseases. We combined data from observational studies and uncontrolled trials and found that the rate of flares and de novo irAEs was substantial, but these events were generally mild in most patients. Tumor response rates in patients with autoimmune disease were similar to those in patients without autoimmune disease. In conclusion, immune checkpoint blockade in patients with known autoimmune diseases is possible but requires careful monitoring due to the potential need of hospitalization due to irAEs. Our findings provide a crucial foundation for oncologists to refine their monitoring and management strategies, ensuring with more frequent clinical assessments, closer communication with other autoimmune disease specialists, and patient education about signs and symptoms of irAEs that the benefits of ICI therapy are maximized while minimizing its risks.

Highlights.

• Having an autoimmune disease (AD) increases the risk of irAEs.

• Over 1/3 of patients with AD on ICIs will have AD flares, but most are mild.

• Those with AD and irAEs need close monitoring as over 1/3 will need hospitalization.

• Studies are needed to quantify irAEs by cancer type and AD treatment at ICI start.