The use of immune checkpoint inhibitors (ICIs) has been increasingly prevalent in cancer treatment. ICIs target immune checkpoint molecules that regulate immune cell activation, such as cytotoxic T lymphocyte associated antigen-4 (CTLA-4) and programmed cell death-1 (PD-1). The binding of CD80/86 to CTLA-4 and PD-1/PD-L1 normally inactivates T-cells and blocks pro-apoptotic pathways, eventually leading to effector T cell exhaustion (1). By overexpressing these ligands, cancer cells block T cell activation and evade the immune system. ICIs block these interactions and allow for immune overactivation to take place. Unfortunately, this same mechanism also blocks the normal negative signaling of T cells, leading to autoimmune-like toxicities known as immune-related adverse events (irAEs) (1).

irAEs can affect nearly every organ system and be life-threatening. While most irAEs are reversible with steroid treatment, some are irreversible. Endocrinopathies account for 4-30% of reported irAEs and are usually permanent (1). The most common immune-mediated endocrinopathy is hypothyroidism, which is present in about 10% of patients treated with anti-PD-1 monotherapy, but easily treatable (1). In contrast, ICI-induced type 1 diabetes mellitus (T1DM) is rare, but associated with significant long-term morbidity.

The mechanism of ICI-induced T1DM is not fully understood. Both autoimmune and ICI-induced T1DM are thought to be caused when autoreactive CD4+ T cells, CD8+ T cells, and B lymphocytes destroy the insulin-producing beta cells in the islets of Langerhans. ICI-induced T1DM is only sometimes associated with positive T1DM-associated autoantibodies, suggesting a unique pathophysiology (2). Common islet cell protein targets of these autoantibodies include glutamic acid decarboxylase (GAD), insulin (IA), insulinoma associated antigen-2 (IAA), zinc transporter 8 (ZnT8), and islet cells (ICA). Given the serious and irreversible nature of ICI-induced T1DM, it is critical to identify biomarkers to predict risk of this rare complication. While genetic associations with human leukocyte antigens (HLA) have been documented in autoimmune T1DM, HLAs specific for irAEs are an exciting new field (3, 4). Predictive biomarkers of ICI-induced T1DM have also been proposed but are not yet fully understood (5).

Despite several published cohorts and case reports of ICI-induced T1DM, impact on treatment outcomes remains unclear given the small number of patients reported thus far (2, 6–12). About half of the patients in these cohorts were treated for melanoma. Nearly all received anti-PD1 or PD-L1 therapy for active treatment in stage IV disease as well as adjuvant therapy in other stages. The clinical course in these cohorts is well described. On average, the time to T1DM diagnosis is approximately 23 weeks, 60% of patients present in diabetic ketoacidosis, 40% of patients have positive autoantibodies, and 80% of patients have low or undetectable c-peptide (2, 6–12). However, treatment response data is reported for only a few studies, with variable response rates. Four cohorts reported positive treatment response data (7–10). These cohorts showed objective response rates (ORR) of up to 80-70% in patients with melanoma although results varied (7–10). These data suggest greater radiographic clinical response in melanoma patients with ICI-induced T1DM compared to historical controls (reported ORR 30-50%). To better characterize clinical outcomes, we reviewed all cases of ICI-induced T1DM at our institution. Here we report clinical and proteomic data for this large cohort of patients after long-term follow-up.

This was a single-center case series of adult patients with solid tumors at Huntsman Cancer Institute at the University of Utah Health. Electronic medical records were searched for patients who received ICIs (ipilimumab, pembrolizumab, nivolumab, cemiplimab, atezolizumab, durvalumab and avelumab) during an 12-year period (2011 to 2023). To identify T1DM diagnoses after ICI therapy, the enterprise data warehouse (EDW) at the University of Utah used ICD codes for T1DM (ICD-10-CM E10 and ICD-9-CM 250.01) or obtained C-peptide levels measured after starting ICI therapy. A chart review of these cases was then conducted to confirm diagnosis of ICI-induced T1DM.

Chart review confirmed ICI-induced T1DM cases if they met the following inclusion criteria: sustained insulin dependence with a low or undetectable C-peptide level through longitudinal follow-up after high dose steroids were stopped if they were started. Patients were excluded due to the following reasons: 1) pre-existing T1DM diagnosis, 2) C-peptide drawn for other reasons in non-diabetic patient, 3) non-insulin-dependent worsening of preexisting T2DM, 4) new T2DM diagnosis, 5) worsened hyperglycemia related to steroid use and 6) insufficient data.

Blood chemistry and complete blood count prior to ICI therapy, at six weeks prior to T1DM development, and three weeks prior to T1DM development were available for 11 of 18 patients. Lab data was analyzed by two-tailed, two-sample t-tests for significant changes from baseline to investigate possible association with the development of T1DM.

Through an IRB-approved protocol, peripheral blood was collected from patients with melanoma prior to starting immunotherapy (baseline levels). Plasma was then isolated and stored at –80C. Baseline plasma from 2 patients who later developed ICI induced T1DM and 5 patients who did not have any irAE were available for study and sent for a multiplexed Proximity Extension Assay (PEA), developed by Olink Proteomics (Uppsala, Sweden). This assay was used to measure protein abundance directly in the plasma obtained from patients. In brief, the PEA is a dual recognition assay, where each protein is recognized by oligonucleotide-labeled antibody probes, then amplified and quantified by PCR. The baseline abundance levels of 1472 plasma proteins were analyzed using 4 pre-defined PEA-panels (Olink® Multiplex Inflammation, Oncology, Cardiometabolic and Neurology). The results were analyzed using the Olink® proteomics program that can be found at https://olink.com/products-services/data-analysis-products/olink-statistical-analysis-app/.

To our knowledge, this is the largest reported case series describing oncologic outcomes, clinical attributes, and proteomic data from patients that developed ICI-induced T1DM at a single institution. While melanoma patients had excellent outcomes (100% ORR), in other solid tumors, outcomes were mixed, with most patients having progression of disease. All patients received anti-PD-1 therapy (nivolumab, cemiplimab or pembrolizumab) either as sole immunotherapy or, for 3 patients, as nivolumab in combination with ipilimumab. This is consistent with prior literature documenting anti-PD-1 therapy as opposed to anti-CTLA-4 (ipilimumab) therapy having a stronger predilection for disease (13). The mechanism underlying this observation is complex, but prior animal studies have shown that the PD-1/PD-L1 axis is more important to the self-tolerance of pancreatic beta cells than CTLA-4, given the low rates of ICI-induced T1DM from ipilimumab (14).

Our case series had important clinical implications. Patients began to show significantly elevated glucose levels 3 weeks prior to T1DM onset compared to pre-ICI baseline ( Figure 3 ). This is a unique finding not reported in prior studies of predictive biomarkers, or prior case series (2, 5–12). Of note, glucose levels were not significantly elevated 6 weeks prior to presentation. This finding, taken in conjunction with some patients presenting in DKA with A1c <6.5%, suggests that the progression of pancreatic beta cell destruction and resulting glucose intolerance is very rapid, with development of the disease in a short period of time, in the span of days to weeks. Any significant increase in glucose on routine testing should prompt further evaluation or, at minimum, increased monitoring for ICI-induced T1DM. Early diagnosis is the only way to prevent further complications such as DKA. Through duration of follow-up, all patients required insulin therapy, highlighting the known irreversibility of this irAE (15). Interestingly, despite a lack of contraindication to resuming therapy, most patients did not receive further ICI therapy after developing T1DM. It is important for clinicians to recognize that ICI-induced T1DM should not preclude patients from further ICI therapy if indicated.

There are several proposed distinct subtypes of ICI-induced T1DM (16). Fulminant T1DM, characterized by rapid onset of T1DM presenting with DKA, usually presents with normal or near normal A1c (16). Two of our patients with A1c <7 may represent this subtype. Another proposed subtype is a phenotype like “decompensated” type 2 diabetes, which presents with a high A1c and detectable C-peptide, but new insulin dependency with or without prior history of type 2 diabetes. Five of our patients presenting with high A1c and low C-peptide levels were consistent with this group. One would expect positive autoantibody prevalence reflecting the proposed mechanism, with a more frequent prevalence in the fulminant variety, and less prevalent in a phenotype like type 2 diabetic decompensation. Another proposed cause is secondary to prior ICI induced pancreatitis, which is possible in our patient with an elevated lipase. Few patients had a lipase level checked, and it would be helpful in future workup. In our case series, only 3 of the 10 tested patients had positive autoantibodies which is similar prevalence to other published cohorts (30% vs 40-50%). It might be explained that most of the subtypes we observed were more like decompensated T2DM rather than a fulminant T1DM which could be present in slightly higher abundance in other studies (2, 6–12). Prior cohorts have suggested some correlation between the number of positive autoantibodies and shortened time from treatment to onset of T1DM (10). The three patients in our cohort did not support any clear conclusions, having time to onset comparable to our average.

Thirteen of 18 patients (72%) experienced at least 1 other irAE, with most experiencing multiple irAEs. Endocrinopathies are relatively common in patients with ICI-induced T1DM. One systematic review found ICI-induced thyroid dysfunction in 24% of cases (2). Eight in our cohort (44%) had other endocrine irAEs, comparable to published literature. It is not clear if prior severe irAEs are a predictor of future severe irAEs (T1DM). Only three of 18 patients had developed a severe (grade 3 or higher) irAE prior to their diagnosis of T1DM. Severe irAEs have been previously cited in one large case series as a poor prognostic factor, as patients are not often able to resume their ICI therapy (17). In our cohort, this was not the case, as none of these patients had progression of disease.

While limited by small sample size, our PEA analysis suggested several interesting relative differences. Decreased levels of proteins involved in inflammatory pathways such as IL-6, OSMR, and AMBP in patients who developed ICI-induced T1DM could help support the excellent response. IL-6 and OSMR play a role in tumor proliferation by activating transcription factors. Therapies targeting these pathways have been shown to be inhibitory to tumor growth (18). The IL-6 signaling pathway has also been described as promoting self-tolerance of pancreatic beta cells in autoimmune T1DM, although this is not well understood (19). Increased expression of proteins in metabolic pathways like GLO1 and DXCR were seen in T1DM patients without clear implications. GLO1 has been shown in several preclinical models to promote tumor cell proliferation, and one study showed GLO1 upregulating PD-L1 expression, suggesting positive treatment response when it is targeted (20, 21). Finally, increased levels of cell signaling proteins such as NPTN and ENPP5 were also noted. ENPP5 has been cited as a biomarker for increased insulin resistance in non-insulin dependent diabetes (22). NPTN is described as an important factor in T-cell activation, helping explain excellent response (23). This interpretation is entirely speculative and subject to confounding factors between patients but could be a clue for future larger studies with subject matching that is required for validation.

Despite ICI-induced T1DM, patients with melanoma displayed excellent clinical response and survival. No patient in this group experienced either progression of disease in the active treatment group or recurrence in the adjuvant group. In comparison, the other malignancies group only had two on active therapy that did not progress. Prior case series note a favorable response in melanoma compared to around 50% response rates to anti-PD-1 therapy in melanoma in general (24). More specifically immunotherapy trials involving patients with advanced (stage III or IV), unresectable disease noted a PFS over a comparable time period in the 40-50% range (25, 26). While patients with lower recurrence stage IIIA or IIIB disease need longer follow up times to determine clinical outcome, the 5 patients with metastatic disease all responded and none have relapsed with a mean follow up time of 39 months. Reasons for this excellent clinical response are difficult to determine although proteomic data may help guide future research. While there is a known association between severe irAEs and good clinical outcomes, in many instances this may be due to immortal time bias (27).

While our search criteria were broad, and our prevalence falls within the expected range, it is possible some patients with ICI-induced T1DM were not included. There was a patient identified during our EDW based screening who could not be included due to lack of documentation of lab testing needed to validate the diagnosis as ICI-induced T1DM. Other patients could have been missed by our search entirely, as our institution is a referral center for one of the country’s largest geographic catchments, and many patients were more likely to be hospitalized acutely closer to home. This also highlights the necessity of further diagnostic testing when patients present with possible irAEs or suspicious history. Further diagnostic testing of great interest would be HLA typing in these patients, which was not available in our study. Our case series is biased heavily towards melanoma, and it is difficult to see trends with other malignancies. While the response in melanoma patients was excellent, it is limited by the small size of the cohort. The follow up time for our patients with stage IIIA or IIIB melanoma may not be long enough to see a difference given the long PFS known for these stages.

As a point of future study, germline, HLA, and immunophenotypic data could be helpful in providing insight and predictive value in determining which patients are at higher risk of ICI-induced T1DM. Clinicians should be aware of the rapid onset and irreversibility of ICI-induced T1DM, with evidence of elevated blood glucose 3 weeks prior to diagnosis. Positive treatment outcomes in patients with melanoma who develop ICI-induced T1DM may help guide treatment considerations. Larger studies are needed to elucidate strategies to identify high-risk patients and improve treatment outcomes for this rare irAE.

The original contributions presented in the study are included in the article/supplementary materials. Further inquiries can be directed to the corresponding author.

The studies involving human participants were reviewed and approved by 00138167. The patients/participants provided their written informed consent to participate in this study.

JM and JPM identified cases. JM and SH-L verified and analyzed patient clinical data. SH-L and MK analyzed proteomic data. JM and JPM drafted the manuscript. All authors contributed to the article and approved the submitted version.