Article Text

Download PDFPDF

1238 Risk evaluation of immune checkpoint inhibitor diabetes through islet autoantibodies & HLA types in a large, real-world cohort
  1. Zoe Quandt1,
  2. Samantha Liang2,
  3. Anastasia Lucas3,
  4. Marshall Thompson4,
  5. Ashis Saha5,
  6. Christian Hammer5,
  7. Kevan Herold6,
  8. Arabella Young7,
  9. Christine Spencer8 and
  10. Mark Anderson1
  1. 1University of California San Francisco, San Francisco, CA, USA
  2. 2Parker Institute for Cancer Immunotherapy, South San Francisco, CA, USA
  3. 3University of Pennsylvania, Philadelphia, PA, USA
  4. 4Parker Institute for Cancer Immunotherapy, Durham, NC, USA
  5. 5Genentech, South San Francisco, CA, USA
  6. 6Yale University, New Haven, CT, USA
  7. 7Huntsman Cancer Institute, University of Utah, Salt Lake City, UT, USA
  8. 8Parker Institute for Cancer Immunotherapy, Traverse City, MI, USA
  • Journal for ImmunoTherapy of Cancer (JITC) preprint. The copyright holder for this preprint are the authors/funders, who have granted JITC permission to display the preprint. All rights reserved. No reuse allowed without permission.


Background From 2019–2022, we prospectively enrolled 1200 pan-cancer patients treated with immune checkpoint inhibitors (CPI) and identified three CPI-diabetes mellitus (DM) cases as part of the Radiohead study.1 Baseline and on-treatment blood samples were collected. From this longitudinal cohort, we evaluated whether islet autoantibodies and HLA typing could be used to predict a patient’s risk of developing CPI-DM.

Methods For islet autoantibody analysis, serum for 829 patients with both a baseline and early on-treatment sample were tested for GAD, IA-A, mIAA and ZnTA-8 autoantibodies using a radiobinding assay. HLA types were imputed from SNP genotypes generated using an Illumina GSAv3 array, using HIBAG2 and population-specific reference panels for 1029 patients. Fisher exact was used for autoantibodies and high-risk type 1 DM HLA types (DR3, DR4, DQ8, DQ2) associations. Logistic regression was used for combined HLA type and autoantibody risk prediction.

Results Of the 829-patient autoantibody subset, GAD was the most common autoantibody at 5.31%, followed by insulin at 3.74%. Two of the three (66%) CPI-DM patients had positive GAD, one prior to CPI initiation and the other emerging early on treatment. For every 35 patients with pre-treatment GAD positivity, one would be projected to develop CPI-DM; this increases to one in 19 patients if GAD positive is determined at either pre or early on treatment. The presence of GAD at baseline and/or early on treatment but not the other autoantibodies was associated with CPI-DM development (relative risk 34.84, p=0.004).

All three CPI-DM subjects were carriers of known HLA risk genotypes for T1DM, compared to 50% of patients in the remaining cohort of 1026 patients. The DR4-DQ8 risk haplotype was present in 2 out of 3 CPI-DM patients (66%), in contrast to 18% in the rest of the cohort. There was insufficient statistical power for association testing, due to the low number of cases.

In the 769-patient combined cohort, for which both HLA type and autoantibody status was determined, presence of GAD at baseline (odds ratio 21.8, p=0.01) was predictive of CPI-DM development after adjustment for HLA DR4-DQ8; results were similar with adjustment for other individual T1DM high risk HLA types.

Conclusions Pre-treatment and/or early on treatment analysis of T1DM autoantibodies and HLA types have the potential to predict development of CPI-DM and might be a critical first step in risk stratification for this rare immune-related adverse event. Given the very small number of cases, these findings must be further validated in high-risk cohorts.

Acknowledgements We would like to thank the Juvenile Diabetes Research Foundation and the Leona M. and Harry B. Helmsley Charitable Trust for their support as well as the patients who participated in our study.


  1. Lucas A, Liang S, Silva DD, et al. 1256 Relationships between irAEs and checkpoint inhibitor response in RADIOHEAD: a large prospective pan-tumor cohort of standard-of-care patients with clinical annotation, molecular and immune profiling. JITC. 2022;10:1256

  2. Zheng X, Shen J, Cox C, et al. HIBAG—HLA genotype imputation with attribute bagging. Pharmacogenomics J. 2014;14:192–200

Ethics Approval Patients provided informed consent to for this study which was sponsored by the Parker Institute for Cancer Immunotherapy and was reviewed by WIRB under protocol # 20182579.

This is an open access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited, appropriate credit is given, any changes made indicated, and the use is non-commercial. See

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.