A higher eosinophil count before immune checkpoint inhibitor therapy was associated with thyroid dysfunction afterward, according to a retrospective cohort study. The result points to eosinophil count as a candidate risk marker, but it does not show that the count causes thyroid dysfunction or that the association can predict what will happen to any individual patient.
The investigators set out to identify pretreatment characteristics associated with thyroid dysfunction after treatment. They reviewed medical records of patients receiving immune checkpoint inhibitors, then compared those who developed thyroid dysfunction with those who did not.
The association after LDH adjustment
One of the study's main results came from multivariable logistic regression, a statistical model used to account for other factors. After adjustment for lactate dehydrogenase, or LDH, each 0.1 × 10³/µL increase in pretreatment eosinophil count was associated with an adjusted odds ratio of 2.18 for thyroid dysfunction. The 95% confidence interval was 1.56 to 3.04, and P was less than 0.001.
An odds ratio is a way of comparing odds, rather than a direct percentage of patients affected. Here, the reported 2.18 applies to each stated increase in eosinophil count after LDH adjustment. It describes the pattern in the reviewed cohort, not a guaranteed outcome for an individual.
A cutoff separated the two groups
Researchers also tested how well the pretreatment count separated patients who developed thyroid dysfunction from those who did not. Receiver operating characteristic analysis, which measures how well a marker separates two groups, produced an area under the curve of 0.812. The 95% confidence interval ranged from 0.722 to 0.902.
The reported optimal cutoff was 0.2 × 10³/µL. At that threshold, sensitivity was 81.0% and specificity was 68.2%. In practical terms, the cutoff identified 81.0% of patients in the thyroid-dysfunction group and correctly classified 68.2% of those in the group without thyroid dysfunction within this analysis.
Those figures describe performance within the cohort that was studied. The authors' recommendation for prospective studies reflects the need to determine whether the marker has clinical utility beyond this reported analysis.
The marker still needs prospective testing
The authors recommend prospective studies that also include thyroid autoantibody status to determine whether eosinophil count has clinical utility as a candidate marker. That would address a central uncertainty left by the retrospective design: whether the association persists when the marker is evaluated prospectively and whether antibody information adds value.
The study therefore supports an association and the reported discrimination metrics in this cohort, not causality or validated clinical utility. On the supplied evidence, eosinophil count remains a candidate marker rather than a validated stand-alone prediction tool.
Research oversight and disclosures
The research received approval from the Ethics Committee of the Faculty of Medicine, University of Tokyo, under approval number 2529. Because the study was retrospective, the committee waived informed consent, and an opt-out method was implemented.
The authors declared no competing interests. The manuscript disclosed that AI tools, including large language models and machine-learning software, were used for preparation, analysis and presentation. Study data are available from the corresponding author upon reasonable request.
The article page identifies the work as an original article published on 21 August 2026.
Paper data and sources
Original title: Risk factors for immune checkpoint inhibitor-related thyroid dysfunction: a retrospective cohort study.
Authors: Takahiro Amemiya, Hiroshi Suzuki
Journal/Repository: Endocrine
Status: Peer-reviewed
First online: 2026-08-21
DOI: 10.1007/s12020-026-04758-6
Original paper · Full text