ICIs used in our clinic are cytotoxic T lymphocyte antigen-4 (CTLA-4) inhibitors (ipilimumab), programmed cell death protein 1 (PD-1) inhibitors (nivolumab and pembrolizumab), and programmed cell death protein 1 ligand 1 (PD-L1) inhibitors (atezolizumab, durubumab, and avelumab). This is a single-center, retrospective, observational study of a total of 466 patients who were treated with ICI at Kawasaki Medical School Hospital from September 1, 2014, to May 31, 2022. Details are as previously described (4). The study protocol, including opt-out informed consent, was approved by the Institutional Review Board of Kawasaki Medical School (No. 5726-00). The study was conducted by the principles of the Declaration of Helsinki. A flowchart of the participants in this study is shown in Figure 1 . Of the 466 patients who were treated with ICIs at our hospital during this period, 19 patients whose thyroid hormone levels were not measured after ICI administration and 25 and 4 patients diagnosed with hypothyroidism or Graves’ disease, respectively, at the time of ICI administration were excluded from the study. According to Common Terminology Criteria for Adverse Events (CTCAE) version 5.0 (http://www.jcog.jp/doctor/tool/CTCAEv5J_20220901_v25_1.pdf), the patients with hypothyroidism were classified as follows: 57 patients with grade 1, 54 patients with grades 2-4, 305 patients with normal thyroid function (NTF), and 3 patients with central hypothyroidism.

Thyroid hormone levels were measured in 419 subjects in all participants. The method of measuring thyroid hormones is as previously described (4). All blood tests after ICI administration were checked and classified as NTF if the TSH was within the institutional reference range. When there was a transient or permanent increase in TSH to 4.12-10.0 μIU/mL and FT4 was not high at that time, the patient was classified as CTCAE grade 1; when TSH was above 10.0 μIU/mL and FT4 below 0.68 ng/mL, or when levothyroxine replacement therapy was started due to hypothyroidism, the patient was classified as CTCAE grade 2-4. Transient thyrotoxicosis was defined as TSH less than measurement sensitivity and FT4 > 1.26 ng/mL that improved spontaneously without antithyroid therapy. The study included 32 participants with transient thyrotoxicosis; 24 participants had Graves’ disease ruled out by thyroid-stimulating antibodies or by evaluation with thyroid ultrasound. None of the participants in this study presented with permanent thyrotoxicosis.

Data are expressed as mean and standard deviation. The primary endpoint was to assess the correlation between the severity of thyroid-related irAEs and the efficacy of immunotherapy. The secondary endpoint was to determine the clinical characteristics of thyroid-related irAEs. The prevalence of thyroid-related autoantibodies in NTF and grade 1 and grades 2-4, the remission rate by RECIST (Response Evaluation Criteria in Solid Tumors) criteria, and the prevalence of malignancy by stage were evaluated with a chi-square test. The frequency of thyroid-related irAEs by each ICI was also used as a chi-square test. Differences in thyroid hormone levels between patients with and without thyroid-related irAEs were evaluated using ANOVA and the Tukey method as a post hoc test. To assess one-year survival by the presence of thyroid-related irAE, patients were divided into groups by the presence or absence of thyroid-related irAE within one year after ICI administration, and calculated by chi-square test for survival or death at one year after ICI administration. The mean observation period with and without thyroid-related irAEs during the study period was calculated using nominal logistic analysis adjusted for age, sex, stage of cancer, and ICI used. The observation period was calculated as the number of days alive from the date of the first ICI administration to March 31, 2023. Patients who were transferred during the treatment was calculated as survivors, with the observation period censored at the date of the last visit. When the date of death was known even after the patient was transferred to a hospital, the date of death was used as the end date of observation. JMP version 16.0.2 (SAS Institute Inc. North Carolina. USA) was used for all statistical analysis, and Microsoft Excel for Mac version 16.71 (Microsoft Co. WA. USA) was used to create tables.

The mean age of the study participants was 68.2 ± 10.3 years, and the percentage of male participants was 72.6%. There was no significant difference between participants with abnormal thyroid function and NTF (p=0.48, 0.56, respectively). ICIs were administered 1.4 ± 3.8 times in NTF, 2.6 ± 5.3 times in Grade 1, and 3.5 ± 4.9 times in Grades 2-4, with a significantly higher number of times in Grades 2-4 compared to NTF (p=0.0024). The frequency of thyroid dysfunction by ICI and the primary disease of the participants are shown in Table 1 . The frequency of thyroid dysfunction in the study participants was 28.2%. The incidence induced by the drug was 54.2% for ipilimumab/PD-1 inhibitor combination therapy, 26.7% for PD-1 inhibitors, and 26.0% for PD-L1 inhibitors, with the highest incidence in participants receiving combination therapy (p = 0.014). There was a significant difference in the proportion of primary disease for which each ICI was administered (p<0.0001). Table 2 shows the treatment details at the time of ICI administration for participants who developed thyroid-related irAE. Of the participants who developed Grade 2-4 hypothyroidism, 61.1% were treated with ICI alone, and 38.9% were also treated with some other anticancer agent besides ICI at the same time. Trends of TSH levels of NTF, grade 1, and grades 2-4 for six months after ICI administration are shown in Figure 2 . TSH levels were significantly higher in Grade 1 and Grades 2-4 when treated with ICI compared to NTF (p<0.0001, p<0.0001, respectively). Transient thyrotoxicosis occurred in 8 patients (2.6%) with NTF, 4 patients (7.0%) with grade 1, and 20 patients (37.0%) with grades 2-4. The mean time from ICI administration to the diagnosis of irAE was 215.9 (95%CI: 153.2-278.7) days for grade 1, and 165.6 (95%CI: 102.9-228.4) days for grades 2-4. In grades 2-4, levothyroxine was administered to 47 patients (82.5%). The positive rates of TgAb and TPOAb in each group are shown in Table 3 . The positive rates of TgAb and TPOAb increased in line with the increase in CTCAE grade, with transient thyrotoxicosis having the highest positive rate (p<0.0001). The mean dose of levothyroxine was 67.1 μg/day (95% CI: 0-200). Four of the 47 patients were able to stop treatment with levothyroxine, and the time to stop was 385.5 days (95%CI: 87-712).

Figure 3A shows the survival rates for each group within 1 year after ICI administration: 43 patients had a grade 1 thyroid-related irAE and 47 patients had a grade 2-4 thyroid-related irAE within 1 year after ICI administration. The 1-year survival rates for participants who developed grade 1 and grade 2-4 hypothyroidism within 1 year after ICI administration were 64.9% and 88.9%, respectively. On the other hand, the 1-year survival rate for participants who had NTF within 1 year after ICI administration was 52.1%, with a significantly longer observation period in those who developed thyroid disorders (p<0.0001). Next, the mean observation period in the participants during the entire post-ICI observation period was significantly longer: grade 2-4 861.8 (95%CI: 726.9-996.7) days and transient thyrotoxicosis 937.8 (95%CI: 760.6-1114.9) days compared to NTF 479.2 (95%CI: 422.5-536.0) days (p<0.0001, <0.0001, respectively) ( Figure 3B ). Cancer stage at the time of ICI administration did not differ among the groups (p=0.68) ( Figure 3C ). Nevertheless, the remission rate assessed by RECIST criteria was significantly higher in grades 2-4 compared to NTF (p<0.0001) ( Figure 3D ).

Past studies have reported possible gender differences in the incidence of irAE (6). Therefore, we evaluated the difference in one-year survival rate and treatment efficacy by gender. In the subjects of this study, there was a trend toward a higher survival rate after one year of ICI administration in both male and female patients with thyroid-related irAE ( Figure 4A ). And in both men and women, there was a trend toward significantly higher remission rates after ICI administration in participants with thyroid-related irAE ( Figure 4B ). Disease frequencies by gender and ICI used are shown in Table 4 . There was a significant difference in disease frequency between male and female who received each ICI (ipilimumab and PD-1 inhibitors: p=0.032, PD-1 inhibitors: 0.071, PD-L1 inhibitors: 0,0038). Next, we evaluated the differences in the severity of thyroid disorders according to gender, and the results are shown in Figure 4C . There was no gender difference in the frequency of Grade 1, Grade 2-4, and transient thyrotoxicosis (p=0.20). Survival rates one year after ICI administration at each ICI are shown in Figures 4D–F . Male with thyroid-related irAE tended to have significantly higher survival rates than subjects without irAE in any ICI. In female, survival was higher for those with thyroid-related irAE compared to subjects without irAE, although this difference was not statistically significant. There was no gender difference in survival for subjects with the appearance of thyroid-related irAE.