Immune checkpoint inhibitor–induced hyperthyroidism: Incidence,risk factors,and clinical outcomes in a real-world cohort study

Abstract

Background

Hyperthyroidism is an immune-related adverse event associated with ICI use. Its underlying risk factors, progression to thyroiditis, and impact on patient outcomes remain less understood.

Methods

Data were obtained from the TriNetX network for patients aged ≥ 18 yr with cancers indicated for ICI therapy between January 1, 2013, and December 31, 2024. Patients with pre-existing thyroid disorders were excluded. Competing risk analyses evaluated hyperthyroidism or thyroiditis versus death within 12 months of ICI initiation. Cox proportional hazards modeling identified significant predictors, and backward elimination retained variables that met significance threshold.

Results

Among 41,629 patients receiving ICIs, 2.2% developed hyperthyroidism and 1.6% developed thyroiditis within 12 months. In those with hyperthyroidism, 17% initiated metoprolol, 8.9% initiated propranolol, and 5.2% initiated atenolol. Mortality rate was 2.9%. Patients who developed hyperthyroidism were more likely to have RCC (p < 0.0001, SMD = 0.27) or endometrial cancer (p < 0.0001, SMD = 0.24) and less likely to have liver (p < 0.0001, SMD = 0.24) or bladder cancer (p < 0.0001, SMD = 0.1575). While Nivolumab and ipilimumab usage was more common in patients who developed hyperthyroidism, ipilimumab showed a higher hazard in CPH (HR: 1.39, p = 0.0001). Patients with type 1 diabetes also showed significantly higher risk (HR 1.70, p = 0.0362), while black or African American patients showed lower hazard (HR: 0.798, p = 0.0305).

Conclusion

ICI-induced hyperthyroidism is associated with significant mortality. Identifying high-risk cancer subtypes and using beta-blockers for symptom control are essential to mitigating adverse effects.

Keywords

hyperthyroidism immune-related adverse events (irAEs)Immune checkpoint inhibitors (ICI)thyroiditis cancer immunotherapy

Introduction

Immune checkpoint inhibitors (ICIs) have revolutionized the treatment of many advanced malignancies by enhancing anti-tumor T-cell activity. Antibodies targeting checkpoints such as CTLA-4, PD-1, or PD-L1 can unleash robust immune responses against cancer, leading to improved survival in diseases like melanoma, lung cancer, renal cell carcinoma, and others.^1,2 However, by removing inhibitory “brakes” on the immune system, ICIs can also break self-tolerance and provoke immune-related adverse events (irAEs) that affect normal tissue.^1,3 Endocrine glands are frequent targets of ICI toxicity, and thyroid dysfunction stands out as the most common endocrine irAE observed in clinical practice.⁴

The spectrum of ICI-induced thyroid abnormalities includes hypothyroidism (thyroid hormone deficiency), thyrotoxicosis (hyperthyroidism due to excess circulating thyroid hormones), and thyroiditis (inflammatory destruction of the thyroid). Notably, hypothyroidism is more commonly reported, whereas hyperthyroidism occurs in a smaller subset of patients.⁵ Published clinical trials and pharmacovigilance studies indicate that approximately 1% to 7% of patients receiving anti–PD-1 or anti–PD-L1 therapies develop thyrotoxicosis (transient or overt hyperthyroidism), depending on the agent and regimen.^6–8 Combination immunotherapy (for example, PD-1 plus CTLA-4 blockade) appears to further increase the risk of thyroid irAEs, with some studies reporting thyrotoxicosis rates exceeding 5–8% under combined treatment.⁹

ICI-induced hyperthyroidism is most often caused by autoimmune (lymphocytic) or painless thyroiditis, wherein immune activation leads to destructive inflammation of the thyroid gland.⁹ This destructive thyroiditis causes the release of preformed thyroid hormones, resulting in transient thyrotoxicosis that typically lasts a few weeks and is frequently followed by hypothyroidism as hormone stores are depleted, and the gland recovers or scars.¹⁰ Most cases are not true Graves’ disease (TSH-receptor antibody–mediated hyperthyroidism), although rare instances of ICI-triggered Graves’ disease have been documented in case reports.¹¹ Clinically, patients may present with symptoms of thyrotoxicosis (palpitations, tachycardia, tremor, anxiety, weight loss, etc.), or hyperthyroidism may be detected on routine thyroid function tests before symptoms become severe. Given that ICI-induced thyroiditis can be asymptomatic or nonspecific in early stages, proactive laboratory monitoring is recommended.¹²

Despite growing recognition of ICI-related hyperthyroidism, several aspects remain incompletely understood. It is not entirely clear which patients are predisposed to this irAE – for instance, whether certain demographic factors or tumor types carry a higher risk – and how the risk varies among different ICI agents. Furthermore, the clinical significance of ICI-induced hyperthyroidism on cancer treatment outcomes merits investigation. Optimal management strategies for ICI-induced hyperthyroidism are based largely on expert opinion and analogy to spontaneous thyroiditis or Graves’ disease, since no randomized trials address irAE thyroiditis treatment.

In this context, we conducted a large real-world cohort study to better characterize ICI-induced hyperthyroidism. Our objectives were to quantify the incidence of hyperthyroidism attributable to ICIs, identify patient or treatment characteristics associated with increased risk, and describe the management and outcomes of these cases. We focused exclusively on hyperthyroid presentations (thyrotoxicosis with or without thyroiditis) and did not address primary hypothyroidism in this analysis, as the clinical considerations for hypothyroidism are distinct.

Methods

Study design and data source

This study is a retrospective cohort analysis using the TriNetX network, a federated electronic health record (EHR) database that aggregates de-identified patient data from multiple healthcare organizations.¹³ The TriNetX Analytics platform was used to identify eligible patients and perform analyses. The study was reviewed and deemed exempt from institutional IRB approval since only de-identified data were used.

Cohort selection

We included adults (age ≥18 years) with a documented diagnosis of one of the following malignancies indicated for which ICI use is indicated: biliary cancer (stage IV or TNM_M1), breast (triple-negative or stage IV or TNM_M1), cervical cancer (stages III-IV or TNM_M1), colorectal cancer (stage IV or TNM_M1), skin cancer (stage IV or TNM_M1, squamous cell histology), endometrial cancer, esophageal cancer (stage IV or TNM_M1), gastric cancer (stage IV or TNM_M1), head, face, and neck cancer (stage IV or TNM_M1, squamous cell histology), liver cancer, Hodgkin lymphoma, mesothelioma, melanoma (stages IIB-IV or TNM_M1), Merkel cell carcinoma, mycosis fungoides or Sézary disease, lung (stages IB-IV or TNM_M1, specific histologies), non-Hodgkin lymphoma, kidney cancer, or bladder cancer. Patients were then restricted to those who initiated treatment with an immune checkpoint inhibitor between January 1, 2013, and December 31, 2024. ICI agents of interest included PD-1 inhibitors (nivolumab, pembrolizumab, cemiplimab), PD-L1 inhibitors (atezolizumab, durvalumab), and CTLA-4 inhibitors (ipilimumab). The first administration of any ICI on or after January 1, 2013, was defined as the index date for each patient. We excluded patients with evidence of pre-existing thyroid disease before ICI initiation – this included any history of autoimmune thyroiditis, hyperthyroidism, or thyrotoxicosis. A full list of inclusion and exclusion criteria, as well as the codes used, can be found in Supplementary Table 1.

Outcomes and definitions

The primary outcome was new-onset hyperthyroidism (thyrotoxicosis) within 12 months, at least one day after the initiation of one of the listed ICIs. We defined hyperthyroidism based on EHR diagnostic codes (ICD-10-CM codes E05 for thyrotoxicosis/hyperthyroidism). We included coding for thyroiditis (e.g., ICD-10 E06) as a related outcome for the cumulative incidence because, in practice, many cases of ICI-induced thyrotoxicosis are due to thyroiditis. Secondary outcomes included the use of beta-blocker medications (metoprolol, propranolol, or atenolol) in the period after hyperthyroidism diagnosis. All-cause mortality was also considered for the cumulative incidence, based on EHR data andexternal mortality sources available within the TriNetX platform, including third-party death records, where applicable.

Statistical analysis

Baseline characteristics of patients who developed hyperthyroidism were summarized and compared to those of the overall ICI-treated cohort using TriNetX's built-in analysis, which employs Chi-square tests for categorical variables and t-tests for continuous variables. We report means (± standard deviation) for continuous variables and counts (percentages) for categorical variables. Cumulative incidence functions for hyperthyroidism and for thyroiditis within 12 months of ICI initiation were estimated using a competing risks framework, with death treated as a competing event. The Aalen-Johansen estimator was used to calculate the probability of each event over the one year following ICI initiation. Results were visualized using cumulative incidence curves.

We built a multivariable Cox proportional hazards model to identify independent predictors of hyperthyroidism within 1 year of ICI. Candidate covariates included: age (as a continuous variable), sex, race (White, Black, Asian), ethnicity (Hispanic vs non-Hispanic), baseline BMI percentile, cancer type (categorized by primary tumor type: e.g., melanoma, lung cancer, renal/kidney cancer, colorectal, breast, lymphoma, etc.), ICI agent (as an individual drug), thyroxine (categorized as low <0.4 mIU/L, normal 0.4–4, or high >4), and baseline free T4 level (classified as low <0.8 ng/dL, normal 0.8–2.0, or high >2.0). We employed backward stepwise elimination with a threshold of p < 0.05 to retain variables in the final model. Hazard ratios with 95% confidence intervals were calculated for each covariate in the final model and reported. Two-sided p-values were reported, with significance set at α = 0.05.

All analyses were conducted within the TriNetX platform, which utilizes R for statistical calculations. Two-sided statistical tests were employed throughout. No imputation was done for missing data; patients with missing covariate data were excluded from multivariable modeling (the number of such exclusions was <5% of the cohort for each variable).

Ethical considerations

This study employed de-identified data and involved no direct patient contact or intervention.

Results

Patient population and baseline characteristics

A total of 41,629 cancer patients met the inclusion criteria. By study design, none of the patients had a known thyroid disorder before starting ICI therapy. Of these, 942 went on to develop hyperthyroidism and 40,687 did not. Baseline characteristics for patients who did and did not develop hyperthyroidism are shown in Table 1. Patients who experienced hyperthyroidism were, on average, slightly younger than those who did not (mean age, 66.9 vs 69.1 years; p < 0.0001, SMD = 0.17), and had a higher baseline BMI (mean, 28.6 vs 27.5; p < 0.0001, SMD = 0.17). The hyperthyroid group had a more balanced sex distribution (51% male vs. 61%, p < 0.0001, SMD = 0.21) compared to the non-hyperthyroid group. In contrast to spontaneous autoimmune thyroid disorders, which predominate in females, ICI-related thyroiditis did not show a strong female bias in this large sample. Considering ethnicity, a greater proportion of hyperthyroid patients were non-Hispanic (78% vs. 71%, p < 0.0001, SMD = 0.15). Racial distributions, however, were generally similar (White: 70% vs. 68%, Black: 11% vs. 10%, Asian: 7% vs. 7%). Autoimmune conditions like lupus and celiac disease were rare (0–1%), making comparison difficult due to TriNetX's rounding policy (counts < 10 are rounded to 10). As a result, caution is warranted in interpreting these comparisons.

Table 1.

Baseline characteristics.

Variable	Hyperthyroid group(n = 942)	Non-Hyperthyroid group(n = 40,687)	P-Value	SMD
Demographics
Age at Index (mean ± SD)	66.9 ± 12.7	69.1 ± 13	< 0.0001	0.1697
Current Age (mean ± SD)	62.7 ± 12.5	65.1 ± 13	< 0.0001	0.1918
Male	476 (51%)	24,721 (61%)	< 0.0001	0.2070
Female	420 (45%)	14,646 (36%)	< 0.0001	0.1758
Race
White	661 (70%)	27,811 (68%)	0.2359	0.0394
Black or African American	104 (11%)	4033 (10%)	0.2526	0.0368
Asian	66 (7%)	2777 (7%)	0.8276	0.0071
Other Race	32 (3%)	1494 (4%)	0.6571	0.0149
American Indian or Alaska Native	≤ 10 (1%)	153 (< 1%)	0.0009	0.0812
Native Hawaiian or Pacific Islander	≤ 10 (1%)	171 (< 1%)	0.0031	0.0748
Ethnicity
Not Hispanic or Latino	749 (79%)	29,263 (72%)	< 0.0001	0.1542
Hispanic or Latino	46 (5%)	2531 (6%)	0.0922	0.0584
Malignancy
Renal Cell Carcinoma	328 (35%)	9194 (23%)	< 0.0001	0.2727
Endometrial Cancer	172 (18%)	4017 (10%)	< 0.0001	0.2430
Liver Cancer	108 (11%)	8176 (20%)	< 0.0001	0.2384
Bladder Cancer	124 (13%)	7700 (19%)	< 0.0001	0.1575
Breast Cancer	32 (3%)	1597 (4%)	0.4086	0.0281
Colon Cancer	38 (4%)	1365 (3%)	0.2535	0.0360
Biliary Tract Cancer	≤ 10 (1%)	942 (2%)	0.0109	0.0974
Melanoma	61 (6%)	3206 (8%)	0.1131	0.0544
Lung Cancer	180 (19%)	8168 (20%)	0.4637	0.0244
Cervical Cancer	34 (4%)	1087 (3%)	0.0788	0.0538
Esophageal Cancer	18 (2%)	712 (2%)	0.7100	0.0120
Stomach Cancer	20 (2%)	819 (2%)	0.8119	0.0077
Skin Cancer	54 (6%)	2994 (7%)	0.0582	0.0658
Head, Face, and Neck Cancer	≤ 10 (1%)	732 (2%)	0.0908	0.0621
Hodgkin Lymphoma	23 (2%)	1485 (4%)	0.0498	0.0704
Mesothelioma	14 (1%)	762 (2%)	0.3857	0.0301
Mycosis Fungoides	≤ 10 (1%)	136 (< 1%)	0.0002	0.0874
Sézary Syndrome	0 (0%)	25 (< 1%)	0.4466	0.0351
Non-Hodgkin Lymphoma	43 (5%)	2216 (5%)	0.2376	0.0404
Merkel Cell Carcinoma	≤ 10 (1%)	550 (1%)	0.4446	0.0266
Other Conditions
Rheumatoid Arthritis (with factor)	≤ 10 (1%)	163 (< 1%)	0.0018	0.0776
Other Rheumatoid Arthritis	12 (1%)	710 (2%)	0.2735	0.0386
Systemic Lupus Erythematosus	≤ 10 (1%)	124 (< 1%)	< 0.0001	0.0920
Celiac Disease	≤ 10 (1%)	77 (< 1%)	< 0.0001	0.1108
Type 1 Diabetes Mellitus	17 (2%)	709 (2%)	0.8855	0.0047
Lab Values
BMI (mean ± SD)	28.6 ± 6.88	27.5 ± 6.56	< 0.0001	0.1667
n (%)	768 (82%)	32,126 (79%)	-	-
TSH (mean ± SD)	5.21 ± 54	3.81 ± 34.8	0.3542	0.0306
n (%)	572 (61%)	20,770 (51%)	-	-
Free T3 (mean ± SD)	285 ± 86.2	270 ± 87.1	0.2135	0.1651
n (%)	59 (6%)	2047 (5)	-	-
Free T4 (mean ± SD)	1.19 ± 0.368	1.16 ± 0.487	0.3541	0.0576
n (%)	335 (36%)	11,166 (27%)	-	-

Results reflect baseline characteristics up to 1-day prior to ICI use. Values ≤ 10 are rounded up to protect patient privacy. Results reflect n (%) unless specified.

Cancer type and hyperthyroid risk

Regarding underlying malignancies, renal cell carcinoma (35% vs. 23%, p < 0.0001, SMD = 0.27) and endometrial cancer (18% vs. 10%, p < 0.0001, SMD = 0.24) were significantly more common in the hyperthyroidism cohort. Conversely, liver cancer (11% vs. 20%, p < 0.0001, SMD = 0.24), bladder cancer (13% vs. 19%, p < 0.0001, SMD = 0.1575), and to a lesser degree, Hodkin lymphoma (2% vs. 4%, p = 0.0498, SMD = 0.0704) were less common in patients who developed hyperthyroidism. Other malignancies, such as lung, breast, colon, melanoma, and others, were present at similar frequencies.

Incidence of ICI-induced hyperthyroidism

Within 12 months of ICI therapy, 904 patients (approximately 2.2%) developed new-onset hyperthyroidism (thyrotoxicosis) (Figure 1). The slight reduction in patient count from the original query (942) to the final analytic cohort (904) reflects case-wise exclusion of patients with missing data needed for time-to-event modeling. The risk of hyperthyroidism rose most steeply in the initial months of therapy and began to plateau around 6 months. At 3 months, the cumulative incidence of hyperthyroidism was 1.4%, increasing to 2.0% by 6 months and 2.3% by 12 months. For comparison, the 12-month cumulative incidence of diagnosed thyroiditis (inflammation of the thyroid) was 1.6%. These two outcomes were not mutually exclusive – in many cases, patients carried diagnoses for both hyperthyroidism and thyroiditis, reflecting the clinical scenario of transient thyroiditis-induced thyrotoxicosis. Indeed, the curves for hyperthyroidism and thyroiditis ran in parallel early on. Importantly, death was a significant competing event: by 12 months, 29% of the overall cohort had died, with a cumulative incidence of 32% (reflecting advanced cancer mortality).

We also noted that most cases of ICI-induced hyperthyroidism occurred early in the course of treatment. The median time to onset of hyperthyroidism was approximately 2.3 months after starting immunotherapy. Over 75% of hyperthyroid events happened within the first 3 months. Very few new hyperthyroid cases were observed beyond 6–8 months of therapy. For the thyroiditis outcome specifically, the median onset was later (around 4 months). No instances of delayed, very late-onset hyperthyroidism (e.g.,>1 year on therapy or after cessation) were captured in our 12-month analysis window.

ICI use and hyperthyroidism

Among the ICIs used, pembrolizumab and nivolumab were the most common, accounting for 80–89% of the total ICI usage across cohorts (Table 2). There was no significant difference in usage between pembrolizumab in patients who developed hyperthyroidism and those who did not (51% vs. 49%). There was significantly higher usage of nivolumab (38% vs. 31%, p < 0.0001, SMD = 0.1350). Ipilimumab usage was also significantly higher in the patients who developed hyperthyroidism (20% vs. 10%, p < 0.0001, SMD = 0.2899). Atezolizumab (10% vs. 6%, p < 0.0001, SMD = 0.1461) and Durvalumab usage (8% vs. 6%, p = 0.0031, SMD = 0.1055) were slightly lower in hyperthyroid patients. The use of cemiplimab and relatlimab was too low to derive meaningful implications due to TriNetX's low-count rounding policy.

Table 2.

ICI distribution across cohorts.

ICI	Non-Hyperthyroid group (n)	Hyperthyroid group (n)	P-value	SMD
Pembrolizumab	19,983 (49%)	476 (51%)	0.3898	0.0283
Nivolumab	12,687 (31%)	354 (38%)	< 0.0001	0.1350
Atezolizumab	4019 (10%)	56 (6%)	< 0.0001	0.1461
Durvalumab	3381 (8%)	53 (6%)	0.0031	0.1055
Ipilimumab	4033 (10%)	190 (20%)	< 0.0001	0.2899
Cemiplimab	168 (< 1%)	≤ 10 (1%)	0.0026	0.0759
Relatlimab	51 (< 1%)	≤ 10 (1%)	< 0.0001	0.1221

Values ≤ 10 are rounded up to protect patient privacy. Results reflect n (%) unless specified. Results reflect values up to the day of the hyperthyroidism diagnosis.

Cox proportional hazards model

In the multivariable Cox proportional hazards model, several variables were found to be significantly associated with an increased risk of developing hyperthyroidism (Table 3) Most notably, use of ipilimumab was linked to a 39% higher hazard (HR: 1.39, 95% CI: 1.18–1.64, p = 0.0001). Patients with type 1 diabetes also had a 70% higher risk (HR: 1.70, 95% CI: 1.04–2.79, p = 0.0362), and those with any report of elevated free thyroxine levels (≥2 ng/dL) had more than double the risk (HR: 2.12, 95% CI: 1.34–3.34, p = 0.0012). Meanwhile, Black or African American patients had a lower risk of hyperthyroidism (HR: 0.798, 95% CI: 0.65, 0.979, p = 0.0305).

Table 3.

Cox proportional hazards model.

Covariate	HR	Coefficient	Standard error	z	p-value	95% CI
Ipilimumab	1.388	0.328	0.084	3.882	0.0001	(1.176, 1.637)
Black or African American	0.798	−0.226	0.104	−2.164	0.0305	(0.65, 0.979)
Type 1 diabetes mellitus	1.699	0.53	0.253	2.094	0.0362	(1.035, 2.79)
T4 ≥ 2 ng/dL	2.117	0.75	0.232	3.229	0.0012	(1.343, 3.339)

Reference cohort: patients with hyperthyroidism. HR > 1 = higher hazard in this group, HR < 1 = lower hazard.

Other covariates, such as age, sex, and race, were not significant independent predictors in the multivariable model. Although univariate comparisons suggested a slightly higher age in the hyperthyroid group, age did not retain significance when adjusting for other factors. Similarly, we did not observe a significant sex predisposition and there was no difference in adjusted hazard between males and females.

Beta-blocker utilization

Within one year of diagnosis, beta-blocker use was significantly more common among patients who developed hyperthyroidism following ICI therapy compared to those who did not (30% vs. 15%; risk difference [RD], 15%; 95% CI, 11%–19%; p < 0.0001) (Table 4). The most frequently prescribed agent was metoprolol, initiated in 17% of hyperthyroid patients versus 13% of those without hyperthyroidism (RD 4.9%, 95% CI: 1.9%–7.8%, p = 0.0003). Propranolol, a non-selective beta-blocker, was used in 8.9% of hyperthyroid patients compared to 1.5% in controls (RD 7.4%, 95% CI: 5.5%–9.3%, p < 0.0001), and atenolol in 5.2% vs. 0.67%, respectively (RD 4.6%, 95% CI: 3.1%–6.0%, p < 0.0001). These numbers exclude patients with prior beta-blocker use before ICI therapy or hyperthyroidism diagnosis. The findings suggest beta-blockers, particularly metoprolol and propranolol, were commonly initiated in response to ICI-induced thyroid dysfunction.

Table 4.

Beta blocker usage up to 1-year after diagnosis.

Medication	Hyperthyroidism after ICI	No hyperthyroidism after ICI	RD	95% CI	p-value
Any Beta Blocker	30%	15%	15%	(11%, 19%)	< 0.0001
Propanolol	8.9%	1.5%	7.4%	(5.5%, 9.3%)	< 0.0001
Atenolol	5.2%	0.67%	4.6%	(3.1%, 6.0%)	< 0.0001
Metoprolol	17%	13%	4.9%	(1.9%, 7.8%)	0.0003

Results exclude patients with beta-blocker usage prior to ICI use and hyperthyroidism diagnosis.

Discussion

In this large, real-world study of over 40,000 ICI-treated cancer patients, we found that approximately 2.2% developed new-onset hyperthyroidism (thyrotoxicosis) within 12 months of starting therapy. This overall incidence is in line with the incidence range reported in clinical trials and meta-analyses for PD-1/PD-L1 inhibitors, which generally cite a rate of about 1–5% of patients experiencing a hyperthyroid event.^6–8 Our findings reinforce that while ICI-induced hyperthyroidism is less common than ICI-induced hypothyroidism, it is a significant and clinically common complication. In absolute terms, 2–3 out of every 100 patients on PD-1/PD-L1 inhibitors can be expected to develop thyrotoxicosis. This is a markedly higher rate than in immunotherapy-naïve populations – for context, spontaneous hyperthyroidism (from autoimmune or other causes) might affect ∼1 in 500 to 1000 persons per year in the general adult population,¹⁴ whereas ICIs raise that risk several-fold.

The observed incidence in various ICIs (approximately 2–3% for PD-1 inhibitors and ∼1–2% for PD-L1 inhibitors) is consistent with prior large-scale analyses. For example, Barroso-Sousa et al. (2018) analyzed 38 trials and reported all-grade hyperthyroidism in 3.2% of patients on anti–PD-1 monotherapy and 8.0% on combination ICI therapy.¹⁵ Another recent meta-analysis focusing on newer ICIs found a pooled thyrotoxicosis incidence of 4.6%¹ with PD-1 inhibitors associated with the highest rates (around 7.5%).¹ Our slightly lower incidence likely reflects the real-world nature and reliance on documented diagnostic codes. Some patients with transient, subclinical lab abnormalities may not have been coded as “hyperthyroidism” in routine practice and thus were not captured. Indeed, single-institution studies that conduct systematic laboratory screening often report higher rates of thyroid dysfunction. For instance, in a dedicated melanoma cohort receiving pembrolizumab, de Filette et al. (2016) found thyroid dysfunction (including asymptomatic lab changes) in about 19% of patients, with overt thyrotoxicosis in ∼8%.¹⁶ Therefore, our 2.2% should be viewed as the incidence of clinically recognized hyperthyroid events. The true subclinical incidence may be higher, but importantly, even clinically significant cases occur in a notable percentage of patients – a frequency high enough to warrant routine monitoring.

We observed that hyperthyroid irAEs overwhelmingly occurred early during ICI therapy. The median onset of 9.9 weeks is slightly longer than prior observations (median ∼6–7 weeks for anti–PD-1 in monotherapy, shorter ∼3 weeks in combination settings).¹⁵ One practical implication is that oncologists and endocrinologists should be particularly vigilant for thyroid dysfunction in the first 2–3 months after starting immunotherapy. Our data support current guidelines that recommend checking thyroid function every 4–6 weeks initially.¹⁷ Beyond 6 months, new hyperthyroid events were rare in our cohort, although late-onset cases have been reported.

A novel finding from our analysis is the association of certain cancer types with differing hyperthyroid risk. Patients with endometrial carcinoma and renal cell carcinoma were more common in the hyperthyroidism cohort, and these associations have not been widely reported before. One hypothesis is that these tumor types may evoke stronger immune stimulation or involve combination treatments that amplify autoimmunity. For example, endometrial cancers that receive pembrolizumab are often mismatch-repair deficient (characterized by a high mutational load) or are treated with pembrolizumab plus Lenvatinib. Both scenarios could heighten immune reactivity and increase the risk¹⁵ of thyroiditis. Renal cell carcinoma patients frequently receive combination immunotherapy (nivolumab + ipilimumab)¹⁸ or sequential therapies that might predispose them to irAEs. These observations will need validation in other cohorts, but if confirmed, they could help clinicians identify which patients may require closer thyroid monitoring. It also raises interesting questions for research into why certain cancer-ICI combinations produce more thyroiditis. Possibilities include differences in T-cell clonal expansion, cross-reactivity between tumor antigens and thyroid antigens, or tissue-specific expression of checkpoints and antigens.

Our findings that pembrolizumab and nivolumab usage was higher in the hyperthyroidism group support a pattern that PD-1 inhibitors tend to carry a higher risk of thyroid immune-related adverse events (irAEs) than PD-L1 inhibitors or CTLA-4 inhibitors, as seen in prior meta-analyses.¹ For instance, a 2018 meta-analysis by Barroso-Sousa et al. reported all-grade hyperthyroidism in ∼3.2% of patients on PD-1 monotherapy vs ∼1.3% on CTLA-4 monotherapy, with combination therapy further increasing the risk.³ Our real-world data are consistent with those trial-based findings: pembrolizumab (PD-1) was used more frequently in the hyperthyroidism group (51% vs. 49%), although this difference was not statistically significant, as was the case with nivolumab (38% vs. 31%, p < 0.0001). In contrast, PD-L1 inhibitors atezolizumab (6% in the hyperthyroidism cohort vs. 10%, p < 0.0001) and durvalumab (6% vs. 8%, p = 0.0031) were used less frequently in the hyperthyroidism group. It is noteworthy that in the CPH model, however, ipilimumab (a CTLA-4 inhibitor) was independently associated with a significantly increased hazard of developing hyperthyroidism (HR: 1.39, p = 0.0001), highlighting the potential additive or synergistic endocrine toxicity of CTLA-4 blockage.¹⁹

Our study also highlights the lack of strong demographic predictors for ICI-hyperthyroidism. Unlike classical autoimmune thyroid disease (Graves or Hashimoto's), which disproportionately affects younger women, ICI-induced thyroiditis affected men and women nearly equally and typically older adults (median ∼65–70 years). This pattern aligns with the prior report²⁰ and suggests that the mechanisms are distinct from those of spontaneous autoimmunity. That said, we did not have data on baseline thyroid autoantibodies in this cohort; other studies have shown that the presence of anti-thyroid peroxidase (anti-TPO) antibodies before or early in treatment is associated with a higher likelihood of thyroid dysfunction.¹⁷ In our clinical practice, we often measure anti-TPO in patients who develop thyroiditis as it may predict progression to permanent hypothyroidism. Future prospective studies incorporating such lab markers could refine risk stratification.

One reassuring finding is that ICI-induced hyperthyroidism was generally manageable based on the low beta-blocker usage, which was given to 30% of the affected patients in our cohort. The predominant use of metoprolol and propranolol is in line with standard recommendations.²¹ Beta-blockers are recommended for any patient with symptomatic hyperthyroidism from thyroiditis or Graves’ disease.²¹ It's possible that some of the remainder of the population had only subclinical or mild thyrotoxicosis, which may not have warranted treatment, or had contraindications to beta-blockers (e.g., severe reactive airway disease).

We acknowledge several limitations in our study. First, it was retrospective and reliant on EHR data and diagnostic codes; thus, it may under-capture subclinical thyroid abnormalities or misclassify some cases. Second, our data did not always allow us to distinguish definitively distinguish between causes of thyrotoxicosis (thyroiditis vs Graves’ disease), as antibody and scan results were not uniformly available. Similarly, we could not reliably distinguish transient destructive thyrotoxicosis from persistent hyperthyroidism, including rare Graves disease, using diagnosis-code–based EHR data and medication records alone; therefore, our analysis focused on incident clinically recognized hyperthyroidism rather than adjudicated long-term disease course. However, given prior studies (and the short duration of hyperthyroidism in most cases), it is reasonable to assume the majority were thyroiditis. Third, although we attempted to control for confounders in the Cox model, unmeasured factors (such as baseline immune status, presence of thyroid antibodies, or concomitant medications that can affect the thyroid) were not available. Fourth, different drugs are used in other cancer populations, although cancer type was considered in the Cox model residual bias. Nonetheless, the general trend of PD-1 inhibitors having higher thyroiditis rates than PD-L1 inhibitors echoes clinical trial reports,¹ lending credibility to our findings.

Conclusion

Clinically apparent hyperthyroidism affects approximately 2–3% of patients on ICI therapy, most often early in treatment, and with certain cancer types appearing more susceptible. Although less than half of these cases require only symptomatic management—typically a short course of beta-blockers—routine thyroid function monitoring and patient education remain essential. Importantly, thyroid irAEs seldom necessitate discontinuation of immune checkpoint inhibitors, allowing patients to continue life-prolonging therapy. As ICIs expand into earlier disease settings and new combinations, heightened awareness and prompt intervention will ensure that ICI-induced hyperthyroidism remains a manageable rather than prohibitive complication.

Supplemental Material

sj-pdf-1-opp-10.1177_10781552261443700 - Supplemental material for Immune checkpoint inhibitor–induced hyperthyroidism: Incidence, risk factors, and clinical outcomes in a real-world cohort study

Supplemental material, sj-pdf-1-opp-10.1177_10781552261443700 for Immune checkpoint inhibitor–induced hyperthyroidism: Incidence, risk factors, and clinical outcomes in a real-world cohort study by Baqir Jafry, Farzeen Syed, Jennifer Collins and Amir Kamran in Journal of Oncology Pharmacy Practice

Footnotes

Acknowledgments

This work was previously presented in abstract form at the 2025 ASCO Annual Meeting, Chicago, Illinois, May 2025: Jafry B, et al. Clinical factors and prognostic outcomes of hyperthyroidism induced by immune checkpoint inhibitor therapy. J Clin Oncol. 2025;43(16_suppl):2620. doi:10.1200/JCO.2025.43.16_suppl.2620

ORCID iD

Baqir Jafry

Ethical considerations and IRB statement

This study used de-identified, aggregated electronic health record data obtained from the TriNetX Research Network. As no identifiable patient information was accessed or used, this study was deemed exempt from Institutional Review Board (IRB) review.

Author contributions

B.J., F.S., J.C., and A.K. contributed to this work as follows:

Study concept and design: B.J., F.S., A.K.

Data acquisition and management: J.C., A.K.

Data analysis and interpretation: B.J., J.C.

Manuscript drafting: B.J.

Critical revision for important intellectual content: F.S., J.C., A.K.

Final approval of the version to be published: B.J., F.S., J.C., A.K.

Accountability for all aspects of the work: All authors.

Funding

The authors received no financial support for the research, authorship, and/or publication of this article.

Declaration of conflicting interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Data availability statement

The data supporting the findings of this study were obtained from the TriNetX Research Network, which provides access to de-identified, aggregated electronic health record data. Due to data use agreements, individual-level data cannot be shared publicly. Researchers may request access to similar data directly from TriNetX () following institutional approval.

Supplemental material

Supplemental material for this article is available online.

References

Yoo

Lee

, et al. Incidence of endocrine-related dysfunction in patients treated with new immune checkpoint inhibitors: a meta-analysis and comprehensive review. Endocrinol Metab (Seoul) 2023; 38: 750–759.

Postow

Callahan

Wolchok

. Immune checkpoint blockade in cancer therapy. J Clin Oncol 2015; 33: 1974–1982.

Barroso-Sousa

Barry

Garrido-Castro

, et al. Incidence of endocrine dysfunction following the use of different immune checkpoint inhibitor regimens: a systematic review and meta-analysis. JAMA Oncol 2018; 4: 173–182.

Karaviti

Kani

Karaviti

, et al. Thyroid disorders induced by immune checkpoint inhibitors. Endocrine 2024; 85: 67–79.

Chera

Stancu

Bucur

. Thyroid-related adverse events induced by immune checkpoint inhibitors. Front Endocrinol (Lausanne) 2022; 13: 1010279.

Curkovic

Johnson

. Updates in toxicities associated with immune checkpoint inhibitors. Expert Rev Clin Immunol 2023; 19: 1117–1129.

Weber

D'Angelo

Minor

, et al. Nivolumab versus chemotherapy in patients with advanced melanoma who progressed after anti-CTLA-4 treatment (CheckMate 037): a randomised, controlled, open-label, phase 3 trial. Lancet Oncol 2015; 16: 375–384.

Robert

Schachter

Long

, et al. Pembrolizumab versus ipilimumab in advanced melanoma. N Engl J Med 2015; 372: 2521–2532.

Baraka

Abosheaishaa

Nassar

. Immunotherapy-induced thyroid dysfunction: an updated review. Egypt J Intern Med 2023; 35: 48.

10.

Yang

Shen

Zhang

, et al. Incidence of thyroid dysfunction caused by immune checkpoint inhibitors combined with chemotherapy: a systematic review and meta-analysis. Int Immunopharmacol 2024; 133: 111961.

11.

Husebye

Castinetti

Criseno

, et al. Endocrine-related adverse conditions in patients receiving immune checkpoint inhibition: an ESE clinical practice guideline. Eur J Endocrinol 2022; 187: G1–G21.

12.

Yamauchi

Yabe

. Best practices in the management of thyroiddysfunction induced by immune checkpoint inhibitors. Eur Thyroid J 2025; 14: e240328.

13.

Ludwig

Anson

Zirpel

, et al. A comprehensive review of methodologies and application to use the real-world data and analytics platform TriNetX. Front Pharmacol 2025; 16: 1516126.

14.

Censi

Salmaso

Ceccato

, et al. Hyperthyroidism incidence in a large population-based study in northeastern Italy. Endocr Connect 2023; 12. doi:10.1530/EC-23-0292

15.

Latif

Abdelgadir

Omara

, et al. Immune checkpoint inhibitors-induced thyroid dysfunction in patients with advanced malignancies. 10/01 2022.

16.

de Filette

Jansen

Schreuer

, et al. Incidence of thyroid-related adverse events in melanoma patients treated with pembrolizumab. J Clin Endocrinol Metab 2016; 101: 4431–4439.

17.

Chieng

JHL

Htet

Zhao

, et al. Clinical presentation of immune-related endocrine adverse events during immune checkpoint inhibitor treatment. Cancers (Basel) 2022; 14: 2687.

18.

Motzer

Tannir

McDermott

, et al. Nivolumab plus ipilimumab versus sunitinib in advanced renal-cell carcinoma. N Engl J Med 2018; 378: 1277–1290.

19.

Grosso

Jure-Kunkel

. CTLA-4 blockade in tumor models: an overview of preclinical and translational research. Cancer Immun 2013; 13: 5.

20.

Bai

Chen

, et al. Immune checkpoint inhibitor-associated thyroid dysfunction: a disproportionality analysis using the WHO Adverse Drug Reaction Database, VigiBase. Eur J Endocrinol 2020; 182: –9.

21.

Haanen

Obeid

Spain

, et al. Management of toxicities from immunotherapy: ESMO clinical practice guideline for diagnosis, treatment and follow-up^☆. Ann Oncol 2022; 33: 1217–1238.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.07 MB

0.00 MB