Characterizing Thyroid Hormone Replacement, Baseline Thyrotropin, and Survival in Immune Checkpoint Inhibitor–Associated Thyroid Dysfunction

Abstract

Background:

Thyroid dysfunction (TD) occurs commonly from immune checkpoint inhibitors (ICI) cancer therapy, but questions remain regarding its predicting factors, appropriate dosing for thyroid hormone replacement, and the strength of association with overall survival (OS). We aim to address these three questions in our study.

Methods:

We performed a retrospective cohort study of adult patients with cancer who received ICIs from December 1, 2012, to December 31, 2019. After excluding 28 patients with preexisting primary hypothyroidism, 811 patients were evaluated for the development of new-onset ICI-TD. Kaplan–Meier survival and log-rank tests were used to compare OS distributions between ICI-TD status groups, following which Cox regression models addressed immortal time bias (ITB).

Results:

Of the 811 included patients with a median follow-up of 19.2 months, 122 (15.0%) patients developed ICI-TD. The median age at initiation of ICIs was 64.8 years; women comprised 42.8% of the cohort. There were no significant differences in age, sex, race, malignancy type, or personal history of autoimmunity in patients who developed ICI-TD versus those who did not. ICI-TD occurred most frequently after combination ICI therapy (32%) compared with CTLA-4 ICI and PD-1/PD-L1 ICI monotherapy (p = 0.002). The median levothyroxine dose was the highest, being 1.41 mcg/kg/day in the overt hypothyroidism group. Patients with ICI-TD had a higher median pre-treatment log2(thyrotropin or TSH) level (1.2, corresponding to TSH 2.3 mIU/L) versus those without (0.79, corresponding to TSH 1.7 mIU/L; p = 0.008); however, the area under the curve was <0.6, hence lacking predictive ability. The survival benefit of ICI-TD was not apparent after addressing ITB and adjusting for other variables affecting patient outcomes.

Conclusions:

The levothyroxine dose needed for overt hypothyroidism from ICI-TD is similar to athyreotic hypothyroidism. While baseline TSH in the upper normal range is associated with an increased risk of ICI-TD, there is no absolute baseline TSH value that accurately predicts ICI-TD in the clinical setting. The link between ICI-TD and OS needs further validation after accounting for ITB.

Introduction

Thyroid dysfunction (TD) has emerged as one of the most common adverse events from immune checkpoint inhibitors (ICI) cancer therapy. Multiple studies have characterized its frequency, onset, natural history, and, more recently, a link to improved survival.¹ However, questions remain regarding predictive factors, appropriate dosing for thyroid hormone (TH) replacement, and the strength of association with survival after accounting for other contributing factors.

ICI-TD is usually characterized by destructive thyroiditis presenting in different phases, where in the setting of overt hypothyroidism, the need for TH replacement is usually permanent.^2,3 Clinical practice for TH replacement in this group has been extrapolated from non-ICI data,⁴ starting slightly lower than weight-based replacement doses in older adults or those with cardiac comorbidities. However, this can lead to frequent dose adjustments, affecting the quality of life of patients and delayed achievement of euthyroid state. Mosaferi et al.⁵ recently demonstrated that adults with ICI-TD who developed overt hypothyroidism after initial thyrotoxicosis required a mean levothyroxine dose of 1.45 mcg/kg/day to achieve a final stable euthyroid state, which was higher than their initially prescribed dose of 1.16 mcg/kg/day. This study challenged some of the usual practices for ICI-TD management. However, these findings must be validated in other cohorts before widely impacting clinical practice. Our primary objective was to characterize the TH replacement dose needed for achieving euthyroidism in ICI-TD.

Another point that remains uncertain but with some supporting evidence is a baseline thyrotropin (TSH) in the high-normal range being associated with higher odds of ICI-TD. Thyroid antibody testing at baseline has some data for its predictive role in ICI-TD^6–9 but does not add to the usual thyroid function screening; it is not currently recommended for screening by the National Comprehensive Cancer Network (NCCN) guidelines.¹⁰ Baseline TSH, on the other hand, is recommended by the NCCN guidelines¹⁰; hence, it may help stratify patients for the likelihood of developing TD and change the clinical threshold for monitoring. Previous studies have demonstrated that TSH in the mid to upper normal range is associated with higher odds of ICI-TD.^11–13 However, to our knowledge, no prior study has evaluated the area under the curve (AUC) for a specific TSH value to predict the development of ICI-TD. Lastly, but importantly, the development of ICI-TD has been linked with improved overall survival (OS) and, in some studies, improved progression-free survival (PFS); however the survival advantage was less substantial after accounting for immortal time bias (ITB).¹ Our secondary objectives were to identify the predictive role of baseline TSH in the development of ICI-TD and the association between the development of ICI-TD and OS after addressing ITB.

Materials and Methods

Patient selection

We performed an institutional review board (IRB)–approved retrospective cohort study evaluating adult patients with cancer who received a Food and Drug Administration–approved ICI from December 1, 2012, to December 31, 2019, at a tertiary referral cancer center. The study was approved by our IRB (PROTOCOL # 0801-22-EP), and the research was completed in accordance with the Declaration of Helsinki as revised in 2013. The IRB provided ethical approval for this study. Waiver for informed consent was approved by the IRB as this was considered a retrospective chart review study. From 839 patients treated with ICIs, cases with preexisting primary hypothyroidism already on TH replacement (n = 28) were excluded; thus, our study cohort comprised 811 patients for analysis to evaluate the development of new-onset primary ICI-TD. Analysis for the predictive role of baseline TSH on ICI-TD was conducted after excluding 178 patients with missing values of TSH—hence, a reduced sample size of 633.

Study measures

All evaluations and treatments were at the discretion of the treating clinicians. Study members reviewed charts to identify the demographic, clinical, laboratory, and radiological data. TSH, free thyroxine (fT4), total triiodothyronine (T3), thyroid peroxidase (TPO) antibodies, and TSH receptor antibody (TRAb) were tested by Roche Cobas immunoassays (Roche, Indianapolis, IN). Protocol-based thyroid function testing occurred before each infusion of ICI and additionally if clinical presentation warranted investigation. Thyrotoxicosis was defined as TSH below the reference range and T4/T3 normal (subclinical) or elevated (overt). Primary hypothyroidism was defined as TSH above the reference range and T4/T3 normal (subclinical) or below the reference range (overt). Central hypothyroidism or nonthyroidal illness with low T4/T3 and inappropriately normal or low TSH was excluded, unless the patient had prior clear diagnosis of ICI-TD. ICI-TD was defined as the first abnormal TSH (either elevated or low) with or without abnormal T4 or T3, consistent with most prior cohort studies. When evaluating for normalization of thyroid function tests, labs were tested at least twice over a period of 6–12 weeks to confirm that they remained normal. Data regarding patient weight and levothyroxine dosing were collected in those with hypothyroidism. Weight and levothyroxine dosing data were collected at the time of euthyroid state, which was defined as two consecutively normal TSH levels separated by at least 6 weeks. We calculated the weight-based tools by dividing the most recent TH dose by the patient’s body weight as microgram per kilogram per day.

Outcomes

We compared the natural history among those patients, including whether they had preceding thyrotoxicosis and then developed hypothyroidism or presented in the hypothyroid phase. The first outcome assessed was the weight-based levothyroxine dosing needed to achieve a euthyroid state in patients who developed hypothyroidism. The second outcome assessed was the predictive role of baseline TSH for developing ICI-TD both as a single measure and as a log2-transformed value. The third outcome was to evaluate the association of the development of ICI-TD with OS before and after accounting for ITB and other confounding variables.

Statistics

Descriptive statistics were used to summarize the characteristics of the patients. Continuous variables were compared between two groups using the nonparametric Wilcoxon rank-sum test and more than two groups using the Kruskal–Wallis test. Categorical variables were compared between groups using the Chi-square test or Fisher’s exact test. Analysis for the predictive role of baseline TSH on ICI-TD was conducted after excluding 178 patients with missing values of TSH—hence, a reduced sample size of 633. Log2 transformation of TSH levels was applied; if the TSH level was 0, then 0.001 was added before Log2 transformation. The Wilcoxon rank-sum test was used to compare log2(TSH) levels between the two groups. Univariate logistic regression was used to examine log2(TSH) alone as the predictor of ICI-TD versus control. The classification ability of the biomarker was evaluated using the receiver operating characteristic curve and leave-one-out cross-validation (LOOCV). The AUC was calculated to provide a summary of overall classifier effectiveness.

Kaplan–Meier survival analysis was performed to estimate OS, and the log-rank test was used to compare OS distributions between ICI-TD status groups. To address ITB due to the time-dependent nature of ICI-TD, univariate and multiple extended Cox regression models incorporating time-dependent variables were used to assess the effects of ICI-TD on the OS of patients. Hazard ratio (HR) and confidence interval (CI) were calculated, and p-values were derived from the models. The assumption of proportional hazards of Cox models for other covariates was assessed using goodness-of-fit tests. The ICI-TD status was initially defined as a binary variable: patients who developed ICI-TD during the study period were assigned a value of 1, and those who did not were assigned a value of 0. Standard Cox proportional hazards models were first used to evaluate the association between ICI-TD and OS. However, the supremum test revealed a significant violation of the proportional hazards assumption for the ICI-TD variable (p < 0.0001). Additionally, ICI-TD had a time-dependent nature: ICI-TD onset occurred after ICI initiation and impacted the hazard only after it occurred. In other words, a patient’s ICI-TD status could be changed during the study period from no TD to TD. To address this, ICI-TD was redefined as a time-dependent variable, z(t), where z(t) = 0 if the patient had not yet developed ICI-TD at time t, and z(t) = 1 after its onset. Extended Cox regression models were then used to assess the effect of the time-dependent variable on the OS of patients. Although the ICI-TD variable was not statistically significant in the extended model, this approach provided a more accurate and unbiased estimation of its effect by correctly aligning exposure status with the appropriate risk period. We conducted a post hoc power analysis using simulated data to assess the ability of an extended Cox proportional hazards model to detect the effect of the time-varying covariate, ICI-TD onset. Each simulation included 811 subjects followed for up to 84 months, with 15% experiencing ICI-TD onset during follow-up and an overall event rate of approximately 52%. For each simulated dataset, we applied the extended Cox model to account for the time-varying nature of the covariate. The estimated empirical power to detect the disease effect at a 5% significance level was approximately 94.9% when OS times were generated by assuming a constant baseline hazard approximated from our data and a true HR of 0.5 estimated from prior studies without addressing ITB,^1,14 indicating that the sample size was sufficient to detect this extent of HR reduction due to ICI-TD onset. However, the power reduced to 48.5% when OS times were generated by assuming a constant baseline hazard approximated from our data and a true HR of 0.7 estimated from a prior study addressing ITB,¹⁴ indicating that the sample size is not sufficient to detect this extent of HR reduction due to ICI-TD onset after addressing ITB. A sample size of 1592 would be needed using HR of 0.7 from the prior study¹⁴ to sufficiently detect moderate HR reduction with an empirical power of 80.6%. SAS software version 9.4 was used for data analysis (SAS Institute Inc., Cary, NC). A p-value <0.05 was deemed statistically significant.

Results

Study population characteristics

As detailed in Table 1, a total of 811 patients with cancer receiving ICIs (after excluding preexisting hypothyroidism) were included for analysis, of which 54 received CTLA-4 ICI monotherapy, 707 received PD-1/PD-L1 ICI monotherapy, and 50 received a combination of CTLA-4 and PD-1/PD-L1 ICIs. The median age at initiation of ICI therapy for the entire cohort was 64.8 years; women comprised 42.8% of the cohort. From this cohort, 122 (15.0%) patients developed ICI-TD. There were no significant differences in age at ICI initiation, sex, race, malignancy type, or personal history of autoimmunity in patients who developed ICI-TD versus those who did not. In this cohort, 15 patients also developed ICI-pituitary dysfunction, as previously published,¹⁵ which was present in 4.1% of those with ICI-TD but in 1.4% of those without (p = 0.06).

Table 1.

Baseline Characteristics of Patients Prior to Initiation of ICI

Baseline characteristics	All ICI-treated patients (n = 811)	ICI-thyroid dysfunction present (n = 122)	ICI-thyroid dysfunction absent (n = 689)	p-Value
Age at ICI initiation (y), median (range)	64.85 (19.19, 91.7)	65.21 (19.26, 91.7)	64.71 (19.19, 91.68)	0.46
Gender, n (%)
Male	464 (57.21)	68 (55.74)	396 (57.47)	0.72
Female	347 (42.79)	54 (44.26)	293 (42.53)
Race,a n (%)
White	748 (92.57)	118 (96.72)	630 (91.84)	0.13
African American	30 (3.71)	1 (0.82)	29 (4.23)
Other	30 (3.71)	3 (2.46)	27 (3.94)
Malignancy,a n (%)
Malignant melanoma	154 (19.04)	29 (23.97)	125 (18.17)	0.09
Renal	71 (8.78)	14 (11.57)	57 (8.28)
Uroepithelial	42 (5.19)	5 (4.13)	37 (5.38)
Non-small cell lung	254 (31.4)	32 (26.45)	222 (32.27)
Small cell lung	45 (5.56)	8 (6.61)	37 (5.38)
Neuroendocrine	10 (1.24)	4 (3.31)	6 (0.87)
Other	233 (28.8)	29 (23.97)	204 (29.65)
Personal history of autoimmunity,a n (%)
No	689 (85.06)	103 (84.43)	586 (85.17)	0.83
Yes	121 (14.94)	19 (15.57)	102 (14.83)
ICI hypophysitis, n (%)
No	796 (98.15)	117 (95.9)	679 (98.55)	0.06
Yes	15 (1.85)	5 (4.1)	10 (1.45)

^a Three had missing values in race, two had missing values in malignancy, and one had missing value in personal history of autoimmunity.

ICI, immune checkpoint inhibitors.

Natural history and management of ICI-TD

As detailed in Table 2, of the 122 ICI-TD cases, 97 had received PD-1/PD-L1 ICI, 16 had received combination ICI, and 9 had received CTLA-4 ICI alone. When evaluating the frequencies of ICI-TD, it occurred most frequently after combination ICI therapy (32%) compared with CTLA-4 ICI (16.7%) and PD-1/PDL-1 ICI monotherapy (13.7%; p = 0.002). The median time from ICI therapy to the diagnosis of ICI-TD was 2.8 months (interquartile range 1.6–5.5). ICI-TD presented as new-onset hypothyroidism (n = 67, 54.9%), subclinical in 38 and overt in 29; or as thyrotoxicosis (n = 55, 45.1%), subclinical in 32 and overt in 23. We suppose that the mild thyrotoxic phase was missed initially in the patients presenting as new-onset hypothyroidism. Among patients who presented with thyrotoxicosis, 32 progressed to overt hypothyroidism and 23 had normalization of thyroid function tests. None of the cases demonstrated persistent hyperthyroidism. Subgroup analysis did not show a difference in the frequency of subclinical hypothyroidism by the class of ICI therapy (p = 0.11).

Table 2.

Frequency of Thyroid Dysfunction Categorized by ICI Therapy Type

ICI therapy type, n (% out of ICI type)	All ICI-treated patients (n = 811)	Thyroid dysfunction (n = 122)	Non-thyroid dysfunction (n = 689)	p-Value
CTLA-4	54	9 (16.67)	45 (83.33)	0.002
Combination of CTLA-4 and PD-1/PD-L1	50	16 (32)	34 (68)
PD-1/PD-L1	707	97 (13.72)	610 (86.28)

The median levothyroxine dose at final visit needed to achieve euthyroidism among the patients with ICI-TD hypothyroid with available data (n = 80) was 1.23 mcg/kg/day. The ICI-TD hypothyroid group consisted of four subgroups depending on their initial presentation: subclinical thyrotoxicosis followed by hypothyroidism, overt thyrotoxicosis followed by hypothyroidism, subclinical hypothyroidism at presentation, and overt hypothyroidism at presentation. When compared between these subgroups, the median levothyroxine dose was highest at 1.41 mcg/kg/day in patients who presented with overt hypothyroidism (n = 28), 1.22 mcg/kg/day in patients who presented with overt thyrotoxicosis followed by overt hypothyroidism (n = 17), 1.14 mcg/kg/day in patients who presented with subclinical thyrotoxicosis followed by hypothyroidism (n = 13), and 0.84 mcg/kg/day in patients who presented with and remained with subclinical hypothyroidism (n = 20; p = 0.07) (Table 3).

Table 3.

Final Median Levothyroxine Dose (mcg/kg/day) in Patients with ICI-TD That Developed Hypothyroidism and Had Levothyroxine Dosage Data Available (n = 78 Out of n = 80 Who Received Levothyroxine During Study Duration)

ICI-TD natural history in hypothyroidism group	N (%)	Median	Minimum	Maximum	p-Value
Subclinical thyrotoxicosis followed by hypothyroidism	13 (16.7)	1.14	0.28	2.34	0.066
Overt thyrotoxicosis followed by hypothyroidism	17 (21.8)	1.22	0.35	2.43
Subclinical hypothyroidism	20 (25.6)	0.84	0.16	3.23
Overt hypothyroidism	28 (35.9)	1.41	0.33	3.13

TD, thyroid disfunction.

Patients with ICI-TD had higher median pre-treatment log2(TSH) level (1.2, corresponding to TSH level 2.3 mIU/L) versus those without (0.79, corresponding to TSH level 1.7 mIU/L) indicated by Wilcoxon rank-sum test (p = 0.008) (Fig. 1). Based on these initial significant findings, we built a logistic model using all data without cross-validation to predict the occurrence of ICI-TD by log2(TSH) levels: ln [p/(1-p)] = −1.94 + 0.14 *log2 (TSH), where the predicted probability of ICI-TD was denoted as p. The AUC estimate was 0.59 for the built model (Fig. 2) and 0.56 using LOOCV (Fig. 3), which indicated that TSH did not achieve acceptable discrimination to discriminate accurately between ICI-TD and non-ICI-TD groups.

FIG. 1.

Wilcoxon rank-sum test to compare log2-transformed baseline thyrotropin (TSH) value between group with immune checkpoint inhibitor (ICI)-thyroid dysfunction (TD) and group without ICI-TD.

FIG. 2.

Area under the curve (AUC) of 0.59 for the receiver operating characteristic (ROC) curve in the logistic model to predict the occurrence of ICI-TD by log2-transformed baseline TSH value.

FIG. 3.

AUC of 0.56 for the leave-one-out cross-validation (LOOCV) in the logistic model to predict the occurrence of ICI-TD by log2-transformed baseline TSH value.

Factors affecting OS

During a median follow-up of 19.2 months [CI 17.93–21.19], 420 (51.8%) patients died. Median OS was significantly higher in the ICI-TD group (38.8 months) compared with the non-ICI-TD group (14.2 months; p = 0.0002) using traditional Kaplan–Meier survival analysis (Fig. 4). However, further multiple extended Cox regression analyses after adjusting for other covariates, including age at ICI initiation, sex, race, and cancer type (melanoma vs. lung cancer vs. other), and using ICI-TD as a time-dependent covariate to account for ITB did not show a significant association between ICI-TD and OS (Table 4), but the study was not adequately powered for this analysis.

FIG. 4.

Kaplan–Meier plots for overall survival (OS) in patients with cancer who developed ICI-TD (n = 122) versus those who did not develop it (n = 689) (p = 0.0002).

Table 4.

Multiple Extended Cox Regression Model Evaluating the Link Between ICI-TD as a Time-Varying Covariate and Hazard Ratio for Mortality After Adjusting for Confounding Variables and Accounting for Immortal Time Bias

Variables	HR [CI] for mortality	p-Value
ICI-TD, present vs. absent	1.08 [0.79, 1.47]	0.65
Age (y) at ICI initiation	1.006 [0.999, 1.014]	0.11
Sex, male vs. female	1.23 [1.01, 1.5]	0.05
Race, white vs. other	1.22 [0.82, 1.82]	0.32
Malignancy
Malignant melanoma vs. lung cancer	0.32 [0.22, 0.47]	<0.0001
Others vs. lung cancer	0.75 [0.6, 0.92]	0.007
ICI therapy type
CTLA-4 vs. PD-1/PD-L1 ICI	2.1 [1.36, 3.24]	0.0008
Combination vs. PD-1/PD-L1 ICI	0.87 [0.54, 1.39]	0.55

CI, confidence interval; HR, hazard ratio.

Discussion

In this study, we investigated three unanswered questions in the field of ICI-TD by investigating a cohort of ICI-treated adult patients with cancer. We confirm that most cases of ICI-TD reported are from PD-1/PD-L1 ICIs when used alone or in combination with a CTLA-4 ICI as compared with CTLA-4 ICI alone, as shown in prior studies.¹⁶ The actual frequency of ICI-TD is higher cohort studies, including ours, as compared with ICI trials likely due to several factors outlined in a review article.² Concordant with the guidelines,¹⁰ screening for TD has been well standardized. Overall, TH replacement was initiated in about two-thirds of those with new-onset ICI-TD (n = 80, 65.5% out of 122 with ICI-TD or 86% out of 93 with hypothyroidism). The median follow-up of 19.2 months in our study is enough to characterize the course of ICI-TD. Since the severity of hypothyroidism did not differ by ICI class, we do not expect the ICI class to impact the magnitude to TH replacement. The NCCN guidelines¹⁰ recommend oral levothyroxine 1.2–1.4 mcg/kg/day for hypothyroidism treatment in ICI-TD. For patients with advanced age, cardiac risk, or prolonged hypothyroidism, the recommendation is to initiate at 0.8–1.0 mcg/kg/day. In our study, the median levothyroxine dose needed to achieve a euthyroid state in the overt hypothyroidism group was 1.4 mcg/kg/day. This is in line with findings from previous studies. Three prior studies evaluated dosing specifically for ICI-TD. Iyer et al. found a 1.2 mcg/kg/day median dose needed to achieve euthyroid status,¹⁷ while Ma et al. had a 1.4 mcg/kg/day median dose achieving euthyroid status.¹⁸ A recent study by Mosaferi et al.⁵ specifically addressed this issue by evaluating 103 adults with ICI-thyroiditis who presented with thyrotoxicosis and then developed overt hypothyroidism and demonstrated that to achieve a final stable euthyroid state, they required a dose mean of 1.45 mcg/kg/day, which was higher than their initially prescribed dose of 1.16 mcg/kg/day. This final dose was also higher than the mean dose in the control Hashimoto’s thyroiditis cohort, which was 1.25 mcg/kg/day but was not different from that of athyreotic patients who required a dose of 1.54 mcg/kg/day. That study restricted its analysis to patients who presented with thyrotoxicosis followed by hypothyroidism compared with those with Hashimoto’s thyroiditis and athyreotic hypothyroidism. They also included patients with preexisting use of levothyroxine therapy if thyroid function changes met the criteria of hyperthyroidism followed by hypothyroidism requiring higher doses of TH replacement. Excluding patients who did not show a thyrotoxic phase and including those with preexisting levothyroxine use may change the dosing evaluation; hence, our study by excluding patients with prior use of TH replacement but including patients who developed any of the thyroiditis phases provides more comprehensive data on the levothyroxine dosing required to achieve euthyroid status in ICI-TD hypothyroidism. Altogether, these findings inform clinicians to start levothyroxine at a dose closer to weight-based for overt hypothyroidism from ICIs in the absence of significant cardiovascular disease or advanced age.

We found that a higher baseline TSH level, even within the normal range, is associated with the development of overt hypothyroidism (2.3 vs. 1.7, p = 0.008). Previous studies have demonstrated that TSH in the mid to upper normal range is associated with higher odds of ICI-TD^11–13 but have not evaluated the actual discriminatory power of the baseline TSH value in predicting ICI-TD. We addressed this in our study by a logistic model where the AUC estimate was 0.59 for the built model and 0.56 using LOOCV, indicating that the baseline TSH did not achieve acceptable discrimination to discriminate accurately between ICI-TD and non-ICI-TD groups. Hence, the initial association of a TSH in the high-normal range did not translate into a strong role of baseline TSH in predicting the development ICI-TD. Altogether, these findings imply that clinicians should not take baseline TSH level as a single factor when predicting the development of ICI-TD, which still needs screening and targeted case finding as recommended by the guidelines.¹⁰

ICI-TD has been associated with improved OS in multiple prior studies and with improved PFS in some studies, as summarized in a meta-analysis.¹ This meta-analysis by Cheung et al.¹ demonstrated a 40–60% reduction in mortality rate in patients with TD when compared with those without. However, after accounting for ITB, which was only done in a few studies, this association with OS was less substantial.¹ This bias describes the phenomenon where patients who remain in the study longer (lived longer) are theoretically more likely to develop an outcome. At the same time, those who die or whose disease progresses earlier are at a lower risk of experiencing an outcome. Our study that accounts for ITB adds to the currently available literature. We found an OS benefit incurred by the ICI-TD group on initial analysis; however, this was not significant after using ICI-TD as a time-dependent covariate to address ITB and adjusting for other confounding variables, including age, sex, race, cancer type, and ICI class. A study that evaluated whether the prolonged OS associated with ICI-induced thyroiditis was a product of ITB showed that the adjusted HR for mortality in patients who develop ICI-TD changed from 0.5 to 0.7 after accounting for ITB.¹⁴ Hence, our results are in alignment with prior studies that showed improved OS in ICI-TD (without accounting for immortal time bias) but some that did account for this bias found that the strength of survival benefit was not as strong.¹⁴ Our study possessed sufficient power to detect a survival advantage prior to accounting for ITB but not after accounting for this. We report these findings to raise awareness that all future studies evaluating a link with survival need to account for ITB to provide a more accurate measure of survival benefit. Altogether, these data suggest that if this adverse event is managed well, patients are expected to continue having a favorable response to their primary cancer treated with immunotherapy. In the future, this association raises the possibility of using ICI-TD as a biomarker when assessing response to immunotherapy, but that requires more validation in larger studies after accounting for ITB.

Limitations of this study include its retrospective nature, which raises the concern for unaccountable bias and residual confounding in the analyses. Clinically, we do not measure antithyroid antibodies prior to ICI initiation as these are not currently recommended by guidelines. Another limitation of chart review studies including ours is the inability of the patient medical records to accurately capture personal history of autoimmunity; hence, the absence of an association with ICI-TD development in our study may be due to that limitation. We did not aim to characterize the differences in TD arising from ICIs in conjunction with tyrosine kinase inhibitors (TKIs), which has been reported previously.¹⁹ Additionally, we are not able to characterize the differences in TD arising from the more recent combinations of ICIs with cytoreductive chemotherapy or Lag3 inhibitors as that would have limited the follow-up duration—hence hindered our ability to appropriately address our main research objectives. In our study, 178 patients did not have TSH level before ICI therapy recorded in their charts raising concern that some of them may have had subclinical hypothyroidism, however, since none were on TH replacement, we included them to assess the dose requirement in ICI-TD hypothyroidism. Baseline TPO antibody status could potentially impact the link between baseline TSH and development of ICI-TD, but this was not measured in our study. Our study did not have enough sample size to statistically compare the levothyroxine dose between each subcategory of ICI-TD, but it is possible that some patients presenting with hypothyroidism could have had an asymptomatic thyrotoxicosis phase in between laboratory screening, and those presenting with hypothyroidism could have had underlying thyroid autoimmunity, which could account for higher TH replacement dose when there is additional destructive thyroiditis from ICI. Lastly, the study sample for the outcomes of mortality was not specifically powered to identify a moderate difference in survival after accounting for ITB; hence, the lack of association with OS after accounting for ITB needs to be validated in more extensive cohort studies with a sample size of approximately 1600 for at least 80% power. As survival analysis was not the primary objective of our study, the sample size was not planned to address this question.

To conclude, we found in this large cohort study that the median levothyroxine dose needed for adequate hormone replacement in ICI-TD characterized by overt TD is closer to that needed for athyreotic hypothyroidism, which can guide clinicians in managing this condition. We also found that while baseline TSH in the upper normal range may be associated with an increased risk of developing ICI-TD, there is no absolute TSH value that has enough predictive power for ICI-TD, highlighting the importance of screening and case detection. Lastly, while ICI-TD appears to be associated with better OS, this association is impacted by ITB. However, it does suggest that most patients can continue their cancer treatment with ICI unhindered if ICI-TD is recognized and managed appropriately.

Authors’ Contributions

D.A.: Writing—conceptualization (supporting); writing—original draft (equal); writing—review and editing (supporting); and methodology (supporting). J.J.: Methodology (supporting) and writing—review and editing (supporting). F.Q.: Formal analysis (lead) and writing—review and editing (supporting). W.G.: Conceptualization (supporting) and writing—review and editing (supporting). A.K.G.: Writing—review and editing (supporting). A.K.: Conceptualization (lead); methodology (lead); formal analysis (supporting); writing—original draft (equal); and writing—review and editing (lead).

Disclaimer

The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.

Author Disclosure Statement

None of the authors has anything to disclose.

Funding Information

The project described is supported by the National Institute of General Medical Sciences, U54 GM115458, which funds the Great Plains IDeA-CTR Network. A.K. received a grant from the National Institute of General Medical Sciences, U54 GM115458 (PI: Matthew Rizzo), which funds the Great Plains IDeA-CTR Research Scholars Program awarded to A.K.