Non-thyroidal illness syndrome predicts poor prognosis in hepatocellular carcinoma patients treated with atezolizumab and bevacizumab

Abstract

Background

Atezolizumab plus bevacizumab (Ate/Bev) is the first-line systemic therapy for unresectable hepatocellular carcinoma (HCC). However, the prognostic impact of immune-related thyroid adverse events (irTAEs), particularly non-thyroidal illness syndrome (NTIS), remains unclear.

Methods

We retrospectively analyzed 70 patients with unresectable HCC treated with Ate/Bev. Thyroid function was monitored at baseline and during the treatment. IrTAEs were classified as NTIS (low T3 and/or free T4 with normal or low TSH levels) or thyroid adverse events (hypothyroidism or thyrotoxicosis). Clinical outcomes, including progression-free survival (PFS) and overall survival (OS), were assessed using Kaplan–Meier and Cox regression analysis. Subgroup analyses were performed to examine the associations between liver function, nutritional status, and HCC etiology.

Results

Among 70 patients (median age, 64 years; 80% male), 24 (34.3%) developed irTAEs: 11 had NTIS and 13 had thyroid AEs. NTIS was independently associated with worse OS (adjusted hazard ratio [HR], 2.51; P = 0.041) and shorter median OS (5.4 months). In contrast, thyroid AEs were associated with a favorable trend in PFS (adjusted HR, 0.37; P = 0.073) and a higher objective response rate (53.8% vs. 9.1%; P = 0.006). NTIS was correlated with deterioration in albumin level, body mass index, and ALBI grade. Viral HCC was associated with fewer thyroid AEs, while alcohol-related HCC trended toward NTIS.

Conclusions

Thyroid dysfunction during Ate/Bev therapy is prognostically significant. NTIS reflects systemic deterioration and portends poor survival, whereas thyroid AEs may suggest favorable immune activation. Routine monitoring of thyroid function may aid in risk stratification.

Supplementary Information

The online version contains supplementary material available at 10.1007/s00262-026-04308-z.

Introduction

Hepatocellular carcinoma (HCC) is the sixth most common cancer and the third leading cause of cancer-related deaths worldwide [1]. The advent of immune checkpoint inhibitors (ICIs) has substantially improved outcomes in patients with unresectable HCC [2, 3]. Atezolizumab plus bevacizumab (Ate/Bev) is the standard first-line systemic therapy for unresectable HCC, demonstrating superior survival outcomes compared to tyrosine kinase inhibitors [4–11]. However, responses to immunotherapy vary widely, and robust prognostic biomarkers are lacking. Immune-related adverse events (irAEs), particularly endocrine toxicities, have emerged as potential indicators of treatment response in various cancers [12]. Among these, thyroid dysfunction is one of the most frequently reported irAEs associated with ICIs [13], including in patients with HCC treated with Ate/Bev [14, 15]. These immune-related thyroid adverse events (irTAEs) are often subclinical and detected through routine thyroid function tests (thyroid-stimulating hormone [TSH] and free triiodothyronine/free thyroxine [FT3/FT4]) [16]. The phenotypic spectrum includes hyperthyroidism (thyrotoxicosis), hypothyroidism, and, less frequently, biphasic thyroiditis transitions [13].

While autoimmune thyroiditis resulting in hypothyroidism or thyrotoxicosis has been associated with favorable treatment responses, the clinical implications of non-thyroidal illness syndrome (NTIS), defined by reduced serum T3 without primary thyroid pathology, remain poorly understood. Unlike classical irTAE, which arises from ICI-induced autoimmunity, NTIS is a non-immune-mediated condition that reflects systemic decompensation, particularly in the context of critical illness and hepatic dysfunction. NTIS frequently occurs in patients with advanced malignancies or severe critical illness [17–19]. Given that the liver is the primary site of T3 production via peripheral deiodination, hepatic impairment in HCC may contribute to NTIS development [20]. Moreover, elevated levels of inflammatory cytokines, such as IL-6, which are common in advanced HCC or nonalcoholic steatohepatitis (NASH)-related diseases, have been shown to suppress TSH and impair T3 synthesis, further promoting NTIS [21, 22].

Despite the potential prognostic relevance, NTIS is often overlooked in oncology due to limited awareness and the lack of routine T3 monitoring. Whether NTIS portends a distinct clinical trajectory in HCC patients receiving Ate/Bev remains to be elucidated. To our knowledge, this is the first study to systematically evaluate the incidence, clinical features, and prognostic significance of NTIS in patients with unresectable HCC undergoing ICI-based therapy. By characterizing NTIS among a broader spectrum of irTAEs, this study seeks to clarify its clinical implications and prognostic significance in patients with unresectable HCC undergoing Ate/Bev therapy.

Material and methods

Study design and patients

This retrospective study enrolled 109 patients with unresectable HCC who received Ate/Bev as first-line systemic therapy at Chang Gung Memorial Hospital, Linkou Medical Center, between September 2020 and June 2023.

A formal sample size calculation was not performed because this was a retrospective observational study. The study cohort was determined by the number of eligible patients who received Ate/Bev during the study period. All patients meeting the predefined inclusion and exclusion criteria were included to maximize statistical power within the constraints of real-world data availability. Eligible patients were adults with unresectable or advanced HCC, either Barcelona Clinic Liver Cancer (BCLC) stage C or BCLC stage B, and not amenable to locoregional therapy (e.g., extensive tumors unsuitable for transarterial chemoembolization). The exclusion criteria were as follows: lack of baseline thyroid function tests, absence of any on-treatment thyroid function monitoring, or pre-existing thyroid dysfunction before starting Ate/Bev. Among the 109 patients initially identified, we excluded nine with no baseline thyroid laboratory examination, 18 with no follow-up thyroid tests, and 12 with abnormal thyroid function at the baseline. Seventy patients met the inclusion criteria and were analyzed (Fig. 1). This study was conducted in accordance with the Declaration of Helsinki and approved by the Institutional Review Board (IRB No. 202200991B0C501). The requirement for informed consent was waived because of the retrospective study design and use of de-identified data.

Fig. 1.

Flowchart of patient selection and analysis

Treatment and outcome evaluation

In Taiwan, prior to reimbursement by the National Health Insurance (NHI), a unique circumstance with the Patient Support Program provided access to atezo/bev for uHCC with lower payment but only with a low dose of bevacizumab (500 mg fixed dose, approximately 5–7.5 mg/kg). In our center, patients who qualified for the subsidy program were uniformly started on low-dose bevacizumab (500 mg fixed dose, approximately 5–7.5 mg/kg) intravenously every three weeks according to the Patient Support Programs in Taiwan (primarily in 2020–2023) because the program mandated a low dose to ensure drug affordability and atezolizumab (1200 mg fixed dose). Treatment continued until radiographic disease progression, unacceptable toxicity, or patient withdrawal. HCC diagnosis was established by established criteria (imaging or pathology) per European Association for the Study of the Liver and European Organization for Research and Treatment of Cancer (EASL/EORTC) guidelines [23]. Clinical and laboratory evaluations (including complete blood counts, liver function tests, and tumor markers) were performed at baseline and before each treatment cycle. Alpha-fetoprotein (AFP) levels were measured every 3 weeks. Imaging studies with dynamic computerized tomography (CT) or magnetic resonance imaging (MRI) were performed approximately 9–12 weeks after therapy initiation and then every 2–3 months to evaluate the tumor response.

Tumor responses were assessed by two independent radiologists using the modified Response Evaluation Criteria in Solid Tumors (mRECIST) for HCC [24]. An objective response was defined as a complete response (CR) or partial response (PR), and disease control was defined as CR/PR plus stable disease (SD). The best overall response during the therapy was recorded for each patient.

Definitions of immune‑related thyroid adverse event

Thyroid function was assessed at baseline and monitored periodically during treatment. Tests included thyroid-stimulating hormone (TSH), total T3, free T4 (FT4), and thyroid peroxidase antibody (TPOAb) when available. Thyroid function was re-evaluated at baseline and before every 1–2 treatment cycles (approximately every 3–6 weeks), or at any time when clinically indicated. All included patients had at least one baseline and one follow-up thyroid function panel. Transient gaps in thyroid monitoring did not influence NTIS classification, as NTIS was defined strictly based on the presence of recorded low total T3 and/or low FT4 with low or low-normal TSH levels. Patients lacking the requisite thyroid measurements were not classified as having NTIS. Based on the follow-up results, the patients were categorized into three groups: NTIS, thyroid AE, and no thyroid dysfunction. We classified immune-related thyroid dysfunction as follows: NTIS was defined as low total T3 and/or free T4 levels (below the normal range of 64–152 ng/dL for T3 and 0.70–1.48 ng/dL for fT4) with a normal or low TSH (0.35–4.95 μIU/mL). The thyroid AE group included overt hypothyroidism, subclinical hypothyroidism, thyrotoxicosis, and subclinical thyrotoxicosis cases. Overt hypothyroidism was defined as an elevated TSH above the normal upper limit (or ≥ 10 μIU/mL) with a concurrently low free T4. Subclinical hypothyroidism was defined as a TSH level above the normal range (< 10 μIU/mL) and a normal free T4 level. Overt thyrotoxicosis was defined by a suppressed TSH below the lower limit of normal with an elevated free T4, and subclinical thyrotoxicosis by a low TSH (below normal range) with normal free T4. Patients who maintained normal T3, free T4, and TSH levels throughout treatment were considered to have no irTAE (thyroid AE( −)/NTIS( −) group). Titers of thyroid peroxidase antibody (TPOAb) were measured by chemiluminescence microparticle immunoassay, with values > 5.6 IU/mL defined as positive.

Statistical analysis

Categorical variables are reported as frequencies (%), and continuous variables are reported as medians (IQR). Between-group comparisons were performed using the Kruskal–Wallis and χ²/Fisher’s exact tests with Bonferroni correction. PFS and OS were assessed using Kaplan–Meier curves and compared by log-rank test. Cox regression was used to identify factors associated with survival. Variables with P < 0.05 in the univariate analysis were entered into multivariate models. PFS was defined as the time from treatment initiation to progression or death; OS was defined as death from any cause. Landmark analyses (6 weeks and 3 months) were performed to account for the timing differences in irTAE onset. Wilcoxon signed-rank tests were used for pre/post comparisons of BMI and albumin levels in patients with irTAEs. Logistic regression was used to explore the predictors of NTIS and thyroid AEs. All tests were two-sided with a significance level of P < 0.05. Statistical analyses were performed using SPSS version 29 (IBM SPSS, Inc., Chicago, IL, USA).

Results

Clinical characteristics and incidence of thyroid AEs

Seventy patients with unresectable HCC, treated with Ate/Bev as first-line therapy, were included in the analysis. The baseline patient characteristics are presented in Table 1. The median age was 64.0 years (IQR 59.3–71.0), and 56 (80.0%) patients were male. Hepatitis B virus was the predominant HCC etiology (50 patients, 71.4%), followed by hepatitis C (8.6%) and alcohol-related or non-viral causes (20.0%). At baseline, liver function was well-preserved in most cases: 26 patients (41.9%) had ALBI grade 1, and 36 patients (58.1%) had ALBI grade 2. The median baseline BMI was 23.31 kg/m² (IQR 21.5–25.97). The majority of patients had advanced tumor burden: 52 patients (74.3%) were BCLC stage C, and 38 (54.3%) had portal vein thrombosis. Extrahepatic metastases were observed in 24 (34.3%) patients. The median largest tumor diameter was 9.30 cm, and 50.0% of the patients had AFP ≥ 400 ng/mL at baseline (median AFP 372.90 ng/mL, IQR 14.58–11,802.13). Patients received a median of 4.50 cycles of Ate/Bev (IQR 3–7, range 2–34), with a median follow-up time of 9.93 months (IQR 5.37–17.58, range 1.33–35.20 months).

Table 1.

Baseline characteristics of patients with unresectable hepatocellular carcinoma treated with atezolizumab plus bevacizumab, stratified by thyroid dysfunction status (NTIS, Thyroid AE, and No thyroid dysfunction)

	All patients (n = 70)	NTIS ( +) (n = 11)	Thyroid AE ( +) (n = 13)	Thyroid AE (-) and NTIS (-) (n = 46)	P value
Age, median (IQR), years	64.04 (59.34, 71.04)	63.96 (61.33, 70.39)	68.11 (60.03, 78.14)	63.07 (57.44, 69.67)	0.324
Male sex, n (%)	56 (80.0)	10 (90.9)	8 (61.5)	38 (82.6)	0.151
BMI, kg/m2	23.31 (21.50, 25.97)	23.11 (21.64, 25.87)	22.86 (20.66, 25.07)	23.80 (21.73, 27.12)	0.368
ALBI grade, n (%)					0.189
1	26/62 (41.9)	2/10 (20.0)	6/10 (60.0)	18/42 (42.9)
2	36/62 (58.1)	8/10 (80.0)	4/10 (40.0)	24/42 (57.1)
Tumor size (cm)	9.3 (5.50, 11.15)	9.40 (7.60, 13.00)	8.05 (4.25, 10.85)	9.30 (5.45, 11.08)	0.304
Barcelona clinical liver cancer stage, n (%)					0.889
A	4 (5.7)	0	1 (7.7)	3 (6.5)
B	14 (20.0)	2 (18.2)	2 (15.4)	10 (21.7)
C	52(74.3)	9 (81.8)	10 (76.9)	33 (71.7)
AFP ≥ 400 ng/mL, n (%)	32/64 (50.0)	6/10 (60.0)	5/10 (50.0)	21/44 (47.7)	0.782
NLR, median (IQR)	3.58 (2.61, 5.39)	3.04 (2.31, 6.76)	4.66 (2.76, 7.68)	3.36 (2.47, 4.65)	0.540
Presence of MVI, n (%)	2 (2.9)	0	0	2 (4.3)	0.584
Presence of PVT, n (%)	38 (54.3)	7 (63.6)	5 (38.5)	26 (56.5)	0.408
Presence of extrahepatic spread, n (%)	24 (34.3)	4 (36.4)	6 (46.2)	14 (30.4)	0.566
Etiology of HCC, n (%)
Hepatitis B	50 (71.4)	8 (72.7)	8 (61.5)	34 (73.9)	0.680
Hepatitis C	13 (18.6)	3 (27.3)	1 (7.7)	9 (19.6)	0.450
Alcohol	14 (20.0)	5 (45.5)	1 (7.7)	8 (17.4)	0.053
Prior local therapy for HCC, n (%)	27/60 (45.0)	1/9 (11.1)	6/10 (60.0)	20 (48.8)	0.070
FT4 (ng/dL), median (IQR)	0.98 (0.89, 1.08)	0.98 (0.94, 1.13)	0.97 (0.87, 1.14)	0.99 (0.89, 1.07)	0.754
T3 (ng/dL), median (IQR)	77.5 (60.50, 93.80)	84.80 (64.30, 119.70)	82.55 (65.05, 93.58)	71.35 (57.55, 93.45)	0.398
TSH (mIU/L), median (IQR)	1.45 (1.02, 2.43)	1.12 (0.88, 2.30)	1.52 (1.26, 2.62)	1.68 (0.96, 2.41)	0.636
TPOAb positive, n (%)	3/51 (5.9)	0	1/9 (11.1)	2/33 (6.1)	0.603

Abbreviations: AFP = alpha-fetoprotein; ALBI = Albumin-Bilirubin; BMI = body mass index; FT4 = free thyroxine, HCC = hepatocellular carcinoma; MVI = macrovascular invasion; NLR = neutrophil-to-lymphocyte ratio; NTIS = non-thyroidal illness syndrome; PVT = portal vein thrombosis; T3 = triiodothyronine; thyroid AE = thyroid adverse events; TPO Ab = thyroid peroxidase antibody; TSH = thyroid stimulating hormone

Thyroid dysfunction during therapy was observed in 24 (34.3%) patients. Specifically, 13 patients (18.6%) developed thyroid AEs (immune-related thyroiditis or hypothyroidism), and 11 patients (15.7%) developed NTIS. The remaining 46 patients (65.7%) had no thyroid abnormalities during the treatment period. Among the 13 patients with thyroid AE, 8 (11.4% of the total) experienced thyrotoxicosis and 5 (7.1%) experienced hypothyroidism as the initial presentation. Most thyrotoxicosis cases were transient; of the eight thyrotoxic patients, three returned to the euthyroid state without treatment, and two subsequently transitioned to overt hypothyroidism (consistent with a thyroiditis burnout phase). Similarly, of the five patients with new-onset hypothyroidism, four had subclinical or mild hypothyroidism that eventually resolved spontaneously (except for one patient who required ongoing levothyroxine replacement). None of the patients developed severe symptomatic hyperthyroidism requiring beta-blockers or antithyroid drugs. All NTIS cases were identified by routine laboratory tests showing low T3 (with or without low FT4) and inappropriately low-normal TSH levels, rather than by overt symptoms.

The timing of irTAE onset varies widely. The median time to NTIS occurrence was 2.7 months (range, 0.9–14.7 months) from the start of therapy. Thyroid AEs tended to occur later in hypothyroidism (median 4.40 months, range 1.34–17.61) and earlier in thyrotoxicosis (median 2.66 months, range 0.88–34.73), although the sample size was small. Notably, some irTAEs developed quite late (beyond 1–2 years) in a few patients, underscoring the need for continued thyroid monitoring, even during extended treatment or follow-up.

We compared the baseline attributes of patients who developed NTIS, thyroid AEs, and no thyroid dysfunction (Table 1). There were no significant differences among the three groups in terms of age, sex, baseline BMI, HCC stage, tumor burden, AFP level, NLR, or baseline thyroid hormone levels (T3, FT4, and TSH) (all P > 0.05).

Therapeutic efficacy and survival outcomes

At the data cut-off (February 7, 2024), 30 deaths (42.9%) had occurred, and 40 PFS events (progressions or deaths) were recorded for the PFS analysis. The median OS for the entire cohort was 27.53 months (95% CI 5.18–49.89), as illustrated in Fig. 2A. The median PFS was 5.40 months (95% CI 3.72–7.08) (Fig. 2B).

Fig. 2.

A Overall survival (OS) and B progression-free survival (PFS) in patients with hepatocellular carcinoma (HCC) undergoing Ate/Bev treatment

Tumor response could not be evaluated in all patients: 7 patients died or progressed clinically before the first imaging assessment (~ 8 weeks), and 3 others lacked confirmatory scans beyond the initial assessment. Therefore, response rates were calculated for 63–67 evaluable patients, as indicated. The initial imaging response (at ~ 8–12 weeks) showed an objective response rate (ORR) of 33.3% (21/63 evaluable patients had CR or PR) and a disease control rate (DCR) of 66.7% (Supplementary Table 1). With ongoing therapy, some patients improved from stable disease to response; the best ORR during treatment was 37.3% (25/67 evaluable patients), and the best DCR was 70.1% (Supplementary Table 1).

Association of irTAE with clinical outcomes of Ate/Bev in HCC

Figure 3 depicts the Kaplan–Meier survival curves for PFS and OS according to thyroid status. There were significant differences in PFS (log-rank P = 0.011) and a trend in OS (log-rank P = 0.053) among the three groups (thyroid AE, NTIS, and no thyroid dysfunction). Patients who developed thyroid AEs had the most favorable PFS; their median PFS was not reached (NR) by the end of follow-up and was significantly longer than that in NTIS patients (median 2.57 months, 95% CI 2.17–2.97, P = 0.007) and also longer than that in patients without any thyroid dysfunction (median 5.03 months, 95% CI 3.52–6.55, P = 0.022). In contrast, OS was poorest in the NTIS group: median OS for NTIS patients was only 5.40 months (95% CI 3.11–7.69), whereas median OS was not reached for the no thyroid dysfunction group and was 27.53 months (95% CI 5.18–49.89) for the thyroid AE group. The OS of patients with NTIS was significantly shorter than that of patients in the other two groups (NTIS vs. no thyroid dysfunction: P = 0.040; NTIS vs. thyroid AE: P = 0.007, log-rank test).

Fig. 3.

A Progression-free survival (PFS) and B overall survival (OS) in patients with hepatocellular carcinoma undergoing Ate/Bev treatment stratified by thyroid function status

Treatment exposure was longest among thyroid irAE patients (median 6 cycles, range 3–34), followed by no-dysfunction (median 5 cycles, range 2–20), and shortest in NTIS (median 4 cycles, range 2–16). Corresponding median follow-up durations were 11.2 months, 10.3 months, and 5.3 months, respectively.

In the univariate Cox analysis for PFS, patients with thyroid AEs had a 68% lower hazard of progression than those without thyroid dysfunction (hazard ratio [HR] 0.321; 95% confidence interval [CI] 0.110–0.937; P = 0.038). Older age (≥ 65 years) was associated with a reduced risk of progression in univariate analysis (HR 0.504; 95% CI 0.262–0.967; P = 0.039). Older age and thyroid dysfunction status were entered into the multivariable Cox regression model. In the adjusted analysis, the association between thyroid AEs and PFS became only marginally significant (adjusted HR, 0.370; 95% CI, 0.125–1.099; P = 0.073), while older age was not no longer a significant risk factor (Table 2).

Table 2.

Univariate and multivariate cox regression analysis of prognostic factors associated with progression-free survival in hepatocellular carcinoma using atezolizumab and bevacizumab

	Univariate		Multivariate
	HR (95% CI)	P Value	Adjusted HR (95% CI)	P Value
Age (≥ 65 vs. < 65)	0.504 (0.262–0.967)	0.039*	0.634 (0323–1.244)	0.185
Sex (male vs. female)	0.716 (0.337–1.522)	0.386
ALBI grade (2 vs. 1)	1.753 (0.884–3.477)	0.108
Tumor size	1.004 (0.932–1.081)	0.926
Barcelona clinical liver cancer stage (C vs. B)	1.383 (0.719–2.660)	0.331
AFP (≥ 400 vs. < 400)	1.676 (0.878–3.199)	0.117
Presence of PVT	1.105 (0.582–2.096)	0.760
Extra-hepatic spread	1.262 (0.654–2.434)	0.488
Etiology of HCC (viral vs. Non-viral)	0.805 (0.335–1.931)	0.627
NLR	1.015 (0.943–1.093)	0.691
Baseline TSH (per 1.0 mIU/L)	0.991 (0.668–1.470)	0.964
Baseline FT4 (ng/mL)	0.782 (0.057–9.376)	0.808
Baseline low T3 (< 64)	0.981 (0.427–2.253)	0.965
Thyroid dysfunction
Thyroid AE ( −) and NTIS ( −)	Reference
Thyroid AE ( +)	0.321 (0.110–0.937)	0.038*	0.370 (0.125–1.099)	0.073
NTIS ( +)	1.809 (0.845–3.872)	0.127	1.709 (0.795–3.672)	0.170

Abbreviations: AFP = alpha-fetoprotein; ALBI = Albumin-Bilirubin; BMI = body mass index; FT4 = free thyroxine, HCC = hepatocellular carcinoma; MVI = macrovascular invasion; NLR = neutrophil-to-lymphocyte ratio; NTIS = non-thyroidal illness syndrome; PVT = portal vein thrombosis; T3 = triiodothyronine; thyroid AE = thyroid adverse events; TPO Ab = thyroid peroxidase antibody; TSH = thyroid stimulating hormone

^*Denotes P value < 0.05

For OS, univariate analysis showed that patients with BCLC stage C disease had significantly worse OS (HR 3.707; 95% CI 1.255–10.946; P = 0.018) than those with stage B disease. NTIS was associated with a 2.614-fold higher hazard of death (HR, 2.614; 95% CI, 1.081–6.323, P = 0.033 vs. no thyroid dysfunction). BCLC stage and thyroid dysfunction status were included in the multivariable model. Bothe retained statistical significance: BCLC stage C remained associated with inferior OS (adjusted HR 4.072; 95% CI 1.336–12.413; P = 0.014) and NTIS emerged as an independent predictor of mortality (adjusted HR, 2.511; 95% CI, 1.036–6.085, P = 0.041). Furthermore, direct comparison of NTIS to the thyroid AE group revealed a striking difference: NTIS patients had an over sixfold higher adjusted hazard of death than patients who experienced thyroid AEs (adjusted HR, 6.464; 95% CI, 1.902–21.964, P = 0.003) (Table 3).

Table 3.

Univariate and multivariate cox regression analysis of prognostic factors associated with overall survival in hepatocellular carcinoma using atezolizumab and bevacizumab

	Univariate		Multivariate
	HR (95% CI)	P Value	Adjusted HR (95% CI)	P Value
Age (≥ 65 vs. < 65)	0.502 (0.230–1.097)	0.084
Sex (male vs. female)	0.648 (0.275–1.526)	0.321
ALBI grade (2 vs. 1)	2.096 (1.027–4.954)	0.071
Tumor size	1.051 (0.970–138)	0.226
Barcelona clinical liver cancer stage (C vs. B)	3.707 (1.255–10.946)	0.018*	4.072 (1.336–12.413)	0.014*
AFP (≥ 400 vs. < 400)	1.384 (0.654–2.927)	0.396
Presence of PVT	2.012 (0.916–4.419)	0.082
Extra-hepatic spread	1.884 (0.913–3.887)	0.087
Etiology of HCC (viral vs. Non-viral)	1.044 (0.399–2.735)	0.930
NLR	1.061 (0.982–1.145)	0.133
Baseline TSH (per 1.0 mIU/L)	1.262 (0.813–1.959)	0.300
Baseline FT4 (ng/mL)	0.846 (0.052–13.681)	0.906
Baseline low T3 (< 64)	1.517 (0.624–3.686)	0.357
Thyroid dysfunction
Thyroid AE ( −) and NTIS ( −)	Reference		Reference
Thyroid AE ( +)	0.693 (0.303–2.211)	0.693	0.640 (0.225–1.816)	0.401
NTIS ( +)	2.614 (1.081–6.323)	0.033*	2.511 (1.036–6.085)	0.041*

Abbreviations: AFP = alpha-fetoprotein; ALBI = Albumin-Bilirubin; BMI = body mass index; FT4 = free thyroxine, HCC = hepatocellular carcinoma; MVI = macrovascular invasion; NLR = neutrophil-to-lymphocyte ratio; NTIS = non-thyroidal illness syndrome; PVT = portal vein thrombosis; T3 = triiodothyronine; thyroid AE = thyroid adverse events; TPO Ab = thyroid peroxidase antibody; TSH = thyroid stimulating hormone

^*Denotes P value < 0.05

Landmark analyses at 6 weeks (n = 68 at risk) and 3 months (n = 61 at risk) yielded consistent findings. NTIS remained significantly associated with poor OS (HR = 4.093, P = 0.007 at 6 weeks; HR = 2.522, P = 0.075 at 3 months), while thyroid irAEs continued to show trends toward improved PFS (HR =0.360, P = 0.065 at 6 weeks; HR =0.331, P = 0.088 at 3 months).

The clinical efficacy endpoints mirrored these survival findings. Patients with thyroid AEs achieved a substantially higher best objective response rate (ORR, 53.8%) compared to 37.0% in patients without irTAE and only 9.1% and those with NTIS (P = 0.006 across groups; Supplementary Table 2). The disease control rate followed a similar pattern (the thyroid AE group had the highest DCR at 92.3%, compared to 69.6% in the no irTAE group and 36.4% in the NTIS group). Notably, patients with NTIS had a very low response to Ate/Bev; only one of 11 patients achieved a PR, and none had CR. In contrast, over half of the patients with thyroid AE had a response (including three CR).

Factors associated with irTAE development

We explored the baseline factors that might predispose the patients to thyroid irAEs (Supplementary Table 3). In the univariate logistic analysis, HCC etiology was significantly associated with thyroid AEs: patients with viral-associated HCC (hepatitis B or C) were less likely to develop thyroid AEs than those with non-viral etiologies. Specifically, viral etiology conferred an OR of 0.24 (95% CI 0.059–0.982, P = 0.047) for thyroid AE occurrence. Thyroid immune events were observed predominantly in patients with alcoholic or cryptogenic HCC, whereas those with HBV/HCV had relatively fewer thyroid AEs. Conversely, heavy alcohol use was associated with higher odds of NTIS: 5 of 11 patients with NTIS had a history of significant alcohol consumption, and alcohol etiology had an OR of 3.958 (95% CI 0.966–16.223, P = 0.056) for developing NTIS (a strong trend, though not reaching 0.05 significance). No other baseline factors (age, sex, BMI, baseline ALBI grade, and performance status) were significantly associated with either type of irTAE. Baseline TSH and T3 values were also not predictive, likely because we excluded those with any baseline thyroid dysfunction.

Association of NTIS occurrence with deterioration of ALBI Grade, BMI, and serum albumin level

To understand why patients with NTIS had such poor outcomes, we examined changes in clinical parameters at the time of NTIS detection. Among the 11 patients with NTIS, we observed a significant decline in BMI from baseline (median, 23.11 kg/m²) to the time of NTIS diagnosis (median, 22.49 kg/m²; P = 0.028). This corresponds to a median weight loss of roughly 1.5–2 kg, often accompanied by cancer cachexia progression. Similarly, serum albumin dropped markedly in NTIS patients, with the median albumin level falling from 3.51 g/dL at baseline to 2.73 g/dL when NTIS occurred (P = 0.004). These changes indicate a deterioration in the nutritional and hepatic synthetic status around the time NTIS was identified. In contrast, patients who developed thyroid AEs showed minimal changes in these parameters; their median BMI was essentially unchanged (from 22.86 to 22.75 kg/m², P = 0.371), and albumin slightly increased on average (from 3.94 to 4.02 g/dL, P = 0.581) during thyroid dysfunction (Supplementary Fig. 1). The stability of nutritional status in patients with thyroid AE suggests that their overall health remains better preserved, aligning with their more favorable outcomes.

Liver function, as assessed by the ALBI grade, also differed between the groups. Supplementary Fig. 2 illustrates that patients with NTIS experienced a shift toward worse ALBI grades concurrent with NTIS onset; several patients moved from ALBI 1 to 2 or from ALBI 2 to 3, indicating declining liver function (worsening bilirubin or dropping albumin). In contrast, the ALBI grade distribution in the thyroid AE group remained similar before and after irTAE, with no significant hepatic decompensation observed.

Discussion

In this study of patients with unresectable HCC receiving Ate/Bev, immune-related thyroid dysfunction occurred in one-third of patients and had divergent prognostic implications. Thyroid adverse events, such as hypothyroidism or thyrotoxicosis, were associated with a favorable trend in PFS and higher tumor response, while NTIS was strongly associated with poor survival, likely as a reflection of underlying disease severity. To our knowledge, this is the first study to identify NTIS as an adverse prognostic factor in patients receiving Ate/Bev.

The incidence of immune-related thyroid dysfunction (34.3%) in our cohort aligns with prior reports. Hypothyroidism (7.1%) and thyrotoxicosis (11.4%) rates were comparable to IMbrave150 (~ 10%) and real-world studies reporting 10.9–17.3% and 4.6–5.8%, respectively [25, 26]. These differences may reflect our strict exclusion of baseline thyroid dysfunction and classification of subclinical or isolated low FT4 cases as NTIS, contributing to the higher NTIS rate (15.7%) [26]. The onset of irTAEs varied widely, occasionally occurring > 1 year after therapy initiation, underscoring the need for continued thyroid monitoring throughout immunotherapy [25–27]. Routine assessment of TSH and FT4 levels every 1–2 cycles (or quarterly after stabilization) is recommended. Early detection not only guides symptom management but may also offer prognostic insight.

A key finding was the association between thyroid AEs and the favorable outcomes. Patients with immune-related hypothyroidism or thyrotoxicosis exhibited higher response rates and longer survival than those without these irTAEs. Although the PFS advantage did not reach significance in the multivariate analysis, the trend aligns with prior studies across multiple cancers, including melanoma, lung, and renal cancers [28–31]. This likely reflects heightened immune activation: the same T-cell or antibody responses targeting tumor cells may also trigger thyroid autoimmunity. Importantly, thyroid AEs are generally manageable with hormone replacement or antithyroid agents and seldom require immunotherapy discontinuation. In contrast to severe irAEs, such as hepatitis or colitis, which may necessitate high-dose steroids or treatment interruption, thyroid AEs rarely lead to immunosuppression. In our cohort, no patient with thyroid AEs required systemic steroids or halted treatment with Ate/Bev owing to toxicity [32]. These findings reinforce the clinical significance of thyroid irAEs as favorable biomarkers. As noted by Kotwal et al., PD-L1–related thyroiditis was associated with improved survival [31]. Clinicians should view thyroid AEs as evidence of immune engagement and continue treatment with appropriate endocrine support.

A key and novel finding of our study was the strong negative prognostic impact of the NTIS. Patients with NTIS had a > sixfold increased mortality risk compared to those with thyroid AEs and > 2.5-fold increased mortality risk compared to those without irTAEs. Their outcomes resembled those of historical pre-immunotherapy HCC, with a median PFS of ~ 7 months and minimal treatment response [33, 34]. Importantly, NTIS itself is likely a surrogate marker of aggressive disease biology and patient frailty, rather than a direct cause of worse outcomes. NTIS likely reflects end-stage disease physiology rather than immunotherapy resistance. Patients with NTIS showed significant clinical deterioration, including weight loss, hypoalbuminemia, and progression to ALBI grade 3 liver dysfunction. These features suggest that NTIS is a surrogate marker of hepatic decompensation and systemic failure. This interpretation is consistent with the classic NTIS literature, where low T3 levels are strongly linked to disease severity and mortality in critical illness and cancer. Prior studies have established NTIS as an independent predictor of poor survival in several malignancies [35–39]. Our findings extend this evidence to HCC, reinforcing the notion that NTIS universally signals an adverse prognosis in oncology.

The pathophysiology of NTIS involves impaired thyroid hormone metabolism during illness. Decreased Type 1 and Type 2 deiodinase activity reduces T4-to-T3 conversion, while increased Type 3 activity diverts T4 to inactive reverse T3, leading to low T3 and high rT3, with inappropriately normal or suppressed TSH [40]. The liver is the main site for peripheral T3 production and plays a central role in this process [20, 21]. In advanced HCC or cirrhosis, hepatic dysfunction can directly impair T3 generation. A recent study demonstrated how liver inflammation and fibrosis disrupt systemic thyroid hormone homeostasis via altered hepatic deiodinase activity [20]. In our NTIS cohort, clinical liver deterioration likely contributed to low T3 levels.

In addition, proinflammatory cytokines, particularly IL-6, have been implicated in NTIS pathogenesis [21]. IL-6 levels correlate inversely with T3 levels in patients, and IL-6 knockout mice show a blunted NTIS response during illness [21]. IL-6 also suppresses TSH and impairs tissue-level thyroid hormone activation. As advanced HCC, particularly NASH-related, is characterized by an IL-6-rich inflammatory state [22], we speculate NTIS reflects both hepatic dysfunction and systemic inflammation. This milieu promotes immunosuppression and cachexia, with impaired anti-tumor immunity via T-cell exhaustion and myeloid-derived suppressor cell activation [41]. Therefore, NTIS in our cohort likely represents an adaptive response to systemic illness rather than an immune-mediated thyroid injury. In contrast, thyroid irAEs reflect organ-specific autoimmunity induced by checkpoint inhibitors. Distinguishing NTIS from irTAEs is important, as they carry divergent implications for patient outcomes and therapeutic responses.

In our cohort, patients who developed NTIS showed notable declines in BMI, serum albumin, and ALBI grade, indicating nutritional and hepatic deterioration at the time of NTIS onset. Rather than a direct endocrine toxicity, NTIS likely represents a systemic adaptive response to critical illness. We hypothesize that it reflects the cumulative effects of advanced liver dysfunction and cancer-related inflammation, identifying patients with aggressive tumor biology or declining hepatic reserve. These factors are independently associated with poor prognosis, reinforcing the interpretation of NTIS as a surrogate marker of systemic decompensation, distinct from immune-related thyroid events.

Whether NTIS contributes directly to adverse outcomes remains unclear. Thyroid hormones support metabolic, cardiovascular, and immune functions, and profound T3 deficiency may worsen cardiac output, renal function, and drug clearance. However, it is debated whether NTIS is purely adaptive or partly maladaptive. Small trials of thyroid hormone replacement in NTIS—mainly in critical care—have yielded inconclusive results, with no clear survival benefit [42]. In oncology, indiscriminate T3/T4 use raises concerns of stimulating tumor metabolism, as preclinical studies suggest thyroid hormones may promote cancer proliferation [43]. Consequently, NTIS treatment is not standard in cancer care, and none of the patients in our study received hormone therapy. While causality cannot be inferred, NTIS often coincided with clinical deterioration, supporting its role as a consequence—not a driver—of disease progression. Taken together, these findings support the view that NTIS is predominantly a consequence of advanced disease and systemic decompensation, rather than a modifiable driver of poor outcomes.

This study has several clinical implications. First, routine monitoring of thyroid function—including T3—is warranted in HCC patients receiving Ate/Bev. NTIS may be missed if only TSH and FT4 are measured, yet it signals poor prognosis and often coincides with hepatic decompensation or cachexia. Early identification should prompt comprehensive disease reassessment and optimization of supportive care. While thyroid hormone replacement is not indicated, nutritional or anabolic support may be considered for these patients. NTIS may also suggest treatment failure, supporting the timely consideration of second-line therapies or clinical trial enrollment.

In contrast, thyroid irAEs (hypo- or hyperthyroidism) may indicate effective immune activation and are typically manageable with standard endocrine care. Immunotherapy need not be discontinued, and continued treatment may benefit these patients most. Emerging evidence supports the use of irAEs as prognostic markers and trial stratification variables. While exploratory, tracking thyroid irAEs offers valuable clinical insight and may inform decisions such as extending therapy despite radiographic progression.

This study had some limitations. The modest sample size (n = 70) and relatively small number of NTIS and thyroid AE cases reduced the power for multivariable and subgroup analyses. As a retrospective analysis, potential selection bias and unmeasured confounders exist, although consistent laboratory monitoring and objective endpoints helped mitigate this. We lacked systematic data on cytokines, thyroid antibodies, or detailed body composition, limiting mechanistic interpretation. In particular, thyroid peroxidase antibody (TPOAb) was only available in a subset of patients, precluding the analysis of its relationship with thyroid irAEs or prognosis. Data on corticosteroid use were also unavailable, though no NTIS cases involved steroid-treated irAEs. Therapy was not randomized, and patients with NTIS may have had earlier treatment discontinuation, contributing to the poor outcomes. Still, the association between NTIS and survival remained significant after adjusting for tumor stage.

Despite these limitations, our findings highlight NTIS as a clinically relevant and under-recognized biomarker for HCC. It may reflect systemic frailty, tumor aggressiveness, and poor treatment response. Future prospective studies are warranted to validate NTIS as a prognostic tool and to explore its relevance in other ICI-based regimens, such as durvalumab plus tremelimumab or PD-1–Tyrosine Kinase Inhibitor combinations.

Conclusion

In conclusion, NTIS is a strong negative prognostic indicator in patients with HCC undergoing Ate/Bev therapy, likely reflecting systemic and hepatic deterioration. In contrast, thyroid AEs may indicate robust immune activation and are associated with favorable outcomes. These findings highlight the importance of routine monitoring of thyroid function, including T3 levels, as part of risk stratification and management decisions in immunotherapy-treated HCC.

Supplementary Information

Below is the link to the electronic supplementary material.

Abbreviations

Ate/Bev

Atezolizumab plus bevacizumab

FT3

Free triiodothyronine

FT4

Free thyroxine

HCC

Hepatocellular carcinoma

irAE

Immune-related adverse event

irTAE

Immune-related thyroid adverse events

mRECIST

Modified response evaluation criteria in solid tumors

NTIS

Non-thyroidal illness syndrome

OS

Overall survival

PFS

Progression-free survival

TACE

Transarterial chemoembolization

TPOAb

Thyroid peroxidase antibody

TSH

Thyroid-stimulating hormone

Author contributions

All authors contributed to the study conception and design. Material preparation, data collection and analysis were performed by IWC, WT and CWL. IWC analyzed the data. The first draft of the manuscript was written by IWC and all authors commented on previous versions of the manuscript. WT and CYL conceived and supervised the project. All authors read, revised and accepted the submitted manuscript.

Funding

This work was supported by Linkou Chang Gung Memorial Hospital under Grade [CMRPG3M1721, CMRPG3M1691]; National Science and Technology Council, Taiwan under Gant [NMRPD1P1141 and NRRPD1Q0171]. The funders had no role in the study design, data collection, analysis, and interpretation, preparation of the manuscript or decision to publish the results. The authors thank the Liver Disease Center and Clinical Trial Center at Chang Gung Memorial Hospital for research support. No pharmaceutical company was involved in the study design, data collection, analysis, or manuscript preparation.

Data availability

Data are not publicly available due to ethical reasons. Further inquiries can be directed to the corresponding authors.

Declarations

Conflict of interest

The authors have no conflicts of interest to declare.

Ethical approval

The study was conducted following the Declaration of Helsinki and approved by Institutional Review Board (IRB No. 202200991B0C501). Due to the retrospective nature of this study, informed consent was waived by the Ethics Committee.