The impact of thyroid hormones on patients with hepatocellular carcinoma

Background & aims Hypothyroidism has recently been proposed as predisposing factor for HCC development. However, the role of thyroid hormones (TH) in established HCC is largely unclear. We investigated the impact of TH on clinical characteristics and prognosis of HCC patients. Methods Of 838 patients diagnosed with nonsurgical HCC at the Division of Gastroenterology and Hepatology/Medical University of Vienna between 1992 and 2012, 667 patients fulfilled the inclusion criteria. The associations of thyroid function tests with patient, liver, and tumor characteristics as well as their impact on overall survival (OS) were investigated. Results Thyroid hormone substitution was more often observed in patients with low thyroid-stimulating hormone (TSH) concentration and in patients with elevated free tetraiodthyronine (fT4). Patients with high TSH (>3.77uU/ml) concentrations had larger tumors, while the opposite was true for patients with low TSH (<0.44uU/ml) concentrations. Subjects with elevated fT4 (>1.66ng/dl) were more likely to have elevated CRP. While TSH was only associated with OS in univariate analysis (≤1.7 vs. >1.7uU/ml, median OS (95%CI), 12.3 (8.9–15.7 months) vs. 7.3 months (5.4–9.2 months); p = 0.003), fT4 (≤1.66 vs. >1.66ng/dl, median OS (95%CI), 10.6 (7.5–13.6 months) vs. 3.3 months (2.2–4.3 months); p = 0.007) remained an independent prognostic factor for OS (HR (95%CI) for fT4>1.66ng/dl, 2.1 (1.3–3.3); p = 0.002) in multivariate analysis. Conclusions TSH and fT4 were associated with prognostic factors of HCC (i.e., tumor size, CRP level). Elevated fT4 concentrations were independently associated with poor prognosis in HCC. Further studies are needed to characterize the role of TH in HCC in detail.


Introduction
Components of thyroid hormone signaling are implicated in the development and progression, as well as in the prevention of various cancers including hepatocellular carcinoma (HCC) [1][2][3]. HCC usually develops in patients with liver cirrhosis [4,5] and represents the second most common cause of cancer-related mortality in men globally [6].
The thyroid hormones (TH) tetraiodthyronine (T 4 ) and, to a lesser extent, triiodthyronine (T 3 ) are produced in the thyroid gland in a complex multi-step process, which is tightly controlled by the hypothalamus-pituitary-thyroid axis. Hypothalamic neurons produce the tripeptide thyrotropin-releasing-hormone (TRH), which uses the pituitary portal venous system to reach the anterior pituitary gland, where it binds to the TRH receptor in thyrotropic cells, and thereby stimulates both the production and pulsatile release of the glycoprotein thyrotropin (thyroid-stimulating hormone, TSH). By binding to the TSH-receptor situated at the basolateral membrane of thyrocytes, TSH activates thyroid hormone production and secretion.
The vast majority of both thyroid hormones is bound to the plasma proteins and therefore unable to bind to receptors in target tissues. This constellation allows preparation of a large pool of hormones, which can quickly be released when needed. Conversely, the smaller, unbound fraction of thyroid hormones is responsible for the actual physiological effects. The thyroid gland predominantly secretes T 4 , most of which is subjected to peripheral metabolism by deiodinase enzymes [7].
Since TH display extensive influence on regulatory mechanisms affecting the control of cellular growth and metabolism [7][8][9][10], there are myriad possible modes of interaction with both the development and progression of HCC. Although hypothyroidism has been shown to be associated with an elevated risk of HCC development [2,11,12], the role of TH in established HCC remains to be elucidated. Several preclinical studies have linked TH signaling to tumorpromoting actions, especially at advanced stages of hepatocarcinogenesis [13][14][15]. However, little is known about both the prevalence and clinical role of TH in patients with manifest HCC. Hence, we aimed to investigate the association of TH with clinical patient and tumor characteristics as well as their impact on the prognosis of HCC patients.

Patient selection
Data from all patients who were diagnosed with HCC by biopsy or radiological imaging according to the European Association for the Study of the Liver (EASL) [16] diagnostic criteria between August 1992 and February 2013 at the Medical University of Vienna were retrospectively collected and incorporated into a database. Patients who had received surgical treatment for HCC at any time after diagnosis were excluded from analysis. Only patients aged !18 years with TSH levels available at the time of diagnosis were eligible. Collection and retrospective analysis was approved by the Ethics Committee of the Medical University of Vienna.

Data acquisition
The date of HCC diagnosis (date of biopsy if available or diagnostic imaging) was considered the baseline of this study. Patient information was collected in a preexisting MS Access 2010 database (HCC database) in a pseudo-anonymous manner. Patient characteristics, laboratory parameters including thyroidal function (TSH, T3, T4, fT3, fT4), tumor characteristics, and parameters representing liver function were recorded from the patients' charts and from the electronic patients' information system. Determination of thyroid status was based on TSH and fT 4 levels. We formed 5 subgroups: euthyreosis, hyperthyreosis (primary, secondary, and euthyroid hyperthyroxinemia), subclinical hyperthyreosis, subclinical hypothyreosis, hypothyreosis (primary and secondary), and thyroid hormone substitution. Liver function was assessed by MELD score and Child-Pugh score. The latter has been incorporated in the Barcelona Clinic Liver Cancer classification, the most widely used staging system for HCC, which has been endorsed by the American and European HCC guidelines [16,17].

Statistics
Baseline characteristics were summarized using descriptive statistics. Chi square test or Fisher's exact test were used to compare nominal data. Overall survival (OS) was defined as the time from date of diagnosis (date of biopsy if available or diagnostic imaging) until date of death or last contact. Survival curves were calculated using the Kaplan-Meier method and were compared by means of the log rank test (univariate analysis). Variables that reached a pvalue of <0.05 in univariate analysis were entered into a multivariate analysis. The multivariate analysis was performed using a Cox proportional hazard regression model. Statistical tests were two-sided and a p-value <0.05 was considered significant. All statistical analyses were performed using SPSS version 17.0 (SPSS Inc., Chicago, IL).

Patient characteristics
Of 667 patients included (Fig 1), 82% of the patients were male, with a male to female ratio of 4.6:1. The median age at diagnosis was 64 years (range, 32-87 years). Six percent of all patients were on thyroid hormone substitution. Detailed patient characteristics are shown in Table 1. Additional information on thyroid function is given in S1 Table. Mean follow-up was 65.5 months. Five hundred and forty-four (82%) patients died during the observation period.
Free T 4 (fT 4 ) levels were available in 313 patients. Since sample size of hypothyroid patients (fT 4 levels below 0.76ng/dl) was too small (n = 4) to allow for robust analysis, fT 4 levels were divided into two groups ( 1.66 and >1.66ng/dl, hereafter referred to as 'normal' and 'elevated', respectively). The proportion of female patients was higher in patients with elevated fT 4 levels (normal vs. elevated fT 4 , 16 vs. 36%; p = 0.025). Thyroid hormone substitution was more frequently observed in patients with elevated fT 4 (normal vs. elevated fT 4 , 6 vs. 32%; p<0.001). Finally, subjects with elevated fT 4 were more likely to have elevated CRP levels (normal vs. elevated fT 4 , 51 vs. 88%; p = 0.001). No association was observed between other variables representing liver function and tumor burden, respectively (S3 Table).
Since information on T 3 , T 4 , and fT 3 was missing in 82-97% of patients due to the retrospective character of this analysis (S1 Table), we could not assess their association with patient, liver, and tumor characteristics.
Given that fT 4 levels were only available in 313 patients, we did not include both TSH and fT 4 in the same multivariate Cox regression model, but analyzed them separately with all other variables that were significantly associated with OS in univariate analysis. Finally, fT 4 (HR (95%CI) for fT 4 >1.66ng/dl, 2.1 (1.3-3.3); p = 0.002) remained an independent prognostic factor for OS (Table 4), while TSH was not significantly associated with OS in multivariate analysis (S4 Table). After

Discussion
In our cohort, 10% of patients had elevated TSH levels similar to the prevalence of hypothyroidism in patients with HCC reported by Hassan et al. (HCC patients vs. controls, 11.7 vs. 8%) [12]. Hypothyroidism has profound effects on metabolism and has therefore been linked to various conditions which either directly constitute an HCC risk factor or have the ability to contribute to the development of known predisposing conditions for HCC, such as obesity [18][19][20][21], diabetes [22][23][24][25], non-alcoholic fatty liver disease (NAFLD) [20,26,27], and hepatitis C infection [12,28,29]. Interestingly, Tarantino and colleagues reported that BMI predicted the presence of spleno-renal shunts and spleno-renal shunts were associated with an increased HCC incidence [30].
However, in our study, high serum TSH concentration was neither associated with type II diabetes mellitus nor BMI. Although etiology of liver disease was not significantly associated with TSH levels, HCV was more frequently observed in patients with high TSH levels (41%) compared to those with normal (29%) or low (20%) TSH. In contrast, Reddy and colleagues [31] found that hypothyroidism was more prevalent in patients with unknown etiology than in those with HCV or alcoholic liver disease. Additionally, we observed NASH only in patients with normal or elevated TSH but not in those with low TSH levels.
Notably, elevated TSH was associated with larger tumors in our study. In contrast, in breast cancer, hypothyroid patients were more likely to be diagnosed with a smaller tumor and at an earlier stage compared to euthyroid patients [32]. Moreover, in prostate cancer, serum T 3 was higher in more advanced clinical stage, even though none of the men had levels above the normal range [33].
In terms of survival, patients with higher TSH levels showed a significantly worse outcome in univariate analysis. However, this effect did not hold true upon multivariate analysis when adjusting for other prognostic factors including Child-Pugh class, tumor size, performance status, macrovascular invasion, extrahepatic spread, tumor treatment, AFP, and CRP levels.
Next, we investigated the prevalence of abnormal fT 4 levels, available in 313 of 667 patients. Elevated fT 4 was found in 25 patients (4% of all patients, n = 667; 8% of those whose fT 4 levels where available, n = 313). In comparison, a large study conducted in the United States reported a prevalence of only 0.5% for clinical hyperthyroidism. Notably, they used total instead of free T 4 concentrations but adjusted for the common T 4 -confounders pregnancy and estrogen therapy [34]. In an epidemiologic study by Hassan et al., the prevalence of hyperthyroidism was 1.9% among HCC patients and 1.3% among controls [12]. In our cohort, elevated fT 4 concentrations were not associated with etiology of liver disease and variables representing liver function or tumor burden. However, patients with elevated fT 4 more frequently had elevated CRP levels which indicate worse prognosis in HCC and could be a reflection of the "inflammatory field effect" [35]. This effect could directly fuel tumor progression as CRP levels are an accepted surrogate marker for the release of Interleukin (IL) -6, an important regulator of CRP secretion, which in turn is associated with both acceleration of HCC development and metastasis [35][36][37][38]. Additionally, thyroid hormone replacement was more common in patients with increased fT 4 levels (fT 4 1.66 vs. >1.66, 6 vs. 32%).
We next investigated the impact of fT 4 on the prognosis of HCC patients and found that elevated fT 4 levels at the time of HCC diagnosis were significantly associated with poor OS. This effect held true even after adjusting for other known prognostic factors for HCC in a multivariate Cox regression model. These results are supported by several preclinical studies showing that thyroid hormone signaling promotes tumor invasiveness and metastasis [2,13,15,39,40], and could be of clinical relevance mainly for two reasons. First, fT 4 is a valuable prognostic parameter in HCC that is widely available, non-invasively collectable, and objective. Second, considering both the association between thyroid hormone substitution and elevated fT 4 levels as well as the strong prognostic relevance of elevated fT 4 , our data could suggest some caution when replacing thyroid hormones in patients with established HCC, as hormone replacement-induced fT 4 elevation might render tumors more aggressive. This should especially be considered in patients with advanced HCC receiving sorafenib therapy where hypothyroidism often occurs [41][42][43], and might prompt physicians to initiate hormone replacement therapy.
Notably, other studies reported that T 3 treatment led to regression of preneoplastic lesions in rodent models of heptocarcinogenesis [44,45].
The main limitation of this study was the lack of patients with low fT 4 levels, which precluded the investigation of effects of overt hypothyroidism on HCC and only allowed analyzing the effect of hypothyroidism using TSH levels. This especially impeded the analysis of the impact of thyroid hormones on HCC with metabolic background. Furthermore, the small sample size of patients with abnormal fT 4 levels did not allow subgroup analyses within the BCLC stages and Child-Pugh class. As thyroid hormones may exert opposing effects at different stages of HCC development and progression some effects could evade detection when analysis is performed without differentiation according to tumor stage [2]. However, after grouping patients into early-intermediate (BCLC A-B) and advanced-terminal stage (BCLC C-D), the negative impact of elevated fT 4 on survival remained significant in both subgroups. Additionally, non-thyroidal illness syndrome (NTIS), characterized by low serum T 3 with normal T 4 levels, is associated with HCC and other malignancies [46,47]. The fact that T 3 and fT 3 levels were missing in most of our patients represents another potential bias we could not address in our analysis. In light of the stated limitations, this study could not adequately investigate the hypothesis of a dual role of thyroid hormones in HCC [2].
In conclusion, high TSH level was associated with larger tumor size but not with survival when adjusted for known prognostic factors for HCC. Elevation of fT 4 resulted in poor survival and remained an independent prognostic factor for OS. These results can be considered as hypothesis-generating paving the path for further work. Prospective studies including clinically hypothyroid patients and subgroup analyses of all tumor stages are needed to further elucidate the role of thyroid hormones in HCC.