Propensity score matching analysis of the prognosis for the rare oxyphilic subtype of thyroid cancer (Hurthle cell carcinoma)

There is controversy regarding the prognosis of patients with oxyphilic thyroid cancer (OXTC). The present study compared the prognoses of OXTC, papillary thyroid cancer (PTC), and follicular thyroid cancer (FTC), in order to provide a new perspective regarding the treatment guidelines for these diseases. We evaluated data from patients with thyroid cancer who were included in the Surveillance, Epidemiology, and End Results database between 2004 and 2013. Patient mortality was evaluated using Cox proportional hazards regression analyses and Kaplan-Meier analyses with log-rank tests. The multivariate Cox regression analysis revealed that the cancer-specific survival rate for OXTC was similar to that for PTC, but higher than that for FTC. However, after propensity score matching for relevant factors, the cancer-specific survival rate for OXTC was higher than that for PTC and FTC. This unexpected result provides new implications for the treatment of patients with OXTC.


INTRODUCTION
The incidence of thyroid cancer has risen rapidly during recent decades [1][2][3]. Papillary thyroid cancer (PTC) and follicular thyroid cancer (FTC) (i.e., differentiated thyroid cancer) account for >90% of all thyroid malignancies and are the most common types of thyroid carcinoma [1,4,5]. However, there are other rare histological variants of thyroid cancer, such as tall cell, solid, or oxyphilic thyroid cancer (OXTC) [6][7][8]. OXTC is also known as Hurthle cell carcinoma, and only accounts for 2-5% of all thyroid cancers [9][10][11]. The current World Health Organization classification system has defined OXTC as a variant of FTC, although Ganly et al. have reported that OXTC may be distinct from PTC and FTC, based on the unique genetic features of OXTC [12]. Moreover, given the rarity of OXTC, few studies have www.impactjournals.com/oncotarget

Risk factors for cancer-specific and all-cause mortality
The univariate Cox regression analyses revealed that cancer-specific mortality was associated with age, sex, race, histological type, TNM stage, extension, radiation treatment, and surgical approach. The multivariate Cox regression model revealed that PTC was an independent risk factor for cancer-specific mortality, compared to OXTC ( Table 3). The univariate Cox regression analyses revealed that all-cause mortality was associated with age, sex, race, histological type, TNM stage, multifocality, extension, radiation treatment, and surgical approach. The multivariate Cox regression model revealed that PTC was also an independent risk factor for all-cause mortality, compared to OXTC (Table 3).

Adjusting for patient characteristics using PSM
The cancer-specific and all-cause mortality rates were significantly different when we compared OXTC to PTC and FTC (both p < 0.001, Figure 1A-1D). Thus, to minimize selection bias, PSM was performed for age, sex, race, TNM stage, multifocality, extension, and radiation treatment. After PSM for age, sex, and race, OXTC was associated with a lower cancer-specific mortality rate, compared to PTC (p < 0.001) and FTC (p = 0.005) (Figure 2A, 2B). After PSM for age, sex, race, TNM stage, multifocality, and extension, OXTC was still associated with a significantly lower cancer-specific mortality rate, compared to PTC (p < 0.001) and FTC (p = 0.005) ( Figure  3A, 3B). After PSM for all relevant factors and radiation and surgery treatment, the cancer-specific mortality rate for OXTC remained significantly lower than that for PTC (p < 0.001) and FTC (p = 0.01) ( Figure 4A, 4B).
After PSM for age, sex, and race, OXTC was associated with a lower all-cause mortality rate, compared to PTC and FTC (both p < 0.001, Figure 5A, 5B). Similar results were obtained after PSM for age, sex, race, TNM stage, multifocality, and extension ( Figure 6A, 6B). After PSM for all relevant factors and radiation and surgery treatment, OXTC was associated with a lower all-cause mortality rate, compared to PTC (p < 0.001) and FTC (p = 0.003) ( Figure 7A, 7B).

DISCUSSION
Oxyphilic thyroid tumors (Hurthle cell tumors) include Hurthle cell adenoma, Hurthle cell thyroid carcinoma, and Hurthle cell papillary thyroid carcinoma. The oxyphilic cells exhibit various features, such as a finely granular eosinophilic cytoplasm and an increased number of mitochondria in the thyroid ultrastructure [13]. However, given the rarity of thyroid malignancies, little is known regarding the long-term survival of patients with OXTC, and conflicting results have been reported in the limited number of single-center studies, which had relatively small samples sizes [14][15][16][17].
In contrast, the SEER database identifies and tracks patients from diverse geographic regions, and is considered the gold standard database for tumor surveillance and survival analysis in the US, as it contains  data from approximately 10% of the American patient population [18]. Therefore, the present study evaluated the prognosis of OXTC using population-based data from the SEER database. Goffredo et al. evaluated SEER data (1998-2009) and reported that OXTC was more aggressive and had a poorer prognosis, compared to the other types of differentiated thyroid cancer [19]. However, that study did not compare the prognoses of OXTC and other thyroid cancers using PSM, and may have been limited by its inability to adjust for confounding factors.
In the present study, patients with OXTC were older (57.69 years old), compared to patients with PTC and FTC. Many studies have indicated that age plays a unique prognostic role in thyroid cancer [18,20,21], and the present study confirmed that age was an independent risk factor for both cancer-specific and all-cause mortality. Furthermore, PSM for age, race, and sex confirmed that OXTC was associated with a better prognosis, compared to PTC and FTC. Therefore, an older age at the diagnosis of OXTC may be associated with a relatively poor prognosis. Previous reports have provided conflicting information regarding whether OXTC has more aggressive clinical features, compared to PTC or FTC [22][23][24][25]. However, it is clear that aggressive clinical characteristics (e.g., nodal metastasis, distant metastasis, and larger tumor size) are associated with more advanced disease and shorter cancer-specific survival. In the present study, PSM for demographic and clinical risk factors revealed that OXTC was associated with lower cancer-specific and all-cause mortality rates, compared to PTC and FTC.
The main purpose of radioiodine (RAI) treatment after total thyroidectomy is the ablation of residual thyroid tissue, which facilitates the early detection of cancer recurrence [26]. However, the current guidelines debate the use of RAI for OXTC [11]. In the present study, patients with OXTC were more likely to receive   OXTC and FTC (B). www.impactjournals.com/oncotarget RAI treatment, compared to patients with PTC and FTC, although there remains insufficient evidence regarding the effectiveness of RAI treatment in this setting.
Ganly et al. have indicated that OXTC may be a unique type of thyroid malignancy, compared to PTC and FTC, as it has unique mutational, transcriptional, cytogenetic, and gene expression changes [12]. For example, OXTC was not associated with alterations in the BRAF, PIK3CA, or PPAR genes, and only 11.1% of OXTC cases involved RAS mutations. Although the Wnt/βcatenin pathway is considered highly activated in OXTC, the SEER database does not contain mutation information and we could not include genetic status in our analyses.
There are several limitations in the present study. First, the SEER database lacks information regarding recurrence and surgery-related comorbidities,  and we could not account for these factors in our analyses. Second, we did not evaluate or consider the patients' family history, vascular invasion status, accurate extension information or other histological characteristics. Third, the SEER database does not include information regarding whether the patients underwent repeat surgery, and this lack of information may have biased our findings.
In conclusion, our results indicate that patients who were diagnosed with OXTC had an unexpectedly good prognosis, compared to patients with PTC or FTC. This information may be useful for selecting appropriate treatments for these patients.

Ethical considerations and data collection
This study's retrospective protocol was approved by our institution's ethical review board and complied with the ethical standards of the Declaration of Helsinki, as well as the relevant national and international guidelines. The SEER project is an American population-based cancer registry that was launched in 1973 and is supported by the Centers for Disease Control and Prevention and the National Cancer Institute. The database contains information regarding cancer incidence, prevalence, and mortality, as well as population-based variables and primary tumor characteristics (i.e., histological subtype) from multiple geographic regions.
The present study included and evaluated SEER data (2004-2013) from patients with thyroid cancer according to their subtype (OXTC, PTC, and FTC) using code C73.9 from the International Classification of Diseases for Oncology (i.e., thyroid, papillary, and/or follicular histology). The eligible diagnostic codes were: "papillary carcinoma", "papillary adenocarcinoma", "oxyphilic adenocarcinoma", "follicular adenocarcinoma", "papillary carcinoma, follicular variant", and "papillary & follicular adenocarcinoma". Cases without American Joint Committee on Cancer staging information (version 6) were excluded to ensure accurate analyses. Cases without information of follow up time were also excluded. The three histological subtypes were compared according to age, sex, race, TNM stage, multifocality, extension, and radiation treatment (i.e., none or refused, external beam radiation therapy, or RAI).

Statistical analyses
The quantitative variables were expressed as mean ± standard deviation (SD), while the categorical ones were presented as percentages. Kaplan-Meier survival curves with the log-rank test were used for determination of all-cause survival and cancer-specific survival. Cox proportional hazard regression analyses were using to estimate hazard ratios and 95% CIs, in order to quantify the effects of the different histological subtypes on cancer-specific and all-cause mortality. PSM was also used to further adjust for potential baseline confounding factors. All p-values were 2-sided, and p-values < .05 were considered significant. Analyses were performed using SPSS version 23.0, Stata/SE version 12 (Stata Corp.), and GraphPad Prism version 6 (GraphPad Software Inc.).