Circulating thyroid stimulating hormone receptor messenger RNA and differentiated thyroid cancer: A diagnostic meta-analysis

Thyroid stimulating hormone receptor messenger RNA (TSHR-mRNA) is over-expressed in thyroid cancer patients, which indicates that TSHR-mRNA is a potential biomarker of thyroid cancer. However, system evaluation for TSHR-mRNA as a diagnostic biomarker of thyroid cancer is deficient. The performance of TSHR-mRNA for thyroid cancer diagnosis was evaluated in this study. Three common international databases as well as a Chinese database were applied for literature researching. Quality assessment of the included literatures was conducted by the QUADAS-2 tool. Totally, 1027 patients from nine studies eligible for the meta-analysis were included in this study. Global sensitivity and specificity for the positivity of TSHR-mRNA in the thyroid cancer diagnosis is 72% and 82%. The value of AUC for this test performance was 0.84. Our meta-analysis suggests that TSHR-mRNA might be a potential biomarker to complete present diagnostic methods for early and precision diagnosis of thyroid cancer. Notably, this findings need validation thorough large-scale clinical studies.


INTRODUCTION
Thyroid cancer is one of the most common cancers in the endocrine system, and the incidence of thyroid cancer is about 1%, and it is also the main cause of death in the endocrine system [1]. There are around 44670 new cases and 1690 deaths of thyroid cancer every year [2]. In the past years, the average annual percentage for thyroid cancer increased 5.3% per year [3]. Currently, the diagnostic methods of thyroid cancer mainly include ultrasound, CT and fine-needle aspiration cytology. In a certain extent, it can improve the accuracy of diagnosis of thyroid cancer. However, surveillance, early and precision diagnosis as well as treatment are still the main problem [4]. Currently there are ThyroSeqV2 next generation sequencing assay to improve the diagnosis of benign and malignant thyroid cancer method [5], which is high accuracy but high cost, and can not be popularized. So new methods is necessary in the diagnosis of thyroid cancer.
Thyroid stimulating hormone receptor (TSHR) is a specific protein of thyroid cells, which is found on the membrane of thyroid gland cells. It plays an important role in the development of thyroid function and thyroid disease. Because TSHR-mRNA is over-expressed in thyroid cancer cells, the measurement of TSHR-mRNA in the circulation is very valuable for the diagnosis of thyroid cancer [4]. In recent years, there have been many reports of TSHR-mRNA in the diagnosis of thyroid cancer. However, a system evaluation of TSHR-mRNA as a diagnostic biomarker for thyroid cancer is deficient, so we made this first diagnosis meta-analysis of TSHR-mRNA on thyroid cancer.

Characteristics and quality of the included studies
The literature search identified 67 relevant articles. Titles and abstracts were preliminarily reviewed and a www.impactjournals.com/oncotarget/ Oncotarget, 2017, Vol. 8, (No. 4), pp: 6623-6629 Research Paper total of 55 articles were excluded for various reasons (including literature type is review, case reports and letter, or studies do not solely focus on TSHR-mRNA and/or not specifically pertain to thyroid cancer). The article selection process is summarized in Figure 1. Finally, nine publications were included for this metaanalysis [6][7][8][9][10][11][12][13][14].
The nine included studies were published between 2005 and 2015, of which 1027 patients were involved. 4 out of 9 articles were published in China. RT-PCR was applied for TSHR-mRNA detection, with histopathological examination as the gold standard for thyroid cancer diagnosis. The detailed characteristics of these studies are listed in Table 1.
The methodological quality of the studies according to the QUADAS-2 tool is summarized in Figure 2, and the methodological quality was generally better.

Overall analysis
There was heterogeneity presented in this metaanalysis, as revealed by the results (Q = 21.979, p =0.000, I 2 =91). Thus, a random effects model was used for the further analysis. As shown in Figure 3, the sensitivity and specificity for the pooled data were 0.72 (0.68 to 0.76) and 0.82 (0.78 to 0.85), respectively. The AUC was 0.84 ( Figure 4), and the DOR was 12.93 (6.89 to 24.24). The results corresponded to a PLR (Positive Likelihood Ratio) of 3.71 (2.51, 5.50) and an NLR (Negative Likelihood Ratio) of 0.31 (0.21, 0.46). High heterogeneity was demonstrated in the sensitivity (I 2 = 88.9%) and specificity (I 2 = 80.5%).

Threshold effect and publication bias
To retrospect the cause of high heterogeneity, the threshold was firstly analyzed. There was no threshold effect presenting in this meta-analysis, which was supported by the spearman correlation coefficient [15] value of 0.100 (p = 0.862; p > 0.05).
Besides threshold effect, heterogeneity can also be contributed by socio-demographic characteristics and geographical location [16]. Then, geographical location and patient characters were analyzed with a metaregression. No significant heterogeneity was demonstrated in terms of geographical location (coefficient = 0.220, p = 0.283) and patient characters (coefficient = 0.000, p = 0.988). Therefore, other factors might contribute to the observed high heterogeneity.
Furthermore, the potential publication bias was evaluated by Deeks' funnel plots. The p-value of 0.73 suggested there is no publication bias ( Figure 5).

Sensitivity analysis
The sensitivity value range is 0.    The risk of bias in terms of flow and timing and patient selection and was met by less than 30% of the studies. Only three studies unfulfilled all criteria. www.impactjournals.com/oncotarget
Thyroid and thyroid cancer cells specifically express TSHR-mRNA, increasing evidences have demonstrated that circulating TSHR-mRNA is involved in cancer diagnosis and prognosis. The latest article shows that TSHR-mRNA is a molecular marker of thyroid cancer cells circulating [25]. However, there is no systematic evaluation for these studies.
In this work, we conducted the first diagnostic meta-analysis to draw a system evaluation for TSHR-mRNA on the diagnosis of thyroid cancer. An AUC of 0.8427 showed that circulating TSHR-mRNA may be a promising biomarker to discriminate thyroid cancer patients from benign thyroid nodules individuals, with a summary sensitivity of 72% and specificity of 83%. As a potential diagnostic biomarker for thyroid cancer, serum or plasma TSHR-mRNA is characterized by minimal invasion and convenience compared with FNA based histopathological examination. In addition, TSHR-mRNA could serve to monitor the disease recurrence after surgery [26].
However, as a novel biomarker, circulating TSHR-mRNA for thyroid cancer diagnosis have been conducted in limited studies [27]. Therefore, no large number of patient can be included in this meta-analysis. The finding in this work need validation in large-scale clinical studies. Further more, we supposed TSHR-mRNA combined with FNA should have better performance than anyone alone [13,[28][29][30]. Unfortunately, these studies were too few for further analysis.
Overall, this meta-analysis indicates that circulating TSHR-mRNA is potential to be a biomarker for early-stage thyroid cancer diagnosis. However, future clinical studies are necessary to validate this finding.

Search strategy and study selection
The method of literature search is the same as we have just published [31,32]. With the Medical Subject Headings or text words 'TSHR mRNA', combined with 'Thyroid cancer' or 'Thyroid neoplasm', a computerized search were conducted for studies that were published between 1972 and February 2015 in the databases of PUBMED, MEDLINE and EMBASE and Chinese National Knowledge Infrastructure. Moreover, the reference lists of retrieved articles were manually searched for additional relevant studies. No language restrictions were applied.
Studies that is inability to obtain individual data, review articles, case reports, using the same sample, absence of case or control groups, using specimens other than human, using specimens other than those obtained exclusively from the blood (for example, thyroid gland tissue and derivatives) were excluded.

Data extraction and quality assessment
The data were independently collected from the included studies and reached a consensus on all items by two researchers. Any disagreements would be resolved by discussion with a third author for a final consensus. The following information was collected: reference, number of patients, technique employed, histological types of neoplasias. In addition, the following data were extracted: first author's name, publication year, country, cutoff value, patients size, TP (true-positive), FP (false-positive), FN (false-negative) and TN (true-negative).
Methodological quality of the primary studies was assessed by the Quality Assessment of Diagnostic Accuracy Studies 2 (QUADAS-2) checklist [33].

Meta analysis
Statistical analysis was conducted with Meta-DiSc 1.4 and Stata 12.0 [15]. In the analysis, benign thyroid cancer served as negative control. P < 0.05 was set as statistical significance. A value of I 2 > 50% and/or a p < 0.05 indicated the between-study heterogeneity [34]. When heterogeneity was presented in the study, a random effects model was applied to sensitivity, specificity, likelihood ratio, and diagnostic odds ratio. The threshold effect was investigated based on the spearman correlation coefficient. A summary receiver operating characteristic curve (SROC) was used for the overall evaluation of diagnostic performance. Publication bias was assessed using Deeks' funnel plot analysis [35].