RASSF1A promoter methylation is associated with increased risk of thyroid cancer: a meta-analysis.

OBJECTIVE: Previous studies have reported that Ras-associated domain family 1A (RASSF1A), the most commonly silenced tumor suppressor via promoter methylation, played vital roles in the development of carcinogenesis. The purpose of this meta-analysis was to determine whether RASSF1A promoter methylation increased the risk of thyroid cancer.
METHODS: PubMed, Embase, ISI Web of Knowledge, and Chinese National Knowledge Infrastructure databases were searched to obtain eligible studies. The pooled odds ratios (ORs) and 95% confidence intervals (CIs) were calculated to estimate the strength of the associations, using Stata 12.0 software. The methodological quality of included studies was evaluated using Newcastle-Ottawa scale table. Egger's test and Begg's test were applied to detect publication biases. TSA 0.9 software was used to calculate the required information size and whether the result was conclusive.
RESULTS: A total of 10 articles with 12 studies that included 422 thyroid cancer patients, identifying the association of RASSF1A promoter methylation with thyroid cancer risk, were collected in this meta-analysis. Overall, RASSF1A promoter methylation significantly increased the risk of thyroid cancer (total, OR=8.27, CI=4.38-15.62, P<0.05; Caucasian, OR=9.25, CI=3.97-21.56, P<0.05; Asian, OR=7.01, CI=2.68-18.38, P<0.05). In the subgroup analysis based on sample type, a significant association between thyroid cancer group and control group was found (normal tissue, OR=9.55, CI=4.21-21.67, P<0.05; adjacent tissue, OR=6.80, CI=2.49-18.56, P<0.05). The frequency of RASSF1A promoter methylation in follicular thyroid carcinoma was higher than in control group (OR=11.88, CI=5.80-24.32, P<0.05). In addition, the results indicated that the RASSF1A promoter methylation was correlated with papillary thyroid carcinoma in Caucasians and Asians (total, OR=8.07, CI=3.54-18.41, P<0.05; Caucasian, OR=11.35, CI=2.39-53.98, P<0.05; Asian, OR=6.67, CI=2.53-17.64, P<0.05). On the basis of the trial sequential analysis, the significant association of RASSF1A promoter methylation with thyroid cancer risk was found, and there was no need to perform further studies.
CONCLUSION: This meta-analysis confirms that RASSF1A promoter methylation is a risk factor for thyroid tumor.

Introduction

It has been reported that thyroid carcinomas, the most frequently reported endocrine neoplasia, account for only 3%–4% in all human tumors, but the incidence of thyroid cancer is steadily increasing and has the highest increase in incidence within the past two decades.1 Currently, the use of neck ultrasonography is widely used in the diagnosis of thyroid cancer and brings light to the detection of many small early-stage tumors.2 However, the incidence of detection of large tumors and advanced stage tumor patients have also increased in these years.3,4 Thyroid cancer can be categorized into four histotypes: follicular thyroid carcinoma (FTC), papillary thyroid carcinoma (PTC), anaplastic thyroid carcinoma, and medullary thyroid carcinoma (MTC). PTC, the most common type of thyroid cancer, represents 80% of patients, followed by FTC (10%).5,6 PTC derives from the follicular cells of the thyroid, and it has a distinctive papillary architecture. In addition, the nuclear membrane of cells in PTC has several distinctive alternations, such as grooves, pseudoinclusions, and optical clearing. In contrast to the papillary carcinomas, in which the change of nuclear membrane alternations is vital, the diagnosis of FTC depends on whether the thyroid tumor has invaded through the surrounding vessels. MTC has a rare incidence accounting for ~3%, which derives from parafollicular C cells.7 Compared with FTC and PTC – well differentiated cancer – MTC has a more aggressive clinical characteristic.8 MTC often occurs in a form of inherited cancer, which is known as multiple endocrine neoplasia type 2.9 In recent decades, although many genetic studies for thyroid cancer have been performed, no specific biomarkers for the large number of sporadic thyroid cancers have been found.10–12

The Ras-associated domain family 1 (RASSF1) family of proteins, representing one type of Ras effectors, can inhibit the development of cancer. RASSF1A, one of the seven iso-forms of the RASSF1 family and located on 3p21, is the most common epigenetically inactivated tumor suppressor genes via promoter methylation in human cancers.13,14 It can bind Ras protein in a guanosine triphosphate-dependent manner to mediate the cell apoptotic, which has a similar function with Nore1 in mouse.14 In many solid tumors, RASSF1A, as a component of crucial cell signal pathways, plays an important role in the Ras/PI3K/AKT, Ras/RAF/MEK/ERK, and Hippo pathways, and the inactivation of RASSF1A can result in the alternations of clinical characteristic in tumors.15,16 It has been reported that the CpG methylation of RASSF1A promoter can lead to the loss-of-expression of RASSF1A.14 Previous studies have found that the RASSF1A promoter methylation can increase the risk of lung cancer, breast cancer, prostate cancer, ovarian cancer, renal cell carcinoma, colorectal cancer, hepatocellular carcinoma, and gastric cancer.17–24 Furthermore, these results indicated that the RASSF1A promoter methylation may be a significant prognostic factor for many human cancers. Despite a number of studies performed on thyroid cancer patients, the relationship of RASSF1A promoter methylation with thyroid cancer risk and pathogenesis remains controversial. To clarify these controversial results, a meta-analysis was performed.

Methods

Publication search

Relevant literature was retrieved from PubMed, Embase, ISI Web of Knowledge, and Chinese National Knowledge Infrastructure databases up until September 2016, without language restrictions. The following keywords were applied: “RASSF1 protein, human (Supplementary Concept),” “Thyroid Neoplasms (Mesh),” “Methylation (Mesh),” “RASSF1A,” “thyroid cancer,” and “hypermethylation” to search eligible studies. Review articles and the references from relevant primary studies were manually searched for identifying additional potential studies.

Inclusion and exclusion criteria

The inclusion criteria were as follows: 1) studies primarily evaluating the frequency of RASSF1A promoter methylation in thyroid cancer with control group, 2) studies that have a case–control design, 3) studies that had sufficient data of frequency, and 4) patients who had been accurately diagnosed according to the diagnostic criteria. Studies were removed if they met one of the items: 1) reviews, meta-analysis, or animal studies; 2) duplicate data; and 3) study performed in the cell lines.

Data extraction

The following data were extracted from the eligible studies: year of publication, country, ethnicity, frequency of RASSF1A promoter methylation, cancer type, method of methylation detection, and sample type. The information was reviewed by two investigators. If the two investigators had divergence, the disagreement was resolved by consulting with the other authors.

Quality assessment

To ensure the high quality of included literatures, the Newcastle–Ottawa scale (NOS) table was applied to assess the methodological quality of observational studies. The NOS is commonly used in the quality assessment of nonrandomized studies, such as case–control and cohort studies. The scores of quality assessment range from 0 to 9. If the study got a score <5 it was excluded.

Trial sequential analysis

In a cumulative meta-analysis, false-positive and false-negative results may arise, which result from the repeated testing and the adding of other studies. Therefore, it was crucial to lower the risk of type I error and type II error in the repeated statistics. In addition, interim analysis was performed to estimate the required sample size according to the data in the common trials and to avoid wasting of the sample. In this present meta-analysis, the trial sequential analysis was performed to estimate the required information size (RIS) and observe whether the result was conclusive through controlling the risk of type I error and type II error. TSA 0.9 software (www.ctu.dk/tsa) was applied to conduct these analyses by combining the RIS with trial sequential monitoring boundaries. The trial sequential analysis was conducted at the level of 5% risk of type I error and 80% power of statistical test (5% risk of type II error). Moreover, adjusted information size was calculated with a relative risk reduction (−50%) according to the incidence of control and case groups.

Statistical analysis

The frequency of RASSF1A promoter methylation in case and control groups was collected to calculate the pooled odds ratios (ORs) and 95% confidence interval (CI),25 evaluating the strength of association of RASSF1A promoter methylation with thyroid cancer risk or thyroid cancer clinical characteristics. Heterogeneity among studies was measured using Cochran’s Q-test and Higgin’s I2.26,27 If heterogeneity existed in studies (P<0.05, I2>50%), random-effects model was applied, otherwise, fixed-effects model was used (P>0.05, I2<50%).28,29 In addition, subgroup analysis based on ethnicity and sample type were conducted to detect the source of heterogeneity and lower the between-study heterogeneity. Begg’s test and Egger’ test were used to check for the publication bias. When the P-value was <0.05, publication bias was considered as significant. The aggregated sensitivity and specificity were observed by conducting the sensitivity analysis. All meta-analyses were performed using Stata software 12.0 (Stata Corporation, College Station, TX, USA).

Results

Characteristics of studies

Through searching PubMed, Embase, ISI Web of Knowledge, and Chinese National Knowledge Infrastructure databases, 25 publications were initially retrieved. Eight duplicate articles were removed. After reading titles and abstracts, four articles were excluded due to uncorrelated contents. Furthermore, full-text was read to eliminate irrelevant literatures, and two articles were eliminated because they had no sufficient data of RASSF1A promoter methylation. Finally, 10 publications with 422 patients and 219 controls were identified in this meta-analysis. Of these studies, seven studies were performed in Caucasians,30–36 while three studies were conducted in Asians.37–39 The detection of RASSF1A promoter methylation applied methylation-specific polymerase chain reaction and quantitative methylation-specific polymerase chain reaction. With regard to the type of control sample, normal tissue was used in five studies,31,33,35,36,39 four studies applied benign tissue,31,32,35,38 and adjacent tissue was applied in two studies34,37 (Figure 1; Tables 1 and S1–S6).

Figure 1

Flow chart of study selection procedure.

Table 1

Characteristics of eligible studies in the meta-analysis

References (year)	Country	Ethnicity	Histology	Control		Tumor		Methods	Control	QA score
				U (n)	M (n)	U (n)	M (n)
Santoro et al30 (2013)	Italy	Caucasian	TC	NR	NR	10	9	MSP	NR	NR
Brait et al31 (2012)	USA	Caucasian	TC	1	13	5	38	QMSP	NT	6
Brait et al31 (2012)	USA	Caucasian	TC	1	43	5	38	QMSP	BT	6
Qu and Xue37 (2012)	People’s Republic of China	Asian	TC	21	7	13	15	MSP	AT	8
Dai et al38 (2011)	People’s Republic of China	Asian	TC	22	10	20	30	MSP	BT	8
Mohammadi et al32 (2011)	Iran	Caucasian	TC	2	23	6	19	MSP	BT	7
Wang et al39 (2009)	People’s Republic of China	Asian	TC	10	0	25	63	MSP	NT	6
Lee et al33 (2008)	Sweden	Caucasian	FTC	16	5	3	18	MSP	NT	7
Nakamura et al34 (2005)	USA	Caucasian	TC	27	0	51	27	MSP	AT	7
Xing et al35 (2004)	USA	Caucasian	TC	14	0	32	19	QMSP	NT	7
Xing et al35 (2004)	USA	Caucasian	TC	14	0	5	4	QMSP	BT	7
Schagdarsurengin et al36 (2002)	Germany	Caucasian	TC	3	1	11	27	MSP	NT	6

Abbreviations: AT, adjacent tissue; BT, benign tissue; FTC, follicular thyroid carcinoma; M, methylation sample; MSP, methylation-specific PCR; NR, not reported; NT, normal tissue; PCR, polymerase chain reaction; QA, quality assessment; QMSP, quantitative methylation-specific PCR; TC, thyroid carcinoma; U, unmethylation sample.

Quantitative data synthesis

According to this meta-analysis, there was a significant association of RASSF1A promoter methylation with thyroid cancer risk (OR=5.94, 95% CI=2.09–16.89, P<0.05). When stratified by the ethnicity, a significant increased risk of thyroid cancer was detected in Asians, but not in Caucasians. In the stratified analysis by sample type, significantly increased risks were found in both normal tissues and adjacent tissues in detection of RASSF1A promoter methylation in thyroid cancer (normal tissue: OR=9.55, CI=4.21–21.67, P<0.05; adjacent tissue: OR=6.80, CI=2.49–18.56, P<0.05). Furthermore, significant association was also found in both FTC and PTC (for FTC: OR=11.88, CI=5.80–24.32, P<0.05; for PTC: OR=8.07, CI=3.54–18.41, P<0.05). In evaluating the association for pathological stage of thyroid cancer, no significant correlations of RASSF1A promoter methylation with distant metastasis and TNM-stage were observed (Figures 2–4; Table 2).

Figure 2

Forest plot for the association between Ras-associated domain family 1A promoter methylation and risk of thyroid cancer.

Abbreviations: CI, confidence interval; OR, odds ratio.

Figure 3

Forest plot for the association between Ras-associated domain family 1A promoter methylation and risk of follicular thyroid carcinoma.

Abbreviations: CI, confidence interval; OR, odds ratio.

Figure 4

Forest plot for the association between Ras-associated domain family 1A promoter methylation and risk of papillary thyroid carcinoma.

Abbreviations: CI, confidence interval; OR, odds ratio.

Table 2

Meta-analysis of association between Ras-associated domain family 1A promoter methylation and thyroid cancer risk

Variables	Controls/cases (n)	OR	95% CI	Heterogeneity
				I2 (%)	P-value
Total	118/347	8.27	4.38–15.62	47.20	0.078
Ethnicity
Caucasian	80/231	9.25	3.97–21.56	46.90	0.110
Asian	38/116	7.01	2.68–18.38	70.30	0.066
Disease type
FTC	104/94	11.88	5.80–24.32	51.50	0.067
PTC	83/166	8.07	3.54–18.41	8.4	0.359
Sample type
Normal tissue	63/241	9.55	4.21–21.67	51.80	0.081
Adjacent tissue	55/106	6.80	2.49–18.56	58.10	0.122
Pathogenesis
Distant metastasis	183/67	1.18	0.64–2.16	34.60	0.191
TNM-stage	64/32	1.79	0.72–4.44	52.60	0.121

Abbreviations: CI, confidence interval; FTC, follicular thyroid carcinoma; OR, odds ratio; PTC, papillary thyroid carcinoma.

Literature qualities

In this meta-analysis, 10 studies were scored by NOS table by two independent authors. The scores of eligible literatures ranged from 6 to 8, suggesting that the quality of included studies were of a high quality and the inclusion criteria were satisfied. The detailed information is shown in Table 1.

Publication bias and sensitivity analysis

The results of Begg’s test and Egger’s test indicated that no significant publication bias existed in the analysis of association between RASSF1A promoter methylation and thyroid cancer risk or pathological characteristics. (For risk of thyroid: Begg’s test P=0.453, Egger’s test P=0.377; for distant metastasis: Begg’s test P=0.624, Egger’s test P=0.669; for TNM-stage: Begg’s test P=0.117, Egger’s test P=0.441). Sensitivity analysis indicated that the results were stable in this meta-analysis, and the pooled ORs did not have a significant change by omitting each study (Figures 5–7).

Figure 5

Funnel plot of publication biases on the association between Ras-associated domain family 1A promoter methylation and thyroid cancer risk.

Figure 6

Funnel plot of publication biases on the association between Ras-associated domain family 1A promoter methylation and follicular thyroid carcinoma risk.

Figure 7

Funnel plot of publication biases on the association between Ras-associated domain family 1A promoter methylation and papillary thyroid carcinoma risk.

Trial sequential analysis indicated that Z-curve crossed the traditional boundary and the monitoring boundary for benefit, which indicated that no further studies were performed to study this association of RASSF1A promoter methylation with thyroid cancer risk. The Z-curve did not reach the boundary of RIS, and the RIS was 1,115. In other associations, no trial sequential analysis was conducted for the small number of samples (Figure 8).

Figure 8

Trial sequential analysis comparing the Ras-associated domain family 1A promoter methylation frequency of control group and thyroid cancer group.

Abbreviation: RIS, required information size.

Discussion

Epigenetic alternations play an important role in the regulation of gene expression, especially in mammalian cells. DNA methylation usually occurs in CpG islands – clusters of CpGs of gene promoter region, which can lead to gene activation or inactivation. The hypermethylation of gene promoter represses gene expression by inhibiting gene transcription and further affects the total cell signal pathway. The inactivation of RASSF1A, one of tumor suppressor genes, is critical in the tumorigenesis of thyroid cancer. The changes in gene structure, deletion, or mutation can result in the irreversible loss of RASSF1A function. Alternatively, the cumulative RASSF1A methylation can also influence the expression and affect the progression of cancers, such as tumor differentiation and metastasis. In previous studies of thyroid cancers, many genes methylation have a significant association with the development of thyroid tumor. For example, p16INK4A/CDKN2A, p21CIP1/CDKN1B, and p27KIP1/CDKN1B, considering as tumor suppressors, can regulate the activity of cyclin–CDK complexes in mammalian cells. However, the expression of these genes in which promoter methylation is detected in 30% of thyroid tumor was commonly downregulated in thyroid cancer patients.40,41 In addition, the hypermethylation of thyrotropin receptor gene, receptor, fibroblast growth factor type 2, and receptor, fibroblast growth factor type 1 existed in thyroid cancers.42–44 RASSF1A protein, lacking enzymatic activity, contains a Ras-association domain and can regulate the cell cycle and apoptosis.13,14 Many studies have found that the RASSF1A promoter inactivation was frequent in thyroid cancers, >30% of thyroid tumors,33,34,37,39 but other reports also observed that the RASSF1A promoter methylation did not have significant correlations with thyroid cancer risk.31,35,36 Thus, it is necessary to conduct a meta-analysis to determine the strength of association between RASSF1A promoter methylation and thyroid cancer. In the literature retrieval, a previous meta-analysis was found.45 However, the meta-analysis only explored the association between RASSF1A promoter methylation and PTC risk, and no clinical information was included. The result of the previous study indicated a significant association of RASSF1A promoter methylation with PTC risk. This was the same result in this meta-analysis. To discuss the association of RASSF1A promoter methylation with FTC risk, distant metastasis, and TNM-stage, the meta-analysis was conducted.

In this study, the results indicated that aberrant methylation of RASSF1A promoter was more frequently detected in thyroid cancer than in noncancerous controls. In the included studies, although the results of Wang et al39 and Qu et al37 demonstrated that the frequency of RASSF1A promoter methylation increased the risk of thyroid cancer, the pooled ORs and 95% CI show that there was a significant association between RASSF1A promoter methylation and the risk of thyroid cancer in Asians. Furthermore, a small degree of heterogeneity among studies was noticed in the overall analysis. Therefore, the stratified analysis, on the basis of sample type and ethnicity, was carried out to reduce the heterogeneity. At the same time, the pooled ORs indicated that the significant association was found in the stratified analysis based on ethnicity and sample type. Furthermore, between-study heterogeneity, to a large extent, disappeared in the subgroup analysis. This phenomenon indicated that the sample type and ethnicity were the primary cause of heterogeneity. Interestingly, no significant association between the frequency of RASSF1A promoter methylation in thyroid malignant tumor and the frequency in thyroid benign tumor existed. In other subgroups, the higher frequency of FTC and PTC group was found, compared with the nontumorous group. From all eligible studies in this meta-analysis, no results indicated that the RASSF1A promoter methylation had a significant influence in the metastasis of thyroid tumor.30,36–39 In the meantime, the result of this statistical analysis was similar to the previous studies (for M0-stage and M1-stage: OR=1.18, 95% CI=0.64–2.16, P<0.05). Moreover, no significant association between the frequency of RASSF1A methylation and TNM-stage of thyroid cancer (for I-stage and II–IV-stage: OR=1.79, 95% CI=0.72–4.44, P<0.05) was found. In the included studies, one study found a significant association,38 but two studies obtained negative results.30,38 These results were inconsistent, but now, to a large extent, we could say that the frequency of RASSF1A promoter methylation might not be related with the TNM-stage of thyroid cancer. Certainly, the small sample size may limit the power of statistics. From the trial sequential analysis, although no studies were continuously performed to evaluate the association of RASSF1A promoter methylation with thyroid cancer risk, further studies, investigating the association between RASSF1A promoter methylation and thyroid cancer pathogenesis, were still needed to be carried out.

In this meta-analysis, the publication bias was not detected in the eligible studies. Begg’s test, Egger’s test, and funnel plot were performed, which indicated that the data did not have a considerable discrepancy among studies. Furthermore, consistent results were found in sensitivity analysis. However, several potential limitations must be emphasized in this study. First, with few included studies in this meta-analysis, the influence of small sample size of case and control cannot be ruled out. Second, the value of cut-off point of methylation cannot be achieved, thus the sensitivity and specificity of methylation about thyroid cancer risks cannot be determined. Third, the clinical information was so little that the analysis of association of RASSF1A promoter methylation with the pathogenesis of thyroid cancer patients cannot be performed.

Conclusion

This meta-analysis showed that RASSF1A promoter methylation might play an important role in thyroid cancer initiation and progression. The RASSF1A promoter methylation might be a promising biomarker for the early diagnosis of thyroid cancer. In addition, RASSF1A promoter methylation was associated with the increased risk of distant metastasis and late TNM-stage of thyroid cancer patients. However, in consideration of the limitations acknowledged above, more large-scale, multicenter, and well-designed case–control or cohort researches will provide more insights into the role of RASSF1A promoter methylation in the risk and pathogenesis of thyroid cancer.

Table S1

Thyroid cancer risk

References	Country	Ethnicity	Histology	U (n)	M (n)	U (n)	M (n)	Methods	Control type	NOS score
Santoro et al1	Italy	Caucasian	Thyroid carcinoma	NR	NR	10	9	MSP	NR	NR
Brait et al2	USA	Caucasian	Thyroid carcinoma	1	13	5	38	QMSP	Normal tissue	6
Brait et al2	USA	Caucasian	Thyroid carcinoma	1	43	5	38	QMSP	Benign tissue	6
Qu and Xue3	People’s Republic of China	Asian	Thyroid carcinoma	21	7	13	15	MSP	Adjacent tissue	8
Dai et al4	People’s Republic of China	Asian	Thyroid carcinoma	22	10	20	30	MSP	Benign tissue	8
Mohammadi5	Iran	Caucasian	Thyroid carcinoma	2	23	6	19	MSP	Benign tissue	7
Wang et al6	China	Asian	Thyroid carcinoma	10	0	25	63	MSP	Normal tissue	6
Lee et al7	Sweden	Caucasian	Follicular thyroid carcinoma	16	5	3	18	MSP	Normal tissue	7
Nakamura et al8	USA	Caucasian	Thyroid carcinoma	27	0	51	27	MSP	Adjacent tissue	7
Xing et al9	USA	Caucasian	Thyroid carcinoma	14	0	32	19	QMSP	Normal tissue	7
Xing et al9	USA	Caucasian	Thyroid carcinoma	14	0	5	4	QMSP	Benign tissue	7
Schagdarsurengin et al10	Germany	Caucasian	Thyroid carcinoma	3	1	11	27	MSP	Normal tissue	6
Normal and adjacent tissue
Brait et al2	USA	Caucasian	Thyroid carcinoma	1	13	5	38	QMSP	Normal tissue	6
Qu and Xue3	People’s Republic of China	Asian	Thyroid carcinoma	21	7	13	15	MSP	Adjacent tissue	8
Wang et al6	People’s Republic of China	Asian	Thyroid carcinoma	10	0	25	63	MSP	Normal tissue	6
Lee et al7	Sweden	Caucasian	Follicular thyroid carcinoma	16	5	3	18	MSP	Normal tissue	7
Nakamura et al8	USA	Caucasian	Thyroid carcinoma	27	0	51	27	MSP	Adjacent tissue	7
Xing et al9	USA	Caucasian	Thyroid carcinoma	14	0	32	19	QMSP	Normal tissue	7
Schagdarsurengin et al10	Germany	Caucasian	Thyroid carcinoma	3	1	11	27	MSP	Normal tissue	6
Benign tissue
Brait et al2	USA	Caucasian	Thyroid carcinoma	1	43	5	38	QMSP	Benign tissue	6
Dai et al4	People’s Republic of China	Asian	Thyroid carcinoma	22	10	20	30	MSP	Benign tissue	8
Xing et al9	USA	Caucasian	Thyroid carcinoma	14	0	5	4	QMSP	Benign tissue	7
Mohammadi et al5	Iran	Caucasian	Thyroid carcinoma	2	23	6	19	MSP	Benign tissue	7

Abbreviations: M, methylation sample; MSP, methylation-specific PCR; NOS, Newcastle–Ottawa scale; NR, not reported; PCR, polymerase chain reaction; QMSP, quantitative methylation-specific PCR; U, unmethylation sample.

Table S2

Meta-analysis table

Variables	Controls/cases (n)	OR	95% CI	I2 (%)	P-value
Heterogeneity
Total	136/365	5.94	2.09–16.89	55.40	0.028
Ethnicity
Caucasian	98/249	5.4	1.40–20.86	59.60	0.03
Asian	38/116	10.02	0.63–159.62	70.30	0.066
Disease type
FTC	104/94	11.88	5.80–24.32	51.50	0.067
PTC	152/133	2.18	0.48–9.83	68.40	0.007
Sample type
Normal tissue	81/259	5.85	1.43–23.91	62.60	0.02
Adjacent tissue	55/106	7.29	0.77–69.05	58.10	0.122
Pathogenesis
Distant metastasis	183/67	1.18	0.64–2.16	34.60	0.191
TNM-stage	64/32	1.79	0.72–4.44	52.60	0.121

Abbreviations: CI, confidence interval; FTC, follicular thyroid carcinoma; OR, odds ratio; PTC, papillary thyroid cancer.

Table S3

TNM-stage

References	Country	Ethnicity	Histology	Thyroid tumor				Methods
				TNM-stage I		TNM-stage II–IV
				U (n)	M (n)	U (n)	M (n)
Santoro et al1	Italy	Caucasian	Thyroid carcinoma	3	3	6	6	MSP
Qu and Xue2	People’s Republic of China	Asian	Thyroid carcinoma	13	7	1	7	MSP
Dai et al4	People’s Republic of China	Asian	Thyroid carcinoma	15	23	5	7	MSP

Abbreviations: M, methylation sample; MSP, methylation-specific PCR; PCR, polymerase chain reaction; U, unmethylation sample.

Table S4

Distant metastasis

References	Country	Ethnicity	Histology	Thyroid tumor				Methods
				M0 stage		M1 stage
				U (n)	M (n)	U (n)	M (n)
Santoro et al1	Italy	Caucasian	Thyroid carcinoma	6	8	3	1	MSP
Qu and Xue3	People’s Republic of China	Asian	Thyroid carcinoma	11	8	3	6	MSP
Dai et al4	People’s Republic of China	Asian	Thyroid carcinoma	16	27	4	3	MSP
Wang et al6	People’s Republic of China	Asian	Thyroid carcinoma	30	43	3	13	MSP
Schagdarsurengin et al10	Germany	Caucasian	Thyroid carcinoma	11	23	11	20	MSP

Abbreviations: M, methylation sample; MSP, methylation-specific PCR; PCR, polymerase chain reaction; U, unmethylation sample.

Table S5

Papillary thyroid carcinoma

References	Country	Ethnicity	Histology	Control		PTC			Methods
				U (n)	M (n)	U (n)	M (n)
Qu and Xue3	People’s Republic of China	Asian	Thyroid carcinoma	21	7	13		15	MSP
Wang et al6	People’s Republic of China	Asian	Thyroid carcinoma	10	0	19		42	MSP
Nakamura et al8	USA	Caucasian	Thyroid carcinoma	27	0	23		11	MSP
Xing et al9	USA	Caucasian	Thyroid carcinoma	14	0	24		6	QMSP
Schagdarsurengin et al10	Germany	Caucasian	Thyroid carcinoma	3	1	5		8	MSP

Abbreviations: M, methylation sample; MSP, methylation-specific PCR; PCR, polymerase chain reaction; PTC, papillary thyroid cancer; QMSP, quantitative methylation-specific PCR; U, unmethylation sample.

Table S6

Follicular thyroid carcinoma

References	Country	Ethnicity	Histology	Control		FTC		Methods
				U (n)	M (n)	U (n)	M (n)
Qu and Xue3	People’s Republic of China	Asian	Thyroid carcinoma	21	7	13	15	MSP
Lee et al7	Sweden	Caucasian	Follicular thyroid carcinoma	16	5	3	18	MSP
Schagdarsurengin et al10	Germany	Caucasian	Thyroid carcinoma	3	1	3	7	MSP
Xing et al9	USA	Caucasian	Thyroid carcinoma	3	9	14	0	QMSP
Wang et al6	People’s Republic of China	Asian	Thyroid carcinoma	1	14	10	0	MSP
Nakamura et al8	USA	Caucasian	Thyroid carcinoma	4	4	27	0	MSP

Abbreviations: FTC, follicular thyroid carcinoma; M, methylation sample; MSP, methylation-specific PCR; PCR, polymerase chain reaction; QMSP, quantitative methylation-specific PCR; U, unmethylation sample.