The present study included 122 patients with conventional PTC who received neck level VI LN dissection. All PTCs were surgically resected at Kangbuk Samsung Hospital, Sungkyunkwan University School of Medicine (Seoul, Republic of Korea) from January 1 to December 31, 2010. Among 122 patients with PTC, central LN metastasis (level VI) was found in 64. To elucidate the clinicopathological significance of the mLNR, clinicopathological parameters, such as age, sex, tumor size, BRAF^V600E mutation, tumor multifocality, and extrathyroidal extension, a review of medical charts and pathological records was performed. All dissected LNs were embedded for microscopic examination, and LN characteristics were evaluated. pTNM staging was performed according to the eighth AJCC TNM classification system [7]. All procedures performed in the current study were approved by the Institutional Review Board (IRB; approval No. KBSMC 2016-04-070). Formal written informed consent was waived by the IRB.

All embedded LNs were evaluated by reviewing glass slides. Evaluated characteristics were number of dissected LNs and mLNs, largest LN size, largest mLN size, and mean LN size, mLNR, and tumor deposit size of mLN. mLNR was defined as number of mLNs to total number of dissected LNs. We categorized the cases into high (≥ 0.44) and low mLNR (< 0.44) and into macrometastasis (tumor deposit size ≥ 0.2 cm) and micrometastasis (< 0.2 cm). In the present study, based on median cut-off value (mLNR, 0.44), eligible studies were divided into high and low mLNR subgroups.

Two array blocks containing 64 PTCs with neck level VI LN metastasis were prepared for immunohistochemistry. A tissue core was obtained from each primary tumor. Core tissue biopsies (2 mm diameter) were collected from individual paraffin-embedded donor PTC blocks and delivered in recipient tissue array blocks using a trephine apparatus. Each block contained nonneoplastic thyroid tissue as an internal control.

Sections 4 μm in size were cut, deparaffinized, and hydrated using routine xylene-alcohol procedures before incubation with 0.01 M citrate buffer (pH 6.0) for 5 minutes in a microwave oven for antigen retrieval and treatment with 3% H₂O₂ to quench endogenous peroxidase. Sections were treated with normal serum from the animal used to generate secondary antibodies to block nonspecific binding and then incubated with anti-nuclear factor κB (NF-κB) RelA (1:100, Santa Cruz Biotechnology, Santa Cruz, CA, USA), anti-cyclin D1 (1:100, Thermo Fisher Scientific, Fremont, CA, USA), or anti-Ki-67 (1:50, Santa Cruz Biotechnology). A compact polymer method from Bond Intense detection kits (Leica Biosystems, Newcastle upon Tyne, UK) was used for immunohistochemical staining and 3,3'-diaminobenzidine (Vector Laboratories, Burlingame, CA, USA) for visualization. To confirm reaction specificity, a negative control without primary antibody was included. For counterstaining, immunostained sections were lightly stained with Mayer’s hematoxylin.

Nuclear expression of NF-κB RelA, regardless of cytoplasmic expression, was considered as NF-κB activation. In the present study, NF-κB positivity was defined as NF-κB stain in ≥ 5% of tumor cell nuclei [22-24]. Regarding interpretation for cyclin D1, patients showing nuclear staining in κ 10% of the tumor cells were considered positive [25]. For Ki-67 staining, we evaluated at least 300 cells and counted cells with nuclear staining for each specimen. Proliferation index (%) was 100 × Ki-67-positive cells/300 cells.

Relevant articles were obtained by searching PubMed and MEDLINE databases up to May 15, 2018. Searches were performed using the keywords ‘papillary thyroid carcinoma’ and ‘metastatic lymph node ratio or lymph node ratio.’ The title and abstract of all searched articles were screened for exclusion. Review articles were also screened to find additional eligible studies. Search results were then scanned based on the following inclusion and exclusion criteria: (1) human thyroid tissue was investigated, (2) available information for the correlation between mLNR and survival in PTC with neck level VI LN metastasis, (3) case reports or non-original articles were excluded, and (4) non-English language publications were excluded. Data from all eligible studies were extracted by two authors. The following data were extracted from each of the eligible studies [8,10-12,14,16,26]: first author’s name, year of publication, number of patients analyzed, and correlation between mLNR and survival based on previous studies [27,28].

SPSS ver. 22.0 software (IBM Corp., Armonk, NY, USA) was used for statistical analyses. The correlations between mLNR and clinicopathological parameters, including the immunohistochemical results of primary tumor, were determined using either chi-square test or the Fisher exact test (two-sided). The relationships between mLNR and tumor size, LN size, mLN size, and tumor deposit diameter of LN were analyzed using a two-tailed Student’s t-test. Linear regression analysis was conducted to investigate correlations between primary tumor size and tumor multifocality, number of tumors, and supplemental tumor size.

In the meta-analysis, data were analyzed using the Comprehensive Meta-Analysis software package (Biostat, Englewood, NJ, USA). The correlation between high mLNR and survival rate in patients with PTC was investigated using meta-analysis. We performed subgroup analysis based on follow-up period and cut-off value for high mLNR. Heterogeneity between studies was analyzed using the Q and I² statistics and presented as p-values. In addition, sensitivity analysis was conducted to assess the heterogeneity of eligible studies and the impact of each study on the combined effect. Because various cut-off values for high mLNR were used for various populations, the application of random-effect model rather than fixed-effect model was more suitable. The assessment of publication bias was performed using Begg’s funnel plot and Egger’s test. The results were considered statistically significant at p < .05.

Among 122 patients with PTCs, neck level VI LN metastasis was found in 64. We first investigated the clinicopathological significance of high mLNR (≥ 0.44) in 64 PTC patients with neck level VI LN metastasis. No significant correlation was observed between high mLNR and clinicopathological parameters, such as age, sex, tumor size, BRAF^V600E mutation, tumor multifocality, and pTNM stage (Table 1). Next, the correlations between mLNR and pathological statuses of examined and mLNs were evaluated. The tumor deposit diameters of mLNs were 0.43 ± 0.31 cm and 0.29 ± 0.41 cm in high and low mLNR, respectively (p = .114) (Table 2). The rate of macrometastasis (> 0.2 cm) in high mLNR was significantly higher than in low mLNR (82.9% vs 44.8%, p = .006). In addition, extracapsular spread of LN was more frequently found in high mLNR than in low mLNR (37.1% vs 6.9%, p = .007). However, high mLNR was not significantly correlated with mean size or largest size of examined and mLNs. The mLNR was significantly increased by the largest LN size, but not the largest mLN size (p = .001 vs p = .364) (Fig. 1). In the high mLNR group, the largest examined LN and the largest mLN were equal to 94.1%; however, in the low mLNR group, the rate was 56.7%.

We previously reported that nuclear NF-κB RelA of the primary tumor significantly correlated with LN metastasis [29]. In the present study, the correlations between mLNR and the immunohistochemical expressions of primary tumor, such as NF-κB RelA, cyclin D1, and Ki-67 labeling index, were investigated. High mLNR was not significantly correlated with nuclear NF-κB RelA (p = .271) or cyclin D1 expression (p > .999). In addition, no significant correlation between high mLNR and ki-67 labeling index was observed (p = .917).

The database search identified 288 reports. Of these, 254 reports were excluded due to lack of sufficient information. Other studies were excluded because they reported the results of other diseases (n = 4), were non-English (n = 6), or were non-original articles (n = 16) (Fig. 2). After applying the inclusion and exclusion criteria, seven reports were finally included for meta-analysis [8,10-12,14,16,26]. Eligible studies used various cut-off values for high mLNR ranging from 0.22 to 0.65. To elucidate the prognostic role of mLNR, the correlation between high mLNR and survival rate was investigated. High mLNR significantly correlated with worse disease-free survival at the 5-year and 10-year follow-up (hazard ratio [HR], 5.408; 95% confidence interval [CI], 3.189 to 9.170 and HR, 6.020; 95% CI, 3.483 to 10.404, respectively) (Fig. 3). To evaluate the optimal cut-off value for high mLNR, subgroup analysis based on a cut-off value of 0.44, was performed. High (≥ 0.44) and low mLNR cut-off (< 0.44) subgroups showed significant correlation between high mLNR and worse disease-free survival at the 5-year follow-up (HR, 5.088; 95% CI, 1.926 to 13.443 and HR, 5.570; 95% CI, 2.833 to 10.952; respectively) and 10-year follow-up (HR, 5.597; 95% CI, 1.806 to 17.350 and HR, 6.158; 95% CI, 2.863 to 13.245, respectively).