Results
Among the 392 patients enrolled in this study, 308 were female and 84 were male, aged 18–83 years with a median age of 47.62 years. The mean diameter of nodules was 7.05 mm (range, 1–10 mm; SD, 1.97). A total of 159 patients (40.6%) were demonstrated to exhibit CLNM by the final pathological examination, while the other 233 patients (59.4%) did not. Table II presents the comparison of clinicopathological and US differences between CLNM-negative (Group I) and CLNM-positive (Group II) groups. Male gender, age <45 years, maximum tumor diameter >5 mm, lower lobe location and multifocal carcinoma with total tumor diameter (TTD) >10 mm were significantly correlated with the incidence of CLNM (P<0.001). ECS was significantly associated with CLNM (P=0.022). Tumor bilaterality, chronic lymphocytic thyroiditis and US characteristics did not predict CLNM in PTMC. A multivariate logistic regression analysis was performed on the factors that were revealed to be statistically significant by the univariate analysis. Male gender (OR, 5.021; P<0.001), tumor size >5 mm (OR, 4.842; P<0.001), age <45 years (OR, 3.911; P<0.001), lower lobe location (OR, 4.652; P<0.001), ECS (OR, 3.885; P=0.002) and TTD >10 mm (OR, 8.553; P=0.016) were identified as statistically significant independent predictive factors for CLNM (Table III). According to the multiple logistic regression analysis, the patients were scored by giving an index point to each characteristic that was significantly associated with CLNM. The index score for each characteristic is presented in Table IV. The mean clinicopathological score was 1.23±0.722 in the CLNM-positive group and 2.34±0.973 in the CLNM-negative group, which was a statistically significant difference (P<0.001). ROC curves were then used to identify that an index point ≥2 was the optimal point to distinguish between patients with and without CLNM (Fig. 1). The sensitivity and specificity reached 86.2% and 70.4%, respectively, with an area under the ROC curve of 0.816.
Discussion
Owing to the widespread use of thyroid US and US-guided FNAC, the incidence of PTMC has increased in Eastern China in previous years (2,11). Although PTMC is generally associated with an excellent prognosis, CLNM is common in PTMC patients with an incidence of 41–64% (12–17). There is no controversy in performing therapeutic neck dissection on cervical lymph node-positive PTMC patients to reduce the chance of persistence and recurrence (18), however debate remains as to whether CND should be performed on cN0 PTMC patients. A number of authors have proposed that there should be routine CND in patients with PTMC since there is evidence demonstrating that residual metastases in cervical lymph nodes represent the most common site of recurrent disease, and repeated operations may increase the incidence of operative complications, including hypoparathyroidism and damage to the laryngeal nerve (19). However, a previous study indicated that CND is not required, as the presence of CLNM is not associated with the disease-free survival rate of patients (20). The present study focused on identifying the preoperative and intraoperative predictive factors for CLNM in cN0 PTMC patients, which may aid in identifying approaches to appropriate surgical treatment for patients in Eastern China. Although the use of thyroid US for detecting CLNM is limited (20), the present study also investigated characteristics of US to identify any potential predictors.
The incidence of CLNM in the present study was 40.6% (n=159), which is consistent with previous studies (21,22). Statistical analyses performed between the CLNM-positive and -negative groups in the present study included eight clinicopathological parameters. The results demonstrated that male gender, age <45 years, maximum tumor diameter >5 mm, lower lobe location and multifocal carcinoma with TTD >10 mm were independent predictors for CLNM from the multivariate analysis. Zeng et al (23) indicated that certain US characteristics (upper pole location, no well-defined margin and presence of calcifications) were predictive factors for lateral LNM in patients with PTMC in Eastern China. Few studies have reported the associations between US characteristics and CLNM. In the present study, no significance was identified for these associations, which is consistent with a previous study by Kim et al (24).
A number of previous studies have reported that age is not associated with CLNM in patients with PTMC (24,25). However, Wang et al (26) demonstrated that there is a significant difference between patients <45 years old and ≥45 years old in cN0 PTMC. The present study identified that age <45 years was significantly associated with CLNM patients. Male gender has been considered to be an important prognostic characteristic of CLNM by a number of previous studies (21,27). The present study demonstrated a significant association between male gender and CLNM, which is consistent with these previous reports. Tumor size has been previously confirmed as a prognostic feature of CLNM (20). Lee et al (28) demonstrated that PTMC with tumor size >7 mm had a stronger association with poor prognosis compared with those <7 mm. Lim et al (25) had reported that tumor size >5 mm was frequently associated with CLNM. The results of the present study demonstrated no difference between the ≤7 and >7 mm tumor size groups. However, the univariate and multivariate analyses confirmed tumor size >5 mm as an independent predictor of LNM. A number of previous studies reported the upper pole location as an independent predictive factor for lateral LNM (23,29). To the best of our knowledge, few reports have defined the statistical correlation between tumor location and CLNM in PTMC. Wang et al (26) proposed that the CLNM rate in the lower/middle pole group was significantly increased compared with that of the upper pole group. A study by Choi et al (30) demonstrated that tumors located in the upper neck had an increased risk of lateral LNM, while tumors in the lower neck had an increased risk of CLNM. Similarly, the present study determined that lower lobe location was significantly associated with CLNM in patients with cN0 PTMC. The association between tumor multifocality and CLNM remains controversial. In the present study, multifocality was not an independent predictor of CLNM, which is not consistent with previous reports (21,24). Therefore, the TTD, which was defined as the sum of the maximal diameter of each lesion, was calculated. Univariate and multivariate analyses revealed that TTD >10 mm was statistically significantly associated with CLNM in the present study, which was similar to the findings of a previous study by Zhao et al (31). In the present study, tumor bilaterality was not an independent predictive factor for CLNM, although this has been demonstrated to be significantly associated with CLNM in a previous study (24). ECS is an important prognostic characteristic of CLNM. ECS in patients with CLNM was significantly increased compared with that in those without CLNM in the present study, which was confirmed by certain previous reports (27,32). In the present study, coexisting chronic lymphocytic thyroiditis was found in 21.7% of patients with PTMC. Patients with chronic lymphocytic thyroiditis are considered to be more likely to exhibit thyroid carcinoma compared with patients without this condition (33,34). A previous report indicated that coexisting chronic lymphocytic thyroiditis in patients with papillary thyroid carcinoma had an inverse correlation with CLNM (35). The present study demonstrated that the frequency of CLNM was 48.2% (41/85 cases) in the chronic lymphocytic thyroiditis group, which was reduced compared with the control group. However, underlying chronic lymphocytic thyroiditis was not statistically significantly associated with CLNM.
To improve the diagnostic accuracy, a scoring system of independent predictors based on the multiple logistic regression analysis was established. The ROC curves had high predictive performance since the area under the ROC curves was 0.85 (95% CI, 0.81–0.90). Using the ROC curves, index point ≥2 was identified as the optimal cut-off point to distinguish CLNM-positive from CLNM-negative, with a sensitivity of 86.2% and a specificity of 70.4%. These results may aid surgeons to tailor the follow-up treatment of each patient accurately.
There are certain potential limitations in the current study. First, not all risk factors are included, such as lateral LNM, since only 12 patients (3.06%) enrolled in the present study underwent lateral neck dissection. Second, tumor recurrence and disease-free survival rate were not investigated. Further studies should be designed to confirm whether the scoring system is useful for treatment outcome.
In conclusion, this scoring system for the prediction of CLNM may be a practical and convenient diagnostic approach for cN0 patients with PTMC in Eastern China. Patients with an index point <2 may be considered as CLNM-negative. For patients with index point ≥2, the present study indicates that a CND may be routinely performed.