乳突狀甲狀腺癌中央區頸部淋巴結轉移預測模型及風險評分系統之建立

Abstract

目的：透過乳突狀甲狀腺癌臨床資料建構中央區頸部淋巴結轉移之羅吉斯迴歸預測模型及風險評分系統。 方法：採用回溯性研究方法，對1984年至2018年間在林口長庚紀念醫院初次行甲狀腺全摘除及中央區淋巴結清除術之650位甲狀腺乳突癌患者臨床資料進行單因子分析與多變量羅吉斯迴歸模型，以探討和確定危險因素。然後，透過迴歸分析建立Logistic評分系統和Additive評分系統，以預測發生中央區頸部淋巴結轉移的風險。 結果：我們的結果顯示，年齡、性別、腫瘤大小及病灶多灶性為甲狀腺乳突癌中央區淋巴結轉移相關危險因子。Logistic評分系統分數為2分時，敏感度為51.3%，特異性為70.4%，陽性預測值為66.8%，陰性預測值為55.5%，ROC曲線下面積（AUC）為0.646（95% CI = 0.606-0.686）。Additive評分系統分數為4.1分時，敏感度為53.0%，特異性為67.1%，陽性預測值為65.1%，陰性預測值為55.2%，AUC為0.645（95% CI = 0.605-0.685）。Logistic評分系統及Additive評分系統透過Hosmer-Lemeshow goodness-of-fit檢定結果分別為 χ2=6.379（P=0.496）、χ2=9.941（P=0.077）。 結論：我們的發現表明，疑似中央區頸部淋巴結轉移之甲狀腺乳突癌患者可透過年齡、性別、腫瘤大小及病灶多灶性來預測淋巴結轉移風險。研究結果，當估算患者相對應之Logistic評分系統score ≥2分或Additive評分系統score ≥4.1分時，則該病患對於中央區頸部淋巴結轉移屬高風險，應考慮及時接受手術。但兩評分系統的鑑別力不佳，期許未來資料庫可增加相關變項資料，例如甲狀腺超音波影像臨床特徵或BRAF基因檢測，以提升兩個評分系統效度，並為乳突狀甲狀腺癌治療計劃提供量化依據，便於臨床醫師參考。

Objective: To construct a Logistic regression prediction model and risk scoring system for central cervical lymph node metastasis using clinical data of papillary thyroid carcinoma. Methods: A retrospective study was conducted using the clinical data of 650 patients with papillary thyroid carcinoma who underwent total thyroidectomy and lymph node dissection in the Linkou Chang Gung Memorial Hospital between 1984 and 2018. Univariate analysis and multivariate logistic model were applied to explore and define the risk factors, followed by regression analysis to establish the Logistic scoring system and the Additive scoring system in order to predict the risk of central cervical lymph node metastasis. Results: Our results showed age, gender, tumor size, and multifocality are risk factors for central cervical lymph node metastasis of papillary thyroid carcinoma. When the score derived from the Logistic scoring system reached 2, the sensitivity was 51.3%, the specificity was 70.4%, the positive predictive value (PPV) was 66.8%, the negative predictive value (NPV) was 55.5%, and the area under the curve (AUC) was 0.646 (95% CI=0.606-0.686). When the score derived from the Additive score system reached 4.1, the sensitivity was 53.0%, the specificity was 67.1%, the PPV was 65.1%, the NPV was 55.2%, and the AUC was 0.645 (95% CI=0.605-0.685). The results of the Logistic scoring system and the Additive scoring system through Hosmer-Lemeshow goodness-of-fit test were χ2=6.379 (P=0.496) and χ2=9.941 (P=0.077). Conclusion: Our findings suggest that the risk of lymph node metastasis in patients with papillary thyroid carcinoma who are suspected of central cervical lymph node metastasis could be predicted by their age, gender, tumor size, and multifocality. According to our results, when patients’ corresponding Logistic scoring system score is greater than or equal to 2 or the Additive scoring system score is greater than or equal to 4.1, they have a higher risk for central cervical lymph node metastasis and should be considered for surgery. However, the discrimination of the two scoring systems were not good. Our findings suggest that future databases can acquire additional variables such as clinical features of thyroid ultrasound imaging or BRAF (B-Raf proto-oncogene) genetic test to improve the effectiveness of the two scoring systems and provide a quantitative basis for the treatment plan of papillary thyroid cancer.