胰头癌肿瘤分期影响因素分析

Analysis of clinical factors associating with improvement of TNM staging accuracy in pancreatic head adenocarcinoma

目的探讨影响胰头癌(pancreatic head adenocarcinoma,PHA)肿瘤分期的临床因素。方法回顾性分析川北医学院附属医院2004年1月至2010年12月间经手术治疗的PHA临床资料。有序多分类Logistic回归模型行术后肿瘤TNM分期(post-operative TNM staging,pTNM)的单因素和多因素分析。交互验证检验多因素分析结果的判别效果。结果 121例PHA纳入本研究。术前影像学TNM分期为:肿瘤未明确28例(23.1%),IA9例(7.4%),IB45例(37.2%),IIA18例(14.9%),IIB15例(12.4%)和III6例(5.0%)。pTNM分别为IA2例(1.7%),IB7例(5.8%),IIA14例(11.6%),IIB6例(5.0%),III63例(52.1%),IV29例(24.0%)。肿瘤最大直径,术前影像学分期,ALT≥84.6U/L,中性粒细胞/淋巴细胞比值和CA19-9界值是pTNM的影响因素。多因素分析结果预测pTNM准确率为71.7%,交互验证准确率66.7%,若pTNM划分为IA+IB+IIA+IIB,III和IV三类,则预测准确率78.6%,交互验证准确率74.6%。结论影像学分期结合临床因素能够提高PHATNM分期准确度。

Objective To explore possible clinical factors contributing to improvement of TNM staging accuracy in pancreatic head adenocarcinoma （PHA）. Methods The retrospective study was conducted on 121 consecutive cases of PHA surgically treated from January 2004 to December 2010 in Affiliated Hospital of North Sicbuan Medical College. Univariate and multivariate analysis using ordinal logistic regression explored clinical factors relating to postoperative tumor staging （pTNM staging）. Cross-validation was done to test discriminant accuracy of muhivariate analysis results when compared with preoperatively radiologically TNM staging evaluation. Results Altogether 121 cases of potentially-resectable PHA entered the study. Preoperatively radiologically TNM staging evaluation revealed as followed： unidentified： 28 cases （23.1%）, IA 9 cases （7.4%）, IB 45 cases （37.2%）, IIA 18 cases （14.9%）, IIB 15 cases （12.4%） and III 6 cases （5.0%）, respectively. Simple radiological TNM staging accuracy was frustrating, since pTNM staging results as followed： IA 2 cases （1.7%）, IB 7 cases （5.8%）, IIA 14 cases （11.6%）, IIB 6 cases （5.0%）, III 63 cases （52.1%） and IV 29 cases （24.0%） , respectively. Multivariate analysis demonstrated that radiologically TNM staging, tumor diameters, ALT ≥ 84.6U/L, neutrophil-lymphocyte ratio and serum CA19-9 values contributed to improvement of accurate TNM staging. Cross-validation based on multivariate results exhibited better discriminant performance, demonstrating the number of observations correctly classified as 71.7% and 66.7% of cross-validated grouped cases correctly classified. When pTNM staging reclassified three categories as IA ＋ IB ＋ IIA ＋ IIB,III and IV, 78.6% of original grouped cases correctly classified and 74.6% of cross-validated grouped cases correctly classified. Conclusion Clinical factors associating with TNM staging could improve accuracy of PHA TNM staging and may be incorporated into current management strategy for PHA.