臨床預(yù)測模型構(gòu)建機(jī)器學(xué)習(xí)r語言進(jìn)階-講義等輔助資料r語言進(jìn)階-第9-13章ch09jco 2013 wang

1、VOLUME 31 NUMBER 9 MARCH 20 2013Prognostic Nomogram for Intrahepatic Cholangiocarcinoma After Partial HepatectomyYizhou Wang, Jun Li, Yong Xia, Renyan Gong, Kui Wang, Zhenlin Yan, Xuying Wan,Guanghua Liu, Dong Wu, Lehua Shi, Wanyee Lau, Mengchao Wu, and Feng ShenAll authors, Eastern Hepatobiliary Su

2、rgery Hospital, Second Military Medi- cal University, Shanghai; Wanyee Lau, Chinese University of Hong Kong, Hong Kong, China. ABSTRACTPurposeThis study aimed to establish an effective prognostic nomogram for intrahepatic cholangiocarci- noma (ICC) after partial hepatectomy.Patients and MethodsThe n

3、omogram was based on a retrospectively study on 367 patients who underwent partial hepatectomy for ICC at the Eastern Hepatobiliary Surgery Hospital from 2002 to 2007.The predictive accuracy and discriminative ability of the nomogram were determined by concordance index (C-index) and calibration cur

4、ve and compared with five currently used staging systems on ICC. The results were validated using bootstrap resampling and a prospective study on 82 patientsPublished online ahead of print at on January 28, 2013.Supported by Grants No. 2008ZX10002- 025,2012ZX10002-016 from the State Key

5、Project on Infectious Diseases of China (S.F.), No. 3091006;30772141from the National Natural Science Foun- dation of China (S.F.), and No.SHDC12010121 from the Shanghai Hospital Development Program (S.F.).operated on from 2007 to 2008 at the sameResultstitution.On multivariate analysis of the prima

6、ry cohort, independent factors for survival were serum carcinoembryonic antigen, CA 19-9, tumor diameter and number, vascular invasion, lymph node metastasis, direct invasion, and local extrahepatic metastasis, which were all selected into the nomogram. The calibration curve for probability of survi

7、val showed good agreement between prediction by nomogram and actual observation. The C-index of the nomogram for predicting survival was 0.74 (95% CI, 0.71 to 0.77), which was statistically higher than the C-index values of the following systems: American Joint Committee on Cancer (AJCC) seventh edi

8、tion (0.65), AJCC sixth edition (0.65), Nathan (0.64), Liver Cancer Study Group of Japan (0.64), and Okabayashi (0.67; P .001 for all). It was also higher (0.74) in predicting survival for the mass-forming type of ICC (P .001). In the validation cohort, the nomogram discrimination was superior to th

9、e five other staging systems (C-index: 0.75 v 0.60 to 0.63; P .001 for all).ConclusionThe proposed nomogram resulted in more-accurate prognostic prediction for patients with ICC after partial hepatectomy.Y.W., J.L., and Y.X. contributed equally to this work.Authors disclosures of potential con- flic

10、ts of interest and author contribu- tions are found at the end of this article.Corresponding author: Feng Shen, MD, Department of Hepatic Surgery, East- ern Hepatobiliary Surgery Hospital, Second Military Medical University, Shanghai, 200438, P.R. China; e-mail: . 2013 by Ame

11、rican Society of Clinical Oncology0732-183X/13/3109-1188/$20.00DOI: 10.1200/JCO.2012.41.5984J Clin Oncol 31:1188-1195. 2013 by American Society of Clinical Oncologyand/or distant metastases.7-9 ICC differs from HCC in carcinogenesis and biologic behaviors.10 It also differs from hilar and distal bil

12、e duct cholangiocar- cinoma in clinical features, imaging manifestations, and therapeutic strategies.11 Therefore, ICC is a dis- tinct hepatobiliary malignancy and requires a dis- tinct prognostic predictivemodel.The most commonly used staging system for ICC is the TNMclassificationsystem. Okabayash

13、iet al12 proposed a new staging system for the MF type of ICC based on the TNM staging system, but the tumor diameter is not used as a parameter because of its irrelevance to prognosis of resectable ICC. However, Yamasaki13 reported that ICC with a di- ameter greater than 2 cm had a relatively poor

14、prog- nosis. Thus, 2 cm was regarded as a cutoff value for the T classification, and this is adopted as an INTRODUCTIONIntrahepatic cholangiocarcinoma (ICC) is a neo- plasm that originates from the endothelial cells of segmental or proximal branches of the bile duct.1 It accounts for 5% to 30% of al

15、l primary liver malig- nancies2,3 and has an incidence inferior only to hepatocellular carcinoma (HCC). Three gross patho- logic types of ICC are defined by the Liver Cancer Study Group of Japan (LCSGJ); these are the mass- forming (MF) type, periductal infiltrating type, and intraductal growth type

16、.1 In recent years, the inci- dence and mortality of ICC are increasing world-wide.4 Partial hepatectomy remathe matay ofcurative treatment for ICC.5,6 Unfortunately, prog- nosis after partial hepatectomy is unsatisfactory, with a high incidence of locoregional recurrence1188 2013 by American Societ

17、y of Clinical OncologyDownloaded from on August 9, 2014. For personal use only. No other uses without permission.Copyright 2013 American Society of Clinical Oncology. All rights reserved.JOURNAL OF CLINICAL ONCOLOGYO R I G I NA LR E P O R TNomogram for Intrahepatic Cholangiocarcinom

18、aimportant parameter by the LCSGJ system.13 The widely used sixth edition of the American Joint Committee on Cancer (AJCC) TNM staging system concludes that patients with ICC with a tumor greater than 5 cm in diameter have a poorer prognosis. However, this system is based on the data from HCC, and i

19、ts applicability to ICC is still uncertain.14,15 Nathan et al16 proposed an improved TNM staging system in which the tumor diameter was excluded in the T classification. This view is further adopted in the seventh edition of the AJCC TNM staging system (Appendix Table A1, online only).17 The lack of

20、 consensus on which staging system to use has led to considerable confusion.Currently, nomograms have been developed in the majority of cancer types.18-20 The use of nomograms has compared favor- ably to the traditional staging systems for many cancers, and thus, it has been proposed as an alternati

21、ve or even as a new standard.21-23 To our knowledge, this study is the first attempt to establish a prognostic nomogram for resectable ICC based on the clinicopathologic data of 367 patients with ICC who underwent partial hepatectomy, to determine whether this model provides more-accurate prediction

22、 of patient survival when compared with the currently available staging systems.patic metastases and newly found intrahepatic nodules identified intraop- eratively were removed whenever possible. Hepaticojejunostomy was carried out in patients with tumor involvement of the primary and second- ary bi

23、le ducts.Histopathologic study of the resected specimens was carried out independently by three pathologists who came to a consensus by discus- sion if there was any controversy. The pathologic diagnosis of ICC was confirmed only if CK7 positive, CK19 positive, MUC1 positive, CK20 negative, HepPar1

24、negative, and glypican-3 negative phenotype was iden- tified in an immunohistochemisty test.25 Pathologic features, such as tu- mor diameter, number, capsule, site, surgical margin, vascular invasion, lymph node metastasis, hepatitis, and cirrhosis, were documented, and the degree of cell differenti

25、ation wasdetermined.Follow-UpPatients were observed once every 2 months in the first 2 years after surgery and then every 3 to 6 monthsthereafter. At each of the follow-up visits, a detailed history and a complete physical examination were carried out. Blood was taken for serum CA 19-9, CEA, -fetopr

26、otein, and liver function tests, and an abdominal ultrasound was carried out. Contrast-enhanced CT or magnetic resonance imaging was performed once every 6 months or earlier when tumor recurrence or metastasis was suspected. Further investigation was carried out when clinically indicated. ICC recurr

27、ence/metastasis was defined as the ap- pearance of a newly detected tumor confirmed on two radiologic images, with or without elevation of serum tumor markers. Overall survival (OS) and time to recurrence (TTR) were used as primary end points. OS was defined as the interval between partial hepatecto

28、my and death or the last date of follow-up. TTR was calculated from surgery to the date when recurrence/metastasis was diagnosed.Categorization of Patients in Different Staging SystemsPatients were categorized according to five current staging systems (the sixth and seventh editions of the AJCC TNM

29、classification and the Nathan, Okabayashi, andLCSGJ staging systems).12-17Statistical AnalysisStatistical analyses to identify risk factors were performed using SPSS15.0 for Windows (SPSS, Chicago, IL). Categorical variables were grouped based on clinical findings, and decisions on the groups were m

30、ade before modeling. The results were compared using the 2 test or Fishers exact test. Continuous variables were compared using the t test or Mann-Whitney U test for variables with an abnormal distribution. Survival curves were depicted using the Kaplan-Meier method and compared using the log-rank t

31、est. Cox regression analysis was used for multivariate analyses.A nomogram was formulated based on the results of multivariate anal- ysis and by using the package of rms26 in R version 2.14.1 (/). A final model selection was performed by a backward step- down selection process

32、 with the Akaike information criterion.27 The perfor- mance of the nomogram was measured by concordance index (C-index) and assessed by comparing nomogram-predicted versus observed Kaplan-Meier estimates of survival probability. Bootstraps with 1,000 resample were used for these activities. Comparis

33、ons between the nomogram and other staging sys- tems were performed with the rcorrp.cens package in Hmisc28 in R and were evaluated by the C-index. The larger the C-index, the more accurate was the prognostic prediction.29 During the external validation of the nomogram, the total points of each pati

34、ent in the validation cohort were calculated according to the established nomogram, then Cox regression in this cohort was per- formed using the total points as a factor, and finally, the C-index and calibra- tion curve were derived based on the regression analysis. P .05 was considered statisticall

35、y significant. The related computerized programs for nomograms with R are listed in the Appendix (online only). PATIENTS AND METHODSPatients and Study DesignA retrospective study was conducted on a primary cohort of patients who underwent partial hepatectomy for ICC between September 2002 and Febru-

36、 ary 2007 at the Eastern Hepatobiliary Surgery Hospital (Shanghai, China). Inclusion criteria included the following: no history of previous anticancer therapy; no history of other malignancies; complete resection of macroscopic liver tumors; and histopathologically proven ICC. Exclusion criteria we

37、re as follows: hilar or extrahepatic cholangiocarcinoma, including intrahepatic me- tastasis of extrahepatic cholangiocarcinoma; tumors of uncertain origin or probable metastatic liver tumor; mixed type of primary liver cancer as con- firmed histopathologically; and perioperativemortality.From Septe

38、mber 2007 to April 2008, an independent cohort of consec- utive patients who underwent partial hepatectomy for ICC was prospectively studied, using the same inclusion and exclusion criteria. These patients formed the validation cohort of this study.The study was censored on October 19, 2011. The stu

39、dy was approvedby thetitutional ethics committee. Informed consent was obtainedbefore surgery.Diagnosis and TreatmentAfter a detailed history and a complete physical examination, blood was taken from the patients for hepatitis B surface antigen, hepatitis B virus DNA level, anti hepatitis C virus (H

40、CV) antibody, serum albumin, total bilirubin, ALT, -glutamyltransferase, -fetoprotein, CA 19-9, and carci- noembryonic antigen (CEA). Other routine investigations included chest x-ray, upper GI endoscopy, abdominal ultrasound, contrast-enhanced computed tomography (CT), and/or magnetic resonance ima

41、ging. In re- cent years, positron emission tomography was performed in patients with clinical or radiologic suspicion of intrahepatic or extrahepatic metastases. A preoperative diagnosis of ICC was based clinically, radiologically, and on raised serum markers.24Partial hepatectomy was carried out ac

42、cording to tumor diameter, location, presence or absence of cirrhosis, and estimated volume of the future liver remnant. Liver resection was carried out based on Couinauds segments, sectors, and hemilivers. Routine dissection of lymph nodes in the hepatoduodenal ligament and retropancreatic and/or p

43、ara-aortic lymph nodes was performed for either preoperatively or intraoperatively diagnosed ICC. Direct invasions of adjacent structures and local extrahe- RESULTSClinicopathologic Characteristics of PatientsIn the primary cohort, of 421 patients with ICC who received partial hepatectomy during the

44、 study period, 367 met the 2013 by American Society of Clinical Oncology1189Downloaded from on August 9, 2014. For personal use only. No other uses without permission.Copyright 2013 American Society of Clinical Oncology. All rights reserved.Wang et alJOURNAL OF

45、CLINICAL ONCOLOGY1190 2013 by American Society of Clinical OncologyDownloaded from on August 9, 2014. For personal use only. No other uses without permission.Copyright 2013 American Society of Clinical Oncology. All rights reserved.Table 1. Demographics and Clinicopathologic Charact

46、eristics of Patients With Intrahepatic CholangiocarcinomaPrimary Cohort (n 367)Validation Cohort (n 82) Demographic or CharacteristicNo. of Patients%No. of Patients%SexMale24667.05263.4Female12133.03036.6Age, yearsMedian5356Range27-7837-79History of hepatitisYes14840.33137.8No21959.75163.2History of

47、 biliary diseaseYes6016.31012.2No30783.77287.8AFP, g/LMedian4.24.2Range0.6-1,2100.6-1,210CEA, g/LMedian2.52.8Range0.1-809.60.3-809.6CA 19-9, U/mLMedian41.236.1Range0.4-1,0000.6-1,000TBIL, mol/LMedian13.213.6Range4.7-333.94.2-304.2ALB, g/LMedian42.243.7Range29.8-64.823.4-61.9ALT, U/LMedian28.624.6Ran

48、ge3.5-2,6493.0-2,048GGT, U/LMedian75.586Range0.9-1,50210-1,000.5HBsAgPositive18751.04352.4Negative18049.03947.6HBeAgPositive369.8910.9Negative33190.27389.1Anti-HCVPositive82.200Negative33197.882100Blood transfusionYes6116.61518.3No30683.46781.7Tumor size, cmMedian5.55.0Range0.4-22.01.2-15.0Tumor num

49、ber34211.411.2332588.68198.8Surgical margin, cmMedian0.50.4Range0.1-3.10.1-3.5CirrhosisYes7821.31821.9No28978.76478.1(continued on following page)Nomogram for Intrahepatic Cholangiocarcinomacriteria be entered onto this study. For the validation cohort, we studied 82 consecutive patients. The clinic

50、opathologic characteristics of patients in the primary and validation cohorts are listed in Table 1. The gross pathologic types of ICC were MF type (n 345; 94.0%), periductal infiltrating type (n 20; 5.5%), and intraductal growth type (n 2; 0.5%).1Independent Prognostic Factors in the Primary Cohort

51、 The results of the univariate analysis are listed in Appendix Table A2 (online only). Multivariate analyses demonstrated that serum CEA, CA 19-9, tumor diameter and number, lymph node metastasis, vascular in- vasion, and direct invasion and local extrahepatic metastasis were inde-pendent risk facto

52、rs for tumor recurrence and OS (Table 2).Tumor Recurrence and OS in the Primary CohortThe median follow-up time was 39.3 months (range, 7.5 to 107.2 months), the median TTR was 13.8 months (range, 1.3 to 63.6 months), and the postoperative 1-, 3-, and 5-year recurrence rates were 48.9%, 62.5%, and 6

53、7.3%, respectively. Themedian OS was 21.0months (range, 1.6 to 105.7 months), and the 1-, 3-, and 5-year OSrates were 61.9%, 40.8%, and 35.2%, respectively.Prognostic Nomogram for OSThe prognostic nomogram that integrated all significant inde- pendent factors for OS in the primary cohort is shown in

54、 Figure 1. The C-index for OS prediction was 0.74 (95% CI, 0.71 to 0.77). The calibration plot for the probability of survival at 3 or 5 year after surgery showed an optimal agreement between the prediction by nomogram and actual observation (Figs 2A and 2B). 2013 by American Society of C

55、linical Oncology1191Downloaded from on August 9, 2014. For personal use only. No other uses without permission.Copyright 2013 American Society of Clinical Oncology. All rights reserved.Table 2. Multivariate Analysis of the Primary CohortRecurrenceOverall SurvivalVariablePHR95% CIPHR

56、95% CICA 19-9.0481.0001.000 to 1.001.0041.0011.000 to 1.001CEA.0111.0101.002 to 1.018.0011.0111.005 to 1.018Vascular invasion (yes v no).0151.5461.088 to 2.196.0051.6051.154 to 2.231Direct invasion and local extrahepatic metastasis (yes v no).0261.5721.055 to 2.342.0251.5921.061 to 2.390Lymph node m

57、etastasis (yes v no).0011.7021.241 to 2.335.0012.0571.531 to 2.764Tumor diameter.0031.0571.018 to 1.096.0011.0781.041 to 1.116Tumor number 1 nodule2-3 nodules.0401.5861.021 to 2.464.0451.5821.011 to 2.474 3 nodules, or periductal invasion.0014.4022.874 to 6.742.0016.0964.059 to 9.155Abbreviations: CEA, carcinoembryonic antigen; HR, hazard ratio.Table 1. Demographics and Clinicopathologic Characteristics of Patients With Intrahepatic Cholangiocarcinoma (continued)Primary Cohort (n 367)Validation Cohort (n 82) Demographic or Charac