| Part 1. Differentiation of pancreatobiliary-type from intestinal-type periampullary carcinomas using 3.0 T MRIBackground and Purpose Typically, most periampullary carcinomas could be divided into either intestinal or pancreatobiliary type, although some other rare types could also be seen. The histological subtyping into intestinal-and pancreatobiliary-types is emerging as one independent prognostic factor, which is more important than tumor origin site. Findings in previous studies indicated that the prognosis of the intestinal type was better than that of the pancreatobiliary type. Therefore, how to differentiate the two types has become the new research hotspot. Recently, due to rapid development of MRI technique, it has been widely used in assessment of abdominal diseases. MR cholangiopancreatography (MRCP) has been reported to be more accurate than computed tomography (CT) for evaluating periampullary lesions, as MRCP provides more accurate information about the anatomy of biliary system and pancreatic duct. The apparent diffusion coefficient (ADC) calculated from diffusion-weighted imaging (DWI) could provide functional information reflecting tissue water mobility. The purpose of the current study was to determine if combined conventional MRI, MRCP and DWI can be beneficial for differentiating pancreatobiliary-type from intestinal-type periampullary carcinomas.Materials and Methods According to strict inclusion and exclusion criteria,43 patients (mean age,56.6 years; range,32-85 years) with periampullary carcinoma that were treated in our hospital from January 2011 to January 2015 were included in our study.All MRI exams were performed with a 3.0-T system (Magnetom Verio, Siemens, Germany). Two kinds of MRCP sequences were applied, including 2D MRCP and 3D MRCP sequence. Conventional MRI included a free-breathing HASTE sequence, a Tl-weighted dual-echo in-phase and out-phased sequence, and a breath-hold turbo spin-echo T2-weighted sequence. For DWI, respiratory-triggered, fat-suppressed, single-shot echo-planar imaging was performed. Each acquisition was obtained using two b values of 0, and 800 s/mm2. The ADC map was generated automatically with a commercially available software workstation system (Syngo Multimodality workplace, Siemens, Germany) base on ADC=-In (S (b=800)/S (b=0))/800. For gadobenate dimeglumine (Gd-BOPTA, MultiHance)-enhanced and unenhanced imaging, arterial phase (25 seconds), portal venous phase (60-70 seconds), equilibrium phase (3 minutes), delayed phase (10 minutes) and 90-minute hepatobiliary phase images were obtained by using a volume interpolated body examination (VIBE) sequence.Two radiologists reviewed MR images blindly and independently. For the analysis of MRCP images, the two observers recorded the following image features: (1) shape of the distal margin of common bile duct (CBD) (smooth or irregular); (2) symmetry of the distal lumen of CBD (symmetric or asymmetric); (3) narrowing of the distal end of CBD (tapering or abrupt); (4) presence of oval filling defect at the distal end of CBD and (5) maximum diameters of CBD and pancreatic duct (PD). For the analysis of enhancement pattern, signal intensity (SI) of each tumor on contrast-enhanced and unenhanced VIBE images was measured. SI ratios (SIR) (SIpost/SIpre×100%) of arterial phase, portal venous phase, equilibrium phase, and delayed phase were subsequently calculated. According to the highest SIR value of each phase, two observers recorded in which phase the enhancement peaked Enhancement pattern was then categoried into two kinds:rapid enhancement (peaked in arterial or portal venous phase) and progressive enhancement (peaked in equilibrium or delayed phase). For each lesion detected at conventional MRI, a ROI was defined at the corresponding location on DW image and ADC map. The ROI was drawn freehand on ADC map to include the area that was hyperintensity on DW image and hypointensity on ADC map, and to encompass as much of the lesion as possible. Two kinds of ADC values were recorded, including mean ADC (ADCmean) value and minimum ADC (ADCmin) value.An independent-sample t-test was used to compare ADC value between the two groups. The Chi-square or Fisher exact test was used to compare MRCP image feature and enhancement pattern between the two groups. Interobserver agreement was assessed by using Cohen ? statistic or intraclass correlation coefficient (ICC). Univariate and multiple logistic regression analysis were performed in order to identify MRI features with predictive values. Diagnostic performance of significant variables for differentiating the two groups was calculated by using receiver operating characteristic (ROC) curve analysis. The optimal cut-off value was defined according to Youden index. Kaplan-Meier method and log-rank test were used for drawing survival curves and comparing survival rates. All statistical analyses were performed using SPSS software (SPSS for Windows, version 17.0; Chicago, USA) and MedCalc software (MedCalc for Windows, version 12.7.0; Mariakerke, Belgium). A P-value< 0.05 was considered to be statistically significant.Results According to WHO pathologic criteria,28 patients were classified as pancreatobiliary-type, and 15 patients were intestinal-type. There was no significant difference in the five MRCP image features that described before between the two groups. For the analysis of tumor enhancement pattern,19 (68%) of 28 pancreatobiliary-type tumors peaked in equilibrium or delayed phase, whereas 13 (87%) of 15 intestinal-type tumors peaked in arterial or portal venous phase (P= 0.001). ADCmin value of pancreatobiliary-type group ([0.96 ± 0.21]×10-3 mm2/s) was significantly lower than that of intestinal-type group ([1.11 ± 0.24] ×10-3mm2/s) (P= 0.042). In our evaluation of interobserver agreement, the ? values and ICCs for the analyses all exceeded 0.8, indicating almost perfect agreement. At multiple logistic regression analysis, enhancement pattern was the only significant independent predictor (P= 0.002, odds ratio [OR]= 0.073).According to the ROC curve analysis, the optimal cut-off value of ADCmin for differentiating the two groups was 1.12 ×10-3 mm2/s, with the sensitivity and specificity being 85.7% and 53.3% correspondingly. When diagnostic predictive values were calculated with the combined model incorporating the enhancement pattern and ADCmin, the sensitivity and specificity reached to 71.4% and 86.7% correspondingly. The area under the curve (AUC) increased from 0.680 (for ADCmin alone) to 0.846 (for combined model).Results of survival analysis showed that the median survival time following resection of pancreatobiliary-type patients was significantly shorter than that of intestinal-type patients (x2= 4.602, P= 0.032).Conclusion and Significance Combined MRCP, contrast-enhanced MRI and DWI could effectively differentiate pancreatobiliary-type from intestinal-type periampullary carcinomas. Progressive enhancement and low ADCmin value are likely to suggest a pancreatobiliary-type periampullary carcinoma.Part 2. Differentiation of benign from malignant periampullary lesions using 3.0 T MRIBackground and Purpose Both malignant and benign periampullary lesions can cause biliary obstruction with similar clinical manifestations. Patients with benign lesions can be managed by more conservative treatment modalities. In this regard, prompt assessment and accurate differential diagnosis of biliary obstruction are very important for determining the most appropriate treatment and improving patient prognosis. Recently, with the rapid development of MRI technique, conventional MRI with MRCP and DWI can provide more comprehensive information, and have been widely used in assessment of biliary obstruction. The purpose of our study was to evaluate imaging features for discriminating malignant from benign periampullary lesions using combined MRCP, contrast-enhanced MRI and DWI, and compare diagnostic performances of different image sets.Materials and Methods Patients with extrahepatic bile duct obstruction down to the level of periampullary region were collected from January 2011 to January 2015. A total of 108 patients who met the inclusion and exclusion criteria were included in our study, and all of them were confirmed by histological proof or follow-up of at least 12 months. MRI sequences and parameters were the same with part 1.Two radiologists reviewed MR images blindly and independently. In the first step, the two observers evaluated different image features of benign and malignant periampullary lesions. For the analysis of the MRCP images, the observers recorded the following features:(1) shape of the distal margin of the CBD (smooth or irregular); (2) symmetry of the distal lumen of the CBD (symmetric or asymmetric); (3) narrowing of the distal end of the CBD (tapering or abrupt); and (4) the maximum diameters of CBD and PD. For the analysis of enhancement pattern, three subgroups were classified:(1) hypovascular enhancement with gradual enhancement; (2) hypervascular enhancement on arterial or portal venous phase with persistent or washout enhancement; and (3) undefined enhancement. In hepatobiliary phase, presence or absence of contrast agent in CBD was recorded. For the assessment of DWI sequence, observers recorded presence or absence of restricted diffusion of each lesion.In the second step, another two experienced radiologists who did not participate in the image analysis part evaluated the added value of combined image sets. They independently reviewed three image sets:(1) MRCP alone; (2) MRCP with contrast-enhanced MR images; and (3) combined sets of MRCP, contrast-enhanced MR images and DW images. During each reading session, observers used a five-point scale to record their confidence level for the diagnosis of benign or malignant periampullary lesion:(1) definitely benign; (2) probably benign; (3) indeterminate; (4) probably malignant; and (5) definitely malignant. The criteria were based on several previous studies.An independent sample t-test was used to compare maximum diameters of CBD and PD between benign and malignant group. Chi-square or Fisher exact test was used to compare differential imaging findings between two groups. Interobserver agreement was assessed using the Cohen's k statistic or ICC. Diagnostic performance of each image set and each observer was calculated by using ROC curve analysis. Area under the ROC curve (AUC) was calculated. Pairwise comparisons of the ROC curves were performed to compare the diagnostic performances of different image sets. All statistical analyses were performed using SPSS software (version 17.0; Chicago, USA) and MedCalc software (version 12.7.0; Mariakerke, Belgium). A P-value< 0.05 was considered to be statistically significant.Results One hundred and eight patients were evaluated, including 43 patients with malignant periampullary lesions (mean age,56.6 years; range,32-85 years) and 65 patients with benign periampullary lesions (mean age,60.9 years; range,26-89 years).In the first step, observers evaluated different image features of benign and malignant periampullary lesions. The results were as follows:In this reading session, the ? values and ICCs for the analyses all exceeded 0.7, indicating good agreement.1. MRCP image features:Fifty (77%) of 65 benign lesions showed smooth margin of the distal CBD, whereas 25 (58%) of 43 malignant lesions showed irregular margin (P< 0.001). Symmetric lumen of CBD was seen in 33 (51%) of 65 benign lesions, whereas 36 (84%) in 43 malignant lesions had asymmetric lumen (P< 0.001). There was no significant difference in narrowing of the distal end of CBD between the two groups, since 27 (42%) in 65 benign lesions and 26 (60%) in 43 malignant lesions showed abrupt narrowing (P= 0.054). The maximum diameters of CBD and PD in malignant group (17.79 ± 4.52 mm and 3.98 ± 2.49 mm, correspondingly) were significantly larger than those of the benign group (14.11 ± 4.61 mm and 2.60 ±1.12 mm, correspondingly) (P< 0.001, P= 0.001, correspondingly).2. Enhancement pattern:Benign lesions showed more hypervascular enhancement on arterial or portal venous phase (85%,55 in 65 patients), whereas 22 (51%) of 43 malignant lesions showed hypervascular enhancement on arterial or portal venous phase, and 20 (47%) of 43 malignant lesions showed hypovascular enhancement with gradual enhancement (P= 0.001). In hepatobiliary phase, contrast agent was present in CBD in 60%(39 in 65 patients) of benign lesions, whereas 70% (30 in 43 patients) of malignant lesions showed no contrast agent in CBD (P= 0.002)3. DW image features:Only three (5%) in 65 benign lesions showed restricted diffusion, whereas 37 (86%) in 43 malignant lesions showed restricted diffusion (P< 0.001).In the second step, observers evaluated the added value of combined image sets. The results were as follows:Interobserver agreement regarding the confidence level was moderate (k= 0.443) for the first reading session, good (k= 0.609) for the second reading session, and excellent (k= 0.806) for the third reading session by two observers.Diagnostic performance for detection and differentiation of benign periampullary lesions from malignant lesions significantly improved for both observers after the combination of MRCP and contrast-enhanced MR images with or without DW images compared with MRCP alone. AUC value improved from 0.776 to 0.981 (without DW images) (P< 0.001) and 0.994 (with DW images) (P< 0.001) for observer 1, and from 0.808 to 0.982 (without DW images) (P< 0.001) and 0.996 (with DW images) (P< 0.001) for observer 2. The difference in AUC value between MRCP with contrast-enhanced MR images and combined sets of MRCP, contrast-enhanced MR images and DW images was not statistically significant for observer 1 (P= 0.065). But for observer 2, the AUC value of combined sets of MRCP, contrast-enhanced MR images and DW images was statistically significantly higher than that of MRCP with contrast-enhanced MR images (P= 0.049).Conclusion and Significance Combined use of MRCP, contrast-enhanced MR images and DW images could effectively help differentiate malignant from benign periampullary lesions, and combined image sets could significantly increase the diagnostic performance.Part 3. Analysis of prognostic factors in periampullary carcinomas following resectionBackground and Purpose Pancreaticoduodenectomy is the most widely used curative surgery for resectable periampullary carcinomas, and is also the only treatment modality to offer a chance for cure. However, long-term survival rates following resection vary greatly. The significant prognostic factors that may predict patient survival are still controversial. We collected and analyzed clinical and histopathological data systematically, and aimed to identify the risk factors for patient survival following resection in northern part of China.Materials and Methods We retrospectively reviewed the clinical and histopathological features of 337 patients with periampullary carcinoma that were treated with curative surgery from June 2006 to July 2013 in our institution. The collected data included patient characteristics, clinical symptoms and signs, results of laboratory tests, treatment modalities, perioperative conditions, as well as histopathological features. Kaplan-Meier method and log-rank test were used for drawing survival curves and comparing survival rates. Independent prognostic factors affecting survival were determined by Cox hazard regression test. All statistical analyses were performed using SPSS software (SPSS for Windows, version 17.0; Chicago, USA). A P-value< 0.05 was considered to be statistically significant.Results A total of 337 patients (mean age,58.15 years; range,25-83 years) with periampullary carcinoma were analyzed, including 228 men (mean age,58.72 years; range,31-83 years) and 109 women (mean age,56.95 years; range,25-78 years). The median survival time of all patients was 26 months. The 3-year and 5-year overall survival rates were 38.0% and 24.3%, correspongdingly. The results of univariate Cox analysis showed that serum CEA level (P= 0.002, RR= 1.699), CA199 level (P= 0.004, RR= 1.584), total bilirubin (TBIL) level (P= 0.014, RR= 1.516), intraoperative blood loss (P= 0.003, RR= 2.032), intraoperative transfusion of packed red blood cells (P< 0.001, RR= 1.417), tumor origin (P< 0.001, RR= 1.555), tumor size (P= 0.001, RR= 1.797), tumor differentiation (P= 0.004, RR= 1.556), T stage (P= 0.002, RR= 1.549), lymph node metastasis (P< 0.001, RR= 1.789), number of positive lymph node (P= 0.003, RR= 1.370), lymph node ratio (LNR) (P < 0.001, RR= 1.294) and vascular invasion (P= 0.021, RR= 1.646) were significantly associated with patient survival following resection. In multivariate Cox analysis, CA199 level> 39U/ml, intraoperative transfusion of packed red blood cells > 4 units, tumor originated from pancreas, T3-T4 stage, and LNR? 0.300 were independent risk factors for patient survival following resection.We continued to carry out subgroup analyses according to tumor origin (pancreas originated tumor vs non-pancreas originated tumor) and lymph node status (number of positive lymph node and number of resected lymph node).In our group, compared with non-pancreas originated tumor, tumor originated from pancreas had larger tumor size (P< 0.001) and higher T stage (P= 0.024). Survival analysis showed that if non-pancreas originated tumor invaded pancreas, the survival rate decreased significantly (P< 0.001). In univariate Cox regression analysis, pancreas invasion was a risk factor for a poor outcome of patients with non-pancreas originated tumor (P< 0.001, RR=2.083).In patients with lymph node metastasis, the survival rate of patients with resected lymph node< 7 was lower than that of the patients with resected lymph node? 7 (P= 0.041). In univariate Cox regression analysis, resected lymph node< 7 was a risk factor for a poor outcome of patients with lymph node metastasis (P= 0.048, RR= 1.580). In patients without lymph node metastasis, the survival rate of patients with resected lymph node< 16 was lower than that of the patients with resected lymph node? 16 (P= 0.036). In univariate Cox regression analysis, resected lymph node< 16 was a risk factor for a poor outcome of patients without lymph node metastasis (P = 0.043,RR= 2.189).Conclusion and Significance In our group, CA199 level? 39U/ml, intraoperative transfusion of packed red blood cells? 4 units, tumor originated from pancreas, T3-T4 stage, and LNR? 0.300 were independent risk factors for patient overall survival following resection. |
PDF Full Text Request