review of literature on clinical pancreatology - The Pancreapedia

understand the utility <strong>of</strong> these devices, one need to first consider mechanical properties suchas radial force (expansion force) and axial force (straightening force) [517].Endoscopic biliary drainage is widely accepted as palliative treatment in patients withpancreatic cancer. One study was designed to compare self-expanding metal stent andpolyethylene stent in a homogeneous patient group with advanced pancreatic cancer. Thestudy included 154 patients initially treated with a metal or plastic stent. Median survival time,stent patency, and stent-associated hospital admissions were evaluated. The mediansurvival time in patients treated with metal and plastic stent was 6 and 4 months,respectively. Self-expanding metal stents have a significantly higher patency rate thanpolyethylene stents. Stent occlusion was observed in 21 (33 %) <strong>of</strong> 63 patients in the plasticstent group after a median period <strong>of</strong> 57 days and in 17 (19 %) <strong>of</strong> 91 patients in the metalstent group after a median period <strong>of</strong> 126 days. The total time <strong>of</strong> hospital stay after initialimplantation <strong>of</strong> metal or plastic stent was 7 and 17 days, respectively. It was concluded thatself-expanding metal stents have a longer patency than polyethylene stents. Additionally, thenumber <strong>of</strong> stent-associated hospital admissions and the total time <strong>of</strong> hospital stay werehigher in the plastic stent group. The median survival time was not significantly different inboth groups [518].Prediction <strong>of</strong> survivalThe pancreatic nomogram, originally developed in the Memorial Sloan-Kettering CancerCenter in the USA, combines clinicopathological and operative data to predict diseasespecificsurvival at 1, 2 and 3 years from initial resection. An external patient cohort from aretrospective pancreatic adenocarcinoma database at the Academic Medical Centre inAmsterdam was used to test the validity <strong>of</strong> the pancreatic adenocarcinoma nomogram. Thecohort included 263 consecutive patients who had surgery between 1985 and 2004. Data forall the necessary variables were available for 256 patients (97 %). At the last follow-up, 35patients were alive, with a median follow-up <strong>of</strong> 27 (range 3-114) months. The 1-, 2- and 3-year disease-specific survival rates were 61, 30 and 16 percent respectively. The nomogramconcordance index was 0.61. The calibration analysis <strong>of</strong> the model showed that the predictedsurvival did not significantly deviate from the actual survival. The Memorial Sloan-KetteringCancer Center pancreatic cancer nomogram provided an accurate survival prediction. It mayaid in counselling patients and in stratification <strong>of</strong> patients for clinical trials [519].Identification <strong>of</strong> defined patient groups based on a prognostic index may improve theprediction <strong>of</strong> survival and selection <strong>of</strong> therapy for patients with pancreatic cancer. Manyprognostic factors have been identified <strong>of</strong>ten based on retrospective, underpowered studieswith unclear analyses. Data from 653 patients were analysed. Continuous variables are <strong>of</strong>tensimplified assuming a linear relationship with log hazard or introducing a step function(dichotomising). Misspecification may lead to inappropriate conclusions but has not beenpreviously investigated in pancreatic cancer studies. Models based on standard assumptionswere compared with a novel approach using nonlinear fractional polynomial transformations.The model based on fractional polynomial-transformed covariates was most appropriate andconfirmed five previously reported prognostic factors: albumin, CA 19-9, alkalinephosphatase, LDH and metastases, and identified three additional factors not previouslyreported: WBC, AST and BUN. The effects <strong>of</strong> CA 19-9, alkaline phosphatase, AST and BUNmay go unrecognised due to simplistic assumptions made in statistical modelling. Weadvocate a multivariable approach that uses information contained within continuousvariables appropriately. The functional form <strong>of</strong> the relationship between continuouscovariates and survival should always be assessed [520].