Abstract： Objective To detect serum biomarkers for pancreatic cancer associated diabetes and establish a model for diagnosis.Methods SELDI-TOF-MS was used to detect the differentially expressed serum proteins from 17 pancreatic cancer associated diabetes patients,17 new-onset type Ⅱ diabetes patients and 17 healthy controls,then a model of biomarkers was constructed and validated by Biomarker Patterns Software 5.0.Results Twelve discriminating m/z peaks were identified in the protein fingerprints in 10 pancreatic cancer associated diabetes patients,10 new-onset type Ⅱ diabetes patients and 10 healthy controls.Among them,the three biomarkers of mass/charge ratio 6116,6695 and 8936 were used to construct the model,which could diagnose 90％ pancreatic cancer associated diabetes form control groups.Blind test of other7 samples of three groups showed that 100％ pancreatic cancer associated diabetes,71％ new-onset diabetes and 86％ healthy controls were correctly classified.After searching protein database,there were metallothionein,pancreatic progenitor cell differentiation and proliferation factor-like protein,and fibroblastic growth factor 1,which were close to the weights of the above mentioned 3 differentially expressed proteins.Conclusions SELDI can identify 3 biomarkers for pancreatic cancer associated diabetes and a reliable model for diagnosis of pancreatic cancer associated diabetes is established.