Preoperative predictors for the nature of lesion in patients with focal masses in the setting of chronic pancreatitis

Abstract

Objective To establish a predictive model and to find the preoperative predictors for the nature of lesion in the setting of chronic pancreatitis. Methods The 121 patients from 7 tertiary medical centers in Shanghai from July 1998 to April 2007 with focal mass lesions in the setting of chronic pancreatitis were selected as the study population. The final diagnosis had to be confirmed histologically by surgical specimens (n =97) or by follow-up (n = 24). A case control study was conducted; the patients were divided into pancreatic cancer group and chronic pancreatitis group. The age, sex, past history, initial clinical presentations, lab results and imaging exams were collected by reviewing the medical records of these patients. χ~2 test and t test was used for univariate analysis, then the factors with P≤0. 25 were selected for further multivariate analysis, and multivariate logistic regression model was used to estimate odds ratio and 95% CI. Results Of 121 , 21 patients had a final diagnosis of pancreatic cancer and other 90 patients had a final diagnosis of chronic pancreatitis. Abdominal tenderness, direct bilirubin, CA19-9 and CEA were independent predictors of cancer in patients with focal mass lesions. Their odds ratios (95% CI) were 5. 691 (1.468, 22.070) , 1.011 (1.001 , 1.021) , 1.003 (1.001, 1.005) , 101.9 (0.988, 1.051) , respectively. Their P values were 0. 012, 0. 030, 0.003 and 0. 23 , respectively. Conclusions The logistic regression model may accurately predict the nature of lesion in the setting of chronic pancreatitis and may have certain clinical implication. Key words: Pancreatitis, chronic; Pancreatic neoplasms; Diagnosis, differential; Logistic regression; Predictive model