Serum cytokine biomarker panels for discriminating pancreatic cancer from benign pancreatic disease

Abstract

Background: We investigated whether combinations of serum cytokines, used with logistic disease predictor models, could facilitate the detection of pancreatic ductal adenocarcinoma (PDAC).

Methods: The serum levels of 27 cytokines were measured in 241 subjects, 127 with PDAC, 49 with chronic pancreatitis, 20 with benign biliary obstruction and 45 healthy controls. Samples were split randomly into independent training and test sets. Cytokine biomarker panels were selected by identifying the top performing cytokines in best fit logistic regression models during multiple rounds of resampling from the training dataset. Disease prediction by logistic models, built using the resulting cytokine panels, was evaluated with training and test sets and further examined using resampled performance evaluation.

Results: For the discrimination of PDAC patients from patients with benign disease, a panel of IP-10, IL-6, PDGF plus CA19-9 offered improved diagnostic performance over CA19-9 alone in the training (AUC 0.838 vs. 0.678) and independent test set (AUC 0.884 vs. 0.798). For the discrimination of PDAC from CP, a panel of IL-8, CA19-9, IL-6 and IP-10 offered improved diagnostic performance over CA19-9 alone with the training (AUC 0.880 vs. 0.758) and test set (AUC 0.912 vs. 0.848). Finally, for the discrimination of PDAC in the presence of jaundice from benign controls with jaundice, a panel of IP-10, IL-8, IL-1b and PDGF demonstrated improvement over CA19-9 in the training (AUC 0.810 vs. 0.614) and test set (AUC 0.857 vs. 0.659).

Conclusions: These findings support the potential role for cytokine panels in the discrimination of PDAC from patients with benign pancreatic diseases and warrant additional study.

Figure 1

Classification of patients with PDAC vs. healthy control individuals. A. Feature finding showing the percentage occurrence of cytokines in models following resampling of training data for PDAC vs. Healthy Control. B. Training set ROC Curves for PDAC vs. Healthy Control for a panel of IL-8, CA19-9 and IL-1b versus CA19-9 alone. C. Test set ROC curves for PDAC vs. Healthy Control for the panel of IL-8, CA19-9 and IL-1b versus CA19-9. D. The resampling performance of the panel of IL-8, CA19-9 and IL-1b versus CA19-9 for the classification of PDAC vs. Healthy Control. The accuracies, sensitivities and specificities of the panel compared to CA19-9 alone over 100 patient-balanced resamples is shown. The quantiles of the performance indicators are summarised as boxplots with individual points superimposed and the significance of Friedman test comparisons is indicated (***p <0.001).

Figure 2

Classification of patients with PDAC vs. patients with benign disease. (A) Feature finding results showing the percentage occurrence of cytokines in models following resampling of training data for PDAC vs. benign disease. (B) Training set ROC Curves for PDAC vs. benign disease for a panel of IL-8, IP-10, IL-6, PDGF and CA19-9 versus CA19-9 alone. (C) Test set ROC curves for PDAC vs. benign disease for the panel of IL-8, IP-10, IL-6, PDGF versus CA19-9. (D) The resampling performance of the panel of IL-8, IP-10, IL-6, PDGF and CA19-9 versus CA19-9 for the classification of PDAC vs. benign disease. The accuracies, sensitivities and specificities of the panel compared to CA19-9 alone over 100 patient-balanced resamples is shown. The quantiles of the performance indicators are summarised as boxplots with individual points superimposed and the significance of Friedman test comparisons is indicated (***p < < 0.001).

Figure 3

Classification of patients with PDAC vs patients with chronic pancreatitis. (A) Feature finding results showing the percentage occurrence of cytokines in models following resampling of training data for PDAC vs. chronic pancreatitis. (B) Training set ROC Curves for PDAC vs. chronic pancreatitis for a panel of IL-8, CA19-9, IL-6, IP-10 versus CA19-9 alone. (C) Test set ROC curves for PDAC vs. chronic pancreatitis for the panel of IL-8, CA19-9, IL-6 and IP-10 and CA19-9 alone. (D) The resampling performance of the panel of IL-8, CA19-9, IL-6, IP-10 versus CA19-9 for the classification of PDAC vs. chronic pancreatitis. The accuracies, sensitivities and specificities of the panel compared to CA19-9 alone over 100 patient-balanced resamples is shown. The quantiles of the performance indicators are summarised as boxplots with individual points superimposed and the significance of Friedman test comparisons is indicated (***p < < 0.001).

Figure 4

Classification of patients with PDAC in the presence of jaundice vs. patients with benign disease in the presence of jaundice. (A) Feature finding results showing the percentage occurrence of cytokines in models following resampling of training dataset for PDAC patients with high bilirubin vs. patients with benign disease and high bilirubin. (B) Training set ROC Curves for PDAC patients with high bilirubin vs. patients with benign disease and high bilirubin for a panel of IP-10, IL-8, IL-1b, PDGF and CA19-9 versus CA19-9 alone. (C) Test set ROC curves for PDAC patients with high bilirubin vs. patients with benign disease and high bilirubin for the panel of IP-10, IL-8, IL-1b, PDGF, CA19-9 compared to CA19-9 alone. (D) The resampling performance of the panel and CA19-9 for the classification of high bilirubin PDAC vs. high bilirubin benign disease. The accuracies, sensitivities and specificities of the panel IP-10, IL-8, IL-1b, PDGF, CA19-9 compared to CA19-9 alone over 100 patient balanced resamples is shown. The quantiles of the performance indicators are summarised as boxplots with individual points superimposed and the significance of Friedman test comparisons is indicated (***p < < 0.001).