Table 3. Diagnostic performance of preliminary models to predict malignant IPMN pathology in the study cohort.

Modela	Variables included	AUC (95% CI)	Sensitivity	Specificity	Positive predictive value	Negative predictive value
Demographic and clinical data	Age at diagnosis, gender, presence of symptoms	0.73 (0.56-0.89)	0.83	0.55	0.63	0.79
Standard imaging data	High risk stigmata	0.84 (0.72-0.96)	0.83	0.85	0.83	0.85
Genomic data	5-miRNA genomic classifier (MGC)	0.83 (0.69-0.97)	0.78	0.80	0.78	0.80
Standard imaging + genomic data	High risk stigmata, MGC	0.95 (0.88-1.00)	0.94	0.90	0.89	0.95
Standard imaging data	Worrisome features	0.54 (0.38-0.69)	0.72	0.35	0.50	0.58
Standard imaging + genomic data	Worrisome features, MGC	0.83 (0.69-0.97)	0.83	0.80	0.79	0.84
Radiomic data	Radiomic PC1 classifier	0.77 (0.61-0.93)	0.83	0.74	0.75	0.82
Radiomic + genomic data	Radiomic PC1 classifier + MGC	0.92 (0.83-1.00)	0.83	0.89	0.88	0.85
Standard imaging+ radiomic+ genomic data	Worrisome features, Radiomic PC1 classifier + MGC	0.93 (0.85-1.00)	0.89	0.89	0.89	0.89

^a Full models include 20 benign and 18 malignant IPMNs.