Performance of MRI-based surveillance for high-risk individuals for pancreatic cancer

Abstract

Purpose

Surveillance is recommended in high-risk individuals (HRIs) for pancreatic cancer (PC), as it seems effective in detecting malignant/premalignant lesions. Most surveillance programs are based on a combination of MRI and endoscopic ultrasound.

This study aimed to evaluate the incidence of relevant imaging findings (PC and lesions with high-grade dysplasia) and the performance of MRI-based surveillance in a single-centre HRIs cohort.

Methods

Single-centre, retrospective study on 274 prospectively collected HRIs (2002–2023). Surveillance was performed exclusively with MRI from 2010; MRI and/or CT in 2002–2009. MRI findings, demographic, clinical, and histopathological data were recorded. Diagnostic accuracy measures for the surveillance program were calculated.

Results

The median follow-up was 5.8 years. Focal lesions were detected in 59 % of the patients: cysts in 56 % (152/274), cystic and solid lesions in 3 % (8/274), and solid lesions in 1 % (2/274). Four patients (1 %, 4/274) had suspicious cysts (3 resected; no high-grade dysplasia at histopathology). Six patients (2 %; 6/274) had PC: two in stage I (0.7 %), one in stage II (0.4 %), three in stage IV (1.1 %). Three out of six patients had liver metastases at diagnosis. Surveillance diagnostic accuracy measures: sensitivity 67 % (95 %CI 22–96 %), specificity 98 % (95 %CI 96–99 %), positive predictive value (PPV) 44 % (95 %CI 14–79 %), negative predictive value 99 % (95 %CI 97–99 %).

Conclusions

The incidence of PC in HRIs was low (0.32 per 100-person-years), with similar proportion of stage I and IV PC. No lesions with high-grade dysplasia were identified. Sensitivity and PPV were low, suggesting that early diagnosis of PC in MRI surveillance setting is challenging.

Keywords (Mesh terms)

Pancreas, Pancreatic Cancer, Early Diagnosis, Surveillance, MRI, Incidence.