Personalized Surveillance Intervals for Intraductal Papillary Mucinous Neoplasm (IPMN): Multicenter Study Using Parametric Models

Abstract

Objective: The aim was to build a calculator for personalized surveillance of BD-IPMNs.

Summary background data: The interval time for surveillance of low-risk branch duct intraductal papillary mucinous neoplasms (BD-IPMNs) has not been established yet.

Methods: The study included an international cohort of BD-IPMNs without worrisome features (WFs) or high-risk stigmata (HRS). IPMN evolution was defined as the occurrence of HRS or WFs. The derivation cohort comprised 60% of patients. The validation group comprised the remaining patients. A parametric survival model was developed in the derivation cohort using Akaike (AIC) and Bayesian (BIC) information criteria and c-index. A "k-fold" validation was used to measure the covariate effect on the accelerated failure time. Two models ("standard" and "conservative") were built and validated using the second cohort.

Results: The derivation and validation cohorts included 1,992 and 1,119 BD-IPMNs. The lognormal distribution best fitted the derivation cohort (AIC=2673; BIC=2718). The pooled c-index was 0.689 (0.668 to 0.718, 95%CI). The factors reducing the time needed for IPMN evolution were age [- 2% (-1% to -3%) for each year] and cyst size [-2% (0% to -3%); for each mm]. The "conservative" model, called PANORAMA, was the only one that correctly classified the validation cohort (c-index 0.712 vs. 0.696; P=0.072).

Conclusion and relevance: The development of WF and HRS is influenced by the patient's age and cyst size. After a prudential first control at six months, repeating a semestral/annual follow-up in this time frame could be too tight.