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

Abstract

The aim was to build a calculator for personalized surveillance of BD-IPMNs. The interval time for surveillance of low-risk branch duct intraductal papillary mucinous neoplasms (BD-IPMNs) has not been established yet. 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. 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). 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.