TO THE EDITOR
We read with great interest the observational study by Hopley et al[1], reporting the outcomes of a 17-year single-centre experience of branch duct intraductal papillary mucinous neoplasm (BD-IPMN) surveillance at a high-volume supra-regional pancreatic referral unit in Liverpool, United Kingdom. The study addresses one of the most clinically debated questions in BD-IPMN management: When, and under what conditions, can surveillance safely be de-escalated or discontinued? The Authors are to be commended for the size of their cohort, the length of follow-up, and the pragmatic, real-world nature of their analysis. Nevertheless, we wish to raise several points of discussion, with a particular focus on the radiological point of view of the study, which we believe needs further consideration.
The study reports that cyst size ≥ 30 mm emerged as one of the two most predictive markers of high-grade dysplasia or malignancy, alongside serum carbohydrate antigen 19-9 (CA19-9) ≥ 43 KU/L, with an area under the curve of 0.76.
Even accepting these findings as robust, some concerns remain regarding the radiological methodology. The Authors reported that the final diagnosis was obtained using magnetic resonance cholangiopancreatography (MRCP) sequences, supported by discussion in multidisciplinary team (MDT) meetings, while communication with the main pancreatic duct (MPD) was not mandatory. However, from a technical radiological point of view, the Authors did not sufficiently describe the imaging protocols, magnetic resonance (MR) field strength(s), or the presence of a radiologist in the MDT (general or subspecialised in pancreato-biliary imaging). These aspects can have an impact on the detection and measurement of BD-IPMN[2,3].
MEASUREMENT REPRODUCIBILITY
It is well known that cyst size measurement has moderate inter- and intra-observer variability. This is not an issue per se, but becomes significant when a specific threshold is used to determine the best management for patients[4,5]. The Authors of the present study used a cut-off value obtained from a longitudinal dataset covering 17 years. However, during the study period, technology has dramatically evolved. In this setting, the Authors did not provide sufficient disclosure on how measurements were standardised, which sequence(s) were used, and how many readers performed the measurements. In fact, it has been reported in the literature that the sequence on which intraductal papillary mucinous neoplasm (IPMN) is measured has an important impact on absolute values and reliability analysis between readers: Axial T1-weighted images during contrast media administration and MRCP can overestimate lesion size in comparison to pathology[6]. Moreover, abbreviated MR protocols can be used in clinical practice for surveillance of known IPMN, only if standardised[7]. When standardisation is lacking, the size threshold should be used with caution.
From a practical standpoint, we believe that all studies focused on these approaches should specify concrete imaging standards to reduce this heterogeneity. These should include: (1) A minimum field strength of 1.5 T with a dedicated pancreatic MRCP protocol, comprising thin-slice (≤ 3 mm) three-dimensional MRCP sequences together with axial T2-weighted and, eventually, post-contrast T1-weighted acquisitions[8]; (2) A standardised measurement method, preferably performed on the sequence with the highest reproducibility (three-dimensional MRCP or axial T2-weighted images) rather than post-contrast T1-weighted images, which tend to overestimate lesion size[6]; (3) Structured, template-based reporting that explicitly records cyst location, size, number, MPD diameter, and the presence or absence of each individual worrisome feature (WF) and high-risk feature (HRF), rather than a narrative summary[9]; and (4) Interpretation by, or in conjunction with, a radiologist with specific expertise in pancreatico-biliary imaging[2,3]. We believe that these elements could substantially improve the reproducibility and external validity of de-escalation criteria such as those proposed by the Authors. These aspects are further compounded by the absence of a clear description of how and when WF and HRF were assessed over time. The study appropriately references international guidelines and the Liverpool IPMN surveillance pathway (LISP), but the radiological criteria for WF and HRF evolved considerably across the study period, particularly with the publication of successive iterations of the Fukuoka, European, and Kyoto guidelines[10]. It is not entirely clear how the Authors managed these evolving criteria in their longitudinal cohort, or whether patients enrolled under earlier versions of the pathway were retrospectively reassessed under updated criteria. Features such as MPD diameter, presence of mural nodules, and cyst wall enhancement have different sensitivity and specificity depending on the imaging modality and technique used, and their assessment at MRI vs computed tomography (CT) or endoscopic ultrasound (EUS) is not equivalent[11]. Some of these limitations are specific to how the present study was reported, such as the incomplete description of field strength and number of readers. Others are broader problems that go beyond this study, such as the fact that the definitions of WF and HRF have changed across successive guideline updates, which affects how de-escalation criteria can be applied across different institutions. This second issue is not unique to this cohort, and addressing it will need consensus efforts well beyond any single study.
On the subject of EUS, the study provides interesting and clinically relevant data regarding the relationship between number of EUS examinations and time to surgery. Of the 153 patients who underwent surgery, 127 (83%) had undergone two or fewer EUS examinations, with a markedly shorter median time to surgery compared to those who had more than two EUS. The Authors interpret this, reasonably, as supporting the notion that actionable findings tend to emerge early in surveillance. However, from a radiological perspective, it is worth noting that EUS and MRI/MRCP provide complementary rather than equivalent information in the assessment of BD-IPMN. EUS offers superior spatial resolution for the detection of mural nodules and assessment of cyst wall features, while MRCP provides better global anatomical overview and is less operator-dependent. Indeed, a multicentre study comparing MRI-MRCP and EUS in BD-IPMN patients demonstrated only minimal concordance between the two modalities for the detection of HRF and WF, with the greatest disagreement observed for lesions in the proximal pancreas[11]. Regarding mural nodule detection specifically, contrast-enhanced EUS has been shown to reach sensitivity exceeding 90%, compared to lower rates for conventional MRI and CT[12,13]. These observations should not be interpreted as advocating for the routine addition of EUS in all patients with low-risk BD-IPMN meeting de-escalation criteria. Rather, EUS is likely to be most useful in a targeted subset of cases, such as when MRI/MRCP findings are equivocal, when cyst morphology changes on serial imaging, when a mural nodule is suspected but not clearly characterised on cross-sectional imaging, or when serum CA19-9 rises during surveillance[8,14]. The decision to proceed to EUS in this cohort was governed by LISP criteria and MDT consensus, yet no detailed breakdown is provided of what radiological findings on cross-sectional imaging prompted EUS referral in each case. Understanding this would be valuable, as it would clarify the relative contribution of MRI vs EUS in identifying the features that led to surgical approach.
Furthermore, the study’s conclusion that surveillance may be de-escalated after two years in patients with cyst size < 30 mm and serum CA19-9 < 43 KU/L, and that discharge could be considered at five years, raises important questions about the imaging modality and frequency that should represent de-escalated surveillance. The Authors acknowledge that younger patients will require longer follow-up, but they do not propose a specific imaging algorithm.
In particular, MRI, CT, and EUS should be considered complementary to one another, as they differ in cost, radiation exposure, and diagnostic accuracy[14,15]. The Authors proposed a de-escalation strategy; however, they do not clearly indicate the best imaging modality and follow-up interval, which risks undermining a standardised process. These concerns are further compounded by the selection criteria used for the study design, which included only patients without WF or HRF. This is acceptable given the primary aim of evaluating patients with “presumed innocent” BD-IPMN. However, the radiological methodology can be applied only to this low-risk population and cannot be widely applied to all patients encountered in clinical practice.
The last radiological consideration concerns the role of emerging imaging techniques that arose during the period of this study, particularly diffusion-weighted MRI, contrast-enhanced EUS, and artificial intelligence-assisted image analysis[15-17]. None of these were incorporated into the study, which is understandable given its retrospective nature. However, it is important to acknowledge that the diagnostic landscape for BD-IPMN has evolved rapidly over time and continues to do so. Further prospective studies with a standardised imaging approach should be considered mandatory to validate the de-escalation process proposed by the Authors, and would more clearly assist clinicians and radiologists in the management of these patients.
