MRI follow-up for pancreatic intraductal papillary mucinous neoplasm: an ultrashort versus long protocol

One major finding from this study is that USP is comparable to longer protocols for the radiological assessment in patients under surveillance for IPMN. When reviewing interobserver agreement of cystic mural nodules, MPD mural nodules, and solid pancreatic tumors, readers reached the same conclusions using USP compared to S-LP in 94.9%, 99.1%, and 99.1%, respectively, of cases. Our study yielded similar results from a previous study by Pozzi-Mucelli et al. [10], where agreement for cystic and MPD mural nodules was 93% and 98%, respectively. Among findings for genuine cystic mural nodules, USP performed as well as S-LP. All matching cystic mural nodules were identified in each case read. Reader 1 suspected a mural nodule in two more cases when using USP than when using S-LP, although reader 2 suspected one less case when using USP than when using S-LP. We also found that the intra-observer agreement for cystic mural nodules was strong for reader 1 and moderate for reader 2, and that the interobserver agreement was moderate for both USP and S-LP. These results indicate that USP yields nearly identical results compared to S-LP and can be used in clinical practice.

When examining WF or HRS, the readers reached similar conclusions when evaluating USP and S-LP in 92.4% of cases. The intra-observer agreement was nearly perfect for reader 1 and was moderate for reader 2, while the interobserver agreement was strong for USP and moderate for S-LP. Reader 1 identified one case with abrupt change in the MPD caliber with distal parenchymal atrophy only using USP, not with S-LP. But in one case reader 1 analyzed different cysts as the largest one by using USP and S-LP leading to different measurements (8 mm and 31 mm). Reader 2’s measurements exhibited mostly clinically insignificant variations in cyst size and MPD measurements. Yet, in one patient, the MPD diameter was measured differently (2–5 mm) and an abrupt change in the MPD caliber was not identified using USP. Reader 1 also did not identify this finding using S-LP, indicating that this specific case was particularly difficult to evaluate clinically. For this specific patient, a follow-up examination at 2 years and 7 months after the initial imaging was analyzed, which indicated stable disease. Because reader 2 identified an insignificant variation between MPD and cyst measurements in a total of five cases and most of the additional cystic mural nodule observations were inconclusive when using USP and S-LP, we consider these findings insignificant. The average difference, then, in the percentage of WF/HRS detected between methods fell below 3% for both readers. In addition, the intra- and interobserver agreement results were comparable regardless of the protocol used and fell within a normal variation for this kind of multivariate image analysis. These results indicate that USP and S-LP are comparable for detecting WF or HRS in IPMN patients.

For the size-related parameters, specifically the mean cyst diameter and MPD, the overall difference between the two protocols was 0.19 mm and 0.08 mm, respectively. These data lie within the defined noninferiority limits. Thus, both S-LP and USP appear to provide noninferior results appropriate for clinical use. Therefore, dimensional changes in both lesion and MPD during surveillance can be reliably detected using USP. The noninferiority limits for a cyst were set to ± 10 mm [14, 15]. Dunn et al. demonstrated that a significant and frequent interobserver variability exists in measuring pancreatic cystic lesions [14]. In our study, the SD for cyst size was only 0.27 mm while it was 3.27 mm in the study by Dunn et al. Also, no difference between measurements exceeded 50% in our study when in their study some measurements exceeded 50% [14]. In the study by Maimone et al., the interquartile range (IQR) for cyst size was 3.5 mm while it was only 2.5 mm in our study [15]. Furthermore, the noninferiority limits for MPD were set to ± 2 mm given that we documented an MPD width within a 1-mm accuracy, leading to rounding from 1.4 to 1 mm and from 2.5 to 3 mm, indicating that a 1.1-mm difference led to a 2-mm difference.

USP should not replace LP for the primary characterization of IPMNs. USP could be used for the surveillance of BD-IPMN patients who have no WF or HRS characteristics and have been carefully evaluated. If WF or HRS characteristics develop during surveillance, performing an additional LP becomes necessary. This study warrants further research to evaluate if USP can also detect solid tumors based on the knowledge that IPMN patients carry a higher risk for cancer elsewhere in the pancreas [16]. We argue that the primary diagnosis should rely on LP. Then, if a patient exhibits no signs of WF or HRS, follow-up could rely on USP.

To calculate the cost and time reduction when substituting S-LP with USP, we used two approaches to demonstrate the magnitude of the effect. Given that the estimated USP cost is about 77% that of SP and 39% that of LP, the total cost reduction associated with using USP seems obvious. In addition, S-LP is more time-consuming than USP. We calculated that USP would occupy an MRI suite for 15 min. The reductions in cost and time resulting from substituting S-LP with USP would allow for a greater number of patient scans. In addition, case reading is faster with fewer sequences.

In previous study by Pozzi-Mucelli et al., one limitation resulted from the different slice thicknesses from differences in MRI protocols used in their study [10]. We avoided this problem because we used the same MRI machine. Our study was standardized by using examinations performed with 1.5 T. The measured qualities are likely equal both in 1.5 T and in 3.0 T machines, but further research is needed. However, our study carries some limitations. First, the primary aim of our study was to determine if readers could report similar information with USP and S-LP in the same patient, rather than analyzing the sensitivity or specificity of these protocols. Second, we conducted a retrospective study that included patients in an IPMN surveillance program for whom a histopathological confirmation was unavailable [17]. Approximately 30% of our patients had WF/HRS but only 14% of the population underwent sample taking. Our patient population consisted of elderly patients which may explain that they were frequently treated conservatively. Third, almost 29% of patients were excluded because of the insufficient diagnostic imaging quality. Especially in elderly patients, the breathing and movement artifacts are an everyday problem. In everyday clinical practice, the evaluation of the image quality is an important step of the image interpretation, and in case of non-diagnostic image quality, renewal of the examination or other imaging technique (e.g., CT) should be considered. We hope that the shorter USP examination time (sequence-based time is only 7 min) will reduce this problem. Finally, the calculated cost reduction from substituting S-LP with USP relied on the cost difference between these three protocols estimated at Helsinki University Hospital, and, therefore, variations in prices may exist between hospitals and countries.

In conclusion, an ultrashort MRI protocol is suitable for IPMN surveillance, which could relieve healthcare system burdens without losing any essential information. An ultrashort protocol should be considered as an alternative for IPMN surveillance when a patient does not have WF or HRS. The goal is to identify patients who develop WF or HRS during surveillance. While possible false-positive findings may occur, it is more important to have a low threshold for suspected WF/HRS than to not suspect them at all.

留言 (0)

沒有登入
gif