In this study, we observed a significant correlation between the pattern of pancreatic atrophy and lateral extension in early pancreatic cancer. In particular, patients with FPPA exhibited significantly longer lateral cancer extensions than those without atrophy. All but one patient with FPPA showed intraductal cancer extension. To the best of our knowledge, this is the first study to evaluate the correspondence between atrophy on preoperative CT images and pathologic cancer extension in early pancreatic cancer in detail.

Nakahodo et al. reported similar results regarding the association between FPPA and cancer extension in patients with advanced pancreatic cancer. They concluded that PDAC with FPPA on “previous” preoperative imaging showed a higher rate of HG-PanIN positivity at the resection margins [14]. On the basis of their results, the authors concluded that PDAC with FPPA involves widespread HG-PanIN and requires a wide surgical margin for surgical excision. However, most of the cases in their study involved advanced cancers, whereas our study focused only on early pancreatic cancers, including CIS and microinvasive cancers. In addition, our study differs significantly in that it used FFPE sections to evaluate the extent of cancer extension and atrophy in detail, whereas the previous study evaluated the presence or absence of cancer only at the resection margins.

As suggested by the previous study and our study, long-axis extension of HG-PanIN is more likely to result in positive surgical margins for cancer and caution should be exercised in surgeries where curative resection is planned. This study included two cases of microinvasive carcinoma in which the resection margins were positive for HG-PanIN during surgery (cases 13 and 26 in Fig. 5). In one case, a 3-mm microinvasive carcinoma with FPPA on preoperative CT images was treated with distal pancreatectomy, and the resected macroscopic specimen showed an HG-PanIN extending over seven segments. At the time of surgery, the resection margin was positive for cancer and an additional resection was performed.

Preoperative prediction of long extension of CIS is useful for determining the extent of resection during surgery. Another case of microinvasive carcinoma requiring additional surgical resection is also presented (Supplementary Fig. 2; case 26 in Fig. 5). The patient underwent distal pancreatectomy with caudal pancreatic duct dilatation and UPA and was diagnosed as showing a 3-mm microinvasive carcinoma. A macroscopic specimen showed HG-PanIN extension over 14 sections and intraoperative resection margins positive for HG-PanIN, and additional resections were performed. Interestingly, in this case, images from 3 years prior to diagnosis were available, and localized atrophy of the pancreatic body was observed once. The atrophy progressed over the years and the imaging findings changed to UPA with dilatation of the main pancreatic duct. Of the 13 cancer cases with UPA, nine showed intraductal cancer extension, while four cases showed short extension with cancer in only one segment. These results suggest that CIS with UPA can be classified into two types: the lateral extension type, which may involve the transition from FPPA to UPA with intraductal cancer extension, and the short-segment type, which involves atrophy of the caudal pancreatic parenchyma due to severe stenosis of the main pancreatic duct. In other words, patients with FPPA at diagnosis or on previous images may show lateral extension of HG-PanIN. Even in cases of UPA, intraoperative resection margins require special attention and previous images, if available, should be reviewed.

In addition, in the lateral extension type of UPA, the cancer sometimes extends both upward (cephalically) and downward (caudally) from the main pancreatic duct stenosis. In the short segment type of UPA, imaging shows a significant change in the pancreatic duct diameter above and below the stenosis. Pathologically, there is a notable difference in adenocyte density in these regions, with greater glandular atrophy observed upstream. In these "short" cases, it is possible that the severe stenosis of the main pancreatic duct caused by the cancer has led to distal dilation and pancreatic parenchymal atrophy. When comparing the change in pancreatic duct diameter (caudal/cephalic) due to stenosis, the median increase was 2.0 times in the long extension type and 3.1 times in the short segment type. Although the change was greater in the latter, the difference was not statistically significant (P = 0.076). In cases of UPA, relatively mild dilation of the distal main pancreatic duct may result in further extension of the cancer, warranting greater caution in the extent of resection.

FPPA is a characteristic finding of intraepithelial carcinoma; pathologically, these cases show desquamation of the adenohypophysis with fibrosis and fat replacement [8]. The mechanism underlying these findings is unknown but it has been suggested to be attributable to localized pancreatitis caused by obstruction of the branching pancreatic ducts. FPPA was found not only around HG-PanIN but also around LG-PanIN. The cases reviewed in this study included eight cases of LG-PanIN. These included one case in which localized pancreatic atrophy was observed even with LG-PanIN, but this case showed "long" lesion extension, with the equivalent of LG-PanIN2 in the old protocol extended to four sections. The presence of FPPA may indicate a long PanIN lesion.

Although the prognosis after surgical resection of pancreatic cancer has improved, the incidence of remnant pancreatic cancer after surgery has also increased. The incidence of remnant pancreatic cancer after surgery is said to range from 0.7 to 26.7% [21]. The incidence of remnant pancreatic cancer has been shown to be higher in early-stage pancreatic cancer [3, 5] and Miyasaka et al. reported that the cumulative incidence of remnant pancreatic cancer was comparable between early- and advanced-stage groups [22]. In the present study, among the 32 cases included, three showed remnant pancreatic cancer during the observation period. These included two cases of CIS and one case of microinvasive carcinoma, with HG-PanIN extension distances of 20, 20, and 25 mm, all of which showed atrophy (one UPA and two FPPA). In contrast, no recurrence was observed in patients without atrophy or intraductal cancer extension. These results suggest that in cases showing long intraductal carcinoma extension with FPPA, precancerous lesions, including LG-PanIN, may remain latent in the residual pancreas after resection and indicate a high risk of recurrence. Therefore, even in early-stage pancreatic cancer, the occurrence of remnant pancreatic cancer after surgery should be noted. A longer PanIN suggests a higher rate of postoperative residual pancreatic recurrence, which is meaningful for evaluation. Thus, in cases of FPPA, special attention may be needed not only for the extent of resection during surgery, but also for postoperative recurrence.

In summary, in cases of FPPA, the cancer is thought to be located in the area of atrophy and often extends beyond it, requiring the inclusion of the atrophic area in the surgical resection to ensure clear margins and minimize recurrence in the remnant pancreas. In cases of UPA, the cancer is centered on the main pancreatic duct stenosis, and the type of intraductal extension can be categorized into “long” extension in the longitudinal axis and “short” segment confined to the main pancreatic duct stenosis. As shown in Supplementary Fig. 2, cases with UPA at diagnosis but FPPA on previous images may show lateral cancer extension and should be referred if previous images are available. The surgical technique is determined by the location of the main pancreatic duct stenosis, but in cases where the pathological diagnosis shows “long” extension, attention should also be paid to recurrence in the remnant pancreas.

On the other hand, in our study, there were six cases of short-segment CIS without atrophy. In a study by M. Ikeda et al., it was reported that non-invasive cancer parts (PanIN-3 lesions) continuously spreading from the invasive cancer area were histologically categorized into three types: flat (F), low papillary (LP), and mixed (flat and low papillary) [13]. The LP type tends to spread horizontally along the pancreatic ducts, while the F type tends to spread vertically with minimal lateral spread along the ducts. In our study, the short-segment type without atrophy included three cases of the F type and three cases of the LP or mixed type. On the other hand, all cases of the long extension type were of the LP or mixed type. This suggests that the short-segment type without atrophy can exhibit two patterns of spread: horizontal and vertical. However, the genetic and developmental differences between the two types of pancreatic cancer, F type and LP type, remain to be elucidated. Further research is required on this issue.

Our study had several limitations. First, this was a retrospective study, and the sample size was small, with a limited number of cases from a limited number of institutions. However, in general, cases of CIS are extremely rare and this study included detailed pathological assessments of a sufficiently large number of cases. Second, the studies were limited to surgical cases to exclude pathologically undiagnosed cases. Additionally, because CIS was examined on the basis of the presence of main pancreatic duct stricture, only cases showing changes in the main pancreatic duct were included. Third, there are discrepancies in the preoperative evaluation. Not all cases underwent both ERCP and MRCP, and the results of the two tests are not entirely consistent. The assessment of changes in the diameter of the main pancreatic duct was completed using ERP findings in cases where ERCP was performed and MRCP was used instead in the three cases where ERCP was not performed. Of the 27 cases where both tests were performed, 92.6% (25/27) showed similar pancreatic duct stenosis on both tests. The concordance of findings between ERP and MRCP for each case is detailed in Supplementary Table 2. Not all MRI images were captured using the same model. Of the 30 cases in which MRI was performed, 15 were obtained using 3 T imaging, while the remaining cases were captured with 1.5 T imaging. However, findings of diffusion reduction on MRI and pancreatic duct stenosis on MRCP could also be identified under 1.5 T imaging conditions. In the comparison between MRCP and ERP for main pancreatic duct stenosis described above, the two cases in which the main pancreatic duct stenosis was difficult to assess on MRCP were imaged at 3 T.

In conclusion, HG-PanIN may extend into the pancreatic duct in CIS and microinvasive carcinomas presenting with FPPA and UPA. In such cases, caution should be exercised while determining the extent of surgical resection.