Risk of the Watch-and-Wait Concept in Surgical Treatment of Intraductal Papillary Mucinous Neoplasm

Key Points

Question  Is it safe to follow the natural history of intraductal papillary mucinous neoplasms (IPMNs) based on the actual recommendation?

Findings  In this cohort study of 1439 resected IPMNs focusing on the time of resection, one-third of IPMNs reached the cancer stage before resection.

Meaning  As per this analysis, a watch-and-wait policy should be applied with caution.

Importance  The natural history of intraductal papillary mucinous neoplasms (IPMNs) remains uncertain. The inconsistencies among published guidelines preclude accurate decision-making. The outcomes and potential risks of a conservative watch-and-wait approach vs a surgical approach must be compared.

Objective  To provide an overview of the surgical management of IPMNs, focusing on the time of resection.

Design, Setting, and Participants  This cohort study was conducted in a single referral center; all patients with pathologically proven IPMN who received a pancreatic resection at the institution between October 2001 and December 2019 were analyzed. Preoperatively obtained images and the medical history were scrutinized for signs of progression and/or malignant features. The timeliness of resection was stratified into too early (adenoma and low-grade dysplasia), timely (intermediate-grade dysplasia and in situ carcinoma), and too late (invasive cancer). The perioperative characteristics and outcomes were compared between these groups.

Exposures  Timeliness of resection according to the final pathological findings.

Main Outcomes and Measures  The risk of malignant transformation at the final pathology.

Results  Of 1439 patients, 438 (30.4%) were assigned to the too early group, 504 (35.1%) to the timely group, and 497 (34.5%) to the too late group. Radiological criteria for malignant conditions were detected in 53 of 382 patients (13.9%), 149 of 432 patients (34.5%), and 341 of 385 patients (88.6%) in the too early, timely, and too late groups, respectively (P < .001). Patients in the too early group underwent more parenchyma-sparing resections (too early group, 123 of 438 [28.1%]; timely group, 40 of 504 [7.9%]; too late group, 5 of 497 [1.0%]; P < .001), while morbidity (too early group, 112 of 438 [25.6%]; timely group, 117 of 504 [23.2%]; too late group, 158 of 497 [31.8%]; P = .002) and mortality (too early group, 4 patients [0.9%]; timely, 4 [0.8%]; too late, 13 [2.6%]; P = .03) were highest in the too late group. Of the 497 patients in the too late group, 124 (24.9%) had a previous history of watch-and-wait care.

Conclusions and Relevance  Until the biology and progression patterns of IPMN are clarified and accurate guidelines established, a watch-and-wait policy should be applied with caution, especially in IPMN bearing a main-duct component. One-third of IPMNs reach the cancer stage before resection. At specialized referral centers, the risks of surgical morbidity and mortality are justifiable.

Intraductal papillary mucinous neoplasms (IPMNs) are considered precursors of pancreatic cancer (pancreatic ductal adenocarcinoma [PDAC]), and timely resection before malignant transformation takes place could improve survival. However, the natural history of potential malignant transformation is still not well understood, and the clinical management of IPMN therefore represents a particular challenge.^1,2 The risks and benefits of watch-and-wait approaches vs aggressive early surgery must be analyzed to enable informed and evidence-based decision-making. In particular, the risk of resecting when malignant transformation has already taken place needs to be balanced against the risk of severe surgical complications in excessively early resection.³ In recent years, many specialized referral centers have progressively restricted the indications for resection, claiming that a watch-and-wait policy avoids unnecessary surgery in subgroups at low risk of malignant transformation. Various guidelines that have been published are far from consistent with regard to risk factors.^4-6 Classic radiological features, tumor markers, and clinical signs (eg, abdominal pain, jaundice, recent diagnosis of diabetes), whether alone or in various combinations, have proved unacceptably inaccurate in estimating the risk of malignant disease.⁷ Moreover, the available evidence on this topic is mostly derived from retrospective surgical series consisting of highly heterogeneous patient groups.⁸ Furthermore, different malignant transformation rates have been described for different clinical subtypes (main-duct [MD], mixed-type [MT], or branch-duct [BD] IPMN), histological subtypes,⁹ grades of dysplasia,¹⁰ and sequence-variation profiles.¹¹ Consequently, the current clinical guidelines are insufficient to allow clear-cut decision-making in this context.

Moreover, the data concerning time to progression of IPMN, as reported in observational studies with limited follow-up duration, are inconsistent.⁸ The proportion of patients who receive surgery too late seems to have been underestimated. The 2 largest studies from Europe and the US providing data on IPMNs monitored for more than 10 years suggest that even small, quiescent BD lesions can harbor a nonnegligible risk of malignant disease¹² and the development of cancer in patients with IPMN is just a matter of time.¹³ Finally, a survey among more than 300 Dutch gastroenterologists reported underestimation of the risk of malignant disease.¹⁴

The best management strategy for IPMN has therefore not yet been established. This study sets out to provide an overview of IPMN treatment and outcomes at a high-volume referral center over 19 years, focusing on the association between the timing of resection and the risk of malignant transformation.

After institutional review board approval by the Ethics Committee of the Medical Faculty, University of Heidelberg, data on all consecutive patients who underwent pancreatic resection for IPMN at the Department of General, Visceral, and Transplantation Surgery at Heidelberg University Hospital in Heidelberg, Germany, between October 2001 and December 2019, were extracted from a prospectively maintained database. Patient consent for data acquisition and analysis was documented in the individual treatment contract on admission.

All preoperative images were reviewed retrospectively in blinded fashion by 2 radiologists with expertise in pancreatic diseases (E.K. and P.M.). Each IPMN was clinically classified as MD, MT, or BD. The size and number of cysts larger than 10 mm were recorded. The main pancreatic duct (MPD) was considered dilated when it measured 5 mm or more. Branch-duct IPMN was defined as 1 or more pancreatic cysts in communication with the duct system, in the absence of MPD dilation. An MPD diameter of 5 mm or more was a criterion for MD IPMN, independently of the presence of 1 or more BD cysts. Mixed-type IPMN was defined as 1 or more BD lesions with a MPD diameter of more than 5 mm. The radiological signs considered to suggest IPMN progression toward malignant disease were enhancing or nonenhancing mural nodules, cyst wall thickness, and solid components, as well as an MPD dilation of 5 mm or more.

According to the degree of cytoarchitectural atypia, IPMNs were classified into low-grade (LGD), intermediate-grade (IGD), and high-grade (HGD) dysplasia.¹⁵ As per the World Health Organization’s 2010 classification,¹⁶ no differentiation was made between HGD IPMN and in situ carcinoma, because neoplastic cellular and glandular atypia (including loss of nuclear polarization) can be detected in both cases but remains confined to the mucosal layer. Conversely, invasive cancer was diagnosed when atypia extended beyond the mucosal layer. Low-grade IPMN were diagnosed as adenoma and included in the too early group; IGD IPMN were diagnosed as borderline and together with in situ carcinoma, composed the timely group. Invasive cancer, including colloid-type carcinoma, PDAC arising from IPMN, and PDAC concomitant with IPMN, belonged to the too late group. To simplify terminology, all of these malignant entities are referred to as PDAC.

Data management and statistical analysis were carried out using SAS software release 9.4 (SAS Institute) and IBM SPSS software, version 25 (IBM). Quantitative variables are presented as medians and interquartile ranges (IQRs), and the Kruskal-Wallis test was used to compare continuous parameters among the groups. For categorical parameters, absolute numbers are shown and percentages compared among the groups using Fisher exact tests. Two-sided P values were computed, and the differences were considered statistically significant at P < .05. The Bonferroni method was used for multiple comparisons bias correction.

A total of 1439 patients (707 men [49.1%]; median [IQR] age of 67.5 [60.6-73.2] years) underwent pancreatic surgery for IPMN during the study period. The distribution of IPMN by type and degree of dysplasia over time is shown in Figure 1. Of 1199 patients with available radiological imaging, 543 patients [45.2%] showed signs suspicious for a malignant condition, according to preoperative radiological assessment. Overall, radiological criteria for malignant conditions were detected in 53 of 382 patients (13.9%), 149 of 432 patients (34.5%), and 341 of 386 patients (88.6%) in the too early, timely, and too late groups, respectively (P < .001). Compared with those without worrisome radiological features, those patients were older (median [IQR] age: without worrisome features, 66.6 [60.5-72.7] years; with worrisome features, 68.3 [60.6-73.9] years; P = .01) and more likely to be male (255 of 656 [38.9%]; 286 of 543 [52.7%]; P < .001), diabetic (126 [19.2%]; 132 [24.3%]; P = .03), or displaying elevated plasma levels of carbohydrate antigen 19-9 (CA 19-9) (median [IQR] levels, 11.0 [6.6-22.0] U/mL; 30.0 [10.9-234.0] U/mL; P < .001) and carcinoembryonic antigen (median [IQR] levels, 1.0 [1.0-2.0] ng/mL; 2.0 [1.0-3.0] ng/mL; P < .001) (eTable in the Supplement). When compared with the final pathology of in situ carcinoma or PDAC, the subjective radiological presumption of malignant transformations showed a sensitivity of 77.8% and a specificity of 50.0%, with an area under the receiver operating characteristic curve of 0.792 (95% CI, 0.761-0.823).

Final histopathologic assessment revealed LGD-IPMN or adenoma in 438 of 1439 resections (30.4%), which were consequently classified as too early. On the other hand, 497 of 1439 resected lesions (34.5%) turned out to be invasive PDAC and were thus classified as too late (Figure 2). The remaining 35.1% of resections were classified as timely, since the lesions were diagnosed as either borderline IPMN (327 [22.7%]) or in situ carcinoma (177 [12.3%]).

A comparison of baseline characteristics of the 3 groups is provided in Table 1. Even though the medical history was unremarkable in most patients, a previously reported presence of pancreatic cysts was less frequent in cases of PDAC (too early group, 144 of 438 [32.9%]; the timely group, 186 of 504 [36.9%]; too late group, 124 of 497 [24.9%]; P < .001). Of 385 patients in whom pathological examination revealed PDAC (too late group), 44 (11.4%) had shown no radiological signs of malignant disease. On final pathology, BD-IPMN were significantly more prevalent in the too early group (too early group, 267 of 438 [61.0%]; timely group, 148 of 504 [29.4%]; too late group, 81 of 497 [16.3%]; P < .001), while MT-IPMN were more frequently associated with PDAC (the too late group; 253 of 497 [50.9%]) and borderline or high-grade dysplasia (the timely group; 289 of 504 [57.3%]) than the too early group (129 of 438 [29.5%]; P < .001).

Furthermore, MPD diameter showed a significant correlation with regard to absolute diameter as well as the proportion of patients showing a dilation of 5 mm or more. In the too early group, the median (IQR) MPD size was 3.0 (2.0-5.0) mm, which increased in the timely and too late groups to 5.0 (3.0-9.0) mm and 8.0 (5.0-11.0) mm, respectively.

Overall, 454 (31.5%) of the 1439 patients had a medical history of pancreatic cysts and were treated with a watch-and-wait strategy (Table 1). Within the too late group, 124 of 497 patients (24.9%) had been subject to a watch-and-wait policy. Almost half of those (49 of 124 [39.5%]) had been diagnosed more than 1 year before surgery. Patients treated via watching and waiting from the too-late group had a median (IQR) age of 69 (62-75) years, the male-to female ratio was 0.70, and the prevalence of diabetes was 28.2% (35 of 124 patients). At preoperative evaluation, 51 of 124 patients (41.1%) reported nonspecific symptoms, such as abdominal pain or nausea.

Most patients with PDAC received a standard resection (492 of 497 [99.0%]), while those with a final diagnosis of LGD or adenoma were more likely to undergo enucleation or atypical partial resection (too early group, 123 [28.1%]; timely group, 40 [7.9%]; too late group, 5 [1.0%]) (Table 2). The frequency of surgical complications was comparable in all 3 groups (too early group, 187 of 438 [42.8%]; timely group, 187 of 504 [37.2%]; too late group, 214 of 497 [43.0%]; P = .24). Medical complications, however, occurred more often in the too late group than in the other groups (too early group, 112 [25.6%]; timely group, 117 [23.2%]; too late group, 158 [31.8%]; P = .002).

Postoperative mortality was highest in the too late group with an observed 30-day mortality rate of 2.6% (13 of 497 patients) compared with 0.9% (4 of 438) in the too early group and 0.8% (4 of 504 patients) in the timely group (P = .03). Likewise, 90-day mortality rates were comparable in the 3 groups (Table 2).

At final pathological examinations, BD-IPMN were significantly more prevalent in the too early group (too early group, 267 of 438 [61.0%]; timely group, 148 of 504 [29.4%]; too late group, 81 of 497 [16.3%]; P < .001), while MD-IPMN and MT-IPMN were more frequently associated with PDAC (299 of 497 [60.2%]) and borderline or HGD tumors (343 of 504 [68.0%]) than LGD or adenoma IPMN (156 of 438 [35.7%]; P < .001).

Figure 3 recapitulates clinical decision-making and the indications for surgical resection in patients from the too early group. Of 438 patients, 118 (26.9%) were radiologically diagnosed with MD-IPMN or MT-IPMN. Of the remaining 320 patients, 248 presented with any additional feature, including other radiological suspicion of malignant disease (a worrisome feature), increased plasma CA 19-9 levels, diabetes, abdominal pain, or pancreatitis.

Although discrepancies among published series² mean that the natural course of IPMN is still not wholly understood, the association of IPMN with pancreatic cancer is evident. Intraductal papillary mucinous neoplasms are not only precursors of pancreatic cancer arising from the cyst itself¹⁷; they also increase the risk of a separate PDAC elsewhere in the gland.¹⁸ Recently, molecular profiling in a series of patients with coexistent HGD-IPMN and invasive PDAC unveiled more tumorigenic pathways according to the proportion of shared sequence variations.¹⁹ The malignant progression can originate from a single aberrant cell clone developing first IPMN and afterwards PDAC (the sequential subtype), from similar clonal origin with later divergence (the branch-off subtype), or from 2 separate sequence-variation profiles causing either IPMN or PDAC (the de novo subtype). In line with the branch-off and de novo mechanisms, additional previous clinical evidence suggests that both HGD-IPMN and LGD-IPMN can be associated with concomitant PDAC.^20-22 Moreover, a recent study revealed that different clones of neoplastic cells with individual sequence-variation patterns arise within a single IPMN lesion.²³

Morphological risk factors and biological markers included in the available guidelines are broadly inaccurate with regard to prognostication of malignant conditions, with areas under the receiver operating characteristics curve ranging from 0.51 to 0.84,⁷ thus leading to potential underestimation of the true malignant potential of IPMN. Notably, with an area under the curve of 0.80, subjective radiological assessment in the present Heidelberg collective was noninferior to that reported in international guidelines. Moreover, the sensitivity of radiological assessment was 77.8%, higher than in the American Gastroenterology Association guidelines (7.3%), the Fukuoka guidelines (73.2%), and the American College of Radiology guidelines (53.7%).²⁴

Importantly, we found that only 35% of more than 1400 patients who underwent resection at our institution were treated in a timely fashion. In other words, almost two-thirds of operations were performed either in an early, premalignant phase or when malignant disease had already progressed to an advanced stage.

It may be argued that some patients in the too early group could have had undergone clinical observation, endoscopic examination, cytological and molecular diagnostic tests, and eventually surgery in the event of positive findings.^5,25 However, almost one-third of them had either MD-IPMN or MT-IPMN, which are associated with a high potential of malignant transformation.^4,5,26-28 Although MPD dilatation to 5 to 10 mm is considered a relative indication for resection,⁴ a previous analysis from our institution reported a 37% risk of HGD or invasive cancer associated with an MPD size of 5 to 9 mm.²⁹ A more recent analysis³⁰ from the US revealed an even higher risk of malignant disease (68%) in this group of patients. These observations are reflected in the present result as well. The increasing MPD diameter (3.0-8.0 mm) and the increasing proportion of patients fulfilling MD-IPMN or MT-IPMN criteria shown in the 3 groups suggests that MPD size may be a very strong and clinically useful feature to indicate a timely surgical intervention.

Cyst size assessment as an aid to clinical decision-making remains a topic of debate, and the growth rate over time, rather than cyst size itself, has gained in importance.^31,32 However, progression has been observed in IPMNs smaller than 3 cm without other suspicious radiological signs.³³ In these cases, alternative clinical observations may indicate IPMN transformation. For instance, elevated CA 19-9 levels are associated with a 4-fold to 6-fold increase in the malignant disease rate.³⁴ The presence of diabetes seems to double the likelihood of HGD or invasive cancer, underlining the reported association between disturbed glucose metabolism and PDAC.^35-37 The presence of pancreatitis has likewise been associated with a 3-fold increase in malignant disease rates.³⁸ Although the significance of smoking habits in IPMN has not been finally resolved,³⁹ smoking has been reported to be associated with malignant progression.⁴⁰ Finally, the presence of cyst-associated symptoms, such as abdominal pain, has been historically linked to an increased risk of malignant findings.⁴¹ Large data series on IPMN under surveillance have demonstrated that the risk of malignancy increases over time.^12,13,32

Under the assumption of different types of IPMN, research in the past 20 years has focused on the identification of subgroups with different risks of malignant disease. Current evidence suggests that these elements barely modulate the time to progression and the fate of all IPMN is eventually to become invasive. Still, the timeline of malignant disease development is unknown today and may not be similar in every individual patient, which reflects the current dilemma of accurate IPMN management. The present study, however, addressed this topic in a large cohort and puts a special focus on the correct time for surgery.

Performing timely or even early resection may reduce the extent of surgery, thus limiting the short-term and long-term sequelae with no increase in recurrence rates.⁴² Indeed, there was a significantly higher rate of parenchyma-sparing resection in the too early group than the other groups of our study, which was accompanied by lower rates of medical complications and less need for reintervention. Interestingly, the rates of postoperative mortality were also significantly lower in the too early and timely groups than in the too late group.

Avoiding delayed resection yields other relevant advantages. Given the linear increase in malignant transformation with the passage of time after initial diagnosis, lifelong follow-up is strongly recommended.¹² In this context, health care–associated costs should be taken into account, as well as the risk of missing signs of progression between consecutive follow-up visits. Two cost-effectiveness studies^43,44 have suggested that immediate surgery would eventually prove more expensive than surveillance strategies. Immediate surgery would, however, be the most effective approach with regard to quality of life and life expectancy.

In this study, almost one-third of patients underwent resection too late. Most of them were found to have radiological and clinical features suggestive of IPMN-associated malignant conditions. However, suspicious radiological features were absent in 12% of patients in the too late group and two-thirds of patients in the timely group. Almost half of these patients would have been improperly allocated to surveillance protocols. On the other hand, almost one-quarter of the patients in the too late group presented with a previous medical history of pancreatic cyst, and around half of them had already been under a watch-and-wait policy for more than 1 year before receiving surgery. Because most patients were referred to our institution after an initial workup, it was not possible to keep track of their medical course and the motivation for the observational practice. The risk inherent to apparently harmless IPMN undergoing observation, the current gaps in our understanding of the natural behavior of IPMN, and the need for accurate tools for safe surveillance programs are highlighted by our findings.

The present study has several limitations. Importantly, the data from the presented collective of patients with resected IPMNs were not matched with data from a comparable cohort of patients under surveillance. The Heidelberg interdisciplinary workup supports an earlier surgical approach, which is not generally agreed on. The higher reported rate of malignant conditions than in other cohorts may likewise be criticized. However, given that the indications for surgery in Heidelberg are more liberal and therefore more patients with IPMN undergo resection, our results clearly show the severe risk of undertreatment and highlight the shortcomings of the existing guidelines. Another limitation of the present study is that no outcomes in long-term overall survival after an index IPMN resection are discussed.

We considered borderline dysplasia as a separate condition, although the new World Health Organization classification¹⁰ puts borderline dysplasia together with LGD-IPMN. However, this choice was based on expert opinion and not high-level evidence¹⁵ and has not been sufficiently evaluated with regard to long-term prognosis, especially recurrence risk. A recent Japanese study⁴⁵ has shown that the percentage of patients who develop high-risk recurrence after IPMN surgery is 4.1% in those with initially LGD lesions vs 6.7% in those with initially HGD lesions at the time of index resection. Therefore, the LGD group, including those with former borderline dysplasia, does not show a prognostically relevant difference to the HGD group in terms of high-risk recurrence; observations such as these challenge the new 2-tier World Health Organization system.

In conclusion, until the true biological nature of IPMN is clarified, a precise estimation of the time to progression is possible and more accurate evidence-based guidelines are established, the blind pursuit of a watch-and-wait policy cannot be justified. Especially for MT-IPMN and MD-IPMN, timely resection should be attempted MPD diameter may serve as an index parameter for this decision. Existing prediction metrics remain inadequate, and at referral centers, the risk of surgery-associated complications can be justified, especially in young patients with a long life expectancy.

Accepted for Publication: January 19, 2021.

Published Online: May 19, 2021. doi:10.1001/jamasurg.2021.0950

Corresponding Author: Markus W. Büchler, MD, Department of General, Visceral, and Transplantation Surgery, Heidelberg University Hospital, Im Neuenheimer Feld 110, 69120 Heidelberg, Germany (markus.buechler@med.uni-heidelberg.de).

Author Contributions: Drs Büchler and Hinz had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

Concept and design: Tjaden, Sandini, Mihaljevic, Kaiser, Khristenko, Gaida, Diener, Mehrabi, Strobel, Hackert, Büchler.

Acquisition, analysis, or interpretation of data: Tjaden, Mihaljevic, Khristenko, Mayer, Hinz, Gaida, Berchtold, Schneider, Mehrabi, Müller, Strobel, Hackert, Büchler.

Drafting of the manuscript: Tjaden, Sandini, Kaiser, Khristenko, Gaida, Diener, Hackert.

Critical revision of the manuscript for important intellectual content: Tjaden, Mihaljevic, Kaiser, Khristenko, Mayer, Hinz, Gaida, Berchtold, Diener, Schneider, Mehrabi, Müller, Strobel, Hackert, Büchler.

Statistical analysis: Sandini, Hinz, Hackert.

Obtained funding: Büchler.

Administrative, technical, or material support: Kaiser, Khristenko, Gaida, Schneider, Strobel, Büchler.

Supervision: Mihaljevic, Khristenko, Diener, Mehrabi, Strobel, Hackert, Büchler.

Other—radiological evaluation and supervision: Khristenko.

Conflict of Interest Disclosures: None reported.

Funding/Support: The work of Dr Sandini was partially supported by the International Hepato-Pancreato-Biliary Association Kenneth W. Warren Fellowship.

Role of the Funder/Sponsor: The funders had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.