Article Text

Download PDFPDF

Original article
Resection of pancreatic cancer in Europe and USA: an international large-scale study highlighting large variations
  1. Lei Huang1,
  2. Lina Jansen1,2,
  3. Yesilda Balavarca3,
  4. Esther Molina-Montes4,
  5. Masoud Babaei1,
  6. Lydia van der Geest5,
  7. Valery Lemmens5,
  8. Liesbet Van Eycken6,
  9. Harlinde De Schutter6,
  10. Tom B Johannesen7,
  11. Claus W Fristrup8,
  12. Michael B Mortensen9,
  13. Maja Primic-Žakelj10,
  14. Vesna Zadnik10,
  15. Nikolaus Becker11,
  16. Thilo Hackert12,
  17. Margit Mägi13,
  18. Tiziana Cassetti14,
  19. Romano Sassatelli14,
  20. Robert Grützmann15,
  21. Susanne Merkel15,
  22. Ana F Gonçalves16,
  23. Maria J Bento16,
  24. Péter Hegyi17,
  25. Gábor Lakatos18,
  26. Andrea Szentesi17,
  27. Michel Moreau19,
  28. Tony van de Velde20,
  29. Annegien Broeks20,
  30. Milena Sant21,
  31. Pamela Minicozzi21,
  32. Vincenzo Mazzaferro22,
  33. Francisco X Real23,24,
  34. Alfredo Carrato25,
  35. Xavier Molero26,27,
  36. Marc G Besselink28,
  37. Núria Malats4,
  38. Markus W Büchler12,
  39. Petra Schrotz-King3,
  40. Hermann Brenner1,2,3
  1. 1 Division of Clinical Epidemiology and Aging Research, German Cancer Research Center (DKFZ), Heidelberg, Germany
  2. 2 German Cancer Consortium (DKTK), German Cancer Research Center (DKFZ), Heidelberg, Germany
  3. 3 Division of Preventive Oncology, German Cancer Research Center (DKFZ) and National Center for Tumor Diseases (NCT), Heidelberg, Germany
  4. 4 Geneticand Molecular Epidemiology Group, Spanish National Cancer Research Centre (CNIO), CIBERONC, ISCIII, Madrid, Spain
  5. 5 Netherlands Cancer Registry (NCR), Netherlands Comprehensive Cancer Organization (IKNL), Utrecht, Netherlands
  6. 6 Belgian Cancer Registry (BCR), Brussels, Belgium
  7. 7 Registry Department, The Cancer Registry of Norway (CRN), Oslo, Norway
  8. 8 Danish Pancreatic Cancer Database (DPCD), Odense, Denmark
  9. 9 Danish Pancreatic Cancer Group, HPB Section, Department of Surgery, Odense University Hospital, Odense, Denmark
  10. 10 Epidemiology and Cancer Registry, Institute of Oncology Ljubljana, Ljubljana, Slovenia
  11. 11 Clinical Cancer Registry, DKFZ and NCT, Heidelberg, Germany
  12. 12 Department of General, Visceral and Transplantation Surgery, Heidelberg University Hospital, Heidelberg, Germany
  13. 13 Estonian Cancer Registry, National Institute for Health Development, Tallinn, Estonia
  14. 14 Pancreatic Cancer Registry of Reggio Emilia Province, Unit of Gastroenterology and Digestive Endoscopy AUSL-RE, Local Health Authority-IRCCS, Reggio Emilia, Italy
  15. 15 Department of Surgery, Friedrich-Alexander-University Erlangen-Nürnberg, Erlangen, Germany
  16. 16 Departments of Epidemiology, Portuguese Oncology Institute of Porto (IPOP), Porto, Portugal
  17. 17 Institute for Translational Medicine, University of Pécs, Pécs, Hungary
  18. 18 Department of Oncology, St. Istvan and St. Laszlo Hospital and Out-Patient Department, Budapest, Hungary
  19. 19 Department of Surgical Oncology, Jules Bordet Institute (IJB), Brussels, Belgium
  20. 20 Biometrics Department, The Netherlands Cancer Institute (NKI), Amsterdam, Netherlands
  21. 21 Analytical Epidemiology and Health Impact Unit, Department of Preventive and Predictive Medicine, Fondazione IRCCS, Istituto Nazionale dei Tumori (INT), Milan, Italy
  22. 22 Hepato-Biliary Surgery Unit, Istituto Nazionale dei Tumori (INT), and University of Milan, Milan, Italy
  23. 23 Epithelial Carcinogenesis Group, Spanish National Cancer Research Centre (CNIO), CIBERONC, Madrid, Spain
  24. 24 Department de Ciencies Experimentals i de la, Universitat Pompeu Fabra, Barcelona, Spain
  25. 25 Department of Oncology, Ramon y Cajal University Hospital, IRYCIS, Alcala University, CIBERONC, Madrid, Spain
  26. 26 Hospital Universitari Vall d’Hebron, Exocrine Pancreas Research Unit and Vall d’Hebron Research Institute (VHIR), Universitat Autonoma de Barcelona, Campus de la UAB, Barcelona, Spain
  27. 27 CIBEREHD and CIBERESP, Madrid, Spain
  28. 28 Dutch Pancreatic Cancer Group, Academic Medical Centre Amsterdam, Amsterdam, Netherlands
  1. Correspondence to Lina Jansen, Division of Clinical Epidemiology and Aging Research, German Cancer Research Center (DKFZ), Heidelberg 69120, Germany; l.jansen{at}


Objective Resection can potentially cure resectable pancreatic cancer (PaC) and significantly prolong survival in some patients. This large-scale international study aimed to investigate variations in resection for PaC in Europe and USA and determinants for its utilisation.

Design Data from six European population-based cancer registries and the US Surveillance, Epidemiology, and End Results Program database during 2003–2016 were analysed. Age-standardised resection rates for overall and stage I–II PaCs were computed. Associations between resection and demographic and clinical parameters were assessed using multivariable logistic regression models.

Results A total of 153 698 records were analysed. In population-based registries in 2012–2014, resection rates ranged from 13.2% (Estonia) to 21.2% (Slovenia) overall and from 34.8% (Norway) to 68.7% (Denmark) for stage I–II tumours, with great international variations. During 2003–2014, resection rates only increased in USA, the Netherlands and Denmark. Resection was significantly less frequently performed with more advanced tumour stage (ORs for stage III and IV versus stage I–II tumours: 0.05–0.18 and 0.01–0.06 across countries) and increasing age (ORs for patients 70–79 and ≥80 versus those <60 years: 0.37–0.63 and 0.03–0.16 across countries). Patients with advanced-stage tumours (stage III–IV: 63.8%–81.2%) and at older ages (≥70 years: 52.6%–59.5%) receiving less frequently resection comprised the majority of diagnosed cases. Patient performance status, tumour location and size were also associated with resection application.

Conclusion Rates of PaC resection remain low in Europe and USA with great international variations. Further studies are warranted to explore reasons for these variations.

  • pancreatic cancer
  • surgical resection
  • variation
  • age
  • TNM stage
  • population-based

Statistics from

Significance of this study

What is already known on this subject?

  • Pancreatic cancer (PaC) survival remains poor with limited improvement over the decades. Resection can potentially cure PaC in some patients and significantly prolong survival in a relevant percentage of patients. However, most tumours are not resectable because of local growth or distant metastasis.

  • Tumour, node, metastasis (TNM) stage I–II PaCs are mostly considered resectable, while stage III–IV tumours, which constitute the majority of PaCs, are deemed mostly unresectable. Most guidelines regard poor performance status but not old age as a resection-limiting factor. Centralisation and specialisation might increase resection rate.

What are the new findings?

  • In 2012–2014, resection rates were from 13.2% (Estonia) to 21.2% (Slovenia) across countries. Even for patients with stage I–II cancers, which are mostly considered resectable, resection rates were only between 34.8% (Norway) and 68.7% (Denmark).

  • Great international variations regarding resection for PaC exist especially in stage I–II cancer.

  • Within the investigated countries, an increasing trend in resection is only observed in a very limited number of countries with national and/or regional implementation of centralisation regarding PaC treatment (eg, the Netherlands).

  • The majority of patients with PaC are diagnosed with advanced-stage cancers and at older ages. At the population level, patient age, performance status, tumour stage, location and size are factors significantly associated with resection application.

How might it impact on clinical practice in the foreseeable future?

  • Our findings, together with the existing evidence, highlight the low resection rates with large variations across countries even for early-stage cancers, warranting policy-makers’ and clinical practitioners’ attention.

  • Further studies are warranted to explore the reasons for these large variations across registries to elucidate whether and how PaC management could be improved and to investigate the treatment patterns in other countries.


Pancreatic cancer (PaC) is one of the most lethal malignancies worldwide, with mortality closely paralleling incidence.1 In Europe, PaC is estimated to have caused a total of 85 600 deaths in 2016, accounting for 6.3% of all cancer-related deaths.2 With 5-year survival of only about 5%, PaC prognosis has not significantly improved over the past decades despite numerous efforts in therapeutic modification.3

PaC usually occurs at older ages, and more than half of the patients are diagnosed with advanced-stage diseases due to usually unspecific early symptoms/signs.4 Surgical resection is the only potentially curative treatment. Resectability is to a large extent determined by vascular involvement. Only patients with favourable conditions and high probability to achieve curative resection are deemed eligible candidates for resection, which is conducted in less than one-fifth of diagnosed cases.5 PaC resectability criteria in the European Society for Medical Oncology (ESMO) guidelines6 follow the US National Comprehensive Cancer Network (NCCN) guidelines.7 According to the current guidelines,6–10 for locally advanced PaCs involving major arteries (T4/stage III according to the American Joint Committee on Cancer (AJCC)/Union for International Cancer Control (UICC) TNM (tumour, node, metastasis) staging system) and metastatic (M1/stage IV) cancers, resection should be mostly avoided. However, the resectability criteria are differentially and arguably defined.3 10 Notably, even most of stage I–II PaCs are not curatively resectable.

While resection offers the only chance to cure PaC, the 5-year postsurgical survival remains low (8%–12%) if no further adjuvant treatment is administered.11 The perioperative mortality is noteworthy. In Germany, it is 10% on the whole population basis and is volume-dependent (5%–6% even in the largest centres), mainly influenced by failure to rescue and surgical expertise.12 13 Furthermore, postsurgical morbidity remains relatively high (30%–40%) even in high-volume centres.14 It is important to well inform a given patient so that he/she could decide whether or not to undergo resection.

Large-scale investigations on surgical resection for PaC are rare.10 Using data from seven population-based and nine institution-based cancer registries, this large-scale study aimed to explore the application of surgical resection for PaC in various European countries and USA.


Data sources

Data from six European national population-based (the Netherlands, Belgium, Norway, Denmark, Slovenia and Estonia) cancer registries and the US Surveillance, Epidemiology, and End Results (SEER)-18 database15 were analysed (table 1). Additionally, resection rates in nine institution-based (Heidelberg, Reggio Emilia Pancreatic Cancer Registry (REPCR), Spanish arm of PanGenEU (PanGenS), Erlangen Cancer Registry (ErCR), Portuguese Oncology Institute of Porto (IPOP), Hungarian Pancreatic Study Group (HPSG), Jules Bordet Institute (IJB), Istituto Nazionale dei Tumori (INT) in Milan and The Netherlands Cancer Institute (NKI) in Amsterdam) registries from seven European countries were explored. Data quality is described in online supplementary methods. This study was conducted according to the Declaration of Helsinki and reported following the STROBE guidelines.

Supplementary file 1

Table 1

General information on participating registries

Patient eligibility criteria

Tumour topography and morphology were based on the International Classification of Diseases for Oncology (third edition). Only patients with confirmed diagnoses of primary invasive malignancies of the exocrine pancreas and without other cancers were selected (see online supplementary table s1). Individuals were included regardless of being eligible for resection. Patients with benign/premalignant tumours, non-PaC neoplasms involving the pancreas, neuroendocrine tumours, carcinoids, sarcomas/stromal tumours, germ-cell neoplasms, lymphomas or periampullar tumours were excluded. Patients with diagnosis based on autopsy/death certificate only were also excluded. Tumour stage was defined according to the AJCC/UICC TNM staging system (sixth or seventh edition). As the fifth and prior editions were incompatible with the versions in effect during 2003–2016,6 only patients with PaC diagnosed from 2003 or the first year when resection status was registered until 2016 or the most recent year of registration were included in each registry.

Collected information

Information on patient (age and sex) and tumour characteristics (pathology, location and TNM stages) and treatment was obtained. Resection was defined as surgical removal of the primary tumour, regardless of being curative or palliative and extents of excision and lymphadenectomy. For patients with multiple records, information was derived from the most complete and recent data source.

Specific covariates were only registered by certain population-based registries. Tumour size was available in USA. Comorbidity information at diagnosis was available in the Eindhoven Cancer Registry (EiCR), part of the national Netherlands Cancer Registry. Belgium and Denmark provided Eastern Cooperative Oncology Group (ECOG) score. Information on neoadjuvant chemotherapy was not available in USA, Norway or Estonia, and information on neoadjuvant radiotherapy was not available in Norway or Estonia. Hospital type was available in the Netherlands and Belgium.

Statistical analyses

Patient age was divided into four groups (<60, 60–69, 70–79 and ≥80 years). Age-standardised resection rates were computed for each population-based registry using the age distribution of the US patients, the largest group of patients included in our analysis. Trends of standardised resection rates over years were evaluated for each country and, for simplicity, rates over two-calendar year periods (2003–2004 until 2013–2014) were displayed. Although not all centres could contribute data for the entire period of investigation, age-standardised resection rates during 2012–2014 could be estimated for all databases.

Multivariable logistic regression was used to investigate associations of resection with sex, age group, tumour location and cTNM (clinical TNM) stage in overall patients, with female,<60 years, pancreas head tumour and stage I–II as the reference category, respectively. Year of diagnosis was also included in the models. Sensitivity analyses were conducted by repeating association analyses after imputing missing stages using multiple imputations16 (variables applied: year of diagnosis, sex, age, tumour location, resection, chemotherapy, radiotherapy and survival time and status; routine: PROC MI). In additional analyses, associations were investigated for tumour subgroups according to stage and location, respectively and were reassessed after adding tumour size, ECOG score, comorbidity or hospital type one by one into the models. SAS software (V.9.4, Cary, North Carolina, USA) was used for analysis. Statistical significance was defined by two-sided P<0.05.


The following results refer to analyses of population-based registries if not otherwise specified. Results for institution-based registries are mostly described in online supplementary material.

Characteristics of overall patients

A total of 147 700 patients from seven population-based registries were analysed (table 2). In USA, the Netherlands, Belgium, Norway and Slovenia, patients diagnosed between 2003/2004 and 2013/2014 were included. In Denmark and Estonia, included patients were diagnosed in 2011–2016 and 2009–2014, respectively. Mean ages were 70–72 years, with patients≥70 years comprising 52.6%–59.5% of the diagnosed cases. Around half of the patients (47.2%–52.6%) were female. Most patients had pancreas head tumours (67.8%–74.7%). Metastatic diseases were most commonly diagnosed with proportions between 54.5% (USA) and 72.8% (Slovenia), whereas stage I–II cancers only comprised 18.8% (Slovenia) to 36.2% (USA). Stage was missing for 10.5%–26.5% of patients in investigated countries except Belgium, and missingness was mostly associated with age and tumour location (see online supplementary table S2). Overall, resection rates ranged from 13.0% (Estonia) to 21.7% (Belgium). Chemotherapy was administered to 15.0% (Estonia) to 57.3% (Belgium) of patients. Radiotherapy was less frequently used (1.9% (Slovenia) to 6.9% (Belgium)).

Table 2

Demographic and clinical characteristics of overall patients with pancreatic cancer in population-based registries*

Characteristics of resected patients

Together only 16.0% (23 683/147 700) of the investigated patients with PaC underwent resection (table 3). Resected patients were younger (mean age, 65–67 years) than total patients. Only 36.9%–43.8% were 70 years or older. Most patients had stage I–II tumours (75.1% (Slovenia) to 92.4% (Denmark)). Pancreas head tumours were more frequent among resected patients (78.7%–87.7%). Patients with pancreas head cancers had the greatest proportion of stage I–II tumours, and those with tail tumours had the largest proportion of metastatic lesions. Neoadjuvant chemotherapy (0.3%–4.2%) and radiotherapy (0.1%–4.3%) were rarely administered in countries with available information. Pancreatoduodenectomy was the most common surgical approach (68.6% (Denmark) to 83.9% (the Netherlands)). Adjuvant chemotherapy utilisation varied strongly with proportions between 12.0% (Estonia) and 55.7% (Denmark). Adjuvant radiotherapy was more frequently used in USA (29.5%) than in Europe, where proportions ranged from 0.1% (Denmark) to 8.9% (Belgium).

Table 3

Demographic and clinical characteristics of resected patients with pancreatic cancer in population-based registries*

Characteristics of resected patients by cancer stage were further described (see online supplementary tables s3 and s4). Operated patients with stage I–II disease were mostly older than those with stage III–IV PaC (mean age, 65–68 vs 64–66 years). There was generally a greater proportion of pancreas head tumour in stage I–II PaC than in stage III–IV disease (79.4%–88.6% vs 66.1%–85.3%). Compared with stage IV cancer, resected stage III PaC was more often located in pancreas head (75.4%–94.1% vs 59.0%–75.8%). Accordingly, pancreatoduodenectomy was more frequently performed for stage III than stage IV PaC (66.9%–80.4% vs 50.9%–66.3%).

Resection trends and rates

As shown in figure 1, overall resection rates increased over time in USA (2003–2004 to 2013–2014: 14.1% to 17.0%; Ptrend<0.001), the Netherlands (2003–2004 to 2013–2014: 8.2% to 17.9%; Ptrend<0.001) and Denmark (2011–2012 to 2013–2014: 12.0% to 17.6%; Ptrend=0.007), while no significant trends were observed in Belgium (Ptrend=0.270), Norway (Ptrend=0.102), Slovenia (Ptrend=0.092) or Estonia (Ptrend=0.406). When focusing on the period 2012–2014, resection rates ranged from 13.2% (Estonia) to 21.2% (Slovenia) in population-based registries. Variations were stronger across European institutions with resection rates ranging from 9.4% (NKI) to 52.6% (Heidelberg). For the subgroup of stage I–II tumours, increasing trends were observed in USA (2003–2004 to 2013–2014: 39.4% to 44.0%; Ptrend<0.001), the Netherlands (2003–2004 to 2013–2014: 32.6% to 58.2%; Ptrend<0.001) and Denmark (2011–2012 to 2013–2014: 60.5% to 70.1%; Ptrend=0.017), while no significant trends were observed in Belgium (Ptrend=0.726), Norway (Ptrend=0.675), Slovenia (Ptrend=0.596) or Estonia (Ptrend =0.406). In 2012–2014, the proportions of resected patients ranged from 34.8% (Norway) to 68.7% (Denmark) in population-based registries. Again, variations were larger across institutions with rates from 19.7% (IJB) to 96.9% (Heidelberg).

Figure 1

Age-standardised trend of surgical resection in population-based registries. Age-standardised trend of surgical resection for overall patients with pancreatic cancer (A) and those with TNM (tumour, node, metastasis) stage I–II tumours (B). The US cancer population was used for age standardisation. Starting from 1973 and 1953 respectively, trends of increasing resection rates were observed in USA (5.5%–17.8%) and Norway (5.1%–18.4%) in overall patients (results not shown).

Association of resection with demographic and clinical parameters

Association of resection with demographic and clinical variables in each country was investigated using a multivariable model including year of diagnosis, sex, age, tumour location and stage (table 4). While resection was not significantly associated with sex, it was less frequently conducted with increasing age and cTNM stage. Specifically, compared with patients<60 years, ORs for resection among patients aged 70–79 and ≥80 years ranged between 0.37 (the Netherlands) and 0.63 (Estonia) and between 0.03 (the Netherlands) and 0.16 (USA), respectively. Compared with stage I–II tumours, ORs of stages III and IV cancers were 0.05–0.18 and 0.01–0.06, respectively. Resection was significantly less frequently conducted in pancreas body tumours than head cancers in all countries except Slovenia and Estonia with ORs ranging from 0.22 (Denmark) to 0.65 (USA). Pancreas tail tumours were significantly more often resected than pancreas head cancers in USA (OR=1.99), the Netherlands (OR=1.47), Norway (OR=1.70), Denmark (OR=2.46) and Estonia (OR=3.18), while no significant associations were observed in Belgium. In Slovenia, even an opposite pattern was detected (OR=0.49). After multiple imputations for missing stages, patterns remained unchanged (see online supplementary table s5).

Table 4

Association of surgical resection versus non-resection with demographic and clinical parameters for patients with pancreatic cancer in population-based registries estimated by multivariable logistic regression

Association patterns for stage I–II PaCs were mostly consistent with those for total cancers (see online supplementary table s6). Within stage III–IV PaCs, pancreas tail tumours were significantly less frequently resected compared with pancreas head cancers in USA (OR=0.83), the Netherlands (OR=0.63) and Belgium (OR=0.53); in Norway, Denmark and Estonia, the original significant associations disappeared. Moreover, in USA, male patients were significantly less often resected (OR=0.89). Resection rates were higher for stage III than stage IV PaC in all countries (see online supplementary figure s1). Resection patterns were mostly similar for both stages (see online supplementary table s7). In USA (OR=2.28) and the Netherlands (OR=2.18), pancreas tail tumours were more often resected compared with head cancers in stage III PaCs, but not in stage IV cancers. In Slovenia, pancreas tail PaCs were less frequently resected compared with head cancers in stage IV tumours (OR=0.39), but not in stage III cancers. Association patterns were similar for pancreas head, body and tail cancers (data not shown).

Associations of resection with tumour size, performance status, comorbidities and hospital type were further explored by adding these factors one by one to the main models with covariates of year of diagnosis, sex, age, tumour location and stage (see online supplementary table s8). Resection was significantly less frequently conducted with increasing tumour size in USA. In countries where performance status was available (Belgium and Denmark), cancers were less frequently resected with increasing ECOG scores. In the Netherlands (OR=2.81) and Belgium (OR=2.13), patients managed in academic hospitals underwent more often resection. Detailed information on comorbidity was available from EiCR in the Netherlands. Cardiac, vascular, neurological and pulmonary diseases were associated with less frequent resection. Patients with ≥2 comorbidities underwent less often resection compared with those without comorbidity (OR=0.60). Additional analyses were further performed for stage I–II and III–IV cancers, respectively. Patterns for both stages were mostly consistent with the overall ones.

Institution-based results

To further explore variations in pattern of PaC resection at institution level which might be different from population-based results, data from nine leading institutions in PaC management in the respective country were further analysed (supplementary results, see online supplementary figure s2 and supplementary tables s9–s11). Association results for four large institutions were mostly consistent with the population-based findings but larger variations were observed.


This international large-scale study described utilisation of surgical resection in PaC in Europe and USA. We found overall low resection rates, although with major variations across registries. Various factors were associated with resection application.

Variations in resection rates between countries and over time have been rarely investigated. Our study showed that overall resection rates were low in all investigated countries. Even among patients with stage I–II PaCs, which however only comprised 19%–36% of diagnosed cases, only 35%–69% were resected during 2012–2014. Increases in resection rates over time were only observed in USA, the Netherlands and Denmark. Compared with overall cancers, variations in resection rates across countries were stronger for stage I–II tumours. Notably, centralisation/specialisation, which potentially partly explains the observed geographical and temporal variations, was implemented in all the three countries that showed increases in resection rates. Centralisation could contribute to increases in resection rates.17 Centralisation started in the Netherlands regionally in 200518 19 and nationally in 2011,20 and in Denmark in 200021. Nationally, the number of hospitals performing pancreatoduodenectomy for PaC decreased from 39 to 23 in 2004 to 2009 in the Netherlands and the proportion of patients resected at medium/high-volume (>10 resections/year) centres increased from 53% to 91%, accompanied by an increase in pancreatoduodenectomy from 258 (11%) to 394 (18%).17 22 In 2011, the Dutch Health Inspectorate set an annual volume standard of 20 pancreatoduodenectomies per hospital.23 In Eindhoven, the number of hospitals performing resection decreased from 6 to 3 in 2005–2008 and the annual number of resections per hospital increased from 5 to 16.18 In western Netherlands, pancreatic surgery was centralised in two high-volume hospitals from 2006.24 In USA, nationwide centralisation in pancreatic surgery is also occurring with state-specific variations.25 26 Between 1992–1994 and 2010–2012 in Florida, the number of pancreatic surgeons decreased from 363 to 196, while the resection number increased from 729 to 1569.26 In Denmark, only four university hospitals are allowed to perform pancreatectomy, two with yearly ≥75 resections and the other two ≥25. In Belgium, a population-based study27 proposed centralisation in 2007. While the number of treating hospitals decreased from 77 to 68 with the average resection number per hospital increasing from 5.8 to 7.2 in 2009–2014, only four hospitals kept performing >15 pancreatectomies yearly. In Slovenia, pancreatectomy is centralised in three centres. Patients receiving surgery in higher-volume centres had better survival.26 28 Morbidity and hospital duration could also be reduced by centralisation.25 27 29

Among the investigated European institutions, resection rates also varied substantially, which might be partly due to the centralisation/specialisation degree.19 In Heidelberg where the European Pancreatic Cancer Centre locates, resection rate was as high as 97% for stage I–II cancers. Notably, patients treated in this specialised pancreas surgical centre were mostly deemed operable and fit for surgery. While population-based results are not comparable to institution-based ones, interestingly and notably, in 2012–2014 the resection rate in IJB which treated approximately 1% of the overall national incident cases was lower than the corresponding national population-based resection rate, which might be explained by patient selection. In IJB, patients were younger and smaller proportions of pancreas body and tail cancers were observed, compared with the national registry. Patients were more often treated with chemotherapy and/or radiotherapy. Only 3% of total patients were diagnosed with stage I–II PaC in the specialised centre NKI, which rarely performs any pancreatectomy. Operable patients are mostly referred to other centres and most of the patients come for non-surgical treatment with more advanced-stage disease or for secondary treatment after having been resected elsewhere.

Although in countries with centralisation resection rates increased, they remained low. We found that resection was less often conducted with more advanced tumour stage, with larger tumour size, with older age, in pancreas body cancer and with poorer performance status. Notably, patients with stage III–IV cancers (64%–81%) and those ≥70 years (53%–60%) comprised the majority of the PaC cases in all included countries, largely contributing to the low resection rates based on the strong associations of resection application with age and stage. Patients≥70 years remained the majority (53%–61%) among those with stage I–II PaCs.

Advanced stage is a negative prognostic factor and often regarded contraindicative to surgery.30 Tumour size might be negatively associated with resection frequency, because larger lesions are more prone to involving vessels and associated with more advanced stages. Many resected ‘unresectable’ tumours are detected unsuspectedly during surgery.31 Pancreatologists’ consensus states localised PaCs without major vessel involvement (mostly stage I–II) to be clearly resectable.6 7 For ‘unresectable’ cancers, resection is seldom recommended.8 10 Per guidelines,6 7 tumours circumferentially encasing celiac axis or superior mesenteric artery (T4/stage III) and metastatic cancers (M1/stage IV) are deemed unresectable, largely due to the high possibility of incomplete resection, which is associated with unfavourable survival.32 With increasing experience in vascular surgery, vessel involvement characterising T4/stage III tumours is less frequently regarded surgery-contraindicative.33 The term ‘borderline resectable’ was developed to define a specific subgroup within locally advanced tumours for which curative resection is potentially applicable. While borderline resectable PaCs might be associated with resection rates higher than the other stage III tumours, they could not be investigated here due to the newly emerged, continuously evolving and non-uniform definition.9 34

Resectability criteria are a hot and key issue in PaC management. While there might be differences in treatment guidelines across countries, regarding resectability the investigated countries all follow the NCCN guidelines.7 35 While there remain discrepancies, major progress has been made in the definition which is becoming more and more standardised and uniform in recent years.3 6 10 36 37 However, since the criteria are relatively complex for routine registration practice especially at the population-based level and were mostly evolving during the investigated period, resectability status was mostly not readily registered in the included registries. We conducted subgroup analyses according to TNM stage, which is commonly used and could hopefully help to identify a subgroup of patients where resection is more likely. Notably, the TNM staging system could not substitute the resectability criteria. Locally advanced, unresectable PaCs defined by the International Study Group of Pancreatic Surgery (ISGPS) and NCCN are different from T4/stage III tumours defined by AJCC and UICC. Even TNM stage II tumours can be locally advanced and unresectable according to the ISGPS and NCCN guidelines. A standardised and uniformed resectability definition is hopefully to be implemented in registry practice in the near future.

Older ages are also a negative prognostic factor30 and are associated with higher prevalence of comorbidities38 and complications.39 Whether older ages should be regarded resection-contraindicative remains debatable. Small-scale studies suggested that resection was associated with better survival in the elderly,40 41 which however could be partly explained by selection of fitter patients for resection. Some studies showed that compared with younger individuals, fit elderly patients might obtain similar survival benefits from resection, which could be safely performed.20 42 However, some other large monocentric series have identified age as a risk factor of operative mortality in scores predicting post-pancreatoduodenectomy mortality.38 43 The operative mortality in octogenarians was 4% in a series of 2000 pancreatoduodenectomies.44 Thus, the general pretreatment condition of aged patients should be carefully assessed to ensure that it allows pancreatectomy to be performed with an acceptable perioperative risk. As more than half of the patients with PaC were 70 years or older at diagnosis, further studies are needed to investigate the benefit and harm of PaC resection for the elderly, which should be balanced.20

Tumour location was another factor influencing PaC resection. Pancreas body cancers were less frequently resected, which could be explained by the fact that pancreas body lesions might be most challenging to manage, due to common involvement of major vasculatures and advanced stage at diagnosis.3 33 We also found that higher ECOG scores, which are associated with higher perioperative morbidity and mortality risks, were negatively associated with resection frequencies. Specific comorbidities were also inversely related to resection rates.

The aspects discussed above potentially explain surgeons’ choice of resection for PaC and the low resection rates. Further reasons especially for the low resection rates for stage I–II PaCs remain to be clarified. Notably, resectability might be largely influenced by surgeons’ capabilities and experience, surgical techniques, skills, procedure and equipment.33 Tumour biology, symptom burden, operative tolerance, patient’s preferences, support systems and quality of life (QoL) are important aspects to consider beyond standard resectability classification. Based on SEER-18, for 96.8% of unresected patients, resection was not recommended by clinicians. QoL decreases considerably in the early postoperative phase and its full recovery might take up to half a year.45 However, overall resection does not worsen and even benefits QoL in most PaC survivors in the longer term.46 Patient’s choice might be influenced by his/her socioeconomic status, health insurance coverage, marriage status and trust in doctor.47 Notably, some people might have limited access to medical care due to distance from treatment facilities. Future studies especially on patient preferences and access to care are warranted.

There are some limitations for our study. Due to the retrospective design, some important variables (eg, tumour size, performance status and comorbidities) were in some registries not available or the missing number was too high to be included in the main analyses. Moreover, heterogeneity in available variables across countries might lead to information bias potentially impacting robust inferences of the data and highlights the need for improving comprehensiveness and standardisation in registration practice. Another limitation was that proportions of patients with missing TNM stages were relatively high. Nevertheless, patterns remained the same after multiple imputations. Differences in resection rates especially across institution-based registries could be due to selective referral as well as selective preferences of treatment modalities which might differ in different centres. Furthermore, treatment patterns in other European countries (eg, UK and France) were not investigated in this study, and no Asian or African registries were included. Treatment patterns in these continents/countries need to be clarified in future studies. Notably, USA and Netherlands registries contributed the largest numbers of cases among included registries. However, results for each registry were presented separately, and no pooled-analysis was conducted, reducing the concern of the potential impact of these large registries on result interpretation. In institution-based registries, only treatment modalities applied in the corresponding centre where patients were diagnosed were reported.

Differences in tumour stage across countries were observed, potentially highlighting variation in quality of PaC staging, as it is often difficult to correctly stage T4 cancers with arterial invasion compared with T1–3 cancers, and there could be relevant interobserver variation. There could be differences at the national level.48 In most of the investigated countries, many of those patients are discussed by multidisciplinary teams (MDTs) which is required by law. In Denmark and Estonia, almost all patients are evaluated by MDTs. In Belgium, the proportion of MDT-discussed patients increased from 57.8% in 2005 to 84.5% in 2012. In the Netherlands, about two-third of patients were discussed by MDTs in 2012.49

Strengths of our study are the use of high-quality data from multiple population-based cancer registries, the large sample size, the strict inclusion criteria, the careful case selection and the uniformly defined and standardised variables across registries.

In conclusion, this international large-scale population-based investigation shows that PaC resection rates are generally low in Europe and USA with great international variations. Further studies are warranted to explore reasons for these variations.


We are very grateful to the staff in Surveillance, Epidemiology, and End Results Program (SEER), Netherlands Cancer Registry (NCR), Dutch Pancreatic Cancer Group, Belgian Cancer Registry (BCR), Cancer Registry of Norway (CRN), Danish Pancreatic Cancer Database (DPCD), Danish Pancreatic Cancer Group, Danish Clinical Registries (RKKP), Cancer Registry of Slovenia (CRS), National Center for Tumor Diseases in Heidelberg (NCT), Pancreatic Cancer Registry of Reggio Emilia (REPCR), Spanish arm of the PanGenEU study (PanGenS, especially the coordinators, field and administrative workers, technicians and study participants of the Spanish centres), Erlangen Cancer Registry (ErCR), Portuguese Oncology Institute of Porto (IPOP), Hungarian Pancreatic Study Group (HPSG), Jules Bordet Institute (IJB), Istituto Nazionale dei Tumori (INT) and The Netherlands Cancer Institute in Amsterdam (NKI) for their kind work in data collection and delivery.



  • Contributors Conception or design: LH, LJ, YB, MB, PS-K, HB. Acquisition, analysis or interpretation of data: LH, LJ, YB, EM-M, M Babaei, LvdG, VL, LVE, HDS, TBJ, CWF, M Mortensen, MPŽ, VZ, NB, TH, M Mägi, TC, RS, RG, SM, AFG, M Bento, PH, GL, AS, M Moreau, TvvV, AB, MS, PM, VM, FXR, AC, XM, M Besselink, NM, M Büchler, PS-K, HB. Drafting of the manuscript: LH. Critical revision of the manuscript for important intellectual content: LH, LJ, YB, EM-M, M Babaei, LvdG, VL, LVE, HDS, TBJ, CWF, M Mortensen, MPŽ, VZ, NB, TH, M Mägi, TC, RS, RG, SM, AFG, M Bento, PH, GL, AS, M Moreau, TvvV, AB, MS, PM, VM, FXR, AC, XM, M Besselink, NM, M Büchler, PS-K, HB. Statistical analysis: LH, YB, EM-M (for PanGenS only). Administrative, technical or material support: M Büchler, PS-K, HB. Supervision: LJ, HB. All authors have given final approval of the manuscript for submission and publication.

  • Funding This study was partly supported by the German Cancer Aid (Deutsche Krebshilfe, #111365), European Cooperation in Science and Technology-COST Action (#BM1204: EUPancreas) and Fondo de Investigaciones Sanitarias (FIS), Instituto de Salud Carlos III-FEDER, Spain (#PI1501573).

  • Disclaimer The funders had no involvement in study design; in the collection, analysis or interpretation of data; in the writing of the report or in the decision to submit the paper for publication.

  • Competing interests None declared.

  • Ethics approval This study was approved by Ethics Committee of the Medical Faculty Heidelberg.

  • Provenance and peer review Not commissioned; externally peer reviewed.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.