Article Text

Download PDFPDF

Original article
A multicentre comparative prospective blinded analysis of EUS and MRI for screening of pancreatic cancer in high-risk individuals
  1. F Harinck1,
  2. I C A W Konings1,
  3. I Kluijt2,
  4. J W Poley1,
  5. J E van Hooft3,
  6. H M van Dullemen4,
  7. C Y Nio5,
  8. N C Krak6,
  9. J J Hermans7,
  10. C M Aalfs8,
  11. A Wagner9,
  12. R H Sijmons10,
  13. K Biermann11,
  14. C H van Eijck12,
  15. D J Gouma13,
  16. M G W Dijkgraaf14,
  17. P Fockens3,
  18. M J Bruno1
  19. on behalf of the Dutch research group on pancreatic cancer surveillance in high-risk individuals
    1. 1Department of Gastroenterology and Hepatology, Erasmus MC, University Medical Center Rotterdam, Rotterdam, The Netherlands
    2. 2Family Cancer Clinic, The Netherlands Cancer Institute, Antoni van Leeuwenhoek Hospital, Amsterdam, The Netherlands
    3. 3Department of Gastroenterology and Hepatology, Amsterdam Medical Center, University Medical Center Amsterdam, Amsterdam, The Netherlands
    4. 4Department of Gastroenterology and Hepatology, University of Groningen, University Medical Center Groningen, Groningen, The Netherlands
    5. 5Department of Radiology, Amsterdam Medical Center, University Medical Center Amsterdam, Amsterdam, The Netherlands
    6. 6Department of Radiology, Erasmus MC, University Medical Center Rotterdam, Rotterdam, The Netherlands
    7. 7Department of Radiology and Nuclear Medicine, Radboud University Medical Center Nijmegen, Nijmegen, The Netherlands
    8. 8Department of Clinical Genetics, Amsterdam Medical Center, University Medical Center Amsterdam, Amsterdam, The Netherlands
    9. 9Department of Clinical Genetics, Erasmus MC, University Medical Center Rotterdam, Rotterdam, The Netherlands
    10. 10Department of Genetics, University of Groningen, University Medical Center Groningen, Groningen, The Netherlands
    11. 11Department of Pathology, Erasmus MC, University Medical Center Rotterdam, Rotterdam, The Netherlands
    12. 12Department of Surgery, Erasmus MC, University Medical Center Rotterdam, Rotterdam, The Netherlands
    13. 13Department of Surgery, Amsterdam Medical Center, University Medical Center Amsterdam, Amsterdam, The Netherlands
    14. 14Clinical Research Unit, Amsterdam Medical Center, University Medical Center Amsterdam, Amsterdam, The Netherlands
    1. Correspondence to F Harinck, Erasmus MC, University Medical Center Rotterdam, s Gravendijkwal 230, Rotterdam 3015 CE, The Netherlands; f.harinck{at}erasmusmc.nl

    Abstract

    Objective Endoscopic ultrasonography (EUS) and MRI are promising tests to detect precursors and early-stage pancreatic ductal adenocarcinoma (PDAC) in high-risk individuals (HRIs). It is unclear which screening technique is to be preferred. We aimed to compare the efficacy of EUS and MRI in their ability to detect clinically relevant lesions in HRI.

    Design Multicentre prospective study. The results of 139 asymptomatic HRI (>10-fold increased risk) undergoing first-time screening by EUS and MRI are described. Clinically relevant lesions were defined as solid lesions, main duct intraductal papillary mucinous neoplasms and cysts ≥10 mm. Results were compared in a blinded, independent fashion.

    Results Two solid lesions (mean size 9 mm) and nine cysts ≥10 mm (mean size 17 mm) were detected in nine HRI (6%). Both solid lesions were detected by EUS only and proved to be a stage I PDAC and a multifocal pancreatic intraepithelial neoplasia 2. Of the nine cysts ≥10 mm, six were detected by both imaging techniques and three were detected by MRI only. The agreement between EUS and MRI for the detection of clinically relevant lesions was 55%. Of these clinically relevant lesions detected by both techniques, there was a good agreement for location and size.

    Conclusions EUS and/or MRI detected clinically relevant pancreatic lesions in 6% of HRI. Both imaging techniques were complementary rather than interchangeable: contrary to EUS, MRI was found to be very sensitive for the detection of cystic lesions of any size; MRI, however, might have some important limitations with regard to the timely detection of solid lesions.

    • PANCREATIC CANCER
    • ENDOSCOPIC ULTRASONOGRAPHY
    • MAGNETIC RESONANCE IMAGING
    • SCREENING
    • SURVEILLANCE
    View Full Text

    Statistics from Altmetric.com

    Significance of this study

    What is already known on this subject?

    • Screening individuals at high risk for developing pancreatic cancer leads to the detection of early-stage pancreatic ductal adenocarcinoma and premalignant lesions.

    • Endoscopic ultrasound (EUS) and MRI are considered the most accurate techniques for pancreatic imaging within a screening setting.

    • Hardly any study has prospectively compared the diagnostic yields of EUS and MRI in a blinded fashion.

    • It is unclear whether one test will suffice or whether both tests are complementary.

    What are the new findings?

    • For detection of pancreatic lesions, both tests are complementary rather than interchangeable.

    • EUS proved to be particularly sensitive for the detection of small solid lesions.

    • MRI proved to be particularly sensitive for the detection of (small) cystic lesions.

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

    • Within a screening setting, in order to maximise the detection rate of clinically relevant lesions, both EUS and MRI should be considered.

    Introduction

    Despite all efforts in past decades, the prognosis of pancreatic ductal adenocarcinoma (PDAC) is still dismal. With a mean survival of <6 months and a 5-year survival of <5%, PDAC ranks among the top 5 causes of cancer-related deaths in the Western world despite its relatively low incidence.1 Survival rates are strongly dependent on the stage at which PDAC is detected. Therefore, there is great interest in pancreatic screening to detect PDAC at an earlier and potentially curable stage or, even more preferable, to detect high-grade precursor lesions.

    Screening of the general population is not feasible as we currently lack a simple, reliable and inexpensive screening tool. However, evidence is starting to accumulate that screening might be worthwhile when offered to individuals at high risk of developing PDAC.2 High-risk individuals include mutation carriers of PDAC-prone gene mutations (eg, CDKN2A, BRCA1, BRCA 2, STK11/LKB1) and relatives of patients with familial PDAC. The risk of developing PDAC within these well-defined populations of high-risk individuals is estimated to be at least 10-fold increased compared with the general population and exceeds 76-fold in selected cases.2 ,3 Previous studies have shown that screening these high-risk individuals leads to the detection of early-stage PDAC and premalignant lesions.4–13

    At present, endoscopic ultrasound (EUS) and MRI are considered the most accurate techniques for pancreatic imaging within a screening setting.2 ,8 Only one study8 has prospectively compared the diagnostic yields of EUS and MRI in a blinded fashion. In this study,8 good concordance for lesion size, number and location between EUS and MRI was seen.

    We conducted a prospective head-to-head blinded comparison between EUS and MRI for the detection of clinically relevant pancreatic lesions at first-time screening in individuals at high risk for developing PDAC.

    Methods

    Study design and sites

    We conducted a multicentre prospective blinded cohort study. Participating centres were Erasmus MC-University Medical Center Rotterdam, Academic Medical Center Amsterdam, University Medical Center Groningen and the Netherlands Cancer Institute-Antoni van Leeuwenhoek Hospital.

    Objective

    A prospective head-to-head blinded comparison between EUS and MRI for the detection of pancreatic lesions at first-time screening in individuals at high risk for developing PDAC.

    Participants

    Data were collected within the framework of our ongoing Familial Pancreatic Cancer Surveillance Study. Eligible for inclusion are asymptomatic individuals with an estimated ≥10-fold increased familial or inherited PDAC risk compared with the general population (see inclusion criteria below). The minimal age for inclusion is 45 years or 10 years younger than the age of the youngest relative with PDAC, whichever occurred first. For patients with Peutz–Jeghers syndrome, the minimal age for inclusion is 30 years or 10 years younger than the age of the youngest relative with PDAC, whichever occurred first. Potential candidates are evaluated and recruited by a clinical geneticist to check whether inclusion criteria are fulfilled. This evaluation includes (1) obtaining a detailed personal and family medical history, (2) verification of clinical diagnoses reported by patients and family members by review of medical and pathological records and revision of histological slides whenever available, and (3) based on the medical information genetic testing for suspected gene mutation(s).

    Inclusion criteria

    1. Carriers of CDKN2A gene mutations, regardless of the family history of PDAC.14

    2. Patients with Peutz–Jeghers syndrome (diagnosis based on a proven LKB1/STK11 gene mutation).3

    3. Carriers of gene mutations in BRCA1, BRCA2, p53 or Mismatch Repair Gene with a family history of PDAC in at least two family members.15–18

    4. First-degree relatives (FDRs) of patients with familial pancreatic cancer (FPC). FPC families were defined as families affected by PDAC in at least (1) two FDRs, (2) three relatives in which the affected cases are FDR or second-degree relatives (SDRs) of each other or (3) two SDRs of whom at least one relative was aged <50 years at the time of diagnosis.2 ,19 Eligible for inclusion in our study were all family members with at least one FDR with PDAC.

    Exclusion criteria

    Individuals with a history of PDAC, age <18, upper GI tract obstruction, severe medical illness (American Society of Anaesthesiologists' score ≥3) or who were unable to provide informed consent due to mental retardation or language barrier were excluded.

    Experimental methods

    Screening techniques

    Endoscopic ultrasonography

    All EUS procedures were carried out by five experienced endosonographers (JWP, PF, MJB, HMvD and JEvH). Both electronic radial (Olympus UC-160 AE, Olympus Europe, Hamburg, Germany with Aloka α 5 ultrasoundprocessor, Zug, Switzerland or Pentax EG-3670 URK, Pentax Medical Europe Headquarters, Hamburg, Germany with Hitachi ultrasoundprocessor, Hitachi Medical Systems Europe, Zug, Switzerland) and curvilinear (Olympus UCT/UCP 160, Olympus Europe, Hamburg Germany with Philips HDI 5000 ultrasoundprocessor, Philips Healthcare Medical Systems, Best, The Netherlands or Aloka α 10 ultrasoundprocessor, Zug, Switzerland) instruments were used according to the personal preference of the endosonographer. Procedures were performed under conscious sedation with midazolam/fentanyl or propofol. Imaging of the pancreas was carried out from the duodenum and stomach and was digitally recorded with lossy compression (Endobase, Olympus, Hamburg). In case a relevant clinical lesion or a lesion of unknown significance was detected, both a case description and video recordings were distributed among all participating endosonographists for independent review. The outcome of this independent review was then presented to the local multidisciplinary hepato-pancreato-biliary team consisting of gastroenterologists, surgeons and radiologists for final decision-making regarding further management.

    MRI

    MRI was performed at a 1.5 or 3.0 Tesla machine (Signa HDxt, Discovery 450 or 750, GE Healthcare, Milwaukee, Wisconsin, USA; Siemens Avanto or Philips). The following sequences were obtained: coronal balanced steady state free precession imaging with 6 mm slices, coronal and axial T2-weighted single-shot fast spin echo series with 6 mm slices, axial respiratory triggered (RT) fat suppressed T2-weighted fast spin echo series with 6 mm slices, 3-D heavily T2-weighted coronal MR cholangiopancreatography with 1.4 mm slices (with subsequent axial reconstructions) and breath-hold axial diffusion weighted imaging series including apparent diffusion coefficient mapping with 6 mm slices, using three different b-values (b=50, 400 and 800 s/mm2). The dynamic sequence involved fat suppressed 3-D T1-weighted spoiled gradient-echo series using 2 or 3 mm slices before and after intravenous administration of gadobutrol (Gadovist 1.0 mmol/mL, Bayer Schering Pharma, Berlin, Germany) at a dose of 0.1 mmol/kg body weight using automated infusion with a power injector at a flow rate of 2 mL/s. Series were timed in the arterial, pancreatic and portal phase using bolus tracking. MRIs were scored by three highly experienced radiologists (CYN, NCK and JJH).

    Image interpretation and reporting

    Participating gastroenterologists and radiologists were blinded to the baseline results of either EUS or MRI. Reporting of imaging findings was standardised across EUS and MRI using a Case Record Form. We specifically looked for clinically relevant abnormalities defined as solid lesions of any size and cystic lesions larger than 10 mm, see also below.20 The imaging diagnosis used for the present analysis was based on the initial description/diagnosis provided by either the attending radiologist or gastroenterologist. Whenever there was a discrepancy between the findings of EUS and MRI with respect to clinically relevant lesions, the EUS video and MR images were reviewed to determine whether the lesion(s) was (were) indeed not detectable by the other technique.

    Clinically relevant lesions

    In this article, we mainly focus on the detection of clinically relevant lesions. These include all solid lesions suspicious for a malignancy as well as all lesions that fulfil the revised Sendai criteria for surgery or close follow-up:20 cysts ≥3 cm, cysts with thickened/enhancing cyst walls and/or mural nodules and/or a solid component, main branch intraductal papillary mucinous neoplasms (IPMNs) with main pancreatic duct ≥10 mm in size and side branch IPMNs with side duct dilations/cysts >10 mm.

    Surgical outcomes considered ‘a success’

    Detection and surgical treatment of (1) invasive cancer ≥T1N0M0 with negative margins, (2) multifocal pancreatic intraepithelial neoplasia (PanIN) 3 lesions and (3) high-grade IPMNs were defined as a successful outcome of surveillance.2

    Follow-up policy

    The follow-up policy was based on the agreement of an expert panel consisting of experienced endosonographists, surgeons, radiologists and pathologists and was as follows:

    1. Annually, when EUS and/or MRI detected no pancreatic abnormalities or cystic lesions <10 mm.

    2. Three months in case EUS and/or MRI detected a lesion for which a morphological diagnosis could not be readily made, hereinafter referred to as lesions with unknown clinical significance.

    3. Six months in case of a detected cyst or side branch IPMN with a diameter >10 mm and <30 mm without malignant features (see below).

    4. Surgical resection in case of the detection of a solid lesion morphologically suspicious for a malignancy, cystic lesion >30 mm, cystic lesions with malignant features (thickened/enhancing cyst walls and/or mural nodules) or main branch IPMN with main pancreatic duct ≥10 mm.20

    Statistical methods

    Descriptive statistics were generated to describe patient and lesion characteristics. To compare both imaging test results, a percentage agreement was calculated for the detection and location of lesions, and a Spearman's rho correlation coefficient was calculated for the size of lesions. We considered an agreement of 0.00 as poor, 0.01–0.20 as slight, 0.21–0.40 as fair, 0.41–0.60 as moderate, 0.61–0.80 as substantial and 0.81–1.00 as almost perfect agreement.21 All analyses were conducted using the Statistical Package for the Social Sciences (V.21, SPSS Institute, Chicago, Illinois, USA).

    Results

    Patient characteristics

    At 1 September 2013, a total of 166 high-risk individuals were prospectively included in this study. Twenty-two individuals underwent some form of pancreatic screening prior to inclusion and were therefore excluded from this blinded baseline analysis. Furthermore, five high-risk individuals were excluded from this analysis because they either had underwent only EUS or only MRI (figure 1). Therefore, a total of 139 individuals from 81 unique families were included in this blinded analysis of whom the baseline characteristics are summarised in table 1. The mean age at inclusion was 51 years (SD 9.7, range 20–73 years). Sixteen individuals (12%) were current smokers at time of inclusion. Forty individuals (29%) had a medical history affected by cancer; in 24 of these individuals (60%), the cancer type was melanoma. Seventy-one individuals (51%) carried a pancreatic cancer-prone gene mutation, whereas the remaining individuals stemmed from FPC families. No fine needle aspiration was performed and no procedure-related adverse events occurred.

    Table 1

    Characteristics of asymptomatic high-risk individuals who underwent baseline screening with EUS and MRI (n=139)

    Figure 1

    Flow chart. EUS, endoscopic ultrasonography; FPC, familial pancreatic cancer.

    Diagnostic yield

    Clinically relevant lesions, as defined previously, were detected by either EUS and/or MRI in 9 out of 139 high-risk individuals (6%). Two of these nine individuals (22%) had two clinically relevant lesions. Therefore, a total of 11 clinically relevant lesions were identified in nine individuals: two solid lesions and nine cysts larger than 10 mm. Further characteristics are summarised in table 2. Additionally, eight hypo-echoic areas with unknown clinical relevance were detected by EUS in eight individuals and two lesions with reduced signal intensity on TI-weighted series were detected by MRI in two individuals. Together with the remaining 58 cysts <10 mm (in 34 individuals) and 9 duct ectasias (in 6 individuals), a total of 88 lesions were identified in 46 out of 139 high-risk individuals (33%). Characteristics of these lesions are summarised in table 3. No difference in findings was seen between individuals that carried a PDAC-prone gene mutation and individuals that stemmed from an FPC family.

    Table 2

    Characteristics of all morphologically clinically relevant lesions detected at baseline screening with EUS and MRI (n=11)

    Table 3

    Characteristics of all detected lesions at baseline screening with EUS and MRI (n=88)

    Of all 11 clinically relevant lesions, 6 (55%) were detected by both modalities. EUS detected a total of eight (73%) and MRI detected a total of nine (82%) clinically relevant lesions. When analysing all lesions (clinically relevant lesions, hypo-echoic areas of unknown clinical relevance, hypo-intense areas of unknown clinical relevance and cysts <10 mm), MRI was very sensitive for the detection of cystic lesions (of all 67 cystic lesions, 60 (90%) were detected by MRI and 26 (39%) by EUS) and in specific for subcentimeter cysts (of all 58 cystic lesions <10 mm, 51 (88%) were detected by MRI and 20 (35%) by EUS). In total, there were 38 cysts <10 mm (mean 5 mm, range 2–9 mm) in 23 individuals that were detected by MRI but not by EUS. In 16 of these 23 individuals, the EUS investigation was performed using the radial scope (70%). The majority of these subcentimeter cysts therefore were missed using the radial scope and this could not be attributed to one single centre or endosonographer. Conversely, EUS detected two solid lesions that were not detected by MRI, also not after re-evaluation of the MRI: (1) a 11 mm solid lesion in the body of the pancreas (table 2, lesion 1 and figure 2A) and (2) a 7 mm solid lesion in the head of the pancreas (table 2, lesion 2 and figure 2C). For both lesions, resection was performed. The former lesion proved to be a 12 mm T1N0M0 moderately differentiated adenocarcinoma (figure 2B). Although postsurgical staging suggested a favourable outcome (R0 resection of a small tumour of 12 mm), the patient developed local disease recurrence with liver and peritoneal metastases a few months later and died within 36 months after initial diagnosis. The 7 mm solid lesion in the head of the pancreas proved to be two separate 3 mm lesions very close to each other and was therefore classified as multifocal PanIN2 (figure 2D). Characteristics of all detected lesions by EUS and MRI are summarised in table 4.

    Table 4

    Characteristics of lesions detected by EUS and by MRI respectively at baseline screening

    Figure 2

    (A) The still endoscopic ultrasound image of a 11 mm solid lesion of the body of the pancreas. (B) The histological image after resection of the lesion shown in (A), which proved to be a 12 mm T1N0M0 moderately differentiated ductal adenocarcinoma. (C) Still endoscopic ultrasound image of a 7 mm solid lesion in the head of the pancreas. (D) The histological image after resection of the lesion shown in (C), which proved to be two separate 3 mm lesions, within 2 mm distance of each other, classified as multifocal pancreatic intraepithelial neoplasia 2.

    Both EUS and MRI detected areas of (yet) unknown clinical relevance; these were lesions that were not cystic in nature and without the distinct morphology according to the consensus panel to be classified as a solid lesion or hypoechoic lobule. Table 5 provides a detailed description of these lesions of unknown clinical relevance. None of these cases had a history of (acute) pancreatitis or chronic ethanol overuse; only one was a heavy smoker (>15 cigarettes per day for over 40 years, case no. 5, table 5). In all cases, except one (case no. 8, in table 5), follow-up showed these lesions to remain stable or being not detectable anymore. In case no. 8, EUS detected two 5 mm hypo-echoic lesions (lesion #8 and #9 in table 5). Interval screening at 3 and 6 months showed no morphological changes. However, at follow-up at 12 months, both lesions had a more solid appearance and one of these lesions discretely increased in size (from 5 to 7 mm). Based on these morphological changes, it was decided to resect both lesions. A partial spleen preserving body/tail resection was performed and pathological examination showed multifocal PanIN2 lesions.

    Table 5

    Characteristics of lesions/areas of (yet) unknown clinical relevance that were detected at baseline screening (n=10)

    A total of 41 out of 139 high-risk individuals (30%) had at least one feature of chronic pancreatitis: lobularity was the most frequently detected feature (19%), as well as hyperechoic pancreatic duct margins (17%) and hyperechoic stranding (15%). Twenty individuals (14%) had three or more features of chronic pancreatitis. No differences in features of chronic pancreatitis were seen between individuals that carried a PDAC-prone gene mutation and individuals that stemmed from an FPC family. Also, no correlation with the presence of cysts, alcohol use or tobacco use was found.

    Agreement between EUS and MRI at baseline screening (blinded analysis)

    The agreement between EUS and MRI for the detection of clinically relevant lesions (n=11) was moderate with a 55% agreement (see table 6). Not surprisingly, the agreement was only fair for detection of all lesions regardless of size (n=88, agreement 26%). However, there was a perfect agreement between EUS and MRI for location of both clinically relevant lesions (n=6) and all lesions (n=26) (agreement 100%). Also, there was a substantial to almost perfect agreement between EUS and MRI on the size of clinically relevant lesions (Spearman's rho correlation coefficient of 0.638) and the size of all detected lesions (Spearman's rho correlation coefficient of 0.859).

    Table 6

    Agreement between endoscopic ultrasonography and MRI for different variables and subsets of pancreatic lesions

    Follow-up 12 months

    A total of 135 out of 139 high-risk individuals underwent repeated surveillance after 12 months; one patient developed metastatic disease (case no. 1 in table 2) and three patients withdrew from the surveillance programme (one patient had emigrated and two patients provided no reason for withdrawal). At 12 months' follow-up, 12 clinically relevant lesions were detected in 8 individuals (6%). Also, 7 of these 12 lesions were unchanged compared with baseline screening (lesion #3, 4, 6, 7, 9, 10 and 11, table 2). Two lesions increased in size: in case no. 6 (table 2) a cyst in the pancreatic head grew from 5 to 10 mm, and in another case, a 9 mm large cyst in the tail of the pancreas grew to 13 mm, both without secondary signs of malignancy. Three newly developed clinically relevant pancreatic lesions were identified: (1) case no. 6 developed a cyst of 13 mm in the body of the pancreas that was detected by both imaging modalities; (2) case no. 2, who had underwent a pancreaticoduodenectomy, developed a new 10 mm large cyst in the pancreatic tail detected by MRI; and (3) in another case, one new 10 mm large cyst in the body of the pancreas was detected by MRI, all without secondary signs of malignancy.

    Agreement between EUS and MRI at follow-up 12 months (unblinded analysis)

    The agreement between EUS and MRI for the detection of clinically relevant lesions increased from 55% at baseline screening (blinded results) to 67% agreement at follow-up 12 months (unblinded results).

    Discussion

    To determine the effectiveness of EUS and MRI in their ability to detect pancreatic lesions in high-risk individuals, we conducted a multicentre prospective study in which we compared baseline results in a blinded fashion. This nationwide, blinded prospective study shows that for detection of pancreatic lesions, in this series both tests were complementary rather than interchangeable. EUS and/or MRI showed a total of 11 morphologically clinically relevant lesions at baseline screening in 6% of participating high-risk individuals.

    To date, results of 12 screening studies for pancreatic cancer have been published.4–13 ,22 ,23 Based on these results, EUS and MRI are currently regarded as the most promising screening techniques as they are relatively widely accessible, have low morbidity rates and, in particular, are superior to any other imaging modality with regard to the detection of small pancreatic lesions. However, data on which of these two imaging techniques is to be preferred for screening purposes are largely lacking since only one of these series was conducted in a blinded fashion.8 In this study,8 good concordance for lesion size, number and location between EUS and MRI was seen.

    In our cohort, however, we found a moderate to fair agreement between EUS and MRI on the detection of both clinically relevant lesions and all pancreatic lesions, but a good to perfect agreement on size and location of detected lesions. The moderate agreement between EUS and MRI on the detection of pancreatic lesions is a reflection of the fact that only 55% of the clinically relevant lesions (6 of 11) were detected by both EUS and MRI. For baseline imaging, both radiologist and endosonographists were blinded to the results of the competing imaging modality. Since both modalities were performed on the same day as much as possible, the order being dependent on availability and logistics, it was not possible to unblind investigators after the initial investigation. For follow-up investigations after 12 months, however, radiologists and endosonographers were aware of the baseline results. The agreement per lesion between both techniques increased from 55% at baseline screening to 67% at follow-up surveillance. The disagreement between EUS and MRI lies mostly in the detection of cysts by EUS, and the detection of solid lesions by MRI. As a result, in this series both techniques were complementary rather than interchangeable.

    A possible explanation for the discrepancy in findings between Canto et al8 and our study is the use of both the radial and the linear scope for EUS investigations in all individuals in Canto's cohort, whereas only one of both scopes was used in our cohort. Performing an EUS investigation with two different endoscopes likely increases the detection of (subcentimeter) cysts, as the authors state themselves.24 Also, since the miss rate for pancreatic lesions in high-risk individuals seems lower for linear EUS than for radial EUS,24 the frequent use of the radial scope in our cohort might have negatively influenced our reported concordance between EUS and MRI. Canto's cohort consisted of a slightly different subset of individuals (older mean age, difference in types of underlying gene mutations); however, this should not affect the comparative analysis of EUS and MRI and thus does not explain the discrepancy in findings. Both cohorts were screened in tertiary high-volume centres and by experienced endosonographers and radiologists only.

    EUS proved to be particularly sensitive for the detection of small solid lesions. Two solid lesions detected by EUS, including a stage I PDAC, were not detected by MRI. When MRI investigations in both cases were re-evaluated, these lesions were indeed not detectable. Our results are in line with the results of previous studies that were conducted in a clinical setting (sporadic cases) that showed EUS has the highest sensitivity for the detection of <20 mm pancreatic cancers compared with other imaging modalities including MRI.25 ,26

    MRI was particularly sensitive for the detection of (small) cystic lesions. All nine cystic lesions sized ≥10 mm were detected by MRI, whereas EUS detected six (66%). There are multiple possible explanations why these lesions were missed by EUS. The 24 mm cyst in the head of the pancreas (table 2, lesion #9) was composed of multiple microcysts (figure 3). This composition influences the penetration of the ultrasound waves with the walls of the microcysts reflecting the ultrasound waves causing the lesion not to appear as a cystic lesion on EUS. However, one still would expect the lesion to be discordant compared with the surrounding pancreatic parenchyma and thus identified as a potential ‘lesion’. Indeed, at follow-up 12 months, a different endosonographer detected both lesion #9 and #10 (table 2, case 8). The location of cyst #11 in the uncinate process (table 2, case 9) could be the reason why this particular lesion was missed. This part of the pancreas is sometimes more challenging to visualise by EUS. Lastly, in both cases a radial scope was used. Although in this multicentre study the choice of the device was left to the discretion of the attending investigator, most endosonographists prefer a linear device to scan the pancreas.

    Figure 3

    A 24 mm cyst in the head of the pancreas composed of multiple microcysts.

    Strengths of our nationwide, multicentre, prospective study are that at baseline screening participating gastroenterologists and radiologists were blinded to the results of either EUS or MRI. Moreover, as a result of the extensive genetic evaluation prior to inclusion in this study and rigid inclusion criteria, our cohort consists of individuals truly at high risk for developing PDAC.

    This study is limited by the fact that we lack a definitive diagnosis of the vast majority of cases in whom an abnormality was detected, particularly if detected by one imaging modality only. As a resultant of this baseline screening, only two of all cases (1.4%) were operated. Consequently, it is yet impossible to make a final judgement with regard to the clinical relevance of the different types and sizes of pancreatic lesions detected. For instance, the importance of the hypo-echoic areas of unknown significance that were detected by EUS but not by MRI remains to be determined. Only longer follow-up will learn whether such findings bare clinical relevance. We are currently conducting a prospective follow-up study to assess the clinical relevance of various lesions detected by EUS and MRI and whether screening high-risk individuals is truly effective in reducing PDAC-related morbidity and mortality.

    The true challenge in pancreatic cancer surveillance is to adequately identify pre-neoplastic lesions to avoid resections of early-stage lesions (eg, PanIN1 and 2 lesions), but timely resect advanced lesions before cancer has developed. Based on the present study, it is not possible to draw definite conclusions about the (potential) merits of surveillance to prevent pancreatic cancer death. To answer this pivotal question, long-term follow-up studies are required in a large number of individuals. In this regard, it should be recognised that it has taken many years to prove that colon cancer screening saves lives.

    In conclusion, for individuals at high risk for developing pancreatic cancer that undergo screening, EUS and MRI are rather complementary than interchangeable imaging modalities in our series. For future screening therefore, we will continue to use both imaging modalities in the follow-up of our cohort of high-risk individuals. We found that, in contrast to EUS, MRI is very sensitive for the detection of even the smallest cysts. EUS seems to be most sensitive for the early detection of (small) solid lesions, which from a clinical perspective is an important property of this imaging modality. Exclusive use of linear devices is likely to improve the overall results of EUS. This should be taken into account at future revisions of recommendations regarding which imaging modality to use for surveillance.

    References

    View Abstract

    Supplementary materials

    • Supplementary Data

      This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.

    Footnotes

    • Collaborators On behalf of the Dutch research group on pancreatic cancer surveillance in high-risk individuals (in alphabetical order): Amsterdam Medical Center: CMA, MGWD, PF, DJG, FH, JEvH, ICAWK, CYN, Theo AM van Os and Ellen MA Smets. Erasmus MC, University Medical Center Rotterdam: KB, MJB, Henny van Duijl, CHvE, FH, ICAWK, NCK, JWP and AW. The Netherlands Cancer Institute-Antoni van Leeuwenhoek Hospital: Eveline MA Bleiker, Annemieke Cats, IK, Anja van Rens and Senno Verhoef. University of Groningen, University Medical Center Groningen: HMvD, RHS. University Medical Center Utrecht: Margreet GEM Ausems and Frank P. Vleggaar.

    • Contributors FW and ICAWK contributed equally. FH and ICAWK—responsible for the planning of the project, acquisition, analysis and interpreting of data. Assisted in drafting and critical review of paper. IK—initiator of the project. Contributed to the design of the work. Assisted in drafting and critical review of paper. As a clinical geneticist, IK was furthermore responsible for the genetic evaluation of all potential candidates. JWP, JEvH, and HvD—endosonographers who performed the EUSs. Members of expert panel. Assisted in the interpretation of the results and in the critical review of the paper. CYN, NCK and JJH—radiologists who scored the MRIs. Member of expert panel. Assisted in the interpretation of the results, and in the critical review of the paper. AW, CMA and RS—clinical geneticists responsible for the genetic evaluation of all potential candidates. Assisted in the interpretation of the results and in the critical review of the paper. KB—pathologist who reviewed the resected specimen. Member of the expert panel. Assisted in interpretation of the results and in the critical review of the paper. CHvE and DJG—surgeons who performed the resections. Member of expert panel. Assisted in the interpretation of the results and in the critical review of the paper. MGWD—methodologist. Contributed to the design of the work. Assisted in the analysis and interpretation of the results and critical review of the work. PF—contributed to the conception and design of the work. Assisted in the analysis and interpretation of the results. Assisted in drafting and critical review of the paper. Furthermore, performed EUSs. MJB—principal investigator. Obtained funding. Contributed to the conception and design of the work. Assisted in the analysis and interpretation of the results. Assisted in drafting and critical review of the paper. Furthermore, performed EUSs.

    • Funding ZonMW, grant number 120520016.

    • Competing interests None declared.

    • Patient consent Obtained.

    • Ethics approval The study protocol was approved by the Ethical Committee of all participating centres and the study was conducted in accordance with the Declaration of Helsinki.

    • 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.