Robot-assisted versus open pancreatoduodenectomy for cancer in the pancreatic head (DIPLOMA-2x2): study protocol for an international multicentre patient- and assessor-blinded randomized controlled non-inferiority trial

Abstract

Background

Robot-assisted pancreatoduodenectomy (RPD) is increasingly performed for cancer in the pancreatic head. Randomised evidence confirming its oncological safety and efficacy is lacking. The DIPLOMA-2x2 trial aims to compare the oncological safety of RPD versus OPD in terms of radicality (microscopically radical resection [R0] resection). We hypothesise that RPD is non-inferior to OPD in terms of radicality and superior regarding time to functional recovery.

Methods

The DIPLOMA-2x2 trial is an investigator-initiated, international, multicentre, patient- and assessor-blinded randomised non-inferiority trial. The trial was conducted as a roll-over of the DIPLOMA-2 trial and includes 20 high-volume tertiary referral hospitals in 7 countries. Minimum surgeon’s experience is 60 RPD and 60 OPD, with a minimum annual centre volume of 30 RPD prior to trial initiation. Eligible patients are adults (≥18 years) with suspected or proven upfront resectable pancreatic ductal adenocarcinoma (PDAC) and distal cholangiocarcinoma (DCC) without any vascular involvement. Participants are randomised in a 2:1 ratio to RPD or OPD, respectively, stratified by tumour indication (proven PDAC versus other) and preoperative pancreatic fistula risk (high versus low). In total, 413 patients will be included, of which 137 previously randomised in DIPLOMA-2 and 276 newly recruited in DIPLOMA-2x2. Patients are blinded up to postoperative day 5. Primary outcome is pathological R0-resection rate (pR0; >1 mm tumour clearance at surgical resection margins and anatomical surfaces, 0 mm clearance at anterior surface), tested for non-inferiority with a −7% margin. Pathological assessment is blinded, and a randomly selected 10% of specimen will undergo central review. The key secondary outcome, tested for superiority, is time to functional recovery. Other secondary outcomes include lymph node retrieval, initiation/completion of adjuvant therapy, time to start of adjuvant therapy, disease-free and overall survival, quality of life, and cost-effectiveness. Safety outcomes include major complications (Clavien–Dindo grade ≥ III), serious adverse events, and 90-day mortality. Primary analyses will be based on the modified intention-to-treat principle, including all patients who underwent resection. Postoperative follow-up is up to 90 days for short-term outcomes and up to 60 months for survival.

Discussion

The DIPLOMA-2x2 trial evaluates whether RPD is oncologically safe compared to OPD in patients with upfront resectable PDAC and DCC in experienced high-volume centres.

Trial registration

ISRCTN27483786. Registered: January 11, 2021 (updated: February 6, 2025). First enrolment: January 3, 2021. Expected completion of enrolment: May 2026.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s13063-026-09692-6.

Background

Pancreatic ductal adenocarcinoma (PDAC) and distal cholangiocarcinoma (DCC) are amongst the most aggressive and lethal malignancies [1]. At diagnosis, only 15–20% of patients present with a resectable tumour, and even after surgical treatment, long-term outcomes remain poor, with reported 5-year survival rates ranging between 10 and 25% depending on tumour stage and receipt of adjuvant therapy [2]. Surgical resection is the only curative treatment, but is associated with considerable postoperative morbidity, which may hinder recovery and delay adjuvant systemic therapy, compromising the overall prognosis [2–4].

In recent years, robot-assisted pancreatoduodenectomy (RPD) has increasingly been adopted in high-volume centres worldwide [5–7]. Compared with traditional open pancreatoduodenectomy (OPD), robotic platforms offer enhanced dexterity, tremor filtration, and enhanced three-dimensional magnification, potentially reducing surgical trauma and facilitating postoperative recovery [8]. Two recent randomised trials on RPD vs OPD, one single-centre randomised trial from Germany [9] and a multicentre trial from China [10], demonstrated comparable short-term safety outcomes of RPD as compared to OPD. However, these trials included heterogeneous periampullary tumour types, both benign and malignant disease, and were not specifically powered to evaluate oncological outcomes in patients with malignancy of the pancreatic head.

Therefore, the oncological safety of RPD for patients with cancer in the pancreatic head remains uncertain. Improved postoperative recovery following RPD could theoretically lead to earlier initiation and better tolerance of adjuvant therapy, potentially improving long-term survival. However, two randomised trials in patients with cervical and rectal cancer raised concerns as minimally invasive surgery failed to demonstrate non-inferiority in oncological outcome compared to open surgery [11, 12]. Furthermore, a recent meta-analysis including individual patient data of 16 studies comparing MIPD versus OPD reported a lower R0-resection rate for minimally invasive procedures in patients with DCC, although estimates varied across studies [13]. These findings highlight that the oncological safety of RPD remains to be determined by high-quality randomised trials before widespread adoption of RPD for cancer in the pancreatic head can be justified.

Therefore, the DIPLOMA-2x2 trial is initiated as a ‘roll-over’ extension of the DIPLOMA-2 trial, an international multicentre randomised trial comparing minimally invasive and open pancreatoduodenectomy for patients with resectable pancreatic head and periampullary neoplasms in high-volume centres [14]. The DIPLOMA-2 trial primarily evaluated short-term postoperative outcome. The DIPLOMA-2x2 trial continues within the same infrastructure and focuses specifically on the predefined subgroup of patients undergoing RPD versus OPD for upfront resectable PDAC and DCC without vascular involvement, to assess oncological outcome.

The primary hypothesis is that RPD is non-inferior to OPD with respect to the pR0-resection rate. Time to functional recovery is evaluated as a key secondary outcome, for which we hypothesise superiority of RPD compared to OPD.

Methods

Design

DIPLOMA-2x2 is an investigator-initiated, international, patient- and primary outcome assessor-blinded, parallel-group randomised controlled non-inferiority trial with a 2:1 ratio in patients with cancer in the pancreatic head. The study is conducted as a ‘roll-over’ extension of the recently completed DIPLOMA-2 trial, an international multicentre randomised controlled trial comparing minimally invasive and open pancreatoduodenectomy.

The DIPLOMA-2x2 trial continues with restricted inclusion criteria to complete enrolment of the pre-specified subgroup population of patients with upfront resectable PDAC and DCC without vascular involvement (Fig. 1). The study population consists of patients previously randomised in the DIPLOMA-2 trial who underwent RPD or OPD for preoperatively suspected or proven PDAC and DCC (N = 137), and newly randomised patients in the DIPLOMA-2x2 trial. The DIPLOMA-2x2 trial uses the existing infrastructure and participating centres of the DIPLOMA-2 trial, allowing efficient continuation with focused inclusion criteria.

Fig. 1.

DIPLOMA-2x2 study flow-chart according to SPIRIT

Continuation of the DIPLOMA-2x2 trial was conditional upon review of safety outcomes in DIPLOMA-2. Criteria for the ‘Go/No Go’ decision included a major morbidity rate (Clavien–Dindo grade ≥ III) below 35% and a 90-day mortality rate below 5% in the RPD group, based on the in-hospital/30-day mortality rate reported by the Dutch Pancreatic Cancer Audit [7, 15].

Inclusion for the DIPLOMA-2x2 trial started in December 2023 after approval of the Medical Ethics Review Committee Board of Amsterdam UMC (Amsterdam, the Netherlands) and Fondazione Poliambulanza Istituto Ospedaliero (Brescia, Italy). This protocol is developed according to the SPIRIT guidelines, and the trial results will be reported in accordance with the CONSORT statement [16, 17]. Patient representatives were involved in the design of the trial, including the selection of patient-relevant outcomes.

Study setting and participating centres

The DIPLOMA-2x2 trial is conducted in 20 tertiary referral centres across the Netherlands, Italy, Belgium, Germany, Spain, Sweden, and the USA. Centres were eligible to participate if they met a predefined minimum annual centre volume of 30 RPD. Participating surgeons were required to have performed at least 60 RPD and 60 OPD procedures prior to trial initiation. Centres not meeting these predefined experience thresholds were not eligible to participate in the trial.

Study population

Overall, 413 adult patients with an indication for elective pancreatoduodenectomy (PD) for suspected or proven upfront resectable PDAC and DCC without vascular involvement, technically eligible for both RPD and OPD, will be included. The study population consists of 137 patients randomised in the DIPLOMA-2 trial and 276 newly randomised patients [18].

Inclusion criteria

Eligible patients must meet all of the following criteria:

• Age at least 18 years;

• Indication for elective PD because of a proven or suspected^# upfront resectable cancer in the pancreatic head consisting of PDAC and DCC*. Suspected malignancy includes radiologic or clinical features highly suggestive of PDAC or DCC, for which PD is indicated according to multidisciplinary team consensus;

• Both RPD and OPD are technically feasible according to the local treatment team;

• Preoperative multiphase CT-scan imaging shows no signs of vascular involvement with the mesenteric, celiac trunk and hepatic arteries and the portomesenteric vein; (3D reconstruction optional) on a CT maximum 28 days old (both at randomization and time of surgery);

• Fit to undergo PD according to the surgeon and anaesthesiologist;

• Written informed consent.

Patients with PDAC and DCC were both included as these tumours are localized in the pancreatic head and difficult to distinguish preoperatively. While tumour biology differs, both require identical surgical resection and margin assessment. Therefore, the inclusion of patients with PDAC and DCC is considered appropriate for evaluating radical resection rate. Patients with ampullary and duodenal cancer are excluded because irradical (R1) resection is here of much less concern and overall survival rates are much better compared to PDAC and DCC [19–23].

Pathological proof for malignancy is not required in patients with clear radiological signs of PDAC and DCC as this is also not considered mandatory in current clinical practice in patients who do not receive neoadjuvant treatment. Hence, requiring a preoperative biopsy for the current study is regarded as unethical.

Both patients with PDAC and DCC are included as both these tumour types are associated with a relatively high R1-resection rate and are often challenging to distinguish from each other prior to surgery [24, 25]. Even in prospective randomised controlled trials only including patients with pathologically proven PDAC, up to 10% of patients are ultimately diagnosed with DCC or another diagnosis [26–28].

Patients receiving neoadjuvant treatment are eligible for the present study, as long as the tumour is upfront resectable without vascular involvement prior to the start of neoadjuvant treatment.

All variants of PD wherein the Da Vinci® (or other) robotic platform is used can be included in the intervention arm of the DIPLOMA-2x2 trial. This includes robot-assisted, robotic, and ‘roboscopic’ procedures, provided that both the resection and reconstruction phase are performed robotically. The start of the procedure, i.e. diagnostic laparoscopy and the initial step of mobilization, is allowed laparoscopically. Details regarding these differences will be registered.

Exclusion criteria

Patients meeting any of the following criteria are excluded from participation:

• Indication for elective PD for a proven or suspected non-malignant neoplasm in the pancreatic head and periampullary region (e.g. intraductal papillary mucinous neoplasm, neuroendocrine tumour, ampullary cancer, duodenal cancer);

• A second primary malignancy requiring resection during the same procedure.

• History of chronic or necrotizing pancreatitis, including post-endoscopic retrograde cholangiopancreatography (ERCP) necrotizing pancreatitis;

• Any vascular involvement with the mesenteric, celiac trunk and hepatic arteries and the portomesenteric vein;

• Distant lymph node involvement on preoperative imaging;

• Pregnancy;

• Body mass index (BMI) > 35 kg/m²;

• Participation in another study with interference of study outcomes;

• Hybrid procedures in which the resection is performed via the laparoscopic or robot-assisted approach and the reconstruction is performed via the open approach.

Recruitment

Eligible patients are identified and approached at the outpatient clinic of participating centres by the local treating surgical team. Written informed consent is obtained prior to randomization.

Randomization

After inclusion, all patients are randomised centrally by the study coordinators using an online computer controlled permuted-block randomization module (Castor EDC, CIWIT B.V., Amsterdam, the Netherlands). Patients are randomly allocated to RPD or OPD in a 2:1 ratio with varying block sizes of 3, 6, and 9 patients. This unequal randomization provides more statistical power for detecting adverse events in the RPD group and allows for maintenance of a higher annual RPD volume of the participating surgeons during the study. The latter is perceived as relevant for patient safety as RPD volume is strongly associated with patient outcome. Stratification is performed for indication (i.e. proven PDAC) and preoperative risk for pancreatic fistula (high versus low, based on BMI > 25 kg/m² and main pancreatic duct size < 3 mm). Randomised patients with disease progression on repeated imaging before surgery, who are no longer eligible for RPD and do not meet the inclusion criteria, are excluded and replaced with newly included patients.

Data management

Patients are coded by a unique numeric randomization code to which only the study coordinators have access. Clinical data are collected using standardised electronic case report forms in Castor EDC (CIWIT B.V., Amsterdam, the Netherlands). Source data are stored digitally at participating centres and retained by the project leader for 15 years after all patient follow-up is completed in accordance with Dutch regulatory requirements and Good Clinical Practice guidelines.

Surgical technique and postoperative management

Because of the pragmatic design of the DIPLOMA-2x2 trial, no specific standard surgical technique in RPD and OPD is required. Pancreatoduodenectomy includes both a resection and reconstruction phase. During the resection phase, the pancreatic head, duodenum, distal bile duct, gall bladder, and if required, the pylorus and distal stomach are resected. During the reconstruction phase, a pancreaticojejunostomy, hepaticojejunostomy, and gastrojejunostomy (or duodenojejunostomy in case of pylorus preservation) are performed. Up to two abdominal drains are placed near the anastomoses. The procedure is performed by one or two surgeons. Each surgeon decides on the intraoperative technical strategy, which will be specified in the online case report forms.

All patients receive the same protocolized postoperative care, based on enhanced recovery principles including very early mobilization and oral intake [29]. The treating team specifies the postoperative management in the online case report forms. There are no restrictions regarding postoperative care, including blood tests, drain management, the use of medication, and other interventions.

Conversion from RPD to OPD

Any incision used for other reasons than trocar placement and/or specimen extraction is defined as a conversion. Patients allocated to RPD but converted to OPD are analysed in the RPD group in the modified intention-to-treat analysis (primary) and in the OPD group in the as-treated analysis (secondary). Converted patients will be excluded from the per-protocol analysis. Reasons for conversion are registered and categorized as urgent or non-urgent conversions [30].

Blinding

In the DIPLOMA-2x2 trial, blinding of the primary outcome assessors (pathologists) is implemented to minimize assessment bias. Surgeons are instructed to omit any reference to the surgical approach when submitting specimens for pathological evaluation. Pathologists do not routinely require access to the electronic patient record and could therefore be blinded to the surgical approach during specimen assessment. In certain instances, pathologists may need to consult the surgical report within the patient file, which could risk unblinding. To monitor the integrity of blinding, pathologists report in the online pathology case report form whether they were aware of the surgical approach during specimen evaluation, and if so, for what reason. Additionally, if successfully blinded, pathologists indicate the suspected surgical approach using a five-point scale: (1) strongly believe it was OPD, (2) somewhat believe it was OPD, (3) somewhat believe it was RPD, (4) strongly believe it was RPD, and (5) do not know.

Furthermore, to reduce bias regarding the key secondary outcome ‘time to functional recovery’, patients are blinded for treatment allocation until 5 days after surgery. Directly after skin closure, while the patient is still under general anaesthesia, the patient receives a firmly taped, 40 × 40 cm abdominal dressing to cover all incision(s) including the Pfannenstiel extraction site, ensuring that the treatment allocation remains unknown (robot-assisted versus open). This abdominal dressing is removed on postoperative day 5, or earlier if all criteria for functional recovery are met or for medical reasons, such as suspected wound infection. If earlier inspection is required, attempts are made to maintain patient blinding. This blinding has proven to be successful in previous multicentre randomised controlled trials [18, 31–33]. The success of blinding in the current study is assessed using the blinding index as proposed by Bang et al. [34]. Patients are asked about the alleged treatment allocation on postoperative day 2 and before removal of the abdominal dressing, based on five categories: (1) strongly believe it was RPD, (2) somewhat believe it was RPD, (3) do not know, (4) somewhat believe it was OPD, and (5) strongly believe it was OPD. Patient blinding is not performed in patients who are intraoperatively diagnosed with irresectable disease, including metastasised disease, due to ethical reasons.

Primary outcome

The primary outcome is the pathological R0-resection (pR0) rate (i.e. microscopically radical resection with >1 mm tumour clearance at the surgical resection margins and anatomical surfaces, except for the anterior surface in which 0 mm clearance applies), according to the recommendations of the International Collaboration on Cancer Reporting (ICCR) and Royal College of Pathologists criteria [35–38]. The R0-resection rate is recorded as a binary outcome based on the definitive postoperative histopathological examination. The R0-resection rate is used as a surrogate outcome measure for oncological safety since a sample size powered for overall survival is not feasible. This histopathological assessment includes both the surgical margins (pancreatic transection plane, common bile duct, proximal duodenal/gastric and distal duodenal/jejunal margins, and superior mesenteric artery dissection margin) and the anatomical surfaces (anterior and posterior pancreatic head, and superior mesenteric/portal vein impression surface), according to the new International Study Group for Pancreatic Surgery (ISGPS) and International Study Group for Pancreatic Pathology (ISGPP) recommendations [39]. Local pathologists from the participating centres report the distance from tumour to each surgical margin and each anatomical surface to ensure uniformity in R0 definition. Additionally, a validation is performed by central review (i.e. random number list) in 10% of the specimens. Specimens retrieved from DIPLOMA-2 patients who will be included in the DIPLOMA-2x2 trial are re-examined by blinded pathologists.

Secondary outcomes

The key secondary outcome is postoperative time to functional recovery in days after surgery, reported as mean ± standard deviation (SD) and median with interquartile range (IQR). Functional recovery is reached when all of the following criteria are met: I. adequate pain control with oral analgesia only, II. restoration of mobility to an independent level (or to preoperative level if previously impaired), III. ability to maintain sufficient caloric intake (minimum of 50% required calories), IV. absence of intravenous fluid administration, and V. no signs of active infection (no fever, decreasing C-reactive protein below 150 mg/L) [31, 32]. Assessment of this outcome is performed locally at each participating centre by the clinical treatment team.

Other secondary outcomes include other pathological outcomes (e.g. histopathological diagnosis and tumour differentiation [categorical], lymphovascular invasion including venous invasion, perineural invasion, vascular involvement, and multivisceral involvement [binary], and lymph node retrieval and involvement [numerical; reported as mean ± SD or median with IQR, as appropriate]) and oncological outcomes (e.g. receipt and completion of adjuvant therapy and pathologically proven disease recurrence [binary], time to start adjuvant therapy [days; reported as mean ± SD or median with IQR], and disease-free and overall survival [1-, 2-, 3-, 5-year and median]). Additionally, intraoperative outcomes (e.g. operative time and blood loss [reported as mean ± SD or median with IQR], and conversion [binary]), postoperative outcomes up to 90 days after surgery (e.g. pancreatic surgery-specific complications according to the ISGPS and International Study Group for Liver Surgery [ISGLS] definitions, and mortality; binary), health-care resource utilization up to 90 days after surgery (e.g. length of stay [days; reported as mean ± SD or median with IQR], reinterventions, readmissions, and intensive care unit (ICU) stay [binary]), and quality of life (ranging from 0 ‘death’ to 1 ‘full health’) up to 60 months after surgery are assessed.

Data collection and patient follow-up

Baseline characteristics include age (years; reported as mean ± SD or median with IQR, as appropriate), sex (male or female), BMI (kg/m², reported as mean ± SD or median with IQR), previous abdominal surgery, preoperative diabetes mellitus and preoperative biliary drainage (binary), performance status (Karnofsky score, ranging from 0% ‘death’ to 100% ‘normal, no complaints or evidence of disease’ [40] and American College of Anaesthesiologists [ASA] score, ranging from 1 ‘healthy patients’ to 4 ‘patients with severe systemic disease that is a constant threat to life’), preoperative diagnosis (categorical), tumour size (mm; reported as mean ± SD or median with IQR), main pancreatic duct size (mm; reported as mean ± SD or median with IQR) and organ involvement on preoperative imaging (binary), serum levels of HbA1c (mmol/ml; reported as mean ± SD or median with IQR), CA 19.9 (kU/L; reported as mean ± SD or median with IQR) and CEA (µg/L; reported as mean ± SD or median with IQR), preoperative postoperative pancreatic fistula (POPF) risk (categorical), neoadjuvant treatment (binary), and baseline quality of life scores will be presented in a table of baseline characteristics. All baseline measurements are captured before randomization. All required clinical data are collected after randomization (i.e. from hospitalization up to 60 months postoperatively) by the local treating team using standardized online case report forms and crosschecked with source data by the study coordinators at 3 and 60 months after the last patient has undergone surgery. For quality of life measurements, the validated EQ-5D-5L, QLQ-C30, and PAN-26 questionnaires are sent (electronically or on paper) to the patients at baseline and 1, 3, 6, 12, 36, and 60 months after surgery [41, 42]. Patients also receive an additional questionnaire at 6 months after surgery focusing on potential scar-related complications and body image concerns. Patients are followed at the outpatient clinic according to the local protocol (including CT-scan and serum levels of CA 19.9 and CEA). Because the intervention consists of a single surgical procedure rather than a prolonged course of treatment, no specific strategies to promote or monitor adherence were implemented. See Fig. 2 for the schedule of interventions and assessments.

Fig. 2.

Schedule of enrolment, interventions, and assessments according to SPIRIT

Quality requirements for participation

As in the DIPLOMA-2 trial, participating centres have a minimum annual volume of 30 RPD and participating surgeons a personal experience of at least 60 RPD and 60 OPD prior to trial initiation. The number of included patients per centre cannot exceed 50% of the total number of required patients. Overall, 13 of 14 centres that participated in the DIPLOMA-2 trial continue patient inclusions within DIPLOMA-2x2, and six new centres joined DIPLOMA-2x2. All RPD surgeons are required to have participated in an endorsed RPD training programme, or alternatively, submit a recorded and anonymized video of a personal RPD procedure for evaluation by the study team before participation in the trial.

Safety monitoring and adverse events reporting

All adverse events are recorded up to 90 days postoperatively. Serious adverse events (SAEs) are reported to the Dutch central committee on research involving human subjects (in Dutch: Centrale Commissie Mensgebonden Onderzoek [CCMO]) and the institutional review board (Medical Ethics Committee of Amsterdam UMC) via a web portal (www.ccmo.nl/onderzoeksportaal). The following SAEs are reported to the study coordinators: unplanned intensive care unit admission, reoperation, readmission (within 48 h), and death (within 24 h), regardless of cause. These safety outcomes were chosen as they correspond to the most severe and clinically meaningful adverse events following pancreatic surgery, reflecting complications that have the greatest impact on patient recovery. The remaining adverse events are recorded in a yearly overview list.

Trial oversight and ethics

Oversight of trial conduct, safety, and compliance is provided by an independent Data Safety Monitoring Board (DSMB) and the relevant ethics committees. The independent DSMB is appointed to evaluate the study safety parameters. The DSMB consists of one independent statistician, one independent methodologist, one independent medical oncologist, and three independent surgeons. The DSMB meets (online) after every 50 patients have completed the 90-day follow-up to evaluate safety parameters. The study coordinators and principal investigators are only present during the start of the DSMB meeting (open discussion) to provide the overall data and background information. The minutes of these meetings are sent to the institutional review board by the study coordinators and the trial steering committee. The DSMB is not blinded and is fully informed on all SAEs. The DSMB can request a full report of specific study outcomes whenever required.

The DIPLOMA-2x2 trial is conducted according to the principles of the Declaration of Helsinki (64th version, October 2013) and in accordance with the local laws and regulations, such as the Medical Research Involving Human Subjects Act in the Netherlands. The local principal investigators are responsible to adhere to local laws and regulations. The independent ethics review board of the Amsterdam UMC (Amsterdam, the Netherlands; NL77750.018.21) and Provincial Ethics Committee Brescia (Brescia, Italy; NP4916) approved the study protocol in September 2023. Furthermore, approval from all local ethics committees of participating centres is obtained before participation in the study. All local ethics committees of participating centres will be kept informed about important protocol modifications.

The study will be monitored by a clinical research associate in the Amsterdam UMC and in centres requiring such monitoring. Monitoring will include periodic review of trial conduct, verification of informed consent, and source data checks to ensure compliance with the protocol, Good Clinical Practice, and applicable regulations. The trial is registered in the ISRCTN registry: ISRCTN27483786.

Statistical aspects

Sample size calculation

The DIPLOMA-2x2 trial is designed as a non-inferiority trial, hypothesizing that the R0-resection rate with RPD is non-inferior to OPD in patients with cancer in the pancreatic head. Based on retrospective data from the Dutch Pancreatic Cancer Audit, including patients undergoing RPD and OPD and meeting the inclusion criteria for the DIPLOMA-2x2 trial, the sample size is calculated for pR0-resection rate with the following assumptions: the one-sided significance level (α) is set at 2.5% and power (1 − β) at 80%, expected R0-resection rate of 54% with OPD and 62% with RPD, non-inferiority margin of 7%, 5% cross-over from RPD to OPD (conversion), 5% metastasized disease, requiring a total of 413 patients to be randomised (275 RPD and 138 OPD) [7]. Based on this sample size, a two-sided Gehan–Breslow–Wilcoxon test in the intention-to-treat (ITT) population (N = 404, 269 RPD and 135 OPD) achieves 90% power with 413 patients under the same assumptions with a two-sided significance level of 5% to detect a 33% reduction of time to functional recovery (i.e. 6 days after RPD and 9 days after OPD; hazard ratio = 1.50), corresponding to a reduction in the median time to functional recovery of 3 days. The Gehan–Breslow–Wilcoxon test is stratified for the same stratification factors used in the randomization procedure (i.e. indication and preoperative risk for pancreatic fistula). The sample size calculation was performed in R, version 4.5.2 using the npsurvSS package.

Statistical analysis

Primary and secondary outcome data of all randomised patients are analysed based on four analysis sets. The ITT population comprises all patients in the group to which they were randomised (i.e. converted patients remain in the RPD group), and serves as the primary analysis set. The primary outcome, pR0-resection rate, can only be determined in patients who underwent resection. Therefore, the primary outcome analysis is performed in the modified ITT (mITT) analysis, excluding patients who did not undergo resection (i.e. patients who are intraoperatively diagnosed with irresectable disease). As such, this analysis targets a principal-stratum estimand. All other outcomes, including the time to functional recovery, target a treatment-policy estimand and are analysed according to the ITT principle, including patients who did not undergo resection. The per-protocol set consists of all patients treated per protocol without major protocol violation and conversion. In addition, the as-treated set is analysed with patients in the group in which they were finally treated (i.e. converted patients in the OPD group). Patients who were excluded prior to surgery are excluded from all analyses as they were replaced according to protocol.

The primary outcome measure ‘R0-resection rate’ will be presented as frequencies (percentages) and the difference between the two groups with corresponding one-sided 97.5% confidence interval will be tested for non-inferiority using the Miettinen–Nurminen test, stratified for the stratification factors used in the randomization process [43]. The lower limit of the one-sided 97.5% confidence interval of the difference in proportions will be reported and compared with the 7% non-inferiority margin. The key secondary outcome ‘time to functional recovery’ will be measured in days after surgery and tested for superiority using the Gehan–Breslow–Wilcoxon test if the primary endpoint shows non-inferiority. This hierarchical testing procedure is intended to control the overall type I error rate. Additionally, the individual functional recovery criteria will be analysed separately.

The primary analysis for non-inferiority will be repeated in the per-protocol population. In addition, in supplementary analyses accounting for between-centre differences, the primary outcome will be compared between the randomization arms in the mITT population using generalized estimating equations models and a Mancl–DeRouen variance estimator, adjusted for the stratification factors and with centre as the clustering variable, followed by regression standardization to estimate the difference in proportions. Supplementary analyses for time to functional recovery will account for between-centre differences using a stratified Gehan–Breslow–Wilcoxon test, with stratification by centre.

The distribution of outcomes will be determined using several plots (boxplot, Q-Q plot, and histogram). Normally distributed continuous variables will be presented as means (standard deviations) and will be compared using linear regression models (adjusted for the stratification factors) to estimate the mean difference and corresponding 95% CI. Continuous non-normally distributed variables will be expressed as medians (IQRs) and will either be compared using bootstrapping to estimate the difference in medians or using linear regression models (adjusted for the stratification factors) after log-transformation for right-skewed variables. Categorical variables will be presented as frequencies (percentages) and compared using logistic regression models (adjusted for the stratification factors) to estimate odds ratios and 95% CIs.

Time-to-event outcomes, such as survival outcomes and time to adjuvant therapy, will be estimated using the Kaplan–Meier method. An exploratory Cox regression analysis will be performed to investigate predictors of postoperative survival. All parameters with a p value < 0.1 in a univariable analysis will be included in the multivariable Cox regression analysis. Additional exploratory multivariable analyses will be performed to identify predictors for the outcomes R0-resection rate and time to functional recovery. Furthermore, predictors for secondary outcomes, for example, for receiving adjuvant chemotherapy, will be assessed. A two-tailed p value < 0.05 will be considered statistically significant. Time-to-event outcomes with death as a competing risk will be analysed using cause-specific Cox regression and will be visualized as cumulative incidence curves using the Aalen–Johansen method.

For participants with missing data on the primary and key secondary outcomes, a sensitivity analysis will be performed to determine best-case and worst-case scenarios.

The following pre-specified sensitivity analyses are performed:

• Comparing RPD and OPD for R0-resection rate of ‘surgical margins’ only (pancreatic transection plane, common bile duct, proximal duodenal/gastric and distal duodenal/jejunal margins, and superior mesenteric artery dissection margin) [39];

• Comparing RPD and OPD for time to functional recovery excluding patients who did not undergo resection (e.g. due to intraoperative diagnosed metastasized disease). Comparing RPD and OPD for time to functional recovery excluding patients who were not blinded.

Additionally, the following pre-specified subgroup analyses are performed:

• Comparing RPD and OPD in patients with a preoperative diagnosis of PDAC and DCC separately;

• Comparing RPD and OPD in patients with PDAC who received neoadjuvant therapy and in patients with PDAC who did not receive neoadjuvant therapy;

• Comparing RPD and OPD in patients with major complications (Clavien–Dindo grade ≥ III) and without major complications (Clavien–Dindo grade 0–II) separately;

• Comparing RPD and OPD in centres with a low annual RPD volume (<20 RPD) and high annual RPD volume (≥20 RPD) during the trial separately;

• Comparing RPD and OPD for the incidence of POPF grade B/C in patients with a POPF risk, according to the ISGPS 3-tier classification [44, 45]:

  • ◦
    A—not soft (hard) pancreatic texture and main pancreatic duct >3 mm
  • ◦
    B—soft pancreatic texture or main pancreatic duct ≤3 mm
  • ◦
    C—soft pancreatic texture and main pancreatic duct ≤3 mm

Interaction tests will be used to compare RPD and OPD in the different subgroups.

Additionally, a cost-effectiveness analysis and assessment of quality of life will be performed using data up to 6 months postoperatively. For this predefined subanalysis, three analyses will be performed for (1) total hospital costs, (2) quality-adjusted life-years (QALYs; i.e. 1 year in perfect health), and (3) cost-effectiveness (i.e. the incremental cost-effectiveness ratio [ICER]). First, the mean total hospital costs (euro) of RPD versus OPD will be analysed using the t-test with non-parametric, bias-corrected and accelerated bootstrapping and presented as means with bias-corrected and accelerated 95% confidence intervals. Second, health-related quality of life (HR-QoL) and QALYs will be calculated using the EuroQol Five Dimensions Five-Level (EQ-5D-5L) health utility index scores, ranging from 0 ‘death’ to 1 ‘full health’. Missing HR-QoL data at baseline will be handled with mean imputation; missing HR-QoL data after baseline will be handled using substantive model-compatible fully conditional specification and will target a while-alive estimand [46–48]. Third, the cost-effectiveness will be estimated based on the difference in total hospital costs divided by the difference in QALYs between approaches. Cost-effectiveness will be reported as the probability that the ICER falls below the willingness-to-pay threshold (i.e. the probability of cost-effectiveness of RPD versus OPD per additional QALY).

The statistical analyses will be performed in R, version 4.5.2 or higher using the boot, glmtoolbox, marginaleffects, metalite.ae, rms, smcfcs, and survival packages [47, 49, 50].

Details on the statistical analysis are reported in the study protocol (version 3.2, October 2025; section 11, page 54–56). Additionally, a detailed statistical analysis plan for the short-term outcomes (up to 90 days postoperatively) is drafted prior to database lock. Despite all prior preventive measures taken, a complex international trial may evoke unforeseen situations after database lock that threaten data integrity and can only be resolved by unlocking the database prior to the final analysis. For the purpose of transparency and reproducibility, the statistical analysis plan therefore also describes the procedure to be followed when such situations occur.

Dissemination policy

The results of this trial are submitted to a peer-reviewed medical journal regardless of the study outcome. The results of the DIPLOMA-2x2 trial will be reported in at least two separate main publications. The first will focus on short-term outcomes (e.g. R0-resection, complications, and functional recovery) up to 90 days after surgery, while the second will address long-term oncological outcomes (e.g. disease-free and overall survival, and quality of life up to 60 months). Authorship is based on the International Committee of Medical Journal Editors (ICMJE) guidelines. Those involved in the study who do not fulfil these criteria are listed as ‘collaborator’.

Discussion

The DIPLOMA-2x2 trial is an international, multicentre, patient- and assessor-blinded randomised controlled trial designed to evaluate the oncological safety of RPD in patients with upfront resectable cancer in the pancreatic head without vascular involvement. As a roll-over of the DIPLOMA-2 trial, this study is performed in high-volume centres with experienced surgeons beyond the learning curve.

To date, the only available high-quality evidence comparing RPD and OPD came from randomised controlled trials in patients with both benign and malignant pancreatic head and periampullary tumours [9, 10, 18]. While these previous studies have demonstrated important short-term benefits of RPD, such as reduced blood loss, shorter hospital stay, and faster time to functional recovery, these trials did not have sufficient power to draw definitive conclusions about oncological outcome after RPD versus OPD. Therefore, the oncological safety specifically in patients with cancer in the pancreatic head remains unknown [9, 10, 18].

The choice of a non-inferiority design for the primary outcome R0-resection reflects the realistic clinical hypothesis that RPD is not inferior to OPD in achieving surgical radicality. While overall survival remains the most definitive measure of oncological efficacy, powering a trial for this outcome would require an unfeasibly large sample size, likely exceeding 10,000 patients. In contrast, R0-resection is chosen as a pragmatic and clinically relevant surrogate outcome measure, as achieving microscopically radical resection is associated with improved survival of cancer in the pancreatic head as demonstrated in multiple studies [51–55]. However, the reported R0-resection rates vary widely, likely due to differences in definitions, specimen handling, and pathological assessment protocols [37, 38, 51].

To address these limitations, the DIPLOMA-2x2 trial includes a centralized and standardized pathology reporting protocol, developed in collaboration with the participating pathologists. Although specimen assessment was performed per local protocol to reflect real-world practice, critical components of specimen handling (e.g. R0-resection definitions and reporting format) were harmonized to ensure comparability of outcomes in a multicentre setting [51]. The external validation of a subset of specimens further enhances consistency and ensures uniformity in R0 resection definition.

By building on the infrastructure of the DIPLOMA-2 trial, DIPLOMA-2x2 is expected to feature efficient patient enrolment, minimising time loss, costs, and redundant patient randomization [18]. If oncological non-inferiority is demonstrated, potential advantages of RPD related to recovery, quality of life, and cost-effectiveness may further justify the adoption of RPD in high-volume centres.

Potential limitations include the inherent challenges of blinding in surgical trials, the risk of bias introduced due to conversion from RPD to OPD, variations in the use and regimen type of (neo)adjuvant chemotherapy across countries, and the use of R0-resection rate as a surrogate outcome for overall survival. However, these limitations are mitigated through robust trial design elements, including the (m)ITT, per-protocol, and as-treated analyses, blinding of patients and pathologists, standardized pathological reporting, and predefined subgroup and sensitivity analyses.

Conclusion

The DIPLOMA-2x2 trial is an international, patient- and assessor-blinded randomiSed controlled trial, designed to investigate oncological outcome after RPD versus OPD in terms of R0-resection rate in patients with upfront resectable cancer in the pancreatic head without vascular involvement. By providing level 1 evidence on oncological outcomes, this trial aims to guide appropriate patient selection, and safe and effective implementation of RPD. If oncological non-inferiority is confirmed, the findings support wider adoption of RPD in high-volume centres, ultimately improving postoperative recovery and long-term outcomes for patients with cancer in the pancreatic head.

Trial status

Confirmation of funding of the trial by Intuitive Foundation (Sunnyvale, USA) and the Dutch Cancer Society (KWF) was received on June 22 and July 16, 2024, respectively. Ethical approval of the DIPLOMA-2 trial and the DIPLOMA-2x2 amendment by the primary review committee (Amsterdam UMC) was received in December 2021 and September 2023, respectively. The most recent version of the study protocol is version 3.2 of October 9, 2025. The DIPLOMA-2 trial and the DIPLOMA-2x2 update were registered in the ISRCTN registry on January 11, 2021, and February 6, 2025, respectively (ISRCTN 27483786). The first DIPLOMA-2 patient underwent surgery on January 3, 2021. The first DIPLOMA-2x2 patient was recruited on December 12, 2023, and underwent surgery on December 19, 2023. At the time of submitting this protocol for publication (November 6, 2025), 17 centres were actively recruiting patients for the trial and 351 out of 413 patients (85%) were randomised, of which 341 of 413 patients (83%) underwent surgery. Recruitment is expected to be completed by May 2026.

Abbreviations

ASA

American Society of Anaesthesiologists

BMI

Body mass index

CCMO

Centrale Commissie Mensgebonden Onderzoek

DCC

Distal cholangiocarcinoma

DSMB

Data Safety Monitoring Board

ICU

Intensive care unit

ICMJE

International Committee of Medical Journal Editors

ISGLS

International Study Group of Liver Surgery

ISGPS

International Study Group of Pancreatic Surgery

ITT

Intention-to-treat

mITT

Modified intention-to-treat

MIPD

Minimally invasive pancreatoduodenectomy

OPD

Open pancreatoduodenectomy

PD

Pancreatoduodenectomy

PDAC

Pancreatic ductal adenocarcinoma

POPF

Postoperative pancreatic fistula

R0-resection

Microscopically radical resection

RPD

Robot-assisted pancreatoduodenectomy

SAE

Serious adverse event

Authors’ contributions

NdG, CLB, and JEM drafted the manuscript. NdG, CLB, JEM, AE, JvH, MA, HB, PM, MAH, and MGB participated in the design of the trial during several meetings. NdG, CLB, JEM, MA, HB, PM, MAH, and MGB performed the sample size calculation. CLB, JEM, MAH, and MGB primarily coordinate the trial. All authors and collaborators critically reviewed the manuscript, approved the final version, and are fully aware of this publication.

Funding

The research team received a private donation and research grants from Intuitive Foundation (Sunnyvale, CA) and the Dutch Cancer Society (KWF) for the investigator-initiated DIPLOMA-2x2 trial. The funders had no influence on the study design and will not have an impact on the data collection, interpretation of data, manuscript, and the decision to publish.

Data availability

Not applicable.

Declarations

Ethics approval and consent to participate

The medical ethics review committee of the Amsterdam UMC (Amsterdam, the Netherlands; NL77750.018.21) and Provincial Ethic Committee Brescia (Brescia, Italy; NP4916) reviewed and approved the DIPLOMA-2x2 trial. The trial is registered in the ISRCTN registry with the identification number ISRCTN27483786. Patients can only participate if written informed consent has been provided.

Consent for publication

Not applicable.

Competing interests

The authors have no competing interests to declare.