We conducted a retrospective study in one high-volume expert center in France (Beaujon Hospital, Clichy, France). We studied the medical records of all consecutive patients with pathologically proven PDAC who were operated on by pancreaticoduodenectomy (PD) following NAT between 01/01/2016 and 01/01/2022. NAT consisted of NAC using FOLFIRINOX, followed or not by NACR. We only included patients with borderline resectable or locally advanced but potentially secondarily resectable PDAC. Patients who received another type of NAC other than FOLFIRINOX and those with other concomitant active malignancies were not included. We excluded patients with early (<1 month) postoperative death, unavailable postoperative data, non-measurable tumor at baseline, and/or who performed baseline CT scan more than 1 month before the first cycle of NAC.

All therapeutic indications were decided during dedicated weekly multidisciplinary tumor boards, following national guidelines. ¹⁴ The clinical tumor stage, that is, borderline resectable or locally advanced but potentially secondarily resectable, was determined on contrast-enhanced CT scan done at initial diagnosis. The absence of liver metastases was assessed on systematic liver MRI, including diffusion-weighted sequences.

One cycle of FOLFIRINOX consisted of a 2-h infusion of oxaliplatin (85 mg/m²) followed by a 2-h infusion of leucovorin (400 mg/m²) concomitantly with a 90-min infusion of irinotecan (150 mg/m²), followed by a 46-h continuous infusion of 5-fluorouracil (2400 mg/m²), every 2 weeks. The decision to administer additional NACR was made during a multidisciplinary team meeting (involving oncologists, surgeons, and radiation oncologists), when imaging following NAC showed persistent vein encasement or deformation, or arterial contact, suggesting that an R0 resection would only be achievable following further treatment. When done, NACR consisted of 54 Gy delivered in 30 daily fractions over 6 weeks with concurrent capecitabine at 800 mg/m² twice daily on days of radiation therapy. The planned target volume was determined on a dedicated simulation contrast-enhanced CT as previously described. ¹⁵ A CT scan was performed within 1 month before NAC initiation and after every six cycles, and 4–6 weeks after NACR if performed. Patients underwent standardized clinical and biological evaluation before every chemotherapy cycle and weekly during NACR.

Surgical resection was decided in a tumor board meeting in patients with clinical (no pain and weight gain) and biological (normalization or CA 19-9 < 200) response, and without radiological progression. All patients had a CT scan, MRI, and, if needed, a PET scan within 1 month before surgery. Venous resection was frequently planned except in patients who had invasion of superior mesenteric vein collaterals and/or complete stenosis of the venous axis with cavernous circulation. In patients with superior mesenteric artery invasion, surgery was considered only in patients with a short segment (<2 cm), lateral invasion (<180), without arterial stenosis, and in whom liberation of the SMA can be done without resection. No case of SMA resection was done. Surgery was done 4–8 weeks after the end of radiotherapy, and laparoscopic exploration was considered in some patients with doubtful extra-pancreatic disease despite complete preoperative exploration. Surgical resection was planned preoperatively, and a few patients had arterial sheath biopsies to rule out tumoral invasion. Some selected cases were operated on by the laparoscopic approach.

Postoperative follow-up consisted of clinical, biological, and CT examinations at least every 3 months for the first 2 years and every 6 months thereafter. All patients with compatible clinical conditions received adjuvant chemotherapy for 3–6 months, as decided during multidisciplinary tumor boards in accordance with national guidelines, starting within 12 weeks after the date of surgery. ¹⁴

We collected all relevant epidemiological, tumor, treatment, and follow-up de-identified data on a predefined standardized electronic chart. Postoperative complications (within 3 months) were scored according to the Clavien-Dindo classification. ¹⁶

Morphological data (PDAC size, initial stage, response, and recurrence) were reassessed centrally on CT scans by two experienced radiologists who were blinded from the other data. Tumor response after NAT (i.e., after NAC and after NACR) was measured according to RECIST 1.1 (as a complete response (CR), partial response [PR], or stable disease [SD]). ¹¹ Objective response refers to either CR or PR. Pathological features (ypTNM stage, R status, response) were assessed centrally by an experienced pathologist, blind to the other data. The pathological response was assessed according to CAP score, as CAP0 (complete response), CAP1 (near complete response: single/rare group(s) of cancer cells), CAP2 (partial response, residual cancer with regression), or CAP3 (poor response, no tumor regression). ¹⁷ The proportion of viable tumor cells (VTC) was measured on the whole specimen by two experienced pathologists.

The study protocol conformed to the ethical guidelines outlined in the Declaration of Helsinki. As a retrospective, monocentric observational study, and in accordance with French legislation, it was exempt from institutional review board approval.

Continuous variables were described as medians with their 25th–75th interquartile range (IQR) and compared using the Mann–Whitney test. Categorical variables were expressed as frequencies and percentages and compared using the Chi² or Fisher’s test.

The primary endpoint was RFS, which was measured from the time of surgery to the date of recurrence. The secondary endpoint was OS, measured from the time of surgery to the date of death. Patients alive without event at their last follow-up were censored at that date. Median RFS and OS were estimated using the Kaplan–Meier method and were compared using the log-rank test.

The association between the proportion of VTC and the 2-year RFS rate was explored using the area under the receiver-operating curve (AUROC) and its 95% confidence interval. The best threshold was determined using the Youden method. The interobserver reproducibility of this threshold (i.e., classification of VTC < 30% or ⩾30%) was determined by computing Cohen’s kappa coefficient on a random sample of 15 patients. Variables associated with RFS, and then those associated with OS, were explored using Cox proportional hazards backward stepwise multivariable regression models, which included non-collinear, clinically relevant variables. Values of p < 0.05 were considered statistically significant. All statistical tests were bilateral. Statistical analyses were performed with the Prism software (version 10, Graphpad).

The reporting of this study conforms to the STROBE guidelines ¹⁸ (Supplemental File: STROBE checklist).

During the study period, 154 patients were operated on by PD for a borderline resectable or locally advanced PDAC after receiving FOLFIRINOX NAC in our institution. Among them, 63 patients were excluded because of inappropriate pre-therapeutic CT scan (non-measurable lesion or too long delay before NAC initiation), absence of postoperative follow-up data, or early postoperative death (Figure 1).

The characteristics at the time of surgery of the 91 patients included are summarized in Table 1. Briefly, the median age was 65.5 years, and the sex ratio was balanced. As expected, most patients were in good general condition with low performance status (PS) and ASA score. Most patients (78%) needed endoscopic biliary stenting to treat obstructive jaundice. At initial staging, most patients (87%) had borderline resectable PDAC, whereas 13% of patients had locally advanced PDAC due to arterial involvement.

Before surgery, all 91 patients had received FOLFIRINOX NAC, consisting of an average of 6 cycles (range 4–13; Table 2). Half of them did not receive 5-fluorouracil bolus (mFOLFIRINOX regimen), and the dose of oxaliplatin and/or irinotecan was reduced in 45% and 35% of patients, respectively. In comparison with the initial CT scan, the post-NAC CT scan identified morphological response (CR+PR) in 21 patients (23.1%) and SD in the 70 others (Figure 2).

Then, 77 patients (85%) received additional NACR as decided during tumor board, to improve resection margins ¹⁰ (Table 2). NACR yielded a further decrease in tumor size in 37 patients (48%, green plots in Figure 2) and an increase in tumor size in 4 patients (5%, red plots in Figure 2). In comparison with the initial CT scan, the post-NACR CT scan identified objective response in 30/77 patients (39%), including 14 patients (18%) who had not achieved objective response after NAC. Overall, NAT (NAC ± NACR) yielded objective morphological response in 35/91 patients (38%).

Surgery was performed a median of 62 days (IQR, 48–73) after the end of NAT. PD was associated with porto-mesenteric vein reconstruction in 60% of cases (Table 2). Severe (Clavien-Dindo grade III/IV) postoperative complications occurred in 13 patients. At the histopathological examination of the resected specimens, most PDAC were classified as well (56%) or moderately (37%) differentiated. Tumor stage mostly consisted of ypT1 or ypT2 (84%), and 29% were associated with lymph-node metastases (median lymph-node ratio 0.11, IQR [0.05–0.16]). Overall, 22 patients (24%) had a pathological complete (CAP0) or nearly complete (CAP1) response. The median proportion of VTC was 30% (IQR, 18–60).

Following surgery, 79 patients (87%) received adjuvant chemotherapy, which was initiated a median of 61 days (IQR, 49–73) after surgery (Table 2). Adjuvant chemotherapy mostly consisted of LV5FU2 (53%) or mFOLFIRINOX (23%). 13% of patients did not receive adjuvant chemotherapy due to a general health status incompatible with the administration of adjuvant treatment within the required timeframe (12 weeks).

The median postoperative follow-up was 32.1 months (95% CI (28.6–35.5)). During this follow-up, 51 patients (56%) had recurrence. The first site of recurrence was the liver, lungs, the peritoneum, local, and cutaneous in 20, 16, 11, 3, and 1 patients, respectively. Median RFS was 26.2 months (IC 95% (22.3–30.1)).

There was no statistically significant association between the percent of tumor size evolution after NAT (as depicted in Figure 2) and the risk of recurrence at 2 years (AUROC 0.565, 95% CI (0.44–0.69), p = 0.31). Accordingly, RFS was similar in patients with or without morphological response (partial or complete) after NAT (p = 0.62; Figure 3(a)).

There was no statistically significant association between the CAP score and the risk of recurrence (Figure 3(b)). However, the CAP score was prognostic for RFS when used as a binary variable (CAP0-1 vs CAP2-3, median not reached vs 18.9 months, respectively, p = 0.038) (Supplemental Figure 1(A)).

Conversely, there was a significant association between the proportion of VTC and the risk of recurrence at 2 years (AUROC 0.671, 95% CI (0.56–0.79), p = 0.007), with the best threshold identified for VTC ⩾ 30, corresponding to a sensitivity of 83.7% (95% CI (70%–91.9%)) and a specificity of 47.6% (95% CI (33.4–62.3)). The determination of this threshold had excellent interobserver reproducibility, with a kappa coefficient of 0.86 ± 0.14. Accordingly, RFS was longer in patients with VTC < 30% than in those with VTC ⩾ 30% (median not reached vs 17.2 months, respectively, p = 0.003; Figure 3(c)).

Of note, although there was a weak but significant correlation between the evolution of tumor size on CT scans and the proportion of VTC (r = 0.33, p = 0.002; Figure 4(a)), objective morphological response per RECIST was poorly concordant with a proportion of VAT < 30% (kappa 0.20 ± 0.11; Figure 4(b)).

On univariable analyses, tumor stage > than pT1 (HR 2.10, 95% CI (1.19–3.82), p = 0.012), lymph-node metastases (HR 2.41, 95% CI (1.33–4.27), p = 0.003), CAP2/3 score (HR 2.11, 95% CI (1.08–4.63), p = 0.043), and VTC ⩾ 30% (HR 2.44, 95% CI (1.29–5.01) p = 0.009) were significantly associated with an increased risk of recurrence (Supplemental Table 1). Of note, morphological response after NAT (HR 0.87, 95% CI (0.48–1.52), p = 0.62) did not significantly influence the risk of recurrence.

On multivariable analysis adjusted for the presence of a biliary stent, NACR, lymph-node metastases, and adjuvant chemotherapy, VTC ⩾ 30% was significantly associated with an increased risk of recurrence (HR 2.28, 95% CI (1.08–5.13), p = 0.037) while objective response after NAT did not (Table 3).

Thirty-four (37%) patients died during the follow-up. The cause of death was PDAC for 32 patients, one patient died from a postoperative late complication (liver ischemia), and one from liver cirrhosis. Median OS from surgery was 59.3 months (95% CI (55.8–62.8). On univariable analyses, tumor stage > than pT1 (HR 2.13, 95% CI (1.06-4.48), p = 0.038), lymph-node metastases (HR 2.42, 95% CI (1.14–4.95), p = 0.017), and VTC ⩾ 30% (HR 2.52, 95% CI (1.15–6.33), p = 0.031) were significantly associated with shorter OS (Supplemental Table 2).

The risk of death was neither influenced by morphological response after NAT (Figure 3(d)) nor by the CAP score (Figure 3(e)), while OS was numerically longer in patients with CAP0-1 than in those with CAP 2–3 (median 58.2 vs 48.1 months, respectively, p = 0.085) (Supplemental Figure 1(B)). Conversely, OS was significantly longer in patients with VTC <30% (62.2 months) than in those with VTC ⩾ 30% (45.1 months, p = 0.025; Figure 3(f)).

On multivariable analysis adjusted on male sex, initial stage, and lymph-node metastases, viable tumor cells > 30% were associated with a 2.45-fold increased risk of death (IC 95%, (1.00–6.62)), although this result was not statistically significant (Supplemental Table 2).