Neoadjuvant chemotherapy has become an established treatment approach for borderline resectable pancreatic ductal adenocarcinoma and is increasingly considered in resectable disease. However, the optimal duration of chemotherapy before surgery remains uncertain.
The second randomization of the phase 3 PACT-21/CASSANDRA trial addressed this question by comparing six months of preoperative chemotherapy with four months of preoperative chemotherapy followed by surgery and two months of postoperative treatment in patients with resectable or borderline resectable PDAC.
The results were published in eClinicalMedicine in September 2026 under the title “Short-term versus long-course neoadjuvant chemotherapy for resectable and borderline resectable pancreatic ductal adenocarcinoma (PACT-21 CASSANDRA): results from the second randomisation analysis of a randomised, open-label, 2 × 2 factorial phase 3 trial.”
Authors: Michele Reni, Giulia Orsi, Catia Carconi, Giuseppe Malleo, Letizia Procaccio, Marina Macchini, Katia Bencardino, Barbara Merelli, Andrea Pretta, Ilario Giovanni Rapposelli, Nicolò Pecorelli, Stefano Partelli, Elisa Sperti, Stefano Crippa, Nicole Liscia, Mariacristina Di Marco, Michele Milella, Sara Lonardi, Diego Palumbo, Valter Torri, and Massimo Falconi.
CASSANDRA Trial Design
PACT-21/CASSANDRA was an Italian multicenter, academic, open-label, 2 × 2 factorial phase 3 trial conducted at 17 academic hospitals across 10 regions. Eligible patients were aged 18 to 75 years, had a Karnofsky performance status of at least 70, and had pathologically confirmed resectable or borderline resectable PDAC.
At the first randomization, patients were assigned to PAXG, consisting of cisplatin, nab-paclitaxel, capecitabine, and gemcitabine, or modified FOLFIRINOX. Patients who remained free of disease progression or limiting toxicity after four months of treatment could undergo a second randomization.
They were assigned to receive: two additional months of the same chemotherapy before surgery, for a total of six months of preoperative therapy; or surgery after four months of preoperative chemotherapy, followed by two additional months of the same regimen postoperatively. The primary endpoint of the second randomization was event-free survival.
Secondary endpoints included radiological and CA19-9 responses, complete pathological response, R0 and N0 resections, treatment toxicity, and overall survival. Chemotherapy dose-intensity and the number of administered cycles were exploratory secondary endpoints. The trial is registered as NCT04793932.
Event-Free Survival
Between February 23, 2021, and September 3, 2024, 171 patients underwent the second randomization. Six were subsequently excluded from the analytical intention-to-treat population because of pre-existing disease progression identified during review or final histology incompatible with PDAC. The ITT population therefore included 165 patients: 86 in the long-course preoperative group and 79 in the short-course group. At a median follow-up of 33.4 months, 116 EFS events had occurred.
No significant difference in EFS was observed between the two groups:
- Median EFS was 13.0 months with long-course preoperative chemotherapy versus 15.2 months with the short-course strategy.
- The unadjusted HR was 0.98 (95% CI, 0.68–1.41; P = .90).
- Three-year EFS was 27.1% versus 23.6%, respectively.
Multivariable analysis confirmed that chemotherapy duration was not significantly associated with EFS, with an adjusted HR of 0.93 (95% CI, 0.64–1.35; P = .69). The primary finding remained consistent in sensitivity analyses. A strict ITT analysis including all 171 randomized patients showed an HR of 1.01, while the per-protocol analysis showed an HR of 0.92, with neither comparison reaching statistical significance.
Prespecified subgroup analyses also did not identify an effect of chemotherapy duration according to chemotherapy regimen or anatomical resectability. The HR for long versus short treatment was 0.96 among patients receiving PAXG and 1.02 among those receiving mFOLFIRINOX, with no significant regimen-by-duration interaction.
Response and Resection Outcomes
Although longer preoperative chemotherapy did not improve EFS, several secondary outcomes differed between the treatment strategies. A CA19-9 reduction of at least 50% was observed in 58 of 60 patients (97%) in the long-course group compared with 41 of 49 patients (84%) in the short-course group (P = .02). Radiological partial responses occurred in 57% versus 47%, respectively, although this difference was not statistically significant (P = .19).
Four patients in the long-course group achieved a complete pathological response compared with none in the short-course group, corresponding to 4.7% versus 0% (P = .05). N0 resections were also more frequent with longer preoperative chemotherapy, occurring in 45% versus 27% of patients (P = .01). No significant differences were observed in overall resection or R0 resection rates.
Sixty-nine patients were resected in each group, corresponding to resection rates of 80% with long-course treatment and 87% with short-course treatment (P = .22). R0 resection rates were 62% versus 58%, respectively (P = .66).
Chemotherapy Delivery
A notable difference between the strategies was the proportion of planned chemotherapy that patients were able to receive. In the per-protocol population, 71% of patients assigned to complete chemotherapy before surgery received all four planned additional administrations compared with 51% of patients assigned to receive those administrations after surgery (P < .01). Relative dose-density was also higher when the final cycles were administered preoperatively. Median relative dose-density in the long- versus short-course groups was:
- 83% versus 54% for fluorouracil
- 73% versus 51% for irinotecan
- 75% versus 51% for oxaliplatin
- 85% versus 61% for cisplatin
- 78% versus 58% for nab-paclitaxel
- 57% versus 40% for capecitabine
- 74% versus 54% for gemcitabine.
The investigators attributed the difference to lower tolerability during the postoperative phase in patients assigned to receive the final treatment cycles after surgery.
Overall Survival
Overall survival data were not mature at the time of the analysis. Seventy-three deaths had occurred among the 165 patients, representing 44% of the ITT population. Median OS was 50.5 months with long-course preoperative chemotherapy and 35.1 months with the short-course strategy. The HR was 0.88 (95% CI, 0.55–1.40; P = .58). Three-year OS was 56.8% versus 46.6%, respectively.
The study prespecified OS maturity after at least 100 deaths. The current OS findings therefore remain immature and cannot establish a survival advantage for either treatment strategy.
Study Limitations
Several limitations should be considered when interpreting the second CASSANDRA randomization. Only patients who remained progression-free after four months of chemotherapy, tolerated treatment, and proceeded to the second randomization were included. The investigators noted that this selected population may limit the generalizability of the findings.
The study also included both resectable and borderline resectable PDAC, lacked central pathological and radiological response review, was conducted in high-volume academic centers, and included patients only up to 75 years of age.
Importantly, the sample size was calculated for the first CASSANDRA randomization because no reliable benchmark was available for the second comparison. With 165 patients in the ITT population, post-hoc power calculations indicated 80% power to detect an HR of 0.65 or lower, or 1.55 or higher. It was therefore underpowered to reliably detect smaller treatment effects that could still be clinically meaningful.
Two different chemotherapy regimens were also included. Although the prespecified analysis found no significant regimen-by-duration interaction, subgroup analyses within each regimen had limited statistical power and should be considered exploratory.
Clinical Implications
CASSANDRA is the first prospective phase 3 randomized trial to specifically evaluate preoperative chemotherapy duration in patients with resectable or borderline resectable PDAC. The second randomization showed that administering six months of chemotherapy before surgery resulted in EFS comparable to four months of preoperative chemotherapy followed by two months of postoperative treatment.
Longer preoperative treatment was associated with greater CA19-9 response, more complete pathological responses, higher N0 resection rates, and greater chemotherapy completion and dose-density. However, these improvements did not translate into an EFS benefit, and OS data remain immature.
The findings support the possibility of personalizing the timing of the final two months of chemotherapy according to individual clinical considerations and patient preferences. Further follow-up will be needed to determine whether differences in chemotherapy delivery and pathological response translate into an overall survival benefit.
