Elraglusib and chemotherapy in metastatic pancreatic ductal adenocarcinoma: a randomized controlled phase 2 trial

Abstract

Metastatic pancreatic ductal adenocarcinoma (mPDAC) is one of the leading causes of cancer-related mortality, but advances in therapeutic treatments remain limited. Elraglusib (9-ING-41), an inhibitor of GSK-3β, exhibits a multimodal mechanism of action based on antitumor activity in preclinical models of cancer, including pancreatic. The efficacy and safety of elraglusib with gemcitabine plus nab-paclitaxel (GnP) were assessed in patients with previously untreated mPDAC. In an open-label, international, multicenter, phase 2 study, patients were randomized 2:1 to weekly elraglusib/GnP or GnP alone. Primary endpoints were median overall survival (OS) and 1-year survival rate. The prespecified modified intention-to-treat population included 155 patients on elraglusib/GnP and 78 on GnP. As of the data cutoff of 27 April 2025, elraglusib/GnP improved median OS by 2.9 months and decreased the risk of death by 38% versus GnP (median OS 10.1 months versus 7.2 months, respectively (hazard ratio 0.62; 95% confidence interval 0.46 to 0.84; P = 0.01)). The 1-year survival rates were 44.1% versus 22.3%, respectively. The safety profile of elraglusib/GnP was manageable. The most common grade 3 or higher treatment-emergent adverse events (TEAEs) with elraglusib/GnP versus GnP alone were neutropenia (52.3% versus 30.8%), anemia (25.2% versus 29.5%) and fatigue (16.8% versus 5.1%). Explorative correlative analyses demonstrated that baseline circulating immune-related factors (that is, CXCL2 and TRAIL ligands) were associated with improved survival in the elraglusib/GnP arm. Treatment was accompanied by increases in intratumoral cytotoxic immune cell populations. Together, these findings support the clinical activity of elraglusib/GnP as first-line treatment in mPDAC and provide a biological context for the observed survival benefit. Based on the results of this phase 2 trial, a phase 3 trial is being planned. ClinicalTrials.gov registration: NCT03678883

In this phase 2 trial, combination treatment with elraglusib, a cell-permeable ATP-competitive inhibitor of glycogen synthase kinase-3β, and gemcitabine plus nab-paclitaxel (GnP), in patients with previously untreated metastatic pancreatic ductal adenocarcinoma led to prolonged overall survival compared with GnP only.

Main

Pancreatic cancer is the third leading cause of cancer-related mortality in the United States, fourth in the European Union and seventh worldwide^1–3. Pancreatic ductal adenocarcinoma (PDAC) represents over 90% of all pancreatic cancers⁴. About 50% to 55% of pancreatic cancer cases are metastatic at diagnosis because of lack of specific symptoms in earlier stages^5,6

The current first-line treatment regimens for metastatic pancreatic cancer consist of 5-fluorouracil, leucovorin, irinotecan plus oxaliplatin (FOLFIRINOX); nanoliposome irinotecan plus 5-fluorouracil, leucovorin and oxaliplatin (NALIRIFOX); or gemcitabine plus nab-paclitaxel (GnP)^4,7,8. Some concerns include toxicity with the FOLFIRINOX and NALIRIFOX regimens, limiting their use to patients with an Eastern Cooperative Oncology Group (ECOG) performance status (PS) of 0 and 1 (considered as good PS), and higher cost with NALIRIFOX^4,9,10. Recent studies also suggested that compared with classical metastatic PDAC (mPDAC), basal mPDAC is less responsive to FOLFIRINOX¹¹. GnP remains one of three approved first-line regimens for patients with good (0–1) or intermediate (2) ECOG PS^4,7. Median OS with GnP treatment ranges from approximately 8 to 10 months in clinical trials and approximately 7 to 10 months in the real-world setting^9,10,12–14.

Despite the increasing incidence over the past decades and the identification of potential molecular driver alterations as targets in pancreatic cancer (for example, BRCA1 and BRCA2, PALB2, KRAS, TP53, CDKN2A, dMMR/MSI-H), advances in therapeutic treatments remain limited¹. Moreover, immunotherapy that led to positive results in other solid organ cancers has failed to improve outcomes in PDAC^15–19. Thus, a strong need remains for new therapeutic agents to help prolong survival and improve clinical outcomes in this difficult-to-manage cancer.

Glycogen synthase kinase-3 beta (GSK-3β), a serine/threonine kinase, has been identified as a potential therapeutic target in human pancreatic cancer²⁰. Elraglusib (9-ING-41) is a first-in-class, cell-permeable, ATP-competitive inhibitor of GSK-3β. In preclinical cancer models, elraglusib has contributed to a range of potential antitumor activities: inhibition of cancer cell proliferation and survival, overcoming drug resistance in tumors, inhibition of epithelial–mesenchymal transition, decrease in tumor immune evasion, activation and expansion of infiltrating CD8⁺ T cells and natural killer (NK) cells, epigenetic control of gene expression and reduction in fibrosis^21–24. In models of pancreatic cancer, elraglusib inhibited proliferation and survival of cancer cells, sensitized cancer cells to chemotherapy agents and prevented chemoresistance^21,25,26.

Early-phase clinical trials revealed promising results with elraglusib. In a phase 1 trial, combinations of elraglusib with a variety of chemotherapy backbones led to a median OS of 6.9 months (95% confidence interval (CI) 5.7 to 8.4 months) and an acceptable safety profile in heavily pretreated patients with relapsed or refractory solid cancers or hematological malignancies²⁷. Additionally, a median OS of 4.5 months was observed among 26 patients with treatment-refractory mPDAC who received elraglusib/GnP as their third-line treatment. In a single-arm, phase 2 trial in 42 patients with previously untreated mPDAC, elraglusib/GnP led to a median OS of 11.9 months (95% CI 7.8 to 16.5) and 1-year survival rate of 48% (95% CI 32.1 to 62.3) in the intention-to-treat (ITT) population²⁸. This trial established a recommended phase 2 dose for elraglusib of 9.3 mg kg⁻¹ administered intravenously twice weekly when combined with GnP. Because of the promising survival trends, this trial transitioned to a phase 2 randomized trial that also explored weekly dosing of elraglusib with GnP. The key objectives of this randomized phase 2 trial were to evaluate the 1-year survival rate and the median OS and correlate efficacy parameters with tumor molecular profiles of elraglusib/GnP versus GnP in patients with previously untreated mPDAC.

Results

Patient disposition and baseline characteristics

The study was conducted at 60 sites across six countries in North America and Europe. During the selection stage, 51 patients were randomized 1:1:1 into one of the three treatment arms with 17 patients treated in each of the once-weekly elraglusib/GnP, twice-weekly elraglusib/GnP and GnP arms. Once-weekly dosing was determined to be equivalent to twice-weekly dosing based on pharmacokinetics profile, 1-year survival rate and objective response rate (ORR); thus, in addition to patient preference and improved patient compliance, elraglusib 9.3 mg kg⁻¹ once weekly was selected as the preferred dosing schedule. The presented results for elraglusib/GnP focus on the once-weekly dosing of elraglusib; patients from the selection and confirmation stages are included.

Between 7 October 2021 and 9 February 2024, 286 patients enrolled in the study (171 patients in the once-weekly elraglusib/GnP arm, 17 patients in the twice-weekly elraglusib/GnP arm and 98 patients in the GnP arm); 250 patients received at least one dose of treatment (Fig. 1). The modified ITT (mITT) population included 155 patients in the once-weekly elraglusib/GnP arm, 17 patients in the twice-weekly elraglusib/GnP arm and 78 in the GnP arm (Fig. 1). At the data cutoff of 27 April 2025, the most common reasons for treatment discontinuations were progressive disease (PD) (113 patients, 45.2%), clinical progression (38 patients, 15.2%) and withdrawal by participant (33 patients, 13.2%; Fig. 1). The most common reason for study withdrawal was death (185 patients, 74.0%). Six patients in the elraglusib/GnP arm were still receiving treatment and 49 patients were in the survival follow-up (41 in the elraglusib/GnP arm and eight in the GnP arm). The rest of the results present findings related to once-weekly elraglusib/GnP and GnP arms in both the mITT and ITT populations.

Fig. 1. Consort diagram with patient disposition.

Patient disposition throughout the study, including randomization, reasons for treatment discontinuation and reasons for study withdrawals.

The key characteristics of treated patients were a median age of 66 years (range 42–86 years), male sex in 52.4% of patients and White ethnicity in 83.2% of patients with 93.2% being non-Hispanic or Latino (Table 1). Almost all patients had metastatic disease at study entry, with approximately 73% of patients presenting with metastatic disease at initial diagnosis. The most common site of metastasis was the liver (in over 70% of patients). Most patients had an ECOG PS of 0 (41.2%) or 1 (57.2%). Elevated baseline CA 19-9 levels were reported in 88% of patients, with median CA 19-9 levels of 1,568 (range 1–381,904) in the once-weekly elraglusib/GnP arm and 1,590 (range 2–501,000) in the GnP arm. Before the study, 22.6% of patients in the once-weekly elraglusib/GnP arm and 18% in the GnP arm received prior neoadjuvant or adjuvant cytotoxic chemotherapy; 27.7% and 21.8%, respectively, received prior radiotherapy.

Table 1.

Patient demographics and baseline characteristics (mITT population)

Characteristic	Once-weekly elraglusib/GnP n = 155	GnP n = 78
Age (years)
Mean (s.d.)	65.1 (9.1)	66.2 (9.9)
Median (range)	65.0 (42.0 to 86.0)	68.0 (42.0 to 85.0)
Sex, n (%)
Female	75 (48.4)	35 (44.9)
Male	80 (51.6)	43 (55.1)
Ethnicity, n (%)
White	128 (82.6)	65 (83.3)
Black or African American	7 (4.5)	6 (7.7)
Asian	5 (3.2)	2 (2.6)
Multiracial	1 (0.6)	0
Unknown/not reported	14 (9.0)	5 (6.4)
Ethnic group, n (%)
Hispanic or Latino	8 (5.2)	0
Not Hispanic or Latino	141 (91.0)	77 (98.7)
Unknown/not reported	6 (3.9)	1 (1.3)
Height (in)
Mean (s.d.)	66 (4)	67 (4)
Median (range)	67 (58 to 76)	67 (59 to 76)
Weight (lb)
Mean (s.d.)	160.4 (37.9)	159.0 (42.2)
Median (range)	156.3 (89.3 to 335.0)	154.4 (85.3 to 343.3)
ECOG PS, n (%)
0	64 (41.3)	31 (39.7)
1	89 (57.4)	45 (57.7)
2	2 (1.3)	2 (2.6)
Disease status, n (%)
Metastatic at initial diagnosis	109 (70.3)	60 (76.9)
Metastatic at study entry	154 (99.4)	77 (98.7)
Site of disease/metastases, n (%)
Pancreas	123 (79.4)	68 (87.2)
Liver	112 (72.3)	61 (78.2)
Lymph node	69 (44.5)	27 (34.6)
Lung	59 (38.1)	26 (33.3)
Baseline CA 19-9 (U ml−1), n (%)	146 (94.2)	76 (97.4)
Mean (s.d.)	20,909 (54,084)	31,076 (88,065)
Median (range)	1,568 (1 to 381,904)	1,590 (2 to 501,000)
<37 U ml−1 (within the reference range)	17 (11.0)	11 (14.1)
≥37 U ml−1	138 (89.0)	67 (85.9)

BMI, body mass index; CA, carbohydrate antigen.

Efficacy

In the mITT population, the median OS was increased by 2.9 months with elraglusib/GnP compared with GnP only (median OS of 10.1 months (95% CI 7.7 to 12.5) with elraglusib/GnP versus 7.2 months (95% CI 5.7 to 9.0) with GnP (HR 0.62; 95% CI 0.46 to 0.84; P = 0.01; Table 2 and Fig. 2a)). A survival benefit was seen as early as 2 months (Fig. 2a) and persisted throughout the study. The hazard curves continued to diverge throughout the entire study. The 1-year landmark survival rate was 44.1% (95% CI 36.0 to 51.8) with elraglusib/GnP and 22.3% (95% CI 13.8 to 32.1) with GnP. Landmark survival rates at 18 and 24 months were 20.5% (95% CI 13.4 to 28.8) and 13.2% (95% CI 6.7 to 21.8) in the elraglusib/GnP arm and 4.4% (95% CI 0.5 to 15.5) and 0 in the GnP arm, respectively.

Table 2.

Efficacy outcomes: OS, PFS and overall response (mITT population)

Outcome	Elraglusib/GnP n = 155	GnP n = 78	Effect size (95% CI)	P
OS
Median OS, months (95% CI)	10.1 (7.7 to 12.5)	7.2 (5.7 to 9.0)	HR 0.62 (0.46 to 0.84)	0.010
Survival rate, % (95% CI)
12 months	44.1 (36.0 to 51.8)	22.3 (13.8 to 32.1)
18 months	20.5 (13.4 to 28.8)	4.4 (0.5 to 15.5)
24 months	13.2 (6.7 to 21.8)	0
PFS
Median PFS, months (95% CI)	5.6 (4.6 to 6.4)	5.1 (3.2 to 6.1)	HR 0.90 (0.68 to 1.20)	0.624
Median time to treatment failure (95% CI)	5.0 (3.6 to 5.6)	3.4 (2.4 to 5.1)	HR 0.79 (0.60 to 1.05)	0.178
Overall response
Median duration of response (95% CI)	5.5 (3.8 to 5.8)	4.0 (3.2 to 6.3)	HR 0.81 (0.43 to 1.49)	0.984
Objective response rate	44 (28.4)	17 (21.8)		0.281
Confirmed CR, n (%)	1 (0.6)	0
Confirmed PR, n (%)	30 (19.4)	13 (16.7)
Unconfirmed PR, n (%)	13 (8.4)	4 (5.1)
Stable disease ≥16 weeks,a n (%)	23 (14.8)	11 (14.1)
Stable disease <16 weeks, n (%)	27 (17.4)	16 (20.5)
PD, n (%)	21 (13.5)	15 (19.2)
Not evaluable	40 (25.8)	19 (24.4)
Disease control rate,a n (%)	62 (40.0)	26 (33.3)		0.323

Statistical tests: log-rank tests for time-to-event variables; Cox proportional hazards model for HRs; and Cochran–Mantel–Haenszel general association tests for responder variables. All tests were two-sided.

^aDisease control is stable disease, PR or CR for at least 16 weeks since the first dose of the study drug, or a confirmed CR or PR.

Fig. 2. Kaplan–Meier estimates of OS in the mITT and ITT populations.

a, Statistical test, log-rank stratified by center; P = 0.010. b, Statistical test, log-rank stratified by center; P = 0.022.

Similarly, elraglusib/GnP led to improvements in median OS in the ITT population, where the median OS was increased by 1.7 months with elraglusib/GnP (Table 3). The median OS was 8.9 months (95% CI 7.3 to 11.9) with elraglusib/GnP and 7.2 months (95% CI 5.7 to 8.7) with GnP (HR 0.68; 95% CI 0.51 to 0.89; P = 0.022) (Fig. 2b). The 1-year landmark survival rate was 42.1% (95% CI 34.4 to 49.6) with elraglusib/GnP and 22.3% (95% CI 14.5 to 31.1) with GnP. Landmark survival rates at 18 and 24 months were 19.3% (95% CI 12.6 to 27.1) and 12.4% (95% CI 6.3 to 20.6) with elraglusib/GnP and 6.2% (95% CI 1.7 to 15.2) and 3.1% (95% CI 0.3 to 12.3) with GnP, respectively.

Table 3.

Efficacy outcomes: OS, PFS and overall response (ITT population)

Outcome	Once-weekly elraglusib/GnP n = 171	GnP n = 98	Effect size (95% CI)	P
OS
Median OS, months (95% CI)	8.9 (7.3 to 11.9)	7.2 (5.7 to 8.7)	HR 0.68 (0.51 to 0.89)	0.022
Survival rate, % (95% CI)
12 months	42.1 (34.4 to 49.6)	22.3 (14.5 to 31.1)
18 months	19.3 (12.6 to 27.1)	6.2 (1.7 to 15.2)
24 months	12.4 (6.3 to 20.6)	3.1 (0.3 to 12.3)
PFS
Median PFS (months) (95% CI)	5.5 (4.2 to 6.0)	5.4 (3.6 to 6.3)	HR 0.99 (0.76 to 1.29)	0.963
Median time to treatment failure (95% CI)	4.5 (3.3 to 5.5)	3.2 (2.1 to 4.1)	HR 0.76 (0.58 to 0.98)	0.069
Overall response
Median duration of response (95% CI)	5.5 (3.8 to 5.8)	4.0 (3.2 to 6.3)	HR 0.81 (0.43 to 1.49)	0.984
Objective response rate	44 (25.7)	17 (17.3)		0.115
Confirmed CR, n (%)	1 (0.6)	0
Confirmed PR, n (%)	30 (17.5)	13 (13.3)
Unconfirmed PR, n (%)	13 (7.6)	4 (4.1)
Stable disease ≥16 weeks, n (%)	23 (13.5)	11 (11.2)
Stable disease <16 weeks, n (%)	27 (15.8)	16 (16.3)
PD, n (%)	21 (12.3)	16 (16.3)
Not evaluable, n (%)	56 (32.7)	38 (38.8)
Disease control rate,a n (%)	62 (36.3)	26 (26.5)		0.102

Statistical tests: log-rank tests for time-to-event variables; Cox proportional hazards model for HRs; and Cochran–Mantel–Haenszel general association tests for responder variables. All tests were two-sided.

^aDisease control is stable disease, PR or CR for at least 16 weeks since the first dose of the study drug, or a confirmed CR or PR.

OS benefit was observed with elraglusib/GnP across the subgroups evaluated for the presence of liver metastases, ECOG status, CA 19-9 levels at baseline, and others (Extended Data Figs. 1 and 2). The analysis of patients with liver metastases at baseline revealed a 1.7-month improvement in median OS for elraglusib/GnP in the mITT population (HR 0.62; 95% CI 0.44 to 0.87; P = 0.008; median OS of 8.3 months (95% CI 7.1 to 11.2) for elraglusib/GnP versus 6.6 months (95% CI 5.1 to 8.0) for GnP alone). The 1-year landmark survival rates were 39.2% with elraglusib/GnP versus 15.2% with GnP alone in this subpopulation (Supplementary Fig. 1). Among patients whose CA 19-9 levels dropped by 50% or more after one treatment cycle (n = 30 patients); median OS displayed numerical benefit without statistical significance toward improvement with elraglusib/GnP versus GnP (32.1 months versus 10.6 months; HR 0.32; 95% CI 0.12 to 0.87; P = 0.089; Supplementary Fig. 2).

Extended Data Fig. 1. Forest Plot for Overall Survival in Subgroups of the mITT population.

Abbreviations: 9-ING-41, elraglusib; CA, carbohydrate antigen; CI, confidence interval; ECOG, Eastern Cooperative Oncology Group; GnP, gemcitabine plus nab-paclitaxel; HR, hazard ratio; mITT, modified intention-to-treat; wk, week.

Extended Data Fig. 2. Forest Plot for Overall Survival in Subgroups of the ITT population.

Abbreviations: 9-ING-41, elraglusib; CA, carbohydrate antigen; CI, confidence interval; ECOG, Eastern Cooperative Oncology Group; GnP, gemcitabine plus nab-paclitaxel; HR, hazard ratio; ITT, intention-to-treat; wk, week.

Elraglusib/GnP displayed numerical benefit without statistical significance toward improvement across most secondary endpoints in the mITT population (Table 2). The median progression-free survival (PFS) was 5.6 months (95% CI 4.6 to 6.4) with elraglusib/GnP and 5.1 months (95% CI 3.2 to 6.1) with GnP (HR 0.90; 95% CI 0.68 to 1.20; P = 0.624; Supplementary Fig. 3). The median time to treatment failure (TTF) was 5.0 months (95% CI 3.6 to 5.6) with elraglusib/GnP and 3.4 months (95% CI 2.4 to 5.1) with GnP. ORR was observed in 44 patients (28.4%) with elraglusib/GnP and 17 patients (21.8%) with GnP (Table 2 and Supplementary Fig. 4). In the elraglusib/GnP arm, one patient achieved complete response (CR) and 43 (27.8%) patients achieved partial response (PR), including 13 patients (8.4%) with unconfirmed PR. In the GnP arm, no patients achieved CR and 17 (21.8%) patients achieved PR (confirmed and unconfirmed). The median duration of response (DOR) was 5.5 months (95% CI 3.8 to 5.8) with elraglusib/GnP and 4 months (95% CI 3.2 to 6.3) with GnP. The disease control rate (DCR) was 40.0% with elraglusib/GnP and 33.3% with GnP. Given that schedules for tumor assessment differed according to institution (assessments performed either every 8 or every 12 weeks), the DCR in our study was defined as SD, PR (including unconfirmed PR) or CR for at least 16 weeks since the first dose of the study drug, or a confirmed CR or PR.

Treatment exposure

The median duration of treatment was longer with elraglusib/GnP: 116 days (range 7–554 days; median of four treatment cycles) with elraglusib/GnP and 88 days (range 7–526 days; median of four treatment cycles) with GnP (Supplementary Table 1). More patients on elraglusib/GnP experienced dose delays or reductions (92.9% of patients on elraglusib/GnP and 84.6 % of patients on GnP). Median relative dose intensities were similar for nab-paclitaxel (70% with elraglusib/GnP and 72% with GnP) and gemcitabine (74% with elraglusib/GnP and 78% with GnP). The median relative dose intensity of elraglusib was 82%.

Subsequent therapies

Subsequent systemic anticancer therapies were administered to 77 patients in the elraglusib/GnP arm and 33 in the GnP arm (Extended Data Table 1). The most common subsequent therapy was FOLFIRINOX in both treatment arms (Extended Data Table 1 and Supplementary Table 2). After transitioning to subsequent therapies, numerical benefit without statistical significance for increased median OS remained in the elraglusib/GnP arm based on the exploratory analysis: 13.3 months (95% CI 11.9 to 14.6) in the elraglusib/GnP arm versus 10.6 months (95% CI 9.0 to 12.9) in the GnP arm (HR 0.68; 95% CI 0.43 to 1.07; log-rank P = 0.094) among patients with subsequent therapies and 5.8 months (95% CI 4.1 to 7.1) in the elraglusib/GnP arm versus 3.4 months (95% CI 2.8 to 6.1) in the GnP arm (HR 0.69; 95% CI 0.46 to 1.03; log-rank P = 0.064) among patients without subsequent therapies (Extended Data Fig. 3). A large proportion of patients who did not receive subsequent therapies is still in the survival follow-up for the elraglusib/GnP arm.

Extended Data Table 1.

Subsequent Anti-Cancer Treatments (mITT Population)

^aAll 4 patients had post-treatment therapy indicated as GnP; however, treatment was determined to be first-line.

^bThis patient also received anti-neoplastic therapy with radiation.

Abbreviations: 5-FU, 5-fluorouracil; FOLFIRI, leucovorin plus 5-fluorouracil plus irinotecan; FOLFIRINOX, leucovorin plus 5-fluorouracil plus irinotecan plus oxaliplatin; FOLFOX, leucovorin plus 5-fluorouracil plus oxaliplatin; FOLFOXIRI, leucovorin plus 5-fluorouracil plus oxaliplatin plus irinotecan; GnP, gemcitabine plus nab-paclitaxel; LV, leucovorin; mITT, modified intention to treat.

Extended Data Fig. 3. Exploratory Analysis for Overall Survival in (A) Patients With Subsequent Treatment and (B) Patients Without Subsequent Treatment.

(A) Statistical test: log-rank test stratified by center; p-value: p = 0.094 (B) Statistical test: log-rank test stratified by center; p-value: p = 0.064 Abbreviations: 9-ING-41, elraglusib; GnP, gemcitabine plus nab-paclitaxel; mITT, modified intention-to-treat; wk, week.

Safety

Any TEAE occurred in 155 patients (100%) in the once-weekly elraglusib/GnP arm and 77 patients (98.7%) in the GnP arm, with serious TEAEs affecting a similar proportion of patients in each treatment arm (86 patients (55.5%) in the elraglusib/GnP arm and 44 patients (56.4%) in the GnP arm; Table 4). The most common TEAEs that occurred in the elraglusib/GnP arm were visual impairment (n = 105; 67.7% versus 9.0% with GnP), fatigue (n = 97; 62.6% versus 50.0% with GnP) and neutropenia (n = 95; 61.3% versus 41.0% with GnP), all of which occurred more frequently with elraglusib/GnP than with GnP. The most common TEAEs that were attributed to elraglusib included visual impairment (66.5%), fatigue (42.6%), nausea (36.1%) and neutropenia (32.9%) (Extended Data Table 2). Elraglusib-related infusion-related reaction was reported in 24 patients (15.5%). Visual impairment and infusion-related reactions were transient.

Table 4.

Overview of TEAEs in 20% or more of patients (mITT population)

TEAE	Patients, n (%)
	Elraglusib/GnP n = 155		GnP n = 78
	Any grade	Grade 3 and higher	Any grade	Grade 3 and higher
Any TEAE	155 (100)	139 (89.7)	77 (98.7)	62 (79.5)
Serious TEAE	86 (55.5)	81 (52.3)	44 (56.4)	43 (55.1)
Leading to treatment discontinuation	42 (27.1)	26 (16.8)	20 (25.6)	16 (20.5)
Leading to death	19 (12.3)	19 (12.3)	13 (16.7)	13 (16.7)
TEAEs of any grade in 20% or more of patients
Visual impairmenta	105 (67.7)	1 (0.6)	7 (9.0)	0
Fatigue	97 (62.6)	26 (16.8)	39 (50.0)	4 (5.1)
Neutropeniab	95 (61.3)	81 (52.3)	32 (41.0)	24 (30.8)
Diarrhea	93 (60.0)	15 (9.7)	38 (48.7)	6 (7.7)
Nausea	90 (58.1)	11 (7.1)	38 (48.7)	4 (5.1)
Anemiac	71 (45.8)	39 (25.2)	35 (44.9)	23 (29.5)
Alopecia	71 (45.8)	0	27 (34.6)	0
Decreased appetite	65 (41.9)	9 (5.8)	19 (24.4)	6 (7.7)
Vomiting	59 (38.1)	5 (3.2)	30 (38.5)	1 (1.3)
Thrombocytopenia	58 (37.4)	17 (11.0)	25 (32.1)	6 (7.7)
Peripheral edema	56 (36.1)	3 (1.9)	25 (32.1)	0
Constipation	50 (32.3)	3 (1.9)	24 (30.8)	1 (1.3)
Abdominal pain	45 (29.0)	14 (9.0)	16 (20.5)	2 (2.6)
Pyrexia	44 (28.4)	2 (1.3)	20 (25.6)	1 (1.3)
Weight decreased	42 (27.1)	5 (3.2)	16 (20.5)	4 (5.1)
Peripheral sensory neuropathy	39 (25.2)	4 (2.6)	18 (23.1)	0
Hypokalemia	35 (22.6)	8 (5.2)	24 (30.8)	4 (5.1)
Asthenia	33 (21.3)	9 (5.8)	19 (24.4)	5 (6.4)
Dysgeusia	32 (20.6)	0	16 (20.5)	0
Infusion-related reaction	31 (20.0)	4 (2.6)	1 (1.3)	0

^aVisual impairment includes the MedDRA preferred terms of vision blurred, dyschromatopsia (color vision change), diplopia (double vision), photopsia and visual impairment. Two patients reported two different TEAEs classified as visual impairment but these were counted only once.

^bNeutropenia includes neutropenia and decreased neutrophil count.

^cThrombocytopenia includes thrombocytopenia and decreased platelet count.

MedDRA, Medical Dictionary for Regulatory Activities.

Extended Data Table 2.

TEAEs Related to Elraglusib Occurring at an Incidence ≥10% Among Elraglusib-treated Patients (mITT Population)

^a MedDRA v23.1

Includes events judged possibly, probably or definitely related; excludes events judged unrelated or unlikely related.

Abbreviations: GnP, gemcitabine plus nab-paclitaxel; MedDRA, Medical Dictionary for Regulatory Activities; mITT, modified intention-to-treat; TEAE, treatment-emergent adverse event; wk, week.

Grade 3 or higher TEAEs were more frequent in the elraglusib/GnP arm: 139 patients (89.7%) in the elraglusib/GnP arm and 62 patients (79.5%) in the GnP arm (Table 4). The most common grade 3 or higher TEAEs with elraglusib/GnP were neutropenia (52.3% versus 30.8% with GnP), anemia (25.2% versus 29.5%) and fatigue (16.8% versus 5.1%). Of these, neutropenia and fatigue occurred more frequently with elraglusib/GnP than with GnP.

The concomitant administration of elraglusib with GnP did not increase plasma levels of nab-paclitaxel or gemcitabine (Supplementary Fig. 5). Among patients with grade 3 to 4 neutropenia, granulocyte colony-stimulating factors were administered to a similar proportion of patients: 38 (46.9%) patients on elraglusib/GnP and 11 (45.8%) on GnP. In an exploratory analysis, median OS was significantly prolonged in elraglusib/GnP-treated patients who experienced two or more grade 3 or 4 neutropenia episodes but not in GnP-treated patients (Supplementary Table 3).

TEAEs leading to treatment discontinuation were observed in 42 patients (27.1%) in the elraglusib/GnP arm and 20 patients (25.6%) in the GnP arm (Extended Data Table 3). The most common TEAEs leading to study drug discontinuation in the elraglusib/GnP arm were peripheral neuropathy (eight patients, 5.2%), infusion-related reactions (seven patients, 4.5%) and fatigue (five patients, 3.2%).

Extended Data Table 3.

TEAEs Leading to Any Study Drug Discontinuation for >1 Patient (mITT Population)

^a MedDRA v23.1

Abbreviations: GnP, gemcitabine plus nab-paclitaxel; MedDRA, Medical Dictionary for Regulatory Activities; mITT, modified intention-to-treat; TEAE, treatment-emergent adverse event; wk, week.

Overall, TEAEs leading to death were reported in 19 patients (12.3%) in the elraglusib/GnP arm and 13 patients (16.7%) in the GnP arm (Supplementary Table 4). The most common TEAE leading to death was disease progression. TEAEs leading to death did not overlap between the elraglusib/GnP and GnP arms except for upper respiratory hemorrhage experienced by one patient in each group. Three patients in the elraglusib/GnP arm suffered from sepsis resulting in death. Only five TEAEs leading to death were considered possibly or probably related to elraglusib: cardiac arrest (one patient), death (not otherwise specified) (one patient), dehydration (one patient), septic shock (one patient) and streptococcal sepsis (one patient). These treatment-related TEAEs leading to death occurred with once-weekly elraglusib.

Biomarker analysis

Immunophenotyping of paired tumor samples

Tumor immunophenotyping of paired pre-dose and post-dose formalin-fixed paraffin-embedded tumor samples from six patients (four patients on elraglusib/GnP and two patients on GnP) demonstrated that the percentage of CD8⁺, granzyme B⁺ and CD56⁺ NK cells in the tumor parenchyma was significantly increased (by seven to 40 times) in elraglusib/GnP-treated tumors but not in GnP-treated tumors (Supplementary Table 5 and Extended Data Fig. 4). The percentage of myeloid cells, including myeloid-derived suppressor cells (MDSCs), was decreased in tumor parenchyma after treatment in both elraglusib/GnP and GnP arms.

Extended Data Fig. 4. Immunophenotyping of Pre- and Post-treatment Tumor Samples.

Representative pictures of tumor immunophenotyping analysis of pre- and post-dose tumor biopsies obtained from the patient treated with elraglusib/GnP (A). The percentage of CD8+ cells (B) and granzyme B+ immune cells (C) was evaluated in tumor parenchyma and compared between pre- and post-dose tumor samples in GnP and elraglusib/GnP arms. Abbreviations: CD8 + , a marker of cytotoxic T cells; CD56 + , a marker of natural killer cells; CK, cytokeratin; GrzB + , granzyme B, a marker of activated cytotoxic T cells and NK cells; GnP, gemcitabine plus nab-paclitaxel; H&E, Hematoxylin and Eosin; NE, not evaluable; OS, overall survival; PD, progressive disease; PR, partial response; SD, stable disease.

GSK-3β expression in tumor samples

Pre-dose GSK-3β expression was analyzed in tumor samples obtained from 139 patients (98 patients on elraglusib/GnP and 41 patients on GnP). Aberrant nuclear expression of GSK-3β in 50% or more cancer cells was found in 22% and 24% of tumors in the elraglusib/GnP and GnP arms, respectively (Supplementary Fig. 6a). GSK-3β cytoplasmic overexpression in 50% or more cancer cells was detected in 78% of tumors in each arm—elraglusib/GnP (76 of 98 patients) and GnP (32 of 41 patients) (Supplementary Fig. 6b). Nuclear expression of GSK-3β or cytoplasmic overexpression of GSK-3β were not significantly correlated with clinical response or survival in the elraglusib/GnP or GnP arms.

Correlative analysis for cytokines, chemokines, soluble cell receptors or growth factors

For the analysis, blood plasma samples were available from 155 patients in the elraglusib/GnP arm and 69 patients in the GnP arm. The pre-dose analysis of 40 plasma cytokines, chemokines, soluble cell receptors or growth factors (CCSGs) identified seven CCSGs that correlated with the 1-year survival rate. When stratified according to the median value for the elraglusib/GnP arm, high levels of CXCL2 significantly correlated with and high levels of TRAIL ligand displayed numerical benefit without statistical significance toward improved OS in the elraglusib/GnP arm but not in the GnP arm (Extended Data Fig. 5)²⁹. A comprehensive analysis, including optimized statistical, multivariate and machine learning models, of CCSG data with patient outcomes will be published separately.

Extended Data Fig. 5. Overall survival of patients treated with elraglusib/GnP (A) or GnP (B) stratified about the median of the elraglusib/GnP arm as CXCL2 low or high from pre-dose samples, and overall survival of patients treated with erlaglusib/GnP (C) or GnP (D) stratified about the median of the elraglusib/GnP arm as TRAIL ligand low or high from pre-dose samples.

Note: P-values were calculated using the log rank test. Abbreviations: OS, overall survival.

Discussion

In this randomized phase 2 study, the combination of once-weekly elraglusib/GnP demonstrated a statistically significant improvement in survival, a 2.9-month improvement in OS in the elraglusib/GnP arm (median OS of 10.1 months versus 7.2 months with GnP; HR 0.63; P = 0.01). The 1-year landmark survival rate was 44.1% with elraglusib/GnP versus 22.3% with GnP in previously untreated mPDAC. A survival benefit was observed with elraglusib/GnP as early as 2 months and persisted throughout the study. Survival benefit was consistent across subgroups evaluated for the presence of liver metastases, ECOG status, CA 19-9 levels at baseline and others.

Additional numerical benefits without significance favored elraglusib/GnP over GnP across multiple endpoints, including median TTF, DOR, ORR and DCR. The OS advantage with elraglusib/GnP was observed in the absence of a significant PFS benefit (Tables 2 and 3) and persisted regardless of subsequent therapy (Extended Data Fig. 3), suggesting durable benefit beyond active treatment. Similar to many trials in advanced cancer evaluating immunomodulators, where OS benefit is often observed in the absence of a PFS benefit, this may be related to the immunomodulatory mechanism of action of GSK-3 inhibitors including elraglusib^30–32. This disconnect may reflect delayed clinical responses but sustained OS benefit of the immune-based therapies, which often take time to mount an adequate immune response to a tumor^33,34. This result also suggests that the survival benefit is a consequence of elraglusib’s activity rather than the receipt of more doses in the elraglusib/GnP arm as the survival benefit was observed beyond treatment termination with elraglusib. For agents that mediate antitumor immunity—such as immune checkpoint inhibitors—traditional endpoints like PFS and ORR often correlate poorly with OS^33,35–37.

Our study reported a lower median OS in the GnP arm compared to historical randomized trials, where GnP typically shows an OS of approximately 8 to 10 months^8,10,13,14. However, this finding is consistent with real-world data, where median OS is closer to 7 months when summarized across multiple studies¹². Several factors may explain this discrepancy. Unlike many mPDAC trials, our exploratory study used broader eligibility criteria, enrolling a cohort more reflective of real-world diversity. Notably, we included patients with low albumin (<3 g dl⁻¹) and substantial tumor burden with over 25% of patients having baseline CA 19-9 > 8,000 U ml⁻¹ (the upper threshold of the range reported in NAPOLI-3), features typically linked to poor prognosis¹⁰. Early mortality was significantly higher in the GnP arm (17.9 % within the first 2 months versus 10.9% in NAPOLI-3), probably driven by these high-risk baseline characteristics¹⁰ (Supplementary Table 6). This early mortality probably affected both arms equally because of well-balanced baseline factors, contributing to lower median OS in the elraglusib/GnP arm also. Importantly, despite early declines, Kaplan–Meier curves diverged early and remained separated, indicating a sustained survival benefit in the elraglusib/GnP arm (Fig. 2). These results highlight the importance of considering baseline patient heterogeneity and real-world factors when interpreting survival outcomes relative to prior randomized trials and highlight the importance of including a control arm in the study to account for anticipated heterogeneity in this patient population.

In this study, the patients had similar median relative dose intensities for nab-paclitaxel (70% with once-weekly elraglusib/GnP and 72% with GnP) and gemcitabine (74% with elraglusib/GnP and 78% with GnP) (Supplementary Table 1). These median relative dose intensities for GnP in our study were slightly lower than the median relative dose intensity for nab-paclitaxel (81%) and similar to the median relative dose intensity for gemcitabine (75%) achieved in the MPACT study¹³. The median number of treatment cycles was four in both treatment arms. To evaluate whether the observed OS benefit reflected a true treatment effect or was confounded by differential treatment duration, we performed exploratory efficacy analyses stratified according to treatment exposure as fewer than four versus four or more cycles. While the OS was shorter among patients receiving fewer than four cycles, the relative treatment benefit of elraglusib/GnP compared with GnP alone was preserved regardless of treatment duration, as reflected by similar HRs across exposure subgroups (Extended Data Table 4). These findings indicate that the OS benefit associated with elraglusib/GnP is not dependent on prolonged therapy.

Extended Data Table 4.

Overall Survival for Patients Who Received ≥4 Cycles Versus <4 Cycles of Treatment – Cox Proportional Hazards Model (mITT Population)

^aCox proportional hazards model that includes treatment, cycle group (>=4 or <4), and treatment-by-cycle group interaction.

Abbreviations: CI, confidence interval; GnP, gemcitabine plus nab-paclitaxel; mITT, modified intention-to-treat; NA, not available.

Several complementary observations support that the OS improvement reflects a treatment effect rather than confounding according to treatment exposure. First, the treatment arms were well balanced at baseline and representative of a pragmatic, real-world mPDAC population, where early attrition is common. Approximately 56% of patients in the GnP arm and 50% in the elraglusib/GnP arm did not receive subsequent therapy (Extended Data Table 1). Notably, even among patients who did not proceed to second-line treatment, a numerical OS advantage was observed with elraglusib/GnP despite a relatively short treatment duration.

Second, fewer early deaths occurred within the first 2 months with elraglusib/GnP versus GnP (16.1% versus 23.1%; Supplementary Table 6), and Kaplan–Meier curves began to diverge after approximately two treatment cycles, with sustained separation thereafter (Fig. 2). Importantly, patients in the GnP arm received a similar or greater number of treatment cycles through cycle 7 (Supplementary Table 1), yet an OS advantage favoring elraglusib/GnP was already evident early in treatment, arguing against cumulative dose exposure as the primary driver of survival differences.

In addition, the OS and PFS were discordant, with no significant difference in PFS despite a statistically significant improvement in OS. Given that PFS is closely linked to treatment duration, this dissociation further supports that prolonged survival cannot be explained solely by longer exposure to therapy.

To address potential immortal time and responder biases, a landmark analysis was performed at 6 months from treatment initiation, including only patients alive at the landmark. As expected, patients remaining on treatment at 6 months demonstrated longer subsequent survival within each treatment arm. However, OS from the landmark remained longer in the elraglusib/GnP arm compared with the GnP arm, supporting a treatment effect beyond time on therapy alone (Supplementary Fig. 7). Taken together, these exploratory analyses consistently demonstrate that the OS benefit observed with elraglusib/GnP cannot be attributed to differential treatment duration, cumulative dosing or survivor bias, and instead support a true treatment-associated survival benefit in this broadly representative mPDAC population.

Overall, elraglusib/GnP provides a manageable safety profile, but elraglusib may exacerbate AEs associated with GnP. Visual impairment (attributed to elraglusib) remains the most common TEAE with elraglusib/GnP, which aligns with findings in earlier studies with elraglusib^27,28. Most visual impairment has mild severity and primarily alters perception of contrast and color tones. This TEAE is transient (typically lasting less than 1 h) and reversible. The complete ophthalmological workup during the phase 1 study showed that elraglusib-related visual impairment did not result in any structural changes to the eye or any permanent effects on vision²⁷.

Furthermore, elraglusib appears to exacerbate the neutropenia associated with GnP as grade 3 or 4 neutropenia is reported more frequently with elraglusib/GnP than with GnP. To date, neutropenia has been observed primarily with elraglusib/GnP and not with elraglusib monotherapy^27,28. The concomitant administration of elraglusib with GnP did not increase plasma levels of nab-paclitaxel or gemcitabine (Supplementary Fig. 5). Our hypothesis is that elraglusib may stimulate the proliferation of neutrophils; GnP preferentially kills these rapidly proliferating cells. In an in vivo study, the administration of a GSK-3 inhibitor improved neutrophil and megakaryocyte recovery, increased progenitors from hematopoietic stem cells (HSCs), and enhanced proliferative reconstitution of HSCs in mice transplanted with human or mouse HSCs³⁸. In humans, use of lithium, an inhibitor of GSK-3, contributes to neutrophilia, which is consistent with this hypothesis³⁹.

Two or greater episodes of treatment-induced grade 3 to 4 neutropenia significantly correlated with better survival in the elraglusib/GnP arm but not in the GnP arm (Supplementary Table 3). Our previous study demonstrated that grade 4 neutropenia significantly correlated with objective clinical response in patients with mPDAC who received elraglusib/GnP²⁸. Most patients who developed grade 3 or 4 neutropenia experienced it during cycle 1 (73.1% with elraglusib/GnP and 60.7% with GnP), which suggests a therapeutic effect rather than a dose accumulation or dose intensity effect affecting the OS.

Mechanistically, GSK-3β inhibition modulates hematopoietic progenitor differentiation but is not associated with direct marrow suppression. Accordingly, the higher rates of neutropenia observed with elraglusib/GnP probably reflect the additive myelosuppressive effects of combination therapy and longer treatment exposure among responders, rather than a specific toxicity or pharmacodynamic marker of elraglusib. Importantly, this association is exploratory and hypothesis-generating; early-onset neutropenia may serve as a pharmacodynamic correlate of adequate drug exposure or biological activity, rather than a consequence of prolonged treatment duration.

mPDAC has multiple immunomodulatory defects that lead to tumor immune resistance, including reduced infiltration with cytotoxic T and NK cells and a microenvironment dominated by helper T cells^40,41. These defects indicate increased exhausted T cell populations, increased MDSCs, reduced ability for antigen presentation and increased immune checkpoint expression. Optimal treatments involve a combination of agents that exerts both tumor cytotoxicity (direct antitumor effect) and immune modulation (indirect antitumor effect). Elraglusib has demonstrated the ability to inhibit growth and survival of cancer cells, attenuate the unfavorable tumor immune microenvironment and decrease tumor cell resistance in a variety of preclinical models^24,30.

Targeting myeloid progenitor cells by GnP leads to a decreased number of MDSCs and neutropenia⁴². In pancreatic cancer, intratumoral MDSCs suppress the activity of T and NK cells and thus contribute to tumor resistance to immunotherapy^43–46. Although the number of MDSCs decreased in posttreatment tumors in both elraglusib/GnP and GnP arms, the depletion of MDSCs in the elraglusib/GnP arm occurred along with concomitant increases in the percentage of tumor-infiltrating cytotoxic T cells, granzyme B⁺ cells and NK cells. Given the limited number of tumor samples analyzed and the heterogeneity of PDAC, these immunophenotyping findings are exploratory and hypothesis-generating rather than statistically inferential. Nonetheless, these observations are consistent with prior preclinical and early clinical data suggesting that elraglusib may favorably modulate the tumor immune milieu when combined with GnP and warrant further evaluation in larger, prospectively designed studies.

CCSG analysis identified high levels of TRAIL (numerical benefits without statistical significance) and CXCL2 as predictors for improved survival in the elraglusib/GnP arm but not in the GnP arm. Systemic cytokines can reflect both the systemic immune/inflammatory environment and the immune status of the tumor microenvironment. Thus, identifying several pre-dose cytokine markers that predict for survival in only patients who received elraglusib/GnP is a unique finding of this study. Further, the cytokines and chemokines identified as correlating with improved survival²⁹ are known to mediate tumor immune suppression. For example, CXCL2 has been shown to increase intratumoral MDSC recruitment⁴⁷, which may promote resistance to checkpoint inhibitors⁴⁸. These associations are presented descriptively with intentions to generate hypotheses regarding potential immunomodulatory effects of elraglusib, rather than to establish causal or mechanistic conclusions.

Initial limitations of this study were its open-label design and regional treatment access as several patients withdrew from the study once randomized to the control arm to pursue alternative treatment options closer to home. Patients who withdrew after randomization were asked to consent to survival follow-up, which allowed for control of selection bias relative to the primary endpoint. Despite the mITT/safety population being selected as the primary analysis set and the best suited population to represent treatment benefit, the ITT population, which includes all patients like those who withdrew after randomization and before receiving treatment, showed a statistically significant improvement in median OS and 1-year survival (Table 3). By design, the trial allowed for the inclusion of patients with poor prognostic features, which potentially led to a higher rate of early deaths across both arms, and consequently, lower median OS for GnP compared with other randomized trials.

This open-label, randomized, phase 2 study revealed improved OS with elraglusib/GnP as first-line treatment in mPDAC when compared to the standard of care (SOC) first-line treatment, GnP. A manageable safety profile of elraglusib/GnP was confirmed and mirrors safety findings from the previous studies with elraglusib^27,28. We hypothesize that the survival benefit will be maintained relative to the control group; therefore, direct comparison of the median OS in the elraglusib/GnP arm with other studies or other regimens (for example, FOLFIRINOX or NALIRIFOX) is inherently difficult. Although the median OS of elraglusib/GnP is relative to the median OS of GnP in this study and may shift if the median OS of GnP differs in another study, the 2.9-month survival advantage would be preserved. Supportive analysis in patients who received at least one cycle of treatment (approximately 75% of patients in this study) demonstrated that this survival improvement margin is maintained in the elraglusib/GnP arm even as survival shifts in the GnP arm (Extended Data Table 5).

Extended Data Table 5.

Kaplan-Meier Analysis for Patients Who Received at Least One Cycle of Treatment

Abbreviations: CI, confidence interval; DCR, disease control rate; GnP, gemcitabine plus nab-paclitaxel; HR, hazard ratio; mPFS, median progression-free survival; NS, not significant; ORR, objective response rate; OS, overall survival.

Elraglusib’s benign safety profile as monotherapy, and a unique mechanism of action with a multifaceted immunomodulatory component, position it as an ideal backbone agent for combination with other first-line chemotherapy regimens such as FOLFIRINOX and NALIRIFOX and immune modulators such as checkpoint inhibitors that have shown little activity to date in mPDAC⁴⁹. A phase 2 investigator-initiated trial (NCT05077800) is exploring the combination of elraglusib with FOLFIRINOX in previously untreated mPDAC. Preliminary data on six patients treated with elraglusib/FOLFIRINOX/losartan demonstrated a PR of 20 or more months in three of six patients with extensive liver metastases and 100% DCR⁵⁰. Other studies are underway to explore elraglusib in combination with immune checkpoint inhibitors, and phase 1 studies are being designed for an oral dosage form of elraglusib. The phase 1 study RiLEY (NCT06896188) will investigate the combination of retifanlimab, an anti-PD-1 monoclonal antibody, with elraglusib and FOLFIRINOX in advanced PDAC. Because of PDAC having multiple defects contributing to resistance to immune modulatory drugs^40,41, combinations of elraglusib with RAS or MEK inhibitors could also be potentially explored as therapeutic strategies to overcome tumor resistance and improve clinical outcomes in mPDAC. Additionally, based on the results of this phase 2 study, a phase 3 study is being planned.

Methods

Study design

This was an open-label, randomized, multicenter, phase 2 study to assess the efficacy and safety of elraglusib/GnP versus GnP as first-line treatment in patients with mPDAC (the 1801 study Part 3B, NCT03678883). The study followed the Pick-the-Winner adaptive trial design, which consists of a selection stage and a confirmation stage⁵¹. Patients enrolled from 7 October 2021 to 9 February 2024. All enrolled patients signed the informed consent form.

Ethics approval and consent

The study was performed in compliance with Good Clinical Practice and the Declaration of Helsinki (v2013). Local institutional review boards or the independent ethics committees of the participating centers approved the study protocol.

Patients

Eligible patients were aged 18 years and older, presented with previously untreated mPDAC and had at least one measurable lesion per Response Evaluation Criteria in Solid Tumors (RECIST) v.1.1⁵² using computed tomography or magnetic resonance imaging. Patients had an ECOG PS of 0 or 1 and adequate bone marrow, liver and renal functions. Any focal radiation therapy or surgery with general anesthesia were completed 7 days before initiating study treatment; any surgery with local anesthesia was completed 3 days before treatment. Any prior chemotherapy treatments in the adjuvant or neoadjuvant setting were completed at least 6 months before study enrollment.

Key exclusion criteria included pregnancy or lactation, presence of endocrine or acinar pancreatic carcinoma or current active malignancy other than pancreatic cancer, a history of cardiovascular disease, major surgery within 7 days of enrollment, participation in another clinical trial, or rapidly progressing brain metastases or leptomeningeal involvement. Patients with stable or slowly progressing brain metastases or leptomeningeal disease were eligible to enroll if no new treatments were initiated in the prior 28 days and anticonvulsants and steroids were at stable doses for the prior 14 days (see full eligibility criteria in the protocol in the Supplementary Information). All enrolled patients signed the informed consent form within 6 weeks of initiating study treatment.

Randomization

During the selection stage of the study, patients were initially randomized in a 1:1:1 ratio to one of the three treatment arms—twice-weekly elraglusib/GnP, once-weekly elraglusib/GnP or GnP. During the confirmation stage of the study after selecting once weekly as the preferred dosing for elraglusib, patients were randomized in a 2:1 ratio to receive once-weekly elraglusib/GnP or GnP alone. Randomization for all stages of the study was stratified according to study center with a block size of three.

Procedures

During the study, intravenous elraglusib 9.3 mg kg⁻¹ was administered on day 1 for the once-weekly schedule or days 1 and 4 for the twice-weekly schedule. The elraglusib’s recommended phase 2 dose of 9.3 mg kg⁻¹ was based on the nonrandomized, single-arm, phase 2 study²⁸. GnP treatment consisted of intravenous nab-paclitaxel 125 mg m⁻² followed by intravenous gemcitabine 1,000 mg m² on days 1, 8 and 15 in a 28-day cycle. Dose reductions were allowed for all treatment agents. For elraglusib, the first dose reduction was to 7.0 mg kg⁻¹ dose and the second to 5.0 mg kg⁻¹. Treatment continued until disease progression (as defined in RECIST v.1.1⁵²), unacceptable toxicities or the investigator’s decision that a patient no longer benefited from the treatment.

Radiological tumor assessments were performed per institutional SOC using computed tomography or magnetic resonance imaging. Serum CA 19-9 levels were measured per SOC in all patients with elevated CA 19-9 at baseline and were followed for 12 months after the last dose of study treatment. During the follow-up, survival was followed in all patients for 12 months after the last dose of study treatment and then every 3 months until all patients died or the last patient was followed for 12 months. Safety assessments were performed up to 4 weeks after the last dose of the study treatment.

Study objectives and assessments

In the selection stage, the superiority of elraglusib dosing frequency was determined based on tolerability, toxicity and patient convenience.

The initial primary endpoint was the 1-year landmark survival rate in the different treatment arms. This allowed to determine the elraglusib schedule during the selection stage without waiting for OS. After the 12-month follow-up after the last treatment dose, the overall primary endpoint was the median OS, defined as the number of months from randomization to death from any cause. Additional sensitivity and exploratory analyses for OS were conducted for different patient subgroups.

Secondary endpoints included PFS, TTF, ORR, DCR, DOR and safety. PFS was defined as the time from randomization to PD or death; TTF as time from randomization to disease progression, clinical progression (for example, clinical deterioration, clinical decision because of complicating disease, worsening disease, etc.) or death; ORR as the percentage of patients with CR or PR; and DOR as the median number of months from the first documented CR or PR to disease progression. DCR was defined as the percentage of patients who achieved stable disease lasting 16 weeks or longer, CR or PR. All responses to treatment were assessed according to RECIST v.1.1⁵². TEAEs were monitored from the first dose of treatment to 30 days after the last dose, using the Common Terminology Criteria for Adverse Events v.5.0 and coded using Medical Dictionary for Regulatory Activities v.23.1⁵³.

The primary analysis set for the efficacy and safety analyses was the safety population (or mITT population), defined as patients who received at least one dose of the study treatment. The ITT population consisted of all randomized patients. The mITT population was selected as the primary analyses set for the efficacy analyses to account for potential patient withdrawal in the ITT population after randomization in this open-label study as described in ref. ⁵⁴.

Biomarker analysis

Blood plasma samples were collected pre-dose and 3 h post-dose on day 1 of cycles 1, 2, 4 and 6 from all patients. Using a Luminex immunoassay, 40 CCSGs were measured across the plasma samples²⁹. For the OS analysis, both arms were stratified according to their pre-dose levels about the median level of each CCSG in the elraglusib/GnP arm.

Optional fresh tumor biopsies were obtained within 6 weeks before the first treatment dose and after cycle 2 of treatment. Biopsy samples were fixed with formalin and embedded in paraffin within 24 h of collection. To determine the difference in the percentage of intratumoral immune cells (CD8⁺, granzyme B⁺, CD56⁺ and CD11b⁺) before and after treatment in elraglusib/GnP and GnP arms, immunophenotyping was performed of these biopsy tumor samples using the OmniVUE 8-Plex immunofluorescence assay (Ultivue). The percentage of intratumoral immune cells was determined by counting the total number of CD8+, granzyme B⁺, CD56⁺ or CD11b⁺ cells within the tumor parenchyma area and by dividing the total number of cells in that same area.

GSK-3β expression was assessed using immunohistochemical staining of pre-dose formalin-fixed paraffin-embedded tumor samples. Immunohistochemical staining, histopathological evaluation and scoring were performed by Mosaic Laboratories.

Statistical analysis

The predefined sample size for the selection stage was up to 25 patients in each of the three cohorts. The planned sample size to evaluate the primary efficacy endpoint of the 1-year survival rate was 207 evaluable patients (130 patients for elraglusib/GnP and 77 patients for GnP), which also included patients in the appropriate treatment arms from the selection stage. Primary analysis for survival and safety included all patients in the mITT analysis set. The calculation assumptions included achieving a 1-year survival rate of 55% with elraglusib/GnP and 35% with GnP (20% difference between the arms) with 80% power using a chi-squared test for equal proportions at a two-sided significance level of 0.05. Accordingly, the primary analysis was originally planned as a comparison of the 1-year survival rates. Following regulatory feedback after study initiation, the primary efficacy analysis was revised to a comparison of survival curves using a log-rank test; the sample size was not reestimated. The log-rank test was selected to appropriately account for censored survival times occurring before 1 year. In accordance with the Pick-the-Winner trial design, statistical significance for the primary endpoint was assessed at an adjusted two-sided significance level of 0.048, corresponding to a log-rank test statistic greater than 3.91 (ref. ⁵⁵).

Descriptive statistics were used to summarize the efficacy and safety parameters. Categorical data were represented as frequency distributions (numbers and percentages of patients), while continuous data as the mean (s.d.) and median (range of minimum to maximum). For all time-to-event efficacy endpoints, the log-rank test statistic was used for hypothesis testing, Kaplan–Meier methodology (including figures, median, 95% CI, number of events and number censored) for the summary of variables, and Cox proportional hazards model with treatment arm as the independent variable for HRs. For continued OS, time censoring occurred on the date of the most recent assessment of any type before the loss of follow-up or the database lock. The Cochran–Mantel–Haenszel test statistic was used for the response rates. Additionally, machine-learning-based multivariate models were used for cytokine and survival correlative analysis²⁹.

All statistical analyses were performed using SAS v.9.4 (see the statistical analysis plan in the Supplementary Information for details on the statistical analyses).

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Online content

Any methods, additional references, Nature Portfolio reporting summaries, source data, extended data, supplementary information, acknowledgements, peer review information; details of author contributions and competing interests; and statements of data and code availability are available at 10.1038/s41591-026-04327-4.

Author contributions

D.M., B.A.C., A.P.M., A.U. and T.W. conceptualized the study and developed the methodology. D.M., A.P.M. and C.S. supervised the project. D.M., R.T.S., B.A.C., Y.J., A.L.C., A.C., V.S., A.P., S.H., N.K.L., I.J.P., C.D.L., S.P., P.K., M.M., R.C., F.J.G., A.P.M. and T.S.B. were responsible for data collection. G.F., M.J. and T.W. performed the statistical analysis. D.M. and A.P.M. wrote the original draft. All authors reviewed and edited the original draft of the manuscript and reviewed the final draft to approve its submission for publication.

Peer review

Peer review information

Nature Medicine thanks David Chang, Mithat Gonen and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. Peer reviewer reports are available. Primary Handling Editor: Ulrike Harjes, in collaboration with the Nature Medicine team.

Data availability

Actuate Therapeutics is committed to sharing access to patient-level data and supporting clinical documents from eligible studies with qualified external researchers. Data will be made available after a Data Usage Agreement has been signed. Investigators and institutions who consent to the terms of the Data Usage Agreement, including but not limited to the use of these data for the purpose of a specific project and only for research purposes, and to protect the confidentiality of the data and limit the possibility of identification of participants in any way whatsoever for the duration of the agreement, will be granted access. All data provided are anonymized to respect the privacy of patients who have participated in the trial in line with applicable laws and regulations. Trial data availability is according to the criteria and process described at www.clinicalstudydatarequest.com. The data described in this publication will be made available for any qualified request. To submit a request for data, please contact info@actuatetherapeutics.com. Response to requests will be provided within 15 business days.

Competing interests

D.M. declares research funding from Amgen, Merck, Oncolytics and Rafael; scientific advisory board work for Actuate Therapeutics and Qurient; an advisory/speaker bureau for Amgen, Bristol Myers Squibb, Eisai and Exelixis; and institutional research funding from Acepodia, Actuate Therapeutics, ADC Therapeutics, Amgen, AVEO, Bayer, Blueprint Medicines, Bristol Myers Squibb, BioNTech, Dialectic Therapeutics, Epizyme, Fujifilm, ImmuneSensor, Immune-Onc Therapeutics, Leap Therapeutics, Lycera Corp, Merck, Millennium, MiNA Alpha, NGM Biopharmaceuticals, Novartis, Oncolytics, Orano Med, Puma, Qurient, Rafael, Repare Therapeutics, Triumvira Immunologics, Vigeo Therapeutics and Warewolf Therapeutics. R.T.S. declares institutional research funding from Lisata, Merck, Exelixis, Bristol Myers Squibb, Immuno Vaccine, Compass, Actuate, Faeth and Seagen; consultation for AstraZeneca, Boston Scientific, Genentech, Elevar, ICON Clinical Research and Foundation Medicine; and participation in advisory boards for Exelixis, Merck, AstraZeneca, Boehringer Ingelheim and Zymeworks. B.A.C. declares institutional research support from AstraZeneca, Abbvie, Actuate Therapeutics, Astellas, Agenus, Bayer, Dragonfly Therapeutics, Mink Therapeutics, Pfizer, Pyxis Oncology, Repare Therapeutics and Regeneron; and participation in advisory boards for Seattle Genetics and Eisai. A.L.C. declares consultation or participation in advisory boards for Halozyme, Seattle Genetics, Merrimack and Abbvie; institutional research funding from XBiotech, Newlink Genetics, Taiho Pharmaceutical, Immunomedics, Onconova Therapeutics, Lilly, Gilead Sciences, Genentech, Seattle Genetics, AbGenomics International, Halozyme, Novocure, Amgen, Actuate Therapeutics, Surface Oncology, Nucana, Nextrast and AstraZeneca; and travel support from Halozyme, Abbvie and Nucana. A.C. declares institutional research funding from Genentech, Merck Serono, Bristol Myers Squibb, Merck Sharp & Dohme, Roche, Beigene, Bayer, Servier, Lilly, Novartis, Takeda, Astellas, Natera and Actuate Therapeutics; and participation in advisory board or receipt of speaker fees from AbbVie, Amgen, Merck Serono, GlaxoSmithKline and Roche. V.S. declares institutional research funding from Actuate Therapeutics, Boehringer Ingelheim, Bristol Myers Squibb, Clovis, Esanik, Exelixis, Fibrogen, Ipsen, MedImmune, the National Cancer Institute, PanCAN, Cornerstone (previously Rafael), Relay, Repare, Servier, Syros and Transthera; consulting fees from Amplity (previously Lynx Group), AstraZeneca, Autem, Delcath Systems, Histosonics, Ipsen, Incyte, Rafael (previously Cornertsone), Servier and Taiho; receipt of equipment and supplies from Beigene and Cornerstone (previously Rafael); and travel support from ASCO, the Cholangiocarcinoma Foundation, Histosonics, Lynx Group/Amplity and the National Comprehensive Cancer Network. N.K.L. declares institutional research funding from Exact Sciences. C.D.L. declares institutional research funding from Servier, AstraZeneca, Genentech/Roche and Taiho. F.J.G. declares stock ownership in Actuate Therapeutics. C.S. is an employee and equity holder in Actuate Therapeutics. A.U. is an employee and equity holder in Actuate Therapeutics. T.W., G.F. and M.J. declare consulting fees from Actuate Therapeutics. A.P.M. is an employee and equity holder in Actuate Therapeutics. T.S.B.-S. declares institutional research funding from Actuate Therapeutics, Boehringer Ingelheim, Bristol Myers Squibb, Clovis, Elicio, Esanik, Exelixis, Fibrogen, Jazz, Ipsen, the National Cancer Institute, PanCAN, Cornerstone (previously Rafael), Relay, Repare, RevMed, Servier and Transthera; consulting fees from AstraZeneca, Autem, Amplity (previously Lynx Group), Delcath, Elevar, HistoSonics, Incyte, Ipsen, Jazz, RevMed, Servier, Tallac and TransThera; provision of equipment/supplies from Cornerstone (previously Rafael); travel and lodging costs from ASCO, the Japanese Society of Clinical Oncology, the BinayTara Foundation, the Cholangiocarcinoma Foundation, HistoSonics, Lynx Group/Amplity, the National Comprehensive Cancer Network and the Chinese Association of Pancreatology. Y.J., A.P., S.H., I.J.P., S.P., P.K., M.M. and R.C. declare no competing interests.

Extended data

is available for this paper at 10.1038/s41591-026-04327-4.

Supplementary information

The online version contains supplementary material available at 10.1038/s41591-026-04327-4.