Metformin boosts antitumor immunity and improves prognosis in upfront resected pancreatic cancer: an observational study

Abstract

Background

Beyond demographic and immune factors, metabolic considerations, particularly metformin’s recognized impact in oncology, warrant exploration in treating pancreatic cancer. This study aimed to investigate the influence of metformin on patient survival and its potential correlation with distinct immune profiles in pancreatic ductal adenocarcinoma (PDAC) tumors.

Methods

We included 82 upfront resected and 66 gemcitabine-based neoadjuvant chemoradiotherapy (nCRT)-treated patients from the PREOPANC randomized controlled trial (RCT). Transcriptomic NanoString immunoprofiling was performed for a subset of 96 available resected specimens.

Results

Disparities in survival outcomes and immune profiles were apparent between metformin and non-metformin users in upfront resected patients but lacking in nCRT-treated patients. Compared to non-metformin users, upfront resected metformin users showed a higher median overall survival (OS) of 29 vs 14 months and a better 5-year OS rate of 19% vs 5%. Furthermore, metformin use was a favorable prognostic factor for OS in the upfront surgery group (HR = 0.56; 95% CI = 0.32 to 0.99). Transcriptomic data revealed that metformin users significantly underexpressed genes related to pro-tumoral immunity, including monocyte to M2 macrophage polarization and activation. Furthermore, the relative abundance of anti-inflammatory CD163+ MRC1+ M2 macrophages in non-metformin users and immune-activating CD1A+ CD1C+ dendritic cells in metformin users was heightened (P < .001).

Conclusion

This study unveils immune profile changes resulting from metformin use in upfront resected pancreatic cancer patients, possibly contributing to prolonged survival outcomes. Specifically, metformin use may decrease the abundance and activity of pro-tumoral M2 macrophages and increase the recruitment and function of tumor-resolving DCs, favoring antitumor immunity.

[PREOPANC trial EudraCT: 2012-003181-40]

In pancreatic ductal adenocarcinoma (PDAC), mortality rates closely parallel its incidence. Projections anticipate PDAC to become the second leading cause of cancer-related death by 2030 (1,2). Despite improvements in survival outcomes among 13% of localized PDAC cases, primarily attributed to surgical resection (3,4), the substantial risk of disease recurrence keeps overall survival suboptimal, with only 4% of resected patients surviving beyond 10 years (5,6). Integrating neoadjuvant and adjuvant chemo(radio)therapies has emerged as a strategic approach to enhance further survival outcomes in resectable and borderline resectable PDAC patients (7,8).

Diabetes mellitus (DM), a prominent PDAC risk factor, is a chronic condition involving dysregulated glucose metabolism due to insufficient insulin or impaired insulin action, leading to elevated blood glucose levels and metabolic disturbances. Increasing the risk of various cancers, including PDAC, by twofold (9), DM may also be a symptom of PDAC (10). Notably, approximately 50% of PDAC patients exhibit either type 2 DM, the most prevalent form (11), or impaired glucose tolerance in the early stages (10,12).

In DM patients, metformin is the most prescribed drug to mitigate hyperglycemia. This hypoglycemic agent suppresses hepatic glucose production, lowers blood glucose levels, and increases insulin sensitivity by promoting glucose uptake in skeletal muscles (13-15). Metformin may play an important role in oncology. Metformin suppressed the proliferation of PDAC cells by inhibiting the mammalian target of rapamycin (mTOR) activation and insulin-like growth factor (IGF-1) receptor signaling pathway (16), and metformin use was associated with improved survival outcomes in various cancers, including PDAC (17,18). However, caution is warranted due to the generally low quality of evidence linking metformin to reduced PDAC mortality, attributed to methodological shortcomings in retrospective analyses (19-22). Additionally, two phase II trials in patients with advanced PDAC reported no improved survival with metformin use (23,24). Given these inconsistencies and methodological limitations of existing studies, further research with prospective designs is crucial to elucidate the impact of metformin on survival in PDAC patients.

Examinations of the immunomodulatory effects of metformin have unveiled its multifaceted impact on both innate and adaptive immune systems (25). The immune system plays a complex role in controlling PDAC progression. Inflammation is associated with PDAC development, growth, and invasion (26), while the immunosuppressive tumor microenvironment is a hallmark of pancreatic cancer (27). This microenvironment allows tumor cells to evade adaptive T-cell responses, conferring protection against the immune system and resisting immunotherapy (28). Notably, a comprehensive exploration of metformin’s exact immunological role in PDAC remains elusive.

In this study, we conducted a post hoc analysis examining the influence of metformin on survival outcomes and tumor immunity in a prospective randomized controlled trial (RCT) cohort of resected PDAC patients. Our emphasis on resected PDAC patients, with available tumor samples, enabled a comprehensive exploration of the multifactorial nature of the disease, encompassing interactions among metabolic factors, the immune system, stromal cells, and tumor cells.

Methods

Patient cohort

This study included patients with pathologically confirmed PDAC who underwent surgical resection as part of the PREOPANC phase III RCT (EudraCT number 2012-003181-40) across 16 high-volume pancreatic surgery centers from the Dutch Pancreatic Cancer Group (DPCG). This trial was conducted according to the guidelines of the Declaration of Helsinki and approved by the Ethics Committees of Erasmus MC (MEC-2012-249; December 11, 2012). Written informed consent was obtained from all patients. Patients were randomly assigned to either upfront surgery or gemcitabine-based neoadjuvant chemoradiotherapy (nCRT). Detailed eligible criteria and clinical procedures are outlined in the protocol (29) and the Supplementary Methods (available online). The long-term results have been previously published (30). This study further stratified patients based on metformin use. Patients prescribed metformin at the time of random assignment for at least six months, adhering to standard metformin dosages ranging from 500 to 2000 mg per day as recommended by their treating physicians, were considered “metformin users.” Patients without any history of metformin use were considered “non-metformin users.” Overall survival (OS), calculated from PDAC diagnosis (histologically or cytologically confirmed) to death, was considered the primary survival outcome, with censoring for patients alive at the last follow-up. Demographic data were obtained through investigator observation.

Transcriptomic immunoprofiling using NanoString technologies

To elucidate tumoral immunological landscapes associated with metformin use, we reanalyzed targeted gene expression profiles of the formalin-fixed and paraffin-embedded (FFPE) surgical specimens tissue that were previously fabricated using NanoString technologies (7) (Supplementary Tables 1 and 2, available online). Detailed descriptions of the NanoString measurements are available in the Supplementary Methods (available online).

Statistical analyses

R Statistical Software (v4.1.2) was used for downstream data exploration, statistical analyses, and visualizations, with detailed descriptions in the Supplementary Methods (available online). P values less than .01 were considered statistically significant, denoted as follows: ■ P value less than .05, *P value less than .01, **P value less than .001.

Results

Patient characteristics

Between April 2013 and July 2017, 164 out of 248 patients with borderline resectable or resectable PDAC in the phase III PREOPANC RCT underwent surgical resection. This study focused on 148 patients after the exclusion of 12 patients without pathologically confirmed PDAC and 4 patients who did not complete the full course of gemcitabine-based nCRT. The distribution of metformin use was comparable between the 82 upfront resected (21 metformin, 61 non-metformin) and 66 nCRT patients (18 metformin, 48 non-metformin). No significant differences were observed in preoperative clinical characteristics or postoperative pathological and surgical outcomes within the metformin subgroups. The median age was higher in the non-metformin group (70 years) compared to the metformin users (60 years) in the nCRT group (P = .021) (Table 1).

Table 1.

Preoperative clinical characteristics of 148 resected PDAC patients

Treatment group	Upfront surgery (n = 82)			nCRT (n = 66)
	Non-metformin	Metformin	P	Non-metformin	Metformin	P
	(n = 61)	(n = 21)		(n = 48)	(n = 18)
Sex
Female	25 (41)	10 (48)	.62	24 (50)	7 (39)	.58
Male	36 (59)	11 (52)		24 (50)	11 (61)
Age at diagnosis (years)
Median [min, max]	68 [40, 80]	65 [45, 78]	.25	67 [42, 80]	64 [51, 76]	.14
BMI (kg/m2)
Median [min, max]	25 [18, 43]	25 [18, 35]	.41	25 [19, 44]	25 [19, 32]	.54
Diabetes mellitus, no. (%)
No	40 (66)	15 (71)	.79	38 (79)	14 (78)	>.99
Yes	21 (34)	6 (29)		10 (21)	4 (22)
Hypercholesterolemia, no. (%)
No	55 (90)	19 (90)	>.99	41 (85)	16 (89)	>.99
Yes	6 (10)	2 (10)		7 (15)	2 (11)
Hypertension, no. (%)
No	46 (75)	15 (71)	.78	34 (71)	13 (72)	>.99
Yes	15 (25)	6 (29)		14 (29)	5 (28)
History of cardiovascular disease, no. (%)
No	44 (72)	16 (76)	.78	32 (67)	11 (61)	.77
Yes	17 (28)	5 (24)		16 (33)	7 (39)
History of cancer, no. (%)
No	53 (87)	21 (100)	>.99	41 (85)	16 (89)	>.99
Yes	8 (13)	0 (0)		7 (15)	2 (11)
History of pancreatitis, no. (%)
No	58 (95)	21 (100)	.11	45 (94)	14 (78)	.082
Yes	3 (5)	0 (0)		3 (6)	4 (22)
Resectability, no. (%)
Borderline resectable	25 (41)	10 (48)	.70	22 (46)	5 (28)	.26
Resectable	36 (59)	11 (52)		26 (54)	13 (72)
CA19-9, preoperative (U/mL)
Median [min, max]	263 [1, 12000]	227 [1, 4274]	.69	147 [2, 4114]	98 [8, 1460]	.29
Missing (% of total)	10 (16)	5 (24)		5 (10)	2 (11)
Involvement of the SMA, no. (%)
Absent	58 (95)	19 (90)	.60	42 (88)	17 (94)	.66
Present	3 (5)	2 (10)		6 (12)	1 (6)
Tumor diameter, before nCRT (mm)
Median [min, max]	30 [15, 60]	35 [20, 55]	.14	31 [13, 64]	28 [17, 40]	.46
Missing (% of total)	2 (3)	0 (0)		2 (4)	0 (0)
Tumor diameter, after nCRT (mm)
Median [min, max]	30 [15, 60]	35 [4, 55]	.18	27 [13, 62]	27 [14, 50]	.81
Missing (% of total)	2 (3)	0 (0)		2 (4)	0 (0)
Regional suspicious lymph nodes, no. (%)
Absent	49 (80)	13 (62)	.14	32 (67)	15 (83)	.23
Present	12 (20)	8 (38)		16 (33)	3 (17)
Tumor location, no. (%)
Corpus or tail	7 (11)	3 (14)	.71	10 (21)	2 (11)	.49
Head	54 (89)	18 (86)		38 (79)	16 (89)
WHO performance status, no. (%)
WHO 0	18 (30)	9 (43)	.28	27 (56)	12 (67)	.40
WHO 1	41 (67)	11 (52)		20 (42)	5 (28)
Missing (% of total)	2 (3)	1 (5)		1 (2)	1 (6)
Response to nCRT (RECIST 1.1), no. (%)
Progressive disease	0 (0)	0 (0)	–	6 (12)	0 (0)	–
Partial response	0 (0)	0 (0)		29 (60)	14 (78)
Stable disease	0 (0)	0 (0)		3 (6)	2 (11)
Missing (% of total)	61 (100)	21 (100)		10 (21)	2 (11)
Institute, no. (%)
Academical hospital	43 (70)	16 (76)	.78	33 (69)	16 (89)	.12
General hospital	18 (30)	5 (24)		15 (31)	2 (11)

BMI = body mass index; CA19-9 = carbohydrate antigen 19-9; nCRT = neoadjuvant chemoradiotherapy; PDAC = pancreatic ductal adenocarcinoma; RECIST = Response Evaluation Criteria in Solid Tumors; SMA = superior mesenteric artery; WHO = World Health Organization.

Metformin users with upfront resected PDAC exhibit prolonged survival outcomes compared to those not using metformin

The median follow-up among all patients was 73 months. In the upfront surgery group, non-metformin users experienced more progression or death events than metformin users (Table 2). Survival disparities were evident, with metformin users showing a higher median OS of 29 months (95% CI = 23 to 46) and a 5-year OS rate of 19% (95% CI = 8 to 46), compared to non-metformin users with a median OS of 14 months (95% CI = 11 to 19) and a 5-year OS rate of 5% (95% CI, = 2 to 15) (Figure 1, A). This survival difference was not observed in the nCRT group, with a comparable median OS of 32 months (95% CI = 16 to 53) for metformin users and a median OS of 30 months (95% CI = 21 to 95) for non-metformin users (Figure 1, A).

Figure 1.

Survival analysis of metformin and non-metformin users. (A) Kaplan-Meier curves stratified by treatment illustrate different OS in months associated with metformin use in treatment-naive patients but not in nCRT-treated patients. Treatment-naive patients using metformin (green) show prolonged OS compared to non-metformin users (red). The x-axis displays the overall survival (months), and the y-axis displays the survival probability (%). (B) Forest plots, stratified by treatment, visualize univariate and multivariate Cox proportional hazard models highlighting metformin use as an independent prognostic factor for OS in treatment-naive patients but not in nCRT-treated patients. Squares highlighted in green denote statistical significance (P.adj < .05) for the respective covariate. CI = confidence interval; nCRT = neoadjuvant chemoradiotherapy; OS = overall survival; P.adj = adjusted P value.

Table 2.

Postoperative clinical characteristics of 148 resected PDAC patients

Treatment group	Upfront surgery (n = 82)			nCRT (n = 66)
	Non-metformin	Metformin	P	Non-metformin	Metformin	P
	(n = 61)	(n = 21)		(n = 48)	(n = 18)
Nodal status (N) postoperative, no. (%)
N0	12 (20)	3 (14)	.75	33 (69)	11 (61)	.57
N1	49 (80)	18 (86)		15 (31)	7 (39)
Perineural invasion postoperative, no. (%)
Absent	10 (16)	4 (19)	>.99	23 (48)	9 (50)	>.99
Present	48 (79)	17 (81)		22 (46)	8 (44)
Missing	3 (5)	0 (0)		3 (6)	1 (6)
Resection classification (R) postoperative, no. (%)
R0	25 (41)	10 (48)	.62	35 (73)	13 (72)	>.99
R1	36 (59)	11 (52)		13 (27)	5 (28)
SAE reported (any grade), no. (%)
No	35 (57)	11 (52)	.80	19 (40)	6 (33)	.78
Yes	26 (43)	10 (48)		29 (60)	12 (67)
Tumor grade, postoperative, no. (%)
Moderately differentiated	33 (54)	10 (48)	.84	18 (38)	10 (56)	.31
Poorly differentiated	14 (23)	4 (19)		16 (33)	3 (17)
Well differentiated	6 (10)	3 (14)		5 (10)	1 (6)
Missing	8 (13)	4 (19)		9 (19)	4 (22)
Tumor stage (T) postoperative, no. (%)
T1/T2	2 (3)	0 (0)	>.99	12 (25)	5 (28)	>.99
T3/T4	59 (97)	21 (100)		36 (75)	13 (72)
Type of resection, no. (%)
Pancreas body and tail resection	2 (3)	1 (5)	.39	7 (15)	1 (6)	.58
Pancreatoduodenectomy	57 (93)	18 (86)		40 (83)	17 (94)
Total pancreatectomy	2 (3)	2 (10)		1 (2)	0 (0)
Vascular invasion postoperative, no. (%)
Absent	22 (36)	7 (33)	>.99	27 (56)	12 (67)	.40
Present	37 (61)	12 (57)		20 (42)	5 (28)
Missing	2 (3)	2 (10)		1 (2)	1 (6)
Adjuvant gemcitabine completed, no. (%)
No	41 (67)	14 (67)	>.99	32 (67)	13 (72)	.77
Yes	20 (33)	7 (33)		16 (33)	5 (28)
Adjuvant gemcitabine cycles
Median [min, max]	4 [0, 6]	4 [0, 6]	.62	3 [0, 4]	3 [0, 4]	.38

nCRT = neoadjuvant chemoradiotherapy; SAE: serious adverse event.

Univariate Cox proportional hazard models investigating covariates associated with OS in both treatment groups corroborated earlier observations (Supplementary Table 3, A, available online). The use of metformin favored OS in the upfront surgery group (hazard ratio [HR] = 0.46; 95% CI = 0.26 to 0.79; P.adj = 0.010) but not in the nCRT group (HR = 1.67; 95% CI = 0.92 to 3.05; P.adj = 0.53) (Figure 1, B). Multivariate analysis, correcting for the potential confounders of “resection classification (R)” and “the number of received adjuvant gemcitabine cycles,” confirmed metformin as a favorable prognostic factor for OS in the upfront surgery group (HR = 0.56; 95% CI = 0.32 to 0.99; P.adj = 0.047) (Figure 1, B). Similar results were found in the survival analysis for progrerssion-free survival (PFS), with an even more pronounced favorable impact of metformin on PFS in upfront resected patients (Supplementary Figure 1, Supplementary Table 3, B, available online). Given the importance of age and body mass index (BMI) as prognostic biological factors, we repeated the multivariate Cox regression models for the upfront surgery group, incorporating these covariates. These analyses revealed consistent results, with metformin retaining its significance as a prognostic factor (Supplementary Figure 2, available online).

Metformin use is associated with unique transcriptomic alterations in upfront resected but not in nCRT-treated PDAC tumors

Among the 148 PDAC patients included in our study, RNA isolation succeeded for 125 surgical specimens. After stringent quality control of tissue RNA and NanoString data, we conducted immunoprofiling analysis on a subset comprising 46 upfront resected patients (13 metformin and 33 non-metformin users) and 50 gemcitabine-based nCRT-treated patients (15 metformin and 35 non-metformin users). Within the NanoString patient subset, preoperatively, a lower incidence of DM, hypercholesterolemia, and hypertension was noted in non-metformin users across both upfront surgery and nCRT groups (Supplementary Table 4, available online). Cox proportional hazard models applied to the NanoString subset exhibited a trend akin to the total cohort, although statistical significance was not achieved, likely due to the smaller sample size (Supplementary Table 5, available online).

The transcriptomic NanoString expression data, encompassing 730 immuno-oncology-related genes, underwent visual exploration using t-distributed Stochastic Neighbor Embedding (t-SNE) dimensionality reduction analysis. Upfront resected patients exhibited distinct clustering based on metformin use, implying that metformin usage prompts a unique transcriptomic profile (Figure 2, A). Correspondingly, differentially expressed (DE) gene analysis (Supplementary Table 6, available online) uncovered that metformin users exhibited significant overexpression of one gene and significant underexpression of 20 genes (P < .01; Figure 2, B). In contrast, gemcitabine-based nCRT-treated patients did not cluster based on metformin use, and only three genes (BCL6, CD99, and MICA) were significantly overexpressed in metformin users (P < .01; Figure 2, B).

Figure 2.

Exploration analysis of the transcriptomic NanoString expression data from the 96 PDAC specimens. (A) t-SNE biplots illustrating the expression profile of 730 immune-related genes, stratified by treatment and grouped by metformin (green) and non-metformin (red) users. Clear clusters of patients based on metformin use are observed in the treatment-naive group, whereas no such clustering is present in gemcitabine nCRT-treated patients. Each dot represents a patient, and the x-axis and y-axis represent the first and second t-SNE dimensions. (B) Volcano plots, stratified by treatment, illustrate the DE genes between metformin (green) and non-metformin (red) users. The treatment-naive group exhibits a greater number of DE genes compared to the nCRT-treated group. The x-axis displays the log2 fold of change, while the y-axis displays the -log10 P value. Each dot represents a gene, and gene names indicate that they have exceeded the significance threshold of P less than   .01. Genes on the right (positive) are overexpressed in metformin users, whereas genes on the left are overexpressed in non-metformin users. DE = differentially expressed; nCRT = neoadjuvant chemoradiotherapy; PDAC = pancreatic ductal adenocarcinoma; t-SNE = t-distributed stochastic neighbor embedding.

Transcriptomic alterations in upfront resected PDAC tumors of metformin users promote anticancer (immunological) properties

The DE genes underwent pathway overrepresentation analysis to unravel the immune and biological processes associated with metformin use in the upfront surgery group (Supplementary Table 7, available online). The hypoglycemic function of metformin was corroborated by the significant enhancement of the low-density lipoprotein (LDL) Gene Ontology (GO) pathway in non-metformin compared to metformin users (P.adj <.01; Figure 3, A). Metformin may exert an indirect upregulating effect on peripheral LDL levels by diminishing glucose production and enhancing insulin sensitivity (31). Furthermore, various GO pathways related to immunity were significantly altered. In non-metformin users, pathways linked to positive regulation of macrophage processes (activity, differentiation, and migration) and T helper 2 cell differentiation were significantly enhanced, whereas that related to CD4 T cell differentiation was significantly diminished (P.adj <.01; Figure 3, A). The significant enhancement in the IgG binding pathway in non-metformin users further underscores the macrophage’s heightened levels of opsonization and phagocytosis facilitated through the signaling of IgG antibodies (P.adj <.01). Intriguingly, BioCarta interleukin (IL) pathways associated with M2 macrophage polarization (IL-4 and IL-6) and anti-inflammation (IL-10) were heightened in non-metformin users (Figure 3, A, P.adj <.01), indicating that the amplification in macrophage-related pathways predominantly involved M2 phenotypic processes.

Figure 3.

Detailed analysis of the transcriptomic NanoString expression data from the 46 upfront resected PDAC specimens. (A) Bar plots illustrate the results of the overrepresentation analysis for pathways characterized by a gene set size of ≤40. Only pathways exhibiting a proportional overlap of ≥ 10% (left of the x-axis) and an enrichment P.adj < .01 (right of the x-axis) are presented. Each bar represents a unique pathway, and all pathways are enhanced in non-metformin users (ie, diminished in metformin users). (B) Boxplots illustrating the mRNA-based immune cell score (y-axis) of tumor-infiltrating cells (x-axis). The relative abundance of DCs, M2 macrophages, and monocytes was higher in non-metformin (red) than in metformin (green) users. (C-G) Boxplots, stratified by metformin (green) and non-metformin (red) users, illustrate the log2 gene expression count (y-axis) of DE genes (P < .01) and several genes with trends (P < .05) (x-axis). In metformin compared to non-metformin users, genes related to monocyte recruitment (C), M2 macrophage polarization (D), M2 macrophage activation (E), and immune checkpoints (F) are overexpressed, whereas a gene related to DC development and function (G) were underexpressed. In panels A and C to G, each dot represents a patient. DC = dendritic cell; DE = differentially expressed; GOBP = gene ontology biological processes; nCRT = neoadjuvant chemoradiotherapy; PDAC = pancreatic ductal adenocarcinoma; ■P value  less than .05, *P value  less than .01, **P value less than .001.

Gene expression-based immune cell type profiling was conducted to quantify the abundance of intra-tumoral immune cells (Supplementary Figure 3, available online). In upfront resected patients, the abundance of pro-tumoral and anti-inflammatory CD163+ MRC1+ M2 macrophages (P < .001) was significantly lower in the tumor microenvironment (TME) of metformin users, with a similar trend toward decreased abundance of CD14+ CD33+ monocytes (P < .05, Figure 3, B). Correspondingly, genes related to monocyte recruitment to inflammatory sites (CCL2, CCR2, CXCR4, ITGB2) were significantly underexpressed in metformin users (P < .01; Figure 3, C). The abundant monocytes in non-metformin users were also more likely to polarize into M2 macrophages, evident from the significant overexpression of CSF1, CSF1R, IL13, IL13RA1, IL34, IL10RA, IL4R, IL6, JAK3, and VEGFA, all associated with M2 macrophage polarization (P < .01, Figure 3, D). Additionally, intra-tumoral M2 macrophages in non-metformin users were more active, suggested by the overexpression of ANXA1, AXL, CD33, FCGR2A, and TNFAIP3 (P < .01, Figure 3, E).

Furthermore, the abundance of immune-activating CD1A+ CD1C+ DCs (Figure 3, B) and the expression of FLT3 (Figure 3, F), crucial in the development and function of DCs, were significantly higher in metformin-using upfront resected patients (P < .01). Lastly, LAG3 and STAT3, genes involved in immune checkpoint inhibitory processes, were significantly underexpressed in metformin-using upfront resected patients (P < .01), with a similar trend toward underexpression of the gene encoding the immune checkpoint inhibitor CD96 (P < .05; Figure 3, G). Cox proportional hazard models exploring the prognostic impact of the relative immune cell counts yielded no significant results (Supplementary Table 8, available online).

To validate that the observed immunological effects were due to metformin rather than differences in BMI (or cachexia), we compared immune profiles between patients with BMI ≤ median and those with BMI more than median (Supplementary Table 9, available online). In the upfront surgery group, we found no significant differences in immune cell presence based on BMI. Furthermore, there was no overlap in differentially expressed genes between the BMI and metformin analyses, suggesting that BMI had minimal impact on the reported immunological effects of metformin.

These immunological findings collectively support the concept of a metformin-induced antitumoral immune response in upfront resected PDAC patients, a phenomenon not observed in nCRT-treated tumors.

Discussion

Conflicting findings surround the impact of metformin on survival outcomes in PDAC patients. This study introduces a post hoc analysis investigating the influence of metformin use, either pre-resection or during the follow-up period, on survival outcomes in resected PDAC patients enrolled in the PREOPANC RCT. Metformin use was associated with prolonged OS and PFS in the upfront surgery group, but this association was absent in the gemcitabine-based nCRT group. Motivated by established connections between immune responses and glycemic balances (32-34), on which metformin exerts its effect, we hypothesize that metformin influences tumor immunity, potentially contributing to the observed favorable survival outcomes. Transcriptomic immunoprofiling revealed an immunostimulatory pancreatic TME in metformin users, characterized by reduced infiltration of pro-tumoral CD163+ MRC1+ M2 macrophages and enhanced infiltration of immune-activating CD1A+ CD1C+ DCs. The potential of metformin to enhance antitumor immunity highlights its promise as an adjunctive therapeutic intervention for PDAC patients.

In PDAC patients with DM, retrospective studies present conflicting results on the therapeutic benefits of metformin (22,35,36). Chaiteerakij et al. (22) already emphasized that divergent findings may arise from methodological limitations. Additionally, variations in study design, such as exclusively focusing on PDAC patients with DM or those with advanced PDAC (23,36), could contribute to differing observations (19-21). This underscores the significance of careful patient selection and appropriate statistical methods in investigating metformin’s impact on survival. Unlike previous studies comparing PDAC patients with DM, our study evaluated the benefits of metformin use irrespective of DM status in resected patients. The distribution of patients with and without DM was equal between metformin user groups in the upfront surgery cohort, with 15 (71%) and 6 (19%) patients in the metformin-using group and 40 (66%) and 21 (34%) patients in the non-metformin group, with and without DM, respectively. In this cohort of upfront resected PDAC patients scheduled for adjuvant gemcitabine, metformin use emerged as a favorable prognosticator for OS and PFS. Intriguingly, this association was absent in patients who received gemcitabine-based nCRT. Conversely, a phase II trial investigating the benefits of adding metformin to gemcitabine and erlotinib for advanced PDAC reported no survival improvements when patients were assessed irrespective of DM status (23). However, this conflicting result may be explained by the notion that the survival benefits of metformin are evident primarily in patients with early-stage cancer (34).

Our transcriptomic immunoprofiling analysis revealed that metformin might modulate the immune landscape in the pancreatic TME, potentially inhibiting tumor progression. Metformin appears to shift the immunosuppressive TME toward an immune-active state with antitumoral properties. First, metformin users exhibited an increased abundance of CD1A+ CD1C+ DCs in the pancreatic TME. DCs are crucial in initiating immune responses against malignancies by presenting tumor antigens to T cells, subsequently activating CD4+ T helper cells or CD8+ cytotoxic T cells in PDAC (37). Second, the infiltration of CD14+ CD33+ monocytes and immunosuppressive CD163+ MRC1+ M2 macrophages decreased in the TME of metformin-treated patients. This observation aligns with the underexpression of genes related to the CCL2-CCR2 and CSF1-CSF1R axes, key immune pathways facilitating the recruitment and polarization of monocytes toward the M2 phenotype (38-41). Notably, increased peripheral levels of CCL2 correlated with reduced survival in PDAC patients (42). Furthermore, the underexpression of genes encoding IL-4, IL-13, and IL-34 may contribute to enhanced M2 macrophage infiltration, given their important role in promoting monocyte-to-M2 macrophage polarization (39-41). Finally, the underexpression of ANXA1, crucial for the acquisition of anti-inflammatory M2-phenotype (43), along with the underexpression of genes related to immune checkpoint inhibitory processes such as CD96, LAG3, and STAT3, further supports the idea that metformin may enhance immune activity in the pancreatic TME of metformin users (44,45). Our findings support previous research echoing diverse antitumor effects of metformin in different cancers. In head and neck squamous cell carcinoma, metformin enhances the activity of CD4+ and CD8+ T cells and natural killer cells (46). Moreover, in squamous cell carcinoma patients, metformin exerts an antitumor response by mitigating the abundance of tumor-promoting macrophages (47).

The lack of significant associations between metformin use and survival outcomes or distinct immune landscapes in the nCRT group may stem from potential interactions and masking effects of gemcitabine-based nCRT on metformin’s mechanisms of action. Neoadjuvant treatments such as chemotherapy, radiotherapy, or combination therapies could directly impact the TME, immune responses, and survival outcomes in PDAC (7), complicating the assessment of metformin’s independent effects. Furthermore, individual differences in tumor biology, immune profiles, and treatment responses add complexity to discerning consistent associations. Further research is needed to unravel the molecular, immunological, and prognostic interplay between metformin and gemcitabine-based nCRT in pancreatic cancer.

Although our study benefits from using samples from an RCT, bolstering the credibility of our findings through temporal clarity, consistent data collection, control over confounding variables, and long-term follow-up, we acknowledge several limitations. First, the distribution of patients with and without metformin was unbalanced, potentially impacting the robustness of our statistical analysis. Nevertheless, this distribution accurately reflects real-world prevalence. Additionally, due to limited tissue availability and the quality of tissue RNA or gene expression data, not all resected patients in our survival analyses (n = 148) underwent NanoString profiling (n = 96). Second, detailed clinical data regarding the durations or dosages of metformin were lacking. Patients were stratified based on their metformin usage at randomization, prompting an investigation into whether metformin influences tumor aggressiveness during its development or provides advantages as an adjuvant therapy. However, all patients received at least six months of standard metformin dosages between 500 and 2000 mg per day as recommended by their treating physicians. Moreover, some patients with DM who did not receive metformin used different hypoglycemic drugs, potentially influencing their immune profile or survival outcomes. Third, our immunoprofiling analyses were confined to transcriptomic data, and incorporating protein-based or spatial data could have aided in understanding metformin-induced immune alterations in specific TME compartments. Lastly, our study included patients receiving gemcitabine-based nCRT, but the optimal neoadjuvant approach for borderline resectable and resectable PDAC remains unclear. Exploring whether metformin induces differences in survival outcomes or immune profiles after alternative treatment approaches, as opposed to the observed lack of differences after gemcitabine-based nCRT, may provide further insights into its therapeutic benefit.

In conclusion, our study emphasizes metformin’s therapeutic potential in extending survival for upfront resected PDAC patients. Notably, this benefit appears diminished after gemcitabine-based nCRT. We propose that the improved survival may be linked to metformin’s capacity to enhance antitumor immunity. Specifically, metformin impeded the infiltration of immunosuppressive M2 macrophages while concurrently increasing the presence of immune-activating DCs within the pancreatic TME. Future investigations should explore metformin’s impact after various neoadjuvant treatments and its effects on the peripheral immune system.

Data availability

The datasets used and analyzed during the current study are available, with permission of the Erasmus Medical Centre Rotterdam, from the corresponding author on reasonable request.

Author contributions

Casper W.F. van Eijck, MSc (Conceptualization; Formal analysis; Investigation; Methodology; Resources; Software; Validation; Visualization; Writing – original draft; Writing – review & editing), Disha Vadgama, MSc (Conceptualization; Formal analysis; Investigation; Methodology; Resources; Software; Validation; Visualization; Writing – original draft; Writing – review & editing), Casper H.J. van Eijck, MD, PhD (Conceptualization; Funding acquisition; Methodology; Project administration; Supervision; Writing – original draft; Writing – review & editing), Johanna W. Wilmink, MD, PhD (Conceptualization; Funding acquisition; Methodology; Project administration; Supervision; Writing – original draft; Writing – review & editing).

Funding

This work was financially supported by the Survival With Pancreatic Cancer Foundation (www.supportcasper.nl) [grant number OVIT17-06].

Conflicts of interests

The authors declare that they have no competing interests.

Acknowledgements

The funder had no role in the design of the study; the collection, analysis, or interpretation of the data; or the writing of the manuscript and decision to submit it for publication.

Author notes

Casper W.F. van Eijck and Disha Vadgama contributed equally and share the first authorship.