Pembrolizumab ± paricalcitol in metastatic pancreatic cancer postmaximal cytoreduction

Abstract

Lessons learned

Intravenous paricalcitol did not improve the efficacy of pembrolizumab, likely related to the short half-life.

Background

Immunotherapy has limited benefit in the treatment of advanced pancreatic cancer with the tumor microenvironment playing a key role in immune resistance. In preclinical studies, vitamin D receptor (VDR) agonists have been shown to sensitize pancreatic tumors to PD-1 blockade.

Methods

This was a randomized, double-blinded, placebo-controlled, phase II trial to evaluate pembrolizumab with or without paricalcitol as maintenance therapy for patients with metastatic pancreatic ductal adenocarcinoma (PDAC). Participants were ≥18 years; histologically or cytologically confirmed metastatic PDAC showing no disease progression after frontline systemic therapy, and achieving maximal cytoreduction (eg, with no further antitumor effect), Eastern Cooperative Oncology Group (ECOG) status of 0 or 1; adequate organ function. Study treatment included: pembrolizumab 200 mg IV every 3 weeks and either paricalcitol 25 mcg IV 3 times per week or placebo. The primary objective was to evaluate 6-month progression free survival (PFS). Secondary objectives include evaluating the toxicity of the combination and overall survival (OS).

Results

There was no significant difference in 6-month PFS, median PFS, median OS, nor treatment-related AEs between the 2 arms.

Conclusions and relevance

Paricalcitol did not improve the efficacy of pembrolizumab likely related to its short half-life of only 5-7 hours. Microbiome analysis revealed significant difference between long-term (>12 weeks) and short-term (<12 weeks) survival groups across treatment arms. Modulation of the tumor microenvironment will likely require more sustained VDR activity.

Trial Registration

Clinicaltrials.gov, ID: NCT03331562.

Introduction

Pancreatic ductal adenocarcinoma (PDAC), projected to be the second leading cause of cancer-related death by 2030, claims over 51 000 US lives annually, with about 90% succumbing to the disease within 5 years.¹ Standard treatments include FOLFIRINOX, NALIRIFOX, and gemcitabine with nab-paclitaxel, but cumulative toxicities limit long-term use, prompting maintenance strategies.

PDAC has faced numerous negative clinical trials with immunotherapy, but its use in a maintenance setting, where the microenvironment may be more favorable post-therapy, remains unexplored. Only in the rare 1%-2% of microsatellite instability (MSI) high pancreatic cancer patients is robust activity with immune checkpoint inhibitors seen.² PDAC is characterized by desmoplasia creating an immunosuppressive microenvironment with stromal fibrosis. Tumor-associated macrophages, myeloid-derived suppressor cells, and regulatory T cells inhibit the patient’s antitumor immune response creating an immune-privileged site.³

Vitamin D, a fat-soluble vitamin, obtained through diet and sunlight, converts to calcitriol, a potent steroid hormone that binds to the vitamin D receptor (VDR), impacting numerous signaling pathways. Vitamin D has been shown to impact proliferation, cellular differentiation, and epithelial-mesenchymal transition. Low levels of vitamin D are linked to a proinflammatory state through lipopolysaccharide-induced production of IL-6 and MKP-1 induction of TNF alpha.^4,5

Activation of pancreatic stellate cells during tumorigenesis leads to a desmoplastic reaction with inflammation and fibrosis promoting pancreatic cancer progression and impairing activity of chemotherapy and the immune system.⁶ The VDR, expressed in the pancreatic tumor stroma, acts as a master transcriptional regulator of pancreatic stellate cells. Activation of the receptor reverts the pancreatic stellate cells to a quiescent state, thus modulating the tumor microenvironment and decreasing fibrosis.⁷ With the normalization of the intra-tumoral vasculature, perfusion is improved, allowing better delivery of therapeutics. The VDR is also expressed in immune cells such as B cells, T cells, and antigen-presenting cells. Vitamin D deficiency can also modulate the innate and adaptive immune response. Preclinical studies suggest that VDR activation leads to the modulation of tumor microenvironment, decreasing fibrosis, and sensitizing pancreatic cancer cells to chemotherapy and the immune system.⁸ We enrolled patients responding to cytotoxic chemotherapy allowing time for potential tumor microenvironment modulation with paricalcitol.

Methods

Study design and patients

This double-blind, placebo-controlled, randomized phase II trial was approved by institutional review boards and conducted in accordance with the protocol (Supplement 1), the principles of CFR, ICH GCP, and the Declaration of Helsinki. Patients provided written informed consent prior to participation.

The primary objective was to evaluate 6-month progression free survival (PFS). Secondary objectives included evaluating overall survival (OS) and comparing toxicities between the arms. Exploratory objectives included assessing improvement in patient-reported disease-related symptoms, tumor texture changes, gut microbiota, and correlating with treatment response.

Eligible patients were 18 years or older with histologically or cytologically confirmed metastatic PDAC, showing no disease progression after frontline systemic therapy, and achieving maximal cytoreduction (eg, with no further antitumor effect), Eastern Cooperative Oncology Group (ECOG) status of 0 or 1, and adequate organ function. Prior adjuvant/neoadjuvant chemotherapy was permitted. Exclusion criteria included prior anti-PD-1, anti-PD-L1, or anti-PD-L2 therapy, and serum vitamin D level≥50 ng/mL. Refer to Supplement 1-Protocol for complete eligibility criteria and study procedures.

Procedures

Patients received intravenous (IV) pembrolizumab 200 mg every 3 weeks and IV paricalcitol 25 mcg (or placebo saline equivalent volume) 3 times weekly during each 21-day cycle prior to the pembrolizumab infusion. Adverse events (AEs) were graded per National Cancer Institute Common Terminology Criteria for AEs (NCI-CTC) version 4.0. Complete blood count and serum chemistries were collected weekly, parathyroid hormone (PTH) every 3 cycles, and CA 19-9 (or CEA or CA -125 for non-expressors of CA 19-9) at baseline and each cycle. Tumor response was evaluated every 9 weeks per RECIST1.1. Treatment continued until disease progression, unacceptable toxicity, or patient/investigator decision, with survival follow-up.

The trial collected tissue, blood, and fecal samples. Blood was taken at baseline, every 9 ± 1 weeks, and end of treatment (EOT) for serum and for peripheral blood mononuclear cells (PBMCs) analysis. Archival tumor samples were obtained when available, and patients could opt to provide fresh tumor tissue for analysis of mutations, transcriptional programs, and cellular VDR targets. Fecal samples were collected before and after pembrolizumab dosing for microbiome analysis. A patient questionnaire assessed self-reported disease symptoms throughout treatment (Supplement 2).

Radiomics analysis

Contrast-enhanced (CE-CT) images of the chest, abdomen, and pelvis obtained for RECIST1.1 review, were used to assess radiomic features in each treatment arm using baseline (BL) and first post-baseline (pBL) CE-CT as described previously.⁹ Radiomic feature values were then used to create radiomic signatures using mathematical modeling to reduce overfitting and improve the model’s generalizability. Once a feature set was identified, a logistic regression model was applied. The feature sets were then tested using statistical metrics for inter-model comparisons including: (1) receiver operating characteristic—area under the curve (ROC-AUC); (2) Bayesian Information Criterion; (3) McFadden’s pseudo R-squared; (4) F-test P-value and accuracy (Supplementary Table S1). For each top performing model, SHAPs (SHapley Additive exPlanations) analysis was conducted to obtain additional information on model performance. Radiomic models with a P-value < .05 were considered of further exploratory interest.

Statistical analysis

The full statistical analyses are described in Supplement 3. Survival and safety analyses included all patients receiving at least one dose of study treatment. Patients unevaluable for progression at 6-months due to a non-progression-related event were excluded. PFS and OS were calculated for each arm using Kaplan-Meier methods. Differences between treatment arms were assessed with a log-rank test, and hazard ratios were estimated using a Cox proportional hazards model. The proportion of patients alive at 6 months was compared between arms using Pearson’s chi-squared test. Sensitivity analysis for PFS included patients who died from causes other than progressive disease, with Fine and Gray¹⁰ methods used for competing events. Changes in PTH and calcium levels were analyzed with an interaction term between time and treatment in a linear mixed-effects model.

Results

Patient characteristics

Between May 17, 2018 and November 14, 2019, 27 patients with stage IV PDAC were randomized across 6 sites in US. Of those randomized, 24 were treated and included in the efficacy and safety analyses (Figure 1). Three patients were deemed ineligible after randomization and did not initiate study treatment due to (1) ongoing biliary tract infection, (2) disease progression, and (3) elevated bilirubin and liver function tests. The median age was 67.6 years, 54.2% (13/24) were female, and most had pancreatic head tumors. Seventeen patients (70.8%) had liver metastases (Table 1). Twenty-one patients (87.5%) discontinued study treatment due to disease progression, with 3 (12.5%) due to AEs. The study treatment concluded in June 2020 with the last survival contact completed in March 2023.

The primary endpoint of PFS at 6 months was assessed by both the Investigator (Figure 2a) and an independent central review (Figure 2b). There was no significant difference in PFS at 6 months between pembro + placebo (16.7%) and pembro + paricalcitol (0.0%) by Investigator assessment (P = .20) nor between pembro + placebo (12.5%) and pembro + paricalcitol (0.0%) by independent assessment (P = .37). Median PFS was also similar: pembro + placebo (2.2 months) vs pembro + paricalcitol (2.0 months) by Investigator t (P = .87) and pembro + placebo (4.0 months) vs pembro + paricalcitol (3.1 months) by independent assessment (P = 0.91). Median OS demonstrated no significant difference between the pembro + placebo (10.2 months) and pembro + paricalcitol (10.4 months) arms (Figure 2c).

Calcium significantly increased with paricalcitol vs placebo (P < .001) while PTH levels did not significantly differ (P = .332.)

Safety and adverse events

The most common AEs in both arms were abdominal pain, ALT increase, AST increase, anemia, and nausea (Table 2). Of 38 treatment-related AEs (TRAE) reported in 14 patients (58.3%); 19 were related to pembrolizumab alone, 9 to paricalcitol/placebo, and 10 to both. Seven AEs (18.45%) were grade 3, of which most (6) occurred in 2 patients on pembro + paricalcitol. No grade 4 or 5 treatment-related AEs were reported and there was no significant difference in TRAEs between arms.

Disease related symptoms

The most common bothersome symptoms were fatigue (50%) and neuropathy (45%) (Supplementary Figure S1). Nineteen patients (79%) reported at least one of these 2 concerns. All 24 patients had improvement in their “most bothersome disease-related symptoms” during treatment. Overall, 63% improved, 29% worsened, and 8% saw no change (Supplementary Figure S2).

Radiomic analysis

Radiomic analysis of CE-CT images was conducted. Two subjects (one in each arm) lacked suitable post-baseline CT scans. Among those with suitable BL and pBL scans, significant differences in radiomic tumor texture patterns between the 2 arms were observed on the first-pBL scan (Supplementary Figure S3). While baseline radiomic textures of the largest RECIST lesion were similar between arms, significant differences emerged in post-baseline scans.

Microbiome changes

We analyzed fecal microbiome composition at 3 separate time points: before treatment (baseline), after treatment, and 30 days after the EOT (EOT30). Significant variability in the fecal microbiome was noted across individual patients, treatment regimens, and survival outcomes. No significant changes in microbiome composition were observed between time points within treatment arms. However, significant differences were seen between survival outcomes. The β-diversity analysis (Supplementary Figure S4) revealed distinct dissimilarities in microbiomes across the arms with different survival lengths, pembro + placebo (P-value .001) and pembro + paricalcitol (P-value .003). The α-diversity analysis (Supplementary Figure S5) revealed that in the pembro + placebo arm, patients with shorter survival (<12 weeks) had higher Shannon index, microbial richness, and evenness (P-value .018). Conversely, in the pembro + paricalcitol arm, longer surviving patients (>12 weeks) exhibited a higher Shannon index (P-value .015). These results suggest that the combination of pembro + paricalcitol has a significant influence on the microbiome composition, associated with the survival length.

Analysis of bacterial genus abundance via ANCOM-BC (analysis of composition of microbiome with bias correction) highlighted discernible differences in the bacterial genera between survival groups (Figure 3). However, when separating baseline (pre-treatment) samples for analysis, no statistically significant differences were found in microbiome composition between patients who survived <12 weeks (n = 12) and those who survived >12 weeks (n = 4).

Discussion

Overall, this was a small well-conducted study, with a balanced distribution between the 2 treatment arms and a median age representative of the real-world patient population, despite yielding a negative result. While the addition of paricalcitol did not impact PFS or OS, the study uncovered noteworthy findings, particularly radiomic changes observed through CE-CT. Radiomics detects changes in tumor/diseased tissues not readily identified by visual inspection.¹¹ Radiomics relies on pixel values extracted from CT/MRI/PET scans and assembles them by pixel intensity (1st order), pixel spatial distribution (2nd order), and shape morphology. Each grouping of pixels can be mathematically modeled to correlate with pharmacodynamic effects, outcomes, and treatment response. Given that paricalcitol can have a profound effect on the tumor microenvironment, any changes in radiomic signaling might be related to this effect. Our preliminary data supports this hypothesis. In particular, lesion radiomics on pBL scans appears to show significant differences between treatment arms. Interestingly, the type of radiomic features most predictive of this difference is related to 1st order CT-CE features; namely, lesions that have higher pixel intensity can be attributable to the paricalcitol-treated subjects. Although it is intriguing to postulate that paricalcitol can alter tumor perfusion due to reductions in tumor fibrosis, such conclusions should be considered cautiously due to the limited number of subjects undergoing radiomic analysis. Despite feature reduction steps to reduce false positive results, overfitting of the data cannot be excluded entirely.

Additionally, microbiome analyses revealed significant differences between long and short-term survival groups (Figure 3), suggesting the gut microbiota as a modulator of therapeutic efficacy in these treatments. Notably, increases in Blautia and Alistipes in long-term survivors, regardless of treatment, suggest these genera may enhance survival with pembrolizumab. However, Akkermansia shows a contrasting pattern: its increase in Akkermansia in long-term survivors in pembro + placebo could be linked to its known beneficial effects on the mucosal barrier and immune modulation,¹² while its decrease in pembro + paricalcitol group suggests paricalcitol may alter microbial dynamics or host responses in a way that diminishes Akkermansia’s positive impact or reflects different metabolic requirements. These data suggest that microbiome composition could inform personalized PDAC treatment. The reliability of these findings is limited by small, uneven sample sizes, especially when comparing baseline microbiomes. Conclusions are preliminary and should be approached with caution. Future studies require larger, more balanced sample sizes, and stratified sampling for better representation. Exploring the microbiome’s role in PDAC survival offers potential for therapeutic interventions and prognostic tools.

Paricalcitol has a short half-life of 5-7 hours in healthy subjects¹³ and in the current study was administered 3 times weekly. Although calcium levels increased significantly with paricalcitol, PTH levels remained unchanged, likely due to the short half-life of the paricalcitol resulting in minimal impact on the tumor microenvironment.

Limitations

Enrolling patients with optimal chemotherapy response was challenging for assessing tumor microenvironment changes during paricalcitol treatment. Limited tumor tissue and funding constraints prevented the sequencing of tumor tissue and PBMC.

Conclusions

VDR is key in modulating the pancreatic tumor microenvironment; however, the VDR agonist, paricalcitol, did not improve the clinical activity of pembrolizumab in a maintenance setting in patients with metastatic PDAC, likely due to the short half-life of the paricalcitol formulation utilized in this trial. Tumor microenvironment changes were noted with CE-CT in patients treated with pembro + paricalcitol vs pembro + placebo. More sustained VDR activation may enhance clinical activity; though daily IV dosing of paricalcitol is impractical. Oral daily dosing is being explored in another combination study for PDAC (NCT03520790). Our findings will inform future research on vitamin D as a cancer treatment. Immunotherapy in the maintenance setting should only be utilized in a clinical trial.

Acknowledgments

The authors thank the patients and their families for participating in this study and the investigators and site personnel who contributed to this important trial. The authors would like to acknowledge the very important contribution of the microbiome analysis completed by a valued colleague, Sarah Highlander, who unfortunately passed away before this study was published. The authors also wish to acknowledge the important contributions of the trial medical monitor, Dr. Tomislav Dragovich, and medical writing and editorial assistance provided by Nina Cantafio and Elissa Heller (all were compensated for their contributions).

Author contributions

Vincent Chung: Provision of study material or patients, collection and/or assembly of data, manuscript writing, final approval of manuscript. Angela Alistar: Provision of study material or patients, collection and/or assembly of data, manuscript writing, final approval of manuscript. Carlos Becerra: Provision of study material or patients, collection and/or assembly of data, manuscript writing, final approval of manuscript. Anup Kasi: Provision of study material or patients, collection and/or assembly of data, manuscript writing, final approval of manuscript. Erkut Borazanci: Provision of study material or patients, collection and/or assembly of data, manuscript writing, final approval of manuscript. Gayle Jameson: Provision of study material or patients, collection and/or assembly of data, manuscript writing, final approval of manuscript. Denise Roe: Formal Analysis; Conception and design, data analysis and interpretation, manuscript writing, final approval of manuscript. Betsy Wertheim: Data analysis and interpretation, manuscript writing, final approval of manuscript. Derek Cridebring: Collection and/or assembly of data, data analysis and interpretation, manuscript writing, final approval of manuscript. Morgan Truitt: Collection and/or assembly of data, data analysis and interpretation, manuscript writing, final approval of manuscript. Michael Downes: Collection and/or assembly of data, data analysis and interpretation, manuscript writing, final approval of manuscript. Michael Barrett: Collection and/or assembly of data, data analysis and interpretation, manuscript writing, final approval of manuscript. Ron Korn: Collection and/or assembly of data, data analysis and interpretation, manuscript writing, final approval of manuscript. Keehoon Lee: Collection and/or assembly of data, data analysis and interpretation, manuscript writing, final approval of manuscript. Haiyong Han: Collection and/or assembly of data, data analysis and interpretation, manuscript writing, final approval of manuscript. Ronald Evans: Collection and/or assembly of data, data analysis and interpretation, manuscript writing, final approval of manuscript. Daniel Von Hoff: Conception and design, manuscript writing, final approval of manuscript.

Funding

Funded by a Stand Up To Cancer (SU2C) AACR Grant No. SU2C-AACR-CT03-16. Merck Sharp & Dohme LLC, a subsidiary of Merck & Co., Inc., Rahway, NJ, USA provided drug (pembrolizumab) for the study.

Conflicts of interest

Vincent Chung reports having received research funding from Merck for Investigator Sponsored Trials (IST). Angela Alistar reports being a speaker for Merck and reviewing ad boards for Merck. Carlos Becerra reports consulting for Ipsen, Bristol Myers Squibb, and DaZen. Anup Kasi reports consulting or advisory role for Ipsen and Cardinal Health; and reports research funding (institutional) from TESARO, Astellas Pharma, Rafael Pharmaceuticals, Geistlich Pharma, Cardiff Oncology, FibroGen, Bavarian Nordic, Novocure, Cend Therapeutics, and Ability Pharma. Erkut Borazanci reports consulting for BPG, Qurient, Taiho, Clearnote, Conjupro, Elevation Oncology, Nanology, and VCN. Gayle Jameson reports having received research funding from Bristol Myers Squibb. Michael Barrett reports grants from National Cancer Institute on pancreas: UH3 CA238926-02-2: HRD-IA signatures in pancreatic ductal adenocarcinoma and R01 CA265050.-1: New Therapeutics for Pancreatic Cancer; and reports being the CEO of a National Cancer Institute funded start-up Binary Genomics. Ron Korn reports being employed by Imaging Endpoints Core Lab; and reports being an adjunct faculty at TGen/City of Hope; and reports being a consultant for HonorHealth Research Institute; and reports being a shareholder for Globavir, Renibus, and Verve; and reports patents for PCT/US2017/047026, US10854338B2, US10332634B2. Haiyong Han is an SAB member of Stromatis Pharma and a consultant for ImproveBio, Inc. Daniel Von Hoff reported Stock and Other Ownership Interests: Medtronic, CerRx, SynDevRx, UnitedHealthcare, Anthem Inc, Stromatis Pharma, Systems Oncology, Stingray Therapeutics, Orpheus Bioscience, AADi, Origin Commercial Advisors, Halia Therapeutics, Lycia Therapeutics, (3+2) Pharma. Consulting or Advisory Role: Imaging Endpoints, Senhwa Biosciences, Alpha Cancer Technologies, CanBas, Lixte Biotechnology, RenovoRx, TD2, Phosplatin Therapeutics, SOTIO, Immunophotonics, Genzada Pharmaceuticals, L.E.A.F. Pharmaceuticals, Oncology Venture, Verily, Atheneex, Novita Pharmaceuticals, Vicus Therapeutics, Agenus, Samumed, BioXCel Therapeutics, Sirnaomics, AiMed, Erimos Pharmaceuticals, Pfizer, ImmuneOncia, Viracta Therapeutics, AlaMab Therapeutics, NeoTx, Xerient, Noxxon Pharma, Lycia Therapeutics, EXACT Therapeutics, ImaginAb, SignaBlok, Compass Therapeutics, OnQuality Pharmaceuticals, Sellas Life Sciences, Catamaran Bio, Thirona Biosciences, Bristol Myers Squibb, Remix Therapeutics, SMP Oncology fka SDP/Tolero, Bessor Pharma, Coordination Pharmaceuticals, Orphagen Pharmaceuticals, Red Arrow Therapeutics, Soley Therapeutics, Invios GmbH, Mekanistic Therapeutics, POINT Biopharma, Peptomyc, Remunity, SIWA Therapeutics, Xenthera, Indaptus fka Decoy, Panavance Therapeutics fka Geistlich, CyMon Bio, Bryologyx, Moleculin Biotech, EnGeneIC, Race Oncology. Research Funding: Lilly, Genentech, Celgene, Incyte, Merrimack, Plexxikon, Minneamrita Therapeutics, Abbvie, Aduro Biotech, Cleave Biosciences, CytRx Corporation, Daiichi Sankyo, Deciphera, Endocyte, Exelixis, Five Prime Therapeutics, Gilead Sciences, Merck, Pfizer, Pharmacyclics, Phoenix Biotech, Samumed, Strategia, Halozyme. Patents, Royalties, Other Intellectual Property: Intramedullary Catheter, Methods of Human Prostate Cancer, Use of 5,6-Dihydro-5-Azacytidine in the Treatment of Prostate Cancer, Targeting Site-2 Protease (S2P) for the Treatment of Pancreatic Cancer (pending), Targeting Ecto-5-Nucleotidase (Cd73) for the Treatment of Pancreatic Cancer, Targeting a Protein Tyrosine Phosphotase-PRL-1 for the Treatment of Pancreatic Cancer (pending), Targeting a Protein PRC1 for the Treatment of Pancreatic Cancer (pending), Targeting Ecto-5-Nucleotidase (CD73) for the Treatment of Pancreatic Cancer (pending), Protein Kinase Inhibitors (pending), Methods, Compounds and Compositions with Genotype Selective Anticancer Activity (pending), Methods and Kits to Predict Therapeutic Outcome of BTK Inhibitors (pending), Muscle Fatigue Substance Cytokines and Methods of Inhibiting Tumor Growth Therewith (pending), 2-aryl-pyridylazoles for the Treatment of Solid Tumors such as Pancreatic Cancer (pending). The other authors declare no conflicts of interest.

Data availability

The study level data that support the findings of this study have been posted on clinicaltrials.gov. The individual participant data (IPD) may be available upon request from the corresponding author. Data availability is subject to institutional approval.

References