Pegargiminase in leiomyosarcoma: ARGSARC ends a Phase 3 where the active arm gave up chemotherapy

What was tried

Polaris Group ran ARGSARC, a randomized, double-blind Phase 3 trial in advanced or metastatic leiomyosarcoma that had already progressed on an anthracycline. The active arm gave pegargiminase (ADI-PEG 20) at 36 mg/m2 weekly plus gemcitabine 600 mg/m2 and docetaxel 60 mg/m2. The control arm gave matched placebo plus gemcitabine 900 mg/m2 and docetaxel 75 mg/m2. The primary endpoint was progression-free survival, with overall survival and objective response as secondary endpoints. The trial enrolled 218 patients and was terminated for "Lack of Efficacy, no safety concern," with no results posted (ClinicalTrials.gov).

The protocol built in an asymmetry that frames everything below. The active arm traded a third of its gemcitabine dose and a fifth of its docetaxel dose to make room for an arginine-depleting enzyme. The control arm kept full-dose chemotherapy.

The biological hypothesis

Pegargiminase is a pegylated arginine deiminase, an enzyme that hydrolyzes circulating L-arginine to citrulline (ChEMBL CHEMBL3137346). Most normal cells recycle citrulline back to arginine using argininosuccinate synthase 1 (ASS1). Many carcinomas and sarcomas silence ASS1 epigenetically and lose that recycling step, which leaves them dependent on arginine drawn from the blood (Phillips 2013). Leiomyosarcoma is one of the tumors where ASS1 loss is common, so depleting serum arginine should starve the tumor while sparing ASS1-intact tissue. The dependency is real and the rationale is clean.

What the dependency is not is genetically validated in the human-genetics sense. In Open Targets the ASS1 association with sarcoma scores 0.098, and the genetic_association component of that score is 0 (Open Targets, ASS1). The evidence comes from expression and literature, not from inherited or somatic variants linking ASS1 loss to sarcoma outcomes. A target can be a true metabolic vulnerability and still carry no causal human-genetic anchor, which raises the bar on clinical proof.

What actually happened

The registry records ARGSARC as terminated for lack of efficacy with no safety concern (ClinicalTrials.gov). The endpoint numbers are not yet public, so the useful comparison is the same drug in a different tumor. In ATOMIC-Meso, pegargiminase added to first-line pemetrexed and platinum in nonepithelioid pleural mesothelioma improved median overall survival to 9.3 months from 7.7 months (hazard ratio 0.71, 95% CI 0.55 to 0.93, P = .02) and median progression-free survival to 6.2 months from 5.6 months (hazard ratio 0.65, 95% CI 0.46 to 0.90, P = .02), with grade 3 to 4 adverse events of 28.8% against 16.9%. The benefit was modest and it was real. The design difference is the point. In mesothelioma both arms received full-dose chemotherapy and pegargiminase was a true add-on. In leiomyosarcoma the active arm received less chemotherapy than the control.

Failure mechanism, best guess

Two facts collide. First, arginine starvation in ASS1-deficient cells is mostly cytostatic rather than cytotoxic. Panda 2025 showed that ADI-PEG 20 lowers CDK2 activity and halts the cell cycle, but BCL-XL holds BAX and BAK inactive so apoptosis never fires. Inhibiting BCL-XL restored killing in their models. Second, ARGSARC asked that cytostatic effect to offset a 33 percent cut in gemcitabine and a 20 percent cut in docetaxel. Gemcitabine and docetaxel kill dividing sarcoma cells. Cutting their dose to add an agent that mainly arrests the cycle, in a fast-growing tumor, is a trade that can lose progression-free survival even when the underlying biology is correct.

The likely failure is not that ASS1 dependence is fake. It is that the regimen subtracted cytotoxicity to add cytostasis. A second contributor is selection. ARGSARC enrolled leiomyosarcoma broadly rather than ASS1-low leiomyosarcoma. The disease is heterogeneous, with separate uterine and non-uterine biology and recurrent fusions in subsets (Momeni-Boroujeni 2024). Without an ASS1 cutoff, arginine-replete tumors dilute any signal.

How to prevent this next time

Three changes, two of them quantitative.

Keep the backbone fixed. Run the experimental agent as a true add-on at full chemotherapy dose in both arms, the design that produced a survival signal in ATOMIC-Meso. A dose-reduced control confounds mechanism with exposure and makes a null result uninterpretable.

Size against the right base rate and stop on a predictive-probability rule. Oncology Phase 3 programs reach approval roughly a third of the time, and the cumulative likelihood of approval from Phase 1 in oncology is about 3.4 percent (Wong, Siah, Lo 2019). A cytostatic add-on in an aggressive sarcoma deserves a low prior, and an interim predictive probability of success makes that prior explicit:

PP(success)=∫θ​P(final success∣θ,interim data)⋅π(θ∣interim data)dθ

When that probability drops below a preset floor at the interim, stop and save the patients. Enrich for the dependency as the third step. Power the trial in ASS1-low tumors defined by a validated immunohistochemistry cutoff, which lowers the number needed to screen by concentrating responders and shrinks the sample size needed to detect a real effect.

The single highest leverage change would have been to hold gemcitabine and docetaxel at full dose in both arms and test pegargiminase as a pure add-on, the way ATOMIC-Meso did.

What this means for similar programs

Arginine deprivation is not finished. The dependency is genuine and ATOMIC-Meso shows it can extend survival when layered onto an intact regimen. The lesson for the class is about combination architecture, not target choice. Cytostatic metabolic agents need cytotoxic or pro-apoptotic partners, and they need those partners delivered at full strength. The most promising direction is pairing arginine depletion with an agent that converts arrest into death, such as a BCL-XL inhibitor, rather than swapping out chemotherapy dose to fit the new drug in.

Open questions

What fraction of enrolled tumors were actually ASS1-low, and did response track with ASS1 status. Did the lower chemotherapy exposure in the active arm correlate with worse progression-free survival, which would confirm the design read of this failure. Would a BCL-XL or MCL1 co-target convert cytostasis to cytotoxicity in leiomyosarcoma the way it did in preclinical models. Do uterine and non-uterine leiomyosarcoma differ in arginine dependence enough to warrant separate trials.

Sources

1. ClinicalTrials.gov, ADI-PEG 20 or Placebo Plus Gem and Doc in Previously Treated Subjects With Leiomyosarcoma (ARGSARC), NCT05712694, 2026, U.S. National Library of Medicine, https://clinicaltrials.gov/study/NCT05712694.

2. Szlosarek PW, Creelan BC, Sarkodie T, et al., Pegargiminase Plus First-Line Chemotherapy in Patients With Nonepithelioid Pleural Mesothelioma (ATOMIC-Meso), 2024, JAMA Oncology,.

3. Panda PK, Paschoalini Mafra AC, Bastos ACS, et al., BCL-XL Protects ASS1-Deficient Cancers from Arginine Starvation-Induced Apoptosis, 2025, Clinical Cancer Research,.

4. Phillips MM, Sheaff MT, Szlosarek PW, Targeting Arginine-Dependent Cancers with Arginine-Degrading Enzymes, Opportunities and Challenges, 2013, Cancer Research and Treatment,.

5. Momeni-Boroujeni A, Mullaney K, DiNapoli SE, et al., Expanding the Spectrum of NR4A3 Fusion-Positive Gynecologic Leiomyosarcomas, 2024, Modern Pathology,.

6. Wong CH, Siah KW, Lo AW, Estimation of Clinical Trial Success Rates and Related Parameters, 2019, Biostatistics,.

7. Open Targets Platform, ASS1 (ENSG00000130707) Target-Disease Associations, 2026, https://platform.opentargets.org/target/ENSG00000130707.

8. ChEMBL, Pegargiminase (CHEMBL3137346), EMBL-EBI, https://www.ebi.ac.uk/chembl/explore/compound/CHEMBL3137346.