Calciphylaxis complicated by NDM-1-producing Pseudomonas aeruginosa: first experience with cefepime–zidebactam alongside phage therapy in the USA

Calciphylaxis: a rare vascular complication

Calciphylaxis is a rare disorder characterised by calcification of the arteriolar microvasculature in the dermis and subcutaneous adipose tissue. Vessel occlusion and tissue ischemia result in severe pain that is nearly universal and often out of proportion to the visible lesions.^1-3 The disease carries high morbidity due to progressive tissue necrosis, risk of secondary infection, and few treatment options.^4,5 Also called calcific uremic arteriolopathy, calciphylaxis typically occurs in patients with end-stage renal disease on dialysis (incidence ~3·5–4·5 per 1000 patient–years).^18,19 But this condition can also arise in individuals with pre-dialysis chronic kidney disease, acute kidney injury, or normal kidney function.^7-9,20 In paediatric and young adult populations calciphylaxis is particularly rare, which can contribute to diagnostic delays.^21-26

Our patient had several established risk factors for calciphylaxis, including female sex, obesity, history of dialysis, and chronic inflammation associated with her underlying malignancy, and prolonged hospitalisation.^1,3,18 Other recognised risk factors include hyperphosphataemia, warfarin use, diabetes, vitamin D supplementation, systemic glucocorticoids, and hypercoagulable states.^1,3,18 The pathogenesis remains incompletely understood but involves microvascular calcification driven by osteogenic signalling, deficiencies in calcification inhibitors (eg, fetuin-A and matrix Gla protein), and chronic inflammation.^3,27-30 These processes lead to arteriolar narrowing from calcium deposition, thrombosis, and subsequent tissue ischaemia and necrosis.

Clinically, calciphylaxis typically presents with extremely painful, violaceous skin lesions that progress to necrotic ulcers with black eschars.³¹ Lesions most commonly affect areas of increased adiposity, including the thighs, abdomen, and buttocks, consistent with our patient’s presentation.³ The diagnosis is primarily clinical in patients with end-stage renal disease presenting with characteristic painful, ulcerated lesions, although skin biopsy might be necessary in atypical cases, as with our patient.³ There are no specific laboratory markers for calciphylaxis. Some patients can have elevated parathyroid hormone, phosphorus, or calcium, but these findings are not consistently observed.³² Our patient, for example, only had hypercalcaemia.

There are no high-quality studies to guide optimal treatment. Management focuses on mitigating risk factors, wound care, and pain control. However, growing evidence suggests that off-label intravenous sodium thiosulphate is associated with improvement or complete resolution of skin lesions in several studies.^33-35 Although not fully understood, postulated mechanisms include vasodilatory and antioxidant properties, as well as inhibition of adipocyte-induced vascular smooth muscle calcification.^36,37 Most experience with sodium thiosulfate involves patients on dialysis who receive the medication three to four times per week.³ Our patient, who had recovered kidney function, received daily intravenous sodium thiosulfate, an approach with little published experience. This dosing regimen necessitated close monitoring for adverse effects, including hypotension, metabolic acidosis, electrolyte abnormalities, and QTc-interval prolongation.³ The appropriate duration of therapy also remains unknown. Experts generally recommend a 3-month treatment course and earlier discontinuation if no clinical response is observed within 4 weeks.³⁸ Given our patient’s extensive wounds and favourable response and tolerability, sodium thiosulfate therapy was continued for 26 weeks (6·5 months).

Vitamin K supplementation can further support vascular health by promoting carboxylation of matrix Gla protein, a key inhibitor of vascular calcification.³⁹ Patients with calciphylaxis have lower fractions of carboxylated matrix Gla protein, suggesting impaired vitamin K-dependent mechanisms.²⁸ This mechanism can explain why warfarin, a vitamin K antagonist, is a recognised risk factor for this condition.^39-42 Our patient received concurrent vitamin K therapy three times weekly alongside sodium thiosulfate, recognising the potential synergistic benefit of restoring calcification inhibitor function. She also received bisphosphonate therapy (intravenous pamidronate followed by oral alendronate) for persistent hypercalcaemia in the setting of immobilisation and calciphylaxis, as part of a comprehensive treatment approach. This strategy is occasionally employed in treatment-resistant calciphylaxis complicated by hypercalcaemia. Although evidence for bisphosphonates remain inconclusive—with meta-analyses showing no clear mortality benefit—observational studies suggest a potential reduction in amputation risk.⁴³

Wound care and pain management are also essential components of treatment. Surgical debridement and negative-pressure wound therapy should be considered, particularly for infected or necrotic lesions, to promote healing.³⁸ Simultaneously, managing the severe pain associated with calciphylaxis often requires a multimodal regimen of analgesics targeting different mechanisms of action.³⁸ Our patient, for example, required fentanyl, hydromorphone, ketamine, and methadone during the acute phase of her illness. But as treatment progressed and tissue perfusion improved, her pain medications were successfully weaned (appendix p 4), reflecting the interdependence of source control, vascular healing, and symptom resolution.

Therapeutic challenges and novel strategies for NDM-producing P aeruginosa

Whereas vascular pathology was the primary driver of lesion progression, local infection with pan-resistant P aeruginosa further complicated management, necessitating novel therapeutic approaches. NDM-producing P aeruginosa represents a major and emerging threat due to its extensive drug resistance and few therapeutic options.¹² This pathogen has multiple resistance mechanisms, including reduced membrane permeability, multidrug efflux pumps, and production of β lactamases across all four Ambler Classes (A–D).¹² Class B metallo-β-lactamases, which use zinc ions to hydrolyse virtually all β-lactams except monobactams, are particularly concerning. The three most clinically relevant metallo-β-lactamase families—imipenemase, Verona integron-encoded metallo-β-lactamase, and NDM—have spread globally, although they remain relatively uncommon among P aeruginosa in the USA compared with other regions.^12,45

Most US cases of NDM-producing P aeruginosa have been linked to international travel, but our patient harboured one of the first domestically acquired isolates.¹⁴ Sequencing analysis traced her infection to the tertiary adult hospital where she received ECMO.^13,14 Among the eight patients in this nosocomial cluster, she was the only one to develop cefiderocol resistance.¹⁴ The emergence of resistance was likely influenced by her exposure to intermittent cefiderocol totalling about 3·5 months between July and December, 2023, for recurrent fevers, infectious complications, and known colonisation with multidrug-resistant organisms (appendix p 2). This resistance prompted initiation of ceftazidime–avibactam plus aztreonam to leverage aztreonam’s stability against metallo-β-lactamases—although reported synergy against metallo-β-lactamase-producing P aeruginosa is low (<10%).^46,47 It was only later that checkerboard testing confirmed a fractional inhibitory concentration index (FICI) of 2, indicating no synergy (FICI >0·5 to ≤4).¹⁷

With conventional options exhausted, our attention shifted to novel agents, including β-lactam enhancers, to address the patient’s evolving resistance in the context of calciphylaxis. Zidebactam, the first clinically available β-lactam enhancer, both inhibits some β lactamases and shows potent binding affinity for penicillin-binding protein 2 (PBP2) in Enterobacterales, P aeruginosa, and Acinetobacter species.^48,49 The drug also inhibits Ambler Class A (including Klebsiella pneumoniae carbapenemases) and Class C β lactamases. Although not stable in vitro against Class B metallo-β-lactamases and Class D oxacillinases, its combination with cefepime demonstrates substantial activity against metallo-β-lactamase-producing P aeruginosa through its β-lactam enhancer mechanism.^50,51

The enhancer effect arises from complementary PBP targeting: cefepime inhibits PBP1 and PBP3, whereas zidebactam engages PBP2, achieving at least 3 log₁₀ bacterial killing.⁵⁰ Animal models show that this β-lactam enhancer activity lowers the cefepime pharmacodynamic target or fraction of time above the MIC from more than 60–100% to just 9–16% for an at least 1 log₁₀ kill effect.^52,53 Notably, this occurs even when both drugs individually have high MICs, as the mechanism is independent of standard susceptibility cutoffs. Complementary PBP targeting also disrupts cell wall synthesis, leading to morphological changes—including spheroplast formation—that compromise bacterial membranes and accelerate cell death.^50,51

These promising mechanistic and preclinical findings have spurred the clinical development of cefepime–zidebactam (Wockhardt 5222), which has entered multinational trials for complicated urinary tract infections in Europe and Asia.⁵⁴ In phase 1 studies, cefepime–zidebactam demonstrated linear pharmacokinetics without accumulation and about 90% renal excretion as unchanged drug.⁵⁵ The safety profile was also generally favourable, with mild adverse events, such as headache, diarrhoea, and infusion reactions. Under an FDA-approved Expanded Access application, our patient received cefepime–zidebactam (cefepime 2 g plus zidebactam 1 g) as a 60 min intravenous infusion every 8 h, based on her creatinine clearance (60–120 mL/min). At the start of therapy, the cefepime–zidebactam MIC was 64 μg/mL, which is at the upper limit but still within the Clinical and Laboratory Standards Institute investigational susceptible breakpoint (≤64 μg/mL). This relatively elevated MIC might have been influenced by the patient’s preceding 3-month exposure to ceftazidime–avibactam (appendix p 2). Similar to zidebactam, avibactam is a diazabicyclooctane β-lactamase inhibitor that targets PBP2.^48,49 Selective pressure from avibactam might have driven PBP2 alterations, contributing to the initial cefepime–zidebactam MICs observed. With subsequent exposure to cefepime–zidebactam over 6 weeks, resistance ultimately emerged (MICs >512 μg/mL), exemplifying this pathogen’s rapid capacity for adaptation. Whole-genome sequencing and analysis of resistance-associated mutations will be the focus of our future work.

Given both the emergence of resistance and constraints in drug supply, our salvage regimen was adjusted to combination therapy with cefiderocol and sulbactam–durlobactam for the remaining 2 weeks. Cefiderocol is a siderophore cephalosporin that exploits bacterial iron transport to cross the outer membrane, exerting its bactericidal activity primarily through inhibition of PBP3. This drug remains stable against most β lactamases, including metallo-β-lactamases.⁵⁶ That said, resistance can emerge through mutations in iron uptake systems or PBP3, and co-production of metallo-β-lactamases and serine β lactamases.^56-58 To address these additional resistance mechanisms, sulbactam–durlobactam was added. This agent combines a narrow-spectrum β-lactam β-lactamase inhibitor with no Pseudomonas activity (sulbactam) and a novel diazabicyclooctane β-lactamase inhibitor (durlobactam) that potently inhibits Class A, C, and D enzymes.^59,60 When used together, in-vitro checkerboard testing showed MIC reductions for cefiderocol (from 16 to 4 μg/mL) and sulbactam–durlobactam (from >16 to 0·50 μg/mL) against our patient’s isolate, yielding an FICI of 0·266. Although such in-vitro synergy (FICI ≤0·50)¹⁷ might not reliably predict clinical efficacy, similar findings have been reported with durlobactam restoring in-vitro imipenem susceptibility in P aeruginosa, alongside positive clinical outcomes.^60,61 This synergy likely reflects complementary targeting of distinct resistance mechanisms, with cefiderocol bypassing some metallo-β-lactamase-mediated resistance and sulbactam–durlobactam inhibiting co-existing non-metallo-β-lactamases.⁶² Bacterial killing can be further enhanced through durlobactam’s PBP2 inhibitory activity.⁶⁰ This effect can be particularly relevant when PBP3 function is compromised. Such a scenario has been observed in cefiderocol-resistant NDM-producing Enterobacterales, where isolates with PBP3 mutations are resensitised by durlobactam.⁵⁸ However, this phenomenon has not yet been reported in P aeruginosa. Whether such PBP2-mediated killing additionally involves β-lactam enhancer mechanisms—beyond complementary multi-PBP targeting—also remains to be determined.

Considering our patient’s underlying immunosuppressive state and extensive lesions, we opted for an 8-week antibiotic course. The optimal duration of antimicrobial therapy for infected calciphylaxis ulcers is not established. Variation in wound healing, host factors, and microbiological findings often necessitates individualised, prolonged courses. In our patient, this extended duration was chosen to support tissue repair while allowing concurrent modified oncological therapy to resume safely (appendix p 1).

Bacteriophage therapy for pan-resistant infections

Given the pan-resistant profile of our patient’s isolate, bacteriophage therapy emerged as a promising therapeutic option. Phages are naturally occurring viruses that specifically target and lyse bacterial hosts, offering a mechanism distinct from traditional antibiotics.⁶³ Phage therapy was first described over a century ago, however, its clinical application was largely abandoned in Western societies with the advent of antibiotics in the 1940s. However, the rise of extensively drug-resistant pathogens has renewed interest in this therapeutic modality.^64,65

Growing clinical experience with bacteriophage therapy has demonstrated encouraging safety profiles.⁶⁶ For instance, in a case series of 100 patients receiving personalised phage therapy, 15 adverse events were reported, including seven non-serious drug reactions suspected to be treatment-related.⁶⁷ Another analysis of 63 patients identified phage-related adverse events in only nine cases, most of which were mild and transient.⁶⁸ The most significant reaction involved fever and dyspnoea occurring 2 h after infusion of 1 × 10¹¹ PFU/mL, which resolved with supportive care and did not recur when the same phage was subsequently administered at a lower dose (1 × 10¹⁰ PFU/mL).⁶⁹

Optimal phage dosing remains undefined, as clinical outcomes depend on multiple variables including phagespecific characteristics (ie, burst size, adsorption rate, and stability), bacterial inoculum, multiplicity of infection or phage-to-bacteria ratios, and pre-therapeutic bacterial genetic diversity—which is particularly pronounced in Pseudomonas.^68,70 As a result, clinical dosing regimens have varied widely, ranging from 1 × 10³ to 1 × 10¹¹ PFU per dose, with treatment durations extending from a single dose to 16 weeks of daily therapy.⁶⁸ Pharmacokinetic considerations are also important as intravenous phages undergo rapid clearance through mononuclear phagocyte uptake and neutralising antibody formation.⁶⁸ Direct administration at infection sites can achieve higher local concentrations and has been successfully used for infections of the skin and soft tissue, prosthetic joints, ventricular assist devices, and the respiratory tract.⁶⁸

For our patient, bacteriophage therapy was coordinated with the Van Tyne Laboratory. The overall process, from initial laboratory contact to treatment initiation, took about 4 weeks. Within roughly 1 week, susceptibility testing identified phage KEN3, a podovirus in the genus Bruynoghevirus.⁷¹ The phage was subsequently manufactured at 1 × 10⁹ PFU per dose, with endotoxin levels of 80 endotoxin units per dose and undetectable Pseudomonas exotoxin A (<0·64 ng per dose). Regulatory clearance followed, including FDA Emergency Investigational New Drug and local Institutional Review Board approvals. Sterility testing was required before treatment authorisation and represented the rate-limiting step, taking approximately 2·5 weeks to complete. To our knowledge, there are no previous reports describing phage therapy for superinfected ulcerated calciphylaxis lesions. Therefore, initial dosing was guided by published clinical experience in wound infections, available manufacturing titres, and expert consultation.^66,68 All dosing was also carefully coordinated to remain within FDA-specified endotoxin limits of five endotoxin units/kg per h. Treatment began with a 10% test dose (1 × 10⁸ PFU intravenously over 10 min), which was well tolerated. Systemic therapy was then administered at 1 × 10⁹ PFU intravenously every 8 h. After 11 days, the dose was increased to 5 × 10⁹ PFU intravenously every 4 h in response to persistent, albeit reduced, P aeruginosa growth. This dose adjustment was not guided by real-time phage resistance testing, as this was only performed retrospectively. Phage therapy was continued for a total of 7 weeks, initially alongside cefepime–zidebactam and later with cefiderocol plus sulbactam–durlobactam. The full phage course was well tolerated, with no treatment-related adverse events.

Beyond systemic therapy, comprehensive wound management was essential. Our patient was taken to the operating room at least weekly for negative-pressure dressing changes, debridement, and application of ACell, a porcine-derived extracellular matrix scaffold. We also attempted topical phage delivery to maximise local antimicrobial activity. Despite various delivery methods described in the literature including phage-immobilised dressings, hydrogels, and liposome-encapsulated formulations,⁷² only liquid phage preparations were available for our clinical use. This formulation constraint made irrigating wound vacuum systems—which allow controlled fluid instillation and dwell periods—an attractive option. Such systems have been shown to enhance biofilm reduction in vitro with phage dwell times of at least 1 h.⁷³ However, only conventional negative-pressure wound therapy was feasible in our case. Surgifoam sponges soaked in phage solution were therefore placed directly on the open wounds, followed by coverage with Mepitel and Adaptic (non-adhering) dressings before reapplying the wound VAC. Yet this approach faced practical limitations. Surgifoam is not optimised for controlled phage release, and concurrent negative-pressure wound therapy likely reduced sustained local delivery. Given these concerns and the need for xenograft placement (ACell), topical phage therapy was discontinued after three doses. Instead, the wound beds were soaked in a hypochlorous acid–containing, saline-based irrigant (VASHE Wound Cleanser Solution) before ACell application.

Throughout treatment, wound cultures remained polymicrobial—most often growing organisms such as E faecalis (cleared with targeted therapy), coagulase-negative staphylococci, and Candida glabrata—with NDM-1-producing P aeruginosa persisting despite our interventions. In this context, distinguishing active infection from colonisation was challenging. In our immunocompromised patient, deep tissue culture positivity together with progressive clinical decline and systemic features—including persistent fevers and elevated inflammatory markers—favoured active infection. After antimicrobial treatment, fevers resolved and inflammatory markers declined substantially. Nevertheless, the relative contribution of antimicrobial therapy versus calciphylaxis-directed interventions to the favourable outcome could not be definitively determined, as the wounds ultimately healed under the comprehensive treatment regimen.

Over the course of treatment, the patient’s P aeruginosa isolate became resistant not only to cefepime–zidebactam (MIC >512 μg/mL) but also to phage KEN3, as evidenced by declining PFU (figure 3A). Although complete bacterial eradication was not achieved, therapy appeared to select for a slow-growing P aeruginosa phenotype (data not shown). In published reports, phage-driven resistance in multidrug-resistant P aeruginosa has been associated with beneficial evolutionary trade-offs, including reduced bacterial virulence or restored antibiotic susceptibility—a phenomenon sometimes referred to as phage steering or the see-saw effect.^74,75 Even with the emergence of phage resistance, many cases have nonetheless resulted in favourable clinical outcomes.^67,76 Such positive outcomes can also reflect synergistic interactions between phages and antibiotics, as shown in vitro against metallo-β-lactamase-producing P aeruginosa.^67,77,78 That said, real-time testing of phage–antibiotic combinations was not done in this case, representing an opportunity for future optimisation. Beyond bacterial resistance, weekly plasma sampling retrospectively revealed the development of plasma neutralisation of phage activity within 13 days of initiating therapy (figure 3B). Yet the effect of this neutralisation remains unclear, as successful outcomes have been reported even in patients with high anti-phage antibody titres.^79,80

Phage and antibiotic resistance exhibited different kinetics after therapy cessation. Phage resistance reverted rapidly to susceptibility—within 2 days of stopping therapy—suggesting a high fitness cost (figure 3A). In contrast, antibiotic resistance declined more gradually (table). Follow-up cultures obtained 2 months after discontinuation of all antimicrobials (June, 2024) showed isolates were again susceptible to cefepime–zidebactam (MICs decreased from >512 to 64 μg/mL) and cefiderocol (MIC 4–8 μg/mL, within the susceptible–intermediate range). Together, these findings suggest that in-vivo selective pressures from prolonged phage and antibiotic exposure might have imposed fitness costs that became disadvantageous once therapy was withdrawn, permitting re-emergence of susceptibility over time.