From:  Bacteriophage therapy in the era of antimicrobial resistance: clinical evidence, translational challenges, and emerging regulatory frameworks (2020–2026)

 Comparing key 2020–2026 clinical trials and high-value clinical datasets.

Authors and YearIndication/pathogen focusDesign & sizeRoute/dosing (high-level)Key results (selected)Critical notesEvidence level (CEBM)Strength of evidence
Petrovic Fabijan et al. (2020)
[56]
Severe S. aureus infection incl. endocarditis/septic shockSingle-arm, non-comparative under expanded access; n = 13IV phage cocktail AB-SA01 adjunct to antibioticsNo adverse reactions reported; dosing rationale suggested for further trialsGMP-quality preparation and pre-specified safety primary outcome strengthen translational credibility within early-phase evidence; however, single-arm design and small n (13) preclude efficacy inference and limit PK/PD generalization; immune neutralization not assessed, leaving a relevant safety dimension uncharacterized; salvage population further restricts external validity.Level 4—Single-arm non-comparative; safety/PK primaryUncontrolled—structured
Pirnay et al. (2024)
[10]
Difficult-to-treat infections, multiple pathogensRetrospective observational; n = 100 BT cases (subset from > 1,000 requests)Personalized products; various routes; often + antibioticsClinical improvement 77.2%; eradication 61.3%; eradication ~70% less likely without antibiotics; 15 adverse events (7 suspected ADR) resolvedConsecutive case ascertainment substantially mitigates the success bias typical of retrospective phage cohorts and strengthens face validity of pooled outcomes; linkage to QC and production metrics adds methodological transparency uncommon in this literature; however, retrospective observational design, patient heterogeneity, variable routes/products, and uncontrolled concomitant therapies preclude causal inference, particularly for the antibiotic synergy signal.Level 4—Retrospective observational cohortUncontrolled—structured
Kim et al. (2024)
[39]
uUTI due to E. coli with history of drug-resistant UTIRandomized open-label dosing study; ITT n = 39Intraurethral + IV LBP-EC01 + TMP-SMX; multiple IV dose regimens testedNo serious AEs; 46% with AEs (more with higher IV dosing); rapid E. coli reduction and symptom resolution by day 10 in evaluable patientsEngineered phage with robust PK characterization and randomized dose-ranging design represents a methodological advancement over compassionate-use studies in the field; tolerability boundaries identified at higher systemic exposure provide actionable safety information for future trials; however, open-label design introduces observer bias risk for symptom resolution endpoints; absence of placebo or standard-of-care comparator limits efficacy inference; co-administration of TMP-SMX confounds attribution of microbiological response to phage activity specifically.Level 2—Randomized uncontrolled phase II dose-ranging trialRandomized uncontrolled—dose-ranging
Chan et al. (2025)
[41]
CF adults with MDR/PDR P. aeruginosaCompassionate cohort; n = 9Nebulized single phage or cocktails; receptor/trade-off selection strategyMedian sputum Pseudomonas decrease (104 CFU/mL); ppFEV1 improvement signal; no adverse events; microbiome not alteredEvolutionary trade-off phage selection and pre-specified multi-endpoint measurement (sputum density, ppFEV1, microbiome) represent methodological advances over earlier case series; however, compassionate-use design, small n (9), salvage population, and patient-specific regimens preclude efficacy inference.Level 4—Compassionate-use case seriesUncontrolled, structured
Petersen et al. (2026)
[7]
Gut E. coli colonization in healthy volunteers (decolonization rationale for HSCT risk reduction)Randomized, placebo-controlled, double-blind; n = 36Oral BID × 7 days; escalating doses (108–1012 PFU/dose)No grade 3–4 AEs; no SAEs in treated groups; phage recovered in stool; microbiome stable; 78% reduction in E. coli vs placebo at day 14 (NS)Randomized placebo-controlled double-blind design with healthy-volunteer cohort and pre-specified safety, PK, and microbiome endpoints represents the most methodologically rigorous phage therapy trial design to date in the field; CRISPR-Cas3-armed engineered phage with structured PK characterization across five orders of magnitude in dose offers an unusually complete biological profile for a first-in-human study; however, the 78% reduction in E. coli at day 14 did not reach statistical significance, limiting interpretation of pharmacodynamic activity; healthy-volunteer design does not characterize drug behaviour in the immunocompromised target population (HSCT recipients), and the gap between asymptomatic colonisation reduction and clinical infection prevention requires dedicated efficacy trials.Level 2—Randomized placebo-controlled double-blind phase 1 dose-escalation trialControlled, exploratory
Gorodnichev et al. (2026)
[8]
ICU VAP with MDR K. pneumoniaeProspective open-label non-randomized with comparator groups; n = 21 (3 × 7)Inhaled phage cocktail BID ×14 days (targeted vs non-targeted vs no phage) + antibioticsEradication at day 14: 86% targeted vs 57% antibiotics-only vs 0% non-targeted; no therapy-related AEsInclusion of a non-targeted phage control arm is rare in phage therapy literature and provides a methodological control for non-specific effects of phage administration that pure phage-vs-no-phage designs cannot; comparative design with concurrent controls represents a meaningful step up from single-arm case series; however, non-randomized allocation introduces selection bias risk between arms; very small per-arm sample size (n = 7) produces wide confidence intervals around the headline percentages, and the 0% eradication in the non-targeted arm cannot be cleanly attributed to phage mismatch without controlling for between-arm differences; open-label design adds observer bias risk.Level 3—Prospective non-randomised controlled study with concurrent comparator armsControlled, exploratory
Paul et al. (2021)
[61]
Abdominal infection with vancomycin-resistant E. faeciumCase reportIV magistral preparation with two enterococcal phagesClinical improvement correlated with CRP reduction and no associated clinical AEsIllustrates the feasibility of magistral pharmacy-prepared phage products as a regulatory pathway for individualized therapy in paediatric salvage settings; however, single-patient observation with no comparator cannot distinguish phage-attributable improvement from antibiotic effect, surgical management, supportive care, or natural disease trajectory; generalisability to other patients is precluded by the inherent limitations of single-case evidence.Level 5—Single case reportUncontrolled, exploratory
Rodriguez et al. (2022)
[62]
Refractory MRSA chronic rhinosinusitisCase report Systemic + intranasal phage + antibioticsReported successful treatment of refractory CRS using combined systemic and topical phage therapy administered alongside antibioticsIllustrates the use of compartment-targeted dual delivery (systemic + intranasal) for an anatomically defined infection site, demonstrating the feasibility of combined delivery strategies in difficult-to-treat upper airway infections; however, single-patient observation with concomitant oritavancin administration cannot distinguish phage-attributable improvement from antibiotic effect or natural disease fluctuation.Level 5—Single case reportUncontrolled, exploratory
Tan et al. (2021)
[63]
MDR lung infection due to carbapenem-resistant A. baumanniiCase reportNebulized personalized phage + antibioticsReported clinical improvementDocuments nebulised delivery of a personalised single-phage preparation matched to patient isolate in a critically complex patient profile (elderly, ventilator-dependent, multiple comorbidities); reports phage-resistance evolutionary trade-off in recovered isolates; however, dual concomitant antibiotic administration (tigecycline + polymyxin E) confounds attribution; single-phage (non-cocktail) approach carries elevated resistance risk; high endotoxin level in the preparation, acknowledged by the authors, raises preparation-quality concerns.Level 5—Single case reportUncontrolled, exploratory

Evidence levels in this table follow the Oxford Centre for Evidence-Based Medicine (CEBM) Levels of Evidence (https://www.cebm.ox.ac.uk/resources/levels-of-evidence/ocebm-levels-of-evidence). Strength of evidence is graded within each tier based on the presence of a comparator arm (‘controlled’ vs ‘uncontrolled’), randomization, and the degree of methodological structure of the study (pre-specified outcomes, protocol-driven enrolment, regulatory framework, and quality verification of the intervention). ‘Structured’ studies have pre-specified outcomes and formal protocols; ‘exploratory’ studies do not. The body of evidence summarized below is predominantly hypothesis-generating; routine clinical adoption should await confirmatory randomized trials. MDR: multidrug-resistant; PDR: pandrug-resistant; GMP: Good Manufacturing Practice; QC quality control; PK/PD: pharmacokinetics and pharmacodynamics.