Phage therapy is the use of bacteriophages—viruses that infect bacteria—to treat bacterial infections. Therapeutic phages are generally selected for their ability to enter susceptible bacteria, reproduce within them, and destroy them through cell lysis. Unlike most conventional antibiotics, an individual phage often acts against only a limited set of strains within a bacterial species. This specificity makes matching the treatment to the infecting organism central to many clinical approaches.
First developed in the early twentieth century, phage therapy remained in use in parts of Eastern Europe and the former Soviet Union while antibiotics became the dominant antibacterial treatment elsewhere. Increasing antibiotic resistance has renewed international interest in phages, particularly for infections that are difficult to eradicate with available drugs and surgery. Modern research combines historical experience with genome sequencing, controlled clinical studies, improved manufacturing, and bacterial susceptibility testing.
Phage therapy is not a single standardized intervention. It encompasses fixed mixtures, individually selected preparations, naturally occurring phages, and genetically modified candidates delivered by several routes. Published clinical reports describe successful treatment in some patients, but controlled trials have produced mixed findings. Its assessment therefore depends on the particular phage preparation, bacterial target, infection site, accompanying treatment, and quality of the supporting evidence.
Biological basis and therapeutic specificity
Bacteriophages attach to receptors on bacterial surfaces and deliver their genetic material into the cell. These receptors may include surface proteins, capsules, lipopolysaccharides, pili, or flagella. Successful attachment is only the first step: the phage must also overcome intracellular defenses and use the bacterium’s resources to produce new particles. In a productive lytic infection, viral components assemble inside the bacterium, and phage-encoded proteins disrupt the cell envelope, releasing progeny. The interval before release and the number of particles produced vary with both the phage and the physiological condition of its host.
Most therapeutic development favors strictly lytic phages. Temperate phages, by contrast, can establish lysogeny, in which their genetic material persists within a bacterial host rather than immediately destroying it. Some temperate phages carry genes that alter bacterial virulence; well-known examples in bacterial disease include phage-associated toxin genes. For this reason, candidate therapeutic phages are examined for genetic features associated with lysogeny, toxins, and antimicrobial resistance. Genome screening reduces identifiable risks, although many phage genes still have incompletely characterized functions.
Host specificity is both a potential advantage and a practical limitation. A phage active against one strain of Pseudomonas aeruginosa may fail against another strain of the same species because of differences in receptor structure or intracellular defenses. A preparation targeting one pathogen may leave much of the surrounding bacterial community unaffected, unlike an antibiotic with broad activity. However, preservation of the human microbiome cannot be assumed for every preparation, and indirect ecological effects may occur when a targeted organism is removed.
Phages do not productively replicate in human cells, but this does not make their behavior in the body biologically simple. They interact with mucus, extracellular material, immune cells, and circulating proteins. Their bacterial hosts also occupy different physiological states: rapidly growing cells can support phage reproduction differently from nutrient-deprived or dormant cells. In a biofilm, bacteria are embedded in a structured extracellular matrix and may be distributed across steep oxygen and nutrient gradients. Some phages or their enzymes can disrupt particular matrix components, but activity against free-living bacteria does not necessarily predict eradication of an established biofilm.
Discovery and historical development
Frederick Twort described a transmissible phenomenon associated with bacterial destruction in 1915. In 1917, Félix d’Hérelle independently reported an agent that lysed dysentery bacteria and introduced the term bacteriophage, meaning “bacteria eater.” D’Hérelle interpreted the agent as a virus that multiplied in bacteria and rapidly promoted its therapeutic potential. These investigations preceded the availability of electron microscopy, so early arguments about the nature of phages relied on filtration, dilution, culture experiments, and the appearance of clear zones in bacterial lawns.
Therapeutic use began soon afterward, including treatment of dysentery and other bacterial infections. Phage products were developed by laboratories and commercial manufacturers in several countries during the 1920s and 1930s. Results were inconsistent. Some preparations contained poorly characterized mixtures, lacked activity against the patient’s organism, or lost potency during production and storage. Many reports also lacked comparison groups or used endpoints that would not meet later standards of evidence-based medicine. Early failures therefore reflected a mixture of biological limitations, variable products, and weak methods of evaluation.
The development of sulfonamides and the widespread introduction of penicillin during the 1940s changed antibacterial medicine. Antibiotics were comparatively straightforward to standardize and often treated infections without requiring strain-specific matching. Interest in phage treatment diminished in much of Western Europe and North America, although phages remained central to fundamental research in genetics and molecular biology. Therapeutic work continued in the Soviet Union, especially in Georgia, and in Poland. The institute established in Tbilisi in 1923 and later named for George Eliava became a prominent center for phage research and production.
Historical continuity did not produce a uniform international evidence base. Much clinical experience appeared in regional publications, observational series, or reports using methods different from those later required for drug authorization. Consequently, a long history of use cannot by itself establish effectiveness for a modern clinical indication. Renewed international interest from the late twentieth century onward has emphasized bacterial identification, genomic characterization, manufacturing quality, and prospective clinical evaluation.
Widely reported compassionate-use cases in the 2010s accelerated public and scientific attention. These included treatment of a severe multidrug-resistant Acinetobacter baumannii infection in the United States in 2016 and a 2019 report describing engineered phages used against disseminated Mycobacterium abscessus. Such cases demonstrated that coordinated phage selection and delivery were possible in complex hospital settings. They also highlighted the distinction between an informative individual treatment experience and evidence capable of establishing average benefit across a patient population.
Pathogen identification and phage matching
Phage selection usually begins with establishing the cause of infection. Clinical specimens are cultured to recover the suspected pathogen, while diagnosis determines whether the organism is causing disease rather than merely colonizing a body site. This distinction matters in chronic wounds, respiratory samples, and infections involving medical devices, where several bacterial species may be recovered. A phage active against a cultured organism will not necessarily improve the patient’s condition if that organism is not the principal cause of tissue damage.
Laboratories assess susceptibility through several complementary methods. In a spot test, a small volume of phage suspension is placed on a bacterial lawn and examined for clearing. This is useful for screening, but clearing alone does not prove productive phage infection: concentrated preparations may damage cells without generating sustained viral reproduction. Plaque assays provide stronger evidence of productive infection by identifying localized zones of lysis originating from infectious particles. Efficiency of plating compares plaque formation on a test strain with that on a reference host, although results depend on culture conditions.
Liquid-culture assays track bacterial growth in the presence of phages over time. They can reveal delayed killing, incomplete suppression, or regrowth associated with resistant subpopulations. Neither plaque formation nor optical-density measurements fully reproduce conditions in infected tissue. Temperature, acidity, oxygen availability, bacterial growth phase, mucus, serum, and simultaneous antibiotic exposure can change performance. Unlike conventional antibiotic susceptibility testing, phage testing does not yet have universally accepted clinical breakpoints linking a particular laboratory result to a defined probability of treatment success.
Two main treatment models follow from these constraints. Fixed products contain a predetermined phage or mixture intended to cover a useful proportion of a target pathogen population. Personalized preparations are assembled or selected after testing the patient’s isolate against a collection. Fixed products are easier to manufacture consistently and evaluate in conventional trials, but may not cover every strain. Personalized matching can accommodate unusual isolates, although it requires access to collections, rapid testing, and procedures for qualifying newly selected material.
A phage cocktail combines multiple phages, potentially broadening coverage and reducing the likelihood that a single bacterial change will defeat the entire preparation. Nevertheless, the number of components is not a sufficient measure of quality. Phages may recognize the same receptor, share vulnerabilities, or interfere with one another. Serial isolates can also differ from the original specimen as infection evolves. Matching is therefore a continuing microbiological process in some treatment programs, rather than a one-time determination that a bacterial species is generally “phage sensitive.”
Production, formulation, and quality control
Therapeutic phages are commonly recovered from environments containing their bacterial hosts, including sewage, surface water, soil, and clinical material. Isolation is followed by purification of individual phage populations and characterization of host range, killing behavior, and genomic sequence. A manufacturing strain must support reliable propagation while limiting unwanted contaminants. Because production requires bacterial growth, the resulting material initially contains a mixture of phage particles, host-cell fragments, proteins, nucleic acids, and medium components.
Purification is especially important for preparations intended for intravenous administration. Gram-negative bacterial hosts contain endotoxin, principally lipopolysaccharide, which can provoke inflammatory reactions if present at excessive levels. Filtration, chromatography, and other separation processes are used to remove bacterial material while preserving phage infectivity. Product release testing examines identity, potency, purity, and microbiological quality appropriate to the route of administration. Genomic characterization also addresses known toxin genes, resistance determinants, and features incompatible with the intended therapeutic design.
Potency is commonly expressed in plaque-forming units per milliliter, abbreviated PFU/mL. This is a measure of infectious activity under specified assay conditions, not simply a count of physical particles. Two samples with similar particle numbers may differ substantially in infectious titer. Conversely, differences in the bacterial assay host or laboratory procedure can affect measured potency. Molecular methods detect phage genetic material, but detection of a genome does not establish that the corresponding particle remains capable of infecting bacteria.
Formulation must account for phage-specific stability. Temperature, pH, ionic composition, freezing, drying, and contact with containers can affect infectivity. Different phages in the same cocktail may deteriorate at different rates, changing the mixture during storage. Aerosol delivery adds mechanical stresses, while oral formulations must address stomach acidity and other gastrointestinal conditions. Stability studies therefore concern the final product in its intended container and delivery system, rather than the phage in an ideal laboratory buffer alone.
Personalized therapy creates a distinctive manufacturing problem: changing a component may improve bacterial coverage but also changes the administered product. A newly substituted phage may differ in its propagation host, purification behavior, stability, and biological activity. Quality systems must distinguish what can be standardized across a platform from what requires separate evaluation. Phage-derived proteins, such as purified cell-wall-degrading enzymes, present a related but distinct manufacturing pathway because they do not reproduce as intact viruses and are characterized more like other biological medicines.
Administration and behavior within the body
Phages have been administered orally, topically, intravenously, by inhalation, and through local delivery into infected spaces. Route selection in clinical studies reflects the location of infection and the ability of viable particles to reach bacteria. A topical preparation may place phages directly onto an accessible wound, whereas bloodstream or deep-tissue infections create additional distribution barriers. Local delivery during surgery can expose an infected site to high concentrations, but procedures performed at the same time make the independent contribution of phages difficult to measure.
Phage pharmacokinetics differ from those of conventional drugs because infectious particles may both disappear from the body and reproduce within susceptible bacteria. Following intravenous administration, particles can be removed from circulation by the liver, spleen, and other components of the immune system. Distribution into a tissue does not guarantee access to bacteria within abscesses, dense biofilms, or poorly perfused material. The concentration measured in blood may therefore be a weak indicator of activity at the actual infection site.
The possibility of self-amplification is frequently described as an advantage, but it is conditional. A phage must encounter enough susceptible bacteria under conditions that allow productive infection. If bacterial density is low, receptors are inaccessible, or host cells are metabolically unsuitable, phage numbers may decline rather than increase. The multiplicity of infection—the ratio of infectious phages to bacterial cells—is readily defined in a controlled experiment but difficult to establish throughout an infected organ. Local bacterial and phage populations can vary greatly over short distances.
Oral administration introduces barriers including acidity, digestive components, and movement through the intestine. Recovered phages in stool may represent passage through the gastrointestinal tract rather than meaningful replication at the disease site. Inhaled preparations must survive nebulization and reach the relevant airway regions; mucus and heterogeneous ventilation complicate deposition. Similar distinctions apply to other routes: detection of phage material is not equivalent to viable delivery, and viable delivery is not equivalent to clinical benefit.
Repeated exposure may induce neutralizing antibodies that reduce phage activity. The magnitude and clinical consequences vary with the phage, route, treatment duration, and patient. Studies may therefore measure infectious titers, bacterial counts, antibody responses, and clinical changes together. These measurements address different questions: whether the phage arrived, whether it remained infectious, whether the bacterial population changed, and whether the patient’s infection improved.
Resistance, antibiotic combinations, and engineered phages
Bacteria can resist phages by altering or losing surface receptors, producing extracellular barriers, or interfering with viral replication after entry. Restriction–modification systems cleave susceptible foreign genetic material, while CRISPR–Cas systems can provide sequence-specific defense. Other mechanisms interrupt infection through diverse molecular pathways, sometimes sacrificing an infected cell while protecting the surrounding population. These defenses help explain why a phage that performs well against one isolate may fail against another closely related strain.
Resistance may pre-exist in a small bacterial subpopulation or emerge during exposure through mutation and selection. Cocktails can reduce the probability of escape when their components act through sufficiently independent mechanisms, but complete prevention is not guaranteed. Spatially protected populations, incomplete exposure, and physiological changes may also permit bacterial survival without stable genetic resistance. Distinguishing inherited resistance from temporary tolerance is important when interpreting bacterial regrowth in laboratory experiments or persistent infection in patients.
Some resistance mechanisms impose trade-offs. For example, a bacterium that changes a phage receptor may also alter a structure involved in virulence, attachment, or antibiotic efflux. Experiments involving particular P. aeruginosa phages have explored whether selection for phage resistance can increase susceptibility to certain antibiotics. Such evolutionary trade-offs are strain- and phage-dependent; they cannot be assumed whenever resistance emerges. A receptor change may instead have little clinically useful effect or create a different set of biological consequences.
Phage–antibiotic combinations can be synergistic, additive, or antagonistic. An antibiotic may alter bacterial shape or physiology in a way that increases phage production, while another exposure may suppress the cellular processes needed for viral replication. Treatment sequence, drug concentration, bacterial growth state, and the identity of both agents influence the outcome. Consequently, activity observed for one combination cannot be generalized to all members of an antibiotic class or all phages against the same species.
Genetic engineering expands the range of possible interventions. Researchers have modified receptor-binding structures, removed genes associated with lysogeny, and developed phages carrying antibacterial genetic payloads. The 2019 M. abscessus treatment report included phages modified to improve their suitability for bacterial killing. Phage-derived endolysins, which break down bacterial cell walls, are also being investigated as antibacterial agents. These proteins differ from intact phages because they do not self-replicate, and their access to target structures depends strongly on bacterial envelope architecture.
Clinical applications and strength of evidence
Clinical investigation has focused on difficult bacterial infections, including chronic wounds, infected burns, bone and joint infections, device-associated infections, urinary tract infections, and respiratory disease. Frequently studied targets include Staphylococcus aureus, P. aeruginosa, A. baumannii, Klebsiella pneumoniae, and selected mycobacteria. These are not interchangeable indications. Infection anatomy, bacterial physiology, access for local treatment, and the feasibility of removing infected material all influence the likelihood of success.
Compassionate-use reports describe improvements in some patients whose infections persisted despite extensive conventional treatment. They can provide detailed evidence about matching, delivery, tolerability, and microbiological changes. However, patients commonly receive antibiotics, drainage, debridement, or removal of infected devices at the same time. Recovery cannot reliably be attributed to phages alone, and unsuccessful cases may be less likely to receive extensive publication. Observational experience is therefore important for developing hypotheses but cannot substitute for adequately controlled clinical trials.
An early modern randomized controlled trial, published in 2009, studied a phage preparation in 24 patients with chronic ear infection caused by antibiotic-resistant P. aeruginosa. It reported improvements in clinical and microbiological measures compared with placebo, but its small size limited precision and generalizability. A trial published in 2016 involving children with acute bacterial diarrhea in Bangladesh did not demonstrate clinical benefit from oral coliphage preparations. Limited coverage of the relevant organisms and insufficient intestinal phage amplification were important issues in interpreting that result.
The PhagoBurn trial, published in 2019, compared a topical anti-P. aeruginosa cocktail with standard topical treatment in infected burn wounds. Bacterial reduction was slower in the phage group. A major limitation was loss of product titer, meaning that patients received substantially lower concentrations than originally intended. This result illustrates why a negative clinical study must be examined in relation to manufacturing and delivered exposure, without treating those limitations as proof that an adequately formulated product would necessarily have succeeded.
A randomized trial published in 2021 evaluated intravesical phages for urinary tract infection in men undergoing transurethral resection of the prostate. The phage treatment did not show superiority over placebo bladder irrigation. Such findings contrast with some favorable individual reports and emphasize the importance of the comparator and clinical context. In respiratory research, people with cystic fibrosis have been a particular focus because chronic bacterial airway infection is difficult to eradicate, but improvements in bacterial counts do not automatically establish durable effects on lung function or exacerbations.
The evidentiary question is therefore indication-specific: whether a defined product, delivered by a defined route to patients with susceptible organisms, improves a meaningful outcome compared with an appropriate alternative. A broad claim that phage therapy either “works” or “does not work” obscures differences among products and settings. Controlled studies remain essential for determining effect size, identifying suitable populations, and separating bacterial suppression from lasting clinical benefit.
Safety and interpretation of treatment outcomes
Published clinical studies and compassionate-use reports often describe phage administration as tolerable, but the available evidence is not sufficient to exclude uncommon adverse effects. Safety depends on more than the intrinsic biology of the virus. Product purity, bacterial contaminants, administration route, dose, coexisting disease, and simultaneous medicines all matter. Findings from a topical preparation cannot automatically establish the safety of intravenous exposure, and observations from a small group cannot reliably detect rare events.
Inflammatory reactions are a central concern. They may arise from residual bacterial material in a preparation, from components released as bacteria are destroyed, or from the infection itself. Fever, hypotension, and changes in inflammatory markers can have several competing explanations in a severely ill patient. Distinguishing these possibilities requires attention to timing, product testing, cultures, other treatments, and the underlying course of sepsis. Bacterial lysis is not unique to phages, and it does not by itself demonstrate that a harmful systemic reaction will occur.
Gene transfer is another biological consideration. Some phages can move bacterial genetic material between cells through transduction. Selection of appropriate phages and production hosts, together with genome characterization and other testing, aims to reduce this risk. Sequence screening is valuable but not exhaustive: unknown gene functions and mechanisms not apparent from sequence alone limit what it can establish. Engineered candidates additionally require characterization of the intended modification and its effects on biological behavior.
Interpreting outcomes requires separating tolerability, microbiological activity, and clinical effectiveness. A decrease in bacterial counts may be useful evidence of target engagement but may not correspond to healed tissue, restored organ function, or freedom from recurrence. Conversely, a negative culture from an accessible specimen may fail to detect bacteria persisting in deeper tissue. Symptoms may improve because of drainage or antibiotics even when phages contribute little. Appropriate clinical endpoints depend on the disease: wound closure, recurrent infection, device retention, pulmonary exacerbations, or survival address different clinical questions.
Longer follow-up is particularly important for chronic and device-associated infections. Temporary suppression can be followed by relapse from protected bacterial reservoirs, while a new infection may involve a different strain. Serial cultures and strain characterization can help distinguish these events. Structured adverse-event collection and pharmacovigilance are likewise necessary because individualized access programs can otherwise produce fragmented safety information that is difficult to compare across institutions.
Regulation, trial design, and access
Regulatory approaches differ across jurisdictions. Phage preparations have a history of clinical use in Georgia and some other countries, while access in the United States and much of Western Europe has generally depended on clinical research or defined mechanisms for patients with serious disease. In the United States, investigational treatment may proceed through an investigational new drug pathway, including expanded access where applicable. These mechanisms permit treatment under specified conditions; they do not amount to general approval of phage therapy for an infection.
Belgium introduced a framework for magistral phage preparations in 2018, allowing individualized preparations within a pharmacy-based legal structure and associated quality requirements. This approach addresses some difficulties created by patient-specific matching. It differs from authorization of a fixed, industrially manufactured medicine and does not remove the need for quality control or clinical evidence. Across systems, changing a cocktail raises questions about whether the change creates a new product and what additional testing is required.
Conventional trial design is complicated by narrow host ranges. A study may need to identify the pathogen and confirm susceptibility before randomization, potentially slowing recruitment and excluding many otherwise eligible patients. If the product is personalized, participants may receive different phages, requiring a clear definition of the intervention being tested. If the product is fixed, results depend partly on how well its components cover the bacterial strains encountered during recruitment. Both strategies can be studied, but they answer somewhat different questions.
The control treatment must also reflect the infection setting. Placebo alone is not an appropriate substitute for established effective care in a serious bacterial infection; many trials therefore evaluate phages added to standard treatment. Such a design tests incremental benefit rather than whether phages can replace antibiotics. Blinding can be complicated by local procedures, preparation schedules, or changes to individualized products. Prespecified susceptibility criteria, delivered potency, rescue treatment, missing data, and recurrence definitions are important for interpreting the results.
Access depends on infrastructure as well as permission. Phage collections, specialist microbiology laboratories, sequencing, qualified manufacturing, transport, and coordination between treating teams all require resources. These constraints raise questions of health equity, particularly when treatment depends on international shipment or exceptional institutional expertise. Shared collections and standardized testing may improve availability, but their clinical value still depends on demonstrable benefit and reliable product quality.
Economic and organizational incentives also differ from those for conventional antibiotics. Naturally occurring phages, engineered derivatives, cocktails, formulations, and manufacturing methods have different intellectual property implications. Personalized preparations may require substantial laboratory effort for relatively few patients, whereas fixed products need sufficient coverage to support broader use. Integrating phages into antimicrobial stewardship requires evidence about when they add benefit, how they interact with antibiotics, and how bacterial and phage resistance are monitored within routine clinical services.