Prevention Without Eradication: What Nine Device-Associated Infection Trials Reveal About the Translational Failure of Antibiofilm Therapy

1. Introduction

Biofilms are surface-attached microbial communities embedded in a self-produced matrix of extracellular polymeric substances, and they tolerate antibiotic concentrations far above those that kill the same organisms growing planktonically .^1,2 Implanted devices provide a substrate for biofilm formation. Once infection is established on a prosthetic joint, catheter or pacemaker lead, antimicrobial therapy alone may be insufficient, and management often requires removal or exchange of the device; however, selected infections can be managed by debridement with implant retention and systemic antimicrobial therapy .^3,4,5

The laboratory literature has generated many proposed antibiofilm approaches, including matrix-degrading enzymes, quorum-sensing inhibitors, bacteriophages, antimicrobial peptides, nanocarriers and photodynamic or photothermal methods .^6,7,8 What these catalogues rarely examine is what happened when such agents were tested in patients.

This review asks that question directly. It takes selected clinical evidence for nine device types as its evidence base and examines, for each, what was tested, what was found, and what the result suggests about translation. The analysis identifies an asymmetry: several successful device-surface prophylactic technologies prevent biofilm formation on clean devices, whereas evidence for novel adjunctive treatment of established infection is less mature. Because this is a selected narrative synthesis rather than an exhaustive systematic review, it cannot establish the relative importance of trial design, pharmacology, biology or commercial factors in the field as a whole.

1.1. Search Strategy

Studies were identified through PubMed, ClinicalTrials.gov and the EU Clinical Trials Register, searched from inception to 2026, combining device terms (catheter, prosthetic joint, cardiac implantable electronic device, endotracheal tube, vascular graft, ventriculoperitoneal shunt, dental implant) with intervention terms (antimicrobial coating, lock solution, bacteriophage, antimicrobial peptide, impregnated) and infection terms (biofilm, device-related infection, colonisation). Randomised controlled trials in humans were prioritised; where no randomised evidence existed, pilot studies, regulatory decisions and sponsor disclosures were used and are identified as such. One or more representative, well-documented examples were selected per device to allow depth of analysis rather than exhaustive enumeration. This is a narrative synthesis, not a systematic review, and no meta-analysis was attempted.

2. The Device Trial Record

Table 1 summarises the nine device settings. Each is considered below in terms of what was tested, what was found, and what the result teaches about trial design and translation.

2.1. Central Venous Catheters: Local Delivery Works When the Drug Suits the Surface

Darouiche et al.⁹ randomised 865 catheters in 817 high-risk patients across twelve hospitals to polyurethane triple-lumen catheters impregnated with either minocycline and rifampin or chlorhexidine and silver sulfadiazine. Among 738 catheters with evaluable cultures, colonisation occurred in 7.9% of minocycline-rifampin catheters versus 22.8% of chlorhexidine-silver sulfadiazine catheters. Catheter-related bloodstream infection occurred in 0.3% versus 3.4%, respectively, with the advantage most marked in patients requiring access for seven days or more.

Two design features may explain the result. The minocycline-rifampin catheters were impregnated on both luminal and external surfaces, whereas the comparator was coated externally only, so the intervention covered the route by which much late catheter colonisation occurs. Catheter colonisation on removal was also a direct microbiological endpoint rather than a downstream clinical proxy. However, because antimicrobial composition and surface coverage differed simultaneously, the trial compares two product strategies; it does not isolate the independent contribution of each component.

2.2. Haemodialysis Catheters: An Approved Antimicrobial Lock

The LOCK IT-100 trial randomised adults receiving maintenance haemodialysis through a central venous catheter at 70 US sites to a taurolidine-heparin lock solution or heparin alone .¹⁰ In the full-analysis population of 795 participants, catheter-related bloodstream infection occurred in 9 of 397 participants receiving taurolidine-heparin compared with 32 of 398 receiving heparin, corresponding to event rates of 0.13 and 0.46 per 1000 catheter days and a 71% reduction in risk. The FDA approved Defencath (taurolidine and heparin) on 15 November 2023 to reduce the incidence of catheter-related bloodstream infection in a defined adult haemodialysis population .¹¹

The trial is instructive for three reasons. The intervention is instilled into the catheter lumen, achieving taurolidine concentrations of 13,500 µg/mL that no systemic route could approach. The population was narrow and at high baseline risk, which made the event rate high enough to power the trial at a feasible size. And taurolidine acts through non-specific chemical activity rather than a conventional target-based antibiotic mechanism, although this does not establish that biofilm tolerance is irrelevant to its activity.

2.3. Cardiac Implantable Electronic Devices: The Cost of a Rare Endpoint

WRAP-IT randomised 6983 patients undergoing device replacement, upgrade, revision or initial cardiac resynchronisation therapy defibrillator implantation to an absorbable envelope eluting minocycline and rifampin, or to standard infection prevention alone .¹² Major infection within 12 months occurred in 0.7% of the envelope group versus 1.2% of controls, a 40% relative reduction, with pocket infection reduced by 61% and no increase in complications.

The arithmetic is the lesson. An absolute risk reduction of 0.5% required almost 7000 patients to demonstrate. This is a structural problem facing prevention trials for any device whose infection rate is already low: the intervention may work, but only a large, often industry-funded trial can demonstrate it. The envelope also illustrates local delivery. Preclinical pharmacokinetic work reports local tissue concentrations above the minimum inhibitory concentration sustained for at least seven days, with serum concentrations orders of magnitude lower and, in some models, at or below the assay limit of detection; human serum pharmacokinetic data are not reported in the trial itself, so the envelope is best described as achieving high local relative to systemic exposure rather than as producing undetectable systemic levels. The trial population was also enriched: eligibility was restricted to generator replacement, upgrade, revision or de novo cardiac resynchronisation therapy defibrillator implantation, that is, patients at increased risk of pocket infection .¹² Observational real-world series in broader populations have reported mixed results, and the WRAP-IT estimate should not be assumed to transfer to unselected recipients.

2.4. Endotracheal Tubes: When the Endpoint Does not Mean What it Appears to Mean

The NASCENT trial randomised 2003 patients to a silver-coated or uncoated endotracheal tube; 1509 patients were intubated for 24 hours or longer and formed the population underlying the principal reported ventilator-associated pneumonia comparison .¹³ Microbiologically confirmed ventilator-associated pneumonia occurred in 4.8% of the silver-coated group versus 7.5% of controls, a relative risk reduction of 36%, with delayed time to occurrence.

The trial met its primary endpoint, yet the intervention did not become standard care. Klompas¹⁴ argued in correspondence that the evidence indicated an effect on endobronchial colonisation rather than on clinically meaningful invasive disease, noting that the trial showed no difference in duration of ventilation, intensive care stay or mortality. The episode illustrates a recurring problem: microbiologically defined endpoints are the natural way to measure antibiofilm activity, and ventilator-associated pneumonia is difficult to adjudicate, so a trial can be simultaneously positive and unconvincing. An antibiofilm intervention that reduces colonisation without changing patient-centred outcomes may not be adopted, however sound its microbiology.

2.5. Urinary Catheters: Failure to Confirm an Anticipated Effect

Meta-analysis of small randomised studies available before 2012 suggested that silver alloy catheters reduced catheter-associated urinary tract infection by about 40% (relative risk 0.60). The CATHETER trial tested this in 7102 randomised adults requiring short-term catheterisation, of whom 6394 were analysed, comparing silver alloy-coated, nitrofural-impregnated and standard PTFE catheters .¹⁵ Symptomatic urinary tract infection within six weeks occurred in 12.5% of the silver alloy group and 12.6% of controls, an absolute difference of -0.1%. The nitrofural catheter reduced infection to 10.6%, a reduction smaller than the prespecified threshold for clinical importance, and caused significantly more discomfort.

The authors concluded that their results gave no support for routine use of silver alloy catheters. The larger pragmatic trial did not confirm the benefit suggested by earlier evidence. For antibiofilm research, this is a cautionary case about evidence hierarchy: a coherent in vitro mechanism, consistent animal data and a favourable meta-analysis of small studies together predicted an effect that was not confirmed in a large trial.

2.6. Prosthetic Vascular Grafts: Strong Animal Data, Unconfirmed Human Benefit

Rifampicin bonding to gelatin-sealed Dacron protects grafts in animal challenge models with striking consistency. In a canine model, no rifampicin-treated graft became infected after bacteraemic challenge with meticillin-resistant Staphylococcus aureus, whereas every graft in both control groups did .¹⁶ In human trials, the anticipated effect was not confirmed. A 600-patient prospective trial found graft infection in 1.7% of the rifampicin-bonded group versus 2.3% of controls, a non-significant difference ,¹⁷ and a randomised trial in extra-anatomic reconstruction, chosen because such grafts carry high infection risk, found a single infection in 257 patients and no advantage from bonding .¹⁸ A subsequent systematic review of graft coatings concluded that neither rifampicin nor silver coating showed significant benefit, and that very large samples would be required to demonstrate one.

The discrepancy does not prove that rifampicin fails against biofilms. Animal challenge models inoculate a large bacterial load onto the graft at a defined moment, generating a high event rate; human graft infection is rare, delayed and multifactorial. The models and the trials may be measuring different phenomena, although this remains an interpretation rather than a demonstrated causal explanation for each null human result.

2.7. Prosthetic Joints: Treatment of Established Infection

Prosthetic joint infection is one setting in which antibiofilm agents have been tested against established infection rather than for prevention. The PhagoDAIR I pilot study randomised patients with staphylococcal hip or knee prosthetic joint infection treated by debridement, antibiotics and implant retention to intra-articular anti-S. aureus phages or placebo. The sponsor reported that, among evaluable patients, infection control at 12 weeks occurred in 14 of 19 patients receiving phages (74%) and 5 of 7 patients receiving placebo (71%). The study was non-comparative by protocol and too small to establish an incremental treatment effect. Phages were selected for activity against the patient’s isolate using a phagogram. As of October 2026, these results remain unpublished in peer-reviewed form and are available only as a conference presentation and company announcement ;^19,20 they have not been subject to external review and are cited with that limitation.

Every feature of the study constrains what can be concluded. It was non-comparative by protocol. It recruited 29 patients against a planned 64, which the sponsor attributed to restrictive inclusion criteria, with 26 evaluable patients. The sponsor also reported an 80% infection-control rate (16/20) after rescue treatment, including a placebo-arm patient subsequently treated with phages; this rescue analysis should not be conflated with the original treatment-arm result. The wider phage literature in prosthetic joint infection has a similar structure: compassionate-use series and case reports describe high remission rates, but few include a control arm .²¹ The follow-on GLORIA phase II trial was designed to address this uncertainty, but the sponsor’s financial difficulties have placed the programme in doubt; the current registry and corporate status should be verified before publication.

This matters beyond phage therapy. Conventional biofilm-active treatment has also been tested in established orthopaedic implant infection. Zimmerli et al. conducted a randomised, placebo-controlled trial of rifampicin combination treatment in 33 patients with staphylococcal orthopaedic implant infection; among 24 patients completing the trial, cure occurred in 12/12 patients receiving ciprofloxacin-rifampicin and 7/12 receiving ciprofloxacin-placebo. The trial was small, included mixed orthopaedic implants rather than exclusively prosthetic joints, and had substantial attrition, but it demonstrates that randomised treatment studies of established implant-associated infection predate the phage programme .²² A later randomised trial of rifampicin in prosthetic joint infection has also been reported .^23,5

The most informative trial design for a novel eradication agent, a randomised placebo-controlled test on a retained implant, remains vulnerable to restrictive eligibility criteria, small recruitment, and financial fragility of a small sponsor. These are not microbiological problems, but they are not the only barriers to translation.

2.8. Cerebrospinal Fluid Shunts: A Comparison of Two Coating Strategies

BASICS randomised 1605 patients of all ages undergoing insertion of a first ventriculoperitoneal shunt at 21 centres in the UK and Ireland; 1594 patients who received a shunt without infection at insertion entered the reported analysis population. They received a standard silicone shunt, a shunt impregnated with rifampicin and clindamycin, or a silver-impregnated shunt .²⁴ Shunt revision for infection occurred in 6.0% of the standard group, 2.2% of the antibiotic group and 5.9% of the silver group. The antibiotic shunt reduced infection substantially (cause-specific hazard ratio 0.38, 97.5% CI 0.18–0.80), whereas the silver shunt did not (hazard ratio 0.99, 97.5% CI 0.56–1.74). Antibiotic shunts were cost-saving, and the result changed UK practice.

The three-arm design makes this an especially informative trial because it holds the device, operation, population and endpoint broadly constant while varying the antimicrobial strategy. Local delivery is not sufficient on its own: the impregnation concept worked with rifampicin and clindamycin and did not work with silver in the same setting. The trial also illustrates how weak the prior evidence had been. The investigators noted that their systematic review found only one randomised trial and no adequately powered study of antibiotic shunts, and no randomised trial of silver shunts at all, yet both were already in clinical use.

2.9. Dental Implants: The Eradication Problem in Miniature

Peri-implantitis is biofilm-associated inflammation of an implant that cannot practicably be removed, so it is treated by mechanical debridement of the implant surface, with or without antimicrobial adjuncts. It is therefore one of the most frequently studied settings for treatment of established device-associated biofilm infection. Grusovin et al.²⁵ found statistically significant improvements in probing pocket depth, bleeding on probing and success rate with adjunctive local antibiotics, with an additional probing-depth improvement of 0.6 mm (95% CI 0.42–0.78 mm). However, disease resolution remained approximately 20–30%, and the small, heterogeneous evidence base did not support routine antibiotic use.

A recent meta-analysis found limited improvements in selected clinical outcomes up to 12 months with adjunctive systemic antibiotics, without consistent marginal-bone-level benefit; longer-term effectiveness remains uncertain .²⁶ A network meta-analysis found systemic metronidazole, alone with non-surgical treatment or combined with amoxicillin alongside open flap debridement, to be among the more effective regimens for reducing probing depth and marginal bone loss, while noting inconsistent microbiological outcomes .²⁷ Reviews reaching differing conclusions from an overlapping trial base indicate heterogeneity in debridement technique, case definition, implant surface and follow-up large enough to obscure the antimicrobial effect being measured.

A representative randomised trial makes the point concretely: adding systemic amoxicillin and metronidazole to non-surgical treatment produced no statistically significant advantage over debridement alone, and 81% of target sites still bled on probing after treatment in both arms .²⁸

Two features of this literature matter for the wider argument. First, mechanical disruption is doing much of the work: antimicrobials supplement debridement rather than replace it. Second, the endpoints are surrogate measures of inflammation, probing depth and bleeding on probing, rather than microbiological eradication, so the trials cannot distinguish an agent that failed to reach the biofilm from one that reached it and was ineffective. A field that has run many randomised trials over two decades and still reaches differing conclusions is describing a measurement and evidence-quality problem, alongside possible pharmacological limitations.

3. Synthesis: Prevention and Eradication

Considered individually, each study above has a conventional explanation. Considered together, they identify a pattern that deserves greater attention in the antibiofilm literature.

3.1. Successful Device-Surface Prophylaxis

Among the selected device-surface prophylactic technologies reviewed here, four interventions are approved, adopted or widely endorsed: minocycline-rifampin impregnated central venous catheters, taurolidine-heparin lock solution, the antibacterial envelope for cardiac devices, and the rifampicin-clindamycin impregnated ventriculoperitoneal shunt. All four are prophylactic. Each is applied to a device that is not yet infected and acts during the window between implantation and biofilm maturation. None is indicated specifically for eradication of an established biofilm.

By contrast, the settings in this review where an established biofilm on a retained implant was the target behave differently. PhagoDAIR I remains a small non-comparative pilot. Peri-implantitis has been studied in many randomised trials, and these have not fully converged: some reviews find that antimicrobial adjuncts add little to mechanical debridement, while others identify specific systemic regimens that improve selected clinical outcomes .^25,26,27 What is not in dispute is that debridement remains indispensable and that no adjunct has been shown to eradicate an established peri-implant biofilm on its own.

Management depends on the device, pathogen, infection duration and implant stability. Many established infections require removal or exchange, while selected cases can be treated using debridement and implant retention with systemic antimicrobial therapy. Rifampicin-containing regimens are an established example of systemic biofilm-active treatment in selected staphylococcal implant-associated infections .^22,5

This asymmetry is partly the inverse of the preclinical literature. In vitro antibiofilm screening often uses mature biofilms, typically 24 to 48 hours old, and reports minimum biofilm eradication concentrations. Reviews are often organised around dispersing, degrading and killing established communities .^6,7 The field’s laboratory effort is therefore often directed at a problem that its clinical successes have not solved, while successful clinical prevention occupies the narrower early window at a device surface.

3.2. Why Prevention Is Easier to Prove

Three features recur in the successful prophylactic trials. First, delivery is local and achieves concentrations unattainable systemically: taurolidine at 13,500 µg/mL within a catheter lumen, and antibiotics eluted into a device pocket. Second, the target is a defined population at high baseline risk, which raises the event rate and makes the trial feasible. Third, the intervention acts before the matrix exists, so the diffusion barrier that defines biofilm tolerance is not encountered.

Eradication inverts these conditions. The agent must reach an established matrix, the population is heterogeneous in organism, duration and implant type, and the comparator, surgical debridement with prolonged antibiotics, is itself often effective, so a trial must demonstrate an increment over an active and variable standard.

3.3. The Silver Anomaly

Silver deserves separate comment because it was tested against four of the nine device settings and offers an unusually clean natural experiment. Silver ions are broadly antimicrobial in vitro, silver coatings reduce biofilm formation on test surfaces, and silver has been incorporated into catheters, shunts, endotracheal tubes and vascular grafts. In adequately powered human trials, the effect is largely absent. Silver alloy urinary catheters produced an absolute difference of -0.1% against standard catheters .¹⁵ Silver-impregnated shunts produced a hazard ratio of 0.99 in a trial where antibiotic shunts in the same population produced 0.38 .²⁴ In aortic vascular grafts, a systematic review and meta-analysis found that both rifampicin and silver coatings showed a trend towards protection but no significant difference, and noted that clinical studies in this position are scarce ;²⁹ current aorto-iliac guidance accordingly does not support routine use of silver-coated or rifampicin-soaked grafts. The one apparently positive result, the silver-coated endotracheal tube, is also the one whose endpoint was contested .^13,14

The pattern suggests that in vitro antimicrobial activity on a surface is a poor predictor of clinical device protection, and that the quantity of free ionic silver released into a protein-rich, flowing biological environment may be insufficient to matter. For the antibiofilm field, the lesson is specific: of the mechanisms reviewed here, the one with the longest history, broadest spectrum and most consistent in vitro data has the weakest clinical record.

3.4. Recurring Technical Failures

Beyond the prevention-eradication divide, four failure modes recur and are, in principle, remediable.

Endpoint validity. NASCENT showed that a microbiologically defined endpoint can be met without convincing clinicians that patients benefited. Antibiofilm interventions are vulnerable here because their direct effect is on colonisation, which is one step removed from clinical disease.

Extrapolation from favourable models. The silver alloy catheter and the rifampicin-bonded graft both entered pivotal trials on the strength of mechanism, animal data and meta-analyses of small studies. Their anticipated effects were not confirmed or were trivial. Animal challenge models, which inoculate a high bacterial load at a defined moment, may overstate what is achievable against sporadic, low-inoculum human device infection.

Event rates and feasible sample size. WRAP-IT required 6983 patients to detect an absolute reduction of 0.5%. Vascular graft infection occurred in 0.4% of patients in one trial. Where baseline rates are this low, a trial may be affordable only to a device manufacturer, which shapes which questions are asked.

Product and sponsor fragility. PhagoBurn, although a wound rather than a device study, remains a clear example: phage titre fell from 10⁶ to 10² PFU/mL during manufacture, so patients received a fraction of the intended dose, and, in the per-protocol ancillary analysis, three of ten phage-treated patients carried P. aeruginosa with reduced phage susceptibility at day 0; isolates from patients in whom treatment failed were resistant to low phage doses, and a fixed cocktail had been used without prior phagogram screening .³⁰ PhagoDAIR’s successor programme lost its sponsor: Phaxiam Therapeutics entered receivership in March 2025 and was placed in judicial liquidation by the Lyon commercial court on 11 June 2025, with no buyer found for its assets .³¹ Biological agents may fail for pharmaceutical and commercial reasons before their biology is adequately tested.

3.5. Susceptibility Testing

None of the selected trials used a standardised biofilm-specific susceptibility measure to select or stratify patients. Minimum biofilm eradication and inhibition concentrations have been described for two decades but remain insufficiently standardised for routine clinical use .³² PhagoBurn’s pre-existing phage resistance would have been detectable by phagogram before randomisation. PhagoDAIR used a phagogram to select phages active against the patient’s isolate, although a phagogram is not necessarily equivalent to a biofilm-state susceptibility assay.

Biofilm susceptibility assays are poorly reproducible between laboratories, depend heavily on the growth model used, and lack agreed breakpoints. A trial that cannot confirm that the target organism is susceptible to the investigational agent in its biofilm state is testing two hypotheses at once, and a null result cannot distinguish between them. Standardising a biofilm assay would not by itself produce better agents, but it could allow trials to enrol patients whose organisms are susceptible to the intervention being tested.

4. Implications for Trial Design

The following recommendations follow from the patterns above rather than from any single trial.

• Enrich for event rate. Prevention trials should recruit populations with the highest achievable baseline infection rates, as LOCK IT-100 did, rather than general device recipients. The alternative is a sample size only industry can fund.
• Separate the two questions. Prevention and eradication require different trials, populations and endpoints. Eradication data should not be used to support prevention products, and the reverse.
• Match the endpoint to the claim. A colonisation endpoint supports a colonisation claim. If adoption depends on patient-centred outcomes such as device retention, reoperation or mortality, those must be the endpoints, with the sample size that implies.
• Use appropriate susceptibility testing as an eligibility criterion. For agents with narrow spectra, phages above all, susceptibility of the patient’s own isolate should be confirmed before randomisation. Where feasible, biofilm-state susceptibility should be assessed.
• Default to local delivery where biologically appropriate. Every successful prophylactic intervention in this review delivers drug at the device surface. Systemic administration of an agent whose target is a surface-attached community should require specific justification.
• Design eradication trials around implant retention. Debridement with implant retention is a clinical situation where eradication of an established biofilm is both necessary and potentially measurable, with a clear endpoint in infection control at a fixed interval and a defined comparator.
• Anticipate sponsor risk. Where a programme depends on a single small company, as phage therapy largely does, academic and public funders should consider mechanisms that allow a trial to survive the sponsor.

5. Conclusions

Across these selected device-associated infection settings, several locally delivered prophylactic interventions reduce infection risk. Evidence for novel adjunctive treatment of established infection is less mature and is often limited by recruitment, endpoint selection, manufacturing or commercial barriers. However, conventional biofilm-active antibiotic regimens and selected implant-retention strategies already form part of clinical management. These cases therefore identify important translational barriers without establishing that trial design alone explains the limited clinical adoption of novel antibiofilm therapies.

The failures reviewed here were not necessarily refutations of antibiofilm pharmacology. They involved endpoints that measured colonisation rather than disease, effects extrapolated from models that may not reproduce human infection, event rates demanding very large samples, products that degraded before reaching patients, and organisms not tested for susceptibility in the relevant state. These are soluble problems, but they are problems of trial design and drug development as well as discovery. Until the field treats them as central difficulties, additional in vitro antibiofilm agents are unlikely to change what happens to patients with infected devices.