Instrumented spinal surgery saw the earliest and broadest clinical uptake, and in routine practice the powder is rarely applied indiscriminately: its use tends to concentrate in selected high-risk strata – revision and long instrumented constructs; multilevel and deformity fusions; and patients with substantial comorbidity, diabetes, prior infection, or prolonged operative time. The seminal cohort of Sweet et al. (2011) reported a reduction in deep SSI from 2.6 % to 0.2 % in 1732 thoracolumbar fusions, a signal reinforced by several subsequent single-centre cohorts (O'Neill et al., 2011; Molinari et al., 2012; Strom et al., 2013; Caroom et al., 2013). The randomised evidence has proved less consistent. Tubaki et al. (2013) found no significant difference in 907 patients (1.61 % vs. 1.68 %), and Salimi et al. (2022) confirmed the neutral finding in 375 patients with a concerning signal toward gram-negative shift. A high-risk cohort in southern Türkiye documented a reduction from 6.54 % to 1.96 % (Oktay et al., 2021), but the retrospective design limits causal inference.

The methodologically most rigorous synthesis is the trial-sequential meta-analysis by Saka et al. (2024), which pooled eight RCTs across major orthopaedic surgery (4307 participants) and calculated a required information size of 19 233 patients. The cumulative Z curve crosses neither the signalling nor the futility boundary; only the subgroup of gram-positive SSI reached significance (RR 0.52; 95 % CI 0.32–0.83), with very low GRADE certainty. Shan et al. (2020) illustrate the same fundamental issue: the overall pooled estimate across 31 studies is RR 0.39, whereas the RCT-only sensitivity analysis yields a non-significant RR of 1.22. A structural limitation underlies these figures and is easily overlooked: the randomised evidence is generated almost entirely in unselected or mixed-risk cohorts and is not stratified by the very risk factors that govern selective use in practice. The pivotal question – whether a benefit undetectable across an unselected population is nonetheless concentrated within a definable high-risk stratum – cannot therefore be resolved from the current data and is more accurately characterised as an evidence gap than as demonstrated inefficacy. This divergence between retrospective and prospective evidence is a recurring feature of the field and is examined in Sect. 3.5.

Retrospective evidence in arthroplasty is predominantly positive. Matziolis et al. (2020) reported a reduction in TKA infection rates from 1.3 % to 0.3 % (P=0.033) with 1 g of intra-articular vancomycin powder; the parallel THA reduction from 1.1 % to 0.5 % did not reach significance. Dial et al. (2018) documented a reduction from 5.5 % to 0.7 % in 265 THA patients but at the cost of 4.4 % sterile-wound complications in the intervention arm versus 0 % in controls. Meta-analyses of mixed retrospective and prospective data yielded pooled odds ratios of around 0.41 (Peng et al., 2021) – estimates that require reassessment in light of the randomised evidence discussed below.

In the randomised setting, a different picture emerges. The VPIP trial (Saba et al., 2025) randomised 1901 high-risk patients – 821 THA and 1080 TKA – at 17 US centres to four arms: 2 g vancomycin powder alone (1 g subfascial plus 1 g suprafascial), 0.35 % DPI lavage alone, combined vancomycin plus DPI, or saline irrigation. No significant differences between arms emerged for any infection-related endpoint (3-month infection rate P=0.14 for THA; P=0.13 for TKA; wound complications 3.2 % and 2.9 %, respectively). Two clarifications by the investigators are important for interpretation. First, the originally planned non-inferiority design (target n=13400) was abandoned after biostatistical consultation because formal non-inferiority testing at this low event rate would have required prohibitively large cohorts. Analysis proceeded by chi-square comparisons and therefore does not constitute a formal non-inferiority proof. Second, the ad hoc analysis at n=1901 showed that extending enrolment to 80 000 patients would not yield meaningful infection trends and that the number needed to treat would remain near 500 for septic reoperation and 250 for any infection requiring reoperation. On these grounds, enrolment was closed for statistical futility, and the investigators place the use of prophylactic vancomycin or DPI in high-risk primary arthroplasty at the discretion of the surgeon or institution.

The situation in revision arthroplasty differs. The dose-stratified meta-analysis by Liao et al. (2022; 14 studies, 35 418 participants) reports significant effects for revision procedures (RR 0.33 (95 % CI 0.14–0.77) for RTKA; RR 0.37 (0.14–0.96) for RTHA), while acknowledging the narrow evidence base and inconsistencies in the 2 g subgroup for primary THA. Gao et al. (2024; 22 studies, 23 363 participants, comprising 19 retrospective cohorts and 3 RCTs) make the methodological problem explicit: the retrospective studies yield a pooled RR of 0.36 (95 % CI 0.22–0.59), whereas the three available RCTs yield a non-significant RR of 0.39 (95 % CI 0.02–6.78; P=0.52). The distinction between primary prophylactic and septic adjunctive-therapeutic use therefore has real evidentiary consequences.

The VANCO trial (O'Toole et al., 2021) randomised 980 patients in modified intention-to-treat analysis – 481 intervention, 499 control – with operatively treated tibial plateau or pilon fractures and at least one high-risk feature (open fracture, compartment syndrome, staged fixation, or NNIS risk index ≥2) across 36 US trauma centres. A single dose of 1000 mg vancomycin powder was applied over the internal fixation construct before closure. The primary endpoint of deep SSI within 182 d narrowly missed the conventional significance threshold: complete case 6.0 % vs. 9.2 %, Kaplan–Meier 6.4 % vs. 9.8 %, and risk difference -3.4 percentage points (P=0.06). The post hoc analysis revealed a significant reduction in gram-positive deep infections from 6.8 % to 3.3 %, while gram-negative deep infections remained numerically comparable (11 vs. 10 cases) – no evidence of a compensatory pathogen shift. The secondary analysis by Joshi et al. (2024) confirmed the selective reduction in gram-positive cocci infections (3.7 % vs. 8.0 %; P=0.01) without a compensatory resistance shift; notably, this reduction was driven predominantly by methicillin-susceptible S. aureus, the incidence of MRSA infection being comparable between arms, so that a specifically MRSA-directed benefit cannot be inferred from the trial. The VANCO investigators interpret these findings cautiously: not all pre-specified analyses reached the 5 % threshold, but the signal supports a clinically meaningful reduction in gram-positive deep infection in high-risk fracture populations.

Septic revision surgery occupies a conceptually distinct position: here, vancomycin powder is applied in the presence of an identified pathogen as part of a comprehensive infection treatment strategy, not as prophylaxis in a clean field. The pharmacodynamic rationale differs accordingly. Biofilm-eroding local concentrations that cannot be achieved by systemic administration without prohibitive toxicity are delivered directly to the site of infection. This adjunctive-therapeutic role aligns with the pharmacodynamics of glycopeptides, whose efficacy depends on time above MIC and local penetration into sessile bacterial populations.

Randomised evidence specific to septic revision is limited. The Liao et al. (2022) meta-analysis provides the most directly relevant synthesis and reports significant reductions in PJI recurrence with adjunctive vancomycin powder in both revision TKA (RR 0.33; 95 % CI 0.14–0.77) and revision THA (RR 0.37; 95 % CI 0.14–0.96). These estimates derive from a small number of predominantly retrospective studies and must be interpreted with caution, but they converge with the pharmacodynamic rationale and with consistent signals from case series using 1000–2000 mg in two-stage exchange arthroplasty. In infected nonunion and chronic osteomyelitis, the same logic applies – local high-concentration delivery as part of multidisciplinary infection management, rather than prophylaxis in isolation. The absence of definitive randomised evidence in these settings should not be read as equivalence to the primary-prophylactic scenario because the therapeutic intent, the microbiological target, and the risk–benefit calculus are fundamentally different. Stewardship principles favour the adjunctive-therapeutic application over routine prophylaxis in clean fields (Sect. 7.3), and this distinction is the conceptual keystone of the recommendation matrix in Sect. 8.

The divergence between positive retrospective cohorts and neutral-to-negative randomised evidence has well-documented methodological causes (Mancino et al., 2023): study heterogeneity, pre–post designs with uncontrolled co-interventions (antiseptic skin preparation, drain management, glycaemic control), publication bias, and surveillance bias. The trial-sequential analysis of Saka et al. (2024) additionally demonstrates that the current randomised evidence lacks the information size to confirm or reject a small-to-moderate effect, and variability in outcome definitions – superficial versus deep SSI, MSIS, or ICM criteria for PJI – compounds the problem. A further, rarely stated observation points the same way: with the singular exception of the VPIP projection (a number needed to treat near 500 for septic reoperation and near 250 for any infection requiring reoperation; Saba et al., 2025), the literature does not report how many patients must be treated to prevent one infection – an omission that is itself instructive, since a preventive measure whose number needed to treat plausibly lies in the hundreds is hard to justify for routine, indication-independent use. A uniform recommendation across all indications is therefore difficult to defend on the current evidence; differentiation by indication is the more rigorous response.

Direct dose-comparison studies remain rare. Kunakornsawat et al. (2019) compared 1 g with 2 g in instrumented spondylodesis and found comparable SSI rates but a higher rate of sterile-wound secretion in the 2 g arm. The dose-stratified meta-analysis by Liao et al. (2022) supports significant effects at 1 g for both TKA (RR 0.38) and THA (RR 0.37), while 2 g showed no benefit in primary THA (RR 1.02). The systemic pharmacokinetics of topical application are well characterised: serum concentrations remain predominantly below the detection limit or well below the nephrotoxic threshold (Sweet et al., 2011; Schär et al., 2021), a finding corroborated by the absence of clinically relevant renal function decline in our institutional elective cohort (Beschauner et al., 2025). The systemic safety margin therefore permits doses up to 2000 mg in larger wound cavities without anticipation of systemic toxicity. The procedural footprint of the technique is minimal – distributing the powder before closure adds well under 2 min, and to our knowledge no trial has reported added operative time as a formal outcome – and the agent itself is inexpensive, at a few euros per gram. Low unit cost should not, however, be mistaken for cost-effectiveness: no formal economic evaluation incorporating the VPIP futility data has been published, and a number needed to treat nearly 500 in primary arthroplasty renders a favourable prophylactic cost-effectiveness ratio in that indication implausible, irrespective of the price per dose.

Brothers et al. (2023) examined the pharmacodynamic consequences of sub-inhibitory vancomycin exposure in planktonic growth assays, biofilm models (titanium rods and fibrinogen-coated plates), and a murine subcutaneous abscess model across three S. aureus strains – MRSA USA300 JE2, MSSA Newman, and MSSA SH1000. At concentrations of 14 MIC (0.25 µg mL⁻¹) and 12 MIC (0.5 µg mL⁻¹), biofilm formation was elevated 1.3- to 2.4-fold relative to untreated controls across all three strains. In the murine model, sub-therapeutic vancomycin pre-exposure raised the infection rate from 41.5 % to 62.5 % (P=0.03; 95 % CI 4.94–37.06; Fig. 1). Cefazolin did not reproduce this effect at any sub-MIC concentration – the finding is substance-specific, not a general sub-inhibitory phenomenon. The pathophysiological basis is partly clarified by Doroshenko et al. (2014): sub-inhibitory pre-exposure enhances extracellular DNA release, which further restricts antibiotic diffusion into the biofilm.

For topical powder application, the implications are direct. Local wound concentrations immediately after application reach 263–2938 µg mL⁻¹ – up to approximately 1000-fold the MIC against MRSA – and subsequently decay in a dose-dependent manner, with standard 1000 mg doses typically falling below the MIC within 3–4 d (Fig. 1A). During this decay phase, the wound milieu inevitably traverses the sub-MIC range, with the consequences described above. When the initial concentration is not raised well above the MIC, the pro-biofilm-forming phase is prolonged relative to the short bactericidal initial phase. A clinically relevant corollary follows: in topical application, under-dosing is potentially more harmful than no application at all. The wide variability in published dose recommendations (500–4000 mg) is, on this reading, a methodological concern in its own right.

At the opposite end of the dose spectrum, sterile-wound complications define the most relevant local safety limit. Dial et al. (2018) observed sterile-wound complications in 4.4 % of intervention cases versus 0 % of controls in primary THA. In vitro data consistently indicate dose-dependent cytotoxicity at high local concentrations: Rathbone et al. (2011) demonstrated impaired osteogenic cell viability and matrix production above 1000 µg mL⁻¹, and Röhner et al. (2021) reported chondrotoxicity at clinically achievable local concentrations. Two practical points frame the dosing decision. First, crystalline powder is a surface-applied adjunct rather than a dead-space management material: it provides no structural fill and elutes within a few days so that where antimicrobial management of dead space is genuinely required – large defect cavities and infected nonunion – carrier-based systems (antibiotic-loaded calcium sulfate, cement beads, bioactive glass) are the appropriate modality and fall outside the scope of this review. Second, the considerations of Sect. 4.1–4.3 converge on a pragmatic dosing logic: 1000 mg as the standard single dose; escalation to 2000 mg reserved for large revision cavities with high dead space and an adjunctive-therapeutic intent; avoidance of doses below 500 mg in larger wounds, where premature sub-MIC decay risks the biofilm-promoting effect (Fig. 1A); and explicit account of wound drainage, which accelerates elution and shortens the supra-MIC window.

The nephrotoxicity of systemic vancomycin has been recognised since 1958; reported incidences of vancomycin-associated acute kidney injury range from approximately 5 % in elective cohorts to above 40 % in intensive care populations (Yasrebi-de Kom et al., 2025), and concomitant piperacillin/tazobactam amplifies this risk (Pan et al., 2025). After topical application, the picture is consistently reassuring: Sweet et al. (2011), O'Toole et al. (2021), and Beschauner et al. (2025) report predominantly non-detectable serum concentrations and no clinically meaningful decline in renal function. In the Halle elective series of 50 patients, only two cases of KDIGO stage 1 acute kidney injury were observed, both resolving without sequelae. Systemic hypersensitivity phenomena are correspondingly exceptional after topical use: red-man syndrome is a rate-dependent, non-allergic reaction tied to systemic vancomycin exposure and is essentially not expected after powder application given negligible absorption, while severe delayed reactions are likewise a concern of systemic rather than topical therapy. Consistent with this, Ghobrial et al. (2015) recorded only 23 complications across 9721 spinal cases (0.3 %). Systemic toxicity is therefore not the limiting factor within standard dose ranges.

Sterile-wound complications – discussed in Sect. 4.3 in the dosing context – represent the principal local safety limit. A further hypothesis-generating signal emerged from the VPIP trial: aseptic revision rates in the THA cohort were 3.0 % in the vancomycin-only arm, 1.9 % in the DPI-only arm, 0 % in the combination arm, and 1.1 % in the saline control (overall P=0.07). The VPIP investigators considered a vancomycin-crystal-related wear mechanism but discounted it with reference to the in vitro wear simulation of Qadir et al. (2014), in which 10 million loading cycles produced no differences between UHMW-polyethylene and cobalt–chromium couplings with or without vancomycin. The signal remains unexplained. While it does not, on its own, mandate a change in practice, it warrants systematic attention in future long-term analyses.

Major international bodies have reached broadly convergent positions. The 2018 WHO Global Guidelines for the Prevention of Surgical Site Infection advise against the use of topical vancomycin powder (conditional recommendation, low-quality evidence; WHO, 2018), and the 2017 CDC SSI guideline (Berríos-Torres et al., 2017) makes no positive recommendation for topical antimicrobial application. Both were issued before the major randomised trials, and their cautious stance has since been reinforced by the VPIP and VANCO results. The North American Spine Society guideline (Shaffer et al., 2013) endorsed topical vancomycin as a low-cost adjunct in spinal surgery; this position predates the current randomised evidence and has not been formally updated. The German S3 guideline on perioperative antibiotic prophylaxis (AWMF 067-009; DGHM, 2025) is the most recent European assessment and concludes, after review of both the randomised and the meta-analytic evidence, that no general recommendation can currently be issued for intrawound vancomycin in spinal surgery – a position reasoned on discrepant study data, rising resistance rates, the lack of gram-negative coverage, and signals of pathogen shift (Ghobrial et al., 2014; 60.7 % gram-negative cultures in the vancomycin group versus 21 % in controls). The dedicated musculoskeletal-infection consensus bodies are consistent with this caution: the International Consensus Meeting on Musculoskeletal Infection, whose diagnostic criteria were used to define PJI in the VPIP trial, addressed local antibiotic delivery in its proceedings (Anemüller et al., 2019) without endorsing routine intrawound vancomycin as a standard of care, and to our knowledge no recommendation specific to vancomycin powder has been issued by EBJIS.

The meta-analytic evidence on pathogen shift under topical glycopeptide exposure is inconsistent. Gande et al. (2019; 28 studies, 19 302 patients) reported a reduction in gram-positive SSI from 70 % to 45.1 % (P<0.05) and a concomitant rise in gram-negative and polymicrobial SSI from 18.5 % to 35.8 % (P<0.05); the relative risk of gram-negative or polymicrobial SSI was elevated by 93.5 %. By contrast, the VANCO secondary analysis (Joshi et al., 2024) found selective reduction in gram-positive cocci (3.7 % vs. 8.0 %; P=0.01) without a compensatory gram-negative increase. The evidence for pathogen shift is therefore not conclusive, but it is sufficient to justify a cautious stewardship stance, particularly in broad indication-independent use. The concern is sharpened by a consideration of spectrum: an agent that extends only gram-positive cover can, by construction, address only that part of the eventual SSI spectrum left uncovered by standard cephalosporin prophylaxis, and where the breakthrough spectrum is thereby displaced toward gram-negative or polymicrobial pathogens – a risk most acute in the lower lumbar and sacral region given proximity to perineal and enteric flora – the empirical treatment of any subsequent infection may demand broader-spectrum agents, compounding rather than relieving selection pressure. No study to our knowledge has directly quantified subsequent broad-spectrum antibiotic consumption after intrawound vancomycin, which is a concrete target for stewardship-oriented research (Sect. 9). Evidence on specific vancomycin resistance is more reassuring: Chotai et al. (2017) studied 2802 patients and found no rise in vancomycin-resistant organisms, though a shift toward gram-negative pathogens and culture-negative seromas was evident. Neither the FDA nor the EMA has approved vancomycin for topical intraoperative wound application; such use is therefore off-label across major regulatory jurisdictions, with implications for patient information, documentation, and institutional governance.

The conceptual distinction between prophylactic use in the absence of proven infection and adjunctive-therapeutic use in the context of established infection – an infected prosthesis, an infected nonunion, and chronic osteomyelitis – is central from an AMS perspective, and it disposes of a common misconception: intrawound vancomycin powder is not itself an instrument of stewardship. In current practice it is invariably given in addition to, never instead of, guideline-concordant systemic prophylaxis or therapy, so it cannot produce the antibiotic-sparing effect that defines a stewardship intervention; the only route by which it might reduce downstream systemic antibiotic use – a consistent, clinically relevant fall in overall SSI or PJI incidence – is precisely what the randomised evidence fails to demonstrate; and its broad, undirected anti-gram-positive activity is conceptually at odds with targeted stewardship. The bearing of stewardship on this technique is therefore as a constraint on use, not as an argument for it. Adjunctive-therapeutic application against a defined pathogen is, by the same logic, more compatible with stewardship principles than routine prophylaxis in non-infected fields, especially in indications where randomised evidence of prophylactic benefit is absent. Two implications follow: topical application must be understood as additive to, not a substitute for, guideline-concordant systemic prophylaxis, and indications should be tied to actual infection risk, not to institutional habit. These principles translate directly into the recommendation matrix of Sect. 8.