Infected wounds needing instillation (NPWTi-d)

CAT: OCEBM 2 · Start High (randomised, 132 patients, three arms, four-year enrolment). For wound closure time, -1 risk of bias: no blinding of any party and no blinded endpoint assessment, with definitive closure decided 'when the wound was considered clean based on the same criteria' - a subjective, unblinded surgeon judgement; complete-case analysis with 12 of 132 (9.1%) excluded post-randomisation; and two eligibility criteria ('failure to collect intraoperative cultures', 'changing of surgical management') that can only be determined after allocation and after the first operation, so they are post-randomisation exclusions presented as eligibility criteria. Precision is not downgraded: the NPWTi-d versus NPWT difference is a mean difference of -3.90 days (95% CI -6.28 to -1.52, Hedges g -0.72), the sample size calculation reproduces exactly (n = 36 per group at alpha 0.025, 80% power, delta 3 days, SD 4), and enrolment exceeded it. Net moderate. For the number of surgical procedures, -3: the outcome is arithmetically determined by an arm-specific, protocol-mandated debridement interval. Dividing each arm's own reported closure time by its own protocol interval and adding the index operation gives 3.03, 3.50 and 6.85 against reported means of 3.0, 3.5 and 6.2 - an exact match to two significant figures in both NPWT arms. The endpoint is a deterministic function of closure time and a co-intervention frequency the protocol set, not an independent outcome. The Discussion concedes the mechanism and then asserts the opposite - that the difference 'reflects the debridement frequency and also the superior effectiveness of this method' - with no analysis separating the two. Compounding this, the NPWTi-d versus NPWT comparison is p = 0.002 by the printed rank test and p = 0.315 by A t-TEST on the paper's own means and standard deviations (mean difference -0.50, 95% CI -1.49 to +0.49). Net very low. For the harm signals, -2 for imprecision and selective reporting: graft loss 20.5% against 4.9% reaches Fisher p = 0.045 whole-arm and 0.035 per graft actually performed on 2x2 tests the paper does not perform and reports only as a non-significant three-group p of 0.109; amputation 15.4% against 7.3%; and reconstructed mortality among all randomised of 6/44 (13.6%) against 1/47 (2.1%), Fisher p = 0.054, risk difference +11.5% (+0.6 to +22.5). None of this appears in the abstract or the conclusion. Net Low - a signal to hold, not a finding to act on.

Figures: Figures checked

• Complete wound closure: OR 2.006 (95% CI 1.315–3.058), P = .023, favouring NPWTi-d.
• Complications: OR 0.421 (95% CI 0.260–0.683), P = .025, favouring NPWTi-d.
• Number of surgical debridements: no significant difference (P = .2146, 95% CI 0.4–1.3).
• Bacterial burden: 5 of 6 studies reporting bioburden or bacterial count favoured NPWTi-d.
• Descriptive (not pooled): mean therapy length 10.3 days with NPWTi-d versus 20.9 days with NPWT; mean hospital stay 21.3 versus 30.5 days. Length of stay was formally investigated in a single study only.
• Cost-effectiveness: the authors state that the literature provides no evidence of the cost-effectiveness of NPWTi-d, and that their own cost data rest on one study.
• The authors conclude that limited study quality means randomised trials are still needed to confirm the benefit and define recommendations.

CAT: OCEBM 3 · Three outcomes were pooled and the headline is not pooled on thirteen studies. Complete wound closure OR 2.006 (95% CI 1.315 to 3.058, P = .023) comes from six of the thirteen studies, stated in the results as 'out of the 13 studies reviewed, six studies were suitable for the meta-analysis regarding the rate of complete wound closure'. The number of patients behind those six is never printed. Complications OR 0.421 (95% CI 0.260 to 0.683, P = .025) is reported without any study count at all. Surgical debridements show no difference — And are reported with a p value and a confidence interval (p = .2146, 95% CI 0.4 to 1.3) but no point estimate. 871 patients is the review total, it reconciles internally (393 + 478 = 871; 511 males + 317 females + 43 unknown = 871), and it is not the denominator of any pooled estimate. Heterogeneity is specified as a method and then never reported. The statistical section states that Cochran's Q and I-squared were computed and that I-squared of 25 per cent or more would trigger a random-effects model — Yet no i-squared value appears anywhere in the results for any of the three pooled outcomes, and tau-squared is never mentioned. Heterogeneity is therefore neither explained nor even disclosed. A unit-of-analysis error is declared and not corrected: one study contributing two NPWTi-d dwell-time arms was 'considered twice', which counts its shared control arm twice in the pooled denominators. GRADE begins Low for a body of predominantly non-randomised comparative evidence and falls to Very low for risk of bias, which the authors themselves grade as limited, and for the undisclosed heterogeneity. The abstract's claim that the meta-analysis 'proves that NPWTi-d is superior' is a causal statement that eleven non-randomised studies cannot carry.

Figures: Figures checked

• NPWTi-d required significantly fewer surgeries / dressing changes than standard NPWT (P<0.05).
• Significantly greater wound-area reduction after treatment with NPWTi-d (P<0.05).
• No significant difference in healing rate, time to heal, length of stay, dehiscence, reinfection, reoperation or readmission.

CAT: OCEBM 1 · Ten outcomes, each with its own denominator, and not one of them is 564. Number of surgeries and dressing changes MD minus 0.43 (95% CI minus 0.80 to minus 0.05) on 5 RCTs and 404 patients. Wound area after treatment MD minus 9.31 (95% CI minus 17.54 to minus 1.08) on only 2 RCTs and 198 PATIENTS. Time to heal 3 RCTs, 251 patients, null. Length of stay 3 RCTs, 198 patients, null. Healed at end of study RR 1.14 (0.94 to 1.39) on 4 RCTs, null. Dehiscence RR 0.82 (0.65 to 1.03) on 4 RCTs and 385 patients, null. Reinfection RR 1.04 (0.61 to 1.78), reoperation RR 0.79 (0.48 to 1.29) and readmission RR 0.77 (0.54 to 1.10), each on 3 RCTs and 353 patients, all null. Pseudomonas at end RR 1.00 (0.22 to 4.47) on 2 RCTs and 68 patients, null. Eight of the ten outcomes are null. The abstract misreports its own primary result. It prints the surgeries and dressing-changes result as 'RR and 95% CI, minus 9.31 [minus 17.54, minus 1.08]' — Which is the wound-area mean difference pasted into the wrong row — and labels a mean difference as a risk ratio. The results section gives the correct figure, minus 0.43. Anyone quoting this paper from its abstract will quote a number twenty-one times too large. A FURTHER DISCREPANCY: the healed-at-end outcome is 294 patients in Table 2 and 394 in the results text. Both significant results are fragile by the authors' own testing. Heterogeneity for the surgeries outcome is significant (P for heterogeneity below 0.05), and omitting any single one of four trials — Davis 2020, Gonzalez 2020, Jurkovic 2019 or Lavery 2020 — abolishes its significance, with I-squared running 68 to 84 per cent in those runs. The authors write that 'there was the possibility that the result may be a false positive'. Tau-squared is never reported. Downgrade from High for risk of bias (blinding impossible and rated high risk across the pool) and for inconsistency: Low.

Figures: Figures checked

• Fewer operative visits with NPWTi: 2.4±0.9 (6-min) and 2.6±0.9 (20-min) vs 3.0±0.9 for standard NPWT (p≤0.05) (p.709, Table 3).
• Shorter hospital stay for the 20-min group: 11.4±5.1 vs 14.92±9.23 days (p=0.034) (p.709, p.711).
• Shorter time to final surgical procedure for both NPWTi groups (7.8±5.2 and 7.5±3.1 days) vs 9.23±5.2 days (p≤0.05) (p.709).
• Higher percentage of wounds closed before discharge with 6-min NPWTi: 94% vs 62% (p=0.0004) (Table 3, p.712).
• Culture improvement (Gram-negatives, Corynebacterium and yeast excluded) significantly greater with 6-min NPWTi: 90% vs 63% (p=0.0001); overall culture improvement not significantly different (p.712-713).
• 30-day post-discharge closure rates were statistically similar between groups, arguing against premature-closure bias (p.714).

CAT: OCEBM 4 · The resource-use results are real and modest; the culture result is null on the analysis the paper pre-specified, and this library has not been carrying that. Every estimate, with denominators. Operative visits 3.0 (SD 0.9) for standard NPWT against 2.4 (SD 0.9) at 6-minute dwell (P = 0.04) and 2.6 (SD 0.9) at 20-minute dwell (P = 0.003). Hospital stay 14.92 (SD 9.2) days against 11.9 (SD 7.8) — P = 0.10, not SIGNIFICANT — and 11.4 (SD 5.1), P = 0.03. Time to final surgical procedure 9.23 (SD 5.2) against 7.8 (SD 5.2), P = 0.04, and 7.5 (SD 3.1), P = 0.002. Wounds closed before discharge 46/74 (62%) against 32/34 (94%), P = 0.0004, and 27/34 (80%), P = 0.08, not significant. The durability of that closure is the finding the abstract does not carry. Remained closed at one month: 28/46 (61%) after standard NPWT, 24/32 (75%) at 6-minute dwell (P = 0.23) and 14/27 (52%) at 20-minute dwell (P = 0.47). On the whole randomised-equivalent denominator that is 28/74 (38%), 24/34 (71%) and 14/34 (41%) — The 20-MINUTE arm, the arm with the shorter stay, ended the month with 41% of its patients closed against 38% for standard NPWT. The bioburden result is null as pre-specified. Overall culture improvement 28/74 (38%) against 20/34 (59%), P = 0.06, and 17/34 (50%), P = 0.30 — Neither instillation arm improved cultures significantly. A difference appears only after excluding gram-negative organisms, Corynebacterium and yeast, which cuts the denominators to 27, 21 and 20, and only in the 6-minute arm: 17/27 (63%) against 19/21 (90%), P = 0.0001, while the 20-minute arm was 13/20 (65%) against 63%, P = 0.77. Cultures were QUALITATIVE swabs graded from heavy to scant growth. Loss to follow-up is not stated and cannot be recovered: the one-month outcome exists only for the 105 patients closed before discharge, so 37 of 142 contribute nothing to it. Disclosure, verbatim: 'Dr. Kim is a consultant for Kinetic Concepts, Inc. Dr. Attinger is an uncompensated consultant for Kinetic Concepts, Inc. Dr. Lavery has received past research funding from Kinetic Concepts, Inc. The other authors have no financial interest to declare in relation to the content of this article.' No funding statement appears.

Figures: Figures checked

• Solution. Normal saline is the recommended first-choice instillation solution for the majority of wounds; instillation of saline achieved outcomes comparable to antiseptics in the reviewed evidence. Several panel members preferred an antiseptic (hypochlorous acid or sodium hypochlorite) as the initial solution where there are clinical signs of infection, stepping down to saline after 24–48 hours. 0.04% polyhexanide should not be instilled directly on cartilage — the restriction is cartilage-specific and conditional: the source says "unless it has been strongly diluted (0.005%) to prevent damage to the cartilage". (Corrected 2026-07-29: this page previously generalised the restriction to "exposed structures" and omitted the dilution proviso. The panel's separate contraindication list covers exposed unprotected organs and vessels, undrained abscesses and split-thickness grafts — those are different statements.)
• Not recommended. The panel recommended against NPWTi-d over skin grafts or dermal substitutes (the instilled fluid disrupts the bolster function), where an intact seal cannot be maintained, and — for most panel members — in acutely ischaemic wounds because of the underlying perfusion deficit.
• Panel members also advised against instillation of topical solutions not compatible with NPWTi-d foam dressings and disposables; use manufacturer-specified dwell times and concentrations.
• Recommendations are framed as a refinement of, not a departure from, the original NPWTi-d parameter set.

AGREE II: Recommended with modifications — Scope 67% · Stakeholders 28% · Rigour 42% · Clarity 94% · Applicability 46% · Independence 25%

Figures: Figures checked

• Mean NPWTi-d duration 13 days; mean total hospitalisation 51 days.
• Fully decontaminated wounds (no organisms detected) in 23% of cases after NPWTi-d.
• Number of different bacterial species fell significantly from 2.38 to 1.16; bacterial count fell from 3.9 to 1.3.
• The reduction held across all subgroups except surgical wounds, in which NPWTi-d did not significantly reduce bacterial count.
• Significant stimulation of granulation tissue; successful reconstruction achieved in 90% of cases.
• The authors state plainly that full justification for NPWTi-d in the literature is still lacking despite the increasing evidence.

CAT: OCEBM 4 · The headline bioburden numbers do not reconcile with the paper's own subgroup means, and the bacterial count is not a count. The Methods define bacterial count as an ORDINAL RANK assigned by microbiologists: 1 few, 2 moderate, 3 several, 4 numerous. So the reported fall from 3.9 to 1.3 is a change in mean rank on a four-point scale, not a change in colony-forming units, and it is analysed by a signed-rank test. Now the arithmetic. The three wound-type subgroups are mutually exclusive and exhaustive: 12 traumatic, 10 surgical, 8 chronic, totalling 30. Number of different bacteria before therapy is 2.18, 2.24 and 2.2 in the three subgroups — Every one below the reported total of 2.38, which is impossible for a weighted average of a partition. The weighted mean is 2.21 [derived: 2.18 + 2.24 + 2.2, weighted 12/10/8 over 30]. After therapy the subgroups weight to 1.00 [derived: 0.77 + 1.15 + 1.17, weighted 12/10/8 over 30] against a printed 1.16. Bacterial count before weights to 3.34 [derived: 2.6 + 4.1 + 3.5, weighted 12/10/8 over 30] against a printed 3.9; after, to 1.55 [derived: 0.9 + 2.2 + 1.7, weighted 12/10/8 over 30] against a printed 1.3. And the surgical-wound baseline count is printed as 4.1 on a scale whose highest category is 4. The traumatic subgroup's post-therapy BACTERIAL COUNT of 0.9 (from 2.6) also falls below the scale minimum of 1, which is possible only if an undefined zero category was used. CORRECTION 2026-08-05, made once the OCRed text became readable: an earlier version of this note also cited 0.77 as breaching the scale minimum. It does not. 0.77 is the traumatic subgroup's post-therapy number of different bacteria (from 2.18) — a count of species, whose mean may take any positive value — not a bacterial-count rank. The scale-minimum objection applies to the ordinal BC measure only, and 0.9 is the single value that breaches it. The subgroup that matters most failed: in surgical wounds the bacterial count fell from 4.1 to 2.2 with P > .05, the largest absolute fall in the paper and the only one that is not significant. Complete decontamination was achieved in 7/30 (23%) — traumatic 4/12 (33%), surgical 2/10 (20%), chronic 1/8 (12.5%) — and that row does reconcile. One postoperative wound showed an INCREASE in the number of different species during therapy. Successful reconstruction 27/30 (90%) is the proportion who received a reconstructive procedure, not the proportion who healed; the three failures were inoperable and went to outpatient care. No healing outcome, no follow-up period and no loss-to-follow-up figure appears anywhere in the paper. Wound size is printed as a mean of 179 cm2 with a range of 16 to 1600 in the text and 6 to 1600 in Table 2, and the three subgroup means weight to 152.7 [derived: 134 + 179 + 148, weighted 12/10/8 over 30], not 179. Hospital stay is 51 days in the abstract and 52.6 in the Results and Table 4. Conflict of interest, verbatim: 'This is an investigator-initiated study. The V. A. C. VerFlo dressings for NPWTi-d were provided free of charge.' The supplier is not named. Funding, verbatim: 'Open access funding enabled and organized by Projekt DEAL', added by a correction on 7 November 2020.

Figures: Figures checked