Safety and harms

Cost-effectiveness (six economic studies, five indications including — new in pub7 — vascular surgery with inguinal incisions): results differ by indication. Moderate-certainty evidence that NPWT in surgery for lower limb fracture was not cost-effective at any willingness-to-pay threshold, and that it probably is cost-effective in obese women undergoing caesarean section. Other indications: low or very low certainty.

CAT: OCEBM 1 · The review applies GRADE itself and this appraisal adopts its per-outcome ratings rather than substituting a global one, because that is the whole point of a review that grades each outcome separately. SSI: Moderate — start High as a systematic review of randomised trials, −1 risk of bias because most included trials had unclear or high risk in at least one key domain; Not downgraded for inconsistency (I² 29%) or imprecision (RR 0.73, 95% CI 0.63–0.85 on 11,403 participants). Dehiscence: Moderate for the finding of no difference (RR 0.97, 0.82–1.16). Mortality, reoperation, seroma, skin blisters, pain: Low. Haematoma: Very low. The page-level rating of Moderate reflects the primary outcome. No upgrade is available and none is warranted: this is an unrestricted, independent, duplicate-assessed Cochrane review, and the ceiling on its certainty is set by the quality of the trials it pools, not by its own conduct.

Figures: Figures checked

Statistics: recomputed from the paper's own figures: 6 of 6 reproduce.

69 male active-duty personnel, mean age 33.4 ± 9.3. Mechanism mine-blast 71%, gunshot 23%. Vessels: brachial 33.3%, superficial femoral 23.2%, popliteal 11.5% — the remaining 32% of injured vessels are never enumerated. Combined arterial-venous injury 40%; fracture 42%; primary wound contamination 56.5%. 97 reconstructive procedures across 69 patients (autovenous grafting 43, end-to-end 14, PTFE 4, patch 8, ligation 10, temporary shunt 11, thrombectomy 7) — so the number of patients carrying an interposition graft versus a primary repair under the dressing is not recoverable.

Mean length of stay 29.1 ± 11.2 days, median 26.

CAT: OCEBM 4 · Start Low (single-arm observational series). -1 risk of bias, and the reasons are structural: Selection is on the exposure - only patients who received NPWT were analysed, so the series cannot separate the effect of NPWT from the effect of being judged suitable for it; the 16 screened-but-not-included patients are never characterised and there is no flow diagram; Primary medical records were missing in more than 50% of cases, on the authors' own statement; the authors were the treating surgeons with no blinding and no independent adjudication; and ethics approval was obtained on 26 May 2025, approximately 2.5 years after the last patient was treated, so no prospective governance existed over data capture. -1 further for the absence of any prognostic-variable set: ischaemia time, tourniquet duration, time from injury to repair, FASCIOTOMY (the word appears nowhere in the paper), injury severity score (the authors state 'Formal injury severity scoring (ISS/NISS) could not be applied'), shock grade, transfusion volume, anticoagulation regimen and antibiotic regimen are all ABSENT, so no adjustment is possible and none was attempted; and because 97 reconstructive procedures are reported across 69 patients, the number of patients carrying an interposition graft versus a primary repair versus a patch under the dressing is not recoverable. Net very low. The decisive attribution problem, which is not a GRADE domain but governs how the page may be used: 5 of the 9 erosion bleeds and 5 of the 6 thromboses occurred at role II-III, where NPWT use is undocumented and where NPWT at Role IV was the inclusion criterion. The NPWT-attributable bleeding rate at the facility where the two-layer protocol was applied is therefore 4/69 (5.8%, Wilson 95% CI 2.3 to 14.0), not 13.0%. Separately, the two claimed risk windows of days 7-10 and 18-30 are a NARRATIVE READING of nine events summarised only by a median of 18 and a range of 6 to 29: there is no histogram, no Kaplan-Meier, no swimmer plot and no event table anywhere in the paper, the earliest event at day 6 falls outside both windows, and no event occurred after day 29 although the upper window extends to day 30.

Figures: Figures checked

• "the same surgical team should ideally perform all VAC revisions until the end of the therapy."
• "A count of the inserted foams should be made at each procedure (similar to the standard practice for swabs, needles, and instruments)."
• "Attempts should be made to use the largest single piece of foam in the wound bed. Whenever possible, the use of more than one piece of foam should be avoided."
• "If foams have to be trimmed ... the trimming itself should take place away from the wound and the patient to reduce the risk of fragments in the wound."
• "The precise number of the inserted foams and their localization(s) should be consistently documented in the operative note, so that even if another team should carry out the next revision surgery, all foams could be easily and reliably identified." … "it is the responsibility of the surgical team and not of the electronic device to provide a precise and errorless documentation of all used materials."
• "In some cases, the foams are so adherent to the surrounding soft tissues that local application of saline or sterile water is recommended to loosen the foam before removal is reattempted."
• "At foam removal, all dressing materials should be examined for integrity. Should any discrepancy be evident with regard to the number or size of the removed foams, the wound should be further explored to resolve this discrepancy."
• Take-home: "At the smallest doubt that all foams are fully removed, the wound should be further explored."

CAT: OCEBM 5 · This is a narrative review of case reports and there is no effect estimate to grade. What can be graded is the INCIDENCE CLAIM, and it does not survive. The abstract states 'initial studies have indicated rates of up to 10% in large collectives being treated by NPWT'. That phrase appears nowhere in the body. The body contains exactly two numbers: Caniano 2005 at 2 of 21 (9.5%) and Dessy 2015 at 11 of 345 (3.2%). Caniano is a paediatric pilonidal subgroup of a 51-child series whose own abstract reports the two retained sponges against the full 51 - 3.9% - without assigning them to that subgroup at all, so the review's attribution is unverifiable from the source. Pooling both denominators gives 13/366 = 3.6%; with Caniano's full denominator, 13/396 = 3.3%. The source fails the abstract's framing on three counts: the cohort is paediatric general and pilonidal surgery, not orthopaedic and not vascular; n = 21 is not a 'large collective', and the review's own body text says the opposite - 'Most identified studies are case reports with no exact statements about the incidence of this complication within a large collective'; and the plural 'initial studies' at that level rests on a single subgroup of a single 2005 retrospective series. The detection claim is assertion plus descriptive tabulation: no bench work, no phantom study, no diagnostic accuracy study and no sensitivity or specificity for any imaging modality appears anywhere; the only supporting text is that six of nine studies used imaging and that in all of them 'the particular imaging method was not indicative of the retained foams but rather of abscess formations or cavities'. No imaging modality is reported to work, and ultrasound is not mentioned once. The authors' own two cases make the point: MRI reported an abscess and the excised objects were a 16 x 4 x 2 cm and an 8 x 4 x 4 cm polyurethane foam. The risk factors are mostly imported OR asserted: emergency or unplanned surgery, unplanned change of procedure and high BMI are derived from a different population entirely - Gawande 2003, a malpractice-claims case-control study of retained sponges and instruments, not NPWT foam (emergency surgery RR 8.8, 2.4-31.9; unplanned change 4.1, 1.4-12.4; BMI per unit 1.1, 1.0-1.2), and no NPWT-specific quantification exists; change of surgical team and absence of standardised documentation are ASSERTED and are not coded variables in any case; the dressing-change dose-response is extrapolated from a single descriptive statistic with no denominator and no comparison group; and the polyurethane-versus-PVA fragmentation claim is bench chemistry, with the review's own words being that PU 'THEORETICALLY has a higher risk of fragmentation' and that 'no study has provided any information about the rate of fragmented foams in clinical practice'.

Figures: Figures checked

An 83-year-old woman, elective right total hip arthroplasty, no dermatological history, one abnormal preoperative value: a mild leukocytopenia.

At both operations the inflammation was confined to skin and subcutaneous fat. Fascia, muscle and the arthroplasty components were macroscopically normal. Cultures gave S. epidermidis in one of four samples after prolonged incubation and C. acnes in one during NPWT after extended incubation, both read as contaminants. The biopsy showed a dense dermal neutrophilic infiltrate with oedema and focal necrosis, no vasculitis, no granulomas and no organisms — a sterile neutrophilic dermatosis.

The withdrawal is recorded in one sentence:

"In light of this diagnosis, the VAC therapy was discontinued to prevent further mechanical trauma and potential exacerbation of the pathergy phenomenon."

And the rule the authors draw from it:

"Repeated debridements, component exchanges, and aggressive negative-pressure wound therapy may paradoxically worsen PG due to pathergy and should be avoided once PG is suspected."

The diagnosis of the skin disease found the systemic one. Bone marrow biopsy showed CD20-positive hairy cell leukaemia, explaining the chronic leukocytopenia. Ten days, two operations and four antibiotic regimens were spent before the first dermatological opinion.

CAT: OCEBM 5 (single case report) / GRADE not applicable — no effect estimate exists to grade, because a single patient generates no comparison. JBI Critical Appraisal Checklist for Case Reports applied instead, with CARE used as the reporting-completeness check. · GRADE cannot be applied and saying so is the point: there is one patient, no comparator and no estimate, so no certainty rating in an effect can be produced. The value of this paper is not an effect estimate. It is that it names a disease in which negative pressure wound therapy is contraindicated by mechanism, and this library has never named it. Pyoderma gangrenosum is a sterile neutrophilic dermatosis in which minor trauma provokes new and larger lesions — the pathergy phenomenon — so debridement and NPWT do not merely fail to help, they extend the ulcer. The paper's Clinical message states it in terms this library can adopt: repeated debridements, component exchanges and aggressive negative-pressure wound therapy 'may paradoxically worsen PG due to pathergy and should be avoided once PG is suspected', and the operative note records the decision as discontinuation 'to prevent further mechanical trauma and potential exacerbation of the pathergy phenomenon'. Every other harm held in this library is the therapy injuring a wound that would otherwise heal — Bleeding, vessel erosion, retained foam, adhesive and port injury. This one is the therapy feeding the disease. Two limits travel with it and must be stated wherever it is cited. First, the harm is inferred and not measured: the report's own CRP series FELL from 223 mg/L on day 10, before NPWT, to 151 mg/L on day 18 at the end of it, and no wound measurement, lesion count or pain score is reported for the interval the therapy was running; the authors write only that surgery and VAC 'likely contributed'. Second, withdrawal and corticosteroids are three days apart in a disease that responds to corticosteroids, so recovery is attributed to two interventions changed together. The case is therefore decisive as a diagnostic and stopping rule and weak as a demonstration of harm, and the contraindication rests on the disease mechanism rather than on this patient's course. Downgraded further for reporting: the NPWT episode carries no pressure setting, no foam type, no interface, no exchange interval and no wound dimensions, so nothing in it can reach a protocol page.

Figures: Figures checked

Comparisons 14 (short-term) and 15 (long-term). Single study throughout, so I-squared is undefined in every analysis and every forest plot prints "Heterogeneity: Not applicable". No sensitivity analysis, no subgroup analysis.

Absolute effect on SSI: 57 per 1,000 falling to 27 per 1,000 (10 to 77); risk difference 2.94 percentage points; NNT about 34 per incision. Adverse events were defined as intolerance of the unit or tubing, pain or discomfort, and disturbance from pump noise — NNH about 16 incisions.

Certainty was downgraded for risk of bias (blinding of participants, personnel and outcome assessment — nurses and physicians were unblinded during the inpatient period when some infections were diagnosed) and for imprecision (few events, wide intervals). It was never downgraded for inconsistency, because with one study there is nothing to be inconsistent with.

CAT: OCEBM 1 · This is the certainty the REVIEW assigns, and it is reproduced here rather than re-derived, because the review's GRADE process is documented and its reasons are stated. For the NPWT comparison the certainty is Low throughout, downgraded on two domains only: Risk of bias (Svensson-Bjork 2022 is High RISK for blinding of participants and personnel and High RISK for blinding of outcome assessment - nurses and physicians were unblinded during the postoperative inpatient period, which is when some infections were diagnosed) and IMPRECISION (few events, wide confidence intervals; for amputation the review notes explicitly that 'no events were reported in either group, and the sample size calculation did not account for the low incidence of postoperative amputation'). It is never downgraded for inconsistency, and the reason is decisive: There is only one trial, so I-squared is undefined in every analysis and every forest plot prints 'Heterogeneity: Not applicable'. For the device-related adverse-event outcome the certainty is Very low, downgraded for risk of bias and for imprecision on 12 events against zero. Two further considerations this library should carry but which do not change the review's own rating: the pooling is a NAIVE UNPAIRED 2x2 across 377 incisions from 209 patients, ignoring the paired structure of the 336 bilateral incisions, which the review acknowledges only as a departure from its protocol unit of analysis - the trial's own paired analysis returned a combined p of 0.49; and the single included trial declares research funds from Smith & Nephew, which also provided 100 PICO dressings, which is worth recording because it makes a null a stronger null rather than a weaker one.

Figures: Figures checked