Skin grafts and graft-bed preparation

The −80 mmHg subgroup is the most valuable line in this paper for a wiki that keeps pressures on protocol pages. Pooling only the trials that used −80 mmHg gives a larger effect than the overall estimate and, unusually, eliminates the heterogeneity entirely (I² 85% → 0%). Four of the six outcomes above have I² of 0%, and the two that do not — overall graft take and complications — are the two that pool across mixed pressures and mixed definitions.

CAT: OCEBM 1 · Start HIGH (systematic review of randomised trials, prospectively registered, no funding and no declared conflicts). −1 risk of bias: 16 trials averaging about 50 patients each, and a graft bolster cannot be blinded, so every trial carries unblinded assessment of an outcome — graft take — that is judged clinically. −1 inconsistency on the PRIMARY outcome: I² is 85% for overall graft take, reflecting mixed pressures, graft types, donor sites and definitions of take. Net low. Important per-outcome split: the −80 mmHg subgroup for graft take has I² of 0% and a LARGER effect (+10.0%, 95% CI 5.69–14.34), and graft success, graft loss and reoperation all have I² of 0% — for those four the inconsistency downgrade does not apply and the certainty is MODERATE. The complications estimate is the weakest (OR 0.36 with an upper bound of 0.99, I² 76%) and should be treated as Very low. No downgrade for indirectness — this is exactly the intervention and population — and none for imprecision on the outcomes whose intervals are tight.

Figures: Figures checked

Sixty patients, thirty per arm, no drop-outs, drop-ins, protocol violations or refusals, analysed by intention to treat.

The only between-group estimate with a confidence interval in the whole paper is the adjusted one: in a stepwise multivariate linear model containing total grafted area, days from injury to surgery, sex, age and aetiology, the coefficient for negative pressure was −6.9 cm2 (95% CI −10.8 to −3.0), p = 0.001; the same model found grafted area itself predicted loss, coefficient 0.2 (95% CI 0.1 to 0.3), p < 0.001. Length of stay held its significance after the same adjustment.

The mechanistic finding is the most useful result and is not the primary outcome. In controls, graft loss scaled with graft size — slope 0.41 (95% CI 0.22 to 0.57), R2 = 0.50, p < 0.001. Under negative pressure that relationship vanished — slope 0.01 (95% CI −0.02 to 0.05), R2 = 0.03, p = 0.387. The bigger the graft, the more the bolster was worth. Losses were distributed evenly between centre and edge, and no patient in either arm had a recorded complication.

Baseline (Table 1) was balanced on every tested variable, including grafted area at 33.8 cm2 (8.8-124.3) versus 31.2 cm2 (5.5-179.7), p = 0.779. The one imbalance worth naming is mechanism — burns 17/30 (56.7%) against 23/30 (76.7%), p = 0.170 — a twenty-point gap that is non-significant only because thirty per arm cannot detect it. Aetiology was carried into the adjusted model, which is the right response to it.

CAT: OCEBM 2 · Start HIGH (randomised trial), assessed per outcome. Graft area loss: no downgrade for risk of bias — this is the best-appraised trial in the entire NPWT secondary-intention pool and the reason is a real design feature rather than a reporting flourish, namely that the control arm received the identical occlusive foam bolster and differed only in whether the tube was connected, which isolates the negative pressure from the dressing in a way almost no other NPWT trial achieves. No downgrade for inconsistency (single trial). No downgrade for publication bias — there is no device-manufacturer relationship of any kind, which is close to unique in this literature, although it must be recorded alongside that no funding source is declared at all. Minus 1 for indirectness, and this is the load-bearing judgment: the endpoint is area of graft loss at day 4, an early surrogate that is not healing, not time to closure and not durability, measured in SMALL GRAFTS (median 33.8 and 31.2 cm2) on wound beds pre-qualified by quantitative biopsy at under 100,000 CFU/g, in a young, 83% male, working-age occupational-injury population with major burns, polytrauma and hypoalbuminaemia excluded by protocol. Imprecision was considered for a second downgrade and NOT applied: n = 60 does not meet any optimal information size, but the single adjusted between-group estimate the paper reports, −6.9 cm2 (95% CI −10.8 to −3.0), has both bounds on the same side of no effect and both are clinically meaningful, and the effect is corroborated by a coherent dose relationship — graft loss scaled with graft size in controls (slope 0.41, R2 0.50) and that relationship was abolished under negative pressure (slope 0.01, R2 0.03). Net moderate. Length of stay: minus 1 further, to LOW, because the abstract reports p less than 0.001 for 13.5 versus 17 days where the results give p = 0.010 UNADJUSTED, the sub-0.001 figure being the multivariate-adjusted one imported without label; and because two different stay outcomes exist in the paper (8 versus 12 days to discharge from the unit in Table 2; 13.5 versus 17 days total stay in the text only) and are readily conflated. Length of stay in a Chilean occupational-injury referral hospital that discharges only on full healing is in any case a health-system variable, not a wound variable, and does not transfer to a South African setting.

Figures: Figures checked

Test for subgroup difference by design, graft take: p = 0.58.

CAT: OCEBM 2 (SR of RCTs and observational studies). GRADE: graft take RCT-only MODERATE; infection, seroma and reoperation LOW (mixed-design pools); haematoma LOW (null). AMSTAR-2 LOW. · START HIGH for the RCT-only graft-take analysis. −1 risk of bias: a graft bolster cannot be blinded and take is judged clinically. No further downgrade for inconsistency on the UNSTANDARDISED RCT figure, because it converges with an independent RCT-only review (lee-2025: +8.3%, 16 RCTs) — two reviews with different inclusion reaching the same number is the strongest form this evidence takes. Moderate. The SMD headline (1.34, i² 96%) should not be quoted — its heterogeneity and its misprinted unit make it uninterpretable. The four dichotomous outcomes START LOW because randomised and non-randomised studies are pooled in a single estimate without a design split; their direction agrees with lee-2025 (reoperation OR 0.31 there, 0.29 here), which supports direction but does not raise certainty. The review contributes one thing lee-2025 does not: infection (OR 0.37, 1,325 patients) and seroma (OR 0.38, 436 patients) as named outcomes.

Figures: Figures checked

• Graft incorporation (take) was >95% in all patients at all pressure settings and at all time points (abstract; p.2 Table 1).
• The study challenges the assumed 125 mmHg standard: pressures as low as 50 mmHg were tolerated without compromising STSG incorporation (abstract; conclusion, p.6).
• Average wound size was 151 cm² (range 28–561 cm²); most common pathologies trauma 44%, oncology 23%, infection 12.5% (p.2).
• No significant complications occurred; one post-radiation wound showed hypertrophic scarring after 50 mmHg, and one repeat-graft patient (diabetes/PVD) was treated at 125 mmHg (p.3–4).
• Authors report less pain and less peripheral skin reaction at lower pressures, and note the physiologic rationale that any pressure above ~25 mmHg capillary pressure should aid angiogenesis/epithelialisation (p.6).

CAT: OCEBM 4 for the pooled comparison (a randomised subset of 26 merged with a surgeon-selected retrospective subset of 22, reported only in combination). ROBINS-I applied to the pooled analysis; overall risk of bias critical. GRADE cannot be applied — there is no effect estimate of any kind in this paper to rate. No test, no p-value and no confidence interval is reported anywhere, by the authors' own statement. · GRADE is refused here for a specific and stated reason: there is nothing to GRADE. GRADE rates certainty in an estimate of effect, and this paper contains none. The authors write it out: 'Statistical significance was not calculated due to the small study population.' There is no comparison, no difference, no interval and no test statistic in the article. What it reports is a single descriptive claim — graft incorporation exceeded 95% in every patient, at every pressure setting, at every time point — which is not an effect estimate but a ceiling observation. RoB 2 is also the wrong instrument, because only 26 of the 48 patients were randomised and the results are reported only for the pooled set. The prospective 26 were allocated to 50, 75, 100 or 125 mmHg by computer algorithm; the retrospective 22 were bolstered at 75 or 125 mmHg at the operating surgeon's discretion, and pressure chosen by a surgeon is chosen for a reason. The paper pools them on the stated ground of 'similar demographic and wound profiles' and never reports the randomised subset separately, so the randomisation is analytically discarded. The descriptive claim also does not match the table it rests on. The abstract and the Results both state incorporation was 'greater than 95% in all study subjects, at all pressure settings', yet Table 1 prints group MEANS of 95% for the 125 mmHg arm at bolster removal and at 2 weeks. A mean of 95% cannot be produced by a set of values all above 95. Separately, the Methods define assessment at bolster removal, 2 weeks and 4 weeks, while Table 1's third row is headed 6 weeks. The low-pressure finding rests on two patients whose mean age was 71, who both carried myocardial infarction, stroke or peripheral vascular disease, and whose mean wound was 70 cm2 against 151 cm2 for the study as a whole. The authors are accurate about their own paper: it is 'a proof of concept rather than concrete clinical evidence'.

Figures: Figures checked

Two separate operations, two separate endpoints, the same answer: −80 mmHg beats −125 mmHg.

This vault has argued the low-pressure position from Evangelista 2013 (>95% take at 50–125 mmHg), Nakamura 2018 (50 mmHg) and Mowers 2026 (contraction rises with pressure).

Huang adds something none of those supply: a comparative clinical finding, in humans, at scale, from an independent guideline panel — that the higher setting is not merely unnecessary but does not work as well.

CAT: OCEBM 1 (national consensus guideline with GRADE appraisal and registered protocol) / GRADE not applicable to a guideline as a whole — AGREE II applied. The guideline's own per-recommendation Evidence Levels and Recommendation Strengths are reported verbatim below · GRADE is not the instrument for a guideline and must not be applied to it as a whole — AGREE II IS, and it is set out below. What can be reported is the guideline's own grading of each recommendation, which it states transparently. The recommendations this vault will actually use: vascular reconstruction and interventional procedures — prophylactic NPWT recommended to reduce inguinal SSI, evidence level II, recommendation strength b, covering endovascular aortic repair, thoracic endovascular aortic repair and lower extremity revascularisation, with the explicit limitation that against conventional dressings NPWT shows NO significant difference in readmission, 30-day mortality, antibiotic cost or length of stay. Skin graft and tendon OR bone exposure — artificial dermis with NPWT then split-thickness graft in two stages, evidence level i, strength a. Sternal — NPWT for sternal infection promotes healing, reduces mortality and shortens stay, LEVEL II STRENGTH B; prophylaxis in cardiac surgery level II strength b. Diabetic foot — recommended with strict pressure monitoring at 50 to 100 mmHg, level i strength a, and current evidence does not support irrigated over non-irrigated NPWT. Abdominal surgery — reduces SSI and seroma but NOT wound-healing problems, seroma formation or readmission, level II strength b. Pressure injuries — LEVEL III STRENGTH C, the weakest recommendation in the document and correctly graded so. The document's central contribution to this vault is the two convergent pressure comparisons favouring minus 80 over minus 125 mmHg, described in the integrity note.

Figures: Figures checked

• Four core mechanisms of action underpin every indication: (1) reduces tissue oedema; (2) increases granulation tissue formation; (3) increased perfusion as a secondary/later effect; (4) with instillation and dwell time can reduce wound contamination.
• Vascular surgery: high-level evidence supports NPWT for infected vessels and prosthetic vascular grafts — strong recommendation for high-risk surgical patients with a fully exposed, infected prosthetic vascular graft, together with debridement and appropriate antibiotics; continuous suction at lower levels (-50 to -100 mmHg) recommended to avoid bleeding.
• Traumatic/surgical wounds: NPWT is first-line for preoperative wound-bed preparation post-debridement, typically 7-10 days before definitive surgery; strong for open fractures that cannot be closed primarily (used between debridement and reconstruction); caution when tendon/bone exposed; do not use if underlying arterial impairment is not addressed (red).
• Diabetic foot ulcers: first-line post-surgical treatment for DFU once ischaemia is excluded (University of Texas Grade A2/A3); consider for chronic DFU not progressing after 4 weeks; contraindicated where ischaemia is due to concomitant peripheral arterial disease (red).
• Sternal dehiscence: NPWT is the method of choice for post-sternotomy mediastinitis (strong); caution in the haemodynamically unstable patient; the open sternum with exposed vulnerable structures is not an indication (red).
• Abdominal / dehisced abdominal wounds: strong for the open abdomen and for dehisced wounds with intact abdominal wall; NPWT with instillation (NPWTi-d) is effective in reducing infection in the infected open/dehisced abdomen (moderate); do not use if size/severity does not merit it.
• Skin grafts, burns, venous ulcers, pressure injuries, enterocutaneous fistulae: NPWT endorsed for wound-bed optimisation before grafting and intra-operative graft stabilisation; acute burn indications supported but not a substitute for grafting; venous ulcers only if <30% area reduction in 4 weeks and ABPI 0.8-1.2 (contraindicated if ABPI <0.6); NPIAP stage 3/4 pressure injuries until surgical closure (moderate); ECF in selected collapsible/low-output cases only, never in unexplored fistulae.
• Closed incisional NPWT (ciNPT): recommended to decrease wound complications, dehiscence, haematoma/seroma and surgical site infection (strong).
• Contraindications listed: clotting disorders, active non-capillary bleeding post-debridement, exposed organs/vasculature/anastomoses, eschar or dry necrosis, neoplastic tissue, untreated osteomyelitis, non-enteric and unexplored fistulae.

AGREE II: Recommended with modifications — Scope 67% · Stakeholders 56% · Rigour 23% · Clarity 67% · Applicability 46% · Independence 25%

Figures: Figures checked