Open fractures

Negative numbers favour the standard dressing. The sign convention is the paper's: NPWT minus control, so a positive difference favours NPWT.

CAT: OCEBM 1 · Randomisation and concealment: strong — computer-generated, trials-unit delivered, stratified by centre and Gustilo grade, assigned intraoperatively. BLINDING: patients and surgeons could not be blinded and the primary outcome is self-reported disability; this is the main source of concern, but its direction is unknowable and the trial's own answer is that the unblinded patients reported no benefit from the visible device, which is not the direction an expectation effect would push. POST-RANDOMISATION CONSENT: 625 randomised, 460 analysed. The authors report that most non-consenters were found ineligible after randomisation (primary closure, permanent cognitive impairment) and that only 29 eligible patients declined; the consented arms are balanced on Table 1. RoB 2 some concerns overall. ATTRITION: 88% (374/427) primary outcome completion, with mixed-effects modelling of the repeated measures. Precision on the primary outcome is the point: 460 patients powered at 90% for an 8-point MCID; the adjusted interval of minus 8.9 to plus 1.2 excludes an 8-point benefit at its upper bound. That is HIGH certainty that a clinically important benefit does not exist. The lower bound does not exclude an 8-point harm; the trial cannot rule that out and the authors say so. DEEP INFECTION is imprecise for a modest effect — 35 events in total, OR 0.85 (0.42-1.70), rate difference 1.0% (minus 4.2 to plus 6.3) — so MODERATE: a large effect is excluded, a small one is not. Healing at 6 weeks (OR 1.0, 0.6-1.6) and UNION (OR 1.1, 0.7-1.9) are flat and adequately precise for the sizes of effect anyone claims. GENERALISABILITY: 85% Gustilo III, 82% tibia, 24 UK major trauma centres with joint orthoplastic services and NPWT as delivered at clinician discretion — this is the population and setting in which NPWT for open fractures is actually used, which is why the result transfers. INDIRECTNESS to a South African trauma service is modest: the intervention is the same, the comparator (a standard dressing between debridements) is the same, and the ortho-plastic pathway may be slower here, which would if anything favour a bridging dressing and did not appear in the UK data. RETRACTION SCREEN: PubMed record for PMID 29896626 checked 2026-09-14 — no Retracted Publication and no Expression of Concern.

Figures: Figures checked

Costs by period (complete cases, unadjusted): the initial inpatient stay was £1,223 higher with NPWT (bootstrap CI 211 to 2,364; p 0.030) — the only significant difference, and it includes the dressings. Every later period and category was flat. Twelve-month total £14,079 vs £14,002; difference £77 (−2,114 to 2,925), p 0.95.

CAT: OCEBM 2 · Drummond / CHEERS. Question and perspective: well defined — NHS and PSS base case, societal in sensitivity, 12-month horizon matching the trial. EFFECTIVENESS: from a large concealed-allocation randomised trial (costa-2018), the appropriate source. COSTS: resource use collected prospectively, unit costs from national tariffs with sources tabulated, dressing and pump costs included in the inpatient-stay category. UTILITIES: EQ-5D-3L, the NICE reference case, with SF-6D as a check — and the SF-6D analysis makes NPWT DOMINATED (higher cost, lower QALYs), so the two instruments disagree in magnitude though not in verdict. UNCERTAINTY: bootstrap intervals, acceptability curves at three thresholds, net monetary benefit — thorough. MISSING DATA: this is the weak point. Complete economic profiles for 144 of 460 (31 percent); multiple imputation as base case is the recommended method and the authors cite the guidance, but the complete-case result is dominant in the opposite direction with probability 71 to 74 percent, and the abstract's claim of robustness omits it. A result that changes sign with the missing-data method carries genuine uncertainty that the ICER does not convey. THE ICER ITSELF: GBP 267,910 is the ratio of GBP 678 (CI spanning zero) to 0.002 QALY (CI spanning zero). It is not wrong, but its apparent precision is an artefact of division; quote the probability of cost-effectiveness (never above 27 percent), not the ratio. SUBGROUP: pre-specified subgroup analysis consistent with the base case (Gustilo grade); no subgroup favoured NPWT. Transferability to south Africa: UK 2014-15 NHS prices, UK thresholds, UK ortho-plastic pathway; the numbers do not transfer, the direction does, and the one significant cost driver — a longer and costlier initial admission with NPWT — is plausible anywhere. RETRACTION SCREEN: PubMed record for PMID 31674241 checked 2026-09-14 — no Retracted Publication and no Expression of Concern.

Figures: Figures checked

Open fracture wounds (four trials, all NPWT 125 mmHg vs standard care)

• Healing at six weeks: RR 1.01 (95% CI 0.81–1.27) — one trial, 460 participants. Moderate certainty. No difference.
• Wound infection: RR 0.48 (95% CI 0.20–1.13), I² 56% — four trials, 596 participants. Very low certainty (downgraded for risk of bias, inconsistency and imprecision). Uncertain.
• Health-related quality of life: probably no clear difference — EQ-5D utility MD −0.01 (−0.08 to 0.06), 364 participants, moderate certainty.
• Cost-effectiveness: moderate-certainty evidence that NPWT is UNLIKELY to be cost-effective for open fractures in the UK. NPWT was on average more costly and conferred few additional QALYs; the incremental cost-effectiveness ratio was £267,910, and NPWT was unlikely to be cost-effective across a range of cost-per-QALY thresholds.

Other open traumatic wounds (two trials)

• Wound infection, 125 mmHg vs standard care: RR 0.61 (95% CI 0.31–1.18), 509 participants, low certainty.
• 75 mmHg vs standard care: RR 0.44 (95% CI 0.17–1.10), 463 participants — uncertain.
• 75 mmHg vs 125 mmHg: RR 1.04 (95% CI 0.31–3.51), 251 participants — uncertain, and the point estimate is flat.

Authors' conclusion: moderate-certainty evidence of no clear difference in healing at six weeks for open fractures, and moderate-certainty evidence that NPWT is not cost-effective for them. Everything else — infection, adverse events, time to closure or coverage surgery, pain, quality of life — is uncertain.

CAT: OCEBM 1 · Per-outcome GRADE is adopted from the review, which applies it properly. The findings rated MODERATE are both NULL or negative: no clear difference in healing at six weeks for open fractures (RR 1.01, 0.81–1.27, downgraded once for imprecision), no clear difference in health-related quality of life, and — the one that matters commercially — moderate-certainty evidence that NPWT is NOT cost-effective for open fractures in the UK, with an ICER of GBP 267,910. The infection findings are much weaker: VERY LOW for open fractures (RR 0.48, 0.20–1.13, I² 56%, downgraded for risk of bias, inconsistency and imprecision) and LOW for other open traumatic wounds (RR 0.61, 0.31–1.18). The head-to-head 75 vs 125 mmHg comparison (RR 1.04, 0.31–3.51) is downgraded for risk of bias and imprecision and establishes nothing about equivalence. Page rated MODERATE on its primary healing and economic outcomes. No upgrade available; the ceiling is set by seven trials in which one 460-participant study dominates.

Figures: Figures checked

"The evidence showed that use of negative pressure wound therapy (NPWT) after wound excision if immediate definitive soft tissue cover has not been performed is not more clinically effective than other dressings and is unlikely to be cost effective. However, the committee agreed that most of the evidence related to the use of NPWT as a therapy rather than as a temporary dressing… and therefore most of the long-term outcomes were not relevant to the UK context… They agreed that, on that basis, it was not appropriate to make a 'do not routinely use' recommendation."

"NPWT is not the only choice."

CAT: Clinical practice guideline — AGREE II applied in outline (rigour high: NICE method, systematic evidence review, health-economic modelling, public consultation; applicability: NHS). GRADE not applicable to the document; the committee's own finding is that NPWT is not more effective than other dressings · A guideline is appraised for its process and read for what it says; this one says, in the committee's own words, that the evidence showed NPWT after wound excision 'is not more clinically effective than other dressings and is unlikely to be cost effective'. That is the WOLLF result (Costa 2018) and its economic evaluation (Petrou 2019) read by a national committee. The committee then declined to issue a 'do not routinely use' because the trials tested NPWT as a therapy with long-term outcomes, whereas UK practice uses it as a temporary dressing for up to 72 hours between excision and cover, and because the cost-effectiveness of that short use was unclear. Recommendation 1.2.31 is therefore dressing-agnostic — avoid desiccation, minimise changes — and the rationale states that NPWT 'is not the only choice'. The committee's experience-based points (fewer changes, fewer bedding changes, less pain at changes, useful for heavy exudate) are stated as committee opinion and are followed by a research recommendation asking which temporary dressing, 'including negative pressure dressings', is most clinically and cost effective for those very outcomes — which is the committee saying the question is open. PROCESS: NICE method, systematic review, economic modelling, consultation — rigour high. The timing recommendations (2016) are the ones the wiki's open-fracture page should carry beside Liu 2012 and Haidari 2021: wound excision within 12 hours for high-energy and 24 hours for other open fractures; fixation and definitive cover at the same sitting where the orthoplastic list allows, otherwise within 72 hours of injury; definitive cover at the same time as internal fixation. For a South African motivation, NG37 is the document a scheme will find, and it must be cited for what it says: NPWT is one acceptable temporary dressing for up to 72 hours, with no demonstrated advantage over others.

Figures: Figures checked

• Total deep infection: 7 of 25 (28.0%) control versus 2 of 37 (5.4%) NPWT, P = 0.024. Relative risk 0.199 (95% CI 0.045–0.874).
• The composite is doing the work. Control had 2 acute (8.0%) and 5 delayed (20.0%) infections; NPWT had 0 acute and 2 delayed (5.4%). There was no significant difference between the groups when acute and delayed infections were evaluated separately. Quote the total, and say it is the total.
• Delayed infections in the control group appeared at a mean of 11 weeks after discharge — a reminder that a trial following patients only to closure would have missed most of them.
• All 9 patients who developed deep infection had type III open fractures.
• Positive quantitative tissue cultures in 8 fractures (13%) — 5 (20%) control, 3 (8%) NPWT — but only 3 of those 8 developed clinical infection, which the authors read as confirming the value of secondary debridement.
• Tibial subgroup: 24 severe open tibial fractures; infection 36% control versus 8% NPWT.
• Hospital stay 11.7 days control versus 9.5 days NPWT.
• Two delayed amputations for unsalvageable soft-tissue loss; neither patient was infected.
• SF-36: no difference in the mental component at any timepoint; no difference in physical component among non-infected patients; significant physical-component benefit among infected patients at 3, 6 and 9 months.

CAT: OCEBM 2 · Start HIGH — an individually randomised parallel-group trial. MINUS 1 for risk of bias: allocation concealment is never described, the outcome assessors are the treating surgeons and are unblinded, one randomised patient was removed from analysis after allocation, and the trial was paid for by a grant from the manufacturer of the device to the first author. Minus 1 for very serious imprecision: Nine infection events in total, the 95 per cent confidence interval on the relative risk runs from 0.045 to 0.874 — a range spanning a twenty-two-fold reduction to a barely detectable one — and recruitment stopped at 59 patients against a stated design target of 100 after five years. MINUS 1 for selective reporting, and this is the finding the wiki had not previously recorded: Wound dehiscence is named as a primary outcome measure in the abstract and methods, no dehiscence figure appears anywhere in the results, and the discussion nonetheless asserts that the data support decreased wound dehiscence. A second reporting defect points the same way: the sample-size calculation is built on a completely different endpoint — the proportion of patients needing three or more debridements — which the results then report as null. Net VERY LOW. The direction of the infection effect is plausible and consistent with the closed-incision literature; the magnitude is not quotable.

Figures: Figures checked