Amputation stump wounds

Risk ratio for the standard dressing versus iNPWT: 3.2 (95% CI 1.33 to 7.65).

By closure timing: primary closure 5/28 (17.85%) versus 12/23 (52.17%), p = 0.0096; delayed closure 0/3 versus 4/8, p = 0.2364.

Below-knee converted to above-knee: 2/19 (10%) versus 2/14 (14%).

Organisms among infected stumps: Klebsiella pneumoniae 28.5%, E. coli 19%, Pseudomonas aeruginosa 14.3%.

The control-arm infection rate of 51% is the finding worth carrying — roughly three times the rate in the European stump trials the wiki holds, in a population with 50% wet gangrene and 40% diabetes.

CAT: OCEBM 2 · Start HIGH (randomised trial). -1 risk of bias: the trial is open-label with no blinding of participants, personnel or outcome assessors, and the primary outcome - surgical site infection assessed on postoperative day 4 - is a judgement made by clinicians who can see which arm the patient is in; the authors state plainly it is 'an open-label study, with no blinding'. Allocation concealment is adequate (sequentially numbered opaque sealed envelopes) and the sequence generation is adequate. -1 imprecision: 62 patients against a calculated requirement of 31 per group derived from an assumed reduction from 48% to 7%, an assumption taken from a review of retrospective series; the observed risk ratio of 3.2 carries a 95% confidence interval of 1.33 to 7.65, wide enough to include an effect a third of the point estimate. Not downgraded for indirectness OR inconsistency, and not downgraded for the baseline imbalance, because the imbalance runs AGAINST the intervention arm's favour in some respects and towards it in others and, decisively, the control arm carried more diabetes (48% against 32%), more hypertension, more tobacco chewing, more alcohol, more trauma (19% against 4%) and more delayed closures (26% against 10%) - a configuration that inflates the observed effect. No multivariable adjustment is performed anywhere, and the authors acknowledge the imbalance without addressing it: 'With these differences in both groups, the application of incisional NPWT reduced the incidence of SSIs in lower limb amputation stumps significantly.' Net low. The effect size is the thing to interrogate: SSI 16% against 51% is roughly three times the separation seen in the European stump trials this wiki holds, and a 51% control-arm infection rate is itself the most transferable number in the paper - it is the baseline risk of a public-sector amputation stump in a middle-income setting, and a large absolute risk reduction from a high baseline is a different funding argument from a small one from a low baseline.

Figures: Figures checked

Device: CCNPWT (Triage Meditech, New Delhi) — silicone contact layer, polyurethane foam, adhesive film, continuous 125 mmHg, applied immediately after skin closure. Duration is stated two ways: removed on the sixth postoperative day (Methods) and 5 days (flow diagram). The comparator is described three different ways — removed after 48 hours (Methods), once or twice daily (figure), twice a day or at least daily (discussion).

CAT: OCEBM 2 · Start HIGH (randomised). -1 risk of bias: the trial is unblinded and, critically, outcome assessment was unblinded for the first six postoperative days - the blinded wound-care specialists took over only from day 7, after the dressing had been removed, whereas surgical site infection by the CDC criteria the trial used can be diagnosed inside that window. Amputation level is reported for the whole cohort (30% above-knee, 70% below-knee) but not per arm and is absent from the baseline table, so the arms cannot be checked for balance on the strongest prognostic variable for stump healing; neither perfusion nor revascularisation is reported per arm. -1 imprecision: 50 patients, no effect estimate and no confidence interval anywhere in the paper, and four primary outcomes plus a secondary plus 25 further tests with no multiplicity adjustment. For infection, -1 further for the test-selection problem, which is decisive rather than pedantic: SSI (3/25 vs 9/25) is reported at p = 0.047 by UNCORRECTED CHI-SQUARE and is the only row in table 5 tested that way - the other three primary outcomes all use Fisher's exact, which on the same 3-versus-9 table returns p = 0.0955, with Yates-corrected chi-square at 0.098 and the relative-risk interval crossing one (RR 0.33, 95% CI 0.10-1.09). Expected counts are all above 5 so the choice is defensible, but it is outcome-specific and unexplained, and the trial's only significant primary outcome turns on it. Net VERY LOW for infection. For time to stump healing and prosthesis eligibility the picture is better: 5.12 ± 1.53 versus 6.8 ± 1.95 weeks, mean difference 1.68 weeks (95% CI 0.71 to 2.65 on my recomputation), p = 0.002 by both pooled and Welch t-test - a continuous outcome with a wide margin over the null. Net low. Note however that the outcome as measured is a COMPOSITE of stump healing AND a subjective eligibility judgement, made by an unstated assessor on unstated criteria in an unblinded trial, reported to two decimal places of a week from a weekly assessment schedule; and Table 2's whole-cohort value of 6.18 ± 2.18 weeks does not reconcile with the arm values, which imply 5.96 ± 1.93 and would require a control mean of 7.24 rather than 6.8.

Figures: Figures checked

• iNPWT was associated with significantly fewer postoperative wound complications: OR 0.52 (95% CI 0.30–0.89, P=0.02).
• 30-day mortality was not significantly reduced (OR 0.81, 95% CI 0.46–1.45, P=0.48).
• No significant difference in readmission or revision surgery between groups.
• Authors recommend iNPWT for surgical-site-infection prevention in major lower-limb amputation "whenever applicable".

CAT: OCEBM 3 · All three pooled studies are retrospective case-control series at OCEBM level III. GRADE therefore begins Low, not High, and the earlier entry on this page that read OCEBM 1 was wrong. Each outcome has a different denominator and none of them is 457 except the first. Overall postoperative complications OR 0.52 (95% CI 0.30 to 0.89, P = 0.02) is pooled on all THREE studies and 457 patients. Wound infection OR 0.24 (95% CI 0.07 to 0.78, P = 0.02) is pooled on TWO studies and 148 patients — Chang 3/23 versus 12/31 and Gantz 1/47 versus 4/47, which reproduce individual odds ratios of 0.238 and 0.234 and a pooled 0.24 exactly; Stenqvist recorded wound complications as not defined. 30-DAY mortality is pooled on two studies and 403 patients, 56 deaths, 12.4 per cent with iNPWT against 15.2 per cent with standard care, which also reproduces exactly (23/186 and 33/217). A motivation that writes 457 next to the infection odds ratio is overstating the denominator by a factor of three. Two internal defects. First, the results section prints 30-day mortality as 'OR = 0.48; 95% CI: 0.46-1.45' while the abstract prints OR 0.81 with the same interval; the geometric mean of 0.46 and 1.45 is 0.817, so 0.81 is correct and 0.48 is a typographical error in the results text. Second, the paper states 'as we found low heterogeneity (I2 = 0%; P = 0.18), a random-effects model was used', which inverts the rule — I-squared of zero is the condition for a fixed-effect model. Tau-squared is never reported. Downgrade one further level for imprecision: three studies, and the upper bound of the composite estimate sits at 0.89. Readmission and revision surgery showed no difference, and mortality was null.

Figures: Figures checked

CAT: OCEBM 2 (randomised trial) / GRADE Low · Start HIGH (randomised, ethics-approved, independent, unfunded). −1 risk of bias: open-label, which the authors correctly describe as unavoidable since a pump on the leg cannot be masked; wound area and healing are assessed by unblinded staff, and pain, quality of life and physical activity are self-reported by unblinded patients. −1 imprecision, and it is decisive: 14 against 17 patients, below the calculated sample size, with the shortfall caused by dropouts and deaths. The authors state this as their first limitation and attribute the non-significant trends to it, which is the correct reading. Net low. Two points of design quality deserve recording because they are rare in this literature. First, both arms followed the dynamic time strategy for wound bed preparation, so the comparator is structured standard care rather than the undemanding 'conventional dressing' that inflates so many NPWT trials — this vault has repeatedly flagged weak comparators, and this one is not weak. Second, the exclusion criteria are strict and clinically sensible — malignancy, chemotherapy or corticosteroids, NPWT or hyperbaric oxygen in the preceding seven days, haematological abnormality, deep vein thrombosis, malnutrition, and exposure of muscle, tendon OR bone. That last exclusion matters for transfer: the trial studied wounds WITHOUT exposed deep structures, which is not the difficult end of the vascular caseload. Multiplicity is unaddressed: healing rate, healing time, area reduction, infection, adverse events, pain, quality of life and physical activity are all examined in 31 patients with no correction and no hierarchy, so any individual significant secondary finding must be read as exploratory.

Figures: Figures checked

29 patients with diabetes, open bone resection in the foot for acutely infected wounds, NPWT with Soft Port technology for up to 112 days.

Main closure method: split-thickness skin graft.

CAT: OCEBM 4 (uncontrolled prospective series with an external historical comparator) / GRADE not applicable — no concurrent comparison group, so no effect estimate exists to grade · GRADE does not apply — there is no concurrent comparator, so nothing is compared and no effect can be estimated. The paper's central claim is non-inferiority and its design cannot deliver it. A non-inferiority conclusion requires three things this study does not have: a PRE-SPECIFIED MARGIN, a concurrently allocated comparator, and a confidence interval tested against that margin. What is offered instead is this cohort's 75% closure-or-readiness at a mean of 69 days set against a published 56% at 56 days for NPWT and 39% at 77 days for standard care. The timing asymmetry alone invalidates the first comparison: a higher proportion closed after a longer interval is not evidence of a better therapy, and comparing 69 days against 56 days favours this study by construction. The same defect appears in evidence/Searle 2017 and evidence/Tabley 2020, and this vault treats cross-study rate comparison as uninterpretable wherever ascertainment, population and follow-up are not matched. The population is nevertheless the right one for this wiki — acutely infected diabetic foot wounds after open bone resection, which is a common vascular referral — And the within-cohort descriptive data are internally consistent: 24% underwent delayed primary closure at a mean of 58 days, 52% progressed sufficiently to change treatment at a mean of 62 days, only one patient reached 112 days without sufficient progress, and split-thickness skin grafting was the main closure method. Those are useful expectations for counselling a patient; they are not evidence of efficacy.

Figures: Figures checked