Design: Systematic review and meta-analysis of 85 randomised trials
Population: 16,980 adults with primarily closed surgical incisions
Primary outcome: Overall surgical site infection
Effect: RR 0.64 (95% CI 0.57–0.72), NNT 21; blistering RR 4.51 (2.37–8.58)

This meta-analysis pooled 85 randomised trials of 16,980 adults with closed surgical incisions, comparing incisional negative pressure wound therapy with a standard dressing.

Negative pressure was associated with fewer surgical site infections overall (RR 0.64, 95% CI 0.57 to 0.72), for both superficial (RR 0.59) and deep (RR 0.66) infection, and with less dehiscence (RR 0.73), seroma (RR 0.77) and reoperation (RR 0.80, 0.64 to 0.99). Organ-space infection, readmission and mortality did not differ. Skin blistering (RR 4.51) and device-related problems (RR 11.87) were more frequent. The number needed to treat to prevent one infection was 21; the number needed to harm was 18 for blistering and 11 for device problems. Certainty was moderate.

At an NNT of 21, the dressing is hard to justify for every wound, particularly where cost matters. It makes most sense where an infection or dehiscence would be costly — obesity, diabetes, long laparotomies, contaminated fields closed primarily, and re-operative abdomens.

Consider incisional negative pressure for closed incisions at high risk of wound complications.
High-risk features include obesity, diabetes, smoking, long midline incisions and contaminated cases closed primarily.
Warn patients about blistering and check the skin at dressing changes.
Do not expect it to prevent organ-space infection or anastomotic problems.
Weigh the device cost against an NNT of about 21 in your own setting.

Why it matters

It turns a contested dressing into a targeted tool with a known trade-off.

The statistics, in plain English

An NNT of 21 means about 21 patients need the dressing for one infection to be prevented. An NNH of 18 for blistering means one extra patient blisters for roughly every 18 treated — so blistering is about as common an effect as the benefit. The reoperation interval just stays below 1.0 (0.99), so that result is fragile.