Results of document retrieval. A total of 80 publications were found; after duplicates were removed, 58 studies remained and of those, 18 were excluded upon abstract review. After reviewing 40 full-text articles, 8 studies were excluded because they were not randomized and 10 studies were excluded due to no available data (ie, the studies did not report the number of patients who developed PU or the incidence of PU in hydrocolloid dressing and control groups). The remaining 22 studies23-44 were included for analysis (see Figure 1), yielding a total of 2519 patients, including 368 NIV patients with PUs. The number of participants in the hydrocolloid dressing and control groups was 1260 and 1259, respectively; 67 PUs developed in the hydrocolloid dressing and 301 developed in the control group.
Participants in 3 studies35,37,43 were children (maximum age 25 months), and 19 studies23-34,36,38-42,44 included adult and geriatric patients (minimum age >19; 1 study26 did not specify exact adult age). Participants in treatment groups were provided hydrocolloid dressings before putting on the NIV face masks. Control group interventions included gauze dressings or regular skin care; in 9 studies,23,27,29,30,32-34,39,44 gauze was applied to skin beneath the mask. In 13 studies24-26,28,31,35-38,40-43 no dressing materials were applied; clinicians of patients who used NIV continuously for more than 2 hours were instructed to relax the headband, take off the mask for 10 to 15 minutes, keep the face clean and dry, and observe the skin condition. Skin care (keeping facial skin clean and dry, relaxing the NIV strap regularly, maintaining proper strap tightness, and observing the skin condition at intervals) was used to prevent PU development. In 4 studies,27,31,38,42 regular skin care involved Comfeel dressings (Coloplast Corp, Humlebaek, Denmark), and in 2 studies36,44 Algoplaque dressings (URGO Company, Paris, France) were provided. The table summarizes the characteristics of the 22 studies assessed in the final meta-analysis.
Main meta-analysis. The incidence of PU caused by NIV was 5.31% in the hydrocolloid dressing group and 23.91% in the control group. No heterogeneity was found among the included studies (χ² = 23.63; P = 0.31; I² =1 1%). Using a fixed-effect model, the summary OR of patients in hydrocolloid dressing group compared with patients in the control group was 0.15 (95% CI: 0.11–0.20; Z = 12.72; P <.00001) (see Figure 2). The overall effect showed hydrocolloid dressing use significantly decreased the development of PUs in patients with NIV. Sensitivity analysis using a random-effect model found the summary OR of patients was 0.16 (95% CI: 0.11–0.22; Z = 10.9; P <.00001). The sensitivity analysis did not materially change the result of the meta-analysis, indicating the outcome of the meta-analysis was robust (see Figure 3). The funnel plot showed asymmetry, suggesting a publication bias in this meta-analysis (see Figure 4).
Results of subgroup analysis. The meta-analysis of pediatric patients showed that in the hydrocolloid dressing group, 4 patients (1.28%) developed PUs versus 38 (12.14%) in the control group (OR = 0.09; 95% CI: 0.01-0.81; P = .03). In the 19 studies23-34,36,38,42-44 that included adult and geriatrics patients, 63 patients (6.65%) developed PUs in the hydrocolloid dressing group versus 263 (27.80%) in the control group (OR = 0.16; 95% CI: 0.12-0.22; P<.00001). Forest plot chart analysis indicated no differences in OR estimates were found across all age groups (see Figure 5).
In the 9 studies23,27,29,30,32-34,39,44 in which gauze was the control, meta-analysis showed 27 patients (6.01%) in the hydrocolloid dressing group developed PUs versus 108 (23.53%) in the control group (OR = 0.17; 95% CI: 0.10-0.28; P <.00001). In the 13 studies24-26,28,31,35-38,40-43 in which regular skin care was the control, 40 patients (4.93%) developed PUs in hydrocolloid dressing group versus 193 (24.13%) in the control group (OR = 0.13; 95% CI: 0.09-0.19; P <.00001). Forest plot chart analysis indicated no differences in OR estimates were found among different interventions in the control group (see Figure 6).