Data Extraction and Analysis
A single reviewer performed all data extraction to ensure consistency
across each study. The results were broadly divided into two categories:
safety and efficacy. After performing a full-text review, the most
frequently discussed variables pertaining to safety (i.e. rates of
infection, bleeding events, and mortality) and efficacy (i.e. adhesion
formation rate and standardized tenacity scores) served as the major
focus for analysis. “Infection” corresponded to any mention of
infectious processes related to superficial site infection, sternal
wound infection, mediastinitis, or sepsis after insertion of the
adhesion barrier. “Bleeding event” corresponded to any mention of
blood loss events intra- or post-operatively. “Mortality” corresponded
to any mention of death in patients receiving a barrier. “Adhesion
formation” corresponded to any mention of post-operative adhesions
identified either at re-operation or through imaging. “Standardized
tenacity score” (TSS) corresponded to any mention of a scaled-method
for reporting adhesion severity.11, 12 No universal
grading scale exists, and the studies assessed ranged from 0-3 to 0-21
point scales. All scales included a value for no adhesions (largely
“0”), and they then varied along a spectrum from mild to severe
adhesions requiring blunt to sharp dissection. A TSS equation that
standardized the reported tenacity score values was developed and is
presented in Figure 1.
A separate reviewer performed all data analysis using Excel. The rates
of infection, bleeding events, and mortality, were determined by
calculating the mean number of events in the patients receiving adhesion
barriers and the available control groups via Excel. The adhesion
formation was determined by calculating the mean number of events in
patients the patients with adhesion barriers that received
re-operation(s) and the available control group patients that received
re-operation(s) via Excel. The TSS was determined by using the equation
described previously (see Figure 1) via Excel. Issues during either data
extraction and/or analysis were resolved by a third reviewer.
RESULTS:
695 articles were identified through the initial database searches
(Scopus: 558; PubMed 137) (see Figure 2). After removing duplicates, 632
articles were screened. 603 were excluded, and the remaining 29
full-text articles were assessed for eligibility. Four studies were then
excluded (three for wrong study design and one for wrong language) with
a remaining total of 25 for review (see Table 1). The risk of bias is
presented in Figures 3 and 4. 21 out of 25 studies had high bias in
blinding of paricipants and personnel, as well as in outcome assessment.
21 out of 25 studies also had high bias in allocation concealment
(Figure 3). About 80% of the studies had low bias in selective
reporting (Figure 4).