Page 78 - Revista Portuguesa - SPORL - Vol 62. Nº1

the assessment of risk of bias, taking into low risk, 4 as having some concerns and 1 paper
consideration: method of randomization;
allocation concealment; blinding; incomplete
outcome data; selective outcome reporting
and overall. We used the Version 2 of
the Cochrane risk-of-bias tool for randomized
trials (RoB 2), which involved describing each
of these domains as reported in the trial
and then assigning a judgment about the
adequacy of each entry as low, high, or unclear
risk of bias. We presented this information in a
‘risk of bias’ summary.
Statistical analysis
Meta-analysis was carried out in compliance
with Cochrane Collaboration standards using
RevMan v5.4 (Cochrane, London, United
Kingdom). A composite effect size estimate was
used to reflect relevant measures in the meta-
analysis. For meta-analytical comparison, the
effect size of each measure, as well as the 95
percent confidence interval and heterogeneity
as defined by the I statistic, were employed.
2
Means and standard deviations were obtained
through descriptive statistics, with statistical
significance set at p < 0.05. The outcomes of
patients from studies that precisely reported
the outcome of interest were compared. A
random effects model was utilized to calculate
the effect sizes. Each branch was treated as a
unique study for studies with several research
branches. Standardization was based on the
standardized mean difference using Cohen's
d values for studies reporting continuous
variables.
Results
Descriptive results
Three hundred and twenty-nine records
were found in the primary search. Titles and
abstracts were reviewed, and 298 papers were
excluded. 31 papers were comprehensively
evaluated for pertinency. After thoroughly
reviewing the article's whole body, 18 articles
were removed. 13 studies were used in the
qualitative analysis (Table 1). Using the RoB 2,
papers were assessed for risk of bias. Following
quality assessment, 8 papers were graded as
78 Revista Portuguesa de Otorrinolaringologia - Cirurgia de Cabeça e Pescoço
was graded as high risk of bias (see Table 2 for
details). The research by Athanasiadis et al. was
the only ‘intra-individual study design’ in the
quantitative analysis , with the other studies
8
assigning individuals to either an intervention
or placebo-controlled arm. Eldaba et al. was the
only work to include a pediatric population .
14
The study from Chhapola & Matta was
22
excluded from the meta-analysis because
statistical data was not adequately reported.
The study from Shehata et al compared HSI
3
and TXA and therefore provided both HSI and
TXA sample´s subgroups.
A total of twelve papers were therefore
included in the meta-analysis: two relating
to HSI , nine relating to TXA 5,6,8–14 , and one
4,7
comparing TXA and HSI . All were prospective
3
randomized controlled studies. A total of
106 HSI patients and 357 TXA patients were
included (463 cases) and matched with
controls (n = 412), making for a total of 875
patients. Of the ten articles which included TXA
samples, intra-operative TXA was used either
intravenously 6,9,10,12–14 or topically 3,5,8,11 . Figure
1 displays the flow diagram of the selection.
Concerning primary outcome measures, vBL
was covered in eleven studies 3–7,9–14 , SFQ in ten
studies 3,4,7–14 , and DS in seven studies 3,4,7,9,12–14 .
Among the 12 included studies in the meta-
analysis, 11 reported significant improvements
in the outcomes of interest in cases against
control groups (91.2%).
Meta-analysis
Blood loss
A random effects model of standard mean
differences showed that patients undergoing
HSI or TXA experienced significant reduction
in vBL compared to controls, yielding a
combined effect size of – 2.82 (p < 0.001)
(Figure 2). Individually, both HSI (p<0.001)
and TXA (p<0.001) were significantly superior
to controls in reducing vBL. Considering the
subgroup analysis (Figure 2), HSI presented
a more significant reduction in vBL than TXA
(effect size of HSI: - 7.55 vs effect size of TXA: -
2.30, p = 0.01).