Statistical analysis
We estimated the effect of orlistat on serum uric acid level by pooling
baseline and final mean differences and standard deviation (SD) values
and 95% CI of the studies in both intervention and control groups. A
random-effects model by DerSimonian and Laird method [14] was used
to estimate the overall effect size. In addition, standard error of mean
(SEM) was converted to SD by following the formula[15]: SD = SEM\(\times\) \(\sqrt{n}\) (n = number of participants in each group). SDs
of the mean differences were calculated as follows:
\(\text{SD}_{\text{Change}}\)=\(\sqrt{{{((SD}_{\text{Baseline}})}^{2}+{{(SD}_{\text{Final}})}^{2})-(2\times R\times\text{SD}_{\text{Baseline}}\times\text{SD}_{\text{Final}})}\)assuming a correlation coefficient (R) 0.5 as it is a conservative
estimate for an expected range of 0-1. Between studies heterogeneity was
examined using Cochrane’s Q test (significance point at P <
0.1) and \(I^{2}\) statistic with values greater than 50% as evidence
of moderate to high heterogeneity.[16]. In order to find the
potential sources of heterogeneity, we performed subgroup analyses to
evaluate whether results were different by intervention period (3 month
and 6 month, sample size (< 100 and ≥ 100 participants),
health status (metabolic syndrome, hypercholesterolemia, PCOS and
NAFLD). We also performed sensitivity analysis by removed each study
from the analysis. Then, potential publication bias was assessed by
using Egger’s test [17] and Begg’s regression test . In case of
detecting potential publication bias P values less than 0.05. Duval &
Tweedie “trim and fill” approach was applied to adjust analysis for
the impacts of publication bias[18]. Random-effects meta-regression
analysis was conducted to evaluate the association between changes in
serum uric acid and potential confounders including duration of
treatment. The collected information from studies was imported into
Comprehensive Meta-Analysis software version 2 (Biostat, Inc.,
Englewood, NJ, USA)[19]. A P value of < 0.05 was
considered as statistically significant.