2.4. Meta-analysis
A minimum of 3 studies were used to perform a meta-analysis. If the studies presented the same type of intervention (LPS) and control, with the same outcome measure, we pooled the results using a meta-analysis of random effects, with standardized mean differences for continuous outcomes, calculating the 95% confidence intervals and two-sided P values for each outcome. In studies where the effects of clustering have not been taken into account, was to adjust the standard deviations for the design effect. Statistical significance was defined as a p-value< 0.05 (Z test). Cochran’s Q was used to assess heterogeneity, Q was used to calculate the excess variance. A p-value can be calculated for Q, indicating whether all studies will share a common effect size (p < 0.05) or not (p > 0.05), We used univariate meta-regression, where sufficient data existed (> 10 comparisons per variable). We also used the I2 test to investigate heterogeneity (p > 0.1, heterogeneity, I2 > 50%). A Stratified meta-analysis was used to explore heterogeneity in effect estimates according to subgroups, and evidence of publication bias was assessed using funnel plot analysis. For the stratified analyses, subgroup analyses were carried out. The following characteristics of the study were examined as a potential source of heterogeneity: time of LPS-exposure (24 hours); interval of nitric oxide/nitrites concentration on a positive control group (20-50 µM/ > 50 µM) (studies showing mM/mL or another scale were converted to µM); and the method used to determine cell density (cells per well or cells per mL). The results of studies reported as zero were used 0.1 to perform a meta-analysis. The sensitivity analysis was performed after removing outliers, according to the assay range of the cytokines ELISA kit: TNF: 10 - 6.000 pg mL-1, IL-1: 3 - 3.000 pg mL-1, and IL-6: 5 - 2.000 pg mL-1.