3. Results
After an exhaustive search in five databases we retrieved a total of 13,073 articles. This number was reduced after automatic application of a filter to remove duplicates (Endnote) and manual removal of duplicates in Rayyan QCRI web and mobile app (Ouzzani et al., 2016), leaving a total of 7,868 articles to be submitted to the first phase round of application of eligibility criteria (Fig.1).
After applying the first-round of exclusion criteria (title and abstract), a total of 6,542 articles were excluded, leaving 1,326 articles to be read in full. In this second selection phase, more specific criteria were applied to further refine the search, leading to the exclusion of more than 1,309 articles (Supplementary Material S3 - https://osf.io/jexud/). The remaining 17 articles were sent for data extraction (Fig.1).
The objective of most of the studies (15) was to assess the biological activity of plants, extracts, or chemical compounds isolated from them. One study worked with a fraction of egg yolk levitins (Meram & Wu, 2017) and another with Celecoxib (Pang et al., 2019), a non-steroidal anti-inflammatory in the class of specific cyclooxygenase-2 enzyme inhibitors. Thirteen of the studies were carried out in Asia; five in China (Guo et al., 2016; He et al., 2019; Pang et al., 2019; Sun et al., 2017; Zhang et al., 2015) six in South Korea (M.-J. Kim et al., 2019; Y. S. Kim et al., 2016; H. A. Lee et al., 2017; S.-B. Lee et al., 2017; Lim, Kim, et al., 2015; Lim, Lee, et al., 2015), one in India (Ghate et al., 2018), and one in Indonesia (Laksmitawati et al., 2016). Three were conducted in the USA, two in Brazil (Da Silva et al., 2019; Mohr et al., 2019) and one in Canada (Meram & Wu, 2017). Only one of the studies included was from Europe (Turkey) (Karatoprak et al., 2019). Altogether, ten authors performed the NO dosage plus three of the cytokines included in our research (TNF-α, IL-1β, and IL-6) (Ghate et al., 2018; Guo et al., 2016; M.-J. Kim et al., 2019; Y. S. Kim et al., 2016; Laksmitawati et al., 2016; Lim, Kim, et al., 2015; Lim, Lee, et al., 2015; Meram & Wu, 2017; Mohr et al., 2019; Sun et al., 2017), four authors measured NO and two of the cytokines (Da Silva et al., 2019; He et al., 2019; S. B. Lee et al., 2017; Zhang et al., 2015) and only three studies included NO and only one cytokine (Karatoprak et al., 2019; H. A. Lee et al., 2017; Pang et al., 2019) (Table 1). As stated above, all the studies included in this review aimed to evaluate the possible anti-inflammatory activity of isolated compounds, plant extracts, or drugs. In some studies, the authors reported have linked this activity not only to the reduction in the production of pro-inflammatory mediators, but also to the inhibition of nuclear transcription factors such as NFκB and protein kinases such (JNK, ERK, and P38 MAPKs).
All the studies included in our research (17) were considered for a meta-analysis of NO. In all the manuscripts, NO was produced significantly in favor of the LPS induced cells compared to the non-induced cells (mean difference: 32.22 µM, CI95%: 25.81-38.63; p-value <0.001; i²=100%) (Fig. 2A), In order to assess whether the difference between cell density of the studies could interfere in the NO levels, we conducted a subgroup analysis. It was found that cell density did not contribute to the increase or decrease in NO levels produced by LPS-induced RAW 264.7 cells (1 - 2.5 x 105 cells mL-1: 30.61 µM; CI95%: 13.54 -47.68; 3 - 5 x 105 cells mL-1: 33.10 µM; CI95%: 28.77 - 37.43; p-value >0.05, i²=0%) (Fig. 3A). Since the placement range is 105 cells mL-1, independently of the plate used (6-96 wells), the NO levels increased significantly when induced by 1 µg mL-1 LPS compared to non-induced cells.
Similar to the NO results, TNF-α, IL-1β, and IL-6 levels production were significantly increased by induced-LPS cells, compared to non-induced cells (Fig. 2B, 2C and 2D), respectively.
TNF-α was the cytokine with the largest number of studies included for meta-analysis (15) (Ghate et al., 2018; Guo et al., 2016; C He et al., 2019; Karatoprak et al., 2019; M.-J. Kim et al., 2019; Y. S. Kim et al., 2016; Laksmitawati et al., 2016; S.-B. Lee et al., 2017; Lim, Kim, et al., 2015; Lim, Lee, et al., 2015; Meram & Wu, 2017; Mohr et al., 2019; Pang et al., 2019; Sun et al., 2017; Zhang et al., 2015). The TNF-α levels were significantly higher in the induced-LPS cells compared to the non-induced LPS cells (mean difference: 47233 pg mL-1, CI95% 46729-47737; p-value:<0.001, i²=100%) (Fig. 2B). Removing the outliers (S. B. Lee et al., 2017; Lim, Kim, et al., 2015; Lim, Lee, et al., 2015; Meram & Wu, 2017; Pang et al., 2019; Sun et al., 2017) results (that exceed the reaction limits) the TNF production was also higher in the LPS-induced cells (mean difference: 1373 pg mL-1, CI95% 884-1862; p-value:<0.001, i²=100%) (Supplementary Material S4 - Fig 5A).
In the subgroup meta-analysis of the effect of cell density, TNF-α production was favored by the increase in cell density (1 - 2.5 x 105 cells mL-1: 8546 pg mL-1; CI95%: 8007-9086; 3 - 5 x 105 cells mL-1: 105412 pg mL-1; CI95%: 102006 - 108818; p-value <0.001, i²=100%) (Fig. 3B). This was also observed when removing the outliers (1 - 2.5 x 105 cells mL-1: 754 pg mL-1; CI95%: 178 - 1329; 3 - 5 x 105 cells mL-1: 1373 pg mL-1; CI95%: 884- 1862; p-value <0.001, i²=93.2%) (Supplementary Material S5 - Fig. 6A).
Regarding NO and cytokine levels, the group with NO production of between 20-50 µM (67704 pg mL-1; CI95%: 67207 – 68201; p-value <0.001; I²=100%) showed a higher production of TNF-α levels compared to those > 50 µM (6204  pg mL-1; CI95%: 5045- 7363; p-value <0.001, I²=100%) (subgroup p-value <0.001, I²=100%) (Fig 4A). Despite this, when removing the outliers, TNF-α production was higher in the NO cells, producing above 50 µM (20-50 µM: 636 pg mL-1; CI95%: 445- 826; >50 µM: 1373 pg mL-1; CI95%: 884- 1862; p-value <0.01, I²=86.3%) (Supplementary Material S6 - Fig 7A).
For the analysis of correlation between NO and IL-6 levels, a total of 13 studies were included in the meta-analysis (Da Silva et al., 2019; Ghate et al., 2018; Guo et al., 2016; M.-J. Kim et al., 2019; Y. S. Kim et al., 2016; Laksmitawati et al., 2016; H A Lee et al., 2017; S. B. Lee et al., 2017; Lim, Kim, et al., 2015; Lim, Lee, et al., 2015; Meram & Wu, 2017; Mohr et al., 2019; Sun et al., 2017). Initially, we observed that the production of IL-6 was significantly higher in the cells induced by LPS (mean difference: 2905 pg mL-1; CI95%: 2788- 3021; p<0.001; i²=100%) (Fig. 2C). Removing outliers (Da Silva et al., 2019; S.-B. Lee et al., 2017; Lim, Kim, et al., 2015; Lim, Lee, et al., 2015; Meram & Wu, 2017), this was also observed (mean difference: 391 pg mL-1; CI95%: 296- 485; p<0.001; i²=100%) (Supplementary Material S4 - Fig. 5B). Regarding cell density, there was no difference between the subgroups (1 - 2.5 x 105 cells mL-1: 2843 pg mL-1; CI95%: 2665 -3021; 3 - 5 x 105 cells mL-1: 2957 pg mL-1; CI95%: 2777- 3138; p-value >0.05, i²=0%) (Fig. 3C). Removing outliers, the highest cell density showed a higher production of IL-6 levels (1 - 2.5 x 105 cells mL-1: 276 pg mL-1; CI95%: 137- 415; 3 - 5 x 105 cells mL-1: 505 pg mL-1; CI95%: 370- 641; p-value <0.05, i²=81.4%) (Supplementary Material S5 - Fig. 6B). Regarding the NO subgroups, the production of IL-6 was higher in the 20-50uM group (Fig. 4B), but with the removal of outliers, there was no significant difference between the groups (Supplementary Material S6 - Fig. 7B).
For IL-1β meta-analysis, 12 studies were initially included (Da Silva et al., 2019; Ghate et al., 2018; Guo et al., 2016; C He et al., 2019; M.-J. Kim et al., 2019; Y. S. Kim et al., 2016; Laksmitawati et al., 2016; Lim, Kim, et al., 2015; Lim, Lee, et al., 2015; Meram & Wu, 2017; Mohr et al., 2019; Sun et al., 2017; Zhang et al., 2015). As previously mentioned, IL-1β production was significantly higher in the LPS-induced cells compared to non-induced cells (mean difference: 1860 pg mL-1; CI95%: 1813- 1907; p<0.001; i²=100%) (Fig 2D). After removing outliers (Sun et al., 2017), IL-1β levels were also increased (mean difference: 548 pg mL-1; CI95%: 517- 580; p-value <0.001; i²=100%) (Supplementary Material S4 - Fig. 5C).
In the analysis by subgroups according to cell density, IL-1β production was higher with the increase in cell density (1 - 2.5 x 105 cells mL-1: 125 pg mL-1; CI95%: 79- 172; 3 - 5 x 105 cells mL-1: 2727 pg mL-1; CI95%: 2636 - 2819; p-value <0.001, i²=100%) (Fig. 3D). After removal of outliers, IL-1β levels increased significantly with cell density (1 - 2.5 x 105 cells mL-1: 125 pg mL-1; CI95%: 79 - 172; 3 - 5 x 105cells mL-1: 790 pg mL-1; CI95%: 732 - 848; p-value <0.001, i²=99.7%) (Supplementary Material S5 - Fig. 6C).
For the subgroup analysis of NO levels, IL-1β production was higher in the cells that produced less NO (20-50 µM) (mean difference: 2678 pg mL-1; CI95%: 2590- 2767; p-value <0.001; i²=100%) compared to those > 50 µM (223 pg mL-1; CI95%: 168 - 279; p-value <0.001; i²=100%) (Fig. 4C). We observed that one study with IL-1β production above 3000 pg mL-1 (Sun et al., 2017) was responsible for this difference. This study was therefore treated as an outlier and removed from the analysis. After this, a significant difference was observed in the group with NO greater than 50 µM (20-50 µM: 734 pg mL-1; CI95%: 679- 789; >50 µM: 223 pg mL-1; CI95%: 168- 279; p-value <0.01, i²=100%) (Supplementary Material S6 - Fig. 7C).
Our selection of studies involved strict inclusion criteria, as specific results of the studied markers were necessary for the inclusion or exclusion of the studies in/from this review. For this reason, and as expected, both for NO and for all the cytokines studied showed a publication bias in a non-heterogeneous (asymmetric way), as demonstrated by the analysis of funnel graphs (Supplementary Material S7 - Fig. 8A, B, C and D).