Results
The study flow diagram of the literature search and screening as per Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) is showed in Figure 1. A total of 2331 journals were identified when using keywords and various databases (Pubmed, Cochrane, Embase) to perform the literature search. In addition, 8 other journals were identified through a reference list of other related journals. After duplicates were removed, 2312 records were left to screen and we identified 2060 non-related journals that were not clinical trials, randomized controlled trials of humans and animals, unavailable full text, or other vegetable oils that were not relevant. The remaining 72 full-text articles were further accessed for eligibility through reading and 41 of them were removed due to the absence of control groups and unavailable data. In the end, a total of 31 articles fulfilled our criteria and was included in the meta-analysis. In this review, we observed 5 health complications which were obesity, atherosclerosis, hypertension, diabetes, and oxidative stress from consuming heated palm oil (HPO), heated corn oil (HCO), heated canola oil (HCNO), heated extra virgin olive oil (HEVOO), heated sunflower oil (HSFO), and heated soybean oil (HSBO). A total of 8 human studies and 23 in vivo studies involving rats and rabbits were included in this meta-analysis. For obesity, data were taken from either their weight gain, BMI (kg/m2), and adipose tissue; meanwhile, for atherosclerosis, we extracted data from low-density lipoprotein (LDL) cholesterol and thickness of the artery wall as it is caused by an abnormally high concentration of LDL‐cholesterol that deposits on the artery. Data of vasodilation and blood pressure were taken for hypertension while oxidative stress includes oxidative markers such as liver thiobarbituric acid reactive substances (TBARS), malondialdehyde (MDA), and the ratio of glutathione to oxidized glutathione (GSH/GSSG); and lastly, for diabetes, data of homeostatic model assessment of insulin resistance (HOMA-IR) was used. Table 1 shows the study design and baseline characteristics of these 31 studies.
Figure 2 shows the comparison between heated (inclusive of multiple reheating) palm oil and control on atherosclerosis, hypertension, and obesity only. Seven parts of data (from five reported journals) (Famurewa et al., 2017; Kamisah et al., 2016; Kamisah et al., 2015; Nkanu et al., 2018; Siti et al., 2017) were statistically significant at the study level and did not overlap the line of null effect. The results were also statistically significant at the meta-analysis level (P < 0.00001), therefore there were no health complications with the HPO consumption. This set of studies showed a high heterogeneity (I2 = 100%) with the random-effects model as the studies had different study designs. The overall mean difference was 0.18 (95% CI: 0.11, 0.25; P < 0.00001). Figure 3 shows the effect of HCNO on obesity and both studies (Bautista et al. 2014a and Bautista et al. 2014b) showed statistically significant and the overall effect estimate was statistically significant (P < 0.00001) which meant there was no adverse effect of HCNO on human health. The heterogeneity of this forest plot was high (I2 = 95%) and the total mean difference was 5.50 [95% CI: 4.12, 6.87]. Figure 4 shows the comparison between HCO and control of hypertension. The total mean difference was 0.62 [95% CI: -4.71, 5.95] with P = 0.82 (not significant). This set of studies had a low heterogeneity (I2 = 0%) as they were of the same paper. The comparison between HEVOO and control can be seen in Figure 5. Four studies (Battino et al., 2002; Rueda-Clausen et al., 2007; Sayon-Orea et al., 2013) which were Battino et al. 2002, Rueda-Clausen et al. 2005a; 2005band Sayon-Orea et al. 2013, respectively were statistically significant on the right side of the forest plot. The overall effect was also statistically significant (P = 0.003) which meant HEVOO was not cause any adverse effects mentioned above. The mean difference was 1.13 [95% CI: 0.39, 1.87] with a relatively high heterogeneity of I2 = 96%. Figure 6 shows a comparison of HSFO and control of obesity and oxidative stress. Three out of five studies (Aruna et al., 2005; Garrido-Polonio et al., 2004; Srivastava e al., 2010) were statistically significant which were Aruna et al. 2005, Garrido-Polonio et al. 2004a, and Srivastava et al. 2010. The overall effect estimate was statistically significant (P < 0.00001) on the left side of the forest plot which meant the effects of consuming HSFO include obesity and oxidative stress. This analysis had a relatively high heterogeneity (I2 = 99%) and the mean difference is -3.79 [95% CI: -5.61, -1.97]. The comparison between HSBO and control is shown in Figure 7 above with 9 statistically significant studies including (Rueda-Clausen et al., 2007, Das et al., 2017; Leong et al., 2010; Yen et al., 2010) Adam et al. 2009b, Awney 2001, Chiang et al. 2011, Chuang et al. 2013, Leong et al. 2010b; 2010c; 2010d, Rueda-Clausen et al. 2005a and Yen et al. 2010. The heterogeneity was relatively high for this set as well (I2 = 98%) as they included various study designs and intervention method and the mean difference was 0.46 [95% CI:0.29, 0.64]. However, the overall effect was statistically significant (P < 0.00001) which means HSBO will not cause any health implications.