*3.4. Meta-Analysis*

#### 3.4.1. Primary Outcomes

All 10 studies with 526 participants evaluated the effects of serum insulin levels and HOMA-IR. The heterogeneity of serum insulin levels and HOMA-IR were moderate (*I* <sup>2</sup> = 68%, *I* <sup>2</sup> = 75%). Pooled results obtained by employing a random-effects model demonstrated that selenium supplementation remarkably lowered serum insulin levels (SMD: −0.53, 95% CI [−0.84, −0.21], *p* = 0.001) and decreased HOMA-IR (SMD: −0.50, 95% CI [−0.86, −0.14], *p* = 0.006) (Figures 4 and 5). To resolve heterogeneity, sensitivity analysis was conducted by excluding the studies one by one. The pooled results were broadly consistent with the above analysis (Supplementary Figures S1 and S2), and the heterogeneity was largely affected by Faghihi 2014 [48], which was excluded. Nine studies [30,41–47,49] remained after the exclusion, with no heterogeneity in the pooled results (all *I* <sup>2</sup> = 0%), indicating that Faghihi 2014 [48] was a major factor in the source of heterogeneity of insulin levels and HOMA-IR. This may be due to a baseline difference in the hypoglycemic drugs taken between the selenium and control groups in Faghihi's study. An analysis was then conducted with the fixed-effects model, and the result confirmed the previous observation

that supplementing with selenium was associated with lower serum insulin levels (SMD: −0.67, 95% CI [−0.86, −0.48], *p* < 0.0001) and HOMA-IR (SMD: −0.67, 95% CI [−0.86, −0.48], *p* < 0.0001) (Figures 6 and 7).

#### **Figure 3.** Risks of bias graph expressed as percentages.


#### **Figure 4.** Forest plot of insulin levels [30,41–49].


**Figure 5.** Forest plot of HOMA-IR [30,41–49].


**Figure 6.** Forest plot of insulin levels after excluding Faghihi 2014 [30,41–47,49].


**Figure 7.** Forest plot of HOMA-IR after excluding Faghihi 2014 [30,41–47,49].

The effect of selenium supplementation on FPG was assessed in 492 participants through nine studies [30,41–48]. The heterogeneity between studies was high (*I* <sup>2</sup> = 91%). Pooled analysis from the random-effects model indicated that the selenium group and the control group had similar effects on FPG (SMD: 0.06, 95% CI [−0.56, 0.68], *p* = 0.86) (Supplementary Figure S3). To resolve heterogeneity, sensitivity analysis was conducted by excluding the studies one by one. The results showed that although the pooled results were stable, the heterogeneity was not resolved (Supplementary Figure S4). Then, subgroup analysis was conducted based on the underlying diseases of the participants. As shown in Figure 8, the FPG levels in patients with cardiovascular disease were significantly lower in the selenium group than in the control group (SMD: −0.42, 95% CI: [−0.77, −0.07], *p* = 0.02), with no heterogeneity (*I* <sup>2</sup> = 0%). However, there was no statistical difference in terms of FPG between the selenium and control groups in the other two subgroups (Figure 8).

Only two studies [43,48], including 114 participants, assessed the effect of selenium supplementation on HbA1c. The heterogeneity between studies was high (*I* <sup>2</sup> = 85%). Thus, we did not perform a meta-analysis of HbA1c. Both studies reported reductions in HbA1c after treatment in both the selenium and control groups, in which Najib 2020 [43] reported a more significant reduction in HbA1c in the selenium group compared to the control group.

**Figure 8.** Subgroup analysis of FPG [30,41–48].

#### 3.4.2. Secondary Outcomes

Nine studies [30,41,42,44–49] with 492 patients evaluated the effects of TC, TG, and LDL-C, and five studies [30,41,44–46] with 259 patients evaluated the effects of VLDL-C. The heterogeneity of TC, TG, and VLDL-C was insignificant (all *I* <sup>2</sup> = 0). Unfortunately, the pooled results from the fixed-effects model demonstrated that selenium supplements did not significantly lower TC, TG, and VLDL-C in patients with CMDs (SMD: −0.07, 95% CI [−0.25, 0.12], *p* = 0.48, SMD: −0.12, 95% CI [−0.30, 0.06], *p* = 0.20, and SMD: −0.08, 95% CI [−0.33, 0.16], *p* = 0.51, respectively) (Figures 9–11). The heterogeneity of LDL-C was high (*I* <sup>2</sup> = 79%). Pooled results from the random-effects model demonstrated no significant difference in LDL-C between the two groups (SMD: −0.35, 95% CI [−0.76, 0.06], *p* = 0.10) (Figure 12).

**Figure 9.** Forest plot of TC [30,41,42,44–49].


#### **Figure 10.** Forest plot of TG [30,41,42,44–49].


#### **Figure 11.** Forest plot of VLDL-C [30,41,44–46].


**Figure 12.** Forest plot of LDL-C [30,41,42,44–49].

A total of nine studies [30,41,42,44–49], including 492 patients, evaluated the effects of HDL-C. The pooled results from the random-effects model indicated that selenium supplementation remarkably increased HDL-C levels (SMD: 0.97, 95% CI [0.26, 1.68], *p* = 0.007), with high heterogeneity across studies (*I* <sup>2</sup> = 92%) (Supplementary Figure S5). To resolve heterogeneity, a sensitivity analysis was conducted by excluding each study separately. The results showed that the pooled results were broadly consistent with the above analysis (Supplementary Figure S6), and the heterogeneity was largely affected by Ghazi 2021 [42]. Eight studies [30,41,44–49] remained after excluding Ghazi 2021 [42], and pooled results showed that the heterogeneity between studies was decreased (*I* <sup>2</sup> = 58%) (Supplementary Figure S7). The participants of Ghazi 2021 [42] were patients with atherosclerosis, and dyslipidemia is closely related to atherosclerosis. Therefore, considering that the source of heterogeneity may be related to the participants' underlying disease, the included studies were divided into subgroups based on the participants' disease type. As shown in Figure 13, the HDL-C levels were significantly increased in the diabetes mellitus subgroup and cardiovascular disease subgroup (SMD: 0.50, 95% CI [0.10, 0.91], *p* = 0.02 and SMD: 4.22, 95% CI [1.06, 7.37], *p* = 0.009), with significant heterogeneity among studies (*I* <sup>2</sup> = 59% and *I* <sup>2</sup> = 97%). However, there was no statistical difference between the selenium and control groups in the other two subgroups (Figure 13).

**Figure 13.** Subgroup analysis of HDL-C [30,41,42,44–49].
