Identification of relevant RCTs
Our initial search yielded 872 studies. Among these, 226 of records were excluded from this meta-analysis for duplicate records. After screening the title and abstracts, 35 publications met the crude inclusion criteria and were selected for further assessment. Among them, 2 were excluded for non-human studies, 8 not randomized trials and 13 insufficient data. Finally, a total of 12 eligible RCTs17-28 with 1272 patients were included for this meta-analysis. The detailed manuscript screening processes are shown in Figure 1.
Characteristics of the included studies
All RCTs were published between 2006 and 2020, the primary characteristics of them are described in Table 1. 10 RCTs were conducted in China, 1 in Iran and 1 in Greece. The maximum sample size is 180 in a RCT23 and the minimum sample size is 60 in another RCT17. Four RCTs reported patients with diabetes mellitus and report the results (Glucose, insulin or ISI). Overall, all enrolled studies examined the effect of fenofibrate in the treatment of patients with NAFLD.
All trials had two parallel groups and were described as randomized, and allocation concealment was not reported in any of the Chinese studies. The detailed risk of bias, as assessed using the tool from the Cochrane Collaboration, is shown in Figures 2 and 3. The overall risk of bias of all included RCTs was high.
Meta-analysis results
All RCTs reported the data of lipids and 12 studies assessed TC and TG in NAFLD patients that treated with fenofibrate compared with control groups (Figure 4). The result showed a statistically significant difference between the experimental and control groups (MD −0.65 and 95% CI [−0.93, −0.36] with P<0.00001 on TG mmol/L; MD −0.97 [−1.29, −0.66] with P<0.00001 on TC mmol/L). 6 RCTs analyzed HDL and LDL also showed significant difference between the experimental group and the control group (MD 0.22 and 95% CI [0.16, 0.29] with P<0.00001 on HDL mmol/L: MD -0.25 and 95% CI [-0.42, -0.09] with P=0.006 on LDL mmol/L).
11 RCTs reported on the effect of fenofibrate on improving liver function of patients, including lowering serum ALT, AST and r-GT (Figure 5). Fenofibrate had significant difference between the experimental group and control group (MD -7.67 and 95% CI [-12.93, -2.42] with P=0.004 on ALT U/L: MD -8.48 and 95% CI [-11.83, -5.13] with P<0.00001 on AST U/L: MD -11.73 and 95% CI [-18.90, -4.56] with P=0.001 on r-GT U/L)
5 RCTs provided sufficient data to compare the Serum liver fibrosis (HA, LN, PCⅢ, CVI) of experimental group with control group (Figure 6). Results showed a significant decrease in the Serum liver fibrosis after the treatment with fenofibrate (MD -43.49 and 95% CI [-68.84, -18.13] with P=0.0008 on HA μg/L: MD -28.58 and 95% CI [-42.23, -14.94] with P<0.0001 on LN μg/L: MD -18.19 and 95% CI [-30.28, -6.10] with P=0.003 on PCⅢ μg /L: MD -2.10 and 95% CI [-3.75, -0.46] with P=0.01 on CIV μg /L)
4 RCTs reported the effect of fenofibrate on Insulin resistance (Glucose, Regular insulin, ISI) (Figure 7). Fenofibrate had no effect on Insulin resistance (MD -0.45 and 95% CI [-1.30, 0.41] with P=0.30 on Glucose mmol/L; MD -0.09 and 95% CI [-2.03, 1.86] with P=0.93 on insulin mu/L; MD 0.09 and 95% CI [-0.03, 0.20] with P=0.13 on ISI)
6 RCTs reported the total effective rate of fenofibrate treatment (Figure 8). The evaluation criteria included relief or disappearance of clinical symptoms, Recovery of liver function and blood lipid levels, and B-ultrasound examination of liver morphology returning to normal or improving. The results revealed that fenofibrate has a significant effect for NAFLD patients compare with control groups (OR 3.32 and 95% CI [2.23, 4.93] with P < 0.00001)
Sensitivity analysis, publication bias, and evidence strength
Sensitivity analysis indicated consistent results. The funnel plot, Begg's rank correlation test, and Egger’s regression test were performed for publication bias and there was no significant publication bias among the included RCTs (Table 2). The evidence strength of meta-analysis was low because of the high risk of bias of included RCTs and the significant heterogeneity.