4.1 Relationship of liraglutide with renal function
In order to estimate the effect of liraglutide on renal function in patients with T2DN, the statistic differences of eGFR, BUN, Scr, UACR and CysC between the liraglutide group and the control group were calculated and visualized by forest maps. The results of our meta-analysis suggested that there were significant differences between the liraglutide group and the control group in the levels of Scr (SMD=-0.81, 95% CI: [-1.22,-0.4], p < 0.0001) (Fig. 3A, Table 2), UACR (SMD=-2.34, 95% CI: [-3.65, -1.03], p = 0.0005) (Fig. 3B, Table 2) and CysC (WMD or MD=-0.70, 95% CI: [-1.01, -0.39], p < 0.0001) (Fig. 3C, Table 2) after treatment. However, no differences between the liraglutide group and the control group were detected in the levels of BUN (WMD=-1.06, 95% CI: [-2.22,0.10], p = 0.07) (Fig. 3D, Table 2) and eGFR(WMD=-0.81, 95% CI: [-1.22, -0.40], p = 0.21) (Fig. 3E, Table 2) .In summary, liraglutide greatly reduced the levels of UACR, Scr and Cysc compared with treatment without liraglutide.
Table 2
Outcome | Number of study | Sample | Hetergeneilty | Analysis model | Statistical method | WMD/SMD (95% CI) | P value |
Test/Control | χ² | I² | P |
FBG | 14 | 624/625 | 141.13 | 91% | < 0.00001 | Random-effects | Inverse Variance | -0.66[-1.04,-0.27] | 0.0009 |
PBG | 8 | 370/370 | 12.94 | 46% | 0.07 | Fixed-effects | Inverse Variance | -1.51[-1.68,-1.34] | < 0.00001 |
HbA1c | 14 | 515/523 | 217.29 | 94% | < 0.00001 | Random-effects | Inverse Variance | -0.61[-0.95,-0.27] | 0.0004 |
BMI | 10 | 378/386 | 98.99 | 91% | < 0.00001 | Random-effects | Inverse Variance | -2.27[-2.98,-1.56] | < 0.00001 |
eGFR | 3 | 94/95 | 15.20 | 87% | 0.0005 | Random-effects | Inverse Variance | 6.46[-3.69,16.61] | 0.21 |
UACR | 10 | 391/397 | 366.83 | 98% | < 0.00001 | Random-effects | Inverse Variance | -2.34[-3.65,-1.03] | 0.0005 |
BUN | 4 | 204/204 | 48.46 | 94% | < 0.00001 | Random-effects | Inverse Variance | -1.06[-2.22,0.10] | 0.07 |
Scr | 11 | 531/532 | 94.16 | 89% | < 0.00001 | Random-effects | Inverse Variance | -0.81[-1.22,-0.4] | < 0.0001 |
CysC | 2 | 32/39 | 0.44 | 0% | 0.5 | Fixed-effects | Inverse Variance | -0.7[-1.01,-0.39] | < 0.0001 |
IL-6 | 5 | 248/248 | 130.43 | 97% | < 0.00001 | Random-effects | Inverse Variance | -2.03[-3.30,-0.77] | 0.002 |
TNF-α | 6 | 303/303 | 38.24 | 87% | < 0.00001 | Random-effects | Inverse Variance | -1.16[-1.66,-0.66] | < 0.00001 |
4.2 Relationship of liraglutide with hypoglycemic related indicators and BMI
For validating the influence of liraglutide on hypoglycemic related indicators and BMI in patients with T2DN, the statistic differences of FBG, PBG, HbA1c and BMI between the liraglutide group and the control group were estimated. As shown in the Fig. 4A and Table 2, the patients in the liraglutide group measured lower levles of FBG (WMD=-0.66, 95% CI: [-1.04,-0.27], p = 0.0009) (Fig. 4A, Table 2), PBG (WMD=-1.51, 95% CI: [-1.68,-1.34], p < 0.00001) (Fig. 4B, Table 2), HbA1c (WMD=-0.61, 95% CI: [-0.95,-0.27], p = 0.0004) (Fig. 4C, Table 2) and BMI (WMD=-2.27, 95% CI:[-2.98,-1.56], p < 0.00001) (Fig. 4D, Table 2) than patients in the control group, suggesting the excellent performance in controlling blood sugar and weight loss of liraglutide.
4.3 Relationship of liraglutide with anti-inflammatory indicators
In order to further confirm whether the liraglutide could reduce inflammatory reaction and thus prevent renal fibrosis, we compared the levels of IL-6 and TNF-αbetween the liraglutide group and the control group after drug intervention. Notably, liraglutide was demonstrated delay the process of renal fibrosis by antiinflammatory in our analysis. Obviously, the liraglutidegroup was detected lower levles of TNF-α(SMD=-1.16, 95% CI: [-1.66,-0.66], p<0.00001) (Fig.5A, Table2) and IL-6 (SMD=-2.03, 95% CI: [-3.30,-0.77], p=0.002) (Fig.5B, Table2) than control group.
4.4 Sensitivity analysis and evaluation of publication bias
In order to verify the sources of heterogeneity, sensitivity analysis byremoving each study gradually was performed using Stata 13 software. The result showed that when removing the studies of Zhengyan [25] and Aiyitan [27], obvious changes of pooled WMD were found (Fig.6A). Therefore, we considered the heterogeneity coming from these two studies. Morever, funnel plots drawn through Revman software was used to display publication bias. Asymmetry was detected in Fig.6B-C, suggesting that there may be publication bias, and the results that are not statistically significant may not be published.