A total of 253 studies were identified in the initial search. Of these, 206 studies were excluded after screening on titles and abstracts. Full-text reading was performed only for 47 potential studies and details of the searches were shown in the flow chart (Figure 1). Two publications [20, 21],written by the same authors, reported same population, so only the study with more participants was included in our meta-analysis. Finally, 25 studies [20, 22-45] that met the inclusion criteria were included in this systematic review. Of these, 14 studies were written in English and others were written in Chinese.
Characteristic of included studies
Totally, 2498 DM patients and 1424 healthy controls were included in our meta-analysis. Of these 25 studies, 12 were conducted for the different serum levels of YKL-40 between DM patients and healthy controls, 5 were analyzed for YKL-40 levels between GDM patients and healthy controls, and 8 were detected for YKL-40 levels between DM patients with different degree of albuminuria and healthy controls. The characteristics of the included publications are shown in Table 1.
Association between serum YKL-40 levels and DM
Totally, 12 studies showed an association between the serum YKL-40 levels and DM. The meta-analysis results indicated that the serum YKL-40 levels were significantly higher in DM patients compared with healthy controls (SMD=1.62, 95%CI,1.08 to 2.25, P=0.000) (Figure 2). The Galbraith plot was used because of the notable heterogeneity. But the major source of heterogeneity could not be found since too many of the studies were outliers (Figure 3). Furthermore, subgroup analyses by type of DM, region and age showed that YKL-40 levels were still higher in DM patients than those in healthy controls. The value of I2 remained high in various subgroups, with the exception of one subgroup for studies based on population of Asia.
Association between serum YKL-40 levels and GDM
Owing to significant heterogeneity, we used the random-effects model. The pooled SMD was 2.85 (95%CI, 1.01 to 4.70, P=0.002), which indicated that the serum YKL-40 concentrations were significantly higher in GDM patients compared with healthy pregnancies (Figure 4). The source of heterogeneity was hard to be found by the Galbraith plot because the studies were too dispersive. However, when performing sensitivity analysis by sequential omission of individual studies, YKL-40 was not associated with GDM when the article by Xun Shengli et al.  was removed. The pooled SMD was 0.64 (95%CI, -0.28 to 1.56) (P>0.05).
Association between serum YKL-40 levels and albuminuria in DM patients
There were 7,8 and 7 studies analyzing the relationship between serum YKL-40 levels and normoalbuminuria, microalbuminuria and macroalbuminuria, respectively. The forest plot with a random-effects model showed that DM patients with different degree of albuminuria had significantly higher levels of YKL-40 compared with healthy controls (normoalbuminuria: SMD=1.58, 95%CI, 0.59 to 2.56, P=0.002; microalbuminuria: SMD=2.57, 95%CI, 0.92 to 4.22, P=0.002; macroalbuminuria: SMD=2.69, 95%CI, 1.40 to 3.98, P=0.000). The Galbraith plot was applied to detect the potential source of heterogeneity. However, we could not find the possible source of heterogeneity because it plotted too many studies as the outliers. In addition, we conducted subgroup analyses by region and type of DM. The results did not change in various subgroups, and the value of I2 remained high in various subgroups, with the exception of one subgroup for studies based on population of Asia. What’s more, serum YKL-40 levels increased with increasing severity of albuminuria among DM patients (microalbuminuria vs normoalbuminuria: SMD=1.49, 95%CI, 0.28 to 2.71, P=0.016; macroalbuminuria vs microalbuminuria: SMD=0.93, 95%CI, 0.34 to 1.52, P=0.002).
We performed a sensitivity analysis by sequential omission of individual studies. When serum YKL-40 levels were compared between DM patients and heathy controls as well as DM patients with different degree of albuminuria and healthy controls, the pooled SMD were not materially altered. However, YKL-40 was not associated with GDM when the study by Xun Shengli et al.  was deleted.
Funnel plot and Egger’s test were conducted to evaluate the potential publication bias. There was no obvious funnel plot asymmetry and all the P values of the Egger’s tests were over 0.05, suggesting that publication bias was not evident in our meta-analysis.