Inclusion criteria:

1) Research type: Included in the study were clinical RCTs of different doses of dapaglilfozin versus placebo in the treatment of T2DM officially published before February 2018; full text is available; languages are limited to Chinese and English.

2) Research object: Age > 18 years, in accordance with the diagnostic criteria for T2DM (WHO Committee of Experts on Diabetes 1999 diagnosis and classification criteria), in accordance with the dapagliflozin guidelines.

3) Interventions: The experimental group was treated with dapagliflozan alone, regardless of dose size; placebo in the control group; or on the basis of other similar interventions.

Exclusion criteria:

Treatment observation time < 12 weeks, type 1 and special types of diabetes, review, non-RCT and so on.

Safety related indicators: hypoglycemia events, hypotension events, adverse renal reactions, genital tract infections, urinary tract infections.

The search strategy follows the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement [4] . Relevant studies for the analysis were selected by searching PubMed, Embase, Cochrance Library, Web of Science, CNKI, Wanfang Data and VIP database up to the end of February 2018. Taking Pubmed database as an example, subject words and free words are combined to search, and the retrieval scheme is as follows: #1 AND #2 AND #3.

#1 ((dapagliflozin) OR forxiga). #2 (((((((((((((((((((((((((((((((Diabetes Mellitus, Noninsulin-Dependent) OR Diabetes Mellitus, Ketosis-Resistant) OR Diabetes Mellitus, Ketosis Resistant) OR Ketosis-Resistant Diabetes Mellitus) OR Diabetes Mellitus, Non Insulin Dependent) OR Diabetes Mellitus, Non-Insulin-Dependent) OR Non-Insulin-Dependent Diabetes Mellitus) OR Diabetes Mellitus, Stable) OR Stable Diabetes Mellitus) OR Diabetes Mellitus, Type II) OR NIDDM) OR Diabetes Mellitus, Noninsulin Dependent) OR Diabetes Mellitus, Maturity-Onset) OR Diabetes Mellitus, Maturity Onset) OR Maturity-Onset Diabetes Mellitus) OR Maturity Onset Diabetes Mellitus) OR MODY) OR Diabetes Mellitus, Slow-Onset) OR Diabetes Mellitus, Slow Onset) OR Slow-Onset Diabetes Mellitus) OR Type 2 Diabetes Mellitus) OR Noninsulin-Dependent Diabetes Mellitus) OR Noninsulin Dependent Diabetes Mellitus) OR Maturity-Onset Diabetes) OR Diabetes, Maturity-Onset) OR Maturity Onset Diabetes) OR Type 2 Diabetes) OR Diabetes, Type 2) OR Diabetes Mellitus, Adult-Onset) OR Adult-Onset Diabetes Mellitus) OR Diabetes Mellitus, Adult Onset). #3 (((randomized controlled trial) OR randomized) OR placebo).

Data were independently abstracted by the two principal reviewers and any discrepancies were resolved by consensus. The data extracted included title, author, and year of publication, country, intervention measures, sample size, patient age, treatment time, main outcome indicators and intentionality analysis.

Jadad scoring criteria were used [5] , including 4 items: random method, assignment concealment, blind method and withdrawal. The RCT score was 1 - 7, 1 - 3 as low quality and 4 - 7 as high quality.

Outcomes were pooled using Review Manager 5.3 software. Dichotomous data were presented as odds ratio (OR) or risk ratio (RR). All results were estimated from each study with 95% confidence intervals (CIs). Heterogeneity was assessed using the I² statistic. If P > 0.1, I² < 50%, a fixed-effect model with the Mantel-Haenszel method was used; otherwise, the random-effect model was adopted. Subgroup analysis was performed according to the different doses of dapagliflozin (1 mg, 2.5 mg, 5 mg, 10 mg and 20 mg). Funnel plot was adopted to evaluate the publication bias. The bilaterally symmetric graph indicates a low risk of publication bias. Otherwise, the risk of publication bias may exist.

A total of 19 RCTs involving 7704 participants was incorporated into the study, including 5304 cases of dapagliflozin group and 2400 cases of the placebo group. All 19 RCTs were performed from 2009 to 2016, and the follow-up duration was ≥12 weeks. Baseline data for both groups were described in all trials, and there was no statistical difference in general between the dapagliflozin and placebo groups before treatment (P > 0.05). The screening process is displayed in Figure 1. The basic information contained in the study is shown in Table 1.

Two reviewers independently applied the Cochrane Risk of Bias tool to assess the risk of bias in the RCTs. The following methodological features relevant to the minimization of bias were assessed: randomization, allocation concealment, blinding of participants and personnel, blinding of outcome assessment, incomplete outcome data, selective reporting, and other bias. The following judgments were used: low risk, high risk, or unclear risk (either lack of information or uncertainty regarding the potential for bias). The bias risk assessment is shown in Figure 2.

A total of 19 studies were implemented [3] [6] - [23] to compare hypoglycemia events. Heterogeneity was detected in study subgroups (2.5 mg, 5 mg, 10 mg, 20 mg), indicating that there was no statistically significant heterogeneity (I² = 0%, P = 0.96), (I² = 0%, P = 0.67), (I² = 0%, P = 0.70), (I² = 0%, P = 0.89) between studies included in the analysis (Figure 3). Therefore, the M-H fixation effect model was used for meta-analysis. Funnel plot of publication bias evaluation is presented in Figure 4. There was no risk of publication bias among subgroups in the study. No significant differences were observed in the comparison results of dapagliflozin with placebo in the incidence of hypoglycemia event [OR = 1.14, 95%CI (0.95, 1.36), P = 0.17], as shown in Figure 3.

A total of 13 studies were implemented [6] - [11] [13] [15] [16] [18] [19] [20] [22] to compare hypotension events. Heterogeneity was detected in the study subgroups (2.5 mg, 5 mg, 10 mg), indicating that there was no statistically significant heterogeneity (I² = 0%, P = 0.60), (I² = 0%, P = 0.85), (I² = 0%, P = 0.79) between studies included in the analysis (Figure 5). Therefore, the M-H fixation

effect model was used for meta-analysis. Funnel plot of publication bias evaluation is presented in Figure 6. There was no risk of publication bias among subgroups in the study. No significant differences were observed in the comparison results of dapagliflozin with placebo in the incidence of hypotension event [OR = 1.43, 95%CI (0.94, 2.17), P = 0.10], as shown in Figure 5.

A total of 13 studies were implemented [3] [9] [10] [12] [13] [15] - [20] [22] [23] to compare renal adverse events. Heterogeneity was detected in study subgroups (2.5 mg, 5 mg, 10 mg), indicating that there was no statistically significant heterogeneity (I² = 0%, P = 0.42), (I² = 0%, P = 0.85), (I² = 0%, P = 0.89) between studies included in the analysis (Figure 7). Therefore, the M-H fixation effect model was used for meta-analysis. Funnel plot of publication bias evaluation is presented in Figure 8. There was no risk of publication bias among subgroups in the study. Compared with placebo, dapagliflozin (10 mg subgroup) significantly increased the incidences of renal adverse events [OR = 1.57, 95%CI (1.17, 2.09), P = 0.002], as shown in Figure 7.

A total of 19 studies were implemented [3] [6] - [23] to compare genital tract infections. Heterogeneity was detected in the study subgroups (2.5 mg, 5 mg, 10 mg, 20 mg), indicating that there was no statistically significant heterogeneity (I² = 0%, P = 0.81), (I² = 0%, P = 0.98), (I² = 0%, P = 0.64), (I² = 0%, P = 0.82) between studies included in the analysis (Figure 9). Therefore, the M-H fixation effect model was used for meta-analysis. Funnel plot of publication bias evaluation is presented in Figure 10. There was no risk of publication bias among

subgroups in the study. Compared with placebo, dapagliflozin (2.5 mg, 5 mg, 10 mg, 20 mg subgroup) significantly increased the incidences of genital tract infections [OR = 3.65, 95%CI (2.93, 4.56), P < 0.00001], as shown in Figure 9.

A total of 19 studies were included [3] [6] - [23] to compare urinary tract infections. Heterogeneity was detected in the study subgroups (1 mg, 2.5 mg, 5 mg, 10 mg, 20 mg), indicating that there was no statistically significant heterogeneity (I² = 0%, P = 0.53), (I² = 0%, P = 0.86), (I² = 0%, P = 0.77), (I² = 0%, P = 0.97), (I² = 0%, P = 0.88) between studies included in the analysis (Figure 11). Therefore, the M-H fixation effect model was used for meta-analysis. Funnel plot of publication bias evaluation is presented in Figure 12. There was no risk of publication bias among subgroups in the study. Compared with placebo, dapagliflozin (5 mg, 10 mg subgroup) significantly increased the incidences of urinary tract infections [OR = 1.36, 95%CI (1.15, 1.61), P = 0.0004], as shown in Figure 11.