Studies were considered to be eligible for inclusion if they met all of the following criteria: 1) The study design is randomized controlled trials (RCTs), whether blinded or not. Exclude documents with incomplete data, incorrect data, and repeated publications; 2) The type of patient is not limited. All cases met the Chinese Medicine Association Neurology Branch “Diagnostic Criteria for Vascular Dementia”, the “Guidelines for Clinical Research of New Drugs for the Treatment of Dementia” issued by the Ministry of Health, or the American Neurological Society’s Handbook of Diagnostics and Statistics for Neuropathy. Diagnostic criteria for vascular dementia; 3) The experimental group was treated with Danhong injection combined with conventional chemotherapy. The control group received conventional chemotherapy, including aspirin, sibutramine, nimodipine, oxiracetam, butylphthalide, and vasodilators, energy mixture, vitamins, etc; 4) The outcome criteria for outcomes are expected to be clear, including at least one of the following outcome measures: the effectiveness of treatment, the score of the Simple Intelligent Mental State Energy Meter (MMSE), and the score for activities of daily living (ADL).

Comprehensive literature searches were conducted in CNKI Database, Chinese biomedical Literature Database, VIP database, Medline, PubMed, SCI Expanded. The search time was built for each database until December 31, 2018. The English search terms are: vascular dementia, Danhong. Chinese search terms are: Xueguanxingchidai, Danhong. Manually retrieve relevant materials and bibliography of each paper.

First, two research members independently read all the literature titles and abstracts to determine whether they meet the inclusion criteria. The documents that are not included in the uncertainty should be read and then judged; then, cross-check the test results. If there is any disagreement, it will be resolved through discussion between the two parties. If there are still differences, the third person will judge it. The extraction of the literature is implicit in the quality assessment of the Cochrane System Evaluator’s Manual (including whether the resulting data is complete, randomized, whether blinded, etc.)

Statistical analysis was performed using Revman 5.3. When the statistical data is a two-category variable, the relative risk (RR) and the 95% confidence interval (95% CI) was the effect analysis statistic. When the analysis index was continuous variable, the mean difference (MD) and 95% CI were used as the efficacy statistic.

Heterogeneity test: Chi-square test. When (P > 0.10, I2 < 50%), there was no statistical heterogeneity, and the fixed effect model was used for meta-analysis; when (P < 0.10, I2 > 50%), there was statistical heterogeneity, if not For clinical heterogeneity or methodological heterogeneity, a random effect model was used for meta-analysis; if clinical heterogeneity was present, a chi-square test was used. When (P > 0.10, I2 < 50%), there was no statistical heterogeneity 2, and a fixed effect model was used for meta-analysis; when (P < 0.10, I2 > 50%), there was statistical heterogeneity. If there is no significant clinical heterogeneity or methodological heterogeneity, a random effects model is used for meta-analysis; when clinical heterogeneity exists, sensitivity analysis or subgroup analysis is performed based on its source. If the heterogeneity is too large to abandon the meta-analysis, a descriptive analysis is used.

At the beginning, 38 articles were first retrieved, including 38 Chinese databases and 0 English databases. After screening by the researchers, 16 studies [4] - [19] were included, which were all published in Chinese, with a total of 1383 cases. Twelve of the studies [6] [7] [8] [9] [11] - [16] [19] took efficient as outcome. Eleven of the studies [1] - [6] [10] [12] [13] [14] [16] [17] [18] took MMSE score as outcome. The basic characteristics of the included studies are shown in Table 1.

①Efficient ②MMSE.

Of the included studies, only four studies described the use of random number table grouping, and the remaining 12 studies were not specified in the randomized protocol. None of the studies reported hidden allocation. All studies did not report whether blinding was used. None of the patients were lost to follow-up in all trials and no patients were withdrawn. The bias analysis of the included studies is shown in Figure 1.

Seven studies reported the effectiveness of a total of 583 patients, of which 299 patients were treated with Danhong injection combined with conventional chemotherapy, and 284 patients were treated with conventional chemotherapy alone. There was no heterogeneity between the groups (P = 0.99, I2 = 0%), and the fixed effect model was used for Meta analysis. Figure 2 shows that compared with conventional chemotherapy alone, Danhong injection combined with chemical drugs was more effective in the treatment of vascular dementia. The effective rate of the two groups was statistically significant (OR = 3.13, 95% CI [2.02, 4.85], P < 0.0001).

Twelve studies reported MMSE scores. There were 1116 patients, of which 566

were treated with Danhong injection combined with conventional chemotherapy, and 550 patients were treated with conventional chemotherapy. There were heterogeneity differences among the components (P < 0.00001, I2 = 87%), but the source of the difference is unknown, and subgroup analysis is not possible, so the random effect model is used for meta-analysis. Figure 3 shows that the MMSE scores of Danhong injection combined with conventional chemotherapy were higher than those of conventional chemotherapy alone, and the two were statistically different (SDD = 1.76, 95% CI [0.67, 2.85], P = 0.002).

Publication bias was used to perform a funnel plot analysis of the included studies. Figure 4 showed that the included studies were asymmetrically distributed, suggesting that there may be publication bias.