Figure 1 presented the study selection process. 1242 articles were found in nine databases following search strategy described above. After removing duplicates, screening the titles and abstracts, and reading entire articles, 21 studies with 1668 patients met the inclusion criteria and were included into our meta-analysis [22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42]. 15 publications were presented in Chinese language and 6 in English. The 17 included trials were conducted in China [22,23,24, 26, 28, 31,32,33,34,35,36,37,38,39,40,41,42], while the other four were conducted in the United States, France, Mexico, and Brazil [25, 27, 29, 30]. Of all the eligible studies, three compared [25, 29, 30] the efficacy of thalidomide with placebo, and seven [23, 24, 26,27,28, 38, 40] compared with other medical measures including prednisone, vitamin B1, montmorillonite powder, dapsone, colchicine, etc. Nine trials [31,32,33,34, 36, 37, 39, 41, 42] evaluated the combination therapy of thalidomide and other drug compared to other drugs alone. And two triple-arms trials compared the efficacy and safety among the combination of thalidomide and TCM, thalidomide alone and TCM alone [22, 35]. The duration of treatment ranged from 5 days to 6 months. The detailed information on the eligible trials is shown in Table 1.

Flowchart of the study selection process RCT = Randomized controlled trial, CNKI = China National Knowledge Infrastructure, CBM = Chinese BioMedical Literature Database

Three independent raters assessed the methodological quality of the eligible RCTs using the Cochrane Risk of Bias tool. All selected studies were randomized controlled, but only two [25, 38] reported allocation concealment and were evaluated as low risk. Only three [25, 29, 30] and four [25, 29, 30, 38] studies were judged as low risk about detection bias and performance bias, respectively. All studies were classified as low-risk about selective reporting and incomplete outcome. There was insufficient information about other bias and the risk was assessed as unclear. The results of the quality evaluation are shown in Fig. 2.

Risk of bias graph of included RCTs

We performed subgroup analysis under different outcomes.

Complete response.

Three studies [25, 29, 30] evaluated the complete response of thalidomide versus placebo for RAS, the meta-analysis showed a significant improvement in those who received thalidomide versus placebo (RR = 6.95, 95% CI: 3.04–15.92, I2 = 39%, p < 0.00001, fixed-effect model, Fig. 3a). Pooled data from eight trials [22,23,24, 26,27,28, 35, 40] showed that thalidomide had a better therapeutic effect than other drug on complete response (RR = 2.18, 95% CI: 1.38–3.45, I2 = 68%, p = 0.008, random-effect model, Fig. 3b). A meta-analysis of ten trials [22, 31,32,33,34,35,36,37, 39, 41] revealed that the combination therapy of thalidomide and other medicine was better than other medicine alone in complete response (RR = 2.44, 95% CI: 1.55–3.84, I2 = 59%, p = 0.0001, random-effect model, Fig. 3c).

Forest plot of thalidomide group versus control group in complete response rate for RAS treatment. RAS = recurrent aphthous stomatitis, CI = confidence interval. a thalidomide vs. placebo, b thalidomide vs. other drugs, c thalidomide + other drugs vs. other drugs alone

In the only study [25] that reported overall response compared with placebo, the thalidomide group had the higher overall response rate (RR = 3.59, 95% CI: 1.87–6.89, Fig. 4a). Seven studies [22, 24, 26,27,28, 35, 40] evaluated the overall response between the thalidomide and other medicine, and although this did not reach statistical significance, it implies that the thalidomide group might be more effective than the control group in terms of overall response (RR = 1.18, 95% CI: 0.99–1.41, I2 = 80%, p = 0.07, random-effect model, Fig. 4b). Ten studies [22, 31,32,33,34,35,36,37, 39, 41] evaluated that the overall response of the combination of thalidomide and other drug, the meta-analysis indicated that the combination of thalidomide and other drug was superior to other drug alone in overall response (RR = 1.32, 95% CI: 1.23–1.42, I2 = 0%, p < 0.00001, fixed-effect model, Fig. 4c).

Forest plot of thalidomide group versus control group in overall response rate for RAS treatment. RAS = recurrent aphthous stomatitis, CI = confidence interval. a thalidomide vs. placebo, b thalidomide vs. other drugs, c thalidomide + other drugs vs. other drugs alone

Three studies [23, 38, 42] reported on RI, pooled data indicated thalidomide can significantly prolong the RI compared to other interventions (MD = 26.91, 95% CI: 2.29–51.54, p = 0.03, I2 = 99%, random-effect model, Fig. 5).

Forest plot of thalidomide group versus control group in recurrence interval changes for RAS treatment. RAS = recurrent aphthous stomatitis, CI = confidence interval

Four studies [31, 32, 35, 38] provided data on the healing time of the ulcers, the analysis results suggested that thalidomide group was better able to accelerate the healing process compared to the control group (MD = -1.31, 95% CI: -2.06 to -0.56, I2 = 70%, p = 0.0006, random-effect model, Fig. 6).

Forest plot of thalidomide group versus control group in healing time for RAS treatment. RAS = recurrent aphthous stomatitis, CI = confidence interval

Number of ulcers was reported in four studies [23, 26, 35,