Objective To comprehensively investigate the incidence of resorption of lumbar disc herniation, and provide reference data for clinical decision-making. Methods Seven electronic databases (PubMed, Embase, the Cochrane Central Register of Controlled Trials, Web of Science, China National Knowledge Infrastructure, Wangfang data and Chongqing VIP database) were searched for relevant studies that might have reported morphologic changes in lumbar disc herniation when reporting the follow-up results of patients with lumbar disc herniation treated non-surgically from inception to March, 2020. Articles were screened according to inclusion and exclusion criteria, and the total number of patients, number of patients with resorption, and other important data were extracted for analysis. Random effect models were used for meta-analysis, and subgroup analysis, sensitivity analysis, meta-regression analysis, and Egger’s test were performed. Results A total of 15712 articles were identified from these databases, and 48 were eligible for analysis. A total of 2880 non-surgically treated patients with lumbar disc herniation were included in the meta-analysis, 1740 of whom presented resorption. Meta-analysis revealed that the incidence of resorption was 0.60 [95% confidence interval (CI) (0.46, 0.72)]. In subgroup analyses, studies that quantitatively measured the resorption of lumbar disc herniation yielded statistically higher pooled incidence [0.73, 95%CI (0.60, 0.85)] than those that used qualitative methods [0.51, 95%CI (0.34, 0.69)] (P=0.0252). The pooled incidence gradually increased in randomized controlled trials (RCTs) [0.50, 95%CI (0.15, 0.85)], non-RCT prospective studies [0.59, 95%CI (0.48, 0.70)] and retrospective studies [0.69, 95%CI (0.36, 0.95)], but the difference was not statistically significant (P=0.7523). The pooled incidence varied from 0.58 [95%CI (0.54, 0.71)] to 0.62 [95%CI (0.49, 0.74)] after the sequential omission of each single study. There was no significant change in the pooled incidence [0.62, 95%CI (0.43, 0.79)] when only low-risk RCTs and high-quality non-RCT studies were included, comparing with original meta-analysis results. Meta-regression showed that measurements partially caused heterogeneity (R2=15.34%, P=0.0858). Egger’s test suggested that there was no publication bias (P=0.4622). Conclusions According to current research, there is an overall incidence of resorption of 60% [95%CI (46%, 72%)] among non-surgically treated patients with lumbar disc herniation. The probability of resorption should be fully considered before making a decision on surgery.