ObjectiveAnalyze and compare the differences in the efficacy and adverse reactions of various ketogenic diet (KD) in the treatment of refractory epilepsy in children.MethodsSystematic search of electronic databases, including PubMed, Embase, Ovid MEDLINE, Web of Science and the Central Register of Cochrane Controlled Trials, published in English January 2000 Relevant research from January to August 2020. Results: Finally, 11 articles were included and 781 cases were included. Meta-analysis (NMA) method was used to compare 6 classic ketogenic diets (Classic ketogenic diet, CKD), Gradual ketogenic diet initiation (GRAD-KD), and the first modified Atkins diet of 20 g carbohydrates/d (Initial 20 g of carbohydrate/day of modified Atkins diet, IMAD), modified Atkins diet (MAD), low glycemic index diet (LGID) and medium-chain fatty acid diet (Medium-chain triglyceride diet, MCT) Therapeutic effect and adverse reactions of 3, 6, and 12 months.ResultsFrom the results of the direct comparative analysis, CKD and MAD showed superior clinical efficacy in 50% seizure reduction at 3 months to CAU, and the difference was statistically significant [OR=10.58, 95%CI (3.47, 32.40), P<0.05; OR=11.31, 95%CI (5.04, 25.38), P<0.05]; the clinical efficacy of 90% seizure reduction at 3 months for MAD was superior to that of CAU with statistical significance [OR=4.95, 95%CI (1.90, 12.88), P<0.05]. The results of further network meta-analysis suggested that for the comparison of 50% seizure reduction at 3 months, IMAD, GRAD-KD, CKD, MAD, and MCT were superior to CAU, and the difference was statistically significant [OR=0.03; 95%CI (0.00, 0.30), P<0.05; OR=0.07; 95%CI (0.01, 0.76), P<0.05; OR=0.11; 95%CI (0.03, 0.35), P<0.05; OR=0.11; 95%CI (0.04, 0.35), P<0.05; OR=0.13; 95%CI (0.03, 0.67), P<0.05; OR=0.11; 95%CI (0.03, 0.35), P<0.05; OR=0.11; 95%CI (0.04, 0.35), P<0.05]. For the comparison of 90% seizure reduction at 3 months, CKD, GRAD-CK, IMAD, MAD, and MCT were superior to CAU, and the differences were statistically significant [OR=0.05; 95%CI (0.00, 0.31), P<0.05; OR=0.22; 95%CI (0.00, 0.39), P<0.05; OR=0.03; 95%CI (0.00, 0.62), P<0.05; OR=0.12; 95%CI (0.01, 0.60), P<0.05; OR=0.09; 95%CI (0.00, 0.91), P<0.05]. It is suggested in the cumulative probability plot that: the optimal clinical regimen for 50% seizure reduction at 3 months was IMAD (Rank1=0.91), the optimal clinical regimen for 50% seizure reduction at 6 months was CKD (Rank1=0.40), the optimal clinical regimen for 50% seizure reduction at 12 months was MCT (Rank1=0.64); the optimal clinical regimen for 90% seizure reduction at 3 months was IMAD (Rank1=0.94), the optimal clinical regimen for 90% seizure reduction at 6 months was LGIT (Rank1=0.44), and the optimal clinical regimen for 90% seizure reduction at 12 months was MCT (Rank1=0.41); the optimal clinical regimen for seizure reduction at 3 months was GRAD-CK (Rank1=0.46), the optimal clinical regimen for seizure reduction at 6 months was LGIT (Rank1=0.58), and the optimal clinical regimen for seizure reduction at 12 months was CKD (Rank1=0.56). It is suggested in the benefit-risk assessment that among the three KDs (CKD, MAD, MCT) with better 50% and 90% seizure reduction at 3 months and 6 months, combining with the incidence of adverse reactions, CKD was the optimal treatment regimen (CF=0.47, CF=0.86); among the two KDs (CKD, MAD) with better seizure reduction at 3 months and 6 months, combining with the incidence of adverse reactions, CKD was the optimal treatment regimen (CF=0.45); among the two KDs (CKD, MCT) with better 50% and 90% seizure reduction at 12 months, combining with the incidence of adverse reactions, CKD was the optimal treatment regimen (CF=0.65).ConclusionsIn this study, IMAD showed the optimal clinical efficacy at 3 months and MCT at 12 months. With stable efficacy and low incidence of adverse reactions in 12 months, CKD was the optimal treatment regimen for children with refractory epilepsy after the comprehensive evaluation.