Ketogenic diet-induced changes in adult lipid metabolism: a comprehensive systematic review and meta-regression of randomized controlled trials.
Level 1 - systematic review of randomized trials
Systematic review and meta-regression of 53 randomized controlled trials
PubMed 42056882 · doi:10.1186/s12872-026-05799-5
What was done
Systematic review and meta-regression of randomized controlled trials (RCTs) searched across PubMed/MEDLINE, Web of Science, Scopus, Embase, and Cochrane Library through June 2025. The authors evaluated the effect of ketogenic diet interventions (>45% fat, <10% carbohydrate) compared to control diets on adult lipid profiles. Random-effects models (DerSimonian-Laird method) were used to calculate pooled weighted mean differences (WMDs) with 95% confidence intervals (CIs). Evidence quality was evaluated using Cochrane RoB 2 and the GRADE framework.
What was found
Across 53 RCTs, ketogenic diets compared to control diets resulted in: - Triglycerides: WMD -22.31 mg/dL (95% CI: -28.90 to -15.72; p < 0.001; I² = 69.8%) - HDL-C: WMD 3.52 mg/dL (95% CI: 1.24 to 5.81; p < 0.001; I² = 92.5%) - LDL-C: WMD 8.22 mg/dL (95% CI: 5.08 to 11.40; p < 0.001; I² = 59.9%) - Total cholesterol: WMD 8.06 mg/dL (95% CI: 3.04 to 13.07; p < 0.001; I² = 72.3%) Meta-regression identified a negative correlation between HDL-C increases and intervention duration (coefficient = -0.125; p = 0.011). Egger's test indicated no publication bias (p > 0.05). GRADE assessment determined moderate certainty for TG and HDL-C, and low certainty for LDL-C and total cholesterol.
Why it matters
This review shows that ketogenic diets produce mixed lipid changes in adults, lowering triglycerides and raising HDL-C while simultaneously increasing atherogenic LDL-C and total cholesterol.
Limits
The total number of participants was not reported in the abstract. Heterogeneity across trials was substantial to very high (I² between 59.9% and 92.5%), GRADE certainty was low for LDL-C and total cholesterol, and hard clinical cardiovascular outcomes were not evaluated.