Supplements · Diets · Metabolic & Cardiometabolic
garlic decreases fasting blood glucose
In plain terms: Does garlic lower blood sugar?
Strong support Supplements 💰 Industry COI noted🔎 Limited evidence — fewer than 12 studies
Part of: 🧪 Garlic
RefutedContestedStrong support
consensus score 1.00
Probably a little, in people with type 2 diabetes — studies consistently point the same way (fasting glucose down ~7–12 mg/dL). But the trials are small, low-quality, and wildly inconsistent in size, so treat it as a modest, uncertain add-on, not a real glucose treatment.
📅 Last reviewed: 2026-07-15 ⓘ
Evidence ladder
How far up the ladder this claim has climbed. A high consensus on a low rung means "consistent so far," not "proven in people."
Top evidence so far: All trials, pooled (Meta-analysis)
MechanismIn-vitroAnimalObservationalRCTMeta-analysis
How the studies fall
8 support 0 contradict 0 tested null 0 mixed · 8 sources, 8 independent groups · 3 superseded — pooled inside a review, counted once
What the evidence shows
Across multiple meta-analyses of type-2-diabetes trials, garlic consistently lowers fasting blood glucose and HbA1c — but the pooled magnitude varies widely by analysis: older MAs read ~7–12 mg/dL while the largest GRADE-assessed dose-response MA (Alsanie 2026, 27 RCTs) finds only −2.76 mg/dL (CI −5.11 to −0.41), with bigger drops at LOWER doses — a small-study-bias signature.
The evidence (12)
| Source | Grade | Stance | Quality | Finding |
|---|---|---|---|---|
| Ebrahimzadeh A, Safargar M, Rajabzadeh-Dehkordi M, Nematolahi F, Samadani M, Saeid L, Mohtashamian A, Izadi MA, Ebrahimzadeh A, Safarpour AR 2026 · Nutr Metab Insights | meta-analysis | supports | low | Random-effects MA (PROSPERO CRD42024628780), 8 RCTs / 13 effect sizes / 607 participants published 2005-2020. The FBS result is NOT from all 8: it pools 5 studies contributing 10 effect sizes, WMD -12.41 mg/dL (95% CI -15.13 to -9.69), I2 = 99.2%. HbA1c -0.5% (95% CI -0.66 to -0.33), I2 = 96.4%, from 3 studies / 7 effect sizes. Direction is real; the magnitude is not trustworthy. Cochrane risk of bias rates 7 of 8 included studies high overall ('one of the studies ... was of high quality, while the remaining studies exhibited low methodological quality'); publication bias is positive on this very outcome ('Eggers test indicated the presence of publication bias for FBS and HbA1c (P < .05)'); one Pakistani trial is entered twice in Table 2, once as five dose-arms against a single shared control, which is a unit-of-analysis error that alone can explain I2 = 99.2%; and the paper's own FBS subgroups all sit near -7 mg/dL, so -12.41 cannot be rebuilt from its own tables. Read at full text 2026-08-18 (PMC12901877). |
| Kumar R et al. 2013 · Diabetes Metab Syndr Obes | RCT | supports | low | Open-label randomized add-on RCT, n=60 T2D+obesity, 12 wk: metformin+garlic 250 mg BD cut FBG 157.10 to 121.34 mg/dL vs 160.32 to 134.97 on metformin alone (between-group P<0.01, same for PPBG); within-group HbA1c fall NOT significant in either arm. Unblinded, no placebo, unregistered, branded capsule, and Table 5 contradicts Tables 3-4 on six values. ↩ SUPERSEDED — pooled in the review above, counted once |
| Singh A et al 2026 · International Journal of Applied and Basic Medical Research | animal | supports | low | HFD+STZ type-2-diabetic male rats (n=6/group) given whole aqueous garlic 60 mg/day by gavage for 28 days -- a human-equivalent dietary dose, not a pharmacological one. At day 28 fasting blood glucose fell to 233.6+/-45.57 mg/dL vs 459.6+/-67.67 in untreated diabetic rats (P<=0.05), and OGTT glucose was lower at 30 and 120 min but NOT at 60 or 90 min: 'significant improvements observed during fasting, 30 min, and 120 min OGTT measurements, but not at the 60 min and 90 min time points.' Garlic also lowered total cholesterol (119.34 vs 169.55) and triglycerides (123.85 vs 200.14) and raised HDL (50.94 vs 25.37). Unblinded single-lab rat study; garlic-group FBG remained above the study's own >200 mg/dL diabetic threshold, so the effect is a partial reduction, not normalization. |
| Hesari NG et al 2025 · J Health Popul Nutr | RCT | supports | moderate | Double-blind placebo RCT, 97 Iranian women with PCOS + metabolic syndrome (non-diabetic; FBS <126 mg/dL required), garlic 500 mg tablet (2-3 mg allicin) twice daily for 8 weeks. Table 1: FBS 102.08 -> 94.35 mg/dL (baseline-adjusted mean) vs placebo 99.71 -> 100.32, F=29.953, P<0.001; the paper states 'In the present study, the mean FBS level in the garlic group was significantly lower than that of the control group after eight weeks.' Direction is clear, but the abstract's figures ('from 107.2 to 95.9 mg/dL', 10.5%) cannot be reconciled with Table 1's baseline of 102.08, the arm sizes are swapped between abstract (49/48) and Table 1 (48/49), and the abstract reports FSFI improvement 'P<0.05 for all' where the Results text says the adjusted analysis 'showed no significant differences between the garlic and placebo groups at week eight' - so quality is moderate, not high. Diet and physical activity were not monitored. |
| Sobenin IA et al. 2008 · Acta Diabetol 45(1):1-6 | RCT | supports | low | Double-blind placebo-controlled RCT, n=60 type 2 diabetics, 4 wk: 'treatment with Allicor resulted in better metabolic control due to the lowering of fasting blood glucose, serum fructosamine and serum triglyceride levels.' Direction only -- the abstract reports no effect size, CI or p-value for any of its four outcomes, and no full text exists to recover them (not open access). Cholesterol was measured but is absent from the improvements list. Proprietary Allicor preparation; abstract closes 'may allow recommending garlic powder tablets Allicor for the treatment of type 2 diabetes mellitus'. ↩ SUPERSEDED — pooled in the review above, counted once |
| Mejía Delgado EM et al 2025 · World J Exp Med | observational | supports | low | Narrative review (PubMed/MEDLINE, EMBASE, Scielo; Spanish and English; no date limit; 122 references) of Allium sativum in T2DM. Zero-weighted as evidence_role: review — it pools nothing and reports no data of its own; its fasting-glucose conclusion rests on the same meta-analyses and trials the vault already counts directly on this claim. Direction as asserted is supportive: 'In patients with T2DM, garlic supplementation (300-1500 mg/day) has been associated with reductions in fasting glucose (-17 to 29 mg/dL), HbA1c (-0.4% to 0.8%) and improved insulin sensitivity', and 'Systematic reviews and meta-analyses have demonstrated that garlic ... can significantly reduce blood glucose levels, HbA1c, and fructosamine'. But the body is more equivocal than the abstract: 'studies that administered raw garlic (300 mg/day or 3 cloves per day for 2 weeks) or aged garlic (1200 mg/day for 4 weeks) did not show significant effects on the reduction of fasting glucose', and the authors concede 'current evidence is still limited by factors such as short intervention periods, small sample sizes, and lack of standardized garlic preparations'. Effect ranges are quoted from primaries, never pooled or CI-bearing; the printed range '-17 to 29 mg/dL' carries one sign and should not be lifted as a magnitude. Unregistered, no eligibility protocol, no risk-of-bias appraisal; journal's own published reviewer grades split A/B/D. |
| Zhao X et al. 2024 · Nutrients | meta-analysis | supports | low | Meta of 22 RCTs/29 trials (1567 adults): FBG -7.01 mg/dL (95% CI -8.53, -5.49; p<0.001) from only 8 studies/12 effect sizes at I2=91.8%; HbA1c -0.66% (-0.76, -0.55) from 'three trials with seven effect sizes' (I2=62.9%), too few for the authors' own subgroup or bias tests. No FBG publication bias (Egger p=0.285), but the effect is confined to the T2DM subgroup (-7.01) and null in 'Other' populations (-3.83, p=0.227) and in trials <=8 weeks (+7.19, p=0.329). The paper says 'The majority of the studies were deemed to have a minimal risk of bias' while its own Table 2 records unclear randomisation in 18/22 and unclear allocation concealment in 16/22; exclusion criteria still name cinnamon (template carry-over). Direction real, magnitude uninterpretable at this heterogeneity. |
| Hou LQ, Liu YH, Zhang YY 2015 · Asia Pac J Clin Nutr | meta-analysis | supports | low | Meta-analysis of 7 RCTs, n=513: 'Pooled analyses showed that garlic intake results in a statistically significant lowering in FBG [SMD=-1.67; 95% CI (-2.80, -0.55), p=0.004]'. Direction is clear, magnitude is not usable: an SMD of -1.67 from seven small trials with a CI spanning 2.25 SD indicates severe between-study dispersion, and no I-squared, risk-of-bias or publication-bias statistic is reachable at abstract grade. HbA1c and PPG were deliberately NOT pooled -- 'only 1 study included in the meta-analysis reported PPG variables and only 2 studies reported HbA1c variables' -- so this source is silent, not null, on those outcomes. Asia Pac J Clin Nutr 2015;24(4):575-582; funding/COI not seen; population composition not stated. |
| Alsanie SA et al 2026 · Diabetes Res Clin Pract | meta-analysis | supports | moderate | GRADE-assessed dose-response meta-analysis, 27 RCTs, n=1,619 adults (PROSPERO CRD420261298302, search to Jan 2026). Garlic reduced FBG (WMD -2.76 mg/dL, 95% CI -5.11 to -0.41) and HbA1c (WMD -0.53%, 95% CI -0.86 to -0.20), with no significant effect on insulin (WMD -1.18 uIU/mL, CI -2.91 to 0.55) or HOMA-IR (WMD -0.30, CI -0.72 to 0.12). Nonlinear dose-response showed GREATER FBG and insulin reductions at LOWER doses (<=0.5 g/day). Authors: 'modest improvements... further high-quality RCTs are needed'. |
| Wang J et al. 2017 · Food Nutr Res | meta-analysis | supports | low | Meta-analysis of 10 articles reporting 9 RCTs, 768 T2DM patients (430 garlic / 338 control); FBG pooled from 7 trials. Random-effects reduction in FBG at every window — 1-2 wk SMD -1.61 (-2.89, -0.32), 3-4 wk -2.87 (-4.74, -1.00), 6-8 wk -0.88 (-1.25, -0.50), 12 wk -9.57 (-12.39, -6.75), 24 wk -21.02 (-32.47, -9.57) — direction consistent in both models, but I2 = 95-99% at every window except 6-8 wk (I2 = 0%) and SMDs beyond about -3 are not physically possible, so magnitude is unusable. NOT dose- or time-dependent: 'no significant associations with both administered duration [coefficient = 0.21, 95%CI (-0.006, 0.43), P = 0.056] and dose [coefficient = 4.22, 95%CI (-3.10, 11.54), P = 0.26] were found'. Mostly an add-on effect, not garlic alone: 'only one trial adopted a monotherapy of garlic compared with placebo, while all the other trials adopted a combination therapy'. Direction supported; effect size not citable. |
| Ashraf R et al. 2011 · Pak J Pharm Sci | RCT | supports | low | Small single-center RCT (n=60 T2D on metformin): modest extra fasting-glucose reduction vs placebo over 24 wk. ↩ SUPERSEDED — pooled in the review above, counted once |
| Behrouz V, Zahroodi M, Clark CCT, Mir E, Atashi N, Rivaz R 2026 · Nutr Rev 2026 Jan 1;84(1):1-35 | meta-analysis | supports | moderate | Meta-analysis of 108 RCTs (n=7137), search to June 2024: fasting blood glucose WMD -2.77 mg/dL (95% CI -5.25 to -0.28), insulin -1.74 mU/L (95% CI -2.26 to -1.22), HOMA-IR -0.37 (95% CI -0.59 to -0.15); effects larger in adults with unfavorable baseline risk. Direction is clear but the effect is clinically negligible and the CI's lower bound (-0.28 mg/dL) is essentially null; the authors close that 'Further research involving larger sample sizes, extended follow-up periods, and enhanced methodological rigor is essential to draw definitive conclusions.' Abstract-grade: I2, risk of bias and the 108-trial list were not readable, and the pooled trials overlap the vault's existing garlic RCTs and meta-analyses. |
Disagree, or know a study we missed?
We grade by evidence, not opinions. The way to weigh in is to point us to a study we haven't cited (check the evidence table above first), or to flag a problem with one we have. Every submission is reviewed; if it holds up, the grade updates and shows in Science Changes Its Mind.
Opens a short form. You'll sign in with Google so submissions are tied to a real account — we don't display your identity, and we only accept a link we can verify (PubMed, DOI, ClinicalTrials.gov).
Educational only, not medical advice. Grades and scores reflect published evidence weighted by study design and quality; see the methodology.