Does a low-carb diet improve blood sugar and insulin resistance?

Yes, in people with type 2 diabetes, by a modest amount that shrinks over time. A 2021 meta-analysis of randomized trials found low-carb diets lowered HbA1c by 0.47 percentage points more than control diets at six months, in 747 people, on high-certainty evidence. By 12 months the gap had roughly halved, to 0.23 points, with an interval that reaches zero. In trials of adults with and without diabetes, cutting carbohydrate also lowered the insulin resistance score HOMA-IR by 0.54.

Established. Over the first six months, a low-carb diet lowers HbA1c in adults with type 2 diabetes more than the comparison diets did, by about half a percentage point, on high-certainty evidence from a meta-analysis of randomized trials.

Unsettled. Whether the benefit lasts is the open part. A 2024 meta-analysis of the six trials that ran longer than 12 months found no significant difference in HbA1c, and in the 2021 review fewer people kept to their carbohydrate targets as time went on.

What the trials found

HbA1c in type 2 diabetes

The 2021 meta-analysis in the BMJ pooled 23 randomized trials in 1357 people, counting a diet as low-carb under 130 g a day or under 26 percent of energy. At six months HbA1c fell 0.47 points more than on control diets, an interval from 0.34 to 0.60. At 12 months, in 8 studies and 489 people, the difference was 0.23 points, an interval from 0.46 down to zero, rated moderate certainty.

A 2025 meta-analysis of 27 randomized trials in 2,870 adults with type 2 diabetes found a smaller effect. HbA1c fell 0.29 points and fasting glucose by 7.12 mg/dL, with the largest effect at three months and benefits fading after that. Its abstract gives no confidence intervals. None of its diets held carbohydrate below 10 per cent of energy, so it describes low and moderate cuts, not very low-carb eating.

The 2024 meta-analysis of trials longer than 12 months found HbA1c 0.11 standard deviations lower on low-carb, an interval from 0.33 lower to 0.11 higher, which is not significant. It pooled only six trials, and the abstract does not give the number of people.

Remission

In the 2021 review, 57 percent of people on low-carb diets reached an HbA1c below 6.5 percent at six months, against 31 percent on control diets, a difference of 32 people in 100, in 264 people across 8 studies. When remission also required being off diabetes medicine, the effect was smaller and not significant. Remission fell markedly in trials that included people on insulin.

The strongest diet in the network

A 2025 network meta-analysis of 80 trials in 9,232 people with type 2 diabetes ranked a very low-carb, high-protein diet that also cut calories first for HbA1c. Against the various comparators it was 1.0 to 1.79 points lower at six months, and 1.25 to 1.30 points lower at 12 months. Calorie restriction was part of that diet, so the effect of eating less carbohydrate cannot be separated from the effect of eating less and losing weight.

Insulin resistance

A 2025 meta-analysis of 149 randomized trials in 9,104 adults in 28 countries, with and without diabetes, found carbohydrate-restricted diets lowered HOMA-IR by 0.54, an interval from 0.33 to 0.75. The effect was largest in women, people with overweight or obesity and people with type 2 diabetes, and low- and moderate-carb diets gave the most consistent results. People without diabetes are pooled with people who have it, so this does not tell you the effect in a healthy adult on its own.

The fiber counterpoint

A 2025 meta-analysis of 10 randomized trials in 499 people with diabetes, almost all type 2, compared diets that changed carbohydrate and fiber together. Diets higher in both lowered HbA1c by 0.50 points more than diets lower in both, an interval from 0.02 to 0.99, on moderate or high certainty evidence. Trials that cut calories were left out. The kind of carbohydrate matters as well as the amount.

Type 1 diabetes

In a six-month open-label randomized trial of 40 young people with type 1 diabetes, HbA1c fell 0.7 points on a low-carb diet against 0.1 on a Mediterranean diet. Time spent in the target glucose range did not differ. It is one small trial.

Who should be careful

Not for everyone. If you take an SGLT2 inhibitor, a class of diabetes drugs that includes empagliflozin, the American Diabetes Association says very low-carb eating should be avoided because of the risk of ketoacidosis. One case report describes a 58-year-old man on empagliflozin who developed severe ketoacidosis needing intensive care one month after starting a ketogenic diet. It is a single case, and the caution rests on reports like it.

If you take insulin or other glucose-lowering drugs, the ADA says very low-carb eating needs regular medical oversight, because the doses may need lowering to prevent hypoglycemia, and blood pressure needs monitoring.

The ADA does not currently recommend very low-carb eating for people who are pregnant or breastfeeding, for children, for people with kidney disease, or for people with or at risk of disordered eating. That is a position without its own analysis.

What expert bodies say

The American Diabetes Association's 2026 Standards of Care advise considering less carbohydrate for some adults with diabetes to improve blood sugar, and name limiting processed foods as an effective way to do it. The recommendation is graded B. The ADA itself notes that the HbA1c benefit is short term, under six months, with less difference beyond a year.

How we searched

What would change this answer

• Trials of two years or more in which people actually keep carbohydrate below 26 percent of energy, reporting HbA1c and remission without medication.
• A head-to-head trial of a low-carb diet against a high-fiber diet with the same weight loss, to separate the amount of carbohydrate from its quality and from weight.
• Trials of insulin resistance in adults without diabetes reported on their own.

Sources

1. ev-lcbs-01 · Meta-analysis or systematic review · systematic review and meta-analysis of 23 RCTs (1357 participants), GRADE · n = 747
   LCDs achieved greater reductions in HbA1c than did control diets (mean difference –0.47%, –0.60 to –0.34; n=747; GRADE=high) (table 2). At 12 months, eight studies reported on HbA1c levels, showing that the effect size had decreased by around half (mean difference –0.23%, –0.46% to 0.00%; n=489; GRADE=moderate) (table 2).
   
   Who: adults with type 2 diabetes; LCD under 130 g/day or under 26% of energy, at least 12 weeks
   Effect: HbA1c MD -0.47% (-0.60 to -0.34), 17 studies, GRADE high, at 6 months; MD -0.23% (-0.46 to 0.00), 8 studies, n=489, GRADE moderate, at 12 months
   Certainty: Effect at 12 months borderline; adherence to carbohydrate targets fell over time.
   BMJ, 2021 · checked 2026-10-07 · we read the fulltext
2. ev-lcbs-02 · Meta-analysis or systematic review · systematic review and meta-analysis of RCTs · n = 264
   At six months, compared with control diets, LCDs achieved higher rates of diabetes remission (defined as HbA1c <6.5%) (76/133 (57%) v 41/131 (31%); risk difference 0.32, 95% confidence interval 0.17 to 0.47; 8 studies, n=264, I2=58%). Conversely, smaller, non-significant effect sizes occurred when a remission definition of HbA1c <6.5% without medication was used.
   
   Who: adults with type 2 diabetes in RCTs of at least 12 weeks
   Effect: remission (HbA1c <6.5%) 57% vs 31%, risk difference 0.32 (0.17 to 0.47), 8 studies, I2 58%; smaller, non-significant when defined without medication
   Certainty: Remission with LCDs fell markedly in trials including insulin users; 12-month remission data sparse.
   BMJ, 2021 · checked 2026-10-07 · we read the abstract
3. ev-lcbs-03 · Meta-analysis or systematic review · systematic review and meta-analysis of RCTs with meta-regression · n = 2870
   LCDs improved glycemia short-term: HbA1c - 0.29 %, with the largest effect at 3 months. Fasting blood glucose decreased overall (-7.12 mg/dL).
   
   Who: adults with type 2 diabetes, RCTs of 12 weeks or more, 7 Eastern and 20 Western
   Effect: HbA1c -0.29%; fasting blood glucose -7.12 mg/dL; benefits diminish over time; no arm sustained carbohydrate below 10% of energy
   Certainty: Confidence intervals not in the abstract; findings apply to low-to-moderate, not very-low, carbohydrate intake.
   Diabetes Res Clin Pract, 2025 · checked 2026-10-07 · we read the abstract
4. ev-lcbs-04 · Meta-analysis or systematic review · systematic review and meta-analysis of RCTs · n = 6 RCTs
   This study did not show significant differences in changes in glycated hemoglobin between long-term LCD and control diets (standardized mean difference -0.11, 95% confidence interval -0.33 to 0.11, P = 0.32).
   
   Who: adults with type 2 diabetes, LCD interventions longer than 12 months
   Effect: HbA1c SMD -0.11 (95% CI -0.33 to 0.11), P=0.32; HDL up (SMD 0.22) and triglycerides down (SMD -0.19)
   Certainty: Only six trials; participant total not in abstract.
   J Diabetes Investig, 2024 · checked 2026-10-07 · we read the abstract
5. ev-lcbs-05 · Meta-analysis or systematic review · Bayesian network meta-analysis of RCTs, GRADE · n = 9232
   A very low-carbohydrate, high-protein, and calorie-restricted diet had the greatest effect on reducing HbA1c (range of mean difference: - 1.0% to - 1.79%), weight (range of mean difference: -5.83 kg to -10.96 kg), and FPG (range of mean difference: - 2.20 mmol/L to - 2.88 mmol/L) at 6-month follow-up, but at 12-month follow-up, the effect remained only for HbA1c (range of mean difference: - 1.25% to - 1.30%) and FPG (range of mean difference: - 1.21 mmol/L to - 1.27 mmol/L).
   
   Who: patients with type 2 diabetes in RCTs of exact macronutrient compositions
   Effect: very low-carb high-protein calorie-restricted: HbA1c MD -1.0% to -1.79%, FPG -40 to -52 mg/dL (-2.20 to -2.88 mmol/L) at 6 months; HbA1c -1.25% to -1.30% at 12 months
   Certainty: Calorie restriction is part of the winning arm, so carbohydrate cut and weight loss are mixed; ranges are versus several comparators.
   J Health Popul Nutr, 2025 · checked 2026-10-07 · we read the abstract
6. ev-lcbs-06 · Meta-analysis or systematic review · systematic review and meta-analysis of RCTs · n = 9104
   CRDs significantly improved glycemic control (including glucose: SMD = -2.94 mg/dL, 95 % CI: -4.19, -1.68; insulin: SMD = -8.19 pmol/L, 95 % CI: -11.04, -5.43; HOMA-IR = -0.54, 95 % CI: -0.75, -0.33), hepatic stress (GGT: SMD = -6.08 U/L, 95 % CI: -9.97, -2.20), renal function (UACR: SMD = -0.19, 95 % CI: -0.28, -0.10), and adipokine concentration (leptin: SMD = -3.25 ng/mL, 95 % CI: -4.91, -1.59), particularly in females, individuals with overweight/obesity, and people with T2DM.
   
   Who: adults in 28 countries, with and without diabetes; ketogenic, low- and moderate-carbohydrate diets
   Effect: HOMA-IR -0.54 (95% CI -0.75 to -0.33); glucose SMD -2.94 mg/dL (-4.19 to -1.68); insulin -8.19 pmol/L (-11.04 to -5.43)
   Certainty: Glucose and insulin reported as SMD with units in the abstract; low- and moderate-carb diets most consistent.
   Clin Nutr, 2025 · checked 2026-10-07 · we read the abstract
7. ev-lcbs-07 · Meta-analysis or systematic review · systematic review and meta-analysis of RCTs, GRADE · n = 499
   Pooled findings indicate that higher fiber higher carbohydrate diets reduced HbA1c (mean difference [MD] -0.50% [95% confidence interval -0.99 to -0.02]), fasting insulin (MD -0.99 μIU/mL [-1.83 to -0.15]), total cholesterol (MD -0.16 mmol/L [-0.27 to -0.05]) and low-density lipoprotein cholesterol (MD -0.16 mmol/L (-0.31 to -0.01) when compared with lower carbohydrate lower fiber diets.
   
   Who: people with diabetes, 98% type 2, RCTs changing both fiber and carbohydrate
   Effect: HbA1c MD -0.50% (-0.99 to -0.02), fasting insulin -0.99 uIU/mL, LDL -6.2 mg/dL (-0.16 mmol/L) for higher fiber higher carbohydrate; moderate or high certainty
   Certainty: Excluded trials with energy restriction; shows fiber quality matters as much as carbohydrate amount.
   Obes Rev, 2025 · checked 2026-10-07 · we read the abstract
8. ev-lcbs-08 · Randomized controlled trial(s) · open-label randomized controlled trial, 6 months, CGM outcomes · n = 40
   The delta HbA1c improved in both groups: -0.7% vs. -0.1% (p = 0.02).
   
   Who: adolescents and youths with type 1 diabetes, median age 18, 70% female
   Effect: delta HbA1c -0.7% vs -0.1% (p=0.02); time in range similar (p=0.98); time <54 mg/dL comparable
   Certainty: Small, open-label; time in range did not differ.
   Acta Paediatr, 2025 · checked 2026-10-07 · we read the abstract
9. ev-lcbs-09 · Position of an expert body · clinical practice guideline (ADA Standards of Care 2026, section 5) · n = —
   5.15 Consider reducing carbohydrate intake for some adults with diabetes to improve glycemia. An effective way to achieve this is by limiting consumption of processed foods.
   
   Who: adults with diabetes
   Effect: recommendation 5.15, evidence grade B
   Certainty: ADA notes the A1C benefit is short term (<6 months) with less difference beyond 1 year.
   American Diabetes Association Professional Practice Committee, 5. Facilitating Positive Health Behaviors and Well-being to Improve Health Outcomes: Standards of Care in Diabetes-2026, Diabetes Care 2026;49(Suppl 1):S89-S131 · checked 2026-10-07 · we read the section
10. ev-lcbs-10 · Position of an expert body · clinical practice guideline (ADA Standards of Care 2026, section 5) · n = —
    Very-low-carbohydrate eating plans should be avoided in those taking sodium–glucose cotransporter 2 (SGLT2) inhibitors because of the potential risk of ketoacidosis (105,106).
    
    Who: people with diabetes taking SGLT2 inhibitors
    Effect: avoid very-low-carbohydrate plans; recommendation 5.26 discourages ketogenic eating with SGLT inhibition
    Certainty: Based on case reports of diabetic ketoacidosis with SGLT2 inhibitors plus ketogenic diets.
    American Diabetes Association Professional Practice Committee, 5. Facilitating Positive Health Behaviors and Well-being to Improve Health Outcomes: Standards of Care in Diabetes-2026, Diabetes Care 2026;49(Suppl 1):S89-S131 · checked 2026-10-07 · we read the section
11. ev-lcbs-11 · Position of an expert body · clinical practice guideline (ADA Standards of Care 2026, section 5) · n = —
    In addition, very-low-carbohydrate eating plans are not currently recommended for individuals who are pregnant or lactating, children, people who have kidney disease, or people with or at risk for disordered eating (46).
    
    Who: pregnant or lactating people, children, people with kidney disease or disordered eating
    Effect: very-low-carbohydrate plans not currently recommended
    Certainty: Position without its own analysis.
    American Diabetes Association Professional Practice Committee, 5. Facilitating Positive Health Behaviors and Well-being to Improve Health Outcomes: Standards of Care in Diabetes-2026, Diabetes Care 2026;49(Suppl 1):S89-S131 · checked 2026-10-07 · we read the section
12. ev-lcbs-12 · Position of an expert body · clinical practice guideline (ADA Standards of Care 2026, section 5) · n = —
    Health care professionals should maintain consistent medical oversight of individuals following very-low-carbohydrate eating plans and recognize that insulin and other diabetes medications may need to be adjusted to prevent hypoglycemia, and blood pressure will need to be monitored.
    
    Who: people with diabetes on glucose-lowering drugs following very-low-carbohydrate plans
    Effect: adjust insulin and other diabetes medications; monitor blood pressure
    Certainty: Position without its own analysis.
    American Diabetes Association Professional Practice Committee, 5. Facilitating Positive Health Behaviors and Well-being to Improve Health Outcomes: Standards of Care in Diabetes-2026, Diabetes Care 2026;49(Suppl 1):S89-S131 · checked 2026-10-07 · we read the section
13. ev-lcbs-13 · Laboratory or animal data · case report · n = 1
    Literature investigation revealed that a strict low carbohydrate diet can rarely lead to DKA in the setting of SLGT-2 inhibitor use.
    
    Who: 58-year-old man with type 2 diabetes on an SGLT2 inhibitor
    Effect: severe DKA needing ICU care; no other precipitant found
    Certainty: Single case; illustrates the ADA caution.
    Clin Med Insights Case Rep, 2022 · checked 2026-10-07 · we read the abstract