Low-Carb Diets: What Replaces Carbs Matters in Type 2 Diabetes

Cutting carbohydrates can improve glucose control in type 2 diabetes, but "low-carb" hides the decision that follows: what goes on the plate instead? Fat, protein, and a mix of both do not produce identical metabolic results. The best available network analysis ranks a very-low-carbohydrate, high-protein, calorie-restricted pattern first for HbA1c reduction at six months. It also finds that the carbohydrate level associated with the largest LDL reduction sits closer to 40% of calories—not at the very-low-carb end.

That is a useful correction to the usual low-carb argument. A diet can lead the HbA1c table and still be a poor default for someone whose LDL cholesterol or cardiovascular risk needs equal attention.

The replacement macronutrient changes the story

A 2025 meta-analysis by Feng et al. pooled 149 randomized controlled trials involving 9,104 adults across 28 countries. It compared carbohydrate-restricted diet types and the macronutrients used to replace carbohydrate calories. Across all carbohydrate-restricted configurations combined, fasting glucose fell by a reported SMD of −2.94 mg/dL (95% CI, −4.19 to −1.68), fasting insulin by −8.19 pmol/L (95% CI, −11.04 to −5.43), and HOMA-IR by −0.54 (95% CI, −0.75 to −0.33).

Those pooled estimates support short-to-medium-term improvement in glycemic biomarkers, but they do not identify the effect of any one replacement strategy. The review reported directionally larger improvements with replacing carbohydrate with both fat and protein than with fat alone or protein alone across glycemic, liver, kidney, and adipokine outcomes. The publication's summary statistics do not provide the size of that combined-replacement advantage, so it would be premature to turn that pattern into a precise prescription.

The same pooled analysis found lower gamma-glutamyl transferase (GGT), a liver enzyme, by a reported SMD of −6.08 U/L (95% CI, −9.97 to −2.20), and lower urinary albumin-to-creatinine ratio (UACR) by −0.19 (95% CI, −0.28 to −0.10). These are biomarker results, not proof that a carbohydrate-restricted diet prevents liver or kidney disease. Several outcomes were standardized measures, so the reported effects cannot be translated directly into clinical thresholds such as UACR mg/g from these figures alone.

The largest effects in Feng et al.'s subgroup analyses appeared in women, people with overweight or obesity, and people with type 2 diabetes rather than non-diabetic participants. That makes the evidence especially relevant to the people most likely to consider dietary carbohydrate restriction for glucose management. It does not mean healthy people should expect the same magnitude of change.

The six-month HbA1c winner

A 2025 Bayesian network meta-analysis by Badrooj et al. compared 80 randomized trials in adults with type 2 diabetes. Rather than asking only whether low-carb diets worked, it ranked dietary configurations defined by carbohydrate intake, protein intake, and calorie restriction. By SUCRA probability—a ranking method that combines direct and indirect comparisons—the very-low-carbohydrate, high-protein, calorie-restricted configuration ranked first for HbA1c reduction at six months.

That ranking gives clinicians and patients a more specific answer than "eat fewer carbs." It does not establish a permanent winner. The glycemic advantage of low-carbohydrate strategies attenuated beyond six months in this analysis. The trials cannot cleanly separate biological adaptation from declining adherence. In real life, those explanations can overlap: a diet only works while someone can follow it.

Calorie-restricted configurations generally ranked above energy-unrestricted versions, suggesting that energy restriction adds glycemic benefit beyond macronutrient composition. The available summary does not establish the exact size of that increment. More importantly, the result leaves room for a simpler explanation of some diet rankings: the diet that produces more sustained weight loss may look metabolically superior even if its macronutrient ratio is not uniquely powerful.

For HbA1c at six months, very-low-carb, high-protein, calorie-restricted diets ranked first. For LDL reduction, the apparent sweet spot was closer to 40% carbohydrate.

LDL changes the recommendation

Badrooj et al. found a carbohydrate dose-response pattern for LDL in which the greatest LDL reduction occurred at about 40% carbohydrate intake. That sits in a moderate-carbohydrate range, not at the very-low-carb extreme that ranked highest for six-month HbA1c reduction. This is a real trade-off between surrogate outcomes, not a technicality to file under dietary fine print.

For people with type 2 diabetes and elevated LDL, established cardiovascular disease, or high cardiovascular risk, a glycemia-first ranking should not settle the dietary choice. A 2023 network meta-analysis by Bonekamp et al. found that Mediterranean and DASH dietary patterns achieved comparable or greater improvements in LDL, blood pressure, and composite cardiovascular risk markers than very-low-carbohydrate diets in people with type 2 diabetes. These comparisons emphasize different endpoints: Badrooj et al. ranked macronutrient configurations for glycemia, while Bonekamp et al. compared named dietary patterns on cardiovascular risk factors. They answer related but different questions.

Neither analysis shows that changing LDL or HbA1c by itself prevents heart attacks, kidney failure, or death. They measure intermediate outcomes. Still, LDL and blood pressure are not decorative biomarkers in type 2 diabetes, where cardiovascular disease drives much of the long-term risk.

Kidneys and adipokines: reassuring averages, uneven individuals

The lower average UACR in Feng et al. argues against a blanket claim that carbohydrate-restricted diets with more fat or protein necessarily harm kidneys. Average effects can conceal different individual trajectories, however. A 2025 post-hoc trajectory analysis by Athinarayanan et al. followed adults with type 2 diabetes receiving continuous remote care that included high-fat, very-low-carb nutritional ketosis. It identified a stable or improving majority subgroup and a minority with declining kidney-function trajectories.

This was a post-hoc latent-class analysis, not a randomized test of a ketogenic diet against another diet. It cannot establish how often the declining pattern occurs or prove that the nutritional-ketosis program caused it. Proposed beta-hydroxybutyrate mechanisms remain hypotheses. The practical point is simpler: people with pre-existing kidney impairment should monitor eGFR and UACR when adopting a high-fat, very-low-carb protocol.

Fat-dominant carbohydrate replacement also affects adipokines, though the picture is less tidy than internet nutrition discourse prefers. A 2024 systematic review of ketogenic-diet randomized trials by Xiao et al. found lower leptin with variable adiponectin responses across populations. A 12-week non-randomized study by Cipryan et al. reported a higher adiponectin-to-leptin ratio in healthy participants following a very-low-carbohydrate, high-fat diet plus structured exercise. Because exercise occurred alongside the diet and participants were not randomized, that study cannot isolate a diet effect. Adipokine changes are interesting biological signals, not a reason to declare a ketogenic diet anti-inflammatory medicine.

Weight loss may explain more than the ranking does

The sharpest challenge to a composition-first story comes from a 2022 umbrella review by Churuangsuk et al.. Its conclusion was that sustained weight loss greater than 10 kg is the dominant predictor of type 2 diabetes remission, with several dietary approaches producing broadly comparable metabolic improvement. Under that interpretation, a very-low-carb, high-protein, calorie-restricted pattern may rank well partly because it helps some people lose more weight during the first six months.

The evidence does not resolve how much benefit comes directly from macronutrient composition versus weight loss, calorie reduction, and adherence. Feng et al. reported analyses of isocaloric and non-isocaloric carbohydrate restriction, but the available findings do not quantify a clean composition-only effect for the combined fat-and-protein strategy. Network rankings also rely partly on indirect comparisons among trials, and the exact credible intervals for specific dietary nodes require the full supplementary analyses.

What to do with this evidence

For an adult with type 2 diabetes whose immediate priority is HbA1c, the strongest ranking evidence favors a very-low-carbohydrate, high-protein, calorie-restricted approach for the first six months. Build the reduced-carbohydrate diet with both fat and protein rather than treating either as the universal replacement, because the 149-trial synthesis found the most favorable directional pattern with the combined strategy.

• Track HbA1c, body weight, adherence, and medications, especially if glucose-lowering drugs can cause hypoglycemia.
• Check LDL before and after a very-low-carb approach; a moderate-carbohydrate or Mediterranean-style pattern may offer a better overall trade-off when cardiovascular risk is high.
• Monitor eGFR and UACR when kidney disease is present, particularly with high-fat ketogenic protocols.
• Choose a pattern that can sustain meaningful weight loss. A six-month ranking does not compensate for abandoning the diet in month seven.

The evidence supports carbohydrate restriction as a useful tool for type 2 diabetes, not a one-size-fits-all identity. The best plate depends on which outcome needs the most help—and whether the person can keep eating that way long enough for the result to matter.