Keto and Low-Carb Diets With Type 2 Diabetes: What the Research Shows and What to Ask First

Why this question matters

Few food topics in diabetes get as heated as carbohydrate. Some people say cutting carbs changed their numbers and their lives. Others say a very strict plan is unnecessary, hard to keep up, or risky if you take certain medicines. Both groups can point to something real, which is why a plain look at the evidence is more useful than picking a side.

This guide is for adults with type 2 diabetes or prediabetes who are curious about low-carbohydrate and ketogenic eating, and for the people who cook for them. It explains what “low-carb” means in different sources, what the research can and cannot say, what the safety questions are, and what a day of low-carb eating adds up to in grams. It is education, not a treatment plan. I am not a physician, and nothing here replaces advice from the person who prescribes your medicines.

It is one of several diet guides on this site. For the bigger picture, see diabetes diets compared, the Mediterranean diet guide, the plant-based diet guide, and our look at intermittent fasting. The general starting point is how to build meals around your numbers.

What counts as “low-carb” and “keto”?

There is no single agreed definition, and that is the first thing to know before reading any headline. A meta-analysis published in Frontiers in Nutrition in 2025 says outright that trials vary widely in how they define the diets, how they measure whether people followed them, and how long they last. A 2022 American Heart Association news summary of a large observational study makes the same point: there is no standard definition of a low-carb diet.

Several sources do give numbers. A 2021 review written for the International Working Group on Remission of Type 2 Diabetes, also in Frontiers in Nutrition, sorts patterns by grams of carbohydrate per day. A family-medicine journal summary I read describes studies using under 26 percent of calories from carbohydrate as “low” and under 10 percent as “very low.” The 2025 meta-analysis describes the American Diabetes Association as treating under 130 grams a day, or under 26 percent of energy, as low-carb. I could not open the original American Diabetes Association report to confirm that wording, so treat that line as how one paper describes it.

These are definitions used in the papers I read, not targets. Percent and gram figures do not line up exactly because they depend on total calories. Different authors draw the lines in different places.

To make that concrete: on a 2,000-calorie day, 130 grams of carbohydrate is about 26 percent of calories, because carbohydrate has 4 calories per gram (130 × 4 = 520 calories, and 520 ÷ 2,000 = 26 percent). Ten percent of those calories would be about 50 grams. So, at that calorie level, the percent-based “very low” line and the 50 grams-a-day line describe roughly the same thing. That is simple arithmetic, not a recommendation, and it changes if you eat more or less than 2,000 calories.

Words like “keto,” “nutritional ketosis,” and “low-carb, high-fat” are used loosely online. Strictly, ketosis means your body is producing ketones from fat as a fuel because carbohydrate and insulin are low. The 2021 review draws the line this way: ketones from eating very little carbohydrate are a normal physiological state, while diabetic ketoacidosis is a dangerous condition in which ketones and acid build up in the blood. The two are not the same thing, but they are related, which is why the safety section below matters.

What the research shows

The big picture from reviews

The most useful single source I found is a 2025 systematic review and meta-analysis in Frontiers in Nutrition. It pooled 17 randomized trials with 1,197 adults who had type 2 diabetes and were overweight or obese: 643 on a low-carb diet and 554 in comparison groups. The trials lasted from 3 to 24 months. Compared with control diets, the low-carb groups had, on average:

• HbA1c lower by 0.36 (95% confidence interval 0.44 to 0.29 lower), across 15 studies.
• Fasting glucose lower by about 10.7 (units as the paper reports them), across 10 studies.
• Triglycerides lower and HDL cholesterol higher.
• Body weight lower by about 3.7 units, though the studies disagreed with each other a good deal on this one.
• No statistically significant change in LDL cholesterol or total cholesterol on average.

Those are averages across many trials. They say that, on average, lower-carb eating did better than the comparison diets over the periods studied. They do not say that any one person will see the same change, and an average LDL result that is not significant can still hide people whose LDL went up or down. The authors also note that they planned to compare results by trial length but did not report pooled numbers by time point, so this paper alone cannot tell you how the picture changes from three months to two years.

What happens over time

A second summary, in a family-medicine journal, covers an earlier meta-analysis of 23 small randomized studies, with 8 of them (264 patients) measuring diabetes remission. It reported that at 6 months, low-carbohydrate diets produced remission (defined there as HbA1c below 6.5 percent or fasting glucose below 126 mg/dL, with or without medication) in about 32 more people per 100 treated, which works out to treating about 4 people for one extra remission. By 12 months, the differences in weight and triglycerides had disappeared, and the summary said longer-term data were not available. It also rated most of the studies as having moderate to high risk of bias. Note that “remission” in this context is a lab threshold in a study, not a promise of staying off medicine.

The 2025 paper makes the same time pattern its own point. In its introduction, citing other reviews, it says reductions in HbA1c at 6 to 12 months tend to fade after 12 to 24 months, and in the discussion it says early weight loss is partly water and glycogen, and that weight benefits may fade over six months to a year. The authors did not test that claim directly in their data, so I am reporting it as what they wrote, not as a finding from their own analysis. It matches the other review, which is a reason to take it seriously and also a reason to be cautious: both are describing the same limits of the underlying trials.

The best-known single study, and who ran it

If you have heard that low-carb can “reverse” diabetes, you have probably heard of the Virta Health study. Its one-year results were published in Diabetes Therapy in 2018. It followed 349 adults with type 2 diabetes. A total of 262 took part in a program built around individualized low-carbohydrate, high-fat eating aimed at nutritional ketosis, with remote coaching and a physician through a mobile app, and 87 comparison participants received usual care from their own doctors and dietitians. The comparison group was not randomly assigned, and the study was open-label.

According to a university news release about the paper, the program group lowered average HbA1c by 1.3 percentage points, lost about 12 percent of body weight, and 83 percent stayed in the program for the year. It also reported that of the program participants on insulin, 94 percent reduced or stopped their insulin, and about 60 percent had HbA1c below the diabetes threshold while taking no medication or metformin only. Usual care showed no significant change. The same release states that Virta Health, which sells the treatment, compiled the study and that its co-founders are among the authors.

None of that makes the results fake, and the numbers are striking. But it is the kind of evidence to hold more loosely than a blinded randomized trial from independent researchers: the people who developed and sell the program ran and reported it, participants chose the program, and the one-year results come from an ongoing five-year study. I read the news release, not the full paper, so I cannot speak to the study’s own limitations section.

All figures as reported in the sources named. Numbers from news summaries may differ from the full papers.

What the evidence does not tell us

Reading across these sources, a few gaps stand out, and they matter for how much weight to put on any one result.

Long-term outcomes. Most trials run for months, not years. The summaries I read say benefits shrink as time passes and long-term data are limited. They do not tell us whether the people who keep eating very few carbs for five or ten years have fewer heart attacks or complications. The 2024 American Diabetes Association Standards of Care, as summarized in a Glycemic Index Foundation article, say there is some evidence for lower-carbohydrate eating, particularly in type 2 diabetes, but not enough evidence to support an optimal amount of carbohydrate, and that questions remain about long-term cardiovascular effects.

Head-to-head comparisons. Many comparisons are against usual care or a standard higher-carb diet. Fewer compare low-carb directly with other well-supported patterns such as the Mediterranean diet or a plant-based diet. A summary of the American Diabetes Association’s 2019 nutrition consensus report says the evidence most strongly supports reducing overall carbohydrate for blood sugar, that the association does not recommend a specific amount of carbohydrate, protein or fat, and that low and very low carbohydrate patterns sit on the list of options with research support. That is a different message from “low-carb is the best diet,” and it is also different from “low-carb is not worth considering.”

Who gets studied. The 2025 meta-analysis covered people who were overweight or obese. People of normal weight, people with kidney disease, and people who are older or frail are less well represented in that evidence, and I did not find a source that settles those cases.

Adherence. The family-medicine summary says that for a very low-carbohydrate diet, 50 to 80 percent of participants could follow it even with behavioral support. The Cleveland Clinic page I read says the diet is hard to stick with long term and that research on long-term effectiveness and risks is still early. A plan that works on paper but that you cannot live with is not a plan.

Quality matters: what the carbs and fats are made of

A diet that is low in carbohydrate is still a diet made of something. A preliminary 2022 analysis of the Nurses’ Health Study, Nurses’ Health Study II and the Health Professionals Follow-Up Study, reported by the American Heart Association, looked at 203,541 adults who did not have diabetes at the start and were followed for up to 30 years. Among people eating the lowest-carbohydrate diets in that group (still about 40 percent of calories from carbohydrate, so this is not keto), eating more plant-based protein and fat was associated with a 6 percent lower risk of developing type 2 diabetes, and also limiting sugar and refined carbohydrate was associated with a 15 percent lower risk. Eating mostly animal protein and fat was associated with a 35 percent higher risk, and 39 percent higher if whole grains were also low.

Please read that carefully. It is an observational study, so it can show an association and cannot show cause. It is about who developed diabetes, not about managing it once you have it. The authors did not study people who consistently ate very-low-carb or keto diets. And the findings were preliminary when presented. What I take from it is modest: two diets with the same carbohydrate number can be very different, so the number alone does not tell you what you are eating. That idea shows up again in the Mediterranean and plant-based guides.

Safety: the part to read before you change anything

Low-carbohydrate eating can lower blood sugar. That is the appeal, and it is also the reason it needs a plan if you take medicines that lower blood sugar. Three safety topics come up repeatedly in the sources I read.

1. Low blood sugar with certain medicines

The Cleveland Clinic page on keto and diabetes says the risk of low blood sugar is higher for people taking certain diabetes medications or insulin, and lists warning signs such as shakiness, sweating, dizziness, confusion and blurred vision. The 2021 review for the International Working Group says the main hypoglycemia risk comes from insulin, sulfonylureas and meglitinides, and describes medication reductions that clinicians have made when patients start low-carb eating. Diabetes Australia says people starting low-carb eating should monitor their glucose and talk to their doctor about whether to adjust diabetes medication to reduce the risk of hypoglycemia. In the 2025 meta-analysis, there were 14 reports of hypoglycemic symptoms and 7 documented hypoglycemic events across the trials, with no severe hypoglycemia.

I am not going to give medicine doses or adjustment steps, and I would be wary of any website that does. The point is that this is a plan between you and whoever prescribes your medicine, made before you cut carbs and checked as you go.

2. SGLT2 inhibitors and ketoacidosis

SGLT2 inhibitors are a class of diabetes medicines that also protect the heart and kidneys in some people. An educator article that cites the American Diabetes Association’s 2024 Standards of Care says very low-carbohydrate diets may increase the risk of euglycemic ketoacidosis by lowering insulin levels, and advises caution when combining them. “Euglycemic” means blood glucose can look normal or only mildly raised while ketoacidosis develops, which can make it easy to miss. That article says that for diets under about 50 grams of carbohydrate a day, stopping the SGLT2 inhibitor should be strongly considered, and that a very low-carb diet can be considered after transitioning off it, following further clinical review. That is a decision for the prescriber, not something to do on your own, and I have no way to know your situation. The 2021 review adds that a low-carbohydrate diet alone cannot cause ketoacidosis in this way, but that SGLT2 inhibitors can, and that lower insulin levels on a low-carb diet may add to that risk.

The Cleveland Clinic page adds that ketoacidosis is mainly a concern in type 1 diabetes but can also occur in type 2 diabetes when it is not well managed. I could not find a source I could open that gave a complete sick-day or ketone-testing protocol, so I will not invent one. The sensible step is to ask your clinician, before you start, what to do on days you are ill or not eating normally, and whether ketone testing makes sense for you. In the Virta study, participants logged glucose and ketone levels using the program’s tools, which tells you that some programs build monitoring in from the start.

3. Who should ask a clinician first

Diabetes Australia states that low-carb eating is not recommended for children, or for pregnant or breastfeeding women, and advises against it for people with certain medical conditions or histories. It also strongly encourages speaking with an experienced low-carb dietitian, doctor or other health professional. The Cleveland Clinic page says the same in other words: talk to your doctor or a registered dietitian before starting, and do not begin on your own.

Beyond what those sources say, there are other groups where I would expect a clinician to want to be involved. I did not find an authoritative page I could open that spells these out, so please read this as general caution, not sourced rules: people with kidney disease, because protein, fluid and medicine choices can all change; people with a history of an eating disorder, because a strict food rule can be a trigger; people with a history of pancreatitis or gallbladder or liver problems; and anyone on blood pressure medicine, because blood pressure can fall as weight and sodium change. The 2021 review does note that antihypertensive medicines may need reducing if symptoms of low blood pressure appear, and that people with LADA or MODY (other forms of diabetes that can be mistaken for type 2) should not have insulin stopped completely.

Cholesterol, side effects and nutrition

Lipid results in the 2025 meta-analysis were mostly favorable on average (lower triglycerides, higher HDL), with LDL not significantly different. One participant on a low-carb, high-fat diet in an included trial developed severe hypertriglyceridemia and withdrew, and the authors note early mild stomach and gut symptoms in some people. The same paper says very low-carbohydrate diets may not be ideal for some people with diabetes and may affect lipids unfavorably in this group, citing one included trial. Because averages hide individuals, a cholesterol panel before and a few months after a big diet change is a reasonable thing to ask your clinician about. The Cleveland Clinic page also lists constipation and the possibility of missing nutrients as risks. Online, a cluster of early symptoms such as tiredness, headache and irritability is often called the “keto flu.” I did not find a source I could read that quantifies it, so I mention it only so you recognize the term.

What a low-carb day adds up to in grams

Numbers are easier to think about when you can see them. The table below shows carbohydrate and fiber for common low-carb foods, from USDA FoodData Central records as shown on the public mirror getfoodfacts.com (which lists the FDC ID for each). I scaled gram weights to the portion shown. These are the foods that appear in the example days that follow.

Source: USDA FoodData Central values as displayed by getfoodfacts.com, accessed October 31, 2026. Portion carbs are my arithmetic from per-100 g values and the portion weight on the page; the avocado is shown per 100 g because I did not find a verified gram weight for a whole avocado. Spinach per cup rounds to about 1 g.

I put together three simple example days, built only from the foods in that table, so every number is traceable. They are arithmetic illustrations of what different carbohydrate levels look like, not meal plans, and not advice on how much you should eat.

• Roughly 50 g day: 3 eggs; 2 cups raw spinach; 1 cup broccoli; 1 oz almonds; 100 g avocado; 2 oz cheddar; 1 cup chopped chicken; 1 cup blueberries. Total carbohydrate about 51.4 g, fiber about 18.9 g, so about 32.5 g net if you subtract fiber.
• Roughly 100 g day: the same foods plus 1 cup of cooked white rice. Total about 96.0 g, fiber about 19.5 g, about 76.5 g net.
• Roughly 130 g day: the 100 g day plus another half cup of rice and another half cup of blueberries. Total about 129.0 g, fiber about 21.6 g, about 107.4 g net.

Two lessons come out of the arithmetic. First, a small amount of fruit or starch moves the total a lot: one cup of blueberries (about 21.5 g) is more carbohydrate than the eggs, spinach, cheese, almonds and avocado combined. One cup of cooked white rice (about 44.6 g) is close to the whole 50 g day by itself. Second, “net carbs” (total minus fiber) is a label and tracking convention, not a lab-defined number. How a given person’s glucose responds to fiber-rich foods can differ, which is one reason checking your own meter or sensor is more informative than trusting a calculation. If you want to see how labels handle this, read how to read nutrition label carbs, and note what the sweeteners guide says about sugar alcohol math on “keto” products and what the bread guide says about low-carb loaves.

How this applies in four common situations

If you have prediabetes

The evidence I found most directly is about adults who already have type 2 diabetes. For prediabetes, the sources I read are mostly observational or about preventing diabetes, such as the AHA-reported study above, which suggests the quality of a lower-carb diet may matter. Because lower-carb eating is one of several options the American Diabetes Association lists as having research support, it is reasonable to ask your clinician or a dietitian whether it fits you, and to compare it with the options in the diets compared guide.

If you have type 2 diabetes and take no diabetes medicine

This is the group where the medication risks are lowest, because there is no medicine pushing blood sugar down on its own. It is also the group the published trials most often describe as seeing improvement. A reasonable approach is still to tell your clinician what you plan, ask whether any lab checks make sense (a cholesterol panel, kidney function and HbA1c are common to ask about), and set a review date. If you want to see your own response, a before-and-after approach with the post-meal glucose tracker can help you collect it.

If you take insulin, a sulfonylurea or an SGLT2 inhibitor

Treat the clinician conversation as the first step, not an extra. The sources I read agree that the main risks, low blood sugar and ketoacidosis, are tied to these medicines, and that doses may need to change when carbohydrate falls. Do not reduce or stop a medicine yourself. Ask the prescriber how they want you to monitor, which symptoms mean call right away, and what to do on days you are sick or not eating.

If you use a continuous glucose monitor

A sensor can show how a day of lower-carb eating looks on your own graph, including overnight lows that a few daily finger-sticks would miss. It does not replace a clinician’s plan for medicines. Our guide to reading CGM data explains what time in range and the shape of the line can and cannot tell you. For very-low-carb eating, remember that a sensor measures glucose, not ketones, so it will not show the ketoacidosis concern described above.

Common mistakes

• Changing diet and medicine at once without a plan. The sources I read link the risk to the combination, not to the diet alone.
• Treating “keto” labels as proof of low carbohydrate. Marketing terms are not regulated like nutrition facts. Read the panel.
• Counting only “net carbs.” Net carbs is a convention, and sugar alcohol and fiber subtraction rules vary. Use your own meter or sensor to see what a food actually does for you.
• Swapping carbs for anything low in carbs. The AHA-reported study shows why the source of protein and fat may matter. Low-carb bacon is not the same dietary pattern as low-carb fish, nuts and vegetables.
• Judging the diet by the first two weeks. Early weight loss is partly water. The reviews I read describe the benefits shrinking with time.
• Skipping labs. Cholesterol and kidney function are reasonable to recheck after a big change, especially with a high-fat plan.
• Making it a pass-fail test. A strict plan that you cannot keep is not better than a moderate one you can.

Myth versus fact

Questions to bring to your clinician or dietitian

1. Which of my medicines can cause low blood sugar if I eat fewer carbohydrates, and what is the plan if my numbers drop?
2. Do I take an SGLT2 inhibitor? If so, what does a lower-carb plan mean for the risk of ketoacidosis, and when should I call you?
3. What should I do on sick days or days when I cannot eat normally?
4. Does ketone testing make sense for me, and if so, when?
5. Which labs should we check before I start and again in a few months (cholesterol, kidney function, HbA1c)?
6. Is my kidney function or any other condition a reason to choose a different eating pattern?
7. Is there a history in my family or my own that makes a strict eating plan a poor idea?
8. How low should I go, for how long, and how will we decide whether to continue?
9. Can you refer me to a dietitian who has experience with lower-carbohydrate eating?
10. How would you compare this with a Mediterranean or plant-based pattern for me?

Frequently asked questions

Is a keto diet safe for type 2 diabetes?

It depends on you and your medicines. Reviews describe improvements in blood sugar, weight and triglycerides in many people, but risks include low blood sugar with insulin or sulfonylureas and ketoacidosis with SGLT2 inhibitors. Cleveland Clinic and Diabetes Australia both advise involving a doctor or dietitian before starting.

What counts as low-carb for diabetes?

Definitions differ. One 2021 review uses about 30 g or less per day for very low-carb ketogenic, 30 to 50 g for low-carb ketogenic, and 50 to under 130 g for reduced-carbohydrate. Other papers use percent of calories, with under 26 percent called low. Check which definition a study uses.

Does the American Diabetes Association recommend low-carb?

As summarized in the sources I read, the association does not recommend a specific amount of carbohydrate, protein or fat. It lists low and very low carbohydrate patterns among eating patterns with research support and says the choice should be individualized.

Can low-carb eating reverse type 2 diabetes?

Some studies report remission, defined by lab thresholds, in a share of participants at 6 months, and one company-run study reported many participants reducing or stopping medicine over a year. Reviews say benefits shrink by 12 months and that long-term data are limited. Whether a person stays in remission is not guaranteed.

Do I have to stop my diabetes medicine on keto?

Do not stop or change any medicine on your own. Some medicines raise the risk of low blood sugar when you eat fewer carbohydrates, and doses are decided by your prescriber. Ask before you change your eating.

What is euglycemic ketoacidosis?

It is ketoacidosis in which blood glucose is normal or only mildly raised, which can make it harder to notice. A source citing the 2024 American Diabetes Association Standards of Care says very low-carbohydrate diets may increase this risk in people who take SGLT2 inhibitors. Ask your prescriber about it before cutting carbs.

Will keto raise my cholesterol?

On average, pooled trials found triglycerides lower, HDL higher and no significant change in LDL. Individual results vary, and one included trial reported a severe triglyceride rise in one participant. A cholesterol panel before and after a big change is reasonable to ask about.

Are net carbs a reliable way to count?

Net carbs, meaning total carbohydrate minus fiber and sometimes sugar alcohols, is a convention, not a lab-defined measurement. Rules for subtracting vary, so what a food does to your own glucose may differ. A meter or sensor is a better test.

Is low-carb better than the Mediterranean or plant-based diet?

The sources I read do not name a winner. The American Diabetes Association lists several patterns as options. See our diets-compared guide and the Mediterranean and plant-based guides for how they differ.

Who should avoid a very low-carb diet?

Diabetes Australia says low-carb eating is not recommended for children, or for pregnant or breastfeeding women, and advises against it for people with certain medical conditions or histories. Other groups, such as people with kidney disease or a history of an eating disorder, should ask a clinician. This is general information, not a rule for you.

Sources checked

• Frontiers in Nutrition, systematic review and meta-analysis of low-carbohydrate diets in type 2 diabetes with overweight or obesity (17 randomized trials), published January 2025
• Cucuzzella, Riley, Isaacs, Adapting Medication for Type 2 Diabetes to a Low Carbohydrate Diet, Frontiers in Nutrition, 2021
• AAFP, Patient-Oriented Evidence That Matters summary of a meta-analysis of low-carbohydrate diets
• Purdue University news release on the Virta Health one-year study (Diabetes Therapy, 2018)
• American Heart Association news, November 2022, on low-carb diet quality and diabetes risk (preliminary)
• Johns Hopkins Diabetes Information, summary of the ADA nutrition consensus report
• Glycemic Index Foundation, summary of the 2024 ADA nutrition recommendations
• DiabetesEd, low-carb diets and SGLT2 inhibitors (citing ADA Standards of Care 2024, section 5)
• Cleveland Clinic, Is the keto diet safe for people with diabetes?
• Diabetes Australia, low carbohydrate eating (March 2024)
• USDA FoodData Central values as displayed on getfoodfacts.com (broccoli, egg, almonds, avocado, spinach, cheddar, chicken, blueberries, rice)

About the author: Charles Kirkland is the founder of Metabolic Clarity Labs and author of The Journey Back. He is not a physician. This is general education, not a medically reviewed treatment plan.

Sources checked October 31, 2026. Discuss personal decisions with a qualified clinician. Do not start, stop, or change prescribed treatment based on this article.