Low‐carbohydrate diets in East Asians with type 2 diabetes: A systematic review and meta‐analysis of randomized controlled trials

ABSTRACT

Aims

Despite the reported success of low‐carbohydrate diets in improving glycemic control in the Western countries, no studies have investigated the effects of such diets in Asians. We aimed to conduct a systematic review and meta‐analysis of randomized controlled trials to examine the effects of low‐carbohydrate diets on glycemic control in East Asian adults.

Materials and Methods

We systematically searched the PubMed, Cochrane Library, and Embase databases from inception to June 28, 2023, to identify randomized controlled trials examining the efficacy of low‐carbohydrate diets in patients with type 2 diabetes (PROSPERO number CRD 42023453007). The primary outcome was the difference in glycated hemoglobin levels between the low‐carbohydrate diet and control groups. The secondary outcome was the difference in body mass index, fasting blood glucose level, blood pressure, and lipid profile.

Results

Six randomized controlled trials met the eligibility criteria. The study duration ranged from 3 to 18 months, with five studies conducted within 6 months. The results showed that low‐carbohydrate diets were more beneficial in lowering glycated hemoglobin levels and body mass index than control diets. The risk of bias for the six studies was minimal for two and moderate for four. The heterogeneity among the studies was low.

Conclusions

Low‐carbohydrate diets improved glycated hemoglobin levels and body mass index in East Asians compared with control diets. Therefore, carbohydrate restriction may be effective for glycemic management in East Asians with type 2 diabetes for at least 6 months.

Low‐carbohydrate diets improved glycated hemoglobin levels and body mass index in East Asians compared to control diets. Carbohydrate restriction may be effective for glycemic management in East Asians with type 2 diabetes, at least within 6 months.

INTRODUCTION

Dietary therapy is vital in treating diabetes; however, the most effective diet for improving blood glucose levels remains unclear. The importance of weight loss has been emphasized; however, the ideal nutrient ratio for such patients remains undefined in the 2023 Standards of Care in Diabetes guidelines of the American Diabetes Association ¹ .

Lipid restriction may be the most effective energy restriction strategy regarding energy contained per mass ² . However, the appropriate proportions of carbohydrate, protein, and fat calories for reducing glycated hemoglobin (HbA1c) levels and weight loss remain unclear; moreover, nutrient ratios should be determined according to an individual's age and medical condition.

Several systematic reviews and meta‐analyses have been conducted to investigate the effects of restricting carbohydrates in type 2 diabetes, especially in Europe and the United States ³ , ⁴ , ⁵ , ⁶ , ⁷ , ⁸ , ⁹ , ¹⁰ , ¹¹ , ¹² , ¹³ , ¹⁴ , ¹⁵ , ¹⁶ , ¹⁷ , ¹⁸ , ¹⁹ , ²⁰ , ²¹ . These studies have reported significant lowering of HbA1c levels in the carbohydrate‐restricted group than in the control group (regular, high‐carbohydrate, and low‐fat diets) within 6–12 months but not after 12–24 months ³ , ⁸ , ¹¹ , ¹³ , ¹⁸ . However, to our knowledge, no meta‐analyses have investigated the effects of low‐carbohydrate diets (LCDs) on glycemic control in Asians. Diets widely vary across regions and races, making it difficult to comprehensively assess the effects of LCDs on individuals from diverse regions and races. Moreover, Asians with type 2 diabetes have a lower average body mass index (BMI) than Europeans and Americans ²² , and their total dietary energy intake widely varies. Therefore, we aimed to conduct a systematic review and meta‐analysis of randomized controlled trials (RCTs) to examine the effects of LCDs on glycemic control in East Asians.

MATERIALS AND METHODS

Literature search and study selection

This study was conducted according to the Cochrane recommendations ²³ and is reported according to the Preferred Reporting Items for Systematic Reviews and Meta‐Analyses statement ²⁴ . The research outline was preregistered with the International Prospective Register of Systematic Reviews (CRD 42023453007) on August 26, 2023.

To identify appropriate articles, we searched the PubMed, Cochrane Library, and Embase databases from inception to June 28, 2023, for literature written in English. Eligible articles included reports of RCTs conducted in adult East Asian patients with type 2 diabetes to determine the safety or efficacy of LCDs, with HbA1c levels as the primary or secondary outcome. The exclusion criteria were studies that enrolled patients with type 1 diabetes or gestational diabetes and those that included patients who were not adults. The keywords “low‐carbohydrate,” “diabetes,” “nutrition,” “nutritional therapy,” “nutritional treatment,” “nutritional management,” “nutritional intervention,” “diet,” “dietary therapy,” “dietary treatment,” “dietary management,” and “dietary intervention” were used in the search. Keywords containing “gestational,” “adolescent,” “child,” and “children” were excluded. In the case of conference abstracts, a hand search was conducted to find if they had been published as articles.

The primary outcome was defined as a difference in HbA1c levels. The secondary outcome was defined as a difference in BMI, fasting blood glucose levels, systolic blood pressure, diastolic blood pressure, and triglyceride, high‐density lipoprotein cholesterol, and low‐density lipoprotein cholesterol levels. Furthermore, differences in HbA1c levels from studies limited to within 6 months were evaluated as a subgroup analysis.

Data extraction and risk‐of‐bias assessment

All retrieved studies were entered into Rayyan, a literature managing tool, and duplicates were removed. The titles and abstracts of included studies were screened. Reviews, meta‐analyses, comments, protocols, and conference abstracts unrelated to the topic were excluded. Subsequently, the studies that met the eligibility criteria were subjected to a full peer review. If only the abstract was listed in the database, the full text was obtained whenever possible, including studies in which all parts other than the abstract were written in languages other than English. We conducted full‐text screening and excluded studies that were not RCTs, did not have data on HbA1c levels, involved patients without diabetes, involved patients with type 1 diabetes, and that did not include East Asians.

Two independent researchers (J.H. and T.I.) selected the studies based on the inclusion criteria; if there were disagreements, a third reviewer (H.O.) was consulted before a final decision was made. These two independent researchers evaluated the extracted literature using version 2 of the Cochrane risk‐of‐bias tool for randomized trials, consulting the third reviewer as needed. The risk of bias was assessed across five domains (D): D1 specified the randomization process, D2 specified deviations from intended interventions, D3 specified missing outcome data, D4 specified measuring the outcome, and D5 specified the selection of the reported result.

Statistical analyses

SPSS statistical software version 29.0 (IBM Inc., Chicago, IL, USA) was used for all the analyses. The mean changes in HbA1c levels, BMI, fasting blood glucose levels, systolic blood pressure, diastolic blood pressure, and triglyceride, high‐density lipoprotein cholesterol, and low‐density lipoprotein cholesterol levels along with the standard deviations (SDs) in the intervention and control groups were extracted for analysis. SDs were calculated for studies that reported only mean differences and 95% confidence intervals (CIs). If the data required for the analysis were not adequately described, the author was contacted to request for the necessary data. The longer term results were adopted for studies that reported results at several time points. A meta‐analysis was performed using the collected data, and the standardized mean difference (SMD) calculated from each study was integrated with the effect size using a random‐effects model. Heterogeneity among the studies was quantitatively assessed using the I ² statistic, with I ² > 50% indicating heterogeneity. We also evaluated publication bias according to the recommendations for testing funnel plot asymmetry, although the small number of studies may have limited a comprehensive assessment.

RESULTS

Study selection

Details of the literature search are shown in Figure 1. We initially searched PubMed, Embase, and the Cochrane Library and obtained 1,484 studies. Subsequently, 404 duplicate studies were removed. We further screened the studies based on their titles and abstracts and excluded 1,044 studies. The full texts of 36 studies were reviewed, and those that were not RCTs, did not include HbA1c in the outcome, whose participants were patients with type 1 diabetes or without diabetes, and that did not include East Asians were excluded. Meta‐analyses were performed on the final six studies ²⁵ , ²⁶ , ²⁷ , ²⁸ , ²⁹ , ³⁰ . One study did not provide the mean and SD of HbA1c levels; therefore, we contacted the authors before including the study in our analysis ²⁵ .

Figure 1.

Preferred reporting items for systematic reviews and meta‐analyses flow diagram of the literature search process. RCT, randomized controlled trial.

Study characteristics

All six studies in the meta‐analysis were RCTs conducted in East Asian adults with type 2 diabetes published between 2014 and 2021. The characteristics of each study are briefly summarized in Table 1. The study sample sizes ranged from 24 to 134 patients, resulting in 400 patients across the six studies. The study durations ranged from 3 to 18 months, with five studies conducted within 6 months. The mean age ranged from 49.6 to 63.9 years, mean BMI ranged from 24.5 to 30.7 kg/m², and baseline HbA1c level ranged from 7.3% to 8.6% (56–70 mmol/mol) across the studies. The amount of carbohydrates in the intervention groups in each study ranged from <50 g up to 130 g per day.

Table 1.

Characteristics of included studies

Study	Study design and country	Participants	Duration	Intervention	Control	Primary outcome	Age (years)	BMI (kg/m2)	HbA1c (%)
Han, 2021	RCT, China	134 patients with type 2 diabetes	6 months	Carbohydrate <50 g/day	Low fat diet	Glycemic control	51.5	24.5	7.6
Chen, 2020	RCT, Taiwan	92 patients with type 2 diabetes	18 months	Carbohydrate <90 g/day	Traditional diabetic diet	Glycemic control status and change in medication effect scores	63.6	26.9	8.6
Wang, 2018	RCT, China	56 patients with type 2 diabetes	3 months	Carbohydrate <50 g/day	Low fat diet	HbA1c	63.9	24.5	7.9
Nishimori, 2018	RCT, Japan	28 patients with type 2 diabetes	3 months	Carbohydrate 70–130 g/day	IBW(kg) × 25 kcal/day	HbA1c (including degree of fatty liver)	49.6	30.7	7.3
Sato, 2017	RCT, Japan	66 patients with type 2 diabetes	6 months	Carbohydrate <130 g/day	IBW(kg) × 25 kcal/day	HbA1c	59.4	26.6	8.2
Yamada, 2014	RCT, Japan	24 patients with type 2 diabetes	6 months	Carbohydrate 70–130 g/day	IBW(kg) × 28 kcal/day	HbA1c	63.3	25.8	7.7

Age, BMI, and HbA1c levels are the characteristics at baseline, whereas total energy is the characteristic at the end of the study. For age, BMI, HbA1c levels, and total energy, the mean of both groups is shown. BMI, body mass index; HbA1c, glycated hemoglobin; IBW, ideal body weight; RCT, randomized controlled trial.

There were no changes in diabetes medication in two studies ²⁸ , ³⁰ , whereas it was changed in the remaining four. In one study, no significant differences were observed between the two groups: in the LCD group, medication use was reduced in two patients and increased in one. Concurrently, there were no changes in medication use in the control group ²⁷ . In another study, medication use was reduced in six patients in the LCD group and one in the control group ²⁹ . In the two remaining studies, diabetes medication reduction was scored as an endpoint, and a significant decrease in medication use was found in the LCD group ²⁵ , ²⁶ .

Exercise therapy was not mentioned in three studies ²⁸ , ²⁹ , ³⁰ . In the studies that mentioned exercise therapy, one reported no significant difference in the exercise frequency in either group ²⁷ , and two provided exercise advice to both groups ²⁵ , ²⁶ .

Dietary interventions

The carbohydrate amounts in the LCDs designed in the included studies were as follows: 70–130 g in two studies ²⁸ , ³⁰ , <50 g in two studies ²⁵ , ²⁷ , <90 g in one study ²⁶ , and <130 g in one study ²⁹ . Protein levels were set at 1.0–1.2 g/kg in four studies ²⁶ , ²⁸ , ²⁹ , ³⁰ , and the amount per body weight was not stated in two studies ²⁵ , ²⁷ .

When comparing the actual calorie intake of the two groups at the end of this study, the total calories in the LCD group were 1,430–1,800 kcal in four studies, which was not significantly different from those in the control group ²⁵ , ²⁶ , ²⁷ , ³⁰ . However, in two studies, the total calories were lower in the LCD group, with 1,371 kcal in the LCD group and 1,605 kcal in the control group in one study ²⁹ , and 1,389 kcal in the LCD group and 1,513 kcal in the control group in the other study ²⁸ . The control diets were as follows: calorie‐restricted diets in three studies ²⁸ , ²⁹ , ³⁰ , low‐lipid diets in two studies ²⁵ , ²⁷ , and a traditional diabetic diet in one study ²⁶ .

Risk of bias

The risk‐of‐bias assessment results are shown in Figure S1. The risk of bias for the six studies was rated as follows: low risk (n = 2) and some risk concerns (n = 4). The judgment in each of the five domains of risk of bias (D1–D5) was as follows:

In two studies, block randomization was used for random sequence creation; therefore, these studies were judged to have some risk concerns in D1. One study did not indicate whether there were dropouts, and we determined that there were some risk concerns in D2. In all studies, neither the study participants nor researchers were blinded to the dietary intervention; however, we determined that the dropouts were not due to the specific intervention and we considered that the other four studies had low risk in D2. All studies were judged to be at low risk for the items related to D3 (missing outcomes) and D4 (outcome measures). Additionally, for three studies, it was not possible to assess the protocols on the web to determine whether the primary outcome matched the predetermined outcome, and the risk in D5 was judged as of “some concern.”

Glycemic control

The changes in HbA1c levels are shown in Figure 2. In the six studies included in this meta‐analysis, LCDs were associated with a higher significant overall reduction in HbA1c levels than control diets (SMD: −0.55%, 95% CI: −0.77 to −0.33). The baseline HbA1c levels in the studies ranged from 7.3% to 8.6%. Three of the six RCTs reported significant reductions in HbA1c levels in the LCD group than in the control diet group ²⁵ , ²⁶ , ²⁹ . whereas the other three RCTs reported no significant differences ²⁷ , ²⁸ , ³⁰ . Heterogeneity between studies was low, with I ² at 0%.

Figure 2.

Forest plot of the change in HbA1c levels of a low‐carbohydrate diet compared with a control diet in patients with type 2 diabetes. The results are obtained from a random‐effects meta‐analysis. CI, confidence interval (expressed as %); HbA1c, glycated hemoglobin; SD, standard deviation; SMD, standardized mean difference.

When limited to within 6 months, five studies were included in this meta‐analysis. Similar to the overall results, LCDs were associated with a more significant reduction in HbA1c levels than control diets (SMD: −0.56%, 95% CI: −0.80 to −0.32; Figure S2).

Secondary outcomes

The changes in BMI are shown in Figure 3. Four studies reported mean differences and SDs in the change in BMI between the LCD and control groups ²⁵ , ²⁶ , ²⁷ , ²⁹ , whereas two did not ²⁶ , ³⁰ . In a meta‐analysis of the results of four studies, LCDs were associated with a significant decrease in BMI compared with control diets (SMD: −0.41 kg/m², 95% CI: −0.65 to −0.17). The heterogeneity in BMI between the studies was also low, with I ² at 0%.

Figure 3.

Forest plot of the change in BMI of a low‐carbohydrate diet compared with a control diet in patients with type 2 diabetes. The results are obtained from a random‐effects meta‐analysis. BMI, body mass index; CI, confidence interval (expressed as kg/m²); SD, standard deviation; SMD, standardized mean difference.

The changes in fasting blood glucose levels are shown in Figure S3. Three studies reported mean differences and SDs in the change in fasting blood glucose levels between the LCD and control groups ²⁵ , ²⁶ , ²⁸ , whereas three did not ²⁷ , ²⁹ , ³⁰ . In a meta‐analysis of the results of three studies, LCDs were associated with a significant decrease in fasting blood glucose levels compared with control diets (SMD: −0.40 mg/dL, 95% CI: −0.65 to −0.15).

The changes in systolic blood pressure, diastolic blood pressure, and triglyceride, high‐density lipoprotein cholesterol, and low‐density lipoprotein cholesterol levels are shown in Figures [Link], [Link]. Two studies reported mean differences and SDs in the change in these items between the LCD and control groups ²⁶ , ²⁸ . LCDs were not associated with a significant decrease in systolic blood pressure, diastolic blood pressure, and triglyceride, high‐density lipoprotein cholesterol, and low‐density lipoprotein cholesterol levels compared with control diets.

Publication bias

Publication bias regarding the effects of LCDs on HbA1c levels compared with those of control diets was evaluated using funnel plots. The results were symmetrical, with no evident publication bias (Figure S9).

DISCUSSION

This study integrated the results of six RCTs involving East Asian patients with type 2 diabetes and found that LCDs improved HbA1c levels compared with control diets in these patients. Additionally, the LCD group demonstrated a more significant reduction in BMI and fasting glucose levels than the control groups as a secondary outcome. However, the LCD group had no improvement in blood pressure and lipid profile compared with the control group. The mean baseline BMI, age, and HbA1c levels among the studies in this meta‐analysis ranged from 24.5 to 30.7 kg/m², 49.6–63.9 years, and 7.3–8.6%, respectively. The mean baseline BMI was relatively high for Asians. The baseline BMI, age, and HbA1c levels varied among the studies; however, heterogeneity was low. Dropout rates were similar in the LCD and control groups in the included studies. This could be attributed to the fact that the duration of most of the studies was short.

In this meta‐analysis, LCDs significantly reduced HbA1c levels compared with control diets in three studies with as large a number as 66–134 patients ²⁵ , ²⁶ , ²⁹ . The other three RCTs had small sample sizes of <60 patients, which may have resulted in less significant differences ²⁷ , ²⁸ , ³⁰ .

The pathophysiology of type 2 diabetes differs between Western and Japanese populations. Asians, even those with milder obesity, are more prone to diabetes than Westerners ²² . After adjusting for age and sex, the rate of type 2 diabetes in Caucasians with a BMI of 30 kg/m² was the same as in Chinese individuals with a BMI of 26.9 kg/m² and South Asians with a BMI of 23.9 kg/m^{2 31} . In a comprehensive analysis of 74 cohort studies measuring insulin sensitivity and acute insulin response using an intravenous glucose tolerance test, East Asians had higher insulin sensitivity and lower insulin secretory capacity than Africans and Caucasians ³² .

A genome‐wide analysis of the genetic information of approximately 400,000 East Asians identified 61 genetic regions that increased the risk of developing type 2 diabetes, including genes involved in muscle and fat differentiation and microRNA ³³ . Genetic factors may underlie the predisposition to diabetes among Asians without severe obesity. Moreover, not only the pathophysiology of type 2 diabetes but also food habits and culture, including carbohydrate intake, widely vary according to race. Therefore, it is essential to examine the effects of carbohydrate restriction in East Asian populations.

This meta‐analysis involving East Asian patients found an improvement in HbA1c levels with carbohydrate restriction, similar to previous studies involving Western populations. However, the degree of improvement in HbA1c levels differed between Westerners and East Asian patients. Although this improvement is not directly comparable, the SMDs were −0.34 to −0.05% ³ , ⁴ , ⁹ , ¹⁰ , ¹¹ , ¹⁴ , ¹⁹ , ²¹ in Western patients and −0.55% (95% CI: −0.77 to −0.33) in the present study. This suggests that carbohydrate restriction may be more effective in improving HbA1c levels in East Asians.

In the analysis of BMI reduction, the LCD group exhibited a more significant reduction than the control group. Previous studies have confirmed the effect of LCDs on weight loss in patients with obesity ⁸ , ¹⁸ . Therefore, the effects of LCDs on HbA1c reduction contributed to weight loss. However, despite a more significant decrease in HbA1c levels in East Asian patients than in Western patients, BMI reduction owing to carbohydrate restriction tended to be lower in East Asians than in Westerners. The SMDs of BMI reduced by −1.79 to −1.35 kg/m² at 3–6 months in Western patients and by −0.41 kg/m² (95% CI: −0.65 to −0.17) in East Asian patients ⁸ , ¹⁸ . In general, carbohydrate restriction leads to weight loss, which improves insulin resistance and lowers blood glucose levels. The more significant improvement in HbA1c levels in East Asians, despite a smaller decrease in BMI, may be due to racial differences in the pathophysiology of diabetes. Western individuals have a higher BMI; hence, their blood glucose levels may improve because of weight loss owing to carbohydrate restriction. However, Asians have a low insulin secretory capacity and low BMI; therefore, the weight loss effect of carbohydrate restriction is small. In Asians, when carbohydrate intake is high, insulin secretion is insufficient to compensate, leading to elevated blood glucose levels. Consequently, carbohydrate restriction may be more effective in East Asians than in Western patients.

In a meta‐analysis evaluating the long‐term effects of a low‐carbohydrate diet on HbA1c over 12 months in individuals with type 2 diabetes, regardless of race, three of the six studies reported an average carbohydrate intake of <90 g at the end of the study, one study reported <100 g, and another reported <140 g ³⁴ . Most studies in the LCD group did not achieve their target carbohydrate levels by the end of the study, but they were able to maintain a low‐carbohydrate diet for a long period ³⁴ . Notably, only the Asian study demonstrated a significant reduction in HbA1c compared to controls in trials over 12 months. Carbohydrate intake was reported to be higher in Asians than in Westerners ³⁵ . Differences in dietary habits and the pathophysiology of diabetes, particularly in insulin secretory capacity, between Asians and Westerners may have contributed to the observed reduction in HbA1c only in the Asian with a long‐term low‐carbohydrate diet. However, only one RCT has examined the effects of a long‐term LCD in East Asians; hence, further studies are required.

In a meta‐analysis assessing the reduction of HbA1c in type 2 diabetes on low‐carbohydrate diets, not limited to Asian populations, a significant reduction in HbA1c was observed with diets containing <130 g of carbohydrates, but not in studies with <50 g ¹⁰ . In our meta‐analysis, two studies focused on strict carbohydrate‐restricted diets containing <50 g of carbohydrates. One study demonstrated a significant reduction in HbA1c compared to the control group, with the LCD group averaging 61 g of carbohydrate intake by the end of the study ²⁵ . In contrast, another study did not show a significant reduction in HbA1c, as it did not adhere to strict carbohydrate restriction, with an average carbohydrate intake of 173 g at the end of study ²⁷ . The lack of a stronger HbA1c reduction with very low carbohydrate‐restricted diets (<50 g) compared to other studies may be attributed to low compliance rates.

This study has some limitations. First, the duration of the studies in our meta‐analysis differed, and there was only one long‐term study. Second, the effect of LCDs in lean East Asian patients with type 2 diabetes was unclear because the baseline BMI in this study ranged from 24.5 to 30.7 kg/m², which is relatively high for Asians. Increased mortality has been reported among elderly Japanese individuals with low BMI ³⁶ , and safety considerations of carbohydrate‐restricted diets that induce weight loss are needed, especially for lean elderly populations. Third, the studies showed differences in low‐carbohydrate intake, control diets, and total energy intake. Moreover, the amount of fiber in the carbohydrates was not examined.

As the optimal intake of carbohydrates in diabetes depends on the amount of physical activity and degree of insulin action, it is difficult to establish a standardized target quantity. Our results suggest that carbohydrate restriction can treat type 2 diabetes in East Asians. However, notably, LCDs are associated with a risk of diabetic ketoacidosis in patients with type 1 diabetes and those using sodium‐glucose cotransporter 2 inhibitors, and the safety of LCDs is unclear in this patient group ³⁷ .

Carbohydrate restriction may be effective for glycemic management in East Asians with type 2 diabetes, for at least 6 months. However, the long‐term effects of LCDs on lean East Asian patients remain unclear. Further studies are necessary to investigate the prolonged effectiveness of and the extended adherence to LCDs in Asian populations with type 2 diabetes.

FUNDING

This work was supported by Ministry of Health, Labour and Welfare Comprehensive Research Project for Measures against Cardiovascular Diseases, Diabetes and Other Lifestyle Related Diseases Program Grant Number JPMH 24FA1008.

DISCLOSURE

Junya Hironaka reports a relationship with Novo Nordisk Pharma Ltd. and Kowa Pharmaceutical Co., Ltd. that includes speaking and lecture fees. Masahide Hamaguchi reports a relationship with AstraZeneca K.K., Ono Pharma Co., Ltd., and Kowa Pharma Co., Ltd. that includes funding grants. Masahide Hamaguchi reports a relationship with AstraZeneca K.K., Ono Pharma Co., Ltd., Eli Lilly, Japan, Sumitomo Dainippon Pharma Co., Ltd., Daiichi Sankyo Co., Ltd., Mitsubishi Tanabe Pharma Corp., Sanofi K.K., K.K., and Kowa Pharma Co., Ltd. that includes speaking and lecture fees. Takahiro Ichikawa reports a relationship with Abbott Japan Co Ltd that includes funding grants. Takahiro Ichikawa reports a relationship with Kowa Pharmaceutical Co Ltd that includes speaking and lecture fees. Hanako Nakajima reports a relationship with Kowa Pharmaceutical Co., Ltd., Kyowa Hakko Kirin Co., Ltd., and Nippon Boehringer Ingelheim Co., Ltd. that includes speaking and lecture fees. Takafumi Senmaru reports a relationship with Eli Lilly Japan K.K., Mitsubishi Tanabe Pharma Co, Daiichi Sankyo Co., Ltd., Kowa Pharma Co., Ltd., Astellas Pharma Inc., Takeda Pharma Co., Ltd., Sanofi K.K., Taisho Toyama Pharma Co., Ltd., Kyowa Kirin Co., Ltd., Kissei Pharma Co., Ltd., MSD K.K., Novo Nordisk Pharma Ltd., Ono Pharma Co., Ltd., AstraZeneca K.K., Mochida Pharma Co., Ltd., and TERUMO CORPORATION, Abbott Japan Co., Ltd. that includes speaking and lecture fees. Hiroshi Okada reports a relationship with Mochida Pharma Co., Ltd., Teijin Pharma Ltd., MSD K.K., Mitsubishi Tanabe Pharma Corporation, AstraZeneca K.K., Sumitomo Dainippon Pharma Co., Ltd., Novo Nordisk Pharma Ltd., Daiichi Sankyo Co., Ltd, Eli Lilly Japan K.K, Kyowa Hakko Kirin Company Ltd, Kissei Pharmaceutical Co., Ltd, Takeda Pharmaceutical Co., Ltd, Kowa Pharmaceutical Co., Ltd, Ono Pharmaceutical Co., Ltd., and Sanofi K.K. that includes speaking and lecture fees. Emi Ushigome reports a relationship with The Japanese Study Group for Physiology and Management of Blood Pressure and Astellas Foundation for Research on Metabolic Disorders Mishima Kaiun Memorial Foundation that includes funding grants. Emi Ushigome reports a relationship with Nippon Boehringer Ingelheim Co., Ltd., Mitsubishi Tanabe Pharma Corporation, Daiichi Sankyo Co., Ltd, MSD K.K., Kyowa Hakko Kirin Co., Ltd., Sumitomo Dainippon Pharma Co., Ltd., Kowa Pharmaceutical Co., Ltd., Novo Nordisk Pharma Ltd., Ono Pharmaceutical Co., Ltd., Taisho Pharmaceutical Co., Ltd., and Sanofi K.K. outside the submitted work. Donated Fund Laboratory of Diabetes therapeutics is an endowment department, supported with an unrestricted grant from Ono Pharmaceutical Co., Ltd., Taiyo Kagaku Co., Ltd., and Taisho Pharmaceutical Co., Ltd. that includes speaking and lecture fees. Naoko Nakanishi reports a relationship with Kowa Pharmaceutical Co., Ltd., Novo Nordisk Pharma Ltd., Nippon Boehringer Ingelheim Co., Ltd., and TERUMO CORPORATION that includes speaking and lecture fees. Michiaki Fukui reports a relationship with Ono Pharma Co., Ltd., Oishi Kenko Inc., Yamada Bee Farm, Nippon Boehringer Ingelheim Co., Ltd., Kissei Pharma Co., Ltd., Mitsubishi Tanabe Pharma Corp., Daiichi Sankyo Co., Ltd., Sanofi K.K., Takeda Pharma Co., Ltd., Astellas Pharma Inc., MSD K.K., Kyowa Kirin Co., Ltd., Sumitomo Dainippon Pharma Co., Ltd., Kowa Pharma Co., Ltd., Novo Nordisk Pharma Ltd., Sanwa Kagagu Kenkyusho Co., Ltd., Eli Lilly, Japan, K.K., Taisho Pharma Co., Ltd., Terumo Corp., Teijin Pharma Ltd., Nippon Chemiphar Co., Ltd., Abbott Japan Co., Ltd., and Johnson & Johnson K.K. Medical Co., TERUMO CORPORATION that includes funding grants. Michiaki Fukui reports a relationship with Nippon Boehringer Ingelheim Co., Ltd., Kissei Pharma Co., Ltd., Mitsubishi Tanabe Pharma Corp., Daiichi Sankyo Co., Ltd., Sanofi K.K., Takeda Pharma Co., Ltd., Astellas Pharma Inc., MSD K.K., Kyowa Kirin Co., Ltd., Sumitomo Dainippon Pharma Co., Ltd., Kowa Pharma Co., Ltd., Novo Nordisk Pharma Ltd., Ono Pharma Co., Ltd., Sanwa Kagaku Kenkyusho Co., Ltd., Eli Lilly Japan K.K., Taisho Pharma Co., Ltd., Bayer Yakuhin, Ltd., AstraZeneca K.K., Mochida Pharma Co., Ltd., Abbott Japan Co., Ltd., Teijin Pharma Ltd., Arkray Inc., Medtronic Japan Co., Ltd., and Nipro Corp., TERUMO CORPORATION that includes speaking and lecture fees. Other authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Approval of the research protocol: N/A.

Informed consent: As patient data were extracted as anonymized data, informed consent was not required.

Registry and the registration no. of the study/trial: Approval date of Registry is August 26, 2023. Registration No. of this study is CRD42023453007.

Animal studies: N/A.

Supporting information

Figure S1. Risk of bias assessment.

Figure S2. Forest plot of the change in HbA1c level of a low carbohydrate diet compared with a control diet in patients with type 2 diabetes within 6 months. The results are from a random‐effects meta‐analysis. CI, confidence interval (expressed as %); SD, standard deviation; SMD, standardized mean difference; HbA1c, glycated hemoglobin.

Figure S3. Forest plot of the change in fasting blood glucose levels of a low carbohydrate diet compared with a control diet in patients with type 2 diabetes. The results are from a random‐effects meta‐analysis. CI, confidence interval (expressed as %); SD, standard deviation; SMD, standardized mean difference.

Figure S4. Forest plot of the change in systolic blood pressure of a low carbohydrate diet compared with a control diet in patients with type 2 diabetes. The results are from a random‐effects meta‐analysis. CI, confidence interval (expressed as %); SD, standard deviation; SMD, standardized mean difference.

Figure S5. Forest plot of the change in diastolic blood pressure of a low carbohydrate diet compared with a control diet in patients with type 2 diabetes. The results are from a random‐effects meta‐analysis. CI, confidence interval (expressed as %); SD, standard deviation; SMD, standardized mean difference.

Figure S6. Forest plot of the change in the triglyceride levels of a low carbohydrate diet compared with a control diet in patients with type 2 diabetes. The results are from a random‐effects meta‐analysis. CI, confidence interval (expressed as %); SD, standard deviation; SMD, standardized mean difference.

Figure S7. Forest plot of the change in high‐density lipoprotein cholesterol levels of a low carbohydrate diet compared with a control diet in patients with type 2 diabetes. The results are from a random‐effects meta‐analysis. CI, confidence interval (expressed as %); SD, standard deviation; SMD, standardized mean difference.

Figure S8. Forest plot of the change in low‐density lipoprotein cholesterol levels of a low carbohydrate diet compared with a control diet in patients with type 2 diabetes. The results are from a random‐effects meta‐analysis. CI, confidence interval (expressed as %); SD, standard deviation; SMD, standardized mean difference.

Figure S9. Funnel plot of publication bias assessment.