Dietary carbohydrate and the risk of type 2 diabetes: an updated systematic review and dose–response meta-analysis of prospective cohort studies

We did this study to clarify the association between carbohydrate intake and the risk of type 2 diabetes (T2D) and potential effect modification by geographical location. PubMed, Scopus and Web of Science were searched to find prospective cohort studies of dietary carbohydrate intake and T2D risk. A random-effects dose–response meta-analysis was performed to calculate the summary hazard ratios (HRs) and 95%CIs. The quality of cohort studies and the certainty of evidence was rated using the Newcastle–Ottawa Scale and GRADE tool, respectively. Eighteen prospective cohort studies with 29,229 cases among 607,882 participants were included. Thirteen studies were rated to have high quality, and five as moderate quality. The HR for the highest compared with the lowest category of carbohydrate intake was 1.02 (95%CI: 0.91, 1.15; I2 = 67%, GRADE = low certainty). The HRs were 0.93 (95%CI: 0.82, 1.05; I2 = 58%, n = 7) and 1.26 (95%CI: 1.11, 1.44; I2 = 6%, n = 6) in Western and Asian countries, respectively. Dose–response analysis indicated a J shaped association, with the lowest risk at 50% carbohydrate intake (HR50%: 0.95, 95%CI: 0.90, 0.99) and with risk increasing significantly at 70% carbohydrate intake (HR70%: 1.18, 95%CI: 1.03, 1.35). There was no association between low carbohydrate diet score and the risk of T2D (HR: 1.14, 95%CI: 0.89, 1.47; I2 = 90%, n = 5). Carbohydrate intake within the recommended 45–65% of calorie intake was not associated with an increased risk of T2D. Carbohydrate intake more than 70% calorie intake might be associated with a higher risk.

www.nature.com/scientificreports/ glycemic properties of the diet including glycemic index and load might be associated with the risk of developing T2D and other chronic diseases 11 . With regard to T2D, three meta-analyses of cohort studies have been undertaken of the association between dietary carbohydrate and the risk of T2D, but the results have been inconsistent [11][12][13] . However, almost all studies included in the published meta-analyses have been conducted in Western countries, where the intake of carbohydrates was lower than that of Asian countries 14,15 . A recent publication from the PURE study in 21 countries across the world indicated that higher rice consumption was associated with a greater risk of developing T2D, with the strongest association in South Asia and a modest, nonsignificant association in other regions 16 .
A number of population-based prospective cohort studies in Asian countries have recently been published that reported significant positive associations between dietary carbohydrates and the risk of T2D [17][18][19] . In the current study, we therefore aimed to update the evidence from prospective cohort studies of the association between dietary carbohydrate intake and low-carbohydrate diet score (LCDS) with the risk of T2D in the general population. Our secondary outcome was to assess this association separately in Asian and Western countries.

Search strategy.
We performed a comprehensive systematic search on all literature issued earlier than April 2021 in online databases including PubMed/Medline, Scopus and ISI Web of Science. We did not exert any limitation in term of language or time of publication. We used search terms relevant to type 2 diabetes, carbohydrate, and study design to find potential eligible cohort studies (Supplementary Table 1). Reference lists of retrieved articles and relevant reviews were also manually searched. Unpublished data was not included.

Inclusion and exclusion criteria.
Relevant articles with all of the following inclusion criteria were included: (1) published prospective cohort studies conducted in the general population; (2) reported carbohydrate consumption (as either g/d or percentage energy) and LCDS as exposure; (3) considered T2D incidence as the outcomes of interest; (4) provided estimates of the effect size in the form of relative risk, hazard ratio (HR) or rate ratio with corresponding 95% confidence intervals (CIs) for ≥ 2 quantitative categories of carbohydrate consumption or LCDS; and (5) provided the numbers of cases and non-cases or person-years in each category of dietary carbohydrate or LCDS. Studies that reported continuous estimation from the associations were also included. For duplicate publications form the same cohort, the one with the greater number of cases was included in our meta-analysis. We excluded letters, comments, reviews and meta-analyses, and ecologic studies. We also did not include studies that were performed on children or adolescences or those that were conducted among patients with type one diabetes. All outcomes were classified based on the World Health Organization's international classification of disease criteria. Data extraction. Data extraction process was executed by two reviewers in duplicate (FH and AJ), and any divergences were resolved by consultation the principal investigator (SS-B). We extracted the following information from the publications identified: name of the first author, publication year, country, age, sex, study participants, number of cases, duration of follow-up, method of assessment of carbohydrate consumption and LCDS, the fully-adjusted estimates and their 95%CI and list of potential confounders entered into the multivariable statistical model. Gender-specific estimates were combined a by fixed-effects model to include each cohort once in the main analysis. We used web plot digitizer (http:// plotd igiti zer. sourc eforge. net/) to extract numerical estimates from graphs.
Data synthesis and analysis. We considered the HR and its 95%CI as the effect size for the present study.
Relative risks were considered equal to HR 21 . We first performed a pairwise meta-analysis by combining the reported effect sizes for the highest compared with the lowest category of dietary carbohydrate or LCDS in each study. Study-specific results were combined with a random-effects model 22 . The Cochran Q 23 and I 2 statistic 24 were used to test for presence of heterogeneity.
Subgroup analyses of dietary carbohydrates were performed based on sex, geographic location, number of cases, duration of follow-up and adjustments for main confounders including body mass index (BMI), smoking status, alcohol drinking, and energy and fiber intakes. P value for subgroup difference was generated using meta-regression analysis. Subgroup analyses of LCDS were performed based on sex, study location, and duration of follow-up. Publication bias was assessed by visual inspection of funnel plot 23 and Egger's 25 and Begg's 26 tests, when at least 10 studies were available. To determine whether the pooled effect size was influenced heavily by a single cohort, sensitivity analysis was done by step-by-step omission of each cohort at a time.
We used the method introduced by Greenland 27 and Orsini 28 for dose-response meta-analysis. We calculated the HRs for a 10% increment in carbohydrate intake or a 10-point increment in LCDS in each study. Studyspecific HRs were combined by a random-effects model. For this purpose, each cohort study must report the number of cases and person-years and median or range of dietary carbohydrate or LCDS across categories of exposures. For studies that reported dietary carbohydrate as g/d, we converted g/d to percentage calorie from carbohydrate by using the average daily energy intake of the study participants. For studies that reported the results per unit increment in dietary carbohydrate (i.e., per 200 g/d increment), we first converted g/d to percentage energy from carbohydrate and then translated it to a 10% increment in energy intake from carbohydrate. For studies that used different units (for example, 5% increase in carbohydrate intake), we calculated the logarithm www.nature.com/scientificreports/ of the HR and its 95%CI, multiplied by the corresponding unit, and then exponentiated the results. For studies that reported carbohydrate intake as a range in each category, we used the midpoint of lower bounds as a proxy of the median. The widths of the open-ended categories were considered equal to the closest categories. Finally, we performed a one-stage weighted mixed-effects meta-analysis to model dose-response associations 29 . This method estimates the study-specific slope lines and combines them to obtain an overall average slope in a single stage. We included all studies in the main analysis. However, due to substantial difference in carbohydrate consumption in Asian and Western countries, we performed separate nonlinear dose-response analyses in Asian and Western countries. Statistical analyses were conducted using STATA software, version 15.0. P < 0.05 was considered statistically significant.
Quality assessments and grading the evidence. The quality of the original studies included in the present meta-analysis was evaluated using a 9-point Newcastle-Ottawa Scale by two independent investigators (FH and AJ) 30 . Accordingly, studies with 1-3, 4-6, and 7-9 points were rated as poor, fair, and high quality, respectively. The certainty in the estimates was rated by the Grading of Recommendations Assessment, Development, and Evaluation (GRADE) approach. GRADE tool is a metric to assess the certainty of the evidence 31 . This tool grades observational studies as low with downgrades for study limitations, inconsistency, indirectness, imprecision, and publication bias, and upgrades for large effect size, dose-response gradient, and attenuation by plausible confounding.

Characteristics of included studies. Characteristics of the included studies are provided in
The association did not reach statistical significance by the stepwise exclusion of each primary study at a time (HR range: 0.99 to 1.05). In the subgroup analyses, there was no significant association across subgroups except for studies conducted in Asia (HR: 1.26, 95%CI: 1.11, 1.44; I 2 = 6%, n = 6; Table 1). Geographical location, number of cases, and adjustment for dietary fiber intake were potential sources of heterogeneity. There was no evidence of small-study effect such as publication bias with Egger's test (P = 0.99) and Begg's test (P = 0.95) ( Supplementary Fig. 2).
Grading the evidence. The certainty in the estimates was rated by the GRADE approach. The certainty of the evidence was rated low for dietary carbohydrate, with a downgrade for imprecision and an upgrade for dose-response gradient (Supplementary Table 7). The certainty in the estimates was rated very low for LCDS, with downgrades for imprecision and inconsistency.

Discussion
This is the most recent and up-to-date meta-analysis of prospective cohort studies that examined the association between carbohydrate intake from diet and risk of T2D. Since the release of the three earlier meta-analyses [11][12][13] , some prospective cohort studies, especially those conducted in Asian countries, have been published that highlighted a need to present updated evidence for this association. We found evidence of a J-shaped relationship between carbohydrate intake and T2D in the non-linear dose-response, with the lowest risk at carbohydrate intake of 50% total calorie and with risk increasing significantly at 70% of total calorie. There appeared to be a marked difference in the association between carbohydrate intake and T2D between Asian and Western countries. Low carbohydrate diet score was not associated with the risk of T2D. In line with ours, a previous meta-analysis on eight prospective studies in 2013 revealed that total carbohydrate intake was not associated with the risk of T2D in the linear dose-response analysis 12 . In addition, some earlier studies, mostly conducted in Western countries, did not find an association between carbohydrate intake from diet and the risk of T2D 32,42,44,45 .
Another recent meta-analysis of cohort studies showed a non-significant association between carbohydrate intake and the risk of T2D in Western countries and in contrast, found a significant positive association in one Asian study 11 . We updated the evidence and included additional recent studies conducted in Asian countries which showed that carbohydrate intake, within the recommended daily intake of 45-65% of total calorie, as reported in Western countries, was not associated with an increased risk of type 2 diabetes, and even was associated with a modest lower risk at 50% carbohydrate intake. However, the nonlinear dose-response meta-analysis www.nature.com/scientificreports/ www.nature.com/scientificreports/ of five Asian studies suggested that carbohydrate intake higher than 70% of total calorie was strongly associated with a higher risk of T2D. A recent meta-analysis of prospective cohort studies found a similar U-shaped association between carbohydrate intake and total mortality, with the lowest risk being found at 50-55% of carbohydrate intake, and an increased risk at an intake of more than 70% carbohydrate intake 54 . Evidence from earlier prospective cohort studies evaluating the association between the quality and quantity of dietary carbohydrates, reflected by dietary glycemic index and load, suggests that both quality and quantity of dietary carbohydrates are associated with the risk of T2D 18,43,55,56 . In addition, there was also evidence of a causal association between dietary glycemic index and load and the risk of T2D 56,57 .
Studies have suggested some mechanisms relating dietary carbohydrates to the risk of T2D. The long-term exposure to dietary carbohydrates may provide a continuous signal to the pancreatic β-cell to secret insulin to reduce blood glucose levels. Consequently, β-cell exhaustion can result in glucose intolerance 58 . Furthermore, excessive carbohydrates intake produces a large amount of acetyl CoA in the metabolic pathways, thus releasing lots of free radical and thereby exacerbating insulin resistance 58,59 .
There are also several explanations for the observed geographical difference found in the present study. First and most importantly, carbohydrate intake is substantially higher in Asian countries (generally > 60%) than in Western countries (generally < 50%) 54 . We found a relatively J-shaped association, wherein the US and European countries mainly represented the left side of the curve and in contrast, Asian countries represented the right side of the curve 54 . Higher carbohydrate intake increases demand for insulin secretion, leading to β-cell exhaustion. Second, Asian populations have a lower capacity of insulin secretion than that of their Western counterparts [60][61][62] . In addition, type of carbohydrate consumed, especially proportion of whole and refined grains, may be different across the globe and this may create a difference in the association between dietary carbohydrates with the risk of T2D. The main source of carbohydrates in most Asian countries is refined carbohydrates such as white rice and bread, reflecting low diet quality 63,64 . White rice, a high glycemic index food, was associated with an increased risk of T2D, especially in Asian societies 65,66 . More recently, a prospective cohort study conducted in 21 countries www.nature.com/scientificreports/ across the globe indicated that higher rice consumption was associated with an increased risk of type 2 diabetes in South Asian countries, and a modest non-significant association in other regions 16 .

Strength and limitations.
We updated previous meta-analyses and included the most recent studies, especially those conducted in Asia. We included new Asian articles and looked at them separately by subgrouping them because of the difference in their diet. Here we showed that higher carbohydrate intake more than the recommended daily intake of 45-65% was strongly associated with the risk of T2D. We applied a newlydeveloped one-stage linear mixed effects meta-analysis that creates more efficient and flexible plots than the conventional two-stage model. Some limitations should be noted in the context of our findings. Due to the observational nature of the studies included, our resulting associations cannot establish causality. According to the GRADE, the certainty of the evidence was rated low for dietary carbohydrate and very low for LCDS. In addition, we had insufficient data for the analysis of LCDS. We used total carbohydrate intake as exposure which represents a large diverse group of foods such as whole and refined grains. The potential difference in foods constituting total carbohydrate intake in Asian and Western countries might confound the association between total carbohydrate intake and T2D.

Conclusion
The results of this updated meta-analysis of 18 cohort studies (607,882 participants with 29,228 cases) showed that carbohydrate intake within the recommended dietary intake of 45% to 65% of total calorie was not associated with a higher risk of T2D and even was associated with a modest lower risk at 50% carbohydrate intake. Carbohydrate intake more than 70% of total calorie, as found in Asian countries, was associated with substantial higher risk of T2D. However, these findings were obtained from observational studies and thus, could not prove causality. More research, especially in Asian countries, is needed to investigate the association between carbohydrate intakes higher than recommended dietary intake with the risk of T2D.

Data availability
The data, codes, analytical syntax, and other additional data used for the present meta-analysis are available from the corresponding author on reasonable request.