Association of carbohydrate intake from different sources with all-cause and cardiovascular mortality among chronic kidney disease populations: assessment of 1999–2018 National Health and Nutrition Examination Survey participation

Abstract This study analysed the data from the NHANES (1999–2018) to examine how different sources of carbohydrate intake affected the all-cause and cardiovascular mortality of 11,302 chronic kidney disease (CKD) patients. The data were adjusted for other factors using various methods. The results showed that CKD patients (stages 1-2 and 3-5) who consumed more carbohydrates from whole grains, fruits, vegetables and less carbohydrates from fruit juice or sauces had lower mortality rates. Replacing fat intake with carbohydrates from whole grains (HR = 0.86[0.78–0.95]), fruits (raw) (HR = 0.79[0.70–0.88]) and non-starchy vegetables (HR = 0.82[0.70–0.96]), but not protein intake, was linked to lower all-cause mortality. The fibre content in carbohydrates might partly account for the benefits of selected carbohydrate intake. This study provided practical recommendations for optimising the carbohydrate sources in CKD patients.


Introduction
Chronic kidney disease (CKD) is a complicated syndrome accompanied by chronic progressive destruction of renal function caused by various factors and is linked to an increased risk of cardiovascular disease (CVD), morbidity and mortality (Go et al. 2004;Romagnani et al. 2017).Globally, CKD directly brought about an estimated 1.23 million deaths in 2017, with an additional 1.36 million deaths ascribe to CVD caused by impaired kidney function (Collaboration 2020).Patients with CKD should follow a particular diet to reduce the accumulation of metabolic products and thus the onset and severity of uraemic symptoms and to prevent disorders of calcium, phosphorus, and potassium metabolism or protein energy wasting (Pereira et al. 2020).Current guidelines for CKD management recommend dietary modifications, such as limiting protein, phosphorus and salt intake (Ikizler et al. 2020;Mottl et al. 2022).Carbohydrates, one of the three dietary macromolecules, merit further investigation.Previous studies indicated that high carbohydrate intake was associated with a higher risk of total mortality and CVD (Dehghan et al. 2017;Kwon et al. 2020;Lee et al. 2021;Mohammadifard et al. 2022;Jo and Park 2023;Qin et al. 2023).Additionally, several studies have documented the association of carbohydrate quality index and protein-to-carbohydrate ratio with all-cause mortality in adult populations (Fernandez-Lazaro et al. 2021a;Wabo et al. 2022).Some reports indicated that both a carbohydrate-rich diet and a low-carbohydrate diet are associated with an increased risk of incident chronic kidney diseases (Gopinath et al. 2011;Zhang et al. 2022).People with CKD require careful dietary control; however, the relationship between carbohydrate intake and all-cause/cardiovascular mortality in this population is poorly understood.Therefore, further studies are warranted to elucidate this issue.
Carbohydrates are a type of fundamental macronutrient widely present in almost every food item and drink that is converted to glucose when the process of digestion takes place.Carbohydrates are commonly classified as either "simple" (monosaccharides and disaccharides) or "complex" (polysaccharides).Generally, simple carbohydrates tend to have a high glycemic index, while complex and fibrous carbohydrates have a low glycemic index.Disease conditions are related to carbohydrate consumption, and in chronic diseases, complete fruits, vegetables, beans and whole grains are the most appropriate carbohydrate sources, most of which are packed with dietary fibre and other potentially cardioprotective components (Mann 2007;Zhang et al. 2018).Previous researches indicated a significant association between low-fibre and high-sugar diets and an elevated risk of cardiovascular events and all-cause mortality (Riccioni et al. 2012;Threapleton et al. 2013;Atkins et al. 2016;Mirmiran et al. 2016;Gao et al. 2021).Notably, fibre and sugar primarily pertain to the forms of carbohydrates.However, since the complexity of carbohydrate sources, forms, and functions, further specific research is still needed to determine the exact association between carbohydrates from different sources and all-cause/cardiovascular mortality in the CKD population.
This study aimed to investigate the association between carbohydrates from different sources as substitutes for other macronutrients and all-cause/cardiovascular mortality among patients with CKD and quantitatively assess the mediated effects of fibre and sugar on this association.

Population and study setting
This study was based on data from the National Health and Nutrition Examination Survey (NHANES) from 1999 to 2018 (Supplemental Figures 1-19).We selected the participant cohort if they met the following criteria: (1) aged > 18 years and completed at least 1 dietary recall; and (2) estimated glomerular filtration rate (eGFR) < 90 mL/min/1.73m 2 (calculated by CKD-Epidemiology Collaboration (EPI) equations) or urinary albumin >200 mg/L.In this study, the data from individuals with CKD stages 1-2 and 3-5 were analysed separately.(3) Participants without follow-up information, with extreme dietary intake (<800 or >4200 kcal/d for men and <600 or >3500 kcal/d for women) and pregnant women were excluded.Finally, the total number of participants involved in the study was 116,876, with 11,302 participants included and 105,574 participants excluded from the analysis.The median follow-up time for the cohort was 107 months.
The study design and data collection were performed according to the NHANES protocol updated in December 2021.The interview weight (2 years and 4 years) was used in our study to account for the complex study design, survey nonresponse, and poststratification adjustment.The data collected in this study were reviewed and approved by the National Centre for Health Statistics (NCHS) Ethics Review Board (ERB).

Collection of dietary information
Dietary information was obtained via a 24-hour dietary recall interview in the NHANES.The first interview was conducted in person at the NHANES Mobile Examination Centre (from 1999-2001), and a second recall was performed by telephone approximately 3-10 days later (after 2002).Data on the total energy intake (kcal/d), carbohydrate intake (g/d), protein intake (g/d), fat intake (g/d), and fibre intake (g/d) were collected.
We traced the carbohydrate sources from dietary information provided by the USDA Food and Nutrient Database for Dietary Studies (FNDDS).Briefly, food ingredients were identified via the FNDDS food code (Six et al. 2011).The carbohydrate content of each ingredient was determined by the National Nutrient Database for Standard Reference.Each ingredient was also recorded as one of 10 categories: milk products, nuts and legumes, starchy vegetables, vegetables, whole grains, other grains, condiments/sweets, fruits, fruit juice, and others.Whole grains were defined as products clearly marked as whole-wheat flour, bulgur, whole barley, whole-barley flour, oatmeal, oats, rye, whole cornmeal, popcorn, wild rice, brown rice, etc.Otherwise, they were classified as other grains."Others" mainly included animal products, wine and non-classifiable food.All the work was manually completed by the authors.We adjusted carbohydrate intake for total energy intake by using the residual method (Brown et al. 1994).

Demographic data and other covariates
Demographic variables were collected, which included age (years), sex (men/women), ethnicity (non-Hispanic white/non-Hispanic black/Mexican-American/other race), education level (college/more than college and less than college) and annual household income (<$20,000, $20,000 to $75,000, and ≥$75,000).Lifestyle behaviours were obtained from a self-report questionnaire that included the metabolic equivalent of the task for physical activity level (MET-PA), smoking (yes/no), and alcohol intake (g/day).Health conditions included self-reported hypertension (yes/no), self-reported diabetes (yes/no), body mass index (BMI, calculated as measured weight (kg) divided by the square of height (m2), urine albumin levels (mg/L), serum sodium (mmol/L), serum phosphorus (mmol/L) and serum potassium (mmol/L).

Outcome
Death from any cause was the primary outcome of this analysis.We identified causes of death by using National Death Index data.Mortality from cardiovascular disease (CVD) was determined by International Classification of Diseases, Tenth Revision (ICD-10) codes for each participant (I I00-I09, I11, I13, I20-I51, I60-I69).Overall, 4058 deaths were recorded in our cohort, including 1385 recorded CVD-related events.

Descriptive statistics
The baseline characteristics of the participants were presented by quartiles of carbohydrate intake.Continuous variables are shown as the means (standard deviations), and categorical variables are shown as numbers (percentages).One-way ANOVA was used to compare the mean values of continuous variables.Chi-square tests were used to compare the percentages of categorical variables.

Cox and restricted cubic spline regression
Cox models were used to assess the association of carbohydrate intake (from different sources) with death events.The hazard ratio (HR) indicated the increase in death risk per carbohydrate increase of 100 g/day.The carbohydrate intake for total energy intake was determined by using the residual method.Model 1 was adjusted for demographic variables (age, sex, ethnicity, education level and annual household income).Model 2 was further adjusted for dietary information and lifestyle behaviours (protein intake, fat intake, carbohydrate intake from other sources, alcohol intake, smoking and MET-PA).Model 3 was further adjusted for health conditions (BMI, diabetes, hypertension, urine albumin, eGFR, serum sodium, serum phosphorus, serum potassium, CVD and cancer).
Restricted cubic spline regression was used to determine the dose-response association between carbohydrate intake and outcome among CKD patients.The first and last knots were placed at the 0.01 quantile and 0.99 quantile of carbohydrate intake, respectively.The middle knots were chosen by piecewise linear regression.The nonlinearity of the association was tested by Frank Harrell's methods.

Nutrient substitution model
We used predicted isocaloric models to assess the extent to which replacing 1 serving of an assigned carbohydrate with 1 serving of another type of carbohydrate/protein/fat would impact total and cardiovascular mortality by maintaining total daily energy.We also calculated the effects of replacing 1 serving of an assigned carbohydrate with 1 serving of carbohydrates from another source on mortality among CKD patients.The HR and the 95% confidence interval (CI) were calculated by Cox regression.

Subgroup and sensitivity analyses
We also performed a subgroup analysis divided by CKD stage, age, sex, diabetes, hypertension, cancer and CVD.B Schneider's regression models were used to assess the interaction between subgroups (Schneider 1989).Sensitivity analysis was conducted to test the robustness of our findings with the following criteria: (1).The specific carbohydrate intake as a percentage of energy from carbohydrate intake was substituted for carbohydrate intake amount; (2).participants whose cause of death was unrelated injuries were excluded; (3).participants with a follow-up time of <2 years were excluded; (4) participants whose cause of death was malignant neoplasms were excluded; and (5) participants with mild CKD (stages 1-2) were excluded.

Mediation effect analysis
To explore whether the effect of carbohydrates from different sources depending on their fibre proportion, we performed an analysis to investigate the mediated effect of fibre intake on the association between carbohydrate intake and the outcome.Briefly, the linear relationship between fibre intake from specific sources and other variables was calculated first.Then, the survival regression model was used to fit the time-dependent outcome and all variables.The average causal mediation effect (ACME) and total effect were determined according to Kosuke Imai's methods (Imai et al. 2010).The 95% CI was identified by the bootstrap method with 2000 replications.
A two-sided P value <0.05 was considered statistically significant.Cox regression was performed with the "survival" package.Cubic spline analysis was performed with the "rcs" package.Subgroup analysis adopted the "Publish" package.The mediation effect utilised the "mediation" package using R version 3.6.1.

Carbohydrate intake and study population characteristics
Among Next, we evaluated whether the baseline study population characteristics varied depending on the quartile of total carbohydrate intake (Supplemental Table 1).Individuals with consumptions in the lower quartiles of total carbohydrates were more possibly to be female (1759 [30.31%]), to be Mexican American (416 [30.59%]), to have lower education and income, to be current smokers (1445 [25.96%]), to have a lower BMI (mean [SD],28.11 [7.8]

Association of carbohydrate intake from different sources with all-cause mortality/cardiovascular mortality
We researched the risk of all-cause mortality and cardiovascular mortality among populations with CKD stages 1-5 with each 100-g increase in total carbohydrate intake or 10% increase in total energy intake, and no significant linear association was observed (p > 0.05, Supplemental Table 2).The association between carbohydrate consumption and the risk of all-cause/cardiovascular mortality was significantly nonlinear, resulting in a U-shaped association, with the lower observed risk associated with a carbohydrate consumption of 50-60% (Figures 1 and 2).Carbohydrate consumption below or above this range appeared to increase the risk of all-cause mortality and cardiovascular mortality.
For cardiovascular mortality, an increase of 100 g/d in carbohydrate intake from vegetables was significantly associated with fewer deaths (Table 1, vegetables: HR = 0.58, 95% CI = 0.37-0.91,p = 0.02).In addition, the P value from the hazard ratio for fruits (raw) was close to the critical value (0.05), which suggested that an increase of 100 g/d in carbohydrate intake from fruits (raw) was also associated with lower cardiovascular mortality (Table 1, fruits (raw): HR = 0.71, 95% CI = 0.51-0.99,p = 0.05).

Examination of the dose-response effect
Restricted cubic spline functions incorporated carbohydrates from nine sources.Figure 3 shows the doseresponse effect examination results for carbohydrates from nine sources and all-cause mortality.The overall dose-response effect examination results showed that the intake of whole grains and fruits (raw) had an L-shaped nonlinear correlation with all-cause mortality, which indicated that there was a negative correlation effect within a certain range, and no further obvious benefit was observed from increased intake (whole grains: p value for nonlinear association <0.001, fruits  (c and d) cKd stages 1-2 patients; and (e and f) cKd stages 3-5 patients.the analysis was conducted by adjusting for demographic variables (age, sex, ethnicity, education level and annual household income) + dietary information and lifestyle behaviours (protein intake, fat intake, carbohydrate intake from other sources, alcohol intake, smoking and Met-Pa) + health conditions (BMI, diabetes, hypertension, urine albumin, eGfr, serum sodium, serum phosphorus, serum potassium, cVd and cancer).the value of the X-axis represents the residual carbohydrate intake adjusted by total energy intake.
(raw): p value for nonlinear association <0.001).Additionally, the intake of nuts and legumes was negatively correlated with an increased risk of all-cause mortality within a certain intake range (nuts/legumes: p value for nonlinear association <0.001).Higher intake of vegetables was negatively correlated with an increased risk of all-cause death (vegetables: p value for linear association = 0.002).In contrast, an increased intake of fruit (juice) and condiments/sweets increased the risk (fruits (juice): p value for linear association = 0.008, condiments/sweets: p value for linear association < 0.001).Figure 4 shows the dose-response effect examination results for carbohydrates from nine sources and Cardiovascular mortality; we observed similar results.and (e and f) cKd stages 3-5 patients.the analysis was conducted by adjusting for demographic variables (age, sex, ethnicity, education level and annual household income) + dietary information and lifestyle behaviours (protein intake, fat intake, carbohydrate intake from other sources, alcohol intake, smoking and Met-Pa) + health conditions (BMI, diabetes, hypertension, urine albumin, eGfr, serum sodium, serum phosphorus, serum potassium, cVd and cancer).the value of the X-axis represents the percentage of energy from carbohydrate intake.

Substitution models for assigned carbohydrates replaced by other macronutrients
To study the effects of carbohydrates from different sources, we established multivariable-adjusted models to investigate the all-cause/cardiovascular mortality associated with a simultaneous increase of 100 g per day of one specific carbohydrate and a decrease in the intake of another daily carbohydrate, protein or fat.We found an increase in carbohydrates from whole grains, fruits (raw) and vegetables, and a decrease in carbohydrates from other sources and fat was related to a lower risk of death among people with CKD ( Next, we further explored the effect of substituting carbohydrates from nine different sources on all-cause mortality.Overall, the results showed that substituting carbohydrates from fruits (raw), whole grains and vegetables for carbohydrates from other sources could reduce the risk of all-cause mortality (Supplemental Table 3, HR values between 0.66 and 0.87, p < 0.05).Most substitutions for carbohydrates from fruits (juice) seemed to increase all-cause mortality (Supplemental Table 3, HR values between 1.25 and 1.51, p < 0.05).We found that substitution of carbohydrates from whole grains, fruits (raw) and vegetables seemed to have better benefits in reducing cardiovascular mortality (Supplemental Table 4, HR values between 0.72 and 0.81, p < 0.05).

Subgroup analyses and sensitivity analyses
The subgroup analyses of carbohydrates from different sources in the CKD group are presented in the Supplemental Materials.In the subgroup analysis of carbohydrates from fruits, people aged under 60 years obtained a better benefit of lower all-cause mortality than those aged above 60 years (Supplemental Figure 4, P value for interaction = 0.018).Carbohydrates from fruits were more successful in reducing all-cause mortality and cardiovascular mortality among people with CKD with BMIs less than 25 (Supplemental Figure 4, P value for interaction =0.012).Additionally, the association of carbohydrates from fruits with a lower risk of total mortality was more obvious among CKD patients with cancer (Supplemental Figure 4, P value for interaction =0.009).In the subgroup analysis of carbohydrates from condiments/ sweets, a higher risk of total mortality was observed among the subjects with CKD stage 5 (Supplemental Figure 9, P value for interaction =0.043).We did not find a significant effect on the benefits of carbohydrates from vegetables in subgroups stratified by sex, age, BMI, diabetes, cancer, CVD, urinary albumin, or CKD stage (Supplemental Figure 6, P value for interaction >0.05).
In sensitivity analyses, the conversion of carbohydrate intake into a percentage and removal of a specific population did not materially alter our findings in the CKD populations (Supplemental Tables 3-8).

Mediation analysis for fibre and sugar
we performed mediation analysis to examine whether and to what extent fibre and sugar (i) milk and dairy products.the analysis was conducted by adjusting for demographic variables (age, sex, ethnicity, education level and annual household income) + dietary information and lifestyle behaviours (protein intake, fat intake, carbohydrate intake from other sources, alcohol intake, smoking and Met-Pa) + health conditions (BMI, diabetes, hypertension, urine albumin, eGfr, serum sodium, serum phosphorus, serum potassium, cVd and cancer).the value of the X-axis represents the residual of carbohydrate intake adjusted by total energy intake.mediated the association between carbohydrate consumption and all-cause mortality and cardiovascular mortality.Supplemental Tables 7 and 8 show the total effect of carbohydrates and the average causal mediated effect (ACME) of fibre and sugar.The results indicated that fibre was an important mediator between carbohydrates from whole grains and fruits and vegetables and all-cause mortality (p < 0.05), accounting for 80.55%, 39.55% and 59.14% of the association, respectively.We also observed similar results for cardiovascular mortality, in which fibre contributed 69.29% of the effect of carbohydrates from fruits and 45.55% of the effect of carbohydrates from vegetables.Nevertheless, an average causal mediated effect of sugar was not observed.

Discussion
In this prospective cohort study of CKD patients, a U-shaped relationship was observed between the intake of total carbohydrates and all-cause mortality/ cardiovascular mortality.Our findings suggested that 50%-60% carbohydrate consumption may be beneficial (i) milk and dairy products.the analysis was conducted by adjusting for demographic variables (age, sex, ethnicity, education level and annual household income) + dietary information and lifestyle behaviours (protein intake, fat intake, carbohydrate intake from other sources, alcohol intake, smoking and Met-Pa) + health conditions (BMI, diabetes, hypertension, urine albumin, eGfr, serum sodium, serum phosphorus, serum potassium, cVd and cancer).the value of the X-axis represents the residual of carbohydrate intake adjusted by total energy intake.
in reducing the risk of all-cause mortality/cardiovascular mortality.In this study, carbohydrates from whole grains, fruits and vegetables and fewer carbohydrates from fruit juices and sauces were associated with lower all-cause/cardiovascular mortality among CKD patients (stages 1-2 and 3-5).Carbohydrate intake from raw fruits and non-starchy vegetables substituted for fat intake but not protein intake was associated with fewer mortality events.Among the carbohydrates with good benefits, such as those from vegetables and fruits, fibre mediated more than 40% of the observed association.These findings suggest that a change in carbohydrate sources (e.g.vegetables or fruits) can reduce all-cause mortality and cardiovascular mortality.
The daily diet of patients with chronic kidney disease has been widely discussed (Goraya and Wesson 2015;Rysz et al. 2017;Di Renzo et al. 2021).In addition to extensive research on protein intake (Piccoli et al. 2015;Kiuchi et al. 2016;Piccoli et al. 2016;Jhee et al. 2020), the effect of carbohydrate intake needs to be investigated.Excessive carbohydrate intake leads to metabolic syndrome, accompanied by obesity and diabetes, which are common risk factors for CKD (Thomas et al. 2011;Liu et al. 2019).Furthermore, a diet high in carbohydrates can raise the risk of chronic kidney disease onset in nondiabetic individuals and can also raise the risk of all-cause mortality in individuals with prediabetes (Nam et al. 2019;Mohammadifard et al. 2022).Numerous studies indicated that consumption of a low-carbohydrate diet was associated with a higher risk of all-cause mortality among those with CKD (Fung et al. 2010;Farhadnejad et al. 2019;Mazidi et al. 2019;Zhang et al. 2022).Multiple studies elucidated a significant correlation between carbohydrate intake, carbohydrate quality index, and protein-tocarbohydrate ratio with overall mortality in adult populations (Kwon et al. 2020; Fernandez-Lazaro  et al. 2021b; Lee et al. 2021;Wabo et al. 2022).The Dietary Guidelines for Americans recommends that carbohydrates make up 45-65% of the total daily calories, which is an important contribution to daily caloric needs.In our study, we found a U-shaped association between carbohydrate intake and all-cause/ cardiovascular mortality and found that carbohydrate intake should be maintained in the range of 50-60% for CKD patients, which is equivalent to a healthy intake.For patients with CKD, it holds paramount significance to uphold an appropriate ratio of carbohydrates within their overall daily caloric intake and to adapt their dietary habits concerning the origins of carbohydrates.
A series of studies, as well as data from large population-based observational studies, have shown that different carbohydrate types perform various functions in chronic disease (Ludwig et al. 2018;Reynolds et al. 2019;Munoz-Cabrejas et al. 2021).A higher intake of energy-dense, nutrient-poor sources of carbohydrates was associated with a risk of developing moderate CKD over 5 years, such as soft drinks, cookies, and cakes (Gopinath et al. 2011).However, the increased intake of carbohydrates rich in fibre can delay the progression of CKD (Lu et al. 2017).To provide practical recommendations for optimising carbohydrate intake for CKD patients, we divided carbohydrates into nine categories and explored their relationship with all-cause/CVD mortality in the CKD population, which similarly showed a heterogeneous effect of carbohydrates from different sources.The dose-effect relationship results showed that carbohydrates from whole grains, vegetables and fruits had a negative relationship with all-cause mortality, whereas carbohydrates from juice and sweets had a nonlinear positive relationship with all-cause mortality.These results suggested that the effect of carbohydrates on all-cause/cardiovascular mortality could vary depending on the sources of carbohydrates in the CKD population.Any assessment of the association between carbohydrates from different sources and total mortality among subjects with CKD must take into account each macronutrient, such as protein and fat.According to our results, carbohydrates from whole grains, raw fruits and vegetables could reduce all-cause mortality after substituting them for an equal amount of carbohydrates from other sources of fat, except protein.As total protein intake in the CKD group was generally in the lower normal range, previous studies have found that reducing protein intake can increase the risk of end-stage renal disease (ESRD) or death in CKD populations, which may be due to less protein intake aggravating the progression of protein energy wasting (Huang et al. 2008;Lee et al. 2019).A previous study has shown that metabolic acidosis in CKD patients will be aggravated by a large intake of meat and refined grains, leading to an increase in dietary acid load, whereas eating fruits and vegetables may neutralise acidosis and its harmful consequences (Cases et al. 2019).Despite the relatively high potassium content of plant-based diets, such as those rich in fruits and vegetables, they do not seem to have a negative impact on the prognosis of CKD patients (Joshi et al. 2019).Further, choosing fruits and vegetables with low potassium content and using cooking techniques that reduce potassium and phosphorus levels in these foods may help lower the risk of hyperkalemia.Altogether, our results recommend that CKD subjects increase high-quality carbohydrate intake but maintain an appropriate proportion of protein.
Carbohydrates are essential nutrients that include fibres, sugar and starches, and fibre is widely studied (Li et al. 2020;Yang et al. 2021;Nakano et al. 2022;Su et al. 2022).Traditional dietary recommendations for CKD patients limit the intake of vegetables and fruits due to their high potassium content, but this paradigm is changing rapidly because the high fibre content in these foods enhances intestinal motility and short-chain fatty acid production, thereby reducing the production of the most harmful uraemic toxins (Cases et al. 2019;Khosroshahi et al. 2019;Kramer 2019).Previous studies suggested that increasing fibre intake in CKD patients by eating fibre-supplemented foods may reduce serum creatinine levels and improve eGFR (Salmean et al. 2013).Moreover, there is clear evidence that a harmful proinflammatory state exists in CKD and that this state is aggravated in end-stage kidney disease (Massy et al. 2009), and several studies show that increased fibre in the diet can reduce blood pressure and inflammation (Ma et al. 2008;Dong et al. 2022;Shivakoti et al. 2022).Compared to fruit juices, whole fruits are rich in fibre and have moderated to low glycemic load (GL), which was associated with a lower risk of type 2 diabetes, cardiovascular disease and all-cause mortality in prospective cohort studies (Du et al. 2016;Aune et al. 2017).In this study, our results suggested that the association of carbohydrates from fruits and vegetables with all-cause/ cardiovascular mortality was approximately half due to fibre, which suggests that there are many other mediations worth exploring, such as simple sugars (fructose) and complex sugars (soluble fibre).
There are several limitations in this study.Firstly, due to the nature of the observational design, causality could be assumed, as residual confounding could not be completely excluded.Secondly, we manually classified carbohydrates into nine categories according to their sources, and the heterogeneity of carbohydrate properties in each source could not be completely removed.Thirdly, the data on dietary intake were collected by a food frequency questionnaire (FFQ), which may be prone to recall bias and may not be representative of habitual dietary intake.Fourth, due to the absence of some ICD-10 codes, cardiovascular death events may have been underestimated and led to type II errors.Nevertheless, our study provides a practical recommendation for optimising carbohydrate sources for CKD patients for the first time.The substitution analysis strengthens the strategy of macronutrient combination.

Conclusion
In summary, the association between carbohydrates from different sources and all-cause/cardiovascular mortality in the CKD population was first explored in this study, filling the gap in carbohydrate benefits in CKD.Given the considerable health burden of CKD, recommendations for appropriate and healthy carbohydrate intake will help prevent further aggravation of kidney disease and cardiovascular injury.Our findings will provide a scientific carbohydrate intake reference for patients with CKD.

Disclosure statement
No potential conflict of interest was reported by the author(s).

Figure 1 .
Figure1.restricted cubic spline models for the relationship between an increase in carbohydrate intake and all-cause/cardiovascular mortality among (a and b) all cKd patients; (c and d) cKd stages 1-2 patients; and (e and f) cKd stages 3-5 patients.the analysis was conducted by adjusting for demographic variables (age, sex, ethnicity, education level and annual household income) + dietary information and lifestyle behaviours (protein intake, fat intake, carbohydrate intake from other sources, alcohol intake, smoking and Met-Pa) + health conditions (BMI, diabetes, hypertension, urine albumin, eGfr, serum sodium, serum phosphorus, serum potassium, cVd and cancer).the value of the X-axis represents the residual carbohydrate intake adjusted by total energy intake.

Figure 2 .
Figure2.restricted cubic spline models for the relationship between an increased percentage of energy from carbohydrate intake and all-cause/cardiovascular mortality among (a and b) all cKd patients; (c and d) cKd stages 1-2 patients; and (e and f) cKd stages 3-5 patients.the analysis was conducted by adjusting for demographic variables (age, sex, ethnicity, education level and annual household income) + dietary information and lifestyle behaviours (protein intake, fat intake, carbohydrate intake from other sources, alcohol intake, smoking and Met-Pa) + health conditions (BMI, diabetes, hypertension, urine albumin, eGfr, serum sodium, serum phosphorus, serum potassium, cVd and cancer).the value of the X-axis represents the percentage of energy from carbohydrate intake.

Figure 3 .
Figure 3. restricted cubic spline models for the relationship between an increase of 100 g in carbohydrate intake and all-cause mortality.(a) Whole grains; (b) other grains; (c) fruits (raw); (d) fruits (juice); (e) vegetables; (f) starchy vegetables; (g) nuts and legumes;(h) condiments/sauces and sweets; (i) milk and dairy products.the analysis was conducted by adjusting for demographic variables (age, sex, ethnicity, education level and annual household income) + dietary information and lifestyle behaviours (protein intake, fat intake, carbohydrate intake from other sources, alcohol intake, smoking and Met-Pa) + health conditions (BMI, diabetes, hypertension, urine albumin, eGfr, serum sodium, serum phosphorus, serum potassium, cVd and cancer).the value of the X-axis represents the residual of carbohydrate intake adjusted by total energy intake.

Figure 4 .
Figure 4. restricted cubic spline models for the relationship between an increase of 100 g of carbohydrate intake from nine sources and cardiovascular mortality.(a) Whole grains; (b) other grains; (c) fruits (raw); (d) fruits (juice); (e) vegetables; (f) starchy vegetables; (g) nuts and legumes; (h) condiments/sauces and sweets;(i) milk and dairy products.the analysis was conducted by adjusting for demographic variables (age, sex, ethnicity, education level and annual household income) + dietary information and lifestyle behaviours (protein intake, fat intake, carbohydrate intake from other sources, alcohol intake, smoking and Met-Pa) + health conditions (BMI, diabetes, hypertension, urine albumin, eGfr, serum sodium, serum phosphorus, serum potassium, cVd and cancer).the value of the X-axis represents the residual of carbohydrate intake adjusted by total energy intake.
This study was supported by a grant from the National Science Foundation of China.(No. 82374406), the Young Elite Scientists Sponsorship Program by CAST (CACM-2021-QNRC2-B30), Natural Science Foundation of Guangdong Province/Guangzhou City (2021A1515011457 and 202102020269), Guangzhou Science and Technology Project (202206080015), and National Clinical Research Base of Traditional Chinese Medicine (No. [2018]131), Guangzhou University of Chinese Medicine's Youth Elite Talents Cultivation "List Unveiling and Leadership" Team Project.

Table 1 .
the association between an increase of 100 g/d in carbohydrates from different sources with all-cause mortality/cardiovascular mortality.was adjusted for demographic variables (age, sex, ethnicity, education level and annual household income).the energy intakes were adjusted by the residual method.the model 2 was adjusted for covariates in model 1 + dietary information and lifestyle (protein intake, fat intake, carbohydrate intake from other sources, alcohol intake, smoking and Met-Pa). the model 3 was adjusted for covariates in model 2 + health conditions (BMI, diabetes, hypertension, urine albumin, eGfr, serum sodium, serum phosphorus, serum potassium, cVd and cancer).

Table 2 .
Statistical Model based on hazard ratios (95% confidence intervals) for all-cause mortality associated with a decrease of 100 g per day of a specific-source carbohydrate and simultaneous increase of 100 g per day of carbohydrates from other sources/fat/protein in cKd population.regression analysis was conducted by adjusting for demographic variables (age, sex, ethnicity, education level and annual household income) + dietary information and lifestyle (protein intake, fat intake, carbohydrate intake from other sources, alcohol intake, smoking and Met-Pa) + health conditions (BMI, diabetes, hypertension, urine albumin, eGfr, serum sodium, serum phosphorus, serum potassium, cVd and cancer).the energy intakes were adjusted by the residual method.results with statistical significance (p < 0.05) have been marked in bold.

Table 3 .
Statistical model based on hazard ratios (95% confidence intervals) for cardiovascular mortality associated with a decrease of one serving per day of a specific-source carbohydrate and simultaneous increase of one serving per day of carbohydrates from other sources/fat/protein in cKd population.