Chronic kidney disease (CKD) represents a critical public health challenge, manifesting through the progressive and irreversible destruction of nephrons(Reference Webster, Nagler and Morton1). Characterised by a sustained decline in glomerular filtration rate (typically less than 60 ml/min per 1·73 m2) or by enduring signs of kidney injury over a minimum duration of 3 months(Reference Webster, Nagler and Morton1,Reference Hill, Fatoba and Oke2) , CKD contributes significantly to the financial strain on individuals and healthcare systems alike(Reference Mills, Xu and Zhang3). This condition consumes approximately 2–3 % of annual healthcare expenditure in numerous developed countries and ranks among the top ten non-communicable diseases globally(Reference Bello, Levin and Tonelli4). Importantly, deaths attributed to CKD rose from 0·9 million in 2005 to 1·2 million by 2017(Reference Bello, Levin and Tonelli4,Reference Bikbov, Purcell and Levey5) .
Considering these statistics, the emphasis on primary preventive measures has intensified, with previous studies highlighting a strong correlation between lifestyle factors such as diet, physical activity and smoking and the risk of developing CKD(Reference Hallan, de Mutsert and Carlsen6,Reference Whitham7) . Dietary habits have shown a greater impact on both the development and prevention of CKD compared with other modifiable risk factors(Reference Strippoli, Craig and Rochtchina8). Traditional research often focused on the impact of specific nutrients (macro or micro) or particular foods(Reference Geng, Jafar and Neelakantan9). However, considering that dietary habits involve a complex interplay of various food groups (and not only specific nutrients), a broader analysis may yield more comprehensive insights into preventing CKD(Reference Estruch, Ros and Salas-Salvadó10,Reference Schwedhelm, Boeing and Hoffmann11) .
Emerging research in this area is promising. For instance, results from a cohort study suggested that there is a negative relationship between high intake of low-fat dairy products and annual decline in the estimated glomerular filtration rate(Reference Herber-Gast, Biesbroek and Verschuren12). Similarly, data from the Multi-Ethnic Study of Atherosclerosis (MESA) linked higher intakes of fruits and whole grains to improved kidney function markers(Reference Nettleton, Steffen and Palmas13). Conversely, studies such as the Northern Manhattan Study (NMS) and the Framingham Heart Study (FHS) have shown mixed results regarding the impact of these food groups on kidney health indicators(Reference Khatri, Moon and Scarmeas14,Reference Foster, Hwang and Massaro15) . Moreover, an elevated risk of CKD has been associated with high intake of red and processed meats. On the other hand, substituting these with healthier options like legumes, low-fat dairy products and nuts correlates with a reduced CKD risk(Reference Haring, Selvin and Liang16). Despite these insights, the association between the risk of CKD and all food groups remains under debate.
The current investigation seeks to provide a systematic review of prospective cohort studies examining the long-term association between the consumption of various food groups and the risk of CKD, aiming to address existing knowledge gaps and guide future dietary recommendations.
Methods
We conducted the current systematic review and meta-analysis in accordance with methodologies delineated in previously published meta-analyses that examined the association between various dietary categories and the risk of hypertension(Reference Schwingshackl, Schwedhelm and Hoffmann17), type 2 diabetes(Reference Schwingshackl, Hoffmann and Lampousi18), CHD, stroke and heart failure(Reference Schwingshackl, Bechthold and Schwedhelm19), colorectal cancer(Reference Schwingshackl, Schwedhelm and Hoffmann20) and all-cause mortality(Reference Schwingshackl, Schwedhelm and Hoffmann21). Preference was given to the analysis of food groups rather than nutrients to facilitate interpretation and enhance relevance for broader populations. This methodological decision was stipulated in advance in a protocol that has been published previously(Reference Schwingshackl, Chaimani and Bechthold22). Compliance with the recommended reporting standards was ensured by adhering to the guidelines provided by the Meta-analyses of Observational Studies in Epidemiology (MOOSE)(Reference Stroup, Berlin and Morton23). Our study protocol was registered under the CRD42023465532 number in the PROSPERO International Prospective Register of Systematic Reviews (www.crd.york.ac.uk/prospero/index.asp)
Search strategy
Two investigators (SS, ST) conducted searches on PubMed, Web of Science and Google Scholar up until January 2025, without any date or language restrictions (online Supplementary Table 1). To ensure a comprehensive search and to minimise the risk of missing relevant studies, we performed a manual systematic search of Google Scholar in addition to our primary database searches. We searched the first twenty pages of results using the search term: ‘Food groups AND CKD’. This manual search was conducted to identify any potential articles that might have been missed by our primary search strategy. Any articles identified in this manual search that met our pre-defined inclusion criteria were then included in the final analysis. Moreover, reference lists of relevant systematic reviews, meta-analyses and original research articles were screened to identify additional studies. A consensus was reached through investigator consultation.
Study selection
All paper titles and abstracts were independently examined by two researchers (SS, ST). Cohort and nested case–control studies were included provided they met the following criteria: (1) reporting on at least one of nine predetermined dietary groups (vegetables, fruits, nuts, legumes, eggs, dairy products, fish and red meat); (2) outcomes described as incidence of CKD, end-stage renal disease (ESRD), estimated glomerular filtration rate decline, estimated glomerular filtration rate or creatinine clearance at end of study follow-up or serum creatinine; (3) participants aged 18 years or older; (4) standardised risk estimates presented as OR, relative risk (RR) or HR. Studies reporting mean differences or linear associations (ß-coefficients) between food groups and CKD risk were excluded from the analysis.
Data extraction
Two investigators (SS, ST) independently extracted relevant data from included studies, including author names, publication years, study locations, cohort names, number of participants, incident CKD cases, baseline characteristics (age, sex), length of follow-up, dietary and outcome assessment methods, multivariable effect estimates (RR, HR or OR including the corresponding CI) and adjusted covariates. Discrepancies were resolved by a third investigator (HM).
Statistical analysis
Using the DerSimonian and Laird technique(Reference DerSimonian and Laird24) to conduct a random-effects meta-analysis, summary RR and 95 % CI were computed to examine the relationships between the categories with highest v. lowest intake as well as the dose–response estimate for each of the nine food groups and the risk of CKD. Initially, a high v. low meta-analysis was performed by comparing the highest and lowest dietary consumption categories. Subsequently, servings per day were utilised to examine the nine food groups as a continuous feature in a linear dose–response meta-analysis, as specified by Greenland and Longnecker(Reference Schwingshackl, Schwedhelm and Hoffmann17–Reference Schwingshackl, Schwedhelm and Hoffmann21,Reference Greenland and Longnecker25) . If, in included studies, exposures were reported in serving size but the amount was not specified, recommended conversions were applied as follows: A serving of dairy product is equal to 200 g and a serving of red meat, total meat, fruit and vegetables to 100 g. Furthermore, 50 g of poultry, 15 g of fish, 30 g of grains, 28 g of nuts, 50 g of legumes and 50 g of egg is equivalent to one serving(Reference Schwingshackl, Schwedhelm and Hoffmann17). For studies that have previously reported on linear dose–response patterns, direct data extraction was employed. In studies where a dose–response correlation was not found, the study-specific slope was computed. To utilise the above method, data on the distribution of cases, person-years and non-cases, as well as quantification for at least three exposure groups and the corresponding RR with matching 95 % CI, are required. If a study did not provide the distribution of person-years in individual categories but provided the number of cases plus total person-years or the total number of participants plus follow-up time, missing data at one-stage were computed as previously indicated(Reference Chêne and Thompson26,Reference Aune, Greenwood and Chan27) . If a study’s exposure category range was known but its mean values were not reported, the midpoint between the lower and higher values was determined. When dealing with open-ended categories, it was presumed that their width equals that of the neighbouring category.
Furthermore, by multiplying the RR and choosing an ideal intake of the previously described risk-reducing (calculated by 1 − RR∗reduced) and risk-increasing foods (calculated by 1 – 1/RR∗increased), we may determine the potential of foods to lower the risk of CKD. According to the criteria, the serving category with the strongest inverse correlation with the risk of CKD and no further change in the risk association qualifies as the optimum intake of a particular food group. We performed a sensitivity analysis except for studies that were not adjusted for caloric intake.
The I 2 statistic and the Q test were used to assess heterogeneity, with values from 0–20 % indicating very low heterogeneity, 20–50 % suggesting moderate heterogeneity, 50–90 % indicating high heterogeneity and above 90 % representing very high heterogeneity, where meta-analysis may not be appropriate(Reference Higgins and Thompson28). We should investigate potential small-study effects, such as publication bias, using Egger’s test and funnel plots(Reference Egger, Smith and Schneider29) if >= 10 studies were included, according to the Cochrane Handbook’s recommendation(Reference Cumpston, Li and Page30). We utilised a weighted, mixed-effects meta-analysis, applying an A1-stage method, to assess the dose–response curves. In cases where restricted cubic splines were not feasible due to insufficient data, we modelled the relationships with the best-fitting second-order fractional polynomial(Reference Crippa, Discacciati and Bottai31). The Stata version/se 17 (StataCorp) software was employed to perform all statistical analyses.
Risk of bias assessment
Included studies were evaluated for bias risk across four categories(Reference Sterne, Hernán and Reeves32): (1) exposure assessment (low risk of bias: validated, calibrated FFQ or 24-h recall, diet history or diet records (multiple days)), (2) assessment of outcome (low risk of bias: accepted clinical criteria, record linkage (International Classification of Diseases codes), self-reported and validated), (3) sufficient duration of follow-up (low risk of bias: ≥ 5 years) and (4) adjusted basic model (low risk of bias, ≥ 2 factors: e.g. sex, age and race-centre; if only one sex included, then ≥ 1 factor) and outcome relevant adjustments (low risk of bias, ≥ 3 factors: e.g. BMI (in kg/m2), smoking, energy intake and physical activity). Studies were considered to have a low risk of bias if no domains were rated as high or uncertain risk, as indicated previously(Reference Schwingshackl, Schwedhelm and Hoffmann21) (online Supplementary Table 2).
Quality of meta-evidence
The GRADE (Grading of Recommendations Assessment, Development and Evaluation) approach was applied to assess the certainty of evidence for each meta-analysis. GRADE assesses the certainty of evidence across four levels: high, moderate, low and very low. The assessment considers five domains that can decrease the certainty: risk of bias, inconsistency, indirectness, imprecision and publication bias. Factors that can increase the certainty of evidence include a large magnitude of effect, dose–response gradient and plausible residual confounding. High-certainty evidence indicates that further research is unlikely to change our confidence in the estimate of effect. Moderate-certainty evidence suggests that further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate. Low-certainty evidence indicates that further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate. Very low-certainty evidence suggests that we are very uncertain about the estimate of effect(Reference Guyatt, Oxman and Vist33).
Results
Literature search
As depicted in Fig. 1, the initial literature search yielded a total of 6460 publications. After the exclusion of 1224 duplicate records, 78 animal studies and 26 non-original articles, 5132 articles remained that met the initial inclusion criteria. These articles underwent a screening of titles and abstracts, resulting in the exclusion of 5074 records due to irrelevance. Subsequent detailed screening excluded an additional thirty-seven studies for various reasons delineated in online Supplementary Table 3, leaving twenty-one prospective cohort studies for final inclusion. These studies involved 29 484 cases across 615 005 participants(Reference Khatri, Moon and Scarmeas14,Reference Haring, Selvin and Liang16,Reference Asghari, Yuzbashian and Mirmiran34–Reference Lee, Park and Kwon52) .
Flow diagram of study selection.
Characteristics of studies
The studies incorporated in the meta-analysis predominantly targeted a healthy adult population. The studies spanned from 2010(Reference Lin, Hu and Curhan45) to 2024(Reference Lee, Park and Kwon52). The follow-up durations ranged from 3(Reference Mirmiran, Yuzbashian and Aghayan48) to 25(Reference Park, Xun and Tsinovoi49) years. The geographic distribution included eight studies from American(Reference Khatri, Moon and Scarmeas14,Reference Haring, Selvin and Liang16,Reference Kim, Caulfield and Garcia-Larsen41,Reference Lee, Howard and Mete43,Reference Lin, Hu and Curhan45,Reference Ma, Jacques and Hwang47,Reference Park, Xun and Tsinovoi49,Reference Rebholz, Crews and Grams50) regions and nine studies from Asian regions(Reference Asghari, Yuzbashian and Mirmiran34–Reference Geng, Jafar and Neelakantan37,Reference Jhee, Kee and Park40,Reference Kubo, Imano and Muraki42,Reference Lew, Jafar and Koh44,Reference Mirmiran, Yuzbashian and Aghayan48,Reference Lee, Park and Kwon52) , alongside two studies each from European(Reference Liu, Ye and Yang46,Reference Rey-García, Donat-Vargas and Sandoval-Insausti51) and Australian regions(Reference Gopinath, Harris and Flood38,Reference Gopinath, Harris and Flood39) . One cohort study included only female participants(Reference Lin, Hu and Curhan45), whereas the others included both genders. All studies, with the exception of one that used diet history for determining dietary intakes of food groups(Reference Rey-García, Donat-Vargas and Sandoval-Insausti51), employed various forms of FFQ. Online Supplementary Table 4 elaborates on the characteristics of the included studies.
A risk of bias assessment using the Risk Of Bias In Non-randomised Studies-of Interventions tool identified twenty studies as having a moderate risk of bias(Reference Geng, Jafar and Neelakantan9,Reference Khatri, Moon and Scarmeas14,Reference Haring, Selvin and Liang16,Reference Asghari, Yuzbashian and Mirmiran34–Reference Gaeini, Bahadoran and Mirmiran36,Reference Gopinath, Harris and Flood38–Reference Rebholz, Crews and Grams50,Reference Lee, Park and Kwon52) and one study as having a serious risk of bias(Reference Rey-García, Donat-Vargas and Sandoval-Insausti51). The potential biases could stem from unaccounted factors influencing the results, incomplete information on participant selection or bias for exposure assessment (online Supplementary Table 2).
Meta-analysis findings
Total dairy products
Analysis of seven cohort studies(Reference Khatri, Moon and Scarmeas14,Reference Gaeini, Bahadoran and Mirmiran36,Reference Geng, Jafar and Neelakantan37,Reference Gopinath, Harris and Flood39,Reference Kim, Caulfield and Garcia-Larsen41,Reference Kubo, Imano and Muraki42,Reference Ma, Jacques and Hwang47) involving 6472 cases of CKD among 81 271 participants revealed no significant association between dairy product intake levels and CKD risk (RR: 1·02; 95 % CI: 0·95, 1·10; n 7) (Table 1). Moreover, a linear dose–response meta-analysis suggested no increased risk with an additional 200 g/d of dairy product consumption (RR: 0·99; 95 % CI: 0·88, 1·12; n 3) (Table 1). The analysis confirmed no departure from linearity (P = 0·15 for non-linearity; P dose–response: 0·36; Fig. 2(a)).
Food group with a risk of chronic kidney diseases
RR, relative risk.
Dose–response associations of dietary dairy product intake and risk of chronic kidney disease. (a) Total dairy products, (b) low-fat dairy products and (c) high-fat dairy products in random-effects models. Solid lines represent the relative risk of the association between dietary dairy product intake and the risk of chronic kidney disease; dashed lines represent 95 % CI.
Low-fat dairy products
Four studies comprising 21 831 participants and 4607 CKD cases found no significant association between low-fat dairy product intake levels and CKD risk (RR: 0·87; 95 % CI: 0·72, 1·04; n 4) (Table 1). There was also no significant association between consuming an additional 200 g of low-fat dairy product each day and the likelihood of developing CKD (RR: 0·98; 95 % CI: 0·96, 1·01; n 3) (Table 1). Low-fat dairy product consumption was non-linearly associated with risk of CKD (P = 0·002 for non-linearity; P dose–response: 0·002). This non-linear association suggests a U–shaped curve with the most substantial risk reduction at intake levels between 500 and 1000 g/d (Fig. 2(b)).
High-fat dairy products
The analysis included two studies(Reference Haring, Selvin and Liang16,Reference Gaeini, Bahadoran and Mirmiran36) with 3139 CKD cases among 14 368 participants, indicating no substantial association between high-fat dairy product intake and CKD risk (RR: 0·86; 95 % CI: 0·71, 1·04; n 2) (Table 1). No linear association between each additional high-fat dairy product intake and risk of CKD was found (RR: 0·88; 95 % CI: 0·72, 1·08; n 2) (Table 1). Furthermore, no evidence of departure from linearity was found between high-fat dairy product intake and the risk of CKD (P = 0·198 for non-linearity; P dose–response: 0·108) (Fig. 2(c)).
Red meat
Examination of three studies involving 77 031 participants and 4196 cases of CKD(Reference Haring, Selvin and Liang16,Reference Lew, Jafar and Koh44,Reference Mirmiran, Yuzbashian and Aghayan48) showed a positive association between red meat consumption and increased CKD risk (RR: 1·39; 95 % CI: 1·13, 1·71; n 3) (Table 1). According to the results of the highest v. lowest category analysis, eating red meat can make people’s risk of getting CKD 39 % higher. In addition, the pooled RR was 1·34 (95 % CI: 1·08, 1·66; n 3) for a 100 g increment of red meat intake. The linear dose–response relationship was significant, with no departure from linearity detected (P = 0·483 for non-linearity; P dose–response: < 0·001) (Fig. 3(a)).
Dose–response associations of dietary meat intake and risk of chronic kidney disease. (a) Red meat, (b) poultry, (c) fish and (d) total meat in random-effects models. Solid lines represent the relative risk of the association between dietary meat intake and the risk of chronic kidney disease; dashed lines represent 95 % CI.
Poultry
Two prospective cohort studies with 72 150 total participants and 3583 cases assessed the association between poultry intake and risk of CKD(Reference Haring, Selvin and Liang16,Reference Lew, Jafar and Koh44) . Individuals in the highest quintile of poultry consumption did not exhibit a significantly increased risk of CKD compared with those in the lowest quintile (RR: 0·93; 95 % CI: 0·84, 1·03; n 2) (Table 1). Furthermore, the consumption of an additional 50 g of poultry daily did not demonstrate a strong association with CKD risk (RR: 0·89; 95 % CI: 0·73, 1·08; n 1), with no significant non-linearity observed (P = 0·057 for non-linearity; P dose–response: 0·085) (Fig. 3(b)).
Fish
Seven studies with a total of 490 744 participants and 15 360 cases of CKD were included in the analysis of fish intake(Reference Khatri, Moon and Scarmeas14,Reference Haring, Selvin and Liang16,Reference Kubo, Imano and Muraki42–Reference Lew, Jafar and Koh44,Reference Liu, Ye and Yang46,Reference Park, Xun and Tsinovoi49) . The pooled data indicated a protective effect of high fish consumption against CKD, with an RR of 0·88 for those in the highest compared with the lowest intake group (RR: 0·88; 95 % CI: 0·80, 0·97; n 7) (Table 1). The linear dose–response meta-analysis of five studies found a significant negative association between an additional 15 g/d of fish consumption and CKD risk (RR: 0·94; 95 % CI: 0·92, 0·96; n 5). In addition, the non-linear dose–response meta-analysis results indicated a negative linear association between fish intake and risk of CKD (P = 0·870 for non-linearity; P dose–response: < 0·001) (Fig. 3(c)).
Egg
Analysis of four studies with a combined cohort of 84 652 individuals and 4708 CKD cases investigated the impact of egg consumption on CKD risk(Reference Haring, Selvin and Liang16,Reference Kubo, Imano and Muraki42,Reference Lew, Jafar and Koh44,Reference Lee, Park and Kwon52) . There was no significant effect on CKD risk for individuals in the highest v. lowest egg consumption categories (RR: 0·99; 95 % CI: 0·94, 1·03; n 4) (Table 1). Furthermore, an increase of 50 g/d in egg consumption was not associated with a significant change in CKD risk (RR: 0·93; 95 % CI: 0·82, 1·06; n 2), and evidence suggested a departure from linearity (P = 0·564 for non-linearity; P dose–response: 0·057) (Fig. 4(f)).
Dose–response associations of the intake of other food groups and risk of chronic kidney disease. (a) Grains, (b) vegetables, (c) fruit, (d) nuts, (e) legumes and (f) egg in random-effects models. Solid lines represent the relative risk of the association between dietary meat intake and the risk of chronic kidney disease; dashed lines represent 95 % CI.
Total meat
A meta-analysis of six prospective cohort studies including 82 199 participants and 5489 CKD cases examined the impact of total meat consumption on CKD risk(Reference Khatri, Moon and Scarmeas14,Reference Asghari, Yuzbashian and Mirmiran34,Reference Kubo, Imano and Muraki42,Reference Lew, Jafar and Koh44,Reference Rebholz, Crews and Grams50,Reference Rey-García, Donat-Vargas and Sandoval-Insausti51) . There was no significant evidence suggesting that high total meat intake increases CKD risk (RR: 1·16; 95 % CI: 0·99, 1·35; n 6) (Table 1). However, a linear dose–response analysis combining two eligible studies revealed that an additional 100 g of total meat consumed daily is associated with a slight increase in CKD risk (RR: 1·02; 95 % CI: 1·01, 1·03; n 2). The non-linear dose–response meta-analysis results indicated a positive linear association between total meat intake and risk of CKD (P = 0·723 for non-linearity; P dose–response: 0·008) (Fig. 3(d)).
Grains
The inclusion of six studies with a total of 77 149 participants and 5329 CKD cases assessed the association between grain consumption and CKD risk(Reference Khatri, Moon and Scarmeas14,Reference Asghari, Yuzbashian and Mirmiran34,Reference Geng, Jafar and Neelakantan37,Reference Gopinath, Harris and Flood38,Reference Ma, Jacques and Hwang47,Reference Rebholz, Crews and Grams50) . A higher intake of grains was associated with a reduced likelihood of developing CKD compared with a lower intake (RR: 0·87; 95 % CI: 0·77, 0·99; n 6) (Table 1). No significant association was found for an additional 30 g of grains consumed daily with CKD risk RR: 0·96; 95 % CI: 0·90, 1·02; n 2), and there was no evidence of non-linearity (P = 0·210 for non-linearity; P dose–response: 0·119) (Fig. 4(a)).
Vegetables
Five prospective cohorts involving 83 776 participants revealed 6920 cases of CKD(Reference Khatri, Moon and Scarmeas14,Reference Bahadoran, Mirmiran and Momenan35,Reference Geng, Jafar and Neelakantan37,Reference Jhee, Kee and Park40,Reference Rebholz, Crews and Grams50) . The risk of CKD did not significantly differ between individuals with the highest and lowest vegetable intake (RR: 0·86; 95 % CI: 0·73, 1·02; n 5) (Table 1). Additionally, a linear dose–response meta-analysis showed that each additional 100 g of vegetables consumed daily did not significantly affect CKD risk (RR: 0·96; 95 % CI: 0·88, 1·04; n 2), with no departure from linearity noted (P = 0·074 for non-linearity; P dose–response: 0·067) (Fig. 4(b)).
Fruit
Analysis of six studies with 85 448 participants and 7003 CKD cases examined the potential impact of fruit consumption on CKD risk(Reference Khatri, Moon and Scarmeas14,Reference Asghari, Yuzbashian and Mirmiran34,Reference Geng, Jafar and Neelakantan37,Reference Jhee, Kee and Park40,Reference Ma, Jacques and Hwang47,Reference Rebholz, Crews and Grams50) . The findings indicated no strong association between high fruit consumption and CKD risk (RR: 0·95; 95 % CI: 0·84, 1·07; n 6) (Table 1). Furthermore, the linear dose–response meta-analysis showed that each additional 100 g of fruit consumed daily did not significantly influence CKD risk (RR: 1·00; 95 % CI: 0·99, 1·01; n 2), and no departure from linearity was observed (P = 0·738 for non-linearity; P dose–response: 0·310) (Fig. 4(c)).
Nuts
Two cohort studies involving 6975 cases of CKD among 26 638 participants were analysed to determine the relationship between dietary consumption of nuts and the risk of CKD(Reference Haring, Selvin and Liang16,Reference Kim, Caulfield and Garcia-Larsen41) . No significant correlation was observed between high and low nut intake (RR: 0·90; 95 % CI: 0·74, 1·10; n 2) (Table 1). A linear dose–response meta-analysis indicated that each additional 28 g of nuts consumed could reduce the CKD risk by 21 % (RR: 0·79; 95 % CI: 0·69, 0·91; n 1), with non-linear analysis revealing a U-shaped association between nut consumption and CKD risk, indicating that 10–20 g/d provided the most significant risk reduction (P = 0·045 for non-linearity; P dose–response: < 0·001) (Fig. 4(d)).
Legumes
After evaluating 4 studies involving 74 872 participants and 3879 CKD cases(Reference Khatri, Moon and Scarmeas14,Reference Haring, Selvin and Liang16,Reference Lew, Jafar and Koh44,Reference Ma, Jacques and Hwang47) , it was found that legume consumption could decrease the likelihood of developing CKD by 17 % (RR: 0·83; 95 % CI: 0·72, 0·92; n 4) (Table 1). Moreover, the pooled RR for a 50 g increment in legume intake was 0·87 (95 % CI: 0·78, 0·97; n 1), indicating a negative linear association between legume intake and CKD risk. There was no evidence of departure from linearity between dietary legume intake and risk of CKD (P = 0·430 for non-linearity; P dose–response: 0·033) (Fig. 4(e)).
Publication bias and sensitivity analyses
The RR was altered upon the exclusion of specific studies in sensitivity analyses for associations between various food groups and CKD risk. For example, removing the study by Gaeini et al. (Reference Gaeini, Bahadoran and Mirmiran36) significantly impacted the analysis of low-fat dairy products and CKD risk, while the exclusion of studies by Asghari et al. (Reference Asghari, Yuzbashian and Mirmiran34), Kubo et al. (Reference Kubo, Imano and Muraki42) and Khatri et al. (Reference Khatri, Moon and Scarmeas14) revealed similar results for the total meat intake and CKD risk analysis. The exclusion of the study by Geng et al. (Reference Geng, Jafar and Neelakantan37) significantly altered the primary analysis regarding vegetable consumption. Furthermore, when the study by Lew et al. (Reference Lew, Jafar and Koh44) is removed, the previously established significant positive association between consuming red meat and the risk of CKD becomes non-significant. The sensitivity analysis also revealed that when the study by Liu et al. (Reference Liu, Ye and Yang46) was excluded, the significant negative association between fish consumption and CKD became non-significant. We found similar results when we did not include the studies by Asghari et al. (Reference Asghari, Yuzbashian and Mirmiran34), Ma et al. (Reference Ma, Jacques and Hwang47), Gopinath et al. (Reference Gopinath, Harris and Flood38) and Khatri et al. (Reference Khatri, Moon and Scarmeas14) for the association between grains and the risk of CKD. However, no specific study significantly affected the other effect sizes.
After assessing publication bias based on Begg’s and Egger’s linear correlation test, we found possible publication bias in the association of red meat and grain intake and risk of CKD (P = 0·296 Begg’s test and P = 0·040 Egger’s test) (P = 0·060 Begg’s test and P = 0·002 Egger’s test). However, adjustment using the trim-and-fill method did not result in any alterations to the average RR. Finally, neither Egger’s nor Begg’s tests showed evidence of publication bias for the other outcomes.
Grading the evidence
We used the GRADE scoring system to rate the quality of evidence (online Supplementary Table 5). The certainty of the assessed evidence for all outcomes ranged from very low to moderate, with various downgrades assigned for serious risk of bias, indirectness, inconsistency and imprecision.
Discussion
To the best of our knowledge, this is the first meta-analysis to examine the association between the consumption of different food groups and the risk of CKD in a dose–dependent manner. Our results showed a statistically significant positive linear association between increased red meat consumption and heightened CKD risk. Specifically, an incremental daily intake of 100 g of red meat and total meat correlates with a substantial rise in the risk of CKD. Conversely, enhanced intakes of fish, dietary grains and legumes are significantly associated with decreased CKD risk. The results from a linear dose–response meta-analysis suggest that every additional 50 g/d of legumes, 28 g/d of nuts and 15 g/d of fish are associated with a 13–21 % reduction in CKD risk. However, non-linear dose–response analyses illustrate that nut consumption within a daily range of 10–20 g maximises CKD risk reduction. Despite these findings, no significant relationships were observed between CKD risk and the consumption of total dairy products, low-fat dairy products, high-fat dairy products, poultry, eggs, total meat, vegetables, fruits and nuts, except in the context of a U–shaped curve for low-fat dairy products, where 500–1000 g/d was negatively associated with the risk of CKD.
Our study revealed a statistically significant positive linear association between increased consumption of red meat and elevated risk of CKD. This observation aligns with the findings of He et al. (Reference He, Wu and Huang53), who (in their recent systematic review and meta-analysis) corroborated the association of red and processed meat consumption with increased CKD risk. Similarly, Van et al. (Reference Van Wymelbeke, Brondel and Bon54) reported a significant positive association between intake of red and processed meats and CKD risk. In contrast, a meta-analysis conducted by Kelly et al. (Reference Kelly, Palmer and Wai55) identified no significant impact of red meat consumption on the incidence of ESRD. Potential factors contributing to these discrepancies include geographical constraints of the study populations (USA and Sweden) and the reliance on self-reported dietary recalls. Moreover, the meta-analysis by Kelly et al. did not evaluate publication bias. The mechanisms hypothesised for the promotion of kidney disease progression by red meat involve its high protein content, which may induce changes in renal hemodynamics (including glomerular hyperfiltration), leading to kidney fibrosis. This process is mediated by elevated levels of transforming growth factor-β and enhanced signalling in mesangial cells(Reference Mafra, Borges and de Franca Cardozo56,Reference Kitada, Ogura and Monno57) . Furthermore, red meat intake may increase the production of uraemic toxins such as trimethylamine N-oxide by the gut microbiota, thereby augmenting the renal workload associated with solute excretion(Reference Mafra, Borges and de Franca Cardozo56).
Our study identified a significant inverse relationship between increased fish consumption and the risk of CKD. The potential mechanisms underlying this association may include the high levels of n-3 fatty acids present in fish, which can enhance creatinine clearance, attenuate inflammation and oxidative damage and inhibit fibrotic processes by downregulating transforming growth factor-β expression(Reference Liu, Ye and Yang46). These processes are critical pathomechanisms in the progression of kidney dysfunction. Supporting our findings, He et al. (Reference He, Wu and Huang53) and Bach et al. (Reference Bach, Kelly and Palmer58) both reported an inverse association between fish intake and the incidence of CKD and albuminuria. Conversely, Kelly et al. (Reference Kelly, Palmer and Wai55) observed no significant effects of fish intake on ESRD in their systematic review and meta-analysis. Possible reasons for these inconsistencies include variability in the adjustment of confounders and the predominance of cross-sectional study designs with inherent dietary recall bias.
Similarly, our outcomes indicate that the consumption of legumes and dietary grains may confer a protective effect against CKD. This advantageous relationship could be attributed to the diminished production of uraemic toxins and acid load, along with the inhibition of apoptosis in podocytes. Additionally, these dietary components may contribute to decreased inflammation and cell death(Reference Ko and Kalantar-Zadeh59). Bach et al. (Reference Bach, Kelly and Palmer58) and He et al. (Reference He, Wu and Huang53) corroborated these findings, noting a significant inverse association between the consumption of legumes and whole grains and the incidence of CKD and albuminuria. However, a meta-analysis by Kelly et al. (Reference Kelly, Palmer and Wai55) reported no significant impact of these dietary elements on ESRD, diverging from our findings. These inconsistencies could be due to disease severity (ESRD is the most severe stage of CKD, which is characterised by critical loss of kidney function), variability in outcome measurements and potential misclassification in dietary reporting over the long term.
Our investigation also highlights non-linear dose–response data indicating that daily nut consumption between 10 and 20 g optimises the reduction in CKD risk. In alignment with our results, Van et al. (Reference Van Wymelbeke, Brondel and Bon54) documented a protective effect of nut consumption against CKD. Nuts contain beneficial compounds such as phenols, fibre, vegetable protein, phytosterols, vitamins and minerals, which are known to ameliorate factors implicated in the pathogenesis of CKD including hyperglycaemia, insulin resistance, hypertension, dyslipidaemia, oxidative stress and inflammation(Reference Ros, Tapsell and Sabaté60). The anti-inflammatory properties of nuts, for instance, might alleviate glomerular injury and slow the progression of CKD. Furthermore, nuts may influence kidney function by maintaining gut microbiota balance and overall bowel health, selectively stimulating gut microbiota species that confer health benefits(Reference Wang, Qian and Hu61).
Strengths and limitations of the study
The present research has several strengths, including the application of both linear and non-linear dose–response meta-analyses, high-quality and consistency of the data and a large sample size from different geographic regions with diverse dietary habits. This study is also pioneering in assessing the impact of various food groups on CKD risk across published prospective cohort studies. Nonetheless, conclusions must be considered in the context of several limitations, including potential lifestyle confounders not accounted for in the included studies, measurement errors and misclassification inherent in FFQ-based dietary assessments. Furthermore, despite adjustments for numerous potential confounders, residual confounding cannot be excluded. Due to the observational nature of the studies, causality cannot be deduced.
Conclusion
This meta-analysis suggests a possible association between higher red meat consumption and an increased risk of CKD. Conversely, dietary patterns rich in fish, grains and legumes may be associated with a reduced risk. Moderate low-fat dairy product intake (500–1000 g/d) also showed a potential benefit, although the evidence suggests a possible non-linear relationship that requires further investigation. Based on these findings, clinicians could consider advising patients at risk for CKD to moderate their red meat consumption and increase their intake of plant-based foods. The potential role of moderate low-fat dairy product intake should also be considered. Public health messaging should promote balanced dietary patterns, highlighting the potential associations observed in this meta-analysis. It is important to note that these findings are based on observational studies, and further research, particularly large, long-term prospective studies, is needed to confirm these associations and to explore potential causal relationships across all stages of CKD. The complex relationship between low-fat dairy products and CKD also warrants further investigation to clarify its potential role.
Supplementary material
For supplementary material/s referred to in this article, please visit https://doi.org/10.1017/S0007114525104005.
Acknowledgements
This study was supported by the Tehran University of Medical Sciences’ Students’ Scientific Research Center (code: 1402.68636)
The authors’ responsibilities were as follows – S. S., H. M., H. I. and S. T. were responsible for conception, design and literature search; S. S. and S. T. performed data collection and analysis; S. S. and S. T. conducted interpretation of data and critical revision of the study material; S. S., F. N., H. I., S. M. and A. W. contributed to the drafting of the article; H. M. supervised the study; and all authors read and approved the final manuscript.
The authors report no conflicts of interest.
This trial was registered at PROSPERO as CRD42023465532.
