Abstract
Background and Aims:
Chronic kidney disease (CKD) is a severe public health burden, characterized by a gradual loss of kidney function over time. Diet is a modifiable lifestyle-related risk factor for CKD. However, there is uncertainty about which specific dietary patterns (DPs) are more beneficial or detrimental in CKD prevention.
We aimed at deriving DPs using an hybrid approach that combines a priori knowledge and data-driven methods to identify dietary factors that may affect kidney function in healthy and diseased subjects.
Method:
We analysed data of 8686 adults participating in the population-based Cooperative Health Research In South Tyrol (CHRIS) study. Kidney function was multiply assessed by the estimated glomerular filtration rate (eGFR), based on serum creatinine, using the 2021 CKD-EPI equation, the urinary albumin-to-creatinine ratio (UACR), and a kidney disease questionnaire. Participants were split between those free of diagnosed kidney disease, hypertension or diabetes (Group1, n=6133) and those diagnosed with any of the three conditions (Group2, n=2553). Diet was assessed through the self-administered and validated GA2LEN food frequency questionnaire (FFQ). The individual consumption of each food group was converted into portions per week and adjusted for total energy intake.
DPs were estimated using reduced rank regression (RRR), based on four FFQ-derived nutrient mediators (total daily dietary protein; potassium; sodium; and phosphorus intake) selected based on known effects on kidney health.
Generalized cross-validation identified an optimal number of 3 DPs. Factor Loading (FL)-based scores, either as continuous or stratified into sex-stratified tertiles (T1-T2-T3), were included in multiple-adjusted linear regression models for eGFR and log(UACR).
Results:
Group1 participants (53.4% females) were younger and presented better kidney health (median, Mdn age 39.7 years; Mdn eGFR 101.8 ml/min/1.73m2, interquartile range, IQR 91.5-112.1; Mdn UACR 5.2 mg/g, IQR 3.5-8.8) than Group2 participants (Mdn age 57.2 years; Mdn eGFR 90.8 ml/min/1.73 m2, IQR 80.4-100.3; Mdn UACR 6.5 mg/g, IQR 4.1-12).
The identified DPs were (Figure 1): DP1, reflecting greater consumption of all nutrients; DP2, reflecting increased potassium and phosphorus intake and lower sodium and protein intake; and DP3, reflecting increased intake of protein and phosphorus and lower intake of potassium and sodium. The 3 DPs presented stable FLs across groups.
In Group1, DP1 was negatively associated with eGFR (larger effect size in males) both as linear score and in the 3rd tertile (Figure 2); DP2 was positively associated with eGFR in males and with UACR both at low and high levels of the score, with heterogeneous effects between males and females; DP3 was positively associated with UACR overall in the 3rd tertile. In Group 2, we observed protective effects of diet on eGFR, especially at lower DP1 and DP3 levels, and at high levels of DP2 (Figure 2). Similar to Group1, also in Group2 DP2 was positively associated with UACR both at low and high level of the score, but not in males. In its 3rd tertile, DP3 was negatively associated with UACR in females.
Conclusion:
Using RRR proved to be a valid approach to integrate a priori knowledge about nutrients in the estimation of kidney function-oriented DPs. Our results showed heterogeneous effects of the DPs across kidney outcomes, possibly reflecting specificity to kidney function or damage. In individuals affected by any kidney disease, hypertension or diabetes, the effects of DPs on eGFR reflected possible benefits of specific diets, suggesting that disease-specific dietary interventions can be a fundamental and effective approach for disease control.
Figure:
Figure 1: DPs in Group1 (blue) and Group2 (orange), described by Factor Loadings (FLs) of the most characterizing food groups.
Figure 2: β-coefficients and 95% Confidence Interval of linear models between each DP and the kidney outcomes. Statistically significant associations are highlighted in green.