The present cross-sectional study was carried out between April 2023 and October 2023 using PD patients from Shanghai Changzheng Hospital in China. Written informed consent was obtained from all participants before enrollment. The study protocol was approved by the Ethics Committee of Shanghai Changzheng Hospital, Ethics Approval No. 2023SL040. All PD patients were diagnosed with CKD by nephrologists and according to the criteria established by the expert group of the Shanghai Nephrology Clinical Quality Control Center, which included clinical characteristics, endoscopic, histological, and radiological examination et al. All PD patients started with intermittent peritoneal dialysis and gradually transitioned to continuous ambulatory peritoneal dialysis treatment after peritoneal dialysis placement.

The inclusion criteria were the following: (I) aged 18 to 80 years; (II) receiving maintenance peritoneal dialysis for more than 3 months; (III) regular follow-up and completion of peritoneal dialysis adequacy assessment. The exclusion criteria were the following: (I) those who were unwilling to participate and cooperate with the study; (II) those with serious data (demographic or laboratory biochemical) deficiencies; (III) those who were receiving weekly maintenance hemodialysis treatment in conjunction with peritoneal dialysis; (IV) those who had renal disease accompanied by malignant tumors, acute heart failure, cirrhosis, or immune system disease; (IV) those with renal disease accompanied with malignant tumors, acute heart failure, cirrhosis and immune system disorders; (V) history of surgery within the last 3 months; (VI) unconsciousness, communication disorders, and cognitive impairment. In addition, those on special diets (e.g., vegan, atkins) were also excluded from this study. Twenty patients were excluded due to missing values on the Food Frequency Questionnaire or other variables, resulting in a final sample of 147 patients. All researchers involved in this study received professional training.

The following necessary information was collected through the interview and medical records: (I) sociodemographic data regarding age (years), gender (male/female), education (less than primary school/middle school/high school or equivalent/college or above), smoking status (current/former/never) and drinking status (yes/no); (II) body mass index was calculated as weight (kilograms) divided by height (m) squared and divided (BMI; kg/m²); (III) hypertension was defined as having been diagnosed by a physician or taking hypertension medication (yes/no), diabetes were defined as self-reported doctor diagnosis of diabetes, HbA1c ≥ 6.5%, FPG ≥7.0 mmol/L, postprandial 2-h plasma glucose ≥11.1 mmol/L from an oral glucose tolerance test or use of insulin or oral hypoglycemic medication (yes/no); (IV) serum albumin, serum creatinine, serum urea, uric acid, total cholesterol (TC), high-density cholesterol (HDL-C), potassium, phosphate, calcium and C-reactive protein (CRP). The estimated glomerular filtration rate (eGFR) scores were calculated using the Chronic Kidney Disease Epidemiology Collaboration algorithm (26).

Diet was assessed using the Food Frequency Questionnaire (FFQ). The questionnaire used in the present study included 27 groups of foods and beverages, which covered the majority of foods commonly consumed in China. Three aspects of each item were listed in the questionnaire, including whether the item was consumed, the usual frequency of consumption (number of times per day/week/month/year), and the estimated amount of food eaten each time, expressed using the local unit liang for weight (1 liang = 50 g) or cup for volume (1 cup = 250 mL). The translated version of the FFQ will be submitted as attachments Besides, nutrients and total energy intake were calculated by multiplying the usual frequency and portion size of each food item by the nutrient content using the 2018 6th Edition Chinese Food Composition (Standard Edition). In the calculation, Wincome’s hospital nutrition diagnosis and treatment system HNTS (V2.3.2.104) was used, and the 6th edition of “Chinese Food Ingredients” (Standard Edition) in 2018 was entered into the HNTS database. Then, the FFQ was added as a template to HNTS. When entering each person’s dietary data, entering whether they consume the item, the frequency of consumption, and the estimated amount of food consumed each time in the template can directly calculate the nutrients and total energy through HNTS.

The 26 foods and dietary components involved in DII were included in this study and are listed as follows: energy, protein, fat, carbohydrate, fiber, cholesterol, vitamins A, B1, B2, B3, B6, B12, C, D, E, iron, selenium, zinc, saturated fatty acids, monounsaturated fatty acids (MUFA), polyunsaturated fatty acids (PUFA), n-3 fatty acids, n-6 fatty acids, folate, and β-carotene. Dietary nutrient contents of each food were entered based on the data in the 2018 6th Edition Chinese Food Composition (Standard Edition).

According to the given DII scoring table and following the calculation method and logic of DII, use a self-designed calculation software to calculate the DII score of the research subjects. The specific calculation method for DII is as follows: for each participant in the study, first calculate the Z-score of DII for a certain dietary component. The Z-score is calculated by subtracting the global daily average intake from the average amount of dietary components and nutrients consumed by each participant during a day (midnight to midnight), and dividing this value by its standard deviation. Next, it was converted to a percentile score. To center the distribution with values, values were doubling and subtracting “1.” Finally, values were multiplied by “overall inflammatory effect score” and then summed to obtain the overall DII score (19). To adjust DII for energy, all dietary parameters were converted to reflect food intake per 1,000 calories (27). Ultimately, E-DII score was calculated using 25 food parameters and energy as a denominator.

The Malnutrition Inflammation Score (MIS) was used to assess the nutritional status of the patients. The MIS consisted of 10 evaluation parameters in 4 parts, covering a range of physiological indexes and clinical manifestations of the patients, with the score of each parameter ranging from 0 (good condition) to 3 (severely poor condition), and the total score ranging from 0 (in normal condition) to 30 (in severely malnourished condition) (28). This study was designed with reference to existing research findings and accordingly set the cut-off value of poor nutritional status of patients. Specifically, when a patient’s MIS scale score was >7, the patient was judged to have poor nutritional status and was categorized into the poor nutritional status group, and when the score was ≤7, the patient was categorized into the good nutritional status group (28, 29).

SPSS Statistics 25.0 (IBM Corporation, New York, United States) and R version 4.3.0 (R Foundation for Statistical Computing, Vienna, Austria) were used for all statistical analyses. The entire statistical analysis process of this study, was conducted under the guidance and supervision of a professional statistician (HW, Department of Nutrition, Shanghai Changzheng Hospital, Naval Medical University, Shanghai 200003, China/attending doctor). HW was involved from the project initiation and experimental design stages, providing crucial support to ensure the rationality of the analysis plan. Continuous variables were expressed as mean ± standard deviation (SD), and categorical variables were described using frequency and percentage. ANOVA and chi-square test were used to identify the significant differences in variables data by the tertiles. Least absolute shrinkage and selection operator (LASSO) regression analysis was used to screen the variables by 10-fold cross validation to select the nutrients that contribute the most to EDII. Univariate and multivariate logistic regression was used to assess the association between the screened nutrients and malnutrition, in which the nutrient intake was classified into three groups from T1′ to T3′ and T1′ was used as the reference group. Univariate and multivariable logistic regression models were established to assess the association between E-DII and malnutrition. Participants were classified into tertiles based on E-DII, tertile 1 (T1) to tertile 3 (T3), T1 was used as the reference group.

Of the 147 PD participants, 76 (51.7%) were female, and the mean age was 50.9 ± 12.7 years. The mean E-DII score was −0.367, ranging from −2.958 to 2.379. The mean MIS was 8.0 ± 3.7, among males, the mean MIS was 6.82 ± 2.70, while among females, the mean MIS was 9.03 ± 4.16; there were 71 patients with MIS >7 who suffered malnutrition, 26 males and 45 females. The mean duration of PD was 47.9 months. The average hemoglobin level was 106.3 g/L, and 121 (82.3%) patients were diagnosed with anemia. In terms of oedema legs, 36 patients (24.5%) had no edema, 39 patients (26.5%) had mild edema, and 72 patients (49%) had edema. The average circumference of the arm muscles was 22.7 mm, and the average skin fold thickness was 14.8 mm.

Table 1 shows the baseline characteristics of PD patients stratified by E-DII tertile. Compared with the patients in the first tertile (T1), patients in the third tertile (T3) who consumed a more pro-inflammatory diet have a higher prevalence of malnutrition and a lower blood potassium (p < 0.05).

Table 2 shows the average daily intake of each food category in PD patients by E-DII tertile. We observed that PD patients in the first tertile had a higher consumption of vegetables (T3 vs. T1 = 212.4 ± 146.8 g/d vs. 452.7 ± 110.0 g/d, p < 0.001) and red meat (T3 vs. T1 = 52.8 ± 33.7 g/d vs. 76.2 ± 44.4 g/d, p = 0.012), compared with the patients in the third tertile.

Table 3 shows the dietary nutrients intake of PD patients by E-DII tertile. We observed that PD patients in the first tertile had a high intake of anti-inflammatory nutrients, such as total fiber (T3 vs. T1: 6.4 ± 2.6 g/d vs. 11.2 ± 3.3 g/d, p < 0.001), n-3 fatty acids (T3 vs. T1: 0.3 ± 0.2 g/d vs. 0.4 ± 0.2 g/d, p = 0.023), folate (T3 vs. T1: 264.0 ± 103.0 μg/d vs. 146.2 ± 71.37 μg/d, p < 0.001), β-carotene (T3 vs. T1: 2487.0 ± 1146.1 μg/d vs. 7094.2 ± 1920.8 μg/d, p < 0.001), and vitamins such as vitamin A (T3 vs. T1: 423.7 ± 199.2 RE/d vs. 784.1 ± 210.4 RE/d, p < 0.001), vitamin C (T3 vs. T1: 109.4 ± 51.0 mg/d vs. 269.1 ± 76.1 mg/d, p < 0.001), vitamin B1 (T3 vs. T1: 0.5 ± 0.2 mg/d vs. 0.6 ± 0.3 mg/d, p = 0.001), vitamin B2 (T3 vs. T1: 0.7 ± 0.3 mg/d vs. 0.9 ± 0.3 mg/d, p = 0.003), vitamin B3 (T3 vs. T1: 13.3 ± 4.6 mg/d vs. 16.7 ± 5.4 mg/d, p = 0.004) and vitamin B6 (T3 vs. T1: 0.7 ± 0.3 mg/d vs. 0.9 ± 0.3 mg/d, p < 0.001), and minerals such as Mg (T3 vs. T1: 217.2 ± 61.3 mg/d vs. 306.0 ± 73.4 mg/d, p < 0.001), Zn (T3 vs. T1: 8.9 ± 2.8 mg/d vs. 10.9 ± 3.1 mg/d, p = 0.003) and Fe (T3 vs. T1: 15.8 ± 4.8 mg/d vs. 21.5 ± 6.0 mg/d, p < 0.001), compared with the patients in the third tertile.

In the LASSO regression model, a total of 25 nutrients (adjusted by energy) were included, and the key nutrients were screened by 10-fold cross-validation. Figure 1A shows that the model error is the smallest when log(λ) = −3.628016, and the model screens out 10 variables. Figure 1B, each line represents a variable, the vertical axis is the coefficient of the parameter, and the horizontal axis is log(λ). The number of variables screened out varies with different values of log (λ), and the variables that have been screened out by LASSO have non-zero coefficients. In Figure 1B at λ = 0.026569, all 10 variables are retained in the model (with non-zero coefficients), and the coefficients of the rest of the variables are approaching zero with the increase of the penalty term. When λ is increased to 0.107259, the model is most appropriate for retaining only 1 variable. The LASSO regression model identified the key dietary nutrient associated with malnutrition was total fiber.

The restricted cubic spline analysis revealed significant linear associations between both total fiber, E-DII intake and malnutrition risk (Figure 2).

The association between E-DII score, total fiber intake and malnutrition was further evaluated by univariate and multivariate logistic regression analysis (Table 4). We used total fiber as a continuous variable and divided the population into three groups using quartiles. The fiber intake of the first group T1′ was 2.4 g to 6.8 g, with an average of 5.4 ± 1.3 g. The fiber intake of the second group T2′ was 6.9 g to 9 g, with an average of 8.0 ± 0.6 g. The fiber intake of the third group T3′ was 9.1 g to 23.1 g, with an average of 12.3 ± 2.7 g. After adjustment for age, gender, smoking status, drinking status and BMI (Model 2), total fiber (as continues variable) was observed as a protective factor against malnutrition (OR = 0.84, 95% CI: 0.73–0.98, p = 0.025), and a positive association was observed between E-DII (as continues variable) and malnutrition (OR = 1.53, 95% CI: 1.07–2.19, p = 0.021). After further adjustment for hypertension, diabetes, dialysis month, eGFR, TC and HDL-C (Model 3), total fiber (as continues variables) remained a protective factor against malnutrition (OR = 0.81, 95% CI: 0.69–0.95, p = 0.010), specifically, the third tertile (T3) was associated with a lower risk of malnutrition in PD patients compared with the first tertile (T1) (OR_tertile 3 vs. 1 = 0.29, 95% CI: 0.10–0.80, p = 0.018); and a positive association was still observed between E-DII and malnutrition, with each unit increase in E-DII score was linked with 1.62 times increased odds of malnutrition (OR = 1.62, 95% CI: 1.11–2.37, p = 0.013), specifically, the third tertile (T3) was associated with a higher risk of malnutrition in PD patients compared with the first tertile (T1) (OR_tertile 3 vs. 1 = 3.64, 95% CI: 1.25–10.64, p = 0.018).