The NHANES is a program conducted to evaluate people's health and nutritional status in the United States. It combines interviews and physical examinations to collect demographics, dietary, medical examination, laboratory, and questionnaire data (18). The National Center for Health Statistics Research Ethics Review Board approved data collection for NHANES, and written informed consent was obtained from participants.

In this study, we used data from 8 NHANES cycles: 1999–2000, 2001–2002, 2003–2004, 2005–2006, 2011–2012, 2013–2014, 2015–2016, and 2017–2018 (18). As shown in Figure 1, we collected 2,569 CKD participants with complete data for DII and sarcopenia, excluding individuals under 20 years old and with an intentional weight-changing history.

We used the Modification of Diet in Renal Disease Study Equation to estimate glomerular filtration rate (eGFR) (19). CKD was defined as eGFR <60 ml/min per 1.73 m² or urine albumin-to-creatinine ratio (UACR) ≥30 mg/g (20).

Dietary inflammatory index was the major exposure variable. The DII score estimates the impact of diet on inflammation based on the pro- and anti-inflammatory properties of 45 different food components, which were developed according to a systematic review of ~2,000 published research articles (11). Higher positive DII scores mean more pro-inflammatory diets, whereas more negative values correspond to more anti-inflammatory (11). We calculate DII based on dietary intake information by 24 h dietary recalls (24 HR) in this study. Moreover, 27 or 28 food components in 24 HR were used to calculate the DII scores: energy, carbohydrate; protein; total fat; dietary fiber; cholesterol; saturated, monounsaturated, and polyunsaturated fatty acids; ω-3 and ω-6 polyunsaturated fatty acids; vitamins A, B1, B2, B3 (niacin), B6, B12, C, D, and E; folic acid; alcohol; beta-carotene; caffeine; iron; magnesium; zinc; and selenium. Previous studies confirmed that DII scores based on only 27 or 28 food parameters did not influence the predictive ability (13, 21). Since we calculate DII scores according to participants' dietary recall information, we performed a consistency test to evaluate the accuracy of the measurement. As shown in Supplementary Table S1, there was no significant difference between DII scores from the first 24 HR and second 24HR (p = 0.2363).

The outcome of interest was sarcopenia, which was assessed by the sum of four limbs' muscle mass (appendicular lean mass, ALM) (22). NHANES used dual-energy X-ray absorptiometry (DEXA) to measure ALM. Participants with a height > 192.5 cm, weight > 136.4 kg, and pregnant individuals were excluded since these participants could not perform the DEXA test (18). We defined sarcopenia by BMI-adjusted ALM (ALM_BMI): men were judged as sarcopenia if ALM_BMI <0.789, and women < 0.512 (22).

Socio-demographic factors included age, gender, race, and income. We categorized race into 4 groups: Mexican American, non-Hispanic white, non-Hispanic black, and others. Income was evaluated by the poverty income ratio (PIR). PIR < 1 was considered as poor, 1–3 near-poor, and ≥3 as not poor (23). Behavioral variables were physical activity, cigarette smoking, and alcohol drinking. Physical activity was divided into inactive and active groups according to average physical activity time. Participants who met the American Physical Activity Council's chronic health conditions recommendation (75 min/week vigorous or 150 min/week moderate activity) were active (24). Individuals whose dietary alcohol intake was more than 0 g were alcohol consumers. Participants who self-reported smoking more than 100 cigarettes in life were identified as smokers.

Since protein-energy wasting and comorbidities could influence muscle atrophy, we considered low energy intake (energy intake <25 kal/kg/day), low protein intake (<0.6 g/kg/day), and hypoalbuminemia (<38 g/L) (25, 26). In total, 8 comorbidities were included in the analysis: diabetes, hypertension, overweight, central obesity, dyslipidemia, cardiovascular diseases, arthritis, and cancer (Supplementary Table S2) (14, 27). Biomarkers of inflammation included C-reactive protein (CRP), white blood cell (WBC) count, and neutrophil-lymphocyte ratio (NLR).

We collected data from 8 NHANES cycles, corresponding to a 20-year cross-section ranging from 1999 to 2018. Based on potential diet changes over 20 years, we also considered NHANES strata. Each NHANES strata contains two cycles: NHANES strata 1 (NHANES 1999–2002), NHANES strata 2 (NHANES 2003–2006), NHANES strata 3 (NHANES 2010–2014), and NHANES strata 4 (NHANES 2015–2018).

The characteristics of participants overall and among different DII groups (tertiles) are summarized in Table 1. Continuous variables with a normal distribution were presented as mean ± standard deviation (SD). Non-normal continuous variables were reported as median with an interquartile range. Categorical variables were presented as frequency and percentage. Comparisons among different groups were performed by one-way ANOVA tests for normally distributed variables, independent-samples Kruskal–Wallis tests for non-normal continuous variables, and chi-square tests for categorical variables.

To evaluate the association between dietary inflammatory potential and sarcopenia, we first conducted smooth curve fitting (penalized spline method) to address the non-linearity of DII and sarcopenia. Then, a generalized linear mixed model (GLMM) was applied to determine the independent association between DII and sarcopenia. Model 1 had no covariate adjusted; model 2 had age, gender, and race adjusted; model 3 had adjusted for age, gender, race, income, physical activity, smoking, alcohol drinking, diabetes, hypertension, overweight, central obesity, dyslipidemia, cancer, arthritis, heart disease, eGFR, ACR, hypoalbuminemia, low energy intake, low protein intake, CRP, WBC, NLR, and NHANES strata. In addition, we performed subgroup analyses. All continuous covariables were converted into tertiles categorical variables and the interaction effect was assessed via the likelihood ratio test. A sensitivity analysis was conducted by converting DII into a tertile categorical variable.

All analyses were conducted with Empower (R) (28) and RStudio (29).

As described in Figure 1, 2,569 participants from 8 NHANES cycles were included in this study. Table 1 summarized the characteristics of participants overall and among DII tertiles groups. On average, participants were 55.6 years old. About 45.11% were men. The overall prevalence of sarcopenia among patients with CKD was 19.11%. The overall DII score ranged from −4.31 to 4.86. We equally divided the subjects into 3 groups based on their DII score: the first tertile group (T1, n = 856; DII = −4.31 to 0.68), the second tertile group (T2, n = 856; DII = 0.68–2.41), and the third one (T3, n = 857; DII = 2.41–4.86). Subjects in the T1 group consumed a more anti-inflammatory diet, whereas participants in the T3 group intake more pro-inflammatory diet. Compared with the T1 group, subjects in the T3 group were more likely to be female, Non-Hispanic black, poor, inactive, and non-alcohol consumers. Additionally, they had a higher incidence of hypertension, arthritis, low energy intake, low protein intake, and hypoalbuminemia than those in the T1 group (p < 0.05). Besides, the level of CRP in the T3 group was significantly higher than in the T1 group, which confirmed that consuming more pro-inflammatory food is positively associated with systematic inflammation.

The smooth curve fitting results showed the DII score is near-linear positively associated with sarcopenia after adjusting for cofounders (Figure 2A). Then, the generalized linear mixed model analysis further confirmed that sarcopenia is positively related with DII scores (Table 2). The odds ratio (OR) is 1.07 (95% CI, 1.02–1.13), 1.15 (95% CI, 1.08–1.22), and 1.17 (95% CI, 1.06–1.29), respectively, in the model 1 (no adjustment), model 2 (adjusted for age, gender, and race), and model 3 (adjusted for all cofounders). To ensure the validity of the results, we also performed a sensitivity analysis by converting DII to a categorical variable (tertiles). The new adjusted OR values were 1.27 (95% CI, 0.88, 1.84) for T2, and 1.98 (95% CI, 1.32, 2.98) for T3 (a reference to T1), and the p for trend was 0.001.

In addition, we also assessed the relationship between DII and ALM. The smooth curve fitting figure displayed that the DII score has a linear negative relationship with ALM (Figure 2B). The GLMM analysis also showed that DII is independently related with ALM (Table 3). The OR value was −0.50 (95% CI, −0.59, −0.40) after adjusting for all cofounders.

Subgroup analyses results are summarized in Table 4. Relatively stronger associations between DII and sarcopenia were observed among women, older adults, active subjects, smokers, non-alcohol drinkers, and patients with hypoalbuminemia, low energy intake, low protein intake, diabetes, overweight, central obesity, non-dyslipidemia, non-cancer, and non-heart disease. Patients with CKD in G3-5 and A3 stages were also more likely to be sarcopenia if they consumed more pro-inflammatory diets. However, only low energy intake, low protein intake, and dyslipidemia have significant interactions with DII (p for interaction < 0.05). Besides, the association between DII and sarcopenia also remains consistent and stronger in subjects with high levels of CRP and NLR. Significant interactions between CRP and DII were also observed, with p for interaction < 0.05.

To figure out the influence of NHANES strata on the association effect between DII and sarcopenia, we also performed a subgroup analysis for NHANES strata. As shown in Supplementary Table S3, significant associations were only observed in NHANES strata 2003–2006 (DII continuous) and NHANES strata1999–2002 (DII tertile). However, p for heterogeneity was over 0.05, meaning there was no significant difference between 4 NHANES strata for relationships between DII and sarcopenia.