A combination of subject terms and free terms was used to fully search PubMed, Embase, Web of Science, Scopus, China National Knowledge Infrastructure, Wanfang Database, VIP Database, and China Biomedical Literature Database for dietary inflammation index and frailty correlation literature, with a timeframe from the construction of the database to January 2025. Use the following search terms to search: (“frailty” OR “frailty syndrome” OR “frail^*” OR “frailty” OR “debilit^*” OR “weakness^*”) AND (“dietary inflammatory index” OR “inflammatory potential of diet” OR “anti inflammatory diet” OR “pro inflammatory diet”).

Inclusion criteria covered: (i) research assessing the relationship of Dietary Inflammatory Index scores with frailty susceptibility; (ii)middle-aged and older adults aged ≥45 years; and (iii) data on the correlation between DII and risk of frailty were reported in the literature and effect estimates were provided in the form of OR or HRs and their 95% CI; (iv) study design was either observational (case-control studies, cohort studies, and cross-sectional studies) or randomized controlled studies.

Exclusion criteria included: (i) literature for which full text was not available or duplicate publications; (ii) literature for which relevant outcome indicators were not reported or relevant data (OR or HR) could not be extracted; (iii)original studies that were not published in English or Chinese; (iv) reviews, dissertations, systematic reviews, conference abstracts, book chapters, and patents; (v)Lower quality literature.

Two authors independently screened studies based on inclusion/exclusion criteria, consulting a third researcher when necessary to resolve disagreements. Data collection covered three domains: study identification details (author, year, country), methodological elements (study design, population, follow-up, assessment tools, covariates), and analytical outcomes (adjusted risk estimates with confidence intervals)

Methodological quality was systematically evaluated using discipline-specific tools: cohort and case-control studies were appraised with the Newcastle-Ottawa Scale (NOS), where studies achieving scores above 5 out of 9 were deemed methodologically sound (27). Cross-sectional studies were evaluated according to the Agency for Healthcare Research and Quality (AHRQ) criteria, employing an 11-item checklist with binary scoring (1 = present/0 = absent or unclear). Based on cumulative scores, cross-sectional studies were categorized into three quality tiers: low: 0–3, moderate: 4–7 and high: 8–11 quality evidence (28).

RevMan 5.3 and Stata 15.0 software were used for statistical processing. Adjusted dominance ratios (ORs) were combined using an inverse variance weighting approach (treating risk ratios (HRs) as OR equivalents), and effect sizes were combined using the model with the highest number of covariates provided by the authors. Study heterogeneity was examined using the χ² statistic, and when there was no significant heterogeneity (I² < 50%, P > 0.10), a fixed-effects model was used; otherwise (I² ≥ 50% or P ≤ 0.10), random-effects models were used to combine effect sizes, and subgroup analyses and meta-regressions were performed to explore sources of heterogeneity based on study geographic region, type of study, sample size, and DII assessment tool and frailty assessment tool. Sensitivity analyses were conducted through sequential exclusion of individual studies to evaluate result stability. Publication bias was assessed visually via funnel plots and quantitatively through Egger's and Begg's tests, with trim-and-fill adjustment implemented when required.

The detailed process of literature search and study selection is shown in the flow chart (Figure 1). A total of 352 papers were obtained through database search, and 229 papers were obtained after removing duplicates. After preliminary screening of the papers by checking and reading the titles and abstracts, excluding 157 studies with irrelevant topics, 13 reviews and 3 conference abstracts, and eliminating 26 papers with different endpoints or exposures, 8 papers with incompatible study subjects and 6 papers with non-DII records, as well as 1 low-quality paper through full-text reading, 15 papers (23–25, 29–40) were finally included, of which 12 were in English (23–25, 32–40), 3 were in Chinese (29–31).

Fifteen (23–25, 29–40) studies were published from 2017 to 2025 involving 42,130 subjects, including nine (25, 29, 30, 32–34, 36, 37, 40) cross-sectional studies, five (24, 25, 35, 38, 39) cohort studies, and one (31)case-control study. DII calculations were all performed using Shivappa's updated scoring system (20), with fourteen (23–25, 29–31, 33–40) reporting outcome indicators in the form of categorical DII, and 6 (23, 25, 31, 36, 37, 40) reporting indicators in continuous DII. Three (25, 29, 36) studies in the literature reported on the risk of developing a pre- frailty stage. Fifteen studies all controlled for the effect of confounding factors on the outcome by means of multifactorial analyses. Quality assessment revealed scores from 6 to 9 across the studies, reflecting a consistently high standard of methodological rigor. Detailed information is shown in the Table 1.

Six included studies (23, 25, 32, 36, 37, 40) examined the correlation between continuously measured DII scores and frailty incidence. Low heterogeneity (I² = 0%, P = 0.86), supporting the use of a fixed-effects model. Pooled estimates showed a statistically significant positive association between higher levels of DII and increased risk of frailty, with a 7% increase in risk of frailty for each point increase in DII [OR = 1.07, 95% CI (1.05, 1.10), P < 0.001, see Figure 2].

Three (25, 29, 36) studies reported the relationship between the categorical variable form DII and the occurrence of pre-frailty conditions, with low heterogeneity across studies (I² = 9%, P = 0.33), so fixed-effects meta-analysis revealed a 54% higher risk of pre-frailty conditions in the highest category of DII individuals compared to the lowest category of DII individuals [OR = 1.54, 95%CI (1.34, 1.76), P < 0.001, see Figure 3].

Fourteen (23–25, 29–31, 33–40) papers reported on the association between categorical variable form DII and the occurrence of frailty, with a high degree of heterogeneity between studies (I² = 67%, P < 0.001), meta-analysis with a random-effects model revealed a 47% higher frailty risk in the highest category of DII compared to the lowest category of DII individuals [OR = 1.47, 95%CI (1.28, 1.69), P < 0.001, see Figure 4].

The funnel plot analysis evaluating the association between categorical DII and frailty incidence demonstrated visually detectable asymmetry upon systematic examination, continuing to show statistical significance using the Begg test for publication bias (P = 0.016), and Egger's test for publication bias (P = 0.007), which suggests the presence of publication bias. Therefore, correction was performed using the trim-and-fill method, and four similar studies were added after three iterations, and the effect sizes were not reversed after the merger [OR = 1.33, 95% CI (1.13, 1.58), P = 0.001, see Figure 5].

Sensitivity analyses using the drop-by-drop method as shown in the Figure 6 showed no significant changes in the combined effect sizes, which demonstrates the robustness and reliability of the study outcomes.

High-level DII increases the risk of pre-frailty in middle-aged and older adults, which is consistent with the previous meta-analysis results of Donya et al. on the correlation between adherence to the Mediterranean diet and pre-frailty (41). However, in a European community of residents over 70 years old adhering to the Mediterranean diet, no beneficial results of the dietary pattern were found, and the levels of inflammatory markers in the body did not change after a 3-year follow-up (42). The differences in the results of the above studies may be related to differences in the geographic and cultural backgrounds, dietary habits and study designs of the included populations, in addition to differences in the studies' adjustments for confounders that may have contributed to the differences in results. Given the relative paucity of studies exploring the effects of DII on preexisting frailty, more extensive future investigations in diverse populations are needed to fully compare the correlation between DII and the risk of developing preexisting frailty.

Dietary patterns promoting inflammation (elevated DII) may predispose older individuals to frailty development, in harmony with the results of previous studies (15, 43, 44). American scholars Jayanama et al. (15) explored the correlation between diet and frailty using several different dietary quality scores, such as the Nutritional Index, Healthy Eating Index, and Mediterranean Diet Score, and found that higher scores across all dietary evaluation tools consistently correlated with greater frailty likelihood; in addition, Italian scholars Dominguez et al. (43) included 11 studies and Guan Junyi et al. (44) included 9 studies in their meta-analysis and found that adherence to the Mediterranean dietary pattern (anti-inflammatory diet) was correlated with a decreased risk of frailty.

The underlying etiology of frailty is the chronic inflammatory response that occurs in the body, and food intake provides the body with the energy and substrates needed to cope with this response, so dietary modification based on dietary inflammation to intervene in the inflammatory response has become an important breakthrough point in the effective prevention and treatment of frailty (45). Diets with high DII levels, i.e., pro-inflammatory diets, include red meat, refined grains, fried and barbecued foods, and foods rich in unsaturated fatty acid (46), which contain certain components that can stimulate the body to cause elevated levels of inflammatory factors affecting protein synthesis and metabolism, resulting in a decrease in muscle mass and quality and thus leading to the development of frailty (14, 47). A low DII level diet, or anti-inflammatory diet, is centered on increasing the omega-3 fatty acid (deep-sea fish, flaxseed), dietary fiber (whole grains, vegetables), polyphenol (berries, green tea), and monounsaturated fatty acid (olive oil, nuts) components, while limiting processed meats, refined sugars, trans fats (e.g., fried foods), and high GI carbohydrate intake (48), to increase the body's antioxidant property components, and maintain the stabilization of the intestinal flora, synergistically reversing or delaying the severity of the weakness and its progression (45, 49, 50). Therefore, in the future, anti-inflammatory diets can be promoted for middle-aged and older community members, emphasizing the “diet-inflammation-frailty” chain, popularizing knowledge of dietary inflammation, and conducting regular dietary inflammation screenings to identify people at high risk of frailty for older intervention.

Subgroup analyses were conducted to investigate potential sources of heterogeneity in the DII-frailty association. Subgroup analyses based on sample size showed that heterogeneity among subgroups was within acceptable limits, suggesting that differences in sample size may be the main source of heterogeneity. In addition, other subgroup analyses based on different study geographies, study types, and dietary assessment tools, each with varying degrees of reduced between-study heterogeneity, revealed similar results regarding the relationship between pro-inflammatory diets and the occurrence of frailty, with high levels of DII increasing the risk of frailty in middle-aged and older adults.

Notably the results of subgroup analyses by geography suggested a stronger correlation between DII and risk of frailty in Asian populations compared to the Americas, analyzing the reasons for this may be that the Americas population is chronically adapted to a high glycemic, highly processed diet, which is partially tolerated through intestinal flora adjustments or down-regulation of inflammatory pathways, and that Asian populations are more sensitive to added inflammatory stimuli, while the Asian populations have a relatively low muscle mass, resulting in a unit of The damage to muscle metabolism from inflammatory load may be more pronounced in Asian populations.

When stratified by assessment methodology, studies utilizing the Fried frailty criteria demonstrated stronger, statistically significant correlations with DII scores compared to those employing the Frailty Index (FI), where associations were attenuated and non-significant. The reason for this analysis may be the inconsistency in the criteria for judging frailty, with the Fried frailty phenotype focusing on physical performance such as body weight, muscle strength, and gait speed (51), whereas the FI rating of frailty emphasizes the number of cumulative physical deficits (52), leading to differences in the prevalence and classification of frailty causing differences in the results of the study. Future studies should work on the standardization of frailty assessment tools, including the clear definition of core indicators, the unification of thresholds and the optimization of operational procedures, in order to improve the accuracy and comparability of screening tools, while reducing selection bias caused by different measurement methods.

This investigation possesses several notable strengths. Primarily, it represents the most current systematic evaluation and quantitative synthesis examining DII-frailty associations in aging populations, thereby extending previous meta-analytic findings. Our results substantiate emerging evidence that anti-inflammatory dietary patterns may mitigate frailty risk among middle-aged and older adults. Notably, we included studies with inconsistent views for statistical analysis, including more frailty cases and participants than in previous meta-analyses (53), while the inclusion of prospective cohort studies vs. the type of case-control studies increased the reliability of the findings. Secondly we conducted subgroup analyses based on various confounders to elucidate the sources of differences between the included studies. In addition, publication bias was corrected for the presence of studies using the cut-and-patch method, and the results further confirmed the robustness and reliability of the findings.

Several limitations should be acknowledged in this investigation. Primarily, the retained research articles were observational, and inferences about causality between the two were not as strong as in interventional studies; second, the calculation of DII relied on self-administered FFQs and dietary recalls, and thus the results may be subject to recall bias; and, although the included studies were all corrected for confounders, they varied with respect to ethnicity, BMI, physical activity, energy intake, and other factors were not identical, and thus bias may have arisen.