Combined residual inflammation and anemia in relation to kidney function among adults with diabetes: a CHARLS-based cross-sectional and longitudinal study

Abstract

Background

The combined phenotype of residual inflammation and anemia on kidney function in diabetes remains unclear. Our study examined their joint association with impaired kidney function and kidney function decline in a nationally representative Chinese sample.

Methods

Our study encompassed 2,363 diabetes patients from the China Health and Retirement Longitudinal Study (CHARLS) for cross-sectional analysis and a sub-cohort of 547 patients for longitudinal analysis (2011–2015). Exposures comprised residual inflammation (hs-CRP ≥ 2.0 mg/L) and WHO-defined anemia. Outcomes were prevalent impaired kidney function in the cross-sectional analysis and rapid kidney function decline (RKFD) in the longitudinal analysis. Logistic and linear regression, restricted cubic splines, and subgroup and sensitivity analyses were applied.

Results

Cross-sectionally, isolated residual inflammation or isolated anemia were each associated with impaired kidney function, while the combined phenotype showed the strongest association (OR = 3.61, 95% CI: 2.16–5.99; P < 0.001). Longitudinally, this combined phenotype conferred the highest risk for RKFD (OR = 6.42, 95% CI: 2.18–18.34; P < 0.001) and was the only group associated with a steeper eGFR decline (β=-8.68 mL/min/1.73 m², 95% CI: -15.72 to -1.64; P = 0.011). Sensitivity analyses and RCS models supported the robustness of these associations.

Conclusion

Coexisting residual inflammation and anemia are associated with a higher prevalence of kidney function impairment and a more rapid kidney function decline among diabetic patients. Addressing both conditions concurrently as an integrated clinical phenotype may better mitigate kidney function decline.

Clinical trial number

Not applicable.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s12902-026-02577-4.

Introduction

Kidney function is a cornerstone of metabolic and cardiovascular homeostasis, and preserving it has become a central therapeutic goal in long-term diabetes patient management. This population is particularly vulnerable to progressive renal impairment, which typically manifests as a declining glomerular filtration rate accompanied by albuminuria and follows a substantially faster trajectory than that observed in the general population [1–3]. Globally, the accelerating aging and the increasing burden of chronic diseases like diabetes impose substantial economic and health burdens on individuals, families, and societies [4, 5]. Diabetes is now the leading global cause of kidney function loss, with roughly 40% of patients eventually developing diabetic kidney disease [6, 7]. China leads the world in the number of diabetes cases, with age-standardized diabetes incidence and prevalence rates showing an overall increasing trend. Concurrently, the proportion of diabetic patients developing severe complications like chronic kidney disease (CKD) is also rising [8, 9]. The pathogenesis of diabetes-related kidney injury involves multiple pathological processes, including insulin resistance, microvascular damage, inflammatory responses, oxidative stress, and anemia, all of which are particularly exacerbated in diabetic patients and contribute to progressive kidney function impairment and decline [10, 11]. Critically, diabetic patients exhibit a higher prevalence of anemia than non-diabetic populations, but only a minority receive targeted anemia management, thereby potentiating the development of impaired kidney function in this population [12–14].

Among the multiple pathways implicated in diabetic kidney injury, residual low-grade inflammation and anemia are particularly common and potentially modifiable. Nevertheless, their joint relevance to kidney function impairment and decline in type 2 diabetes mellitus (T2DM) remains unclear. Chronic low-grade inflammation, a common feature of diabetes, is linked to cumulative metabolic stressors, including persistent hyperglycemia, lipotoxicity, and adipokine dysregulation [15, 16]. This sustained inflammatory state is often referred to as residual inflammation and is commonly operationalized by an elevated high-sensitivity C-reactive protein level (hs-CRP ≥ 2 mg/L) [17–20]. Inflammation has been associated with anemia in diabetes through proposed mechanisms involving suppressed erythropoiesis and hepcidin-mediated alterations in iron homeostasis [21, 22]. Anemia, in turn, has been reported to coexist with and potentially reinforce systemic inflammatory activation [23, 24]. Although these observations suggest that the two conditions may share overlapping biological pathways, whether they act jointly on the kidney in real-world diabetic populations has not been adequately quantified.

While the individual impacts of inflammation and anemia on kidney function have been explored, the extent to which these two conditions concurrently contribute to kidney function impairment and decline in people with diabetes remains insufficiently understood. The progressive novelty of our study lies in shifting the paradigm from single-risk factor analysis to evaluating the cumulative prognostic value of this combined phenotype. Leveraging nationally representative China Health and Retirement Longitudinal Study (CHARLS) data—drawn from a population that now carries the largest absolute diabetes burden globally—we integrated cross-sectional burden estimation with longitudinal analysis of kidney function change under a unified analytic framework. It is hypothesized that coexisting residual inflammation and anemia are associated with a higher risk of impaired kidney function and accelerated kidney function decline than either condition alone. These findings may help elucidate the interplay between inflammation and anemia in DKD and inform early risk stratification as well as integrated prevention strategies targeting both pathways.

Methods

Study design and participants

CHARLS is a longitudinal survey on the social, economic, and health status of middle-aged and older adults residing in Chinese communities. Utilizing a multistage, stratified, probability proportional to size sampling methodology, it collects high-quality data through personal interviews conducted by uniformly trained investigators [25, 26]. Analyses were restricted to participants aged > 45 years, consistent with the target population of the CHARLS, which was designed to investigate health transitions among middle-aged and older adults. This restriction also ensured a study population in which type 2 diabetes, chronic inflammation, anemia, and kidney function impairment associated with chronic metabolic conditions are more clinically relevant while reducing potential heterogeneity arising from differences in diabetes etiology and disease trajectories among younger individuals [27, 28]. Two analyses were conducted using different analytic samples (Fig. 1): a cross-sectional analysis of impaired kidney function based on Wave 3 (2015) data and a longitudinal analysis of kidney function decline using Wave 1 (2011) as baseline and Wave 3 (2015) as follow-up.

Fig. 1.

Flowchart of participant selection for the cross-sectional and longitudinal analyses of kidney function among adults with diabetes mellitus in the CHARLS database. Note: Both analytic samples were derived from the CHARLS. The cross-sectional analysis was based on the 2015 wave. Smoking, drinking, hypertension, and diabetes were defined using an ever-positive definition across the 2011 and 2015 waves after linkage by unique participant IDs, whereas all exposure and outcome variables were obtained exclusively from the 2015 wave. After excluding participants who did not meet the eligibility criteria or had missing data, 2,363 participants aged > 45 years with diabetes were included. The longitudinal analysis used the 2011 wave as baseline and the 2015 wave as follow-up. After excluding participants with baseline eGFR < 60 mL/min/1.73 m², those who did not meet the eligibility criteria, and those with missing data, 547 participants aged > 45 years with diabetes were included. eGFR, estimated glomerular filtration rate

For the cross-sectional analysis, the 2015 wave served as the primary analytic dataset (n = 21,095). To improve the completeness of covariate ascertainment, smoking, drinking, hypertension, and diabetes were defined using an ever-positive rule, whereby participants were classified as positive if the criterion was met in either the 2011 or 2015 wave after linkage across waves by participant ID. Among the 21,095 participants in the 2015 wave, 14,574 (69.1%) were successfully linked to their 2011 records. Within the final analytic sample (n = 2,363), this cross-wave approach incorporated information from the 2011 wave for 865 participants for smoking, 125 for drinking, 446 for hypertension, and 614 for diabetes. All primary exposures (high-sensitivity C-reactive protein [hs-CRP] and hemoglobin) and outcome-related variables (serum creatinine and cystatin C for estimated glomerular filtration rate [eGFR] calculation) were obtained exclusively from the 2015 wave. For the cross-sectional analysis, the following were excluded: (1) participants lacking measurements for essential data (649 without necessary demographic data; 4,703 without necessary behavioral and physical examination data; 742 without necessary health status and functioning data; 3,603 without laboratory blood data); (2) those aged ≤ 45 years (n = 652); and (3) those who did not meet the diagnostic criteria for diabetes mellitus (n = 8,383). Finally, 2,363 diabetic patients were included for analysis.

For the longitudinal analysis, 2011 data were used as baseline data (n = 17,708), and 2015 data were used as follow-up data. Exclusion criteria were: (1) participants who did not complete the 2015 follow-up examination (n = 3,134); (2) participants lacking measurements for essential data (43 without necessary demographic data; 5,667 without necessary behavioral and physical examination data; 3,898 without necessary health status and functioning data; 1,408 without laboratory blood data); (3) those aged ≤ 45 years or without diabetes mellitus (n = 2,945); (4) those with baseline eGFR < 60 mL/min/1.73 m² (n = 66). Finally, 547 diabetic patients with preserved baseline kidney function were included for analysis. The diagnosis of diabetes was based on at least one of the following: fasting blood glucose > 125 mg/dL, HbA1c ≥ 6.5%, self-reported physician diagnosis, or current glucose-lowering medication taking [29, 30]. CHARLS gained approval from the Peking University Research Ethics Committee (IRB 00001052–11015). Every participant gave informed consent. Details of the study design and data collection are available on the CHARLS website (http://charls.pku.edu.cn/en).

Exposure

Residual inflammation was defined as a serum hs-CRP level ≥ 2 mg/L and considered a primary exposure, reflecting persistent systemic low-grade inflammation associated with renal tissue injury [17, 19]. Anemia, indicated by hemoglobin < 13 g/dL in men or < 12 g/dL in women, was a co-primary exposure in accordance with WHO criteria [31]. Based on the combined status of residual inflammation and anemia, participants were categorized into four groups: (1) Without residual inflammation and anemia (n = 950); (2) With residual inflammation but without anemia (n = 982); (3) Without residual inflammation but with anemia (n = 232); (4) With both residual inflammation and anemia (n = 199). Researchers collected fasting blood samples for the measurement of hs-CRP and hemoglobin. All samples were processed by qualified personnel and subsequently transported to the China Center for Disease Control and Prevention in Beijing.

Covariates

Covariates were confirmed based on prior studies and directed acyclic graphs (Fig. S1), encompassing sociodemographic characteristics, anthropometric measures, and chronic conditions. Sociodemographic data included age, gender, education (categorized as primary school or lower vs. secondary school or higher), and marital status (currently married vs. unmarried). Lifestyle factors assessed were smoking and drinking, determined by self-reported history [32]. Anthropometric measurements were systolic blood pressure (SBP), diastolic blood pressure (DBP), as well as waist circumference (WC), while weight-adjusted waist index (WWI) was WC (cm)/weight (kg) [33, 34]. Hypertension was confirmed via SBP ≥ 140 mmHg, DBP ≥ 90 mmHg, self-reported physician diagnosis, or current antihypertensive medication taking, and was included as a covariate [35, 36]. Additional adjustment for arthritis, cardiovascular disease (CVD), CVD medications, HbA1c, lipid-lowering drug (LLD), and malignancy was performed in sensitivity analyses.

Assessment of kidney function outcomes

In this study, kidney function outcomes were assessed specifically for the cross-sectional and longitudinal analyses. The primary outcome for the cross-sectional analysis was impaired kidney function, defined by an estimated glomerular filtration rate (eGFR) below 60 mL/min/1.73 m² [37, 38]. CHARLS did not collect standardized urinary albumin or protein measurements in the relevant waves; therefore, albuminuria could not be incorporated into the present analyses. Additionally, to account for physiological, age-related declines in kidney function and to mitigate the risk of overdiagnosis in the elderly, an age-specific definition of impaired kidney function was applied as a secondary outcome in sensitivity analyses, defined as eGFR < 60 mL/min/1.73 m² for individuals aged < 65 years and eGFR < 45 mL/min/1.73 m² for those aged ≥ 65 years [39, 40]. eGFR was derived as per the Chronic Kidney Disease Epidemiology Collaboration (CKD-EPI) creatinine-cystatin C equation, which is more accurate at relatively preserved kidney function than creatinine-only equations [41] and improves the identification of early-stage kidney function impairment in diabetic populations [42]. For the longitudinal analysis, the longitudinal kidney outcome was rapid kidney function decline (RKFD), defined as an annual eGFR decline rate > 3 mL/min/1.73 m² or a > 30% reduction in eGFR from baseline [43, 44]. In a sensitivity analysis, RKFD was redefined using a clinically relevant composite outcome comprising either an annual eGFR decline > 5 mL/min/1.73 m²/year or incident impaired kidney function (eGFR < 60 mL/min/1.73 m²) at follow-up [45, 46].

Statistical analysis

Participant characteristics were summarized based on descriptive statistics. Continuous variables with a non-normal distribution were shown as median (interquartile range [IQR]), whereas categorical ones were shown as frequency counts (percentages). In the cross-sectional analysis, baseline characteristics were compared between participants with and without impaired kidney function. In the longitudinal analysis, baseline characteristics were stratified according to inflammatory-anemic status: neither, residual inflammation only, anemia only, or the combined phenotype. Multivariable logistic regression models were leveraged to estimate the associations of residual inflammation and anemia with prevalent impaired kidney function and age-adjusted impaired kidney function among diabetes participants in the cross-sectional analysis, and the risk of RKFD in the longitudinal analysis. Additionally, the association of the combined exposure with total eGFR change (mL/min/1.73 m²) during follow-up was examined via multivariable linear regression analysis. To evaluate the potential multiplicative interaction, a product term of residual inflammation and anemia was incorporated into the regression models. Results are reported as odds ratios (OR) or regression coefficients (β) with 95% confidence intervals (CIs). All analyses were adjusted uniformly for age, gender, education, marital status, smoking, drinking, hypertension, as well as WWI. Restricted cubic spline (RCS) models were fitted to explore the potential non-linear dose-response association of hs-CRP and hemoglobin with impaired kidney function risk. The consistency of these associations in patients with diabetes was examined via subgroup analyses by age, gender, marital status, smoking, drinking, and hypertension. Additive interaction was assessed utilizing attributable proportion (AP), relative excess risk due to interaction (RERI), and synergy index (SI).

The robustness of our primary findings was rated via sensitivity analyses by (1) redefining residual inflammation using a higher hs-CRP threshold (≥ 3 mg/L); (2) redefining RKFD using clinically relevant thresholds: annual eGFR decline > 5mL/min/1.73 m²/year or incident impaired kidney function (eGFR < 60 mL/min/1.73 m²) at follow-up; (3) conducting a quartile-based analysis of hs-CRP levels to assess its association with impaired kidney function risk; (4) repeating the cross-sectional analysis after excluding participants with eGFR < 45 mL/min/1.73 m² and stratifying the longitudinal analysis according to baseline CKD stage (G1: eGFR ≥ 90 and G2: 60–89 mL/min/1.73 m²); and (5) incorporating additional covariates including HbA1c, CVD, malignancy, arthritis, and medications. A two-sided P-value < 0.05 signified statistical significance. Statistical analyses were completed via R 4.5.1 (R Foundation for Statistical Computing, Vienna, Austria).

Results

Study population characteristics

A total of 2,363 adults with diabetes aged > 45 years were included in the cross-sectional analysis, among whom 191 (8.1%) had impaired kidney function (Table 1). In comparison to the group without impaired kidney function, participants with impaired kidney function were older, less likely to be currently married, and had a lower proportion with higher education. They also showed more adverse anthropometric and clinical profiles, including a higher WWI, elevated levels of inflammatory and renal biomarkers (hs-CRP, creatinine, and cystatin C), lower hemoglobin concentrations, and a markedly higher prevalence of hypertension. They also had a higher prevalence of CVD (30.9% vs. 20.5%, P = 0.001) and malignancy (16.2% vs. 11.0%, P = 0.042), as well as greater use of antihypertensive medications (63.4% vs. 38.0%, P < 0.001), LLD (24.1% vs. 17.0%, P = 0.019), CVD medications (26.7% vs. 15.0%, P < 0.001), and glucose-lowering medications (35.6% vs. 27.3%, P = 0.017), whereas the prevalence of liver disease and arthritis did not differ significantly between the two groups. Among the 191 participants with impaired kidney function, 139 (72.8%) had CKD stage G3 and 52 (27.2%) had stage G4/G5. The coexistence of residual inflammation and anemia was more common among participants with G4/G5 disease than among those with G3 disease (28.8% vs. 14.4%, P = 0.034), whereas median hs-CRP and hemoglobin levels did not differ significantly between the two stages (Table S1).

Table 1.

Baseline characteristics of study participants according to impaired kidney function status

	Overall (N = 2363)	Impaired kidney function		P value
		No (n = 2172)	Yes (n = 191)
Age, years	63.00 [57.00, 69.00]	62.00 [56.00, 68.00]	72.00 [66.00, 78.00]	< 0.001
Weight, kg	61.80 [53.90, 70.20]	61.80 [54.10, 70.23]	60.70 [51.35, 69.40]	0.069
Waist, cm	90.20 [82.60, 97.40]	90.10 [82.70, 97.40]	92.00 [82.45, 99.40]	0.205
WWI	1.45 [1.33, 1.58]	1.44 [1.33, 1.57]	1.49 [1.38, 1.64]	< 0.001
Hs-CRP, mg/L	1.90 [1.00, 3.60]	1.90 [1.00, 3.50]	2.50 [1.30, 5.05]	< 0.001
Hemoglobin, g/dL	13.70 [12.60, 14.80]	13.70[12.60, 14.80]	13.00 [11.90, 14.10]	< 0.001
Creatinine, mg/dl	0.75 [0.63, 0.90]	0.73 [0.63, 0.85]	1.19 [1.00, 1.47]	< 0.001
Cystatin C, mg/L	0.86 [0.74, 0.98]	0.84 [0.73, 0.95]	1.35 [1.22, 1.56]	< 0.001
Gender (%)				0.765
Male	1045 (44.20)	963 (44.30)	82 (42.90)
Female	1318 (55.80)	1209 (55.70)	109 (57.10)
Education (%)				0.018
Primary	227 (9.60)	199 (9.20)	28 (14.70)
Illiterate	182 (7.70)	162 (7.50)	20 (10.50)
Middle school	90 (3.80)	81 (3.70)	9 (4.70)
High school+	1864 (78.9)	1730 (79.70)	134 (70.2)
Marital (%)				< 0.001
Married	2013 (85.20)	1869 (86.00)	144 (75.40)
Single	350 (14.80)	303 (14.00)	47 (24.60)
Smoking (%)				0.377
Yes	918 (38.80)	850 (39.10)	68 (35.60)
No	1445(61.20)	1322(60.90)	123 (64.40)
Drinking (%)				0.130
Yes	662 (28.00)	618 (28.50)	44 (23.00)
No	1701 (72.00)	1554(71.50)	147 (77.00)
Hypertension (%)				< 0.001
Yes	1403 (59.40)	1251 (57.60)	152 (79.60)
No	960 (40.60	921 (42.40)	39 (20.40)
HbA1c (%)	6.60 [6.10, 7.50]	6.60 [6.10, 7.50]	6.70 [6.20, 7.45]	0.432
CVD (%)				0.001
Yes	505 (21.4)	446 (20.5)	59 (30.9)
No	1858 (78.6)	1726 (79.5)	132 (69.1)
Liver disease (%)				0.111
Yes	127 (5.4)	122 (5.6)	5 (2.6)
No	2236 (94.6)	2050 (94.4)	186 (97.4)
Malignancy (%)				0.042
Yes	271 (11.5)	240 (11.0)	31 (16.2)
No	2092 (88.5)	1932 (89.0)	160 (83.8)
Arthritis (%)				0.925
Yes	803 (34.0)	737 (33.9)	66 (34.6)
No	1560 (66.0)	1435 (66.1)	125 (65.4)
AHT (%)				< 0.001
Yes	946 (40.0)	825 (38.0)	121 (63.4)
No	1417 (60.0)	1347 (62.0)	70 (36.6)
AD (%)				0.017
Yes	660 (27.9)	592 (27.3)	68 (35.6)
No	1703 (72.1)	1580 (72.7)	123 (64.4)
LLD (%)				0.019
Yes	416 (17.6)	370 (17.0)	46 (24.1)
No	1947 (82.4)	1802 (83.0)	145 (75.9)
CVD drug (%)				< 0.001
Yes	377 (16.0)	326 (15.0)	51 (26.7)
No	1986 (84.0)	1846 (85.0)	140 (73.3)

Continuous variables with non-normal distributions were shown in median (IQR), whereas categorical ones were in frequency counts (percentages). Between-group comparisons were completed via Wilcoxon rank-sum tests for continuous variables and Pearson chi-square tests for categorical ones. Abbreviations: WWI, Weight-Adjusted Waist Index; Hs-CRP, high-sensitivity C-reactive protein; Hb, hemoglobin; HY, hypertension; AHT, antihypertensive agent; AD, antidiabetic agents; CVD, cardiovascular disease; LLD, lipid-lowering drug

Association of residual inflammation and anemia with impaired kidney function

As shown in Table 2 (cross-sectional analysis), residual inflammation (hs-CRP ≥ 2 mg/L) was associated with higher odds of impaired kidney function compared with no residual inflammation (Model 1: OR = 1.44, 95% CI: 1.04–1.99). Anemia was also associated with increased odds of impaired kidney function (Model 2: OR = 2.44, 95% CI: 1.71–3.46). In the combined exposure analysis, participants with both residual inflammation and anemia had markedly greater odds of impaired kidney function than those with neither condition (Model 3: OR = 3.61, 95% CI: 2.16–5.99). When residual inflammation and anemia were included simultaneously in the same model, both remained statistically associated with impaired kidney function (residual inflammation: OR = 1.49, 95% CI: 1.07–2.07; anemia: OR = 2.48, 95% CI: 1.74–3.52) (Table S2). In quartile analyses, relative to the lowest quartiles, the highest quartile of hs-CRP (> 3.6 mg/L) was linked to increased odds of impaired kidney function (OR = 2.07, 95% CI: 1.31–3.33), while the highest quartile of hemoglobin was associated with reduced odds (OR = 0.48, 95% CI: 0.29–0.78) (Table S3). Interaction on the multiplicative scale was assessed by including a product term between residual inflammation and anemia, and no significant multiplicative interaction was observed (P for interaction = 0.658). Additive interaction indices also suggested no significant interaction (RERI = 0.35, 95% CI: -1.52-2.22; AP = 0.10, 95% CI: -0.40-0.59; SI = 1.15, 95% CI: 0.53–2.49) (Table S4).

Table 2.

Association of residual inflammation and anemia with prevalent impaired kidney function among diabetic patients

Characteristic	OR (95% CI)	P value	P interaction
Model 1
Non-residual inflammation	Ref.	Ref.
Residual inflammation	1.44 (1.04, 1.99)	0.028
Model 2			0.658
Non-anemia	Ref.	Ref.
Anemia	2.44 (1.71, 3.46)	< 0.001
Model 3
Non-residual inflammation and non-anemia	Ref.	Ref.
Residual inflammation and non-anemia	1.56 (1.05, 2.34)	0.027
Non-residual inflammation and anemia	2.70 (1.61, 4.49)	< 0.001
Residual inflammation and anemia	3.61 (2.16, 5.99)	< 0.001

Data are presented as odds ratios (95% confidence intervals)

The definition of residual inflammation: hs-CRP ≥ 2.00 mg/L

Anemia was defined as hemoglobin < 13 g/dL (males) and < 12 g/dL (females)

Model 1 adjusted for age, gender, education, marital status, smoking, drinking, hypertension, and WWI

Model 2 adjusted for age, gender, education, marital status, smoking, drinking, hypertension, and WWI

Model 3 adjusted for age, gender, education, marital status, smoking, drinking, hypertension, and WWI

Abbreviations: Ref, reference; hs-CRP, high-sensitivity C-reactive protein; OR, odds ratio; CI, confidence interval; WWI, Weight-Adjusted Waist Index

Association of residual inflammation and anemia with RKFD

In the longitudinal analysis involving 547 participants with preserved baseline kidney function (eGFR ≥ 60 mL/min/1.73 m²), 89 individuals (16.3%) developed RKFD over a 4-year follow-up period. The incidence of RKFD differed significantly across exposure groups (P < 0.001): 11.8% (38/322) in the neither group, 15.7% (25/159) in the residual inflammation-only group, 36.6% (15/41) in the anemia-only group, and 44.0% (11/25) in the combined phenotype group (Table S5). Participants with residual inflammation (hs-CRP ≥ 2 mg/L) had higher odds of RKFD compared to those without residual inflammation (Model 1: OR = 2.12, 95% CI: 1.26–3.56). Anemia was also associated with higher odds of RKFD (Model 2: OR = 2.04, 95% CI: 1.05–3.46). The coexistence of residual inflammation and anemia was associated with the highest odds of RKFD (Model 3: OR = 6.42, 95% CI: 2.18–18.34) relative to participants with neither condition. No significant multiplicative interaction was observed (P for interaction = 0.222) (Table 3). When residual inflammation and anemia were entered simultaneously, both remained statistically associated with RKFD (residual inflammation: OR = 2.13, 95% CI: 1.26–3.58; anemia: OR = 2.04, 95% CI: 1.02–4.03) (Table S6). For additive interaction, AP was 0.64 (95% CI: 0.22–1.07; P = 0.001). Other additive indices were not statistically significant (RERI = 4.13, 95% CI: -2.46-10.73); SI = 4.21, 95% CI: 0.77–22.96) (Table S7).

Table 3.

Association of residual inflammation and anemia with the odds of RKFD among diabetic patients

Characteristic	OR (95% CI)	P value	P interaction
Model 1
Non-residual inflammation	Ref.	Ref.
Residual inflammation	2.12 (1.26,3.56)	0.004
Model 2
Non-anemia	Ref.	Ref.
Anemia	2.04 (1.05,3.46)	0.043
Model 3			0.222
Non-residual inflammation and non-anemia	Ref.	Ref.
Residual inflammation and non-anemia	1.86 (1.05,3.26)	0.032
Non-residual inflammation and anemia	1.43 (0.53,3.48)	0.450
Residual inflammation and anemia	6.42 (2.18,18.34)	< 0.001

Data are presented as odds ratios (95% confidence intervals)

The definition of residual inflammation: hs-CRP ≥ 2.00 mg/L

Anemia was defined as hemoglobin < 13 g/dL (males) and < 12 g/dL (females)

Model 1 adjusted for age, gender, education, marital status, smoking, drinking, hypertension, and WWI

Model 2 adjusted for age, gender, education, marital status, smoking, drinking, hypertension, and WWI

Model 3 adjusted for age, gender, education, marital status, smoking, drinking, hypertension, and WWI

Abbreviations: Ref, reference; hs-CRP, high-sensitivity C-reactive protein; OR, odds ratio; CI, confidence interval; eGFR, estimated glomerular filtration rate; RKFD, rapid kidney function decline; WWI, Weight-Adjusted Waist Index

We further modeled the continuous 4-year change in eGFR using multivariable linear regression. Participants with the combined phenotype showed a significantly steeper decline in eGFR (β=-8.68 mL/min/1.73 m², 95% CI: -15.72 to -1.64). In contrast, isolated residual inflammation (β=-2.35 mL/min/1.73 m², 95% CI: -5.61 to 0.90) and isolated anemia (β=-0.88 mL/min/1.73 m², 95% CI: -6.34 to 4.61) did not show a statistically significant association with the total eGFR change (Table S8) (Fig. S2).

Dose-response relationships of residual inflammation and anemia with impaired kidney function

In the total cross-sectional population (n = 2,363), RCS models indicated significant overall associations of hs-CRP (P-overall = 0.005) and hemoglobin (P-overall < 0.001) with impaired kidney function risk (Fig. 2). No evidence of non-linearity was detected for either biomarker (hs-CRP: P-nonlinear = 0.417; hemoglobin: P-nonlinear = 0.284).

Fig. 2.

Dose-response relationships of residual inflammation and anemia with prevalent impaired kidney function. Note: RCS (three knots) was used to model the continuous, nonlinear association between hs-CRP (Fig. A) and hemoglobin (Fig. B) with prevalent impaired kidney function. The solid line represents the multivariable-adjusted OR for impaired kidney function, with shaded areas indicating 95% CIs. A reference point marks hs-CRP=-0.511 (log-transformed), corresponding to OR = 1.0. Adjusted covariates included age, gender, education, marital status, smoking, drinking, hypertension, and WWI. Abbreviations: hs-CRP, high-sensitivity C-reactive protein; OR, odds ratio; CI, confidence interval

Subgroup analyses

Subgroup analyses based on the cross-sectional dataset were conducted by age (< 65 vs. ≥ 65), gender, marital status, smoking, drinking, as well as hypertension status (Fig. 3). No statistically significant effect modification was detected for the association between combined residual inflammation and anemia exposure with impaired kidney function risk in any subgroup (all P for interaction > 0.05). In the longitudinal analysis stratified by baseline eGFR category, the combined phenotype was associated with higher odds of RKFD in both G1 (OR = 6.34, 95% CI: 1.60-25.04) and G2 (OR = 5.80, 95% CI: 1.30–22.90), although the wide confidence intervals reflected limited precision because of the small subgroup sample sizes. No significant multiplicative interaction between residual inflammation and anemia was observed within either stratum (P for interaction = 0.379 and 0.674, respectively) (Table S9).

Fig. 3.

Subgroup analyses of the association of residual inflammation and anemia with prevalent impaired kidney function. Note: Subgroup analyses assessed the association between combined residual inflammatory risk (defined by hs-CRP) and anemia status and the prevalence of impaired kidney function. Low and high hs-CRP levels were defined as < 2 mg/L and ≥ 2 mg/L, respectively. Anemia was defined as hemoglobin < 13 g/dL (males) and < 12 g/dL (females). The reference group comprised individuals with low hs-CRP without anemia. Multivariable logistic regression models were adjusted for age, gender, education, marital status, smoking, drinking, hypertension, and Weight-Adjusted Waist Index (WWI). Abbreviations: hs-CRP, high-sensitivity C-reactive protein; CI, confidence interval. Significance levels: *P < 0.05, **P < 0.01, ***P < 0.001

Sensitivity analysis

After excluding participants with eGFR < 45 mL/min/1.73 m², the associations of residual inflammation (OR = 1.27, 95% CI: 1.04–1.56) and anemia (OR = 2.09, 95% CI: 1.39–3.10) with impaired kidney function remained significant. In the combined exposure analysis, participants with both residual inflammation and anemia had higher odds of impaired kidney function than those with neither condition (OR = 2.93, 95% CI: 1.49–5.76). No significant multiplicative interaction was observed (P for interaction = 0.186) (Table S10).

When the analysis was restricted to participants with complete 2015 covariate data (without supplementation from the 2011 wave), the associations were essentially unchanged (residual inflammation: OR = 1.71, 95% CI: 1.12–2.61; anemia: OR = 2.22, 95% CI: 1.45–3.39; combined phenotype: OR = 4.21, 95% CI: 2.23–7.93 for both vs. neither) (Table S11).

In the cross-sectional analysis, additional adjustment for HbA1c, comorbidities including CVD, malignancy, and arthritis, as well as lipid-lowering and CVD-specific medications, yielded consistent estimates. The combined phenotype was associated with an OR of 3.62 (95% CI: 2.16–6.03), with no evidence of multiplicative interaction (P for interaction = 0.556) (Table S12). In the longitudinal analysis, a simplified covariate model retaining core confounders, including medication use, yielded a similarly elevated risk for the combined phenotype (OR = 6.10, 95% CI: 2.05–17.52), with no evidence of interaction (P for interaction = 0.235) (Table S13). Furthermore, when RKFD was redefined using an alternative composite outcome comprising either an annual eGFR decline > 5 mL/min/1.73 m²/year or incident impaired kidney function (eGFR < 60 mL/min/1.73 m²) at follow-up, the association with the combined phenotype remained directionally consistent (OR = 3.21, 95% CI: 1.16–9.34; P = 0.028), with no evidence of multiplicative interaction (P for interaction = 0.511) (Table S14).

When the eGFR threshold (eGFR < 45 mL/min/1.73 m² for participants aged ≥ 65) was applied to define impaired kidney function, results aligned with those in the primary analysis (Table S15). Compared to the reference group, the odds of age-specific impaired kidney function were higher among participants with isolated residual inflammation (OR = 1.81, 95% CI: 1.09–3.09) or isolated anemia (OR = 3.18, 95% CI: 1.70–5.93). The combined phenotype was associated with the highest odds (OR = 4.78, 95% CI: 2.62–8.72). Using a higher hs-CRP cutoff to redefine residual inflammation (≥ 3 mg/L), participants with isolated residual inflammation demonstrated greater odds of impaired kidney function (OR = 1.51, 95% CI:1.09–2.10), while those with concurrent residual inflammation and anemia showed substantially higher odds (OR = 3.74, 95% CI: 2.21–6.22) (Table S16). Neither multiplicative (P for interaction = 0.683) nor additive interactions (RERI = 1.04, AP = 0.28, SI = 1.62) (Table S17) were statistically significant under this definition. Furthermore, subgroup analyses using this stricter threshold remained consistent with the primary findings, with insignificant interactions across strata (all P for interaction > 0.05) (Fig. S3).

Discussion

Our study unveiled the joint associations of residual inflammation and anemia with kidney function impairment and decline among Chinese middle-aged and older T2DM adults. In comparison to either condition alone, the combined phenotype showed a cumulative association with both impaired kidney function at baseline and rapid kidney function decline over follow-up. These findings suggest that integrated clinical consideration of residual inflammation and anemia, rather than either factor in isolation, may help identify a particularly vulnerable subgroup of the diabetes population and support earlier kidney-protective strategies.

Previous studies have linked hs-CRP with adverse kidney outcomes in diabetic populations. Residual inflammation, featuring persistently elevated levels of hs-CRP, influences the pathogenesis of varied chronic metabolic diseases [17–20]. Elevated hs-CRP has also been identified as a predictive biomarker associated with kidney function decline and DKD risk in individuals with diabetes or impaired glucose metabolism [47–50]. In line with these findings, the present results show that residual inflammation alone was associated with higher odds of kidney function impairment and more rapid kidney function decline in diabetic adults.

Anemia is highly prevalent and tends to occur earlier in individuals with diabetes, representing a common comorbidity in both DKD and CKD. Rather than representing a single causal pathway, anemia in diabetes may reflect the cumulative burden of chronic hypoxia, systemic inflammation, metabolic stress, and impaired erythropoiesis. Clinical trials indicate that correcting anemia mitigates cardiovascular risks but yield limited evidence for halting kidney disease progression. Therefore, anemia may be better viewed as a clinical risk marker [12, 51, 52]. Observational studies have further linked lower hemoglobin concentrations with ESRD, more severe renal histopathology, and the onset or progression of DKD [53–55]. Consistent with these reports, our study found that anemia alone was associated with higher odds of impaired kidney function and kidney function decline in diabetic adults.

Several previously proposed biological pathways may provide context for these associations. Anemia may reduce oxygen delivery and has been implicated in profibrotic responses through hypoxia-inducible factor-related pathways [56]. Concurrently, cytokine-mediated systemic inflammation, involving IL-1, IL-6, and TNF-α, may increase metabolic stress, promote endothelial dysfunction, and impair erythropoiesis through the IL-6-hepcidin axis by limiting iron availability [22, 57]. These potential mechanisms should be interpreted cautiously, as the observational design cannot determine whether residual inflammation and anemia are causal contributors to kidney dysfunction or overlapping manifestations of shared underlying processes. Future studies incorporating detailed inflammatory, erythropoietic, and iron-metabolic profiling are needed to clarify these relationships. Nevertheless, the stronger association observed for their coexistence suggests that this dual-risk phenotype may have potential clinical relevance.

Our findings support the need for integrated management strategies that address both residual inflammation and anemia in diabetes, rather than considering them in isolation. Routine assessment of hs-CRP and hemoglobin may help identify patients with this high-risk combined phenotype. Notably, SGLT2 inhibitors have been shown to attenuate systemic inflammation and lower the incidence of anemia in clinical studies [58–60], and are now recommended as first-line therapy for slowing the progression of DKD [61]. Emerging evidence also suggests that non-steroidal mineralocorticoid receptor antagonists and hypoxia-inducible factor prolyl hydroxylase inhibitors may affect related inflammatory, renoprotective, or erythropoietic pathways [62, 63]. It should be noted that the study period (2011–2015) predates the widespread clinical adoption in China of novel renoprotective therapies such as SGLT2 inhibitors and non-steroidal MRAs. This temporal context may help characterize the natural trajectory of this combined phenotype before newer therapies were widely used. Future prospective studies are needed to determine whether targeting these pathways can reduce the risk of DKD onset or progression in this vulnerable subgroup.

Several strengths of this study are prominent. First, leveraging a nationally representative CHARLS sample increases the external validity and generalizability of our findings to the broader middle-aged and older diabetic population. Second, exposures and outcomes were assessed using objective laboratory-based biomarkers measured under standardized protocols, supplemented by physician-diagnosed diabetes and medication use, thereby improving measurement validity. Third, given that age-related physiological decline in eGFR is common in cohorts of this age range, we further validated the cross-sectional findings using an age-adjusted threshold for impaired kidney function. The persistence of a strong association under this stricter criterion supports the interpretation that the observed signal reflects pathological rather than purely age-related decline. Fourth, the combination of cross-sectional and longitudinal analyses allowed simultaneous assessment of disease burden and progression. Finally, multiple sensitivity analyses, including RCS models and alternative cut-offs, yielded consistent results, reinforcing the robustness of the findings.

Several limitations should be acknowledged. First, a substantial proportion of participants were excluded from both analyses because of missing data or loss to follow-up, which may have introduced selection bias. In the longitudinal analysis, the relatively small sample size and limited number of RKFD events may have reduced statistical power and resulted in wide confidence intervals, particularly for the combined phenotype group. In addition, some covariates were defined using an ever-positive rule, which may have led to misclassification if participants’ status changed between waves. The generalizability of the findings may therefore be limited, and replication in larger and more diverse cohorts is warranted. Second, RKFD was defined based on eGFR measurements at baseline and at a single follow-up time point, which may not adequately distinguish sustained kidney function decline from transient fluctuations. Furthermore, urinary albumin data were unavailable, limiting the assessment of kidney dysfunction to its GFR-defined component. Although eGFR captures functional impairment, the absence of albuminuria measurements precluded characterization of albuminuric phenotypes. Future studies incorporating repeated eGFR measurements at multiple time points and albuminuria data are needed to provide a more comprehensive characterization of kidney outcomes in this population. Third, residual confounding from unmeasured factors, including time-varying medication use and subclinical inflammatory or hematologic conditions, cannot be excluded. Despite these limitations, the consistency of findings across multiple analytic approaches supports a clinically meaningful dual-risk phenotype of coexisting residual inflammation and anemia in adults with diabetes.

Conclusion

In this nationwide cohort of Chinese adults with T2DM, the combined phenotype of residual inflammation and anemia was associated with higher impaired kidney function prevalence and risk of subsequent RKFD compared with either condition alone, suggesting a cumulative burden on renal outcomes. Integrated screening and management of both inflammation and anemia may therefore represent an important strategy for kidney function preservation in this population.

Supplementary Information

Below is the link to the electronic supplementary material.

Abbreviations

CKD

Chronic kidney disease

CHARLS

China Health and Retirement Longitudinal Study

eGFR

Estimated glomerular filtration rate

CKD-EPI

Chronic Kidney Disease Epidemiology Collaboration

RKFD

Rapid kidney function decline

SBP

Systolic blood pressure

DBP

Diastolic blood pressure

WC

Waist circumference

WWI

Weight-adjusted waist index

OR

Odds ratios

CI

Confidence interval

RCS

Restricted cubic spline

RERI

Relative excess risk due to interaction

AP

Attributable proportion

SI

Synergy index

AHT

Antihypertensive agent

AD

Antidiabetic agents

CVD

Cardiovascular disease

LLD

Lipid-lowering drug

Author contributions

Wenhao Yang: Conceptualization, Methodology, Writing-Original draft preparation. Jinkui Zhao: Investigation. Mingyan Yu: Visualization. Qianyu Li: Data curation. Rui Li: Supervision. Xihai Gao: Resources. Changyong Lin: Validation. Xiao Huang: Funding acquisition. Zihan Zhai: Software. Bo Wu: Project administration, Writing-Reviewing and Editing, and all authors commented on previous versions of the manuscript. All authors read and approved the final manuscript.

Funding

This study was funded by the Research Projects Fund of Binzhou Polytechnic University (2024yjkt03).

Data availability

Publicly available datasets were analyzed. CHARLS data can be accessed at http://charls.pku.edu.cn/en and are de-identified. Respondents are identified solely by unique ID numbers.

Declarations

Ethics approval and consent to participate

Our study used de-identified secondary data from CHARLS, which obtained ethical approval from the Peking University Research Ethics Committee (IRB 00001052–11015). All participants gave informed consent in writing during original data collection.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.