Introduction

Chronic kidney disease (CKD) is a major public health problem affecting more than 10% of the general population in many countries worldwide [1]. This equals roughly 850 million individuals, a number that is twice as high as the number of people with diabetes [2]. CKD is considered a systemic inflammatory disease with many causes [3]. This proinflammatory state contributes to vascular and myocardial remodelling and accelerated aging resulting in atherosclerotic lesions and vascular and myocardial fibrosis. Therefore, CKD is—amongst other factors—associated with an exceptionally high cardiovascular disease (CVD) morbidity and mortality including a high risk for heart failure [4-6]. Recent research suggests that limiting inflammation in particular in patients with moderate CKD might be beneficial [7]. The leading biomarker for inflammation is C-reactive protein (CRP). However, CRP is a signal for inflammation but has no causal role in atherosclerosis development [8].

Apolipoprotein A-IV (apoA-IV) is a 46 kDa glycoprotein involved in reverse cholesterol transport modulating lipid metabolism and glucose homeostasis [9-11]. It is produced in enterocytes and serves as a constituent of chylomicrons, very-low-density lipoprotein and high-density lipoprotein (HDL), or remains not associated with lipoproteins [12-14]. ApoA-IV has antiatherogenic, antioxidative, antithrombotic and anti-inflammatory properties [15-23]. The experimental evidence for the anti-inflammatory properties is quite strong: in a mouse model for acute colitis, apoA-IV significantly and specifically delayed the onset and reduced the severity and extent of inflammation induced by dextran sulfate sodium. ApoA-IV knockout mice exhibited a significantly greater inflammatory response to dextran sulfate sodium, which was reversed by exogeneous administration of apoA-IV [23]. Furthermore, apolipoprotein E–deficient mice that overexpress human apoA-IV showed significantly less atherosclerotic lesions. Lymphocytes of these animals treated with lipopolysaccharides to mimic chronic infection produced less proinflammatory cytokines and recombinant apoA-IV blocked the lipopolysaccharide-induced stimulation of monocytes [18].

Mice transgenic for human apoA-IV are protected against atherosclerosis by a mechanism that does not involve an increase in HDL cholesterol concentration [16]. Recombinant apoA-IV treatment was reported to be associated with a more stable plaque phenotype [17]. Data on apoA-IV and atherosclerosis in humans are sparse, but suggest an inverse association of apoA-IV concentrations with CVD [24-28]. Recent studies described a relationship between apoA-IV glycation and coronary artery disease severity in patients with type-2 diabetes mellitus [29].

Kidney function has a major influence on apoA-IV concentrations. Concentrations increase with decreasing estimated glomerular filtration rate (eGFR) [26, 27], and apoA-IV was shown to be an early marker of kidney impairment [30]. Furthermore, higher concentrations predicted the progression of CKD [31]. ApoA-IV concentrations in dialysis patients have been reported to be twice as high compared to controls without CKD [32, 33]. In a study including more than 1200 haemodialysis patients with type-2 diabetes mellitus, we reported that low apoA-IV concentrations were strongly associated with a higher risk for all-cause mortality, congestive heart failure and sudden cardiac death [34]. Prospective data on patients with CKD with moderately reduced eGFR and/or albuminuria, the overwhelming majority of patients with CKD, are lacking. We therefore aimed to examine the association between apoA-IV and all-cause mortality, incident CVD and heart failure in a large, prospective cohort study of 5141 patients with CKD, the German Chronic Kidney Disease (GCKD) study [35]. These patients can be considered as high-risk patients for CVD.

Methods GCKD study

The GCKD study is a prospective cohort study with ongoing follow-up investigations including 5217 Caucasian patients with CKD enrolled between the years 2010 and 2012. The design and details of the study have been published previously [35, 36]. Briefly, the study aimed to enrol patients with an eGFR of 30–60 ml/min/1.73m2 (Kidney Disease Improving Global Outcomes [KDIGO] stage G3, A1–3) or an eGFR >60 ml/min/1.73m2 in the presence of overt proteinuria (KDIGO stage G1–2, A3) under regular care by nephrologists. Exclusion criteria were non-Caucasian ethnicity, solid organ or bone marrow transplantation, active malignancy within 24 months prior to screening, heart failure New York Heart Association Stage IV, legal attendance or inability to provide consent. Patients were enrolled according to a standardised procedure by trained personnel. Blood and spot-urine samples were collected, processed and shipped on dry ice to a central biobank where routine laboratory parameters were measured centrally, as reported previously [35]. eGFR was estimated using the CKD-EPI formula.

The study was approved by the ethics committees of all participating institutions and registered in the national registry for clinical studies (DRKS 00003971). All methods were carried out in accordance with approved guidelines and the Declaration of Helsinki. Written informed consent was obtained from each study participant. Data are collected and managed using Askimed (https://www.askimed.com) as a cloud-based web platform.

Patients are followed on a yearly basis by trained personnel who collect data on hospitalisations, outcome events and medical history using a structured interview. Any hospital discharge reports are collected from the treating physicians and/or hospitals. Endpoints are continually extracted from these reports by an endpoint committee. The prospectively collected endpoints considered in the present analysis refer to the first 6.5 years of follow-up, based on data export from 8 October 2020.

Definition of outcomes

CVD at baseline comprised a history of nonfatal myocardial infarction, coronary artery bypass grafting, percutaneous transluminal coronary angioplasty, stroke and interventions at the carotid arteries (carotid endarterectomy and/or carotid balloon angioplasty or stent implantation).

The prospectively collected endpoints considered in the present analysis refer to the first 6.5 years of follow-up. All-cause mortality was defined as the primary endpoint. Major adverse cardiovascular events (MACE), hospitalisations and death due to heart failure and further specific causes of death were considered secondary endpoints. MACE were either defined as 3-point MACE including death from cardiovascular causes including myocardial infarction, coronary heart disease, decompensated congestive heart failure, sudden cardiac death and ischemic stroke as well as acute nonfatal myocardial infarction (ST-elevation myocardial infarction, STEMI and non-ST-elevation myocardial infarction, NSTEMI), nonfatal stroke or as 4-point MACE additionally including fatal peripheral ischemia, amputation due to peripheral vascular disease and surgical or percutaneous revascularisation due to peripheral vascular disease. Hospitalisation due to heart failure was defined as follows: hospitalisation with either evidence of a reduced left ventricular ejection fraction (<35%) or radiological evidence of pulmonary venous congestion, alveolar oedema or presence of bilateral or right-sided pleural effusion with a presumed cardiac cause.

Measurement of apoA-IV plasma concentrations

Plasma apoA-IV concentrations were measured at the Medical University of Innsbruck with a double-antibody ELISA using an affinity-purified polyclonal rabbit antihuman apoA-IV antibody for coating. This antibody coupled with horseradish peroxidase was used for detection. Plasma with a known content of apoA-IV served as the calibration standard. Patient samples were diluted 1:12500. In case of the rare situation of very low or very high concentrations outside of the linear range, samples were repeated using a lower or higher dilution, respectively. The intra- and inter-assay coefficients of variation were 2.7% and 6%, respectively [31, 33, 34, 37].

Statistical analysis

Baseline characteristics of the GCKD study participants are provided by quartiles of apoA-IV concentrations. Jonckheere–Terpstra tests for trend were performed for comparison of continuous variables between quartile groups of apoA-IV, and linear by linear χ2-square tests were applied for comparison of categorical variables. Linear regression analysis was used to identify variables associated with apoA-IV concentrations. Association with prevalent CVD was assessed with logistic regression analysis. Cause-specific hazard ratios (HRs) and corresponding 95% confidence intervals (CIs) were calculated for all-cause mortality, noncardiovascular mortality, incident cardiovascular endpoints and death or hospitalisation due to heart failure using Cox proportional hazards regression models. Within this approach, for the cause-specific endpoints, all deaths from other causes were treated as censored observations. The proportional hazards assumption was tested by χ2-test based on Schoenfeld residuals. In addition, for noncardiovascular mortality, incident cardiovascular endpoints and death and hospitalisation due to heart failure subdistribution hazard ratios (SHRs) based on competing risks survival regression were calculated treating all other causes of death as competing events. For both baseline and follow-up analyses, model 1 was adjusted for age, sex, eGFR and urine albumin–creatinine ratio (UACR). In model 2 serum albumin, low-density lipoprotein (LDL) cholesterol, smoking status, diabetes mellitus, statin use, triglycerides, body mass index (BMI), systolic and diastolic blood pressure (and CVD at baseline in the Cox regression analyses) were considered as additional covariables. This selection was based on clinical reasons and by taking into account differences of variables between quartiles of apoA-IV, as shown in Table 1, and associations of variables with apoA-IV, as displayed in Table 2. Forest plots of HR for quartiles of apoA-IV for all-cause mortality, 3-point MACE, 4-point MACE and death and hospitalisation due to heart failure were generated based on model 2. Based on the continuous prospective net reclassification index (NRI) for a median follow-up time of 6.5 years, it was evaluated if apoA-IV concentrations contributed to a better risk classification of individuals in terms of major endpoints as compared to a model already including renal and cardiac risk factors or parameters. The NRI was considered significant if the 95% CIs, determined empirically across 100 subsampling runs, excluded zero. The advantage of the continuous NRI is that it does not depend on the arbitrary selection of specific risk categories, and any change in predicted risk in the correct direction is appropriate [38]. Analyses on NRI were done with the function nricens in R. Due to the skewed distribution UACR, high-sensitivity C-reactive protein (hs-CRP) and triglycerides were log transformed based on the natural logarithm (ln). Estimates were given for an increment of 10 mg/dl in apoA-IV concentrations, as well as for quartile groups of apoA-IV (using quartile 1 as reference). Further, a formal mediation analysis applying the product method approach [39] was performed, assuming hs-CRP as a mediating variable. These models were adjusted as in model 2 and 95% CIs were derived via bootstrap sampling using 1000 iterations. Finally, a time-dependent Cox proportional hazards model was applied as a sensitivity analysis. All variables where several measurements for a patient during the complete observation time were available were taken further for analysis (for more details, see [40]).

Table 1. Baseline characteristics of German Chronic Kidney Disease study patients stratified by quartiles of apolipoprotein A-IV Apolipoprotein A-IV quartiles Quartile 1 (n = 1285) Quartile 2 (n = 1286) Quartile 3 (n = 1285) Quartile 4 (n = 1285) p-value for trend

ApoA-IV (mg/dl): range

 Mean ± SD

 25th, 50th and 75th percentile

5.2–22.0

18.1 ± 3.0

(16.2; 18.6; 20.5)

22.0–27.6

24.8 ± 1.6

(23.4; 24.8; 26.1)

27.6–34.0

30.6 ± 1.8

(29.0; 30.4; 32.1)

34.0–100.2

42.0 ± 7.6

(36.5; 39.8; 45.1)

- Age (years) 60 ± 12 (52; 63; 70) 61 ± 12 (55; 65; 70) 61 ± 12 (55; 64; 70) 59 ± 12 (51; 62; 69) 0.005 Female gender, n (%) 588 (46) 501 (39) 484 (38) 477 (37) <0.001 Body mass index (kg/m2)

30.3 ± 6.1

(26.2; 29.7; 33.7)

30.1 ± 6.0

(25.9; 29.1; 33.5)

29.8 ± 6.0

(25.9; 28.8; 33;1)

29.0 ± 5.6

(25.0; 28.2; 32.4)

<0.001 Smoker and ex-smoker, n (%) 716 (56) 726 (56) 773 (60) 808 (63) <0.001 Diabetes, n (%) 385 (30) 441 (34) 477 (37) 526 (41) <0.001 Hypertension, n (%) 1195 (93) 1242 (97) 1242 (97) 1271 (99) <0.001 Cardiovascular disease, n (%) 338 (26) 323 (25) 363 (28) 298 (23) 0.26 eGFR (ml/min/1.73 m2) 58 ± 20 (44; 54; 67) 51 ± 17 (40; 49; 59) 47 ± 16 (36; 44; 54) 42 ± 14 (31; 40; 48) <0.001 Statin use, n (%) 549 (43) 568 (44) 631 (49) 693 (54) <0.001 UACR (mg/g)

198 ± 585

(6; 19; 105)

274 ± 712

(7; 31; 193)

428 ± 973

(11; 55; 383)

814 ± 1284

(39; 281; 1075)

<0.001 Serum albumin (g/L)

39.0 ± 4.0

(36.8; 39.3; 41.2)

38.8 ± 4.0

(36.9; 39.1; 41.1)

38.3 ± 4.6

(36.3; 38.5; 40.6)

37.3 ± 4.8

(35.2; 37.8; 40.3)

<0.001 Haemoglobin (g/dl)

13.9 ± 1.7

(12.8; 13.8; 15.0)

13.8 ± 1.6

(12.8; 13.8; 14.9)

13.6 ± 1.6

(12.5; 13.5; 14.7)

13.2 ± 1.7

(12.1; 13.1; 14.3)

<0.001 Hs-CRP (mg/L)

6.7 ± 12.8

(1.3; 2.9; 7.1)

4.7 ± 6.9

(1.1; 2.3; 5.1)

4.2 ± 6.2

(1.0; 2.2; 4.6)

3.4 ± 4.7

(0.8; 1.9; 4.0)

<0.001 Total cholesterol (mg/dl)

204 ± 47

(171; 203; 232)

210 ± 50

(176; 208; 240)

209 ± 50

(174; 206; 238)

221 ± 61

(181; 213; 251)

<0.001 LDL cholesterol (mg/dl)

116 ± 39

(88; 112; 141)

119 ± 42

(91; 115; 143)

116 ± 42

(89; 113; 141)

122 ± 50

(89; 115; 148)

0.14 HDL cholesterol (mg/dl) 50 ± 16 (38; 47; 58) 51 ± 16 (39; 47; 59) 51 ± 18 (39; 48; 61) 56 ± 21 (42; 52; 67) <0.001 Triglycerides (mg/dl)

183 ± 104

(113; 160; 223)

199 ± 119

(122; 168; 239)

205 ± 138

(120; 175; 249)

209 ± 143

(117; 171; 257)

<0.001 Note: Values are provided as mean ± standard deviation and (25th; 50th [median]; and 75th percentiles) or as number of patients, n (%). In the total group, for all variables displayed, the number of missing values are ≤2.0% (n = 5141). eGFR (estimated glomerular filtration rate) calculated according to the CKD-EPI equation. Hs-CRP (high-sensitivity C-reactive protein) and urine-albumin values that were below the lower detection limit (LOD) were replaced by LOD/√2. Body mass index was corrected for amputation. UACR (urine albumin–creatinine ratio) was calculated according to the following equation: albumin in urine (mg/l) × 100/creatinine in urine (mg/dl), and is given in mg/g. Hypertension was defined as systolic blood pressure ≥140 mmHg and/or diastolic blood pressure ≥90 mmHg, and/or receiving antihypertensive treatment. Cardiovascular disease was defined as myocardial infarction, coronary artery bypass grafting, percutaneous transluminal coronary angioplasty, stroke and/or interventions at the carotid arteries. Table 2. Linear regression analysis investigating the influence of various clinical parameters on apolipoprotein A-IV concentrations. All variables listed are included in the analysis at the same time β-estimate SE p-value Age −0.150 0.011 <0.001 Female sex −1.705 0.261 <0.001 eGFR −0.236 0.007 <0.001 Urine albumin–creatinine ratioa 0.002 0.0001 <0.001 Body mass index −0.125 0.021 0.013 Current and ex-smoker 0.733 0.241 0.002 Diabetes 3.019 0.263 <0.001 Statin use 0.783 0.234 0.001 HDL cholesterol 0.145 0.008 <0.001 Triglyceridesa 0.007 0.001 <0.001 Hs-CRPa −0.147 0.014 <0.001 Note: Increment one unit for continuous variables. Triglycerides and hs-CRP were ln-transformed due to their skewed distribution. Abbreviations: eGFR, estimated glomerular filtration rate; HDL, high-density lipoprotein; Hs-CRP, high-sensitivity C-reactive protein; SE, standard errors.

Statistical analysis was performed using SPSS for Windows, version 26.0 (IBM Corp., Armonk, New York, NY, USA) and R for Windows, version 4.0.4 (Vienna, Austria) (https://www.r-project.org). For all analyses performed, a two-sided test p-value <0.05 was considered statistically significant.

Results Baseline analysis

Baseline apoA-IV concentrations of patients were available in 5141 of 5217 patients included in the GCKD study with a mean plasma concentration of 28.9 ± 9.8 mg/dl. Clinical characteristics of patients at study baseline according to quartiles of apoA-IV are presented in Table 1. The most striking difference was the decreasing eGFR (and increasing UACR) alongside increasing apoA-IV levels. This became even more evident when apoA-IV concentrations were assessed in the context of the KDIGO risk classification system based on cross categories of eGFR and UACR ('heat map', Fig. 1).

Mean (± standard deviation) apolipoprotein A-IV (apoA-IV) concentrations and number of patients stratified by estimated glomerular filtration rate (eGFR) and urine albumin–creatinine ratio (UACR) risk categories (including nephrotic range albuminuria >2220 mg/g) according to Kidney Disease Improving Global Outcomes guidelines in the German Chronic Kidney Disease (GCKD) study. Increasing concentrations of apoA-IV are displayed with cell backgrounds in increasingly darker shades of blue (increment: 2.5 mg/dl in apoA-IV concentration). Note: numbers of patients do not add up to the total number from GCKD with available apoA-IV values due to missing values for eGFR and UACR.

The median hs-CRP concentrations in patients in the highest quartile of apoA-IV concentrations were significantly lower as in the quartile with the lowest apoA-IV concentrations (1.9 mg/L interquartile range [IQR] [0.8; 4.0] vs. 2.9 mg/L IQR [1.3; 7.1], p < 0.001) (Table 1). This was surprising considering that this quartile included patients who were already markedly more advanced in terms of kidney impairment and other clinical characteristics.

Variables independently associated with apoA-IV concentrations are shown in Table 2. Besides age and sex, a lower eGFR, an increased UACR as well as increased HDL-C and triglyceride concentrations were significantly associated with increasing apoA-IV concentrations. The presence of diabetes mellitus was associated with 3 mg/dl higher apoA-IV concentrations, whereas increased hs-CRP concentrations and an elevated BMI were associated with decreasing apoA-IV concentrations.

CVD at baseline

We analysed the association between apoA-IV concentrations and prevalent CVD. A total of 1289 of the 5141 patients had already experienced a CVD event before baseline and there was a strong inverse association with apoA-IV concentrations: in the age-, sex-, eGFR- and UACR-adjusted model, each increment of apoA-IV by 10 mg/dl revealed 15% lower odds for prevalent CVD (odds ratio [OR] = 0.85, 95% CI 0.78–0.92, p = 0.0001). This association remained stable with an extended adjustment for serum albumin, LDL cholesterol, smoking status, diabetes mellitus, statin use, triglycerides, BMI and systolic and diastolic blood pressure: OR = 0.80, 95% CI 0.72–0.86, p = 0.0000003 (Table 3). This holds even true when hs-CRP was added to the analysis (Table S1). When apoA-IV was categorised in quartiles, the probability for prevalent CVD decreased for apoA-IV values above the 25% percentile for all three models, with the most significant result when the fourth quartile was compared to the first (Table 3 and Table S1).

Table 3. Association of apolipoprotein A-IV with prevalent cardiovascular disease (1289 out of 5141 patients) OR 95% CI p-value Calculations per 10 mg/dl increment of apoA-IV concentrations Model 1 0.85 0.78–0.92 0.0001 Model 2 0.80 0.72–0.86 0.0000003 Calculations per quartile of apoA-IV concentrations Model 1 Quartile 1 1.00 Quartile 2 0.80 0.66–0.97 0.02 Quartile 3 0.90 0.74–1.09 0.28 Quartile 4 0.72 0.58–0.89 0.003 Model 2 Quartile 1 1.00 Quartile 2 0.80 0.65–0.99 0.04 Quartile 3 0.87 0.70–1.07 0.18 Quartile 4 0.63 0.50–0.79 0.00008 Note: Model 1: adjusted for age, sex, estimated glomerular filtration rate and ln-urine albumin–creatinine ratio. Note: Model 2: as model 1 plus serum albumin, low-density lipoprotein cholesterol, smoking status, diabetes mellitus, statin use, ln-triglycerides, body mass index and systolic and diastolic blood pressure. Note: Prevalent cardiovascular disease was defined as myocardial infarction, coronary artery bypass grafting, percutaneous transluminal coronary angioplasty, stroke and/or interventions at the carotid arteries. Abbreviations: CI, confidence interval; OR, odds ratio. All-cause mortality and noncardiovascular mortality

During a median follow-up of 6.5 years, 600 out of 5041 patients (11.9%) with available apoA-IV concentrations, eGFR, UACR and follow-up data died. There was no indication for violation of the proportional hazards assumption in primary and secondary outcome analyses. Table 4 shows the results from Cox regression analyses. Each increase of apoA-IV by 10 mg/dl was associated with a 17% lower risk to die during the observation period in the age-, sex- and kidney function-adjusted model 1 (HR = 0.83, 95% CI 0.75–0.92, p = 0.0003). This association remained significant in extended model 2 additionally adjusted for serum albumin, LDL cholesterol, smoking status, diabetes mellitus, statin use, triglycerides, BMI, systolic and diastolic blood pressure and CVD at baseline (HR = 0.81, 95% CI 0.73–0.89, p = 0.00004). An analysis according to quartiles of apoA-IV mirrored the results from the continuous analysis, which, however, reached statistical significance only in quartile 4 and barely missed significance in quartiles 2 and 3 (Fig. 2, adjustment according to extended model 2). ApoA-IV concentrations were also significantly associated with noncardiovascular mortality (Table 4).

Table 4. Association of apolipoprotein A-IV with outcomes during the prospective follow-up For each increase of apoA-IV by 10 mg/dl Adjustment model HR (95% CI) p-value All-cause mortality (n = 600)a Model 1 0.83 (0.75–0.92) 0.0003 Model 2 0.81 (0.73–0.89) 0.00004 Noncardiovascular mortality (n = 433)a Model 1 0.80 (0.71–0.90) 0.0003 Model 2 0.76 (0.68–0.86) 0.00001 3-point MACE (n = 506)a Model 1 0.89 (0.80–0.99) 0.03 Model 2 0.88 (0.79–0.99) 0.03 4-point MACE (n = 681)a Model 1 0.90 (0.82–0.99) 0.03 Model 2 0.88 (0.80–0.97) 0.01 Death and hospitalisation due to heart failure (n = 346)a Model 1 0.83 (0.72–0.95) 0.006 Model 2