Combination of biomarkers of vascular calcification and sTWEAK to predict cardiovascular events in chronic kidney disease

Background and objectives: Vascular calcification and atherosclerosis have been related with an excess of cardiovascular mortality associated with chronic kidney disease (CKD). Different proteins such as osteoprotegerin (OPG), osteopontin (OPN) are involved in both conditions. In addition, soluble tumor necrosis factor-like weak inducer of apoptosis (sTWEAK) is a proinflammatory cytokine that has been related to cardiovascular disease. We hypothesized that circulating levels of OPG, OPN and sTWEAK may relate to a higher prevalence of cardiovascular outcomes in patients with CKD. Design, setting, participants, & measurements: Baseline circulating levels of OPG, OPN and sTWEAK were measured in 565 patients with CKD stages 3-5D [age, 61(48-68) years old, median (IQR); 359 men] without any previous CV event from The National Observatory of Atherosclerosis in Nephrology (NEFRONA) Study. Patients were followed for cardiovascular outcomes (follow-up of 3.11±1.26 years). Results: After the follow-up, 30 fatal and 33 nonfatal cardiovascular events occurred. At baseline, OPG and OPN levels were increased and sTWEAK concentrations were decreased in CKD patients suffering a fatal or nonfatal cardiovascular event. In a Cox model, after controlling for potential confounding factors, patients with OPG or OPN above and sTWEAK below optimal cut-off points obtained from ROC analysis had a higher risk of fatal and nonfatal cardiovascular events [HR: 2.62 (1.59-5.04); p<0.01; 2.88 (1.59-5.24); p<0.005; 2.00 (1.04-3.84), p<0.05; respectively]. In addition, when CKD patients were grouped according to the number of biomarkers above (OPG and OPN) or below (sTWEAK) of their cut-off points, the combination of the three biomarkers had the highest risk for fatal and nonfatal cardiovascular events [HR: 10.


Introduction
Chronic kidney disease (CKD) is associated with a high incidence of cardiovascular events (CVE) and mortality (1). In fact, CVE and mortality increase progressively as glomerular filtration rate decreases (2-3). Traditional risk factors such as hyperlipidemia, hypertension, diabetes and smoking fail to fully explain the increased CV risk in CKD patients (4). Tools able to improve CV risk assessment are needed in CKD patients. In this context, the addition of vascular calcification (VC) scores to traditional risk factors improves CV risk assessment in CKD patients (5).
Vascular calcification is the result of an accumulation of calcium and phosphate salts within the arterial wall as well as in cardiac valves. Data from animal models have identified different factors such as osteopontin (OPN) and osteoprotegerin (OPG) as factors that may regulate calcification in the arterial wall (6-7). OPG is a key cytokine that belongs to the tumor necrosis factor (TNF) receptor superfamily, which has a range of pleiotropic effects on bone metabolism, endocrine function and the immune system (8). OPG inhibits osteoclastic bone resorption by binding to the receptor activator of nuclear factor-κB ligand (RANKL), acting as a decoy receptor to competitively inhibit RANKL interaction with its receptor, RANK (9). Circulating OPG levels have been associated with the presence of VC and all-cause mortality in CKD patients (10). In addition, elevated serum OPG levels increased the risk of CVD and all-cause mortality in elderly women, and the association was more evident in women with poorer renal function (11). On the other hand, OPN belongs to the small integrin-binding ligand N-linked glycoprotein family (12).
This protein is produced and secreted by different cell types such as macrophages, T cells, renal and vascular smooth muscle cells (VSMCs) as well as osteoblast and osteoclast (13). In CKD, elevated levels of OPN predicted overall and cardiovascular mortality, although this effect was lost after adjustment for inflammatory biomarkers (14).
Tumor necrosis factor-like weak inducer of apoptosis (TWEAK) is a proinflammatory cytokine of the TNF-superfamily that circulates in serum as a soluble form (sTWEAK). Different studies have demonstrated a key role of this cytokine in atherosclerotic plaque development, progression and rupture (15)(16)(17). Recently, it has been demonstrated that TWEAK also participates in 3 VSMC calcification (18). In addition, loss-of function experiments have shown that TWEAK increases atherosclerotic plaque calcification (15). Finally, circulating soluble TWEAK (sTWEAK) levels have been associated with cardiovascular outcomes in CKD patients (19)(20).
The National Observatory of Atherosclerosis in Nephrology (NEFRONA) Study is a multicenter, observational, prospective study designed to analyze the prevalence of atherosclerosis and its associated risk factors in patients with CKD (21). In this study, we evaluated the association between selected calcification biomarkers, sTWEAK and their combination with cardiovascular outcomes in the NEFRONA population.
Briefly, the study included male and female without history of CVD (acute myocardial infarction, angina pectoris, hemorrhagic or ischemic stroke, atherosclerosis, and abdominal aortic aneurysm). CKD patients in this substudy were enrolled within 57 Spanish primary care centers distributed in 32 different regions from Spain. The exclusion criteria included previous CV events, pregnancy, having received any organ transplantation, active infections and, having a life expectancy of <1 year.

Events
Primary outcomes were CVD events according to the International Classification of Diseases of the World Health Organization, which included myocardial infarction, unstable angina, transient ischemic attack, cerebrovascular accident, arrhythmia, congestive heart failure, peripheral artery disease or amputation for vascular disease, and aorta aneurism. CV mortality was defined as cerebrovascular accident (ischemic or hemorrhagic), myocardial ischemia and infarction, hyperkalemia or arrhytmia, sudden death, hemorrhage due to aneurysm rupture and mesenteric infarct. Non-CV mortality causes included neoplasia, accident, infection, uremic, nondetermined or unknown.
The local Ethics Committee of the Hospital Arnau de Vilanova approved the protocol. The authors adhere to the declaration of Helsinki and patients were included after providing informed consent.

Clinical and biochemical data
Patients were asked to complete a questionnaire at recruitment including clinical history of hypertension, dyslipidemia and diabetes, CV risk factors and medication use. Biochemical parameters were obtained from a routine fasting blood test. Serum sTWEAK levels was determined in duplicate with a commercially available ELISA kit (Bender MedSystems), and OPG and OPN levels by multiplex kits (Merck Millipore).

Statistical Analysis
Statistical analyses were performed using SPSS 11.0 (SPSS Inc, Chicago, IL) statistical package. Normally distributed variables were expressed as mean±SD, and non-normally distributed variables were expressed as medians (IQR, expressed as the 25 th and 75 th percentile). Between-group comparisons were assessed for nominal variables with the chi-squared test and Mann-Whitney U test. Spearman rank correlation was used to determine correlations between variables. A receiver operator characteristic curve analysis was done to determine the OPG, OPN and sTWEAK cut-off points and maximum sensitivity and highest specificity for prediction of cardiovascular event. A categorical variable was generated containing the information of the number of biomarkers that were over (or under, in the case of sTWEAK) the cut-off point. Time to event analysis of CV outcomes was done using the Cox proportional hazards model, including adjustment for potential confounding factors. Data are presented in the form of Hazard ratios (HRs) and 95% confidence intervals (95% CIs). Statistical differences in c statistics were compared using the method by DeLong et al. (25); 95% CIs were calculated for each comparison. Kaplan-Meier curves were used to compare time to outcome according to a multimarker score. P value <0.05 was considered statistically significant.

Patient characteristics
Characteristics of patient population according to cardiovascular outcomes are summarized in Table 1 There were significant differences among CKD patients suffering a CV event regarding age, SBP, glucose concentrations, eGFR, diabetes, and circulating OPG, OPN and sTWEAK levels (Table 1).
Univariate association of OPG, OPN and sTWEAK are given in Table 2. The three biomarkers correlated with cholesterol and LDL-c concentrations, eGFR and ABI. In addition, OPG was positively correlated with OPN. OPG and OPN were negatively associated with sTWEAK.  Table 1. There were significant differences among CKD patients suffering a CV event regarding age, SBP, glucose concentrations, eGFR, diabetes, and circulating OPG, OPN and sTWEAK levels. level of >11.4 ng/mL and sTWEAK threshold concentration <307 pg/mL had the highest sensitivity (73% and 67%, respectively) and specificity (57% and 58%, respectively) for the identification of patients suffering a CV event.

OPG, OPN, sTWEAK and CV outcomes
Kaplan-Meier curves showed a significant association of OPG or OPN above and sTWEAK below cut-off points with probability to suffer a CV event (p<0.001 for all) (Fig. 1B).
The predictors for time-to-cardiovascular event (N=63, including a composite of fatal and nonfatal) were studied by univariate and multivariate Cox analysis.
In univariate Cox, age, diabetes, eGFR, being on dialysis, insulin treatment, OPG or OPN above and sTWEAK below of their cut-off points obtained from ROC analysis were significant predictors of outcome (Table 3). Multivariate Cox was used to study the effect of variables that were statistically significant in the univariate analysis. After that, only OPG or OPN above and sTWEAK below of their cut-off points persisted as independent predictors of CV events (Table 3).
Since the three biomarkers correlate with each other and are independent predictors of CV events, a multimarker combination was developed according to the number of biomarkers whose values were above (OPG and OPN) or below (sTWEAK) of their cut-off points. Thus, a patient could have a multimarker score of 0 to 3. The score value was 0 in 31.2% patients, 1 in 8 34.5%, 2 in 23.0% and 3 in 11.3%. Kaplan-Meier curve showed a significant association of multimarker score with probability to suffer a CV event ( Fig.   2A). Using this score in the Cox proportional hazards model, score 2 or 3 were significant predictors of outcome (Table 3). After adjustment by variables statistically significant in the univariate analysis, both score 2 or 3 persisted as independent predictors of CV events (Table 3).
Finally, to assess the clinical usefulness of multimarker score in predicting CV outcomes in CKD patients, we created multivariable regression models with or without multimarker score. The model with conventional CV risk factors included age, sex, hyperlipidemia, hypertension, diabetes, and smoking status as well as eGFR. According to c statistics, the model including the multimarker score has shown a significant improvement from HR (95%CI): 0.71 (0.65-0.78) to 0.79 (95%CI, 0.72-0.85) in accuracy of CV events prediction (p=0.013) (Fig. 2B).

Discussion
In this work, we investigated OPG, OPN and sTWEAK serum levels as predictors of CV outcomes in CKD patients with or without CV risk factors but without any history of CVD. Specifically, we have observed that higher levels of calcification biomarkers OPG or OPN while lower levels of sTWEAK were associated with significantly greater risk of fatal and non-fatal CV events. In addition, we demonstrated that a combination of these biomarkers improves  (14,26). Accordingly, we have demonstrated that OPG levels are negatively associated with eGFR and that high concentrations of OPG can predict CV events in the NEFRONA population. The discrepancy between the potential beneficial effects of OPG and their high levels observed in CKD patients could be explained by a compensatory mechanism to counteract vascular calcium deposition in the arterial wall of CKD patients.
OPN is also expressed and secreted from the vascular wall and bone tissue, plays a role in atherosclerotic plaque development (27) and contributes to kidney damage in mice (28). Furthermore, OPN has been used as a marker However, effects of TWEAK and OPG in vascular calcification should be independent since it has been demonstrated that TWEAK induces macrophage differentiation into osteoclasts in the presence of OPG, indicating a different mechanism of action (29). Confirming data of previous works (19)(20), we observed that sTWEAK levels are independently associated with CV events in CKD patients.
The most important finding of our study is the association of a combination of biomarkers of vascular calcification (OPG and OPN) and inflammation (sTWEAK) with the risk of CV events. This combination showed a good discriminative power, suggesting that marker combinations integrating different biological mechanisms might better stratify CKD patients. Despite the independent association that we observed between multimarker score and 11 risk of fatal and non-fatal CV events, the inclusion of this multimarker score improves clinical risk prediction in our population. Our findings support the potential inclusion of these biomarkers in risk prediction algorithms.
Finally, we want to highlight some limitations of our study for a correct interpretation of the results. Only CKD patients without history of CV events were included in the study. This was a necessary intentional bias because the study was aimed to primary prevention of CV events. A relative low number of fatal and non-fatal CV events was reported during the follow-up, which should limit the statistical power of our analysis.
In conclusion, OPG, OPN and sTWEAK impacted the predictability of cardiovascular outcomes. The information provided by these biomarkers was additive because the risk of developing a CV event increased along with the number of them altered. Thus, a multimarker score including OPG, OPN and sTWEAK concentrations was independently and statistically associated with CV outcomes in CKD patients.           Multivariable Cox analysis included variables that were statistically significant in the univariate analysis. HR, Hazard ratio; 95% CI, 95% confidence interval; Model 1 adjusted also by OPG, OPN or sTWEAK based in their optimal cut-off points. Model 2 adjusted also by number of biomarkers above (OPG and OPN) or below (sTWEAK) of their optimal cut-off points. BMI: Body mass index; SBP: systolic blood pressure; DBP: diastolic blood pressure; LDL: lowdensity lipoprotein; HDL: high-density lipoprotein; eGFR: estimated glomerular filtration rate; hs-CRP: high-sensitivity C-reactive protein.