Clinical Biochemistry 47 (2014) 564–569
Contents lists available at ScienceDirect
Clinical Biochemistry journal homepage: www.elsevier.com/locate/clinbiochem
Factor analysis of risk variables associated with iron status in patients with coronary artery disease Vesna Spasojevic-Kalimanovska a,⁎, Natasa Bogavac-Stanojevic a, Dimitra Kalimanovska-Ostric b, Lidija Memon c, Slavica Spasic a, Jelena Kotur-Stevuljevic a, Zorana Jelic-Ivanovic a a b c
Institute of Medical Biochemistry, Faculty of Pharmacy, University of Belgrade, Belgrade, Serbia Faculty of Medicine, University of Belgrade, Clinical Center of Serbia, Belgrade, Serbia Clinical Chemistry Laboratory, Clinical Center “Bezanijska Kosa”, Belgrade, Serbia
a r t i c l e
i n f o
Article history: Received 27 November 2013 Received in revised form 9 March 2014 Accepted 19 March 2014 Available online 30 March 2014 Keywords: Coronary artery disease Risk factors Iron status parameters Factor analysis
a b s t r a c t Objectives: Epidemiological evidence concerning the role of iron, a lipid peroxidation catalyst, in atherosclerosis and coronary artery disease (CAD) is inconsistent. Design and methods: Exploratory factor analysis was used to examine the potential clustering of variables known to be associated with CAD using data from 188 patients with angiographically-approved disease. The resulting factors were then tested for their association with serum ferritin and soluble transferrin receptor (sTfR) as indicators of body iron status. Results: Factor analysis resulted in a reduction of a variable number from the original 15 to 5 composite clusters. These factors were interpreted as (1) “proatherogenic factor” with positive loadings of TC, LDL-C, apoB and TG; (2) “inflammatory factor” with positive loadings of hsCRP, fibrinogen and MDA; (3) “antiatherogenic factor” with positive loadings of HDL-C and apoA-I; (4) “obesity factor” with positive loadings of weight and waist; and (5) “antioxidative status factor” with positive loadings of SOD and age and negative loading of superoxide anion. “Inflammatory”, “obesity” and “antiatherogenic” factors predicted high ferritin values and the “proatherogenic factor” predicted high sTfR values. We compared the ability of the “proatherogenic factor” with that of a multivariable logistic model that included the “proatherogenic factor” and sTfR values in predicting significant stenosis in patients. The area under the ROC curve was 0.692 vs. 0.821, respectively. Conclusions: “Inflammatory”, “obesity”, “antiatherogenic” and “proatherogenic” factors were associated with increased parameters of body iron status. The measurement of sTfR improves the prediction of CAD based on clustered cardiovascular risk factors. © 2014 The Canadian Society of Clinical Chemists. Published by Elsevier Inc. All rights reserved.
Introduction The most plausible mechanism by which iron may promote the formation of the atherosclerotic plaque lies in its well-known ability to catalyze the production of reactive oxygen species, lipid peroxidation and LDL-oxidation. In 1981 Sullivan proposed the “iron hypothesis”, suggesting that greater levels of stored iron in men and postmenopausal women may explain the higher incidence of heart disease in these groups [1]. Since then, numerous studies have assessed the association Abbreviations: CAD, coronary artery disease; CVD, cardiovascular disease; TC, total cholesterol; HDL-C, high-density lipoprotein cholesterol; LDL-C, low density lipoprotein cholesterol; TG, triglycerides; ApoAI, apolipoprotein AI; ApoB, apolipoprotein B; Lp(a), lipoprotein(a); hsCPR, high sensitivity C-reactive protein; MDA, malondialdehyde; O•− 2 , superoxide anion; SOD, superoxide dismutase; sTfR, soluble transferrin receptors. ⁎ Corresponding author at: University of Belgrade, Faculty of Pharmacy, P.O.BOX 146, 11221 Belgrade, Serbia. Fax: +381 11 3972840. E-mail address: [email protected] (V. Spasojevic-Kalimanovska).
between iron status and the risk of coronary heart disease, but findings have been inconsistent [2–4]. Various laboratory parameters have been used to assess body iron stores and iron availability when attempting to establish a link between iron and coronary artery disease [5]. Among, them ferritin and sTfR are considered to be the most reliable estimates of iron status; ferritin is considered the best measure of body iron stores whereas sTfR is considered to reflect the functional iron compartment. Most in vitro studies have indicated that iron has prooxidant/ proatherogenic properties but such properties have not always been confirmed. Evidence from several prospective epidemiological studies does not support the iron theory [6–8]. Moreover, many studies have failed to find a positive association before and after adjusting for a wide range of cardiovascular risk factors. Increased serum ferritin levels have been shown to represent significant risk factors of myocardial infarction and atherosclerosis owing to iron-mediated oxidative damage [9]. Ferritin is also an acute-phase reactant and its concentration may be increased by myocardial damage
http://dx.doi.org/10.1016/j.clinbiochem.2014.03.014 0009-9120/© 2014 The Canadian Society of Clinical Chemists. Published by Elsevier Inc. All rights reserved.
V. Spasojevic-Kalimanovska et al. / Clinical Biochemistry 47 (2014) 564–569
565
and inflammation [10]. It is possible that ferritin plays a role through other risk factors such as blood pressure and cholesterol [11]. Further studies, particularly investigating interactions/synergies between serum ferritin and other risk factors, are required to establish a role for ferritin in development of CAD. The inconsistency from these studies may be the result of the different parameters that have been utilized for measuring tissue iron stores, differences in samples sizes and recruitment biases. A crucial question that remains to be answered is whether serum ferritin and sTfR are independent atherosclerotic risk factors, or, whether the increased risk attributed to these iron status parameters is actually a result of their complex interactions with other traditionally recognized risk factors. So in this study we evaluated the influence of other biochemical parameters on iron status parameters using factor analysis. Factor analysis reduces a large number of intercorrelated variables to a smaller subset of underlying “independent” variables (factors). The factors represent statistically independent and physiologically distinct phenotypes and may reveal unifying communalities between physiological domains [12]. Although previous studies [9,11] have described the individual contribution of iron status parameters to atherosclerosis and CAD, the potential role of clustering cardiovascular risk factors has been little explored. To investigate the value of iron status parameters in coronary risk assessment in a Serbian population with a high prevalence of atherosclerosis [13], we evaluated the relationship between serum ferritin, sTfR and the presence of CAD. We used exploratory factor analysis to examine potential clustering of variables known to be associated with atherosclerosis and CAD. The resulting factors were then tested for their association with iron status parameters. We also explored whether there was an independent association between iron status parameters and CAD after adjusting for different confounding “clustered factors”. In addition, we tested the accuracy of iron status parameters and clustered variables in CAD detection.
our study protocol, thereby following local biomedical research regulations.
Materials and methods
Statistical analysis
Subjects
Because the distributions of TG, Lp(a), hsCRP and ferritin values were skewed, logarithmic transformation of the data was performed. The results are expressed as arithmetic mean (X) ± standard deviation (SD) for normally-distributed variables and as geometrical mean and 95% confidence interval (CI) for the mean for TG, Lp(a), hsCRP and ferritin. Statistical differences were evaluated according to the Student t-test for continuous variables, whereas proportions were compared using the chi-square test for contingency tables. TG, Lp(a), hsCRP and ferritin values were compared after logarithmic transformation. Factor analysis was conducted using the FACTOR procedure of MedCalc for Windows statistical software Version 9.6.3 (Mariakerke, Belgium) and SPSS 18 statistical software (SPSS Inc, USA). We used principal component analysis, a technique for reducing a number of original variables into fewer summary factors or principal components [18]. The Kaiser–Meyer–Olkin test was used to examine sampling adequacy. Only factors with eigenvalues N1 were extracted for the subsequent orthogonal (varimax) rotation of the factor matrix. Eigenvalues measure the amount of the variation explained by each factor. An eigenvalue greater than 1 indicates that factors account for more variance than accounted by one of the original variables in standardized data. Variables that shared ≥ 25% variance with a summary factor were used for interpretation (corresponding to a loading factor of ≥ 0.50) [12,16]. Factor scores, representing individual subjects' predicted eigenvalues for each factor, were calculated and modeled as independent variables in a multiple logistic regression model in which high ferritin and sTfR concentrations, defined as upper tertile, were the dependent variables. Using multivariate logistic regression analysis we examined the risk for development of significant stenosis in subjects with high ferritin and sTfR concentrations. After that, we calculated the area
The patient group consisted of 188 subjects who were undergoing coronary angiography for suspected coronary artery disease at the Institute of Cardiovascular Disease located in the Clinical Centre of Serbia in Belgrade. All patients with myocardial infarction within the last 6 months, those with unstable angina who had angina pain at rest within one month, or those with a history of prior coronary revascularization were excluded. All patients had a history of stable angina defined on the presence of chest pain that did not change its pattern during the preceding 2 months. Two cardiologists unaware that the patients were enrolled in the study reviewed all the angiograms. Patients were categorized into one of two groups based on the extent of their CAD, as assessed by coronary angiography. One hundred forty-eight patients had significant stenosis (≥50% in one or more coronary arteries), CAD(+) and 40 patients had no stenosis in any artery, CAD(−). All enrolled patients completed a questionnaire that incorporated numerous risk-related issues. For every patient, weight and waist circumference were recorded. We excluded patients with a history of recent clinical infection, concurrent major renal, hepatic or malignant disease, surgery or major trauma during the month prior to study entry. Diabetic patients (patients with a fasting glucose level ≥7.0 mmol/L or patients who were receiving oral hypoglycemic agents or insulin medication) and patients receiving any anti-hyperlipidemic medication were excluded from the study. We excluded individuals with hsCRP ≥10 mg/L, a level considered to be indicative of a clinically relevant inflammatory condition [14]. All patients gave informed consent prior to their enrolment in the study. The study was planned according to ethical guidelines following the Declaration of Helsinki. The institutional review committee approved
Methods Blood samples were obtained from the patients after overnight fasting. Peripheral venous blood was drawn into collection tubes containing ethylene diamine-tetraacetic acid (EDTA), citrate or serum separator gel. Lipid, apolipoprotein and oxidative stress status parameters were determined in EDTA plasma, high sensitivity serum C-reactive protein (hsCRP), ferritin and sTfR in sera and fibrinogen in citrate plasma. All the assays were performed blindly. Most of the biochemical markers were measured using an ILAB 600 analyzer (Instrumentation Laboratory, Milan, Italy). Total cholesterol (TC) and triglycerides (TG) were assayed using routine enzymatic methods. HDL-C was measured using the same enzymatic method after precipitation of the plasma with phosphotungstic acid in the presence of magnesium ions and LDL-C was calculated using the Friedewald formula. Apolipoprotein A-I (apoA-I), apolipoprotein B (apoB), ferritin and sTfR were measured using immunoturbidimetry (Dialab, Vienna, Austria). The concentrations of serum hsCRP were also measured using immunoturbidimetry (BIOKIT, Barcelona, Spain). Fibrinogen was measured by the Clauss method employing ACL 200 Instrumentation laboratory and original reagents. Plasma malondialdehyde (MDA) concentration was measured using the thiobarbituric acidreactive substances (TBARS) assay, previously described by Girotti [15]. In our hands the intra-assay CV was 4.8% and the inter-assay CV was 7.2%. The rate of nitroblue tetrazolium reduction was used to measure the level of O2•−, as described by Auclair and Voisin [16]. The intra-assay CV was 5.6% and the inter-assay CV was 9.5%. Plasma SOD activities were measured according to the previously published method by Misra and Fridovich [17]. One unit of SOD activity is defined as the activity that inhibits the auto-oxidation of adrenalin by 50%. The intra-assay CV was 6.3% and the inter-assay CV was 9.2%.
566
V. Spasojevic-Kalimanovska et al. / Clinical Biochemistry 47 (2014) 564–569
Table 1 Basic clinical and laboratory characteristics of the patients. Variable
Patients (N = 188)
CAD(−) patients (N = 40)
CAD(+) patients (N = 148)
Female gender (n [%]) Age (years) Weight (kg) Waist (cm) TC (mmol/L) TG (mmol/L) HDL-C (mmol/L) LDL-C (mmol/L) ApoAI (g/L) ApoB (g/L) hsCRP (mg/L) Fibrinogen (g/L) MDA (μmol/L) SOD (U/L) O•− 2 (μmol/min/L) Ferritin (μg/L), males
67 (35.6)
21 (52.5)
46 (31.1)*
56.0 ± 9.1 81.1 ± 2.7 97.7 ± 11.9 5.39 ± 1.11 2.06 (1.94–2.19) 0.82 ± 0.21 3.71 ± 1.08 1.38 ± 0.29 1.39 ± 0.36 4.31 (3.64–5.10) 5.61 ± 1.62 3.51 ± 0.76 108.20 ± 40.90 162.50 ± 45.50 113.24 (97.50–131.83) 43.85 (34.51–55.72) 1.44 ± 0.37
53.6 ± 10.2 81.4 ± 15.9 166.2 ± 9.0 4.94 ± 1.13 1.92 (1.65–2.24) 0.88 ± 0.22 3.31 ± 1.09 1.46 ± 0.26 1.24 ± 0.33 3.49 (2.44–4.99) 5.31 ± 1.61 3.47 ± 0.66 113.03 ± 42.89 154.94 ± 40.86 132.43 (82.79–211.84) 46.77 (29.51–73.96) 1.36 ± 0.36
56.6 ± 8.6 81.01 ± 11.9 171.4 ± 9.3** 5.52 ± 1.08** 2.11 (1.98–2.25) 0.80 ± 0.20* 3.81 ± 1.06* 1.35 ± 0.29* 1.42 ± 0.36** 4.55 (3.74–5.52) 5.70 ± 1.61 3.52 ± 0.79 107.02 ± 40.58 164.71 ± 46.80 110.92 (94.19–130.62) 42.46 (31.62–56.75) 1.45 ± 0.37
Ferritin (μg/L), females sTfR (mg/L)
Data are shown as mean ± SD for normally distributed continuous variables, geometrical mean (95% CI for the mean) for continuous variables with skewed distribution and absolute (relative) frequencies for categorical variables. ⁎ Significantly different from the value in the CAD(−) group, with P b 0.05. ⁎⁎ Significantly different from the value in the CAD(−) group, with P b 0.01. ⁎⁎⁎ Significantly different from the value in the CAD(−) group, with P b 0.001.
under the receiver–operator characteristic curve (AUC) for independent predictors of significant stenosis as well as for a multivariable logistic model that included independent predictors of significant stenosis and ferritin or sTfR values. The minimal level of statistical significance was set at P b 0.05. We estimated that a sample of 188 subjects with a two-sided significance level of 0.05 would be required to provide statistical power of 83%. We used power and sample size calculations for generalized regression models with covariate measurement error [19].
Results Demographic characteristics and laboratory data of all CAD patients, CAD(+) and CAD(−) patients investigated in the study are shown in Table 1. The CAD(+) patients had significantly higher TC, LDL-C and apoB concentrations and lower prevalence of female. HDL-C and apoAI concentrations were significantly lower in CAD(+) than in CAD(−) patients. Because the difference in ferritin levels, between genders was
significant P b 0.001, we compared ferritin levels between CAD(−) and CAD(+) groups separately for both genders. No significant difference was found between CAD(−) and CAD(+) for male P = 0.415 and for female P = 0.707 patients. In order to reduce the number of variables factor analysis was performed among all subjects. The Kaiser–Meyer–Olkin measure of sampling adequacy was 0.655. Table 2 shows that factor analysis resulted in a reduction of variable number from the original 15 to 5 composite clusters. All five factors had an eigenvalue ≥1 and they explained 73.95% of the total variance (22.40% factor 1, 15.32% factor 2, 14.06% factor 3, 12.87 % factor 4 and 9.29% factor 5). These factors were interpreted as (1) “proatherogenic factor” with positive loadings of TC, LDL-C, apoB and TG; (2) “inflammatory factor” with positive loadings of hsCRP, fibrinogen and MDA; (3) “antiatherogenic factor” with positive loadings of HDL-C and apoAI; (4) “obesity factor” with positive loadings of weight and waist; and (5) “antioxidative status factor” with positive loadings of SOD and age and negative loading of O•− 2 . The association between factor scores and the presence of high ferritin and sTfR values were assessed using logistic regression analysis. Serum ferritin and sTfR concentrations were categorized into tertiles based on the cut-off points of the entire distribution. The subjects with ferritin and sTfR concentrations between the two cut-off values were excluded from analysis. Because the difference in ferritin levels between genders was significant we defined tertiles separately for females and males. Males with ferritin concentrations ≤85.67 μg/L were coded 0 and with ferritin concentrations ≥169.18 μg/L were coded 1. Similarly, females with ferritin concentrations ≤34.00 μg/L were coded 0 and with ferritin concentrations ≥70.60 μg/L were coded 1. sTfR concentrations were coded as 0 for the lowest tertile (≤1.15 mg/L) and 1 for the highest tertile (≥1.69 mg/L). In separate univariable analysis “inflammatory”, “obesity” and “antiatherogenic” factors predicted a high ferritin value and “proatherogenic factor” predicted a high sTfR value (Table 3). Using univariable and multivariable logistic regression analysis we examined the risk for development of significant CAD (≥ 50%) in patients with high ferritin and sTfR values. Ferritin and sTfR were used as independent variables and CAD variable as dependent variable. The CAD(+) subjects were coded 1 and CAD(−) subjects were coded 0. Table 4 indicates that high sTfR levels (≥1.69 mg/L) increased the risk for significant stenosis whereas a high ferritin value had no effect on risk in the univariate model. We performed additional analysis to determine whether the “proatherogenic factor” was responsible for the association between high sTfR levels and significant CAD risk. Adjustment for the “proatherogenic factor” eliminated the risk, suggesting that this factor is a strong confounder in the model. We compared the ability of the “proatherogenic factor” and sTfR with that of a multivariable logistic model that included the “proatherogenic factor” and sTfR values in predicting the presence of significant stenosis
Table 2 Results of factor analysis of proatherogenic, antiatherogenic, inflammatory, obesity and antioxidative variables. Factors
TC (mmol/L) LDL-C (mmol/L) ApoB (g/L) TG (mmol/L) Fibrinogen (g/L) hsCRP (mg/L) MDA (μmol/L) HDL-C (mmol/L) ApoAI (g/L) Waist (cm) Weight (kg) SOD (U/L) Age (years) O•− 2 (μmol/min/L) a
Proatherogenic
Inflammatory
Antiatherogenic
Obesity
Antioxidative
0.966a 0.951a 0.921a 0.546a −0.027 0.158 −0.050 −0.044 0.009 −0.005 −0.032 −0.019 0.032 0.345
−0.014 0.048 0.076 −0.082 0.831a 0.830a 0.723a −0.136 −0.145 0.090 0.022 −0.081 0.315 0.275
0.096 0.002 −0.063 −0.413 −0.037 −0.112 −0.101 0.884a 0.862a −0.133 −0.322 −0.186 0.285 0.012
−0.063 −0.041 0.005 0.158 −0.011 −0.033 0.184 −0.253 −0.180 0.923a 0.843a −0.185 0.193 −0.071
−0.067 −0.052 0.027 −0.226 −0.114 −0.091 0.145 0.029 −0.125 −0.004 −0.030 0.799a 0.544a −0.502a
Factor loadings higher than 0.50 or lower than −0.50.
V. Spasojevic-Kalimanovska et al. / Clinical Biochemistry 47 (2014) 564–569 Table 3 Logistic regression analysis for factor scores with ferritin and sTfR. Independent variable
Ferritin OR (95% CI)
sTfR OR (95% CI)
Proatherogenic factor
0.856 (0.533–1.373) P = 0.519 2.479 (1.266–4.854) P = 0.008 0.442 (0.237–0.823) P = 0.010 2.016 (1.146–3.546) P = 0.015 0.986 (0.609–1.596) P = 0.955
3.455 (1.582–7.545) P = 0.002 1.243 (0.787–1.963) P = 0.350 0.728 (0.439–1.206) P = 0.218 0.695 (0.416–1.162) P = 0.165 1.404 (0.817–2.415) P = 0.220
Inflammatory factor Antiatherogenic factor Obesity factor Antioxidative factor
in the study population. The AUC for the “proatherogenic factor” was 0.692 (0.578–0.807) and AUC for sTfR was 0.649 (0.514–0.784), which were lower than that of the multivariable logistic model [c-statistics 0.821 (0.705–0.937)]. Discussion In evaluating the relationship between iron status and risk of CAD, it is important to rule out non-iron-related factors that can influence the biochemical measures of iron status. In the present study, we used factor analysis to investigate the underlying correlation structure of variables that were associated with iron status parameters in patients with CAD. To our knowledge, such a factor analysis has not been used in previous studies of association between iron status parameters and risk for CAD. Five factors emerged in the present analysis (Table 2). The first factor accounting for 22.40% of variance comprised TC, LDL-C, apoB and TG, a proatherogenic lipid profile and this factor was considered the “proatherogenic factor”. The second factor was defined by an increase of hsCRP, fibrinogen and MDA. Clustering of hsCRP and fibrinogen as parameters of inflammation and MDA as a marker of oxidative status is consistent with the hypothesis of “oxidative response to inflammation” being the generation of oxidative species as a consequence of the inflammatory process [20]. Our previous study [21] suggested that the relationship between oxidative stress parameters and inflammatory species was due to their strong mutual involvement in atherosclerosis development that leads to CAD progression. The third factor was defined by the increase in HDL-C and apoAI and was considered to represent antiatherogenic parameters. “Obesity factor” was defined with positive loadings of weight and waist circumference. Obesity is a crucial risk factor for atherosclerosis and CAD and substitution of body mass index and waist: hip ratio for weight and waist, respectively, resulted in better factor loading. The fifth “antioxidative status factor” clustered parameters of oxidative stress (O•− 2 ) and antioxidative defense SOD. Although age is not a specific oxidative/antioxidative marker, the results of factor analysis indicate that the age has communality with SOD and O•− 2 . After clustering all variables into five factors, they were analyzed as confounders of iron status. The parameters of iron status examined in our study were: serum ferritin and sTfR. Predictors of high sTfR values were atherogenic lipid parameters LDL-C, apoB, TG and TC (“proatherogenic factor”) in our study. If we hypothesize that high sTfR concentrations reflect an elevated demand for iron in macrophage cells in the atherosclerotic lesion it is logical that high values of proatherogenic lipid parameters are predictors of high sTfR concentration [22] and this finding supports the iron hypothesis. In this study inflammatory factors hsCRP and fibrinogen were strongly associated with high serum ferritin. There is evidence from a recent study [23] that inflammation contributes to the pathogenesis of cardiovascular disease and many prospective studies have shown an association between elevated levels of CRP and the risk of cardiovascular disease. CRP and ferritin are acute-phase reactants, which are increased with inflammation and infection. Serum ferritin increases with inflammation
567
because ferritin synthesis is up-regulated by proinflammatory cytokines. Consequently, the association between serum ferritin and CAD could be confounded by inflammation and the ferritin levels in patients' sera may be a surrogate measure of inflammation [23]. This may explain the failure of some previous studies to find an association between ferritin and CAD. Ferritin as a marker of inflammation rather than iron stores may also explain the apparent synergistic association between ferritin and cholesterol levels and the risk of cardiovascular disease reported in the Brunneck study [4]. Alternatively, the association between ferritin and CRP may be explained by infection. The rise in ferritin levels and decrease in iron levels with acute infection are an important defense against proliferation of microorganisms. This is suggestive evidence that the infection may contribute to the pathogenesis of atherosclerosis [24]. In our study we eliminated patients with higher levels of hsCRP (≥10 mg/L) which can indicate acute infection and the association obtained between inflammatory clustered parameters and ferritin really supports the iron hypothesis. In the current study, the obtained association between high ferritin and obesity, measured by anthropometric surrogates is in agreement with some previous studies [23,25]. In the study of Jehn et al. [25] serum ferritin was significantly associated with waist measurement in both men and women and an increased risk of cardiovascular disease within persons with fat distribution in the abdominal region. Adipose tissue is a major regulator of chronic, low grade inflammation, characterized by abnormal cytokine production, increased acute phase reactants, activation of inflammatory signaling pathways and up regulation of ferritin synthesis. Therefore, the high ferritin levels observed in abdominal obesity could better reflect an inflammatory phenomenon as part of an acute-phase reaction than an increase of the iron stores. In our study, logistic regression analysis did not suggest associations of “antioxidative factor” and parameters of iron status. Ferritin expression is regulated by oxidative stress [26] but several studies have shown that ferritin protects cells from oxidative stress by maintaining iron homeostasis [27] so it is still uncertain whether ferritin is an antioxidative or a pro-oxidative parameter. Although the lipid oxidation theory of atherosclerosis is based on the iron-catalyzed oxidation of LDL, our results suggest that serum ferritin is an inappropriate marker of free (redox-active) iron. There are two possible explanations for this observation. First, the applied statistical method, factor analysis grouped the parameters of oxidative stress O•− 2 , and an enzyme of antioxidative defense, SOD, in one cluster. Second, ferritin does not reflect the free (redox-active) iron. Van Tits et al. [28] measured non-transferrinbound iron (NTBI), which represents the metal catalytically-active form of iron. They verified the proposed atherosclerotic mechanism of iron by measuring serum levels of NTBI in relation to markers of LDL oxidation and markers of inflammation. In our research we were not able to measure NTBI as a marker of iron status. In the second part of our study we evaluated the predictive power of iron status parameters in detecting CAD. We hypothesized that high serum ferritin and sTfR levels would be independently associated with CAD. To test those hypotheses, we measured ferritin and sTfR concentrations in patients with clinically significant CAD (stenosis ≥ 50%). assessed by coronary angiography and in CAD-free subjects (stenosis ≤ 50%). In the current study, we demonstrated that high serum sTfR Table 4 Logistic regression analysis of the associations of high ferritin and sTfR with significant stenosis. Variables
OR
95% CI
P
Univariable logistic model Ferritin sTfR
0.568 3.643
0.231–1.396 1.178–11.265
0.218 0.025
Multivariable logistic model Proatherogenic factor sTfR
2.883 5.139
1.103–7.540 0.953–27.697
0.031 0.057
568
V. Spasojevic-Kalimanovska et al. / Clinical Biochemistry 47 (2014) 564–569
was associated with the presence of significant CAD (≥ 50%). High serum ferritin values were not able to predict CAD and these findings are partially in agreement with the observation of Braun et al. [22], who investigated the association between ferritin and sTfR with the extension of CAD. They found a significant association between sTfR values and the extension of CAD. Mechanisms that can explain the association between sTfR and the extension of CAD are largely unknown. Experimental studies reported [29] an increased need for iron supply in atherosclerotic lesions as cellular replication is a constant feature of atherogenesis. Also, the amount of iron has been reported to be higher than in a normal arterial wall [30]. Furthermore, it has been demonstrated [31] that elevated iron levels in macrophages increase rather than decrease TfR mRNA. The increased number of membrane-associated TfR may increase the capacity of cells in atherosclerotic lesions to take up iron from circulation. An excess iron amount due to a higher density of TfR in cell membranes promotes the formation of toxic oxygen radicals and lipid peroxidation, potentiating the progression of atherosclerotic lesions in arterial walls. In the same study, Braun et al. [22] found that in the group of patients with CAD, ferritin concentrations were higher in patients with acute coronary syndromes than in those with stable angina. This study [22] supports the concept that increased ferritin levels may be a marker of instability in patients with existing CAD. In our study ferritin was not a predictor of significant CAD and this could be explained by the clinical status of our patients as they had stable angina. Multivariable logistic regression indicated the “proatherogenic factor” as a strong confounder in sTfR prediction of significant CAD. The observed association was dependent on the “proatherogenic factor”, suggesting that the increased sTfR concentration is related to coexisting changes in the classical risk factor defined as proatherogenic. To our knowledge, this is the first study to explore whether the measurement of serum sTfR could predict significant CAD independently of clustered cardiovascular risk factors. In the third part of our study, we employed C statistics to assess the clinical usefulness of laboratory markers. In our logistic regression model, the “proatherogenic factor” was an independent risk factor for CAD prediction and the AUC for this model was 0.692. According to Hosmer–Lemeshow's criteria [32] it was considered to be poor, but in the multivariable logistic model that included the “proatherogenic factor” and sTfR, the value was 0.821 considered to be excellent. The level of sTfR as a single parameter was not particularly helpful in discriminating CAD(+) patients from CAD(−) patients but as an additional CAD marker had a significant diagnostic value. The obtained results in the present study partly support the iron hypothesis of atherosclerosis. Although serum ferritin concentration is considered the best indicator for body iron stores, in the present study ferritin was not a predictor of significant coronary artery stenosis. Serum sTfR concentration that reflects the functional iron compartment is a better predictor of CAD but with a reasonable accuracy and more important as an additional CAD marker. Conclusions Factor analysis is an excellent tool for creating a summary risk measure and also defines relationships between variables related to underlying pathophysiologic mechanisms. In our study applied factor analysis identified five underlying factors among a group of 15 variables that have been associated with atherosclerosis and CAD. These data showed that factors interpreted as “inflammatory”, “obesity”, “proatherogenic” and “antiatherogenic” were associated with increased parameters of body iron status, partly confirming previous epidemiological studies. Our results showed that body iron status assessed by measurement of sTfR concentrations could be useful in the prediction of CAD. Despite this, the low accuracy of sTfR determination suggests that a single marker cannot be an adequate parameter for CAD screening. Satisfactory discriminative prediction abilities are achieved only in combination
with the proatherogenic risk factor. These findings highlight the importance of a global approach to assessing risk and the need for studies that elucidate how these various risk domains interact over time to create clinical disease. Further studies with a greater number of patients are required to support our results and to investigate the nature and significance of the relationship among serum iron indices, tissue iron concentrations and CAD. Conflict of interest The authors have no conflict of interest to declare. Acknowledgment This work was supported by a grant from the Ministry of Education, Science and Technological Development, Republic of Serbia (Project No. 175035) and by the European Cooperation in Science and Technology (COST) BM0904 Action. References [1] Sullivan JL. Iron and sex difference in heart disease risk. Lancet 1981;1:1293–4. [2] Toumainen TP, Punnonen K, Nyyssonen K, Salonen JT. Association between body iron stores and the risk of acute myocardial infarction in man. Circulation 1998;97:1461–6. [3] Kiechl S, Aichner F, Gerstenbrand F, Egger G, Mair A, Rungger G, et al. Body iron stores and presence of carotid atherosclerosis. Results from the Bruneck study. Atherioscler Thromb 1994;14:1625–30. [4] Kiechl S, Willeit J, Egger G, Poewe W, Oberhollenzer F. Body iron stores and the risk of carotid atherosclerosis. Prospective results from Bruneck study. Circulation 1997;96:3300–7. [5] Danesh J, Appleby P. Coronary heart disease and iron status. Meta-analyses of prospective studies. Circulation 1999;99:852–4. [6] Liao Y, Cooper RS, McGee DL. Iron status and coronary heart disease: negative findings from the NHANES I epidemiologic follow-up study. Am J Epidemiol 1994;139:704–12. [7] Moore M, Folsom AR, Barnes RW, Eckfeldt JH. No association between serum ferritin and asymptomatic carotid atherosclerosis. The Atherosclerosis Risk in Communities (ARIC) study. Am J Epidemiol 1995;141:719–23. [8] Bozzini C, Girelli D, Tinazzi E, et al. Biochemical and genetic markers of iron status and the risk of coronary artery disease: an angiography-based study. Clin Chem 2002;48:622–8. [9] Ref DW. Ferritin as a source of iron oxidative damage. Free Radic Biol 1992;12:417–27. [10] Gabay C, Kuushener I. Acute-phase proteins and other systemic responses to inflammation. N Engl J Med 1999;340:448–54. [11] Knuiman MW, Divitini ML, Olynyk JK, Cullen DJ, Bartholomew HC. Serum ferritin and cardiovascular disease: a 17-year follow-up study in Busselton, Western Australia. Am J Epidemiol 2003;158:144–9. [12] Cureton EE, D’Agostino RB. Factor analysis, an applied approach; 1983 [N.J. Hillsdale, L. Erlbaum associates]. [13] Causes of deaths. In: Vukomirovic D, editor. Statistical yearbook of Serbia 2005. Belgrade: SCG: Statistical Office of the Republic of Serbia; 2005. p. 90. [14] Pearson TA, Mensah GA, Alexander RW, Anderson JL, Cannon III RO, Criqui M, et al. Markers of inflammation and cardiovascular disease: application to clinical and public health practice: a statement for healthcare professionals from the Centers for Disease Control and Prevention and the American Heart Association. Circulation 2003;107:499–511. [15] Girotti MJ, Khan N, McLellan BA. Early measurement of systematic lipid peroxidation products in plasma of major blunt trauma patients. J Trauma 1991;31:32–5. [16] Auclair C, Voisin E. Nitroblue tetrazolium reduction. In: Greenwald RA, editor. CRC handbook of methods for oxygen radical research. Boca Raton, Florida: CRC Press; 1985. p. 123–32. [17] Misra HP, Fridovich I. The role of superoxide anion in the autooxidation of epinephrine and a simple assay for superoxide dismutase. J Biol Chem 1972;247:3170–5. [18] Stevens J. Applied multivariate statistics for the social sciences. 3rd ed. N.J. Mahwah: Lawrence Erlbaum Associates; 1996. [19] Tosteson T, Buzas J, Demidenko E, Karagas M. Power and sample size calculations for generalized regression models with covariate measurement error. Stat Med 2003;22:7. [20] Libby P, Ridker PM, Maseri A. Inflammation and atherosclerosis. Circulation 2002;105:1135–43. [21] Memon L, Spasojevic-Kalimanovska V, Bogavac-Stanojevic N, KalimanovskaOstric D, Jelic-Ivanovic Z, Spasic S, et al. Association of C-reactive protein with the presence and extent of angiographically verified coronary artery disease. Tohoku J Exp Med 2006;209:197–206. [22] Braun S, Ndrepepa G, von Beckerath N, Vogt W, Schomig A, Kastrati A. Value of serum ferritin and soluble transferrin receptor for prediction of coronary artery disease and its clinical presentations. Atherosclerosis 2004;174:105–10. [23] Williams MJA, Poulton R, Williams S. Relationship of serum ferritin with cardiovascular risk factors and inflammation in young men and women. Atherosclerosis 2002;165:179–84.
V. Spasojevic-Kalimanovska et al. / Clinical Biochemistry 47 (2014) 564–569 [24] Leinonen M, Saikku P. Evidence for infectious agents in cardiovascular disease and atherosclerosis. Lancet Infect Dis 2002;2:11–7. [25] Jehn M, Clark JM, Gaullar E. Serum ferritin and the risk of the metabolic syndrome in U.S. adults. Diabetes Care 2004;27:2422–8. [26] Cairo G, Tacchini L, Pogliaghi G, Anzon E, Tomasi A, Bernelli-Zazzero A. Introduction of ferritin synthesis by oxidative stress. Transcriptional and posttranscriptional regulation by expansion of the “free” iron pool. J Biol Chem 1995;270:700–3. [27] You SA, Wand Q. Ferritin in atherosclerosis. CCA 2005;357:1–16. [28] van Tits LJH, Jacobs EMG, Swinkels DW, Lemmers HLM, van der Vleuten GM, de Graaf J, et al. Non-transferrin-bound iron is associated with plasma level of soluble
[29] [30] [31]
[32]
569
intercellular adhesion molecule-1 not with in vivo low-density lipoprotein oxidation. Atherosclerosis 2007;194:272–8. Shwartz SM, Ross R. Cellular proliferation in atherosclerosis and hypertension. Prog Cardiovasc Dis 1984;26:355–72. Sullivan JL. Iron in arterial plaque: a modifiable risk factor for atherosclerosis. Biochim Biophys Acta 2009;1790:1014–20. Testa U, Petrini M, Quaranta MT, Pelosi-Testa E, Mastroberardino G, Camagna, et al. Iron up-modulates the expression of transferin receptors during monocyte–macrophage maturation. J Biol Chem 1989;264:13181–7. Hosmer DW, Lemeshow S. Assessing the fit of the model, applied logistic regression. NY: John Wiley and Sons; 2000 160–4.
Contents lists available at ScienceDirect
Clinical Biochemistry journal homepage: www.elsevier.com/locate/clinbiochem
Factor analysis of risk variables associated with iron status in patients with coronary artery disease Vesna Spasojevic-Kalimanovska a,⁎, Natasa Bogavac-Stanojevic a, Dimitra Kalimanovska-Ostric b, Lidija Memon c, Slavica Spasic a, Jelena Kotur-Stevuljevic a, Zorana Jelic-Ivanovic a a b c
Institute of Medical Biochemistry, Faculty of Pharmacy, University of Belgrade, Belgrade, Serbia Faculty of Medicine, University of Belgrade, Clinical Center of Serbia, Belgrade, Serbia Clinical Chemistry Laboratory, Clinical Center “Bezanijska Kosa”, Belgrade, Serbia
a r t i c l e
i n f o
Article history: Received 27 November 2013 Received in revised form 9 March 2014 Accepted 19 March 2014 Available online 30 March 2014 Keywords: Coronary artery disease Risk factors Iron status parameters Factor analysis
a b s t r a c t Objectives: Epidemiological evidence concerning the role of iron, a lipid peroxidation catalyst, in atherosclerosis and coronary artery disease (CAD) is inconsistent. Design and methods: Exploratory factor analysis was used to examine the potential clustering of variables known to be associated with CAD using data from 188 patients with angiographically-approved disease. The resulting factors were then tested for their association with serum ferritin and soluble transferrin receptor (sTfR) as indicators of body iron status. Results: Factor analysis resulted in a reduction of a variable number from the original 15 to 5 composite clusters. These factors were interpreted as (1) “proatherogenic factor” with positive loadings of TC, LDL-C, apoB and TG; (2) “inflammatory factor” with positive loadings of hsCRP, fibrinogen and MDA; (3) “antiatherogenic factor” with positive loadings of HDL-C and apoA-I; (4) “obesity factor” with positive loadings of weight and waist; and (5) “antioxidative status factor” with positive loadings of SOD and age and negative loading of superoxide anion. “Inflammatory”, “obesity” and “antiatherogenic” factors predicted high ferritin values and the “proatherogenic factor” predicted high sTfR values. We compared the ability of the “proatherogenic factor” with that of a multivariable logistic model that included the “proatherogenic factor” and sTfR values in predicting significant stenosis in patients. The area under the ROC curve was 0.692 vs. 0.821, respectively. Conclusions: “Inflammatory”, “obesity”, “antiatherogenic” and “proatherogenic” factors were associated with increased parameters of body iron status. The measurement of sTfR improves the prediction of CAD based on clustered cardiovascular risk factors. © 2014 The Canadian Society of Clinical Chemists. Published by Elsevier Inc. All rights reserved.
Introduction The most plausible mechanism by which iron may promote the formation of the atherosclerotic plaque lies in its well-known ability to catalyze the production of reactive oxygen species, lipid peroxidation and LDL-oxidation. In 1981 Sullivan proposed the “iron hypothesis”, suggesting that greater levels of stored iron in men and postmenopausal women may explain the higher incidence of heart disease in these groups [1]. Since then, numerous studies have assessed the association Abbreviations: CAD, coronary artery disease; CVD, cardiovascular disease; TC, total cholesterol; HDL-C, high-density lipoprotein cholesterol; LDL-C, low density lipoprotein cholesterol; TG, triglycerides; ApoAI, apolipoprotein AI; ApoB, apolipoprotein B; Lp(a), lipoprotein(a); hsCPR, high sensitivity C-reactive protein; MDA, malondialdehyde; O•− 2 , superoxide anion; SOD, superoxide dismutase; sTfR, soluble transferrin receptors. ⁎ Corresponding author at: University of Belgrade, Faculty of Pharmacy, P.O.BOX 146, 11221 Belgrade, Serbia. Fax: +381 11 3972840. E-mail address: [email protected] (V. Spasojevic-Kalimanovska).
between iron status and the risk of coronary heart disease, but findings have been inconsistent [2–4]. Various laboratory parameters have been used to assess body iron stores and iron availability when attempting to establish a link between iron and coronary artery disease [5]. Among, them ferritin and sTfR are considered to be the most reliable estimates of iron status; ferritin is considered the best measure of body iron stores whereas sTfR is considered to reflect the functional iron compartment. Most in vitro studies have indicated that iron has prooxidant/ proatherogenic properties but such properties have not always been confirmed. Evidence from several prospective epidemiological studies does not support the iron theory [6–8]. Moreover, many studies have failed to find a positive association before and after adjusting for a wide range of cardiovascular risk factors. Increased serum ferritin levels have been shown to represent significant risk factors of myocardial infarction and atherosclerosis owing to iron-mediated oxidative damage [9]. Ferritin is also an acute-phase reactant and its concentration may be increased by myocardial damage
http://dx.doi.org/10.1016/j.clinbiochem.2014.03.014 0009-9120/© 2014 The Canadian Society of Clinical Chemists. Published by Elsevier Inc. All rights reserved.
V. Spasojevic-Kalimanovska et al. / Clinical Biochemistry 47 (2014) 564–569
565
and inflammation [10]. It is possible that ferritin plays a role through other risk factors such as blood pressure and cholesterol [11]. Further studies, particularly investigating interactions/synergies between serum ferritin and other risk factors, are required to establish a role for ferritin in development of CAD. The inconsistency from these studies may be the result of the different parameters that have been utilized for measuring tissue iron stores, differences in samples sizes and recruitment biases. A crucial question that remains to be answered is whether serum ferritin and sTfR are independent atherosclerotic risk factors, or, whether the increased risk attributed to these iron status parameters is actually a result of their complex interactions with other traditionally recognized risk factors. So in this study we evaluated the influence of other biochemical parameters on iron status parameters using factor analysis. Factor analysis reduces a large number of intercorrelated variables to a smaller subset of underlying “independent” variables (factors). The factors represent statistically independent and physiologically distinct phenotypes and may reveal unifying communalities between physiological domains [12]. Although previous studies [9,11] have described the individual contribution of iron status parameters to atherosclerosis and CAD, the potential role of clustering cardiovascular risk factors has been little explored. To investigate the value of iron status parameters in coronary risk assessment in a Serbian population with a high prevalence of atherosclerosis [13], we evaluated the relationship between serum ferritin, sTfR and the presence of CAD. We used exploratory factor analysis to examine potential clustering of variables known to be associated with atherosclerosis and CAD. The resulting factors were then tested for their association with iron status parameters. We also explored whether there was an independent association between iron status parameters and CAD after adjusting for different confounding “clustered factors”. In addition, we tested the accuracy of iron status parameters and clustered variables in CAD detection.
our study protocol, thereby following local biomedical research regulations.
Materials and methods
Statistical analysis
Subjects
Because the distributions of TG, Lp(a), hsCRP and ferritin values were skewed, logarithmic transformation of the data was performed. The results are expressed as arithmetic mean (X) ± standard deviation (SD) for normally-distributed variables and as geometrical mean and 95% confidence interval (CI) for the mean for TG, Lp(a), hsCRP and ferritin. Statistical differences were evaluated according to the Student t-test for continuous variables, whereas proportions were compared using the chi-square test for contingency tables. TG, Lp(a), hsCRP and ferritin values were compared after logarithmic transformation. Factor analysis was conducted using the FACTOR procedure of MedCalc for Windows statistical software Version 9.6.3 (Mariakerke, Belgium) and SPSS 18 statistical software (SPSS Inc, USA). We used principal component analysis, a technique for reducing a number of original variables into fewer summary factors or principal components [18]. The Kaiser–Meyer–Olkin test was used to examine sampling adequacy. Only factors with eigenvalues N1 were extracted for the subsequent orthogonal (varimax) rotation of the factor matrix. Eigenvalues measure the amount of the variation explained by each factor. An eigenvalue greater than 1 indicates that factors account for more variance than accounted by one of the original variables in standardized data. Variables that shared ≥ 25% variance with a summary factor were used for interpretation (corresponding to a loading factor of ≥ 0.50) [12,16]. Factor scores, representing individual subjects' predicted eigenvalues for each factor, were calculated and modeled as independent variables in a multiple logistic regression model in which high ferritin and sTfR concentrations, defined as upper tertile, were the dependent variables. Using multivariate logistic regression analysis we examined the risk for development of significant stenosis in subjects with high ferritin and sTfR concentrations. After that, we calculated the area
The patient group consisted of 188 subjects who were undergoing coronary angiography for suspected coronary artery disease at the Institute of Cardiovascular Disease located in the Clinical Centre of Serbia in Belgrade. All patients with myocardial infarction within the last 6 months, those with unstable angina who had angina pain at rest within one month, or those with a history of prior coronary revascularization were excluded. All patients had a history of stable angina defined on the presence of chest pain that did not change its pattern during the preceding 2 months. Two cardiologists unaware that the patients were enrolled in the study reviewed all the angiograms. Patients were categorized into one of two groups based on the extent of their CAD, as assessed by coronary angiography. One hundred forty-eight patients had significant stenosis (≥50% in one or more coronary arteries), CAD(+) and 40 patients had no stenosis in any artery, CAD(−). All enrolled patients completed a questionnaire that incorporated numerous risk-related issues. For every patient, weight and waist circumference were recorded. We excluded patients with a history of recent clinical infection, concurrent major renal, hepatic or malignant disease, surgery or major trauma during the month prior to study entry. Diabetic patients (patients with a fasting glucose level ≥7.0 mmol/L or patients who were receiving oral hypoglycemic agents or insulin medication) and patients receiving any anti-hyperlipidemic medication were excluded from the study. We excluded individuals with hsCRP ≥10 mg/L, a level considered to be indicative of a clinically relevant inflammatory condition [14]. All patients gave informed consent prior to their enrolment in the study. The study was planned according to ethical guidelines following the Declaration of Helsinki. The institutional review committee approved
Methods Blood samples were obtained from the patients after overnight fasting. Peripheral venous blood was drawn into collection tubes containing ethylene diamine-tetraacetic acid (EDTA), citrate or serum separator gel. Lipid, apolipoprotein and oxidative stress status parameters were determined in EDTA plasma, high sensitivity serum C-reactive protein (hsCRP), ferritin and sTfR in sera and fibrinogen in citrate plasma. All the assays were performed blindly. Most of the biochemical markers were measured using an ILAB 600 analyzer (Instrumentation Laboratory, Milan, Italy). Total cholesterol (TC) and triglycerides (TG) were assayed using routine enzymatic methods. HDL-C was measured using the same enzymatic method after precipitation of the plasma with phosphotungstic acid in the presence of magnesium ions and LDL-C was calculated using the Friedewald formula. Apolipoprotein A-I (apoA-I), apolipoprotein B (apoB), ferritin and sTfR were measured using immunoturbidimetry (Dialab, Vienna, Austria). The concentrations of serum hsCRP were also measured using immunoturbidimetry (BIOKIT, Barcelona, Spain). Fibrinogen was measured by the Clauss method employing ACL 200 Instrumentation laboratory and original reagents. Plasma malondialdehyde (MDA) concentration was measured using the thiobarbituric acidreactive substances (TBARS) assay, previously described by Girotti [15]. In our hands the intra-assay CV was 4.8% and the inter-assay CV was 7.2%. The rate of nitroblue tetrazolium reduction was used to measure the level of O2•−, as described by Auclair and Voisin [16]. The intra-assay CV was 5.6% and the inter-assay CV was 9.5%. Plasma SOD activities were measured according to the previously published method by Misra and Fridovich [17]. One unit of SOD activity is defined as the activity that inhibits the auto-oxidation of adrenalin by 50%. The intra-assay CV was 6.3% and the inter-assay CV was 9.2%.
566
V. Spasojevic-Kalimanovska et al. / Clinical Biochemistry 47 (2014) 564–569
Table 1 Basic clinical and laboratory characteristics of the patients. Variable
Patients (N = 188)
CAD(−) patients (N = 40)
CAD(+) patients (N = 148)
Female gender (n [%]) Age (years) Weight (kg) Waist (cm) TC (mmol/L) TG (mmol/L) HDL-C (mmol/L) LDL-C (mmol/L) ApoAI (g/L) ApoB (g/L) hsCRP (mg/L) Fibrinogen (g/L) MDA (μmol/L) SOD (U/L) O•− 2 (μmol/min/L) Ferritin (μg/L), males
67 (35.6)
21 (52.5)
46 (31.1)*
56.0 ± 9.1 81.1 ± 2.7 97.7 ± 11.9 5.39 ± 1.11 2.06 (1.94–2.19) 0.82 ± 0.21 3.71 ± 1.08 1.38 ± 0.29 1.39 ± 0.36 4.31 (3.64–5.10) 5.61 ± 1.62 3.51 ± 0.76 108.20 ± 40.90 162.50 ± 45.50 113.24 (97.50–131.83) 43.85 (34.51–55.72) 1.44 ± 0.37
53.6 ± 10.2 81.4 ± 15.9 166.2 ± 9.0 4.94 ± 1.13 1.92 (1.65–2.24) 0.88 ± 0.22 3.31 ± 1.09 1.46 ± 0.26 1.24 ± 0.33 3.49 (2.44–4.99) 5.31 ± 1.61 3.47 ± 0.66 113.03 ± 42.89 154.94 ± 40.86 132.43 (82.79–211.84) 46.77 (29.51–73.96) 1.36 ± 0.36
56.6 ± 8.6 81.01 ± 11.9 171.4 ± 9.3** 5.52 ± 1.08** 2.11 (1.98–2.25) 0.80 ± 0.20* 3.81 ± 1.06* 1.35 ± 0.29* 1.42 ± 0.36** 4.55 (3.74–5.52) 5.70 ± 1.61 3.52 ± 0.79 107.02 ± 40.58 164.71 ± 46.80 110.92 (94.19–130.62) 42.46 (31.62–56.75) 1.45 ± 0.37
Ferritin (μg/L), females sTfR (mg/L)
Data are shown as mean ± SD for normally distributed continuous variables, geometrical mean (95% CI for the mean) for continuous variables with skewed distribution and absolute (relative) frequencies for categorical variables. ⁎ Significantly different from the value in the CAD(−) group, with P b 0.05. ⁎⁎ Significantly different from the value in the CAD(−) group, with P b 0.01. ⁎⁎⁎ Significantly different from the value in the CAD(−) group, with P b 0.001.
under the receiver–operator characteristic curve (AUC) for independent predictors of significant stenosis as well as for a multivariable logistic model that included independent predictors of significant stenosis and ferritin or sTfR values. The minimal level of statistical significance was set at P b 0.05. We estimated that a sample of 188 subjects with a two-sided significance level of 0.05 would be required to provide statistical power of 83%. We used power and sample size calculations for generalized regression models with covariate measurement error [19].
Results Demographic characteristics and laboratory data of all CAD patients, CAD(+) and CAD(−) patients investigated in the study are shown in Table 1. The CAD(+) patients had significantly higher TC, LDL-C and apoB concentrations and lower prevalence of female. HDL-C and apoAI concentrations were significantly lower in CAD(+) than in CAD(−) patients. Because the difference in ferritin levels, between genders was
significant P b 0.001, we compared ferritin levels between CAD(−) and CAD(+) groups separately for both genders. No significant difference was found between CAD(−) and CAD(+) for male P = 0.415 and for female P = 0.707 patients. In order to reduce the number of variables factor analysis was performed among all subjects. The Kaiser–Meyer–Olkin measure of sampling adequacy was 0.655. Table 2 shows that factor analysis resulted in a reduction of variable number from the original 15 to 5 composite clusters. All five factors had an eigenvalue ≥1 and they explained 73.95% of the total variance (22.40% factor 1, 15.32% factor 2, 14.06% factor 3, 12.87 % factor 4 and 9.29% factor 5). These factors were interpreted as (1) “proatherogenic factor” with positive loadings of TC, LDL-C, apoB and TG; (2) “inflammatory factor” with positive loadings of hsCRP, fibrinogen and MDA; (3) “antiatherogenic factor” with positive loadings of HDL-C and apoAI; (4) “obesity factor” with positive loadings of weight and waist; and (5) “antioxidative status factor” with positive loadings of SOD and age and negative loading of O•− 2 . The association between factor scores and the presence of high ferritin and sTfR values were assessed using logistic regression analysis. Serum ferritin and sTfR concentrations were categorized into tertiles based on the cut-off points of the entire distribution. The subjects with ferritin and sTfR concentrations between the two cut-off values were excluded from analysis. Because the difference in ferritin levels between genders was significant we defined tertiles separately for females and males. Males with ferritin concentrations ≤85.67 μg/L were coded 0 and with ferritin concentrations ≥169.18 μg/L were coded 1. Similarly, females with ferritin concentrations ≤34.00 μg/L were coded 0 and with ferritin concentrations ≥70.60 μg/L were coded 1. sTfR concentrations were coded as 0 for the lowest tertile (≤1.15 mg/L) and 1 for the highest tertile (≥1.69 mg/L). In separate univariable analysis “inflammatory”, “obesity” and “antiatherogenic” factors predicted a high ferritin value and “proatherogenic factor” predicted a high sTfR value (Table 3). Using univariable and multivariable logistic regression analysis we examined the risk for development of significant CAD (≥ 50%) in patients with high ferritin and sTfR values. Ferritin and sTfR were used as independent variables and CAD variable as dependent variable. The CAD(+) subjects were coded 1 and CAD(−) subjects were coded 0. Table 4 indicates that high sTfR levels (≥1.69 mg/L) increased the risk for significant stenosis whereas a high ferritin value had no effect on risk in the univariate model. We performed additional analysis to determine whether the “proatherogenic factor” was responsible for the association between high sTfR levels and significant CAD risk. Adjustment for the “proatherogenic factor” eliminated the risk, suggesting that this factor is a strong confounder in the model. We compared the ability of the “proatherogenic factor” and sTfR with that of a multivariable logistic model that included the “proatherogenic factor” and sTfR values in predicting the presence of significant stenosis
Table 2 Results of factor analysis of proatherogenic, antiatherogenic, inflammatory, obesity and antioxidative variables. Factors
TC (mmol/L) LDL-C (mmol/L) ApoB (g/L) TG (mmol/L) Fibrinogen (g/L) hsCRP (mg/L) MDA (μmol/L) HDL-C (mmol/L) ApoAI (g/L) Waist (cm) Weight (kg) SOD (U/L) Age (years) O•− 2 (μmol/min/L) a
Proatherogenic
Inflammatory
Antiatherogenic
Obesity
Antioxidative
0.966a 0.951a 0.921a 0.546a −0.027 0.158 −0.050 −0.044 0.009 −0.005 −0.032 −0.019 0.032 0.345
−0.014 0.048 0.076 −0.082 0.831a 0.830a 0.723a −0.136 −0.145 0.090 0.022 −0.081 0.315 0.275
0.096 0.002 −0.063 −0.413 −0.037 −0.112 −0.101 0.884a 0.862a −0.133 −0.322 −0.186 0.285 0.012
−0.063 −0.041 0.005 0.158 −0.011 −0.033 0.184 −0.253 −0.180 0.923a 0.843a −0.185 0.193 −0.071
−0.067 −0.052 0.027 −0.226 −0.114 −0.091 0.145 0.029 −0.125 −0.004 −0.030 0.799a 0.544a −0.502a
Factor loadings higher than 0.50 or lower than −0.50.
V. Spasojevic-Kalimanovska et al. / Clinical Biochemistry 47 (2014) 564–569 Table 3 Logistic regression analysis for factor scores with ferritin and sTfR. Independent variable
Ferritin OR (95% CI)
sTfR OR (95% CI)
Proatherogenic factor
0.856 (0.533–1.373) P = 0.519 2.479 (1.266–4.854) P = 0.008 0.442 (0.237–0.823) P = 0.010 2.016 (1.146–3.546) P = 0.015 0.986 (0.609–1.596) P = 0.955
3.455 (1.582–7.545) P = 0.002 1.243 (0.787–1.963) P = 0.350 0.728 (0.439–1.206) P = 0.218 0.695 (0.416–1.162) P = 0.165 1.404 (0.817–2.415) P = 0.220
Inflammatory factor Antiatherogenic factor Obesity factor Antioxidative factor
in the study population. The AUC for the “proatherogenic factor” was 0.692 (0.578–0.807) and AUC for sTfR was 0.649 (0.514–0.784), which were lower than that of the multivariable logistic model [c-statistics 0.821 (0.705–0.937)]. Discussion In evaluating the relationship between iron status and risk of CAD, it is important to rule out non-iron-related factors that can influence the biochemical measures of iron status. In the present study, we used factor analysis to investigate the underlying correlation structure of variables that were associated with iron status parameters in patients with CAD. To our knowledge, such a factor analysis has not been used in previous studies of association between iron status parameters and risk for CAD. Five factors emerged in the present analysis (Table 2). The first factor accounting for 22.40% of variance comprised TC, LDL-C, apoB and TG, a proatherogenic lipid profile and this factor was considered the “proatherogenic factor”. The second factor was defined by an increase of hsCRP, fibrinogen and MDA. Clustering of hsCRP and fibrinogen as parameters of inflammation and MDA as a marker of oxidative status is consistent with the hypothesis of “oxidative response to inflammation” being the generation of oxidative species as a consequence of the inflammatory process [20]. Our previous study [21] suggested that the relationship between oxidative stress parameters and inflammatory species was due to their strong mutual involvement in atherosclerosis development that leads to CAD progression. The third factor was defined by the increase in HDL-C and apoAI and was considered to represent antiatherogenic parameters. “Obesity factor” was defined with positive loadings of weight and waist circumference. Obesity is a crucial risk factor for atherosclerosis and CAD and substitution of body mass index and waist: hip ratio for weight and waist, respectively, resulted in better factor loading. The fifth “antioxidative status factor” clustered parameters of oxidative stress (O•− 2 ) and antioxidative defense SOD. Although age is not a specific oxidative/antioxidative marker, the results of factor analysis indicate that the age has communality with SOD and O•− 2 . After clustering all variables into five factors, they were analyzed as confounders of iron status. The parameters of iron status examined in our study were: serum ferritin and sTfR. Predictors of high sTfR values were atherogenic lipid parameters LDL-C, apoB, TG and TC (“proatherogenic factor”) in our study. If we hypothesize that high sTfR concentrations reflect an elevated demand for iron in macrophage cells in the atherosclerotic lesion it is logical that high values of proatherogenic lipid parameters are predictors of high sTfR concentration [22] and this finding supports the iron hypothesis. In this study inflammatory factors hsCRP and fibrinogen were strongly associated with high serum ferritin. There is evidence from a recent study [23] that inflammation contributes to the pathogenesis of cardiovascular disease and many prospective studies have shown an association between elevated levels of CRP and the risk of cardiovascular disease. CRP and ferritin are acute-phase reactants, which are increased with inflammation and infection. Serum ferritin increases with inflammation
567
because ferritin synthesis is up-regulated by proinflammatory cytokines. Consequently, the association between serum ferritin and CAD could be confounded by inflammation and the ferritin levels in patients' sera may be a surrogate measure of inflammation [23]. This may explain the failure of some previous studies to find an association between ferritin and CAD. Ferritin as a marker of inflammation rather than iron stores may also explain the apparent synergistic association between ferritin and cholesterol levels and the risk of cardiovascular disease reported in the Brunneck study [4]. Alternatively, the association between ferritin and CRP may be explained by infection. The rise in ferritin levels and decrease in iron levels with acute infection are an important defense against proliferation of microorganisms. This is suggestive evidence that the infection may contribute to the pathogenesis of atherosclerosis [24]. In our study we eliminated patients with higher levels of hsCRP (≥10 mg/L) which can indicate acute infection and the association obtained between inflammatory clustered parameters and ferritin really supports the iron hypothesis. In the current study, the obtained association between high ferritin and obesity, measured by anthropometric surrogates is in agreement with some previous studies [23,25]. In the study of Jehn et al. [25] serum ferritin was significantly associated with waist measurement in both men and women and an increased risk of cardiovascular disease within persons with fat distribution in the abdominal region. Adipose tissue is a major regulator of chronic, low grade inflammation, characterized by abnormal cytokine production, increased acute phase reactants, activation of inflammatory signaling pathways and up regulation of ferritin synthesis. Therefore, the high ferritin levels observed in abdominal obesity could better reflect an inflammatory phenomenon as part of an acute-phase reaction than an increase of the iron stores. In our study, logistic regression analysis did not suggest associations of “antioxidative factor” and parameters of iron status. Ferritin expression is regulated by oxidative stress [26] but several studies have shown that ferritin protects cells from oxidative stress by maintaining iron homeostasis [27] so it is still uncertain whether ferritin is an antioxidative or a pro-oxidative parameter. Although the lipid oxidation theory of atherosclerosis is based on the iron-catalyzed oxidation of LDL, our results suggest that serum ferritin is an inappropriate marker of free (redox-active) iron. There are two possible explanations for this observation. First, the applied statistical method, factor analysis grouped the parameters of oxidative stress O•− 2 , and an enzyme of antioxidative defense, SOD, in one cluster. Second, ferritin does not reflect the free (redox-active) iron. Van Tits et al. [28] measured non-transferrinbound iron (NTBI), which represents the metal catalytically-active form of iron. They verified the proposed atherosclerotic mechanism of iron by measuring serum levels of NTBI in relation to markers of LDL oxidation and markers of inflammation. In our research we were not able to measure NTBI as a marker of iron status. In the second part of our study we evaluated the predictive power of iron status parameters in detecting CAD. We hypothesized that high serum ferritin and sTfR levels would be independently associated with CAD. To test those hypotheses, we measured ferritin and sTfR concentrations in patients with clinically significant CAD (stenosis ≥ 50%). assessed by coronary angiography and in CAD-free subjects (stenosis ≤ 50%). In the current study, we demonstrated that high serum sTfR Table 4 Logistic regression analysis of the associations of high ferritin and sTfR with significant stenosis. Variables
OR
95% CI
P
Univariable logistic model Ferritin sTfR
0.568 3.643
0.231–1.396 1.178–11.265
0.218 0.025
Multivariable logistic model Proatherogenic factor sTfR
2.883 5.139
1.103–7.540 0.953–27.697
0.031 0.057
568
V. Spasojevic-Kalimanovska et al. / Clinical Biochemistry 47 (2014) 564–569
was associated with the presence of significant CAD (≥ 50%). High serum ferritin values were not able to predict CAD and these findings are partially in agreement with the observation of Braun et al. [22], who investigated the association between ferritin and sTfR with the extension of CAD. They found a significant association between sTfR values and the extension of CAD. Mechanisms that can explain the association between sTfR and the extension of CAD are largely unknown. Experimental studies reported [29] an increased need for iron supply in atherosclerotic lesions as cellular replication is a constant feature of atherogenesis. Also, the amount of iron has been reported to be higher than in a normal arterial wall [30]. Furthermore, it has been demonstrated [31] that elevated iron levels in macrophages increase rather than decrease TfR mRNA. The increased number of membrane-associated TfR may increase the capacity of cells in atherosclerotic lesions to take up iron from circulation. An excess iron amount due to a higher density of TfR in cell membranes promotes the formation of toxic oxygen radicals and lipid peroxidation, potentiating the progression of atherosclerotic lesions in arterial walls. In the same study, Braun et al. [22] found that in the group of patients with CAD, ferritin concentrations were higher in patients with acute coronary syndromes than in those with stable angina. This study [22] supports the concept that increased ferritin levels may be a marker of instability in patients with existing CAD. In our study ferritin was not a predictor of significant CAD and this could be explained by the clinical status of our patients as they had stable angina. Multivariable logistic regression indicated the “proatherogenic factor” as a strong confounder in sTfR prediction of significant CAD. The observed association was dependent on the “proatherogenic factor”, suggesting that the increased sTfR concentration is related to coexisting changes in the classical risk factor defined as proatherogenic. To our knowledge, this is the first study to explore whether the measurement of serum sTfR could predict significant CAD independently of clustered cardiovascular risk factors. In the third part of our study, we employed C statistics to assess the clinical usefulness of laboratory markers. In our logistic regression model, the “proatherogenic factor” was an independent risk factor for CAD prediction and the AUC for this model was 0.692. According to Hosmer–Lemeshow's criteria [32] it was considered to be poor, but in the multivariable logistic model that included the “proatherogenic factor” and sTfR, the value was 0.821 considered to be excellent. The level of sTfR as a single parameter was not particularly helpful in discriminating CAD(+) patients from CAD(−) patients but as an additional CAD marker had a significant diagnostic value. The obtained results in the present study partly support the iron hypothesis of atherosclerosis. Although serum ferritin concentration is considered the best indicator for body iron stores, in the present study ferritin was not a predictor of significant coronary artery stenosis. Serum sTfR concentration that reflects the functional iron compartment is a better predictor of CAD but with a reasonable accuracy and more important as an additional CAD marker. Conclusions Factor analysis is an excellent tool for creating a summary risk measure and also defines relationships between variables related to underlying pathophysiologic mechanisms. In our study applied factor analysis identified five underlying factors among a group of 15 variables that have been associated with atherosclerosis and CAD. These data showed that factors interpreted as “inflammatory”, “obesity”, “proatherogenic” and “antiatherogenic” were associated with increased parameters of body iron status, partly confirming previous epidemiological studies. Our results showed that body iron status assessed by measurement of sTfR concentrations could be useful in the prediction of CAD. Despite this, the low accuracy of sTfR determination suggests that a single marker cannot be an adequate parameter for CAD screening. Satisfactory discriminative prediction abilities are achieved only in combination
with the proatherogenic risk factor. These findings highlight the importance of a global approach to assessing risk and the need for studies that elucidate how these various risk domains interact over time to create clinical disease. Further studies with a greater number of patients are required to support our results and to investigate the nature and significance of the relationship among serum iron indices, tissue iron concentrations and CAD. Conflict of interest The authors have no conflict of interest to declare. Acknowledgment This work was supported by a grant from the Ministry of Education, Science and Technological Development, Republic of Serbia (Project No. 175035) and by the European Cooperation in Science and Technology (COST) BM0904 Action. References [1] Sullivan JL. Iron and sex difference in heart disease risk. Lancet 1981;1:1293–4. [2] Toumainen TP, Punnonen K, Nyyssonen K, Salonen JT. Association between body iron stores and the risk of acute myocardial infarction in man. Circulation 1998;97:1461–6. [3] Kiechl S, Aichner F, Gerstenbrand F, Egger G, Mair A, Rungger G, et al. Body iron stores and presence of carotid atherosclerosis. Results from the Bruneck study. Atherioscler Thromb 1994;14:1625–30. [4] Kiechl S, Willeit J, Egger G, Poewe W, Oberhollenzer F. Body iron stores and the risk of carotid atherosclerosis. Prospective results from Bruneck study. Circulation 1997;96:3300–7. [5] Danesh J, Appleby P. Coronary heart disease and iron status. Meta-analyses of prospective studies. Circulation 1999;99:852–4. [6] Liao Y, Cooper RS, McGee DL. Iron status and coronary heart disease: negative findings from the NHANES I epidemiologic follow-up study. Am J Epidemiol 1994;139:704–12. [7] Moore M, Folsom AR, Barnes RW, Eckfeldt JH. No association between serum ferritin and asymptomatic carotid atherosclerosis. The Atherosclerosis Risk in Communities (ARIC) study. Am J Epidemiol 1995;141:719–23. [8] Bozzini C, Girelli D, Tinazzi E, et al. Biochemical and genetic markers of iron status and the risk of coronary artery disease: an angiography-based study. Clin Chem 2002;48:622–8. [9] Ref DW. Ferritin as a source of iron oxidative damage. Free Radic Biol 1992;12:417–27. [10] Gabay C, Kuushener I. Acute-phase proteins and other systemic responses to inflammation. N Engl J Med 1999;340:448–54. [11] Knuiman MW, Divitini ML, Olynyk JK, Cullen DJ, Bartholomew HC. Serum ferritin and cardiovascular disease: a 17-year follow-up study in Busselton, Western Australia. Am J Epidemiol 2003;158:144–9. [12] Cureton EE, D’Agostino RB. Factor analysis, an applied approach; 1983 [N.J. Hillsdale, L. Erlbaum associates]. [13] Causes of deaths. In: Vukomirovic D, editor. Statistical yearbook of Serbia 2005. Belgrade: SCG: Statistical Office of the Republic of Serbia; 2005. p. 90. [14] Pearson TA, Mensah GA, Alexander RW, Anderson JL, Cannon III RO, Criqui M, et al. Markers of inflammation and cardiovascular disease: application to clinical and public health practice: a statement for healthcare professionals from the Centers for Disease Control and Prevention and the American Heart Association. Circulation 2003;107:499–511. [15] Girotti MJ, Khan N, McLellan BA. Early measurement of systematic lipid peroxidation products in plasma of major blunt trauma patients. J Trauma 1991;31:32–5. [16] Auclair C, Voisin E. Nitroblue tetrazolium reduction. In: Greenwald RA, editor. CRC handbook of methods for oxygen radical research. Boca Raton, Florida: CRC Press; 1985. p. 123–32. [17] Misra HP, Fridovich I. The role of superoxide anion in the autooxidation of epinephrine and a simple assay for superoxide dismutase. J Biol Chem 1972;247:3170–5. [18] Stevens J. Applied multivariate statistics for the social sciences. 3rd ed. N.J. Mahwah: Lawrence Erlbaum Associates; 1996. [19] Tosteson T, Buzas J, Demidenko E, Karagas M. Power and sample size calculations for generalized regression models with covariate measurement error. Stat Med 2003;22:7. [20] Libby P, Ridker PM, Maseri A. Inflammation and atherosclerosis. Circulation 2002;105:1135–43. [21] Memon L, Spasojevic-Kalimanovska V, Bogavac-Stanojevic N, KalimanovskaOstric D, Jelic-Ivanovic Z, Spasic S, et al. Association of C-reactive protein with the presence and extent of angiographically verified coronary artery disease. Tohoku J Exp Med 2006;209:197–206. [22] Braun S, Ndrepepa G, von Beckerath N, Vogt W, Schomig A, Kastrati A. Value of serum ferritin and soluble transferrin receptor for prediction of coronary artery disease and its clinical presentations. Atherosclerosis 2004;174:105–10. [23] Williams MJA, Poulton R, Williams S. Relationship of serum ferritin with cardiovascular risk factors and inflammation in young men and women. Atherosclerosis 2002;165:179–84.
V. Spasojevic-Kalimanovska et al. / Clinical Biochemistry 47 (2014) 564–569 [24] Leinonen M, Saikku P. Evidence for infectious agents in cardiovascular disease and atherosclerosis. Lancet Infect Dis 2002;2:11–7. [25] Jehn M, Clark JM, Gaullar E. Serum ferritin and the risk of the metabolic syndrome in U.S. adults. Diabetes Care 2004;27:2422–8. [26] Cairo G, Tacchini L, Pogliaghi G, Anzon E, Tomasi A, Bernelli-Zazzero A. Introduction of ferritin synthesis by oxidative stress. Transcriptional and posttranscriptional regulation by expansion of the “free” iron pool. J Biol Chem 1995;270:700–3. [27] You SA, Wand Q. Ferritin in atherosclerosis. CCA 2005;357:1–16. [28] van Tits LJH, Jacobs EMG, Swinkels DW, Lemmers HLM, van der Vleuten GM, de Graaf J, et al. Non-transferrin-bound iron is associated with plasma level of soluble
[29] [30] [31]
[32]
569
intercellular adhesion molecule-1 not with in vivo low-density lipoprotein oxidation. Atherosclerosis 2007;194:272–8. Shwartz SM, Ross R. Cellular proliferation in atherosclerosis and hypertension. Prog Cardiovasc Dis 1984;26:355–72. Sullivan JL. Iron in arterial plaque: a modifiable risk factor for atherosclerosis. Biochim Biophys Acta 2009;1790:1014–20. Testa U, Petrini M, Quaranta MT, Pelosi-Testa E, Mastroberardino G, Camagna, et al. Iron up-modulates the expression of transferin receptors during monocyte–macrophage maturation. J Biol Chem 1989;264:13181–7. Hosmer DW, Lemeshow S. Assessing the fit of the model, applied logistic regression. NY: John Wiley and Sons; 2000 160–4.
