OUP user menu

The phosphatonin fibroblast growth factor 23 links calcium–phosphate metabolism with left-ventricular dysfunction and atrial fibrillation

Sarah Seiler, Bodo Cremers, Niko M. Rebling, Florian Hornof, Jana Jeken, Sylvie Kersting, Charlotte Steimle, Philipp Ege, Michael Fehrenz, Kyrill S. Rogacev, Bruno Scheller, Michael Böhm, Danilo Fliser, Gunnar H. Heine
DOI: http://dx.doi.org/10.1093/eurheartj/ehr215 2688-2696 First published online: 6 July 2011

Abstract

Aims High serum phosphate is linked to cardiovascular morbidity and mortality in the general population. Fibroblast growth factor 23 (FGF-23) is a critical phosphate regulating hormone, potentially reflecting phosphate load better than a single serum phosphate measurement. Recent pioneering echocardiographic studies associated FGF-23 with left-ventricular morphology. However, the association between FGF-23 and left-ventricular function is unknown, prompting us to investigate this relationship in our HOM SWEET HOMe study.

Methods and results We studied the association between C-terminal FGF-23, coronary artery disease, and left-ventricular function in 885 subjects undergoing elective coronary angiography. Left-ventricular function was assessed with ventriculography. More, pro-brain natriuretic peptide (pro-BNP) plasma levels were measured. The presence of left-ventricular hypertrophy and atrial fibrillation was assessed by electrocardiography. Patients with an ejection fraction <40% had significantly higher FGF-23 levels compared with patients with the ejection fraction >40% (P< 0.001). In multivariable regression analysis, the observed relationship between FGF-23 and left-ventricular function remained significant after adjustment for estimated glomerular filtration rate, presence of left-ventricular hypertrophy, and other confounding variables. In accordance, FGF-23 significantly correlated with pro-BNP plasma levels (r = 0.31; P< 0.001). Prevalent atrial fibrillation was associated with elevated FGF-23 levels, while the presence of coronary artery disease was not.

Conclusions Fibroblast growth factor 23 levels are associated with left-ventricular function and atrial fibrillation even in the absence of renal function impairment. Of note, these cross-sectional data cannot prove causality; therefore, future studies will have to discern whether FGF-23 exerts a direct untoward effect on the myocardium, or rather represents an ‘innocent bystander’ which reflects a high phosphate burden.

  • FGF-23
  • Calcium–phosphate metabolism
  • Heart failure
  • pro-BNP

Introduction

Cardiovascular diseases are a growing burden to health-care systems around the globe. As conventional risk factors only partly explain incident and prevalent cardiovascular disease, a broader understanding of non-traditional risk factors is eagerly awaited.

Inspired by earlier epidemiological studies in patients with chronic kidney disease (CKD),16 a disturbed calcium–phosphate metabolism has recently been acknowledged as a significant contributor to cardiovascular disease in the general population: among apparently healthy individuals, higher phosphate levels—albeit within apparently normal ranges—emerged as independent predictors of coronary artery calcification over time.7,8 Concomitantly, epidemiological studies found higher serum phosphate—again mostly within physiological ranges—to predict future cardiovascular events,912 development of symptomatic heart failure,10 cardiovascular mortality,9,12 and all-cause mortality9,10,12 among subjects with intact renal function.

Similarly, low circulating levels of active vitamin D [1,25-(OH)2-D3] were identified as predictors of adverse outcome in subjects with intact renal function.13

Serum phosphate and 1,25-(OH)2-D3 levels are under control of the recently discovered phosphate-regulating hormone fibroblast growth factor 23 (FGF-23)14: acting via its co-factor Klotho, FGF-23 reduces gastrointestinal phosphate absorption,15 stimulates renal phosphate elimination,16 and lowers renal conversion of 25-(OH)-D3 into 1,25-(OH)2-D3.14

While high-dietary phosphate intake is generally considered to be the principle physiological stimulus for FGF-23 expression,17 short-term changes in phosphate ingestion do not immediately impact serum FGF-23 levels.1820 Therefore, FGF-23 holds promise as a biomarker of long-term phosphate burden, comparable with Hba1c measurements for assessing long-term glucose control in diabetes mellitus.17

Circulating levels of FGF-23 surpass serum phosphate as predictors of mortality in dialysis patients21 and in patients with less advanced CKD.22 Nonetheless, as FGF-23 levels disproportionally increase in CKD in order to combat hyperphosphataemia, such findings are of limited relevance to individuals with intact renal function.

While we came to understand a deranged calcium–phosphate metabolism as an emerging cardiovascular risk factor in the general population, the impact of FGF-23 measurement in the general population remained poorly defined until recently. In 2009, pioneering echocardiographic studies reported an association between FGF-23 levels and left-ventricular (LV) hypertrophy.23,24 Left-ventricular hypertrophy predisposes to LV dysfunction and heart failure,25,26 both of which are major predictors of mortality2729 and potentially linked to a deranged calcium–phosphate metabolism.10

The HOM SWEET HOMe study, which recruited patients undergoing elective coronary angiography, allowed us to investigate the association between FGF-23 levels, electrocardiogram (ECG) parameters (LV hypertrophy and atrial fibrillation), coronary artery disease, and LV function in 885 subjects, of whom the majority had intact renal function.

Methods

Study cohort

Between May 2007 and January 2010, we recruited subjects who were admitted at Saarland University Hospital for elective coronary angiography. The study was approved by the local Ethics Committee; all patients provided written informed consent. Patients undergoing renal replacement therapy were excluded. A standardized questionnaire was used to record a history of smoking, diabetes, current drug intake, and cardiovascular co-morbidity. Additionally, co-morbidity was assessed by chart review. Patients were categorized as active smokers if they were current smokers or had stopped smoking <1 month before entry into the study. Patients with self-reported or physician-reported diabetes mellitus, with a non-fasting blood glucose level of >200 mg/dL, with a fasting blood glucose level of >126 mg/dL, or with current use of hypoglycaemic medication were categorized as diabetic. Body mass index was calculated as weight (kg)/(height (m))2.

All patients underwent electrocardiographic readings at admission for assessment of LV hypertrophy, and the presence of atrial fibrillation. According to the criteria proposed and validated by the LIFE trial group,30 electrocardiographic LV hypertrophy was defined in patients who had Cornell voltage-duration product [product of QRS duration times Cornell voltage (RavL ± SV3, with 6 mm added in women)] > 2440 mm*ms, and/or a Sokolow-Lyon voltage (SV1± RV5–6) > 38 mm.

Coronary angiography was performed on the day of hospital admission. Coronary artery disease was defined in patients who had a stenosis of ≥50% of a major coronary artery in the present coronary angiography and/or who had a history of coronary revascularization for coronary artery stenosis. The ejection fraction (EF) was measured by ventriculography.

In order to complement the data on the association between FGF-23 levels and EF, we additionally assessed serum levels of pro-brain natriuretic peptide (pro-BNP). Data on pro-BNP levels were available in 513 out of 885 patients of our initial cohort, and further in 135 patients who—during the same time period—underwent elective coronary angiography without ventriculography. Most of these additional 135 patients had had assessment of LV function either by ventriculography or by alternative imaging methods recently before admission to our hospital, and/or had high risk of contrast media nephropathy, so that indication for ventriculography was left to the discretion of the treating physician.

Laboratory measurements

Blood samples were obtained under standardized conditions. The samples were centrifuged at 3300 g for 5 min at room temperature. Supernatants were stored in aliquots at −80°C until further use. C-terminal FGF-23 levels were measured from plasma samples by ELISA (Immutopics, San Clemente, CA, USA; low cut-off value 3 rU/mL, high cut-off value 2000 rU/mL). Blood levels of creatinine, calcium, phosphate, pro-BNP, total cholesterol, HDL-cholesterol (HDL-C), LDL-cholesterol (LDL-C), and C-reactive protein were measured by standard methods. Estimated glomerular filtration rate (eGFR) was calculated using the MDRD study equation (4), and CKD was defined as an eGFR <60 mL/min/1.73 m2.

Statistics

Data management and statistical analysis were performed with PASW Statistics 18. Two-sided P-values <0.05 were considered significant. Categorical variables are presented as percentage of patients and compared by χ2 test. Continuous data are expressed as means ± standard deviation, and compared by Mann–Whitney test, or by Kruskal–Wallis test, as appropriate. Fibroblast growth factor 23 and C-reactive protein levels are presented as median [inter-quartile range (IQR)] because of skewed distribution. The association between continuous variables was assessed by Spearman's rank correlation testing.

Subsequently, we used linear regression analysis to test the association between log-transferred FGF-23 levels and EF, and logistic regression analysis to test the association between log-transferred FGF-23 levels and atrial fibrillation. Multivariable models were used to adjust for confounding. A priori we decided to adjust for those baseline variables which were associated with FGF-23 levels at baseline.

Results

Patients' characteristics

A total of 885 patients were included in the study cohort, of whom 385 patients had undergone previous coronary revascularization, and were admitted for angiographic re-evaluation. In the remaining 500 patients, cardiac catheterization was performed for exclusion of coronary artery disease (mainly in the course of evaluation of chest pain). Patients' characteristics are depicted in Table 1. Estimated glomerular filtration rate was <60 mL/min/1.73 m2 in 170 subjects (classified as CKD patients), the remaining 715 patients had intact renal function with an eGFR ≥60 mL/min/1.73 m2. As expected, individuals with eGFR <60 mL/min/1.73 m2 were older; they were more likely to be female and to be non-smokers. More, they had a higher prevalence of diabetes mellitus, higher plasma levels of C-reactive protein, calcium, phosphate, pro-BNP, and FGF-23. Finally, patients with eGFR <60 mL/min/1.73 m2 were more likely to receive angiotensin receptor blockers and beta-blockers (BBs) (see Supplementary material online, Table S1).

View this table:
Table 1

Patients' characteristics (stratification by ejection fraction)

Total cohort (n= 885)EF <40% (n= 56)EF 40–59% (n= 208)EF ≥60% (n= 621)P-value
Age (years)64.8 ± 10.163.9 ± 10.664.5 ± 10.365.0 ± 10.00.641
Women (%)286 (32.3)12 (21.4)42 (20.2)232 (37.4)<0.001
Smokers (%)127 (14.4)14 (25.0)38 (18.3)75 (12.1)0.006
Diabetes mellitus (%)328 (37.1)22 (39.3)91 (43.8)215 (34.6)0.058
CAD (%)599 (67.7)48 (85.7)159 (76.4)392 (63.1)<0.001
1 vessel disease143 (16.2)7 (12.5)33 (15.9)103 (16.6)
2 vessel disease186 (21.0)12 (21.4)42 (20.2)132 (21.3)
3 vessel disease270 (30.5)29 (51.8)84 (40.4)157 (25.3)
C-reactive protein (mg/L)1.9 (0.9–4.0)2.8 (1.3–5.3)1.8 (0.9–3.9)1.9 (0.9–4.0)0.041
Total cholesterol (mg/dL)191.6 ± 46.2181.0 ± 45.2183.4 ± 41.9195.3 ± 47.20.001
HDL cholesterol (mg/dL)50.3 ± 15.046.8 ± 14.547.4 ± 13.951.6 ± 15.3<0.001
LDL cholesterol (mg/dL)117.3 ± 40.0110.3 ± 37.0113.0 ± 36.1119.3 ± 41.40.091
Body mass index (kg/m2)28.6 ± 4.729.2 ± 5.728.6 ± 4.528.6 ± 4.70.795
Plasma calcium (mmol/L)2.4 ± 0.12.4 ± 0.12.4 ± 0.12.3 ± 0.10.850
Plasma phosphate (mg/dL)3.3 ± 0.53.4 ± 0.63.3 ± 0.53.3 ± 0.50.530
eGFR (mL/min/1.73 m2)76.3 ± 19.472.5 ± 19.875.8 ± 19.576.9 ± 19.30.200
BB (%)645 (72.9)46 (82.1)163 (78.4)436 (70.2)0.058
ACE-I (%)484 (54.7)44 (78.6)127 (61.1)313 (50.4)<0.001
ARB (%)180 (20.3)6 (10.7)37 (17.8)137 (22.1)0.110
Statin (%)479 (54.1)33 (58.9)124 (59.6)322 (51.9)0.299
  • Variables are presented as percentage, or as mean ± SD, as appropriate. C-reactive protein levels are presented as median (inter-quartile range) because of skewed distribution. EF, ejection fraction; CAD, coronary artery disease; eGFR, estimated glomerular filtration rate; BB, beta-blockers; ACE-I, angiotensin-converting enzyme inhibitors; ARB, angiotensin receptor blockers. Data on medication are missing in seven patients.

The median plasma FGF-23 level was 38.6 rU/mL (IQR: 16.3–69.3 rU/mL). Fibroblast growth factor 23 levels were correlated with lower renal function, higher serum phosphate, and C-reactive protein in the total cohort and in subcohorts stratified by eGFR. The correlation coefficients between FGF-23 and other cardiovascular risk factors were of negligible magnitude (Table 2).

View this table:
Table 2

Spearman correlation coefficients of patients' characteristics with plasma FGF-23 levels

Total cohort (n= 885)eGFR <60 mL/min/1.73 m2 (n= 170)eGFR ≥60 mL/min/1.73 m2 (n= 715)
rP-valuerP-valuerP-value
Age0.0980.004–0.0250.7490.0510.172
C-reactive protein0.194<0.0010.2530.0010.162<0.001
Total cholesterol–0.0840.013–0.0900.245–0.0900.017
HDL cholesterol–0.1030.002–0.0920.232–0.1080.004
LDL cholesterol–0.0770.023–0.1050.176–0.0720.055
Body mass index0.0520.1200.0430.5790.0480.196
Plasma calcium0.0860.0100.2160.0050.0310.406
Plasma phosphate0.138<0.0010.1520.0490.1170.002
eGFR–0.251<0.001–0.337<0.001–0.145<0.001
  • Indicated are correlations coefficients (r), and levels of significance (P); eGFR: estimated glomerular filtration rate.

Fibroblast growth factor 23 levels were higher in females than in males (see Supplementary material online, Figure S1). More, FGF-23 levels were higher in diabetic patients than in non-diabetic individuals, and they were higher in patients on BB and/or angiotensin-converting enzyme inhibitor (ACE-I) treatment than in patients not receiving these medications. These differences remained significant after exclusion of patients with eGFR <60 mL/min/1.73 m2 (see Supplementary material online, Table S2).

Smokers tended to have higher FGF-23 levels in the total cohort; this difference reached statistical significance after exclusion of patients with eGFR <60 mL/min/1.73 m2. In contrast, FGF-23 levels did not differ after stratifying patients for intake of angiotensin receptor blockers, or for statin intake, respectively (see Supplementary material online, Table S2).

Fibroblast growth factor 23, left-ventricular function, left-ventricular hypertrophy, and coronary artery disease

Patients with impaired systolic function had a higher prevalence of nicotine abuse, diabetes mellitus, and coronary artery disease, higher intake of ACE-Is and higher C-reactive protein levels. More they tended to have worse renal function (Table 1).

As shown in Figure 1, patients with EF <40% had significantly higher FGF-23 levels than patients with EF ≥40%. This association remained significant after exclusion of patients with eGFR <60 mL/min/1.73 m2.

Figure 1

Fibroblast growth factor 23 levels after stratifying for ejection fraction (EF; categorized as <40, 40–59, and ≥60%) in the total study cohort (left columns), and in individuals with intact renal function (eGFR ≥60 mL/min/1.73 m2; right columns). Data are presented as mean ± SEM.

In multivariable regression analysis, the observed relationship between FGF-23 and EF remained significant after step-wise adjustment for age and gender (model 1), phosphate and eGFR (model 2), prevalence of diabetes mellitus, LV hypertrophy and C-reactive protein (model 3), and intake of BBs and ACE-Is (model 4; Table 3).

View this table:
Table 3

Multivariable regression analysis with left-ventricular function as dependent variable

Model 1Model 2Model 3Model 4
BSEMP-valueBSEMP-valueBSEMP-valueBSEMP-value
Constant61.303.23<0.00163.535.79<0.00160.755.64<0.00162.075.63<0.001
FGF-23 (log)–2.930.71<0.001–2.370.710.001–2.010.700.004–2.030.700.004
Age (years)0.090.050.0490.130.050.0140.170.050.0010.170.050.001
Gender (female)5.551.05<0.0017.341.10<0.0017.001.07<0.0016.981.07<0.001
Plasma phosphate (mg/dL)–3.410.93<0.001–2.590.910.004–2.450.910.007
eGFR (mL/min/1.73 m2)0.070.030.0070.060.030.0180.060.030.020
Diabetes mellitus (yes)–0.860.970.380–0.350.980.719
C-reactive protein (log)0.061.080.954–0.001.070.998
Hypertrophy (yes)–8.891.16<0.001–8.671.16<0.001
BB intake (yes)–0.981.100.374
ACE-I intake (yes)–2.670.970.006
  • B, regression coefficient; SEM, standard error of the mean; P, level of significance; FGF-23, and C-reactive protein values were log transformed to approximate a normal distribution before multivariate regression analysis. eGFR, estimated glomerular filtration rate; BB, beta-blockers; ACE-I, angiotensin-converting enzyme inhibitors. Data on medication are missing in five patients.

We further analysed the interaction between EF, LV hypertrophy, and FGF-23 by dichotomizing patients by the presence or absence of LV hypertrophy. Highest FGF-23 levels were associated with impaired LV function in both patients with LV hypertrophy as well as in patients without hypertrophy before (Figure 2) and after exclusion of patients with impaired renal function (eGFR <60 mL/min/1.73 m2; see Supplementary material online, Figure S2).

Figure 2

Fibroblast growth factor 23 levels after stratifying for the ejection fraction and for presence of left-ventricular hypertrophy in individuals in the total study cohort. Data are presented as mean ± SEM.

Surprisingly, we found no difference in FGF-23 levels between patients with and without coronary artery disease (Figure 3). The same holds true when we stratified patients by the number of major coronary arteries affected (see Supplementary material online, Figure S3).

Figure 3

Fibroblast growth factor 23 levels after stratifying for the presence of coronary artery disease in the total study cohort (left columns), and in individuals with intact renal function (eGFR ≥60 mL/min/1.73 m2; right columns). Data are presented as mean ± SEM.

In accordance, high FGF-23 levels were significantly associated with elevated pro-BNP levels in the total patients cohort (r = 0.31; P< 0.001). Again, this association remained significant in subgroups categorized by eGFR (r = 0.20; P< 0.001 in intact renal function; r = 0.35; P< 0.001 in CKD). After stratifying the study cohort by pro-BNP levels in quartiles, patients in the highest pro-BNP quartile showed most elevated FGF-23 levels (Figure 4).

Figure 4

Fibroblast growth factor 23 levels after stratifying by quartiles of pro-BNP in the total study cohort (left columns), and in individuals with intact renal function (eGFR ≥60 mL/min/1.73 m2; right columns). Data are presented as mean ± SEM.

Fibroblast growth factor 23 and atrial fibrillation

In 67 out of 885 patients recruited, ECG at admission showed atrial fibrillation. As expected, patients with atrial fibrillation were older, and had higher comorbidity (Table 4).

View this table:
Table 4

Patients' characteristics (stratification by the presence of atrial fibrillation)

Atrial fibrillation (n= 67)No atrial fibrillation (n= 817)P-value
Age (years)70.0 ± 7.664.4 ± 10.2<0.001
Women (%)24 (35.8)261 (31.9)0.500
Smokers (%)6 (9.0)120 (14.7)0.274
Diabetes mellitus (%)34 (50.7)294 (36.0)0.018
CAD (%)36 (53.7)563 (68.9)0.041
1 vessel disease12 (17.9)131 (16.0)
2 vessel disease11 (16.4)175 (21.4)
3 vessel disease13 (19.4)257 (31.5)
C-reactive protein (mg/L)2.8 (1.4–7.1)1.8 (0.9–3.8)0.001
Total cholesterol (mg/dL)185.3 ± 44.1192.1 ± 46.40.249
HDL cholesterol (mg/dL)48.0 ± 12.450.5 ± 15.20.355
LDL cholesterol (mg/dL)114.1 ± 37.6117.6 ± 40.20.565
Body mass index (kg/m2)28.5 ± 4.730.2 ± 4.80.002
Plasma calcium (mmol/L)2.4 ± 0.12.4 ± 0.10.681
Plasma phosphate (mg/dL)3.3 ± 0.63.3 ± 0.50.532
eGFR (mL/min/1.73 m2)70.2 ± 21.676.8 ± 19.10.001
BB (%)53 (79.1)592 (72.5)0.453
ACE-I (%)45 (67.2)439 (53.7)0.096
ARB (%)15 (22.4)165 (20.2)0.780
Statin (%)33 (49.3)446 (54.6)0.516
EF (%)59.5 ± 15.765.4 ± 14.40.002
  • Variables are presented as percentage, or as mean ± SD, as appropriate. C-reactive protein levels are presented as median (inter-quartile range) because of skewed distribution. CAD, coronary artery disease; EF, ejection fraction; eGFR, estimated glomerular filtration rate; BB, beta-blockers; ACE-I, angiotensin-converting enzyme inhibitors; ARB, angiotensin receptor blockers. Data on medication are missing in six patients, and the presence of atrial fibrillation was unknown because of missing EGG in one patient.

Patients with atrial fibrillation had significantly higher plasma levels of FGF-23 compared with patients without atrial fibrillation (Figure 5). The same holds true after exclusion of patients with eGFR <60 mL/min/1.73 m2 (Figure 5, right columns), as well as after exclusion of patients with EF <60% (see Supplementary material online, Figure S4).

Figure 5

Fibroblast growth factor 23 levels after stratifying for the presence of atrial fibrillation in the total study cohort (left columns), and in individuals with intact renal function (eGFR ≥60 mL/min/1.73 m2; right columns). Data are presented as mean ± SEM.

In multivariable regression analysis, the elevated FGF-23 levels remained significantly associated with the presence of atrial fibrillation after step-wise adjustment for age and gender (model 1), phosphate and eGFR (model 2), prevalence of diabetes mellitus and C-reactive protein (model 3), EF and presence of LV hypertrophy (model 4), and intake of BBs and ACE-Is (model 5; Table 5).

View this table:
Table 5

Multivariable regression analysis with the presence of atrial fibrillation as dependent variable

Model 1Model 2Model 3Model 4Model 5
OR95% CIP-valueOR95% CIP-valueOR95% CIP-valueOR95% CIP-valueOR95% CIP-value
Constant0.00<0.0010.00<0.0010.00<0.0010.00<0.0010.00<0.001
FGF-23 (log)3.612.16–6.04<0.0013.602.12–6.10<0.0013.161.87–5.33<0.0012.941.74–4.97<0.0013.131.82–5.40<0.001
Age (years)1.071.03–1.10<0.0011.061.03–1.10<0.0011.071.03–1.11<0.0011.081.04–1.11<0.0011.081.04–1.11<0.001
Gender (female)0.850.49–1.470.5580.860.48–1.570.6280.880.48–1.600.6621.130.60–2.130.7041.090.57–2.070.794
Plasma phosphate (mg/dL)0.850.51–1.430.5490.830.49–1.400.4750.730.43–1.250.2460.720.42–1.240.238
eGFR (mL/min/1.73 m2)1.000.98–1.010.6251.000.98–1.010.7911.000.99–1.020.8631.000.99–1.020.951
Diabetes mellitus (yes)1.510.89–2.560.1281.570.92–2.670.0991.510.87–2.600.140
C-reactive protein (log)2.151.20–3.830.0102.141.18–3.880.0122.211.21–4.020.010
EF (%)0.980.96–0.990.0060.980.96–1.000.016
Hypertrophy (yes)1.090.59–2.020.7921.130.61–2.110.696
BB intake (yes)0.810.41–1.590.541
ACE-I intake (yes)1.360.76–2.410.302
  • OR, odds ratio; CI, confidence interval; P, level of significance. FGF-23 and C-reactive protein values were log transformed to approximate a normal distribution before multivariate regression analysis. eGFR, estimated glomerular filtration rate; EF, ejection fraction; BB, beta-blockers; ACE-I, angiotensin-converting enzyme inhibitors. Data on medication are missing in six patients.

Discussion

While numerous studies identified hyperphosphataemia and hypovitaminosis D as major cardiovascular risk factors in CKD patients in the last decade,16 the association between calcium–phosphate metabolism and cardiovascular disease in the general populations was long ignored. Recent data suggest that low-active vitamin D (1,25-(OH)2-D3) levels and high-normal phosphate levels are associated with adverse outcome even in the absence of CKD, and point to novel strategies of prevention and treatment.8,13,31 Nonetheless, those serum phosphate levels that predicted adverse outcome in previous studies8,31 were well within the seemingly normal range, thereby diminishing the discriminative power of this potential biomarker. Hence, a more sensitive parameter of a positive phosphate balance is needed.

Fibroblast growth factor 23 is a negative regulator of serum phosphate and 1,25-(OH)2-D3: in the presence of its co-factor Klotho, FGF-23 lowers serum phosphate levels—directly by reducing gastrointestinal absorption15 and promoting renal elimination of phosphate,16 and indirectly by inhibiting 1α-hydroxylase, thus lowering circulating levels of 1,25-(OH)2-D3.14

First evidence linking FGF-23 and cardiac disease was presented in two cross-sectional studies by Mirza et al. and Gutierrez et al.23,24 Both groups reported an association between LV hypertrophy, assessed by echocardiography, and serum FGF-23 levels in cohorts that comprised both CKD patients and individuals with intact renal function. As both cohorts deliberately excluded subjects with LV dysfunction, no data on the association between FGF-23 and LV function are currently available.

Given the long-term detrimental effect of LV hypertrophy on myocardial function25,26,32 and the immense prognostic impact of myocardial dysfunction,2729 we see an urgent need to extend the findings by Gutierrez et al. and Mirza et al.23,24 from subjects with normal LV function to patients with LV dysfunction. We therefore decided to investigate FGF-23 levels in a large unselected cohort of subjects admitted for elective coronary angiography. Comprising both patients with intact renal function and with CKD, the HOM SWEET HOMe study reports an association of circulating FGF-23 levels with pro-BNP levels, and with LV function—measured by ventriculography. Furthermore, we found a novel relationship between elevated FGF-23 and atrial fibrillation.

Taken together, our findings might help to explain the association of elevated FGF-23 levels and detrimental outcome in selected patient populations such as prevalent33 and incident34 dialysis patients. Notably, in dialysis patients levels of FGF-23 are increased by >1000-fold20 either because of increased production35,36—in a futile attempt to combat hyperphosphataemia—or because of diminished renal clearance of FGF-23.20

We and others recently extended these prognostic data from dialysis patients to other study cohorts, reporting FGF-23 levels to predict future cardiovascular events and death in subjects with earlier stages of CKD,22 and among subjects with coronary artery disease, but intact renal function.37

At present, it remains controversial whether FGF-23 should be considered as a risk marker, or an independent risk factor.

In the former case, elevated FGF-23 levels would represent an ‘innocent bystander’, reflecting a high phosphate burden. This assumption is supported by our pathophysiological understanding of oral phosphate intake and hyperphosphataemia as major inducers of FGF-23 expression, and by the correlation of FGF-23 levels and serum phosphate consistently found across diverse patient cohorts.14 Notably, FGF-23 levels remained associated with LV hypertrophy in previous trials,23,24 and with LV function in the present cohort even after adjustment for serum phosphate levels. Similarly, in prospective studies, higher FGF-23 independently predicted cardiovascular events and mortality after adjusting for serum phosphate22,33,34 and even outperformed serum phosphate as an outcome predictor.21,22

These epidemiological findings nonetheless do not preclude phosphorus from being the pathophysiological mediator of cardiac hypertrophy and dysfunction in subjects with higher FGF-23 levels. Given their weak correlation with oral phosphate intake38 and their high circadian variability,39 serum phosphate levels are only crude indicators of total phosphate burden. Measuring FGF-23 levels might therefore allow us to identify those subjects who have a positive phosphate burden, despite seemingly normal spot serum phosphate values. This concept may be viewed in analogy to glycaemic control in diabetic patients, which is better reflected by Hba1c values than by a single plasma glucose measurement.40

However, we cannot exclude that FGF-23, beyond being a biomarker of phosphate burden, might exert a direct untoward effect on the myocardium, although the absence of its co-factor Klotho in myocardial cells stands against this hypothesis. At least, increased FGF-23 levels may indirectly cause harm by inducing hypovitaminosis D via inhibition of 1α-hydroxylase, given the large body of evidence for vitamin D deficiency as a cardiovascular risk factor.41

Whichever the underlying pathophysiological pathways might be the notion of a close links between structural heart disease and FGF-23 is further supported by the novel finding of an association between atrial fibrillation and FGF-23 levels.

Interestingly, we found no association between the prevalence of coronary artery disease and FGF-23 levels. This finding is in accordance with data from Roos et al.,42 who did not find higher coronary artery calcification in subjects with intact renal function and elevated FGF-23 levels. Among CKD patients, Gutierrez et al.34 reported an increased risk for coronary artery calcification in the highest tertile of FGF-23 at univariate analysis, which lost significance after adjustment for potential confounders. Moreover, in a large Swedish cohort of elderly subjects, FGF-23 levels were associated with endothelial dysfunction,43 but not with the presence of stenosis of at least one major vessel section at whole body magnetic resonance angiography in a multivariate analysis.44

As a limitation, we assessed LV hypertrophy via electrocardiography. Compared with echocardiography, ECG has a moderate sensitivity, but a high specificity for detecting LV hypertrophy.30 Nonetheless, focusing our work on LV function, we deliberately decided to choose ventriculography rather than echocardiography as the imaging method in the HOM SWEET HOMe trial.

Owing to its nature as a cohort study, our trial can only show associations, but not prove causal relationship. Thus, cause and consequence cannot be dissolved, and a myocardial origin of elevated FGF-23 in patients with prevalent myocardial dysfunction cannot be ruled out. In line, myocardial synthesis of FGF-23 has been reported,45 albeit in low quantity. We therefore hope that our observational data might stimulate prospective clinical studies and laboratory investigations which shall help to discern the temporal relation between elevation of FGF-23 and development of myocardial failure.

Despite the prognostic benefits achieved by modern pharmacological intervention—including inhibitors of the renin–aldosterone system and BBs—long-term outcome in patients with CKD remains grim. It is therefore tempting to speculate that a positive phosphate balance with elevated levels of FGF-23 might become a new therapeutic target in these patients. First pioneering trials in both rodents46 and humans47,48 found phosphate binders and calcimimetics49 to lower FGF-23 levels in patients with and without CKD, and larger trials are currently ongoing.

Conclusions

We report for the first time an association between elevated levels of the phosphatonin FGF-23 and LV function in a large cohort study. This findings is complemented by the association between FGF-23 and atrial fibrillation. Future clinical and laboratory studies should aim to elucidate the exact pathophysiological pathways of elevated FGF-23 in CKD. Subsequently, pharmacological interventions aiming at normalizing positive phosphate balance and lowering FGF-23 might become new therapeutic options in treatment of CKD.

Acknowledgements

We are grateful to Martina Wagner and Marie-Theres Blinn for their excellent technical assistance.

Conflict of interest: S.S. has received travel grants from Shire. D.F. has received speaker fees from Shire, Genzyme and Amgen. G.H.H. has received travel grants and speaker fees from Shire.

Footnotes

  • Both authors contributed equally to the submitted work.

References

View Abstract