|Home | About | Journals | Submit | Contact Us | Français|
Fibroblast growth factor 23 (FGF23) is elevated in chronic kidney disease and associated with increased mortality, but data on FGF23 in humans with acute kidney injury (AKI) are limited. Here we tested whether FGF23 levels rise early in the course of AKI following cardiac surgery, and if higher postoperative FGF23 levels are independently associated with severe AKI and adverse outcomes. Plasma C-terminal FGF23 (cFGF23) levels were measured preoperatively, at the end of cardiopulmonary bypass, and on postoperative days 1 and 3 in 250 patients undergoing cardiac surgery. We also measured intact FGF23 (iFGF23), parathyroid hormone, phosphate, and vitamin D metabolites in a subgroup of 18 patients with severe AKI and 18 matched non-AKI controls. Beginning at the end of cardiopulmonary bypass, cFGF23 levels were significantly and consistently higher in patients who developed AKI compared to those who did not. The early increase in cFGF23 predated changes in other mineral metabolites. The levels of iFGF23 also increased in patients who developed severe AKI, but the magnitude was lower than cFGF23. In analyses adjusted for age, preoperative eGFR, and cardiopulmonary bypass time, higher cFGF23 levels at the end of cardiopulmonary bypass were significantly associated with greater risk of severe AKI and the need for renal replacement therapy or death. Thus, cFGF23 levels rise early in AKI following cardiac surgery, and are independently associated with adverse postoperative outcomes.
Acute kidney injury (AKI) is one of the most devastating complications of cardiac surgery. As defined by consensus criteria, AKI occurs in up to 30% of patients following cardiac surgery and is associated with a 6 to 18-fold increased risk of death compared to individuals who do not develop AKI (1–4). Severe AKI that requires dialysis occurs in ~1% of patients who undergo cardiac surgery and is associated with in-hospital mortality rates that exceed 40% (1). Novel markers of early AKI and novel therapeutic targets are thus needed to improve clinical outcomes following cardiac surgery.
Disordered mineral metabolism is a common complication of chronic kidney disease (CKD), but it has been studied less extensively in AKI. Decreased 1,25-dihydroxyvitamin D levels and hypocalcemia associated with reduced or elevated parathyroid hormone (PTH) levels have been reported in small studies of patients with established AKI (5–8), but the mechanisms are poorly defined. Elevated levels of the osteocyte-derived hormone, fibroblast growth factor 23 (FGF23), may contribute to some of these laboratory abnormalities, since FGF23 normally stimulates urinary phosphate excretion and inhibits activation of 25-hydroxyvitamin D (9). In the setting of CKD, FGF23 rises early in the course of disease (10), contributes to development of secondary hyperparathyroidism (11), and is strongly associated with increased risks of cardiovascular disease (12–14) and death (15, 16).
Considerably less is known about FGF23 in AKI. In an animal study of AKI induced by either folate or glycerol injection, we reported that FGF23 levels rose soon after the onset of renal injury, independently of changes in serum phosphate, calcium, or PTH- or vitamin D-dependent signaling (17). In the same study, pilot data from 14 patients undergoing cardiac surgery revealed similar findings (17).
The primary purpose of this study was to comprehensively define the kinetics of changes in FGF23 and associated mineral metabolites before and soon after AKI onset in a larger prospective cohort study of adults undergoing cardiac surgery. We tested the hypotheses that FGF23 levels rise early in the course of AKI, and that higher postoperative levels are independently associated with an increased risk of severe AKI and other postoperative complications.
Cardiac surgery is an ideal clinical model to investigate the evolution of changes in FGF23 and other mineral metabolites before and during human AKI because the timing of renal injury can be pinpointed. We enrolled 250 patients who underwent cardiac surgery and were at high risk of developing postoperative AKI but had no evidence of AKI at baseline. The median (interquartile range [IQR]) age was 79 (72–83) years, the median (IQR) baseline estimated glomerular filtration rate (eGFR) was 50 (41–66) ml/min/1.73m2, and the median (IQR) baseline cFGF23 level was 124 (74–317) RU/ml. Additional baseline and operative characteristics are shown in Table 1.
AKI was defined as an increase in serum creatinine ≥ 0.3 mg/dl within 48 hours or ≥ 50% in 7 days (18). Patients meeting criteria for AKI were further classified as having mild or severe AKI. Mild AKI was defined as an increase in serum creatinine less than two times the baseline; severe AKI was defined as doubling of serum creatinine or need for renal replacement therapy (RRT). Using these definitions, we identified incident AKI in 73 (29%) patients, 55 (22%) of which were mild and 18 (7%) of which were severe.
Levels of cFGF23 over time were significantly higher in patients who developed AKI compared to those who did not (P<0.001 for global comparison of cFGF23 curves in patients with or without AKI, Figure 1A). Evaluation of cFGF23 levels at individual time points revealed that levels were similar at baseline, became significantly increased at the end of cardiopulmonary bypass in those with AKI compared to those without AKI (P≤0.01), and remained significantly higher on postoperative (POD) days 1 (P<0.001) and 3 (P<0.001) (Figure 1A). Baseline cFGF23 levels were higher among patients with pre-existing CKD, defined as an eGFR<60 ml/min/1.73m2, versus those without CKD (median [IQR] 140 [79–349] versus 98 [69–173] RU/ml, P<0.01). However, postoperative cFGF23 levels increased in similar patterns among patients who developed AKI with and without pre-existing CKD (P>0.05 for interaction between pre-existing CKD status, postoperative AKI group, and time; Figure 1B and 1C).
Figure 2 shows plasma cFGF23, serum creatinine, and urinary tubular injury markers in patients with no AKI, mild AKI, and severe AKI. We found a graded association between severity of AKI and cFGF23 levels over time (P<0.001, Figure 2A). Among patients with severe AKI, we found rapid and marked increases in cFGF23 levels such that median levels at the end of cardiopulmonary bypass were ~10-fold higher than preoperative levels (P=0.04), and median POD1 levels were ~40-fold higher (P<0.01). Among patients without AKI, median cFGF23 levels increased by ~3 fold at the end of cardiopulmonary bypass (P<0.001) and ~4-fold on POD1 (P<0.001). The group of patients with mild AKI had intermediate elevations in cFGF23 levels between the groups with no AKI and severe AKI. Serum creatinine (Figure 2B) and urinary neutrophil gelatinase-associated lipocalin (NGAL) (Figure 2C) showed a similar graded association, with higher levels associated with greater severity of AKI. Urinary kidney injury molecule-1 (KIM-1) and n-acetyl-β-(D)-glucosaminidase (NAG) were not associated with AKI (Figure 2D and 2E).
To further evaluate FGF23 levels in AKI, we measured levels of iFGF23 in a nested case-control subgroup analysis of the 18 patients with severe AKI and 18 patients without AKI, who were one-to-one matched on age (± 10 years) and preoperative eGFR (± 10 ml/min/1.73m2). In this subgroup analysis, longitudinal levels of both cFGF23 (P<0.001) and iFGF23 (P=0.04) were significantly higher in patients with AKI than those without AKI (Figure 3A and 3B); however, the magnitude of increase was far greater for cFGF23, as demonstrated by the increased ratio of cFGF23 to iFGF23 (P=0.04, Figure 3C). Preoperatively, the ratio of cFGF23 to iFGF23 was similar among patients who developed AKI compared to those who did not (median [IQR] 3.6 [1.9–17.6] and 3.0 [0.9–14.2], P=0.84), but by POD3 the ratio increased to a median of 81.2 (IQR 52.9–144.1) among patients with AKI compared to 24.4 (IQR 7.3–58.8) among those without AKI (P=0.08). These data indicate more marked increases in cFGF23 versus iFGF23 in AKI following cardiac surgery.
To compare the time course of changes in FGF23 levels versus other mineral metabolites following cardiac surgery, we also measured levels of PTH, phosphate, and vitamin D metabolites in the nested case-control subgroup. Longitudinal PTH, 25-hydroxyvitamin D, and 24,25-dihydroxyvitamin D levels were similar in patients with versus without AKI (Figure 4A, 4C, and 4E). Longitudinal phosphate levels were higher (P=0.02) and 1,25-dihydroxyvitamin D levels were lower (P=0.001) in patients with versus without AKI. When phosphate and 1,25-dihydroxyvitamin D levels were evaluated at individual time points, only POD3 levels were significantly different between groups (Figure 4B and 4D). These data suggest that the elevations in cFGF23 and iFGF23 that were evident by the end of cardiopulmonary bypass and on POD1 among patients with AKI occurred independently of any alterations in PTH, phosphate, or vitamin D metabolites and predated the modest changes in phosphate and 1,25-dihydroxyvitamin D levels observed on POD3.
Next, we compared the association between cFGF23 levels and risk of severe AKI following cardiac surgery relative to other proposed AKI biomarkers at the end of cardiopulmonary bypass and on POD1. Table 2 shows unadjusted and adjusted odds ratios and the area under the receiver operating characteristic curve (AUC-ROC) for risk of incident severe AKI according to levels of cFGF23 and urinary tubular injury markers. In unadjusted analyses, higher cFGF23 levels at the end of cardiopulmonary bypass and on POD1 were associated with an increased risk of severe AKI. These associations were slightly attenuated but remained significant when adjusted for age, preoperative eGFR, and cardiopulmonary bypass time (Table 2). In contrast, urinary NGAL and KIM-1 were associated with severe AKI in adjusted analyses on POD1, but not at the end of cardiopulmonary bypass, while urinary NAG was neither associated with severe AKI at the end of cardiopulmonary bypass nor on POD1 in adjusted analyses. Based on nearly complete non-overlap of their respective confidence intervals, the strengths of association and AUC-ROCs appeared greater at both time points for cFGF23 than the urinary injury biomarkers tested (Table 2).
Table 3 shows unadjusted and adjusted odds ratios that quantify risks of adverse postoperative outcomes according to cFGF23 levels at the end of cardiopulmonary bypass. In unadjusted analyses, cFGF23 levels were associated with the composite of renal replacement therapy or in-hospital mortality (RRT/death), postoperative vasopressor requirement, sepsis, and myocardial injury. After adjusting for age, preoperative eGFR, and cardiopulmonary bypass time, each of these associations remained significant except postoperative myocardial injury.
Median (IQR) duration of mechanical ventilation and hospital length of stay was 1 (1–2) and 13 (9–20) days, respectively. In both unadjusted and adjusted analyses, higher cFGF23 levels were associated with fewer ventilator- and hospital-free days (Table 3), implying longer duration of mechanical ventilation and hospital length of stay.
In this prospective cohort study we report that plasma cFGF23 levels rise early in patients who develop AKI following cardiac surgery, reaching concentrations four-fold greater than baseline by the end of cardiopulmonary bypass. The increases in cFGF23 were independent of alterations in PTH, phosphate, or vitamin D metabolites. These findings are consistent with our prior report in a murine model of AKI (17) in which we reported a similarly large magnitude of FGF23 elevation beginning early in the course of AKI that occurred independently of changes in other mineral metabolites. Further, we found that higher cFGF23 levels at the end of cardiopulmonary bypass are independently associated with increased risk of developing severe AKI and several other important postoperative complications including RRT/death and longer hospital length of stay. These findings suggest that elevated cFGF23 may be an early and novel marker of AKI and other adverse clinical outcomes following cardiac surgery.
Multiple studies demonstrated that elevated FGF23 levels are associated with major cardiovascular events and mortality in CKD and end stage kidney disease (12–16). However, only few studies with limited sample sizes evaluated FGF23 levels in AKI (19, 20). In a pilot study, we reported higher cFGF23 levels in 30 patients with established AKI compared to 30 controls without AKI (8). Additional preliminary data on cFGF23 levels were obtained from 14 adult patients undergoing cardiac surgery, in whom we found a 15.9-fold increase in cFGF23 on POD1 in patients with AKI versus without AKI (N=4 and N=10, respectively) (17).
Interestingly, we found elevations in cFGF23 levels in patients both with and without AKI, though the magnitude of rise was far greater in patients with AKI and even greater in severe AKI. The finding that cFGF23 increased even in patients without AKI suggests that these patients may have experienced subclinical renal injury or, alternatively, that cFGF23 may function as an acute-phase reactant after exposure to operative stress. The latter mechanism is consistent with elevated postoperative cFGF23 levels among patients undergoing hip arthroplasty (N=55), which occurred independent of AKI status (21), and a recent report demonstrating that acute inflammation markedly increases FGF23 levels in mice, particularly cFGF23 levels, as we observed here (22).
In addition to AKI, we also evaluated the association between cFGF23 and several secondary end points and found significant independent associations with RRT/death, persistent vasopressor requirement, sepsis, and longer duration of mechanical ventilation and hospital length of stay. Thus, as in CKD and end stage kidney disease, cFGF23 appears to be associated with adverse outcomes following cardiac surgery. Whether cFGF23 is simply a severity-of-illness marker or directly contributes to adverse outcomes is a key question that could not be answered by this observational study. Elevated FGF23 has been implicated in the pathogenesis of left ventricular hypertrophy in CKD (23, 24), thus raising the possibility that other harmful effects of FGF23 could contribute to our findings.
The association we found between cFGF23 and postoperative sepsis is particularly intriguing given the known inhibitory effects of FGF23 on vitamin D metabolite activation (9, 25, 26), the important effects of vitamin D metabolites on immune function (27–29), and abundant epidemiologic data linking decreased vitamin D metabolite levels with increased incidence and severity of sepsis (30, 31). The higher FGF23 levels coupled with lower 1,25D levels among patients with AKI raises the possibility of FGF23-mediated reduction in 1-alpha hydroxylase activity, a process that occurs both within the kidney and in extra-renal tissues (26). Decreased 1,25D levels, in turn, may at least partially explain the increased risk of adverse outcomes that we observed in association with elevated FGF23 levels.
The immunometric assay for measurement of cFGF23 uses two antibodies directed against different epitopes within the C-terminal portion of FGF23, and thus detects both the intact hormone and C-terminal cleavage products. In contrast, the iFGF23 assay detects only the intact molecule (32). Although both cFGF23 and iFGF23 rose in AKI, cFGF23 levels were approximately 25 to 75-fold higher, suggesting that concentrations of C-terminal FGF23 fragments were especially elevated in AKI. This could theoretically result from reduced clearance of FGF23 fragments in AKI, or reduced processing of FGF23 by the injured kidney in AKI, which was demonstrated in a recent bilateral nephrectomy model in rats (33). Arguing against reduced clearance of FGF23 fragments, we found that cFGF23 levels were also increased among patients without AKI, albeit to a lesser extent. Furthermore, AKI in mice was associated with only mild impairment in the clearance of injected recombinant FGF23 (17). An alternative explanation is that both FGF23 production and cleavage were markedly increased postoperatively. Among patients without AKI, a moderate increase in FGF23 production accompanied by commensurately increased FGF23 cleavage would result in normal iFGF23 levels and absence of hypophosphatemia. In contrast, among patients who developed AKI, markedly increased FGF23 production with less of a relative increase in FGF23 cleavage could explain our observation of a major increase in cFGF23 levels that was accompanied by a moderate increase in iFGF23 levels. These hypotheses are supported by the similar cFGF23 and iFGF23 responses in mice exposed to acute inflammatory stimuli (22).
We acknowledge several limitations, including single-center and observational design. Our cohort was designed to capture patients at high risk of AKI after cardiac surgery, and thus may not be generalizable to healthier patients. We did not measure all potentially relevant biomarkers such as soluble Klotho, vitamin D-binding protein, and inflammatory biomarkers. We did not have access to data on urine output, which might have allowed earlier identification of AKI and would have been another clinical biomarker with which to compare cFGF23. PTH, phosphate, and vitamin D metabolites were measured only in a subgroup, and thus could not be included in mediation analyses of cFGF23 and adverse outcomes. We could not discern the mechanisms responsible for increased FGF23 production or cleavage in this observational study. Finally, we cannot exclude residual confounding from unmeasured variables as an explanation for the associations between cFGF23 and adverse outcomes.
In conclusion, plasma cFGF23 levels rise early in patients with AKI following cardiac surgery and are independently associated with severe AKI and other adverse postoperative outcomes. These changes appear to be independent of changes in PTH, phosphate, or vitamin D metabolites. Whether FGF23 has acute toxic effects on nontraditional target organs such as the cardiovascular or immune systems and may thereby contribute directly to adverse outcomes in AKI as it does in CKD is an intriguing possibility that will require additional study.
We conducted a prospective cohort study in 250 patients who underwent cardiac surgery. Patients were recruited from Brigham and Women’s Hospital between August, 2007 and March, 2012. All patients provided written informed consent and all protocols were approved by our hospital’s Institutional Review Board.
The inclusion and exclusion criteria were chosen to capture patients at high risk of AKI and other adverse outcomes following cardiac surgery. Inclusion criteria were baseline eGFR ≤ 30 ml/min/1.73m2 or any two of the following: baseline eGFR 31–60 ml/min/1.73m2; diabetes mellitus; left ventricular ejection fraction ≤ 40%; previous cardiac surgery; combined coronary artery bypass/valve procedure; urgent procedure; and preoperative intra-aortic balloon pump. Exclusion criteria were preoperative AKI (defined as a 0.3 mg/dL rise in serum creatinine over 24 hours or a 0.5 mg/dL rise over 48 hours) (34); recent aminoglycoside use; serum creatinine > 4.5 mg/dL; end stage kidney disease receiving dialysis; renal transplantation; and pregnancy.
We collected and stored plasma and urine aliquots at −80ºC within 2 hours of collection. We measured analytes in plasma samples at four time points: preoperatively; at the end of cardiopulmonary bypass; and on POD1 and POD3. We measured analytes in urine samples at the first three of these four time points.
Investigator DEL adjudicated all outcomes by reviewing electronic medical records, and was blinded to all study measurements at the time of adjudication. The pre-specified primary endpoint was incident severe AKI, defined as doubling of serum creatinine postoperatively or need for RRT, corresponding to stages 2 and 3 of the criteria established by the Kidney Disease Improving Global Outcomes Work Group (KDIGO) (18).
Secondary end points were incident AKI, defined as an increase in serum creatinine ≥ 0.3 mg/dl within 48 hours or ≥ 50% in 7 days (18); the composite of need for RRT or in-hospital mortality (RRT/death); postoperative vasopressor requirement beyond 24 hours; postoperative sepsis, defined according to consensus defintion (35); and postoperative myocardial injury. We defined postoperative myocardial injury using cut points for creatine kinase myocardial b fraction (CK-MB) specific for valvular or non-valvular surgery, given intrinsic differences in degree of surgical trauma induced by the procedures (36). For non-valvular surgery, postoperative myocardial injury was defined as a greater than 10-fold rise in CK-MB above the upper reference limit of 5 ng/ml within 48 hours of surgery (37). For valvular surgery, a greater than 20-fold rise in CK-MB within 48 hours defined myocardial injury.
Additional end points included duration of mechanical ventilation and hospital length of stay. To avoid the confounding effect of mortality, we calculated ventilator-free days and hospital-free days as 28 minus the number of ventilator-dependent days or hospitalization days, respectively, assuming survival to 28 days or discharge from the hospital. Patients who died before 28 days were assigned a score of zero (38).
We measured plasma C-terminal FGF23 (cFGF23) levels in duplicate in all 250 patients at all four time points using a commercial ELISA kit (Immutopics, San Clemente, CA). We also performed additional analyses in a nested case-control subcohort of 18 participants with severe AKI and 18 participants without AKI. Cases were matched 1:1 to controls based on age (± 10 years) and preoperative eGFR (± 10 ml/min/1.73m2). In this nested case-control subcohort we measured plasma levels of the following analytes at all four time points: intact FGF23 (iFGF23), intact PTH, phosphate, 25-hydroxyvitamin D, 1,25-dihydroxyvitamin D, and 24,25-dihydroxyvitamin D. iFGF23 was measured in duplicate using a commercial ELISA kit (Immutopics, San Clemente, CA). PTH was measured using a two-site immunoassay on a Beckman Access2 automated immunoassay analyzer. Phosphate was measured using the phosphomolybdate method monitored by timed-rate calorimetry on a Beckman DxC automated chemistry analyzer. Vitamin D metabolites were measured using immunoaffinity enrichment and liquid chromatography-tandem mass spectrometry (39). Serum creatinine was measured for clinical purposes using a modified kinetic Jaffe method.
We measured urinary levels of neutrophil gelatinase-associated lipocalin (NGAL), kidney injury molecule-1 (KIM-1), n-acetyl-β-(D)-glucosaminidase (NAG), and creatinine in duplicate in all 250 participants. We normalized urinary injury markers to the urinary creatinine concentration to account for the influence of urinary dilution on biomarker concentrations. NGAL and KIM-1 were measured using a microbead-based sandwich ELISA. NAG was measured using an enzymatic assay (Roche, San Francisco, CA). Creatinine was measured using a Randox Daytona bench top analyzer.
We calculated interassay coefficients of variation (CVs) for each assay using blinded replicate samples from study patients. Interassay CVs were as follows: cFGF23 (6.1%), iFGF23 (9.8%), PTH (2.4%), phosphate (0.2%), 25D (7.2%), 1,25D (14.7%), and 24,25D (3.9%).
Statistical analysis was performed with SAS Version 9.4 (Cary, NC), and MedCalc Version 13.3.1 (Ostend, Belgium). Data are reported as median and IQR (25th–75th percentiles). Baseline/operative characteristics were compared in patients with no AKI, mild AKI, and severe AKI using Kruskal Wallis and χ2 tests for continuous and categorical variables, respectively. We used mixed linear models for repeated measures to test for statistical differences in cFGF23 levels over time between patients with AKI versus those without AKI. As the dependent variable in the mixed models, cFGF23 levels were natural log-transformed due to their skewed distribution. To enhance interpretability, we present the untransformed values graphically. In the models, time represented the repeated measures factor, individuals were represented as a random-effects term, and AKI status (absent or present, or absent, mild, or severe) was treated as a fixed-effects factor. If the P-value comparing cFGF23 curves in patients with AKI versus without AKI was <0.05, we also analyzed between-group differences in cFGF23 levels at individual time points. Comparison of cFGF23 levels at individual time points was assessed using the Wilcoxon rank sum test (for the two-group comparison of AKI versus no AKI) or Spearman’s rank correlation coefficient (for the correlation between presence/severity of AKI and cFGF23 levels). We used an identical analytic strategy to test for differences in other analytes based on AKI status, first testing for overall effects over time followed by testing individual time points only if the overall effect was significant. A similar strategy was also applied to the nested case-control subcohort of 18 participants with severe AKI and 18 participants without AKI. Specifically, we first used mixed linear models for repeated measures to test for longitudinal differences in analyte levels in patients with AKI versus without AKI. If the P-value comparing analyte curves in patients with AKI versus without AKI was <0.05, we also analyzed between-group differences at individual time points using the Wilcoxon signed-rank test for paired data.
Logistic regression was used to assess the association between cFGF23 and urinary injury biomarker levels at the end of cardiopulmonary bypass and on POD1 with incident severe AKI. Multivariate models were used to adjust for age, preoperative eGFR, and cardiopulmonary bypass time >120 min. Logistic and linear regression were used to assess the association between cFGF23 levels at the end of cardiopulmonary bypass and secondary binary and continuous end points, respectively, adjusting for the same covariates as above. In these models, cFGF23 and urinary injury markers were both natural log-transformed, due to their skewed distribution, and normalized to one standard deviation to allow comparison across biomarkers. All comparisons are two-tailed, with P<0.05 considered significant.
This work was supported by grants R21DK100754 and K24DK093723 (to MW) and F32DK100040 (to DEL) from the National Institute of Diabetes and Digestive Kidney Diseases.
ES reports previous consulting agreements with Alere and Abbvie. JVB is a coinventor on KIM-1 patents that are assigned to Partners HealthCare and licensed by Partners HealthCare to Johnson & Johnson, Sekisui Chemical Company, Biogen Idec, Astute Medical, and a number of research reagent companies. MW has received research support or honoraria from Keryx, Luitpold, Pfizer and Shire. SSW served as a consultant to Abbvie, CVS Caremark, Harvard Clinical Research Institute, and Takeda, and has received grants from the National Institute of Diabetes and Digestive Kidney Diseases, Genzyme, Merck, Otsuka, Pfizer, and Satellite Healthcare.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.