Original Report: Patient-Oriented, Translational Research American
Journal of
Nephrology
Am J Nephrol 2013;37:462–471 DOI: 10.1159/000350537
Received: November 26, 2012 Accepted: March 6, 2013 Published online: April 27, 2013
Determinants of Fibroblast Growth Factor-23 and Parathyroid Hormone Variability in Dialysis Patients Ting Jia a, b Abdul Rashid Qureshi a, b Vincent Brandenburg g Markus Ketteler h Peter Barany a Olof Heimburger a Fredrik Uhlin d, e Per Magnusson f Anders Fernström d, e Bengt Lindholm b Peter Stenvinkel a, c Tobias E. Larsson a, c a
Division of Renal Medicine, Department of Clinical Science, Intervention and Technology, and b Renal Medicine and Baxter Novum, Karolinska Institutet, and c Department of Nephrology, Karolinska University Hospital, Stockholm, d Department of Medical and Health Science, Faculty of Health Sciences, Linköping University, e Department of Nephrology, County Council of Östergötland, and f Division of Clinical Chemistry, Department of Clinical and Experimental Medicine, Faculty of Health Sciences, Linköping University, Linköping, Sweden; g Department of Cardiology, Aachen University Hospital, Aachen, and h Division of Nephrology, Klinikum Coburg, Coburg, Germany
Key Words Fibroblast growth factor-23 · FGF23 · Parathyroid hormone · PTH · Chronic kidney disease
Abstract Background/Aims: Treatment strategies for abnormal mineral metabolism in chronic kidney disease are largely based on achieving target ranges of biomarkers that vary considerably over time, yet determinants of their variability are poorly defined. Methods: Observational study including 162 patients of three dialysis cohorts (peritoneal dialysis, n = 78; hemodialysis, n = 49; hemodiafiltration, n = 35). Clinical and biochemical determinants of parathyroid hormone (PTH) and fibroblast growth factor-23 (FGF23) variability were analyzed in the peritoneal dialysis cohort. All cohorts were used for comparison of PTH and FGF23 intra-subject variability (intra-class correlation), and their intra-subject variability in different modes of dialysis was explored. Results: High PTH variability was independently associated
© 2013 S. Karger AG, Basel 0250–8095/13/0375–0462$38.00/0 E-Mail
[email protected] www.karger.com/ajn
with lower 25-hydroxyvitamin D concentration and factors of lipid and glucose metabolism, whereas high FGF23 variability was mainly associated with lower baseline serum phosphorous. These results were consistent in multivariate and sensitivity analyses. The intra-subject variability of FGF23 was lower than for PTH irrespective of dialysis mode. Conclusions: Baseline vitamin D status and serum phosphorous are independent determinants of the longitudinal variation in PTH and FGF23, respectively. The clinical utility of FGF23 measurement remains unknown, yet it appears favorable based on its greater temporal stability than PTH in dialysis patients. Copyright © 2013 S. Karger AG, Basel
Introduction
Patients with end-stage renal disease (ESRD) receiving dialysis treatment generally suffer from abnormalities in mineral metabolism including hyperphosphatemia and Tobias E. Larsson Clinical Research Center (KFC), 6th Floor, Novumhuset Karolinska University Hospital Huddinge SE–14186 Stockholm (Sweden) E-Mail tobias.larsson @ ki.se
secondary hyperparathyroidism [1, 2]. These biochemical deviations are part of a syndrome termed chronic kidney disease-mineral and bone disorder (CKD-MBD), which also entails vascular calcification, bone disease and mortality [3]. The optimal treatment of CKD-MBD remains controversial since therapeutic strategies are predominantly based on monitoring surrogate biomarkers including calcium, phosphorous and parathyroid hormone (PTH) to achieve set target ranges defined by clinical guidelines such as KDIGO (Kidney Disease: Improving Global Outcomes) [3] and NKF-KDOQI (National Kidney Foundation Disease Outcomes Qualitative Initiative) [4]. Monitoring circulating PTH is a cornerstone in clinical care of CKD-MBD due to the established relation between PTH, cardiovascular complications and mortality [5, 6]. High PTH concentrations are associated with increased bone turnover and release of calcium and phosphorous from bone, which could precipitate in the vasculature and trigger an array of cellular mechanisms that promote vascular calcification [7]. On the other hand, inappropriately low PTH concentrations (e.g. in relation to the degree of skeletal and renal PTH resistance) are associated with low bone turnover and attenuated skeletal buffering of serum minerals and subsequent exacerbation of vascular calcification [8]. Consequently, therapeutic decisions regarding the usage and dosing of phosphate binders, vitamin D analogues and calcimimetics are commonly justified by dynamic changes in PTH. In addition to PTH, fibroblast growth factor-23 (FGF23) is another pivotal factor in CKD-MBD [9]. High FGF23 concentrations are associated with adverse cardiovascular outcomes across the spectrum of CKD [10– 14], and routine clinical measurement of FGF23 has been proposed as a screening tool for earlier identification of individuals with emerging CKD-MBD [10]. Combined monitoring of FGF23 and PTH may improve current treatment algorithms for CKD-MBD, although any clinical benefits are yet to be unraveled. There are several fundamental limitations of using biomarkers as groundwork for clinical decision-making. A major caveat in CKD-MBD is the large temporal intraindividual variation of PTH, implying that treatment strategies could be modified based on biological variability rather than true pathophysiological processes [15]. A better understanding of clinical determinants of PTH variability is therefore important to identify individuals with increased likelihood of high variability and subsequent risk of misclassification. Herein, we evaluated determinants of intra-subject variability of PTH and FGF23 in a cohort of peritoneal
dialysis (PD) patients and performed a comparative analysis of PTH and FGF23 variability in patients receiving either PD, hemodialysis (HD) or online hemodiafiltration (HDF).
FGF23 and PTH Variability
Am J Nephrol 2013;37:462–471 DOI: 10.1159/000350537
Materials and Methods PD Cohort All 84 patients in a cohort named ‘Mapping of Inflammation Markers in Chronic Kidney Disease, Part 2 (MIMICK)’ receiving PD treatment at Karolinska University Hospital in Huddinge, Sweden, were invited to participate in the study. All participants were prevalent PD patients who had been on continuous ambulatory peritoneal dialysis or automated peritoneal dialysis for at least 3 months. Patients were recruited from March 2008 to April 2011. The study duration was a 3-month observation period. All PD patients were clinically stable without any recorded mortality, hospitalizations, severe cardiovascular events, cancer or acute bleeding throughout the study period. Prescription of PD and drugs for treatment of CKD-MBD (phosphate binders, vitamin D analogues, cinacalcet) were unchanged throughout the study. Biochemical parameters were collected once weekly. Patients who had less than three separate time-points of PTH measurements (n = 6) were excluded. Thus, 78 patients (25 women, 32%; age 63 ± 13 years) were included in the present study. The study was approved by the local ethical committee (Dnr: 273/94) and all participants provided written informed consent. The residual GFR was measured by calculating the mean of the urinary clearance of urea and creatinine GFR [16]. The CKD etiology was chronic glomerulonephritis (14%), diabetic nephropathy (13%), polycystic kidney disease (9%), vascular disease/nephrosclerosis (12%), and miscellaneous/unknown cause (52%). Medication used was as follows: angiotensin-converting enzyme inhibitors or angiotensin II receptor blockers (43%), β-blockers (71%), α-blockers (7%), calcium-channel blockers (32%), diuretics (87%), statins (49%), vitamin D receptor activators (83%), phosphate binders (70%), calcimimetics (13%) and warfarin (7%). All blood samples were drawn in the morning after overnight fasting to minimize the impact of food intake and diurnal variation. Serum samples were immediately analyzed for standard serum analytes, PTH, 25-hydroxyvitamin D [25(OH)D] and 1,25-dihydroxyvitamin D [1,25(OH)2D] and then stored at –80 ° C until analysis of FGF23. Biochemical measurements were performed at the Department of Clinical Chemistry and the Clinical Research Center at Karolinska University Hospital in Huddinge. Assay analyses were performed with the same batch in order to minimize analytical variation. To minimize analytical variation, analyses of a specific parameter were performed using an assay with the same product number. 25(OH)D and 1,25(OH)2D were measured by commercially available two-site enzyme immunoassay EIA kits (Immunodiagnostic Systems, Sweden). Serum intact FGF23 was measured by enzyme-linked immune sorbent assay (ELISA) (Kainos Laboratories, Tokyo, Japan). PTH was analyzed in serum by an immunometric assay on an Immulite 1000 Analyzer (Siemens Healthcare Diagnostics, Los Angeles, Calif., USA) according to the instructions of the manufacturers. The between-batch CVs were 10). To study the within-subject variation of PTH and FGF23 in different modes of dialysis (PD, HD, HDF), we calculated the intraclass correlation (ICC) from estimates of between-subject (σ2b) and within-subject variance (σ2w), derived from two-way mixed effects models, using the following formula: σ2b/(σ2b + σ2w). Statistical significance was set at the level of p < 0.05. All statistical tests were performed using SAS statistical software, version 9.2 (SAS Institute, Inc., Cary, N.C., USA).
Results
Baseline Characteristics We focused on the PD cohort to examine predictors of PTH and FGF23 variability. The CVi for FGF23 was not normally distributed, thus lgFGF23 variation was used in subsequent analyses. Baseline characteristics according to the tertile of CVi of PTH and lgFGF23 are shown in table 1A and 1B. There was no difference with regard to exposure of activated vitamin D, cinacalcet or phosphatebinding treatment among variability groups of FGF23 and PTH (data not shown). Determinants of PTH and FGF23 Variability PTH variability was negatively correlated to baseline PTH (ρ = –0.27, p = 0.02) with a similar pattern for lgFGF (ρ = –0.32, p = 0.002), which may reflect regression to the mean. Significant correlations between PTH and lgFGF23 variability and biochemical markers determined by Spearman’s correlations and ROC curves are shown in table 2. Factors associated with high PTH variability were lower age, fat mass, fasting glucose, triglycerides, and Jia et al.
Table 1A. Clinical characteristics of PD patients according to tertile of FGF23 variability
Profiles/variables General Gender, female/male Age, years Davies score Creatinine, μmol/l Urea, mmol/l eGFR, ml/min/1.73 m2 CRP, mg/l MBD PTH, pg/ml FGF23, pg/ml Calcium, mmol/l Phosphorous, mmol/l 1,25(OH)2D, pg/ml 25(OH)D, ng/ml Low25(OH)D, % Nutrition PEW (SGA >1), % Fat mass, kg Lean mass, kg Bone mass, kg Body weight, kg Diabetes Diabetes, % Glucose, mmol/l HbA1c, % Insulin, pmol/l Lipid, mmol/l Triglycerides Total cholesterol HDL cholesterol LDL cholesterol Anemia Iron, μmol/l Ferritin, μg/l Transferrin, g/l Hemoglobin, g/l Thyroid TSH, mU/l T4, pmol/l T3, pmol/l T4/T3 ratio
Low CVi tertile (n = 26)
Median CVi tertile (n = 26)
High CVi tertile (n = 26)
p for trend
8/18 65.0 ± 14.1 1.3 ± 1.2 746 ± 193 21.0 ± 7.0 2.7 (1.4 – 4.6) 5.1 (2.2 – 19.9)
5/21 63.6 ± 15.0 1.3 ± 0.9 721 ± 184 21.1 ± 6.0 2.8 (1.8 – 4.4) 3.9 (1.4 – 7.8)
12/14 63.2 ± 12.8 1.4 ± 1.3 684 ± 152 19.4 ± 5.7 2.7 (1.2 – 5.1) 5.6 (1.3 – 13.4)
0.09 0.60 0.94 0.16 0.44 0.35 0.58
319 (141 – 423) 8,531 (3,169 – 18,063) 2.31 ± 0.18 1.88 ± 0.52 13 (9 – 23) 26 (15 – 34) 30
265 (106– 510) 8,012 (1,314 – 19,816) 2.29 ± 0.19 1.70 ± 0.46 13 (8.9 – 19) 24 (18 – 43) 30
200 (145 – 332) 2,770 (1,744 – 4,416) 2.30 ± 0.20 1.59 ± 0.44 10 (9 – 13) 25 (17 – 35) 29
0.40 0.004 0.89 0.02 0.04 0.90 0.94
36 24.8 ± 6.8 46.5 ± 6.9 2.4 ± 0.6 73.7 ± 10.2
36 25.6 ± 7.2 47.9 ± 10.6 2.6 ± 0.5 76.9 ± 13.2
50 22.2 ± 10.0 46.5 ± 7.3 2.4 ± 0.6 70.1 ± 15.1
0.30 0.32 0.61 0.73 0.16
24 5.9 ± 2.3 5.0 ± 1.0 55 (31 – 100)
16 6.2 ± 3.0 4.9 ± 0.8 79 (47 – 103)
35 6.0 ± 3.0 5.0 ± 1.1 41 (24 – 101)
0.37 0.51 0.40 0.60
2.13 ± 1.01 4.83 ± 1.39 1.27 ± 0.62 2.87 ± 1.08
1.79 ± 0.66 4.94 ± 1.20 1.24 ± 0.42 2.88 ± 1.02
1.96 ± 0.90 5.32 ± 1.20 1.38 ± 0.46 3.00 ± 1.07
0.56 0.38 0.17 0.68
11.8 ± 4.8 354 (119 – 537) 2.01 ± 0.41 118 ± 12
14.3 ± 5.1 333 (175 – 544) 1.93 ± 0.39 119 ± 11
13.5 ± 5.1 270 (179 – 526) 1.92 ± 0.38 119 ± 11
0.26 0.59 0.37 0.68
1.8 (1.2 – 2.9) 13.6 ± 3.5 3.8 ± 0.6 3.5 ± 0.8
1.8 (1.4 – 3.3) 13.5 ± 2.8 4.1 ± 0.7 3.5 ± 1.0
2.8 (1.9 – 3.3) 13.3 ± 2.8 3.7 ± 0.7 3.6 ± 0.9
0.03 0.64 0.62 0.32
Data are shown as mean ± SD, percentages or median (interquartile), as appropriate. CRP = C-reactive protein; T4 = thyroxine; T3 = triiodothyronine. CVi was defined as the coefficient of variation and calculated as the ratio between the SD and mean. Low 25(OH)D is defined as 1), % Fat mass, kg Lean mass, kg Bone mass, kg Body weight, kg Diabetes Diabetes, % Glucose, mmol/l HbA1c, % Insulin, pmol/l Lipid, mmol/l Triglycerides Total cholesterol HDL cholesterol LDL cholesterol Anemia Iron, μmol/l Ferritin, μg/l Transferrin, g/l Hemoglobin, g/l Thyroid TSH, mU/l T4, pmol/l T3, pmol/l T4/T3 ratio
Low CVi tertile (n = 26)
Median CVi tertile (n = 26)
High CVi tertile (n = 26)
p for trend
10/16 69.1 ± 9.5 1.3 ± 1.1 713 ± 184 19.7 ± 6.5 2.7 (1.4– 4.6) 6.4 (1.5 – 16.5)
4/22 61.2 ± 15.8 1.2 ± 1.0 756 ± 165 22.4 ± 5.9 2.8 (1.8 – 4.4) 3.7 (1.1 – 6.0)
11/15 62.5 ± 14.9 1.3 ± 1.3 701 ± 178 18.9 ± 6.0 2.7 (1.2 – 5.1) 5.1 (1.9 – 18.5)
0.84 0.06 0.81 0.72 0.40 0.35 0.87
305 (213 – 450) 3,169 (2,149 – 11,989) 2.32 ± 0.18 1.63 ± 0.42 10 (9 – 16) 26 (21 – 39) 20
300 (144 – 484) 7,003 (2,208 – 18,197) 2.29 ± 0.19 1.73 ± 0.45 13 (9 – 22) 24 (15 – 53) 25
201 (85 – 339) 4,383 (1,761 – 9,704) 2.30 ± 0.18 1.77 ± 0.61 9 (9 – 18) 24 (15 – 39) 36
0.02 0.89 0.15 0.02 0.57 0.18 0.07
16 27.5 ± 0.7 45.8 ± 6.4 2.5 ± 0.6 77.2 ± 10.2
38 26.2 ± 0.3 49.4 ± 7.6 2.4 ± 0.6 77.0 ± 11.3
62 19.4 ± 0.8 45.5 ± 7.3 2.4 ± 0.6 67.7 ± 13.9
0.001 0.03 0.87 0.55 0.001
22 6.3 ± 3.0 5.0 ± 1.0 67 (41 – 110)
22 5.9 ± 1.9 4.9 ± 1.0 73 (42 – 98)
25 5.8 ± 3.0 4.9 ± 1.1 38 (23 – 112)
0.80 0.15 0.08 0.15
2.39 ± 1.38 5.33 ± 1.16 1.22 ± 0.50 3.15 ± 1.16
1.80 ± 0.68 4.72 ± 1.39 1.34 ± 0.50 2.75 ± 0.97
1.83 ± 0.86 5.22 ± 1.17 1.29 ± 0.50 3.03 ± 1.00
0.04 0.59 0.40 0.71
12.9 ± 4.4 341 (228 – 515) 1.94 ± 0.40 121 ± 12
13.4 ± 5.0 392 (188 – 560) 2.05 ± 0.42 116 ± 11
13.5 ± 5.9 249 (115 – 561) 1.85 ± 0.45 117 ± 11
0.87 0.23 0.95 0.19
2.2 (1.4 – 2.7) 14.1 ± 3.5 3.9 ± 0.5 3.5 ± 0.8
2.2 (1.3 – 3.0) 13.9 ± 2.6 4.0 ± 0.7 3.5 ± 1.0
2.5 (1.5 – 3.4) 12.0 ± 2.6 3.8 ± 0.7 3.6 ± 0.9
0.34 0.18 0.48 0.32
See footnote in table 1A for details and abbreviations.
25(OH)D but higher insulin concentration and presence of PEW. In contrast, high FGF23 variability was more prevalent in females and associated with lower serum phosphorous but higher HDL cholesterol and thyroidstimulating hormone (TSH). 466
Am J Nephrol 2013;37:462–471 DOI: 10.1159/000350537
In a multivariate stepwise regression model including significant parameters from Spearman’s correlations, presence of PEW and lower fat mass predicted high PTH variability, whereas lower triglycerides and 25(OH)D concentrations were borderline significant. A lower seJia et al.
Table 2. Predictors of FGF23 and PTH variability as determined by Spearman’s correlations and ROC curves for selected parameters that significantly differed in their CVi for lgFGF23 and PTH at baseline
Dependents/profiles lgFGF23 variability General MBD Lipid Thyroid PTH variability General MBD Nutrition Diabetes Lipid
Variables
Correlation coefficient ρ
AUC
SE
Gender, female/male Phosphorous, mmol/l HDL cholesterol, mmol/l TSH, mU/l
0.23a –0.29a 0.19a 0.21a
0.64 0.72a 0.63 0.65
1.28 1.13 –0.53 –0.14
Age, years 25(OH)D, ng/ml PEW (SGA >1), % Fat mass, kg Glucose, mmol/l Insulin, pmol/l Triglycerides, mmol/l
–0.20a –0.20a 0.35b –0.36b –0.21a 0.20a –0.20a
0.61 0.75a 0.70a 0.71a 0.63 0.63 0.67
0.03 0.06 –1.18 0.07 0.07 0.003 0.53
All models were adjusted for age, gender and Davies scores. See footnote in table 1A for further details. a p < 0.05; b p < 0.01.
Table 3. Stepwise regression model showing significant predictors of variability for lgFGF23 and lgPTH in PD
patients Profiles
MBD Nutrition Lipid
Variables
Phosphorous, mmol/l 25(OH)D, ng/ml PEW (SGA >1), % Fat mass, kg Triglycerides, mmol/l
IgFGF23 variation
PTH variation
standard β value
p
–0.23
0.03
standard β value
–0.20 0.30 –0.38 –0.22
p
0.08 0.04
