1,25-Dihydroxyvitamin D to PTH(1–84) Ratios Strongly Predict Cardiovascular Death in Heart Failure

Objectives Vitamin D deficiency and hyperparathyroidism are common in patients with heart failure (HF). There is a growing body of evidence supporting the role of vitamin D and parathyroid hormone (PTH) in cardiac remodeling and worsening of HF. Lack of reliable automated testing of 1,25-dihydroxyvitamin D (1,25(OH)2D), the biologically active metabolite of vitamin D, has limited its contribution to the prognostic assessment of HF. Here, the association of 1,25(OH)2D and PTH(1–84) levels was evaluated for prediction of cardiovascular death in chronic HF patients. Methods We conducted a single center prospective cohort including 170 chronic HF patients (females n = 36; males n = 134; NYHA II-IV; mean age: 67 years; etiology: ischemic n = 119, dilated cardiomyopathy n = 51; mean LVEF: 23%). The primary outcome was cardiovascular death. Results Serum levels of 1,25(OH)2D decreased markedly with increased HF severity. Medians were 33.3 pg/mL for NYHA-II patients, 23.4 pg/mL for NYHA-III, and 14.0 pg/mL for NYHA-IV patients (p<0.001). Most patients had levels of 25(OH)D below 30ng/mL, and stratification by NYHA functional class did not show significant differences (p = 0.249). The 1,25(OH)2D to PTH(1–84) ratio and the (1,25(OH)2D)2 to PTH(1–84) ratio were found to be the most significantly related to HF severity. After a median follow-up of 4.1 years, 106 out of 170 patients reached the primary endpoint. Cox proportional hazard modeling revealed 1,25(OH)2D and the 1,25(OH)2D to PTH(1–84) ratios to be strongly predictive of outcomes. Conclusions 1,25(OH)2D and its ratios to PTH(1–84) strongly and independently predict cardiovascular mortality in chronic HF.

patients reached the primary endpoint. Cox proportional hazard modeling revealed 1,25 (OH) 2 D and the 1,25(OH) 2 D to PTH(1-84) ratios to be strongly predictive of outcomes.
Background Cardiovascular (CV) diseases remain a leading cause of death around the world [1]. Among CV diseases, heart failure (HF) represents a major health concern because of increasing prevalence worldwide with major human, societal and economic impacts [2][3][4][5][6][7]. The need for biomarkers for the prognosis of HF is well established, and different biomarkers from several pathophysiological pathways have been evaluated in this setting [8][9][10][11][12][13].
There is a growing body of evidence supporting the role of vitamin D and parathyroid hormone (PTH) in cardiac remodeling and worsening HF [14][15][16][17]. Furthermore, PTH together with aldosterone and fibroblast growth factor 23 (FGF-23), may also be part of a vicious and deleterious cycle which compromises CV function [18]. Markedly elevated levels of FGF-23 and PTH were observed in patients with CV disorders and HF, and have been related to adverse CV events [14;15;19-21].
Like PTH and FGF-23, 1,25-dihydroxyvitamin D (1,25(OH) 2 D, calcitriol) is an important regulator of calcium and phosphate homeostasis [21][22][23]. Recently, a novel fully-automated 1,25(OH) 2 D assay with improved analytical performance, sensitivity, and reliability has emerged [22;24]. The imprecision at low levels of existing 1,25(OH) 2 D measurement has precluded the ability to identify meaningful clinical correlates of HF progression so far. The aim of this study, therefore, was to assess the impact of sensitive, precise, accurate 1,25(OH) 2 D measurement and its ratios to PTH(1-84) on CV survival in HF patients.

Study population
We prospectively assessed CV death of 170 consecutive fully treated patients with chronic HF and reduced left ventricular ejection fraction (LVEF) followed at the Cliniques Universitaires Saint-Luc, an academic hospital of Brussels, Belgium, between March 30 th 2004 and June 16 th 2006. Patients with left ventricular systolic dysfunction and ejection fraction of 35% or less were eligible for the study. Ejection fraction was measured by radionuclide technique or contrast ventriculography, the latter being associated with coronary arteriography to confirm ischemic etiology. Exclusion criteria were age <18 years, LVEF higher than 35%, abnormal liver function test (AST/ALT 2 times the upper limit of the reference interval), anaemia or iron reserve deficiencies, genetic hypertrophic cardiopathy, severe pulmonary diseases (COPD gold 3-4), patients under dialysis and primary hyperparathyroidism. Survival status was obtained by phone contact with patients, their relatives, or their physicians.

Ethics statement
Competing Interests: Claudia Zierold, Frank Blocki, and Fabrizio Bonelli are employees of DiaSorin Inc., a provider of in vitro diagnostic products. For the current study, their contribution was to provide advice for the best use and monitoring of novel assays and scientific input for results interpretation as well as editing of manuscript. The authors are also working on a provisional patent. This does not alter the authors' adherence to all the PLOS ONE policies on sharing data and materials.
willingness to participate. The ethics committee of the Catholic University of Louvain approved this study as well as the consent procedure.

Clinical outcomes
Patient history and treatment was retrieved from medical files and review of hospital visitation records. Follow-up events including CV mortality and cardiac transplantation were 100% complete. Cardiac death was defined as death attributable to congestive HF, myocardial infarction, sudden death, or death occurring pursuant to revascularization procedures.

Laboratory measurements
Routine laboratory measurements and blood samples for biomarker analyses were obtained at hospital admission. Venous blood samples were obtained at enrollment, processed, and stored at -80°C until time of assay. Levels of 1,25(OH) 2 D were determined at baseline with a fully automated and sensitive immunoassay that uses a recombinant fusion construct of the vitamin D receptor ligand binding domain for specific capture of 1,25(OH) 2 D (DiaSorin, Saluggia, Italy). The limit of quantitation for this 1,25(OH) 2 D assay is 5 pg/mL and the reference interval determined in healthy volunteers ranged between 25.0 and 86.5 pg/mL with a median of 48.1 pg/mL. Plasma Ct-FGF23 concentrations were determined with a second-generation C-terminal human enzyme-linked immunosorbent assay (Immutopics, San Clemente, CA, USA). Levels of 25(OH)D, PTH(1-84) (both DiaSorin), B-type natriuretic peptide (BNP, Beckman Coulter, Fullerton, CA, USA; Alere reagents), N-terminal proBNP (NT-proBNP, Roche Diagnostics, Mannheim, Germany), Chromogranin A (CgA, Dako, Glostrup, Denmark), and Galectin-3 (Gal-3, BG Medicine, Waltham, MA, USA) were also determined. Glomerular filtration rate (eGFR) was estimated by the Modification of Diet in Renal Disease formula.

Statistical analysis
Independent determinants of baseline 1,25(OH) 2 D levels were assessed using multivariable linear regression methods with Log-transformed levels of 1,25(OH) 2 D as the dependent variable. The discriminatory power between biomarkers was assessed by fitting Cox proportional hazard (CPH) models using each of them, along with patient characteristics of age, sex, clinical variables, etiology of the disease, previous admission to hospital and treatment to predict outcomes. Biomarkers with highly skewed distributions -1,25(OH) 2 D, PTH(1-84), NT-proBNP, 25(OH)D, FGF-23, and creatinine were Log-transformed for use in the CPH models. The strength of association of each of these predictors can be summarized by the chi-squared statistic of the fitted CPH. Along with the individual markers and patient characteristics, 1,25(OH) 2 D to PTH(1-84) ratios were also determined.
Several of these predictors have large chi-squared values, indicating strong associations with survival. For these biomarkers, receiver operating curve (ROC) analyses were performed to explore the trade-off between sensitivity and specificity as assessed at the seven-year mark, and to locate a cut point suitable for use in dichotomizing the patients into high-and low-risk groups. This cut point was defined as that value maximizing the difference between sensitivity and false positive rates. The Kaplan-Meier survival curve of patients split by this dichotomy was found and tested using the Log-Rank test. P-values < 0.05 were considered significant. Statistical analysis was performed using R version 3.1.0 (The R Foundation for Statistical Computing) and JMP software version 11.00 (SAS institute, NC, USA). It may be mentioned that the CPH and the ROC analyses involve somewhat different perspectives. The ROC calculations use a snapshot of the patients at the end of the monitoring period and do not distinguish between earlier and later deaths. The CPH and Kaplan-Meier analyses, on the other hand, use the entire survival function and thereby do distinguish earlier from later deaths. Table 1 shows the baseline characteristics and laboratory values of the study population, according to 25(OH)D and 1,25(OH) 2 D levels. 170 chronic HF patients were included (mean age 67±14 years; females n = 36; males n = 134; New York Heart Association (NYHA) II-IV; etiology: ischemic n = 119, dilated cardiomyopathy n = 51; mean LVEF 23±7%).

Baseline characteristics
Most patients had levels of 25(OH)D below 30 ng/mL (58% of HF patients were below 15 ng/mL, 21% between 15 and 20 ng/mL, 15% between 20 and 30 ng/mL, and only 6% were higher than 30 ng/mL, and stratification by NYHA functional class did not disclose significant differences (p = 0.249). In contrast, median serum levels of 1,25(OH) 2 (23), we examined the ratio of these two hormones and found them to be significantly related to HF severity: NYHA class II ratio = 1.14, NYHA class III ratio = 0.47, and NYHA class IV ratio = 0.38 (p<0.001). The median ratio for all patients was 0.62 (range: 0.02-9.3). The 1,25(OH) 2 D to PTH(1-84) ratios were higher in patients with 1,25(OH) 2 D levels above the median, while PTH was not different between the two groups.

1,25(OH) 2 D level and 1,25(OH) 2 D to PTH(1-84) ratio outcomes in chronic HF
Over a median follow-up time of 4.1 years (occurrence of a first endpoint and maximum follow-up times were 7 days and 7.4 years, respectively), 106 HF patients met the endpoint of which 94 died, and 12 underwent heart transplant. Biomarker concentrations as well as clinical variables between patients that developed the outcome and those who remained stable are presented in Table 2. CPH analysis predicting survival from the Logs of 1,25(OH) 2 D and PTH  show that each of these biomarkers is highly significant even when the other is used (their p-values in the CPH using both being 1.2x10 -7 and 1.5x10 -3 ); in other words, both contribute separate, highly significant, predictive power. The coefficient of Log 1,25(OH) 2 D is roughly -2 times that of Log PTH(1-84), suggesting that the two might be summarized by a score function -2 Log 1,25 (OH) 2 D + Log PTH(1-84), which is the negative Log of the ratio of the square of the 1,25 (OH) 2 D assay to the PTH. This composite score marker is assessed in Table 3 as "Square ratio", and indeed it has the second-highest chi-squared value of all the biomarkers, being outperformed only (and very modestly) by BNP. A simpler approach using the 1,25(OH) 2 D to PTH(1-84) ratio, without the doubling on the 1,25(OH) 2 D, is assessed under the label "Ratio". As Table 3 might lead one to expect, this simpler ratio does not perform quite as well, but  Youden's index and the associated criterion based on the ROC curve analysis were used for the stratification of the Kaplan-Meier curves (35.4 pg/mL for 1,25(OH) 2 D, 1.06 for the ratio, 32.0 pg/mL for the Square ratio). Kaplan Meier survival curves for patients stratified by 1,25 (OH) 2 D levels diverged significantly (Log-rank test: p <0.001; Fig 2A). Survival curves for patients stratified by the 1,25(OH) 2 D to PTH(1-84) ratios also diverged significantly (Logrank test: p <0.001; Fig 2B and 2C). Table 4 shows the coefficients and the statistical significance of each term in a multivariate CPH model including all the variables listed in the table. In this table, the contribution of each predictor is adjusted for the common predictive information of all other predictors. The Square ratio has a highly significant p-value (0.011) showing that this predictor is significant even after all other predictors are taken into account; its information therefore is largely unique and additive to that of the other predictors.

Discussion
The main objective of our study was to investigate the prognostic value of 1,25(OH) 2 D, measured with the new generation assay, and its ratios to PTH(1-84) for long-term CV death in patients with chronic HF. These results clearly demonstrate the relationship between decreased levels of 1,25(OH) 2 D and long-term CV death in HF, and most importantly, the novel use of the 1,25(OH) 2 D to PTH(1-84) ratios as superior biomarkers for the prognosis of HF patients. Vitamin D deficiency was previously shown to be associated with cardiovascular diseases [25;26], however the biologically active metabolite 1,25(OH) 2 D was not routinely assessed due in part to the lack of assays that are accurate and precise at low 1,25(OH) 2 D concentrations. Decreased 1,25(OH) 2 D levels in HF patients were previously reported [25;27]; once again, the concentrations were obtained with less sensitive and reliable assays. The development of a novel automated, extraction-free 1,25(OH) 2 D immunoassay, with precision and sensitivity superior to LC-MS/MS, has allowed the exploration of new biological relationships that until now were beyond the reach of current methodologies [22]. The 1,25(OH) 2 D test used here, in contrast to the tests for existing biomarkers, offers several advantages: automation, standardization, small sample size and short turn-around-time. In addition, the pre-analytical extraction step, on samples routinely in excess of 1 mL that is normally required for other 1,25(OH) 2 D assays LC-MS/MS inclusive, is not required. This greatly diminishes the imprecision of the assay, especially at low concentrations [22]. The effect on imprecision that results from combining values from two assays into a single ratio was also examined by running two different lots each of 1,25(OH) 2 D and PTH(1-84) kits, and determining the ratio for the four resulting kit combinations. The coefficient of variation of the ratio was on average 2.8% for ratios >0.5 with a maximum CV of 6.3% (data not shown).
These data clearly demonstrate the relationship between decreased levels of 1,25(OH) 2 D and long-term CV death in HF. Most importantly these data indicate that the relationship between the 1,25(OH) 2 D and PTH(1-84) hormones afford potent biomarkers for the prognosis of HF patients, as depicted by the Kaplan Meier curves, which disclose markedly early changes in survival according to the 1,25(OH) 2 D to PTH(1-84) ratios. The association of 1,25 (OH) 2 D levels and the risk of adverse outcomes in CV diseases was previously observed [25;27], however the 1,25(OH) 2 D to PTH(1-84) ratios are novel biomarkers not previously investigated. The predictive value of 1,25(OH) 2 D to PTH(1-84) ratios were equivalent to established biomarkers of HF severity such as BNP, NT-proBNP and FGF-23 as evidenced by the AUC. These are all the more remarkable given the modest sample size. In contrast, 1,25 (OH) 2 D to PTH(1-84) ratios were superior to Gal-3, PTH(1-84) and 25(OH)D. In comparison to the other biomarkers, 25(OH)D levels were not predictive of the outcome. Several studies have related 25(OH)D levels to survival in HF patients [15;16;28]. The differences might be related to a lower median level of 25(OH)D with a high number of HF patients below 15ng/mL but also to length of follow-up which was longer in our study in comparison to the other reports. However, our results are in agreement with some other reports showing that 25(OH)D deficiency was not related to HF in contrast to PTH or FGF-23 [29;30].
In addition, that the 1,25(OH) 2 D to PTH(1-84) ratios were shown to be statistically superior to 1,25(OH) 2 D and PTH(1-84) alone allowed for the integration of interrelated confounders. This now presents the possibility for treatment based on two modulable factors. Though the square ratio appears to be slightly better than the simple ratio, future studies with larger cohorts are needed to determine which of the two ratios will provide the stronger biomarker. The use of squared ratios, like that employed in the calculation of the widely accepted Body Mass Index, is not without precedent [31]. The 1,25(OH) 2 D to PTH(1-84) ratios are potent biomarkers that combine the contribution of the active form of vitamin D, 1,25(OH) 2 D, with circulating PTH (1-84), which was previously shown to be associated with mortality [16;32]. In addition, studies have shown that PTH contributes to the pathophysiology and worsening of HF [14;33]. Secondary hyperparathyroidism was previously observed in patients with untreated and treated HF with reduced left ventricular ejection fraction with second and third generation immunoassays [33][34][35]. In addition, PTH was shown to have several negative direct and indirect effects on the heart and cardiac cells [18;36]. Furthermore, increased circulating concentrations of PTH might stimulate adrenal aldosterone synthesis, initiating a vicious cycle between hyperparathyroidism and hyperaldosteronism leading to more pro-inflammatory, pro-oxidant and pro-fibrotic actions [37][38][39]. Like 1,25(OH) 2 D and PTH(1-84), FGF-23 is a key regulator of mineral and phosphorus homeostasis [40;41]. Interestingly, a significant relationship between FGF-23 and PTH was previously documented in chronic kidney disease and HF patients [42][43][44][45]. Previous studies have found significantly higher mortality in HF patients with FGF-23 levels >172 RU/mL [45]. FGF-23 and FGF receptors are both expressed in the myocardium, and it was hypothesized that FGF-23 may have a direct effect on the heart and participate in the physiopathology of CV diseases and HF [46;47]. As 1,25(OH) 2 D participates in the regulation of bone and mineral metabolism with PTH and FGF-23, the potential for significant diagnostic interrelations between 1,25(OH) 2 D, PTH and FGF-23 in the physiopathology of the cardio-renal syndrome related to HF is increasing. Our data demonstrate that decreased 1,25(OH) 2 D levels are significantly related to HF severity and to the rise of serum PTH and FGF-23 levels. The decrease of the biologically active 1,25(OH) 2 D concentrations according to NYHA classes might be related to the worsening HF and associated related worsening renal failure which can decrease the 1α hydroxylase activity and therefore leading to a reduction of the 1,25(OH) 2 D levels. The fall of 1,25(OH) 2 D might trigger the rise of PTH and FGF-23 contributing to a vicious cycle for cardiovascular function. On the other hand, worsening of HF and renal failure in HF patients leads to secondary hyperparathyroidism and increases of circulating FGF-23, both of which impact the synthesis of 1,25(OH) 2 D. Measurement of the ratios appears therefore as an efficient tool to assess these interrelated players independently of the kidney function.
The therapeutic rational for testing 1,25(OH) 2 D and its ratios to PTH in HF patients is that, not only could it enable more reliable risk stratification, it also could serve to guide treatment selection and monitor the efficiency of other medical devices [51]. Previous studies showed that 1,25(OH) 2 D supplementation has protective effects on myocardial fibrosis of diabetic rats [52], that it is effective in preserving endothelial function in hypertension [53], and it improved cardiac function in patients on hemodialysis that had controllable hyperparathyroidism [54]. Indeed, more tailored treatment selection might be guided by aiming for higher 1,25(OH) 2 D/ PTH(1-84) ratios. Lower levels of 1,25(OH) 2 D in HF patients could be increased with calcitriol supplementation (or other active metabolite). On the other hand, PTH could be decreased with other pharmacological treatments such as aldosterone antagonists, known to prevent hyperparathyroidism and its consequences [55;56].
In conclusion, based on the present data, 1,25(OH) 2 D to PTH(1-84) ratios are novel biomarkers that strongly and independently predict CV mortality in chronic HF, stronger than 1,25(OH) 2 D alone. In addition, the 1,25(OH) 2 D to PTH(1-84) ratios are comparable or better than currently used biomarkers such as BNP, NT-proBNP, Galectin-3, CgA and FGF-23 for prognosis, but with the potential additional advantage of providing determinants for therapy guidance and patient monitoring.
Supporting Information S1 Dataset. data set of the studied HF patients. (XLSX)