Determinants of circulating calcitriol in cardiovascular disease

Circulating calcitriol may contribute to the risk of cardiovascular disease (CVD), but its regulation in patients with CVD is poorly characterized. We therefore aimed to assess determinants of circulating calcitriol in these patients. We analyzed 2183 independent samples from a large cohort of patients scheduled for coronary angiography and 1727 independent samples from different other cohorts from patients with a wide range of CVDs, including heart transplant candidates, to quantify the association of different parameters with circulating calcitriol. We performed univariable and multivariable linear regression analyses using the mathematical function that fitted best with circulating calcitriol. In the multivariable analysis of the large single cohort, nine parameters remained significant, explaining 30.0 % (32.4 % after exclusion of 22 potential outliers) of the variation in circulating calcitriol (r = 0.548). Log-transformed 25-hydroxyvitamin D [25(OH)D] and log-transformed glomerular filtration rate were the strongest predictors, explaining 17.6 % and 6.6 %, respectively, of the variation in calcitriol. In the analysis of the combined other cohorts, including heart transplant candidates, the multivariable model explained a total of 42.6 % (46.1 % after exclusion of 21 potential outliers) of the variation in calcitriol (r = 0.653) with log-transformed fibroblast growth factor-23 and log-transformed 25(OH)D explaining 29.0 % and 6.2 %, respectively. Circulating 25(OH)D was positively and FGF-23 inversely associated with circulating calcitriol. Although significant, PTH was only a weak predictor of calcitriol in both analyses ( < 2.5 %). In patients with CVD, FGF-23 and 25(OH)D are important independent determinants of circulating calcitriol. The relative importance of these two parameters may vary according to CVD severity. Future studies should focus on the clinical importance of regulating circulating calcitriol by different parameters.


Introduction
Numerous in-vitro and experimental animal studies have provided evidence for beneficial effects of the vitamin D hormone 1,25-dihydroxyvitamin D 3 (calcitriol) on the cardiovascular system [1].Briefly, calcitriol can decrease thrombogenicity, oxidative stress, atherogenesis, and foam cell formation, and improves endothelial repair, vascular relaxation, and dilatation.Deletion of the calcitriol receptor, i.e. the vitamin D receptor (VDR), results in elevated synthesis of renin and angiotensin II, leading to hypertension and cardiac hypertrophy.Experimental diets low in vitamin D content as well as deletion of the VDR stimulate osteoblast-like cell formation in vascular smooth muscle cells (VSMCs), and thus vascular calcification [1].In addition to these experimental data, huge data sets in humans have reported nonlinear inverse associations of circulating 25-hydroxyvitamin D (25(OH)D), the generally accepted indicator of vitamin D status, with cardiovascular morbidity and mortality [2,3], indicating that adequate vitamin D status might play a substantial role in the prevention of cardiovascular events.However, large randomized controlled trials provided no beneficial results regarding vitamin D administration and the prevention of nonfatal and fatal cardiovascular events [4][5][6].Even in strata with deficient 25 (OH)D concentrations, i.e. values < 12 ng/mL, nonlinear Mendelian Randomization analyses reported null effects of genetically predicted 25 (OH)D on coronary heart disease, stroke, and cardiovascular mortality [3].
It is however noteworthy that calcitriol and not 25(OH)D is the most active hormonal form of vitamin D.Besides 25(OH)D, circulating calcitriol has also been reported to be independently and inversely associated with mortality in cardiovascular disease [7,8].Likewise, in two human populations at high and moderate risk for ischemic heart disease, serum levels of calcitriol were inversely correlated with the extent of vascular calcification [9].Nevertheless, the effect of calcitriol on the cardiovascular system seems to be a double-edged sword, as elevated calcitriol concentrations can promote vascular calcification either by direct hypercalcemic (plasma calcium >2.6 mol/L) and hyperphosphatemic (plasma phosphate >1.61 mmol/L) effects or by trans-differentiation of VSMCs into osteoblast-like cells [10,11].In addition, an increase in 1α-hydroxylase expression in VSMCs [12], and loss-of-function mutation of fibroblast growth factor (FGF)-23 [13] can induce ectopic calcification.
Parathyroid hormone (PTH) and FGF-23 are the principal regulators of circulating calcitriol to maintain plasma calcium and phosphate concentrations within a narrow range [14].Synthesis of calcitriol is activated by PTH and suppressed by the phosphaturic hormone FGF-23 [14].Circulating calcitriol is considered to be tightly regulated [15], but a wide range of calcitriol concentrations has been reported in health and disease [16], probably as a result of the body's demand of calcium and phosphate metabolism and the impact of various diseases and conditions on mineral metabolism [14,16].Although there are numerous studies published on the homeostasis of mineral metabolism and the effects and interactions of its regulators, evidence is limited on the extent to which different factors determine circulating calcitriol concentrations in cardiovascular disease (CVD) patients.
The present study, therefore, aimed at quantifying the association of different parameters with circulating calcitriol in patients with CVD.

Patients and data collection
For this investigation, we included 4450 samples from patients with CVD of Caucasian ethnicity.We used data from five different studies, the major results of which have already been published elsewhere: The LUdwigshafen RIsk and Cardiovascular Health study (designated LURIC) provided 3316 samples [7,17,18]; 305 samples were made available from a clinical study on Mineral Metabolism and Clinical Outcome (designated MMCO) [19]; 69 samples were provided by a clinical study in Ventricular Assist Device recipients (designated VAD) [20]; the Effect of VItamin D on MorTAality in Heart Failure Trial (designated EVITA) delivered 165 samples [21,22], and 521 samples were retrieved from the Styrian Vitamin D Hypertension Trial (designated SVDH) [23].LURIC was performed at the Herzzentrum Ludwigshafen (Cardiac Center Ludwigshafen) in southwest Germany (geographic latitude: 49.5 • N) and included patients scheduled for coronary angiography.EVITA, MMCO and VAD were performed at the Heart Failure unit and VAD unit, respectively, of the Clinic for Thoracic and Cardiovascular Surgery, Heart-and Diabetes Center NRW, in East Westphalia, Bad Oeynhausen (geographic latitude: 52.2 • N), Germany.These three latter studies exclusively included patients with end-stage heart failure considered for heart transplantation.SVDH included hypertensive patients and was performed at the Department of Internal Medicine, Medical University of Graz, (geographic latitude: 47.4 • N), Austria.Written informed consent for use of biological samples for scientific research was received from all patients.Moreover, study procedures were approved by the respective ethics committees, as stated in earlier publications [7,[19][20][21]23].

Parameters
In the five studies, we collected baseline characteristics such as anthropometric data, concomitant diagnoses, kidney function, medication use, vitamin D metabolites, and hormones, which might be related to circulating calcitriol such as c-terminal FGF-23 [c-FGF-23] and intact PTH.In addition, parameters of liver function and plasma mineral concentrations were available in some studies.

C-terminal fibroblast growth factor-23
We measured c-FGF-23 using an enzyme-linked immuno assay (ELISA) test kit provided by Immutopics (San Clemente, CA) in four studies.In one study (SVDH), c-FGF-23 was analyzed using a test kit provided by BIOMEDICA (Medizinprodukte GmbH & CO KG, Vienna, Austria).For these latter values, a correction factor of 15.4 was used to convert pmol/l to RU/mL [24].

25-hydroxyvitamin D and 24,25-dihydroxyvitamin D
In a subset of LURIC samples (n=2477), 25(OH)D was (re-)analyzed with liquid chromatography tandem mass spectrometry (LC-MS/MS), as previously described [25].In these samples, 24,25-dihydroxyvitamin D (24,25[OH] 2 D) was also measured with LC-MS/MS [25].In addition, the vitamin D metabolite ratio (VMR), i.e. 24,25(OH) 2 D divided by 25 (OH)D, was determined.A low vitamin D metabolite profile was assumed when 24,25(OH) 2 D was below 1.2 ng/mL and the VMR was below 4 % [26].Individuals that satisfied only 1 of the 2 functional criteria were classified as having a suboptimal vitamin D metabolite profile.Circulating 25(OH)D was also measured by the aforementioned LC-MS/MS method in SVDH samples.In the remaining LURIC and the MMCO samples, serum levels of 25(OH)D were measured by radioimmunoassay or the DiaSorin autoanalyzer (VAD, EVITA) (all from Dia-Sorin, Stillwater, MN, USA).According to the Institute of Medicine [27], 25OHD levels < 12 ng/mL (multiply by 2.496 to convert into nmol/l) are classified as deficient, and values between 12 and 20 ng/mL as inadequate.

Calcitriol
Levels of calcitriol were measured by radioimmunoassay from Nichols Institute Diagnostika GmbH (LURIC) (Bad Nauheim, Germany), ELISA test kit (MMCO) from Immundiagnostik (Bensheim, Germany), chemiluminescence assay (VAD, EVITA) with an autoanalyzer from DiaSorin (Stillwater, MN, USA), and chemiluminescence immunoassays (SVDH) from Immunodiagnostic Systems Ltd.(Boldon, UK).For samples that tested lower than the limit of detection, the value used for analysis was the lowest standard divided by two [28].

Glomerular filtration rate
Glomerular filtration rate (GFR) was calculated according to the Chronic Kidney Disease Epidemiology Collaboration (CKD-EPI) formula [29].

Reference ranges and quality control
Table 1 presents proposed reference ranges by the manufacturers of the assays used for the determination of calciotropic and phosphaturic hormones and, in case of 24,25(OH) 2 D, by those who analyzed these samples.Likewise, intra-and inter-assay variation coefficients, provided by the manufacturers or, in case of LC-MS/MS measurements, the researchers, are presented.All labs contributing data to this investigation participated successfully in legally required external quality controls regarding biochemical analytes.The Medical University of Graz, which quantified 25(OH)D and 24,25(OH) 2 D using the LC-MS/MS method, has satisfactorily participated in the Vitamin D External Quality Assessment Scheme (DEQAS), with target values assigned by the Centers of Disease Control and Prevention (CDC) reference measurement procedure using certified reference material from the National Institute of Standards and Technology.In addition, a highly significant correlation was noted between 25(OH)D levels obtained by radioimmunoassay and by LC-MS/MS in the LURIC samples [7].The Heart and Diabetes Center NRW has successfully participated in the external quality control of the Reference Institute for Bioanalysis of the German Accreditation Office for both 25(OH)D and calcitriol.With regard to calcitriol, no external quality control was available for the LURIC and SVDH samples.

Statistics
To prevent potential bias by using different assays for the measurement of each biochemical parameter, we first restricted our analysis to the LURIC samples.To further strengthen the reliability of the results, we restricted this analysis to the subgroup of LURIC samples with available LC-MS/MS measurements of 25(OH)D (n=2470), since this method is considered to be the preferential method of analyzing 25(OH) D.
We assessed the relationships between circulating calcitriol and potential predictors using univariable and multivariable linear regression analysis.The following parameters were selected as independent variables: age, sex, body weight, body height, BMI, eGFR, 25(OH)D, eGFR, albumin-corrected calcium (calcium corr ), sodium, phosphate, PTH, c-FGF-23, CRP, bilirubin, diabetes mellitus, ß-Blockers, diuretics, and ACE Inhibitors/ARBs.They were considered because they (i) have been shown to be associated with vitamin D status, i.e. circulating 25(OH)D (age, sex, BMI, body weight and height, diabetes mellitus), (ii) are associated with calcitriol synthesis (age, eGFR, 25(OH)D, phosphate, PTH, c-FGF-23, CRP), (iii) may theoretically influence synthesis of vitamin D metabolites (bilirubin, albumin-corrected calcium (calciumcorr )), or (iv) are associated with plasma volume and mineral metabolism (sodium, diuretics, ß-blockers, and ACE Inhibitors/ARBs) and may therefore also influence concentrations of vitamin D metabolites.
Each variable was first tested in a univariable analysis using different mathematical models to determine whether a linear, logarithmic, quadratic, cubic, inverse, exponential or power function provided the best fit.Variables that showed a significant relationship (P values < 0.05) with circulating calcitriol were then tested in a multivariable regression model using the mathematical function that fitted best with circulating calcitriol.In the multivariable analysis, variables were removed stepwise if no significant influence could be detected.Only the variables that were significantly associated with circulating calcitriol remained in the multivariable model.To determine the reliability of our model, we performed several analyses: we tested linearity between variables, checked for potential outliers, looked for auto-correlation and multicollinearity, inspected normality of residuals, and explored homogeneity of variance.To this end, we used scatterplots, histograms, P-P-plots, the Durbin-Watson statistic, and Pearson's correlation coefficient (threshold for multicollinearity: 0.7) where appropriate.Cases with studentized residuals below or above 3 were considered as potential outliers.We also performed bootstrapping by resampling the prediction population 1000 times with replacement to allow for a more robust estimation of the parameters that might otherwise be biased by a lack of homogeneity of variance.
In the LURIC samples, we also present data of independent calcitriol determinants below and above the suggested lower limit of the reference range by calcitriol category (< 18 pg/mL and ≥ 18 pg/mL; see Table 1).For assessment of group differences, the Mann-Whitney U-test was used.The P-values < 0.05 were considered statistically significant.Of the available 2477 samples, 294 had to be excluded because of missing values, leaving 2183 for statistical analysis.
In a second step, we analyzed the relationships between circulating calcitriol and potential predictors in the four other cohorts and the LURIC samples without LC-MS/MS measurements.This resulted in a total number of 1920 samples.For this analysis, the following independent variables were available from all patients and were therefore included in the multivariable analysis as independent variables: age, sex, body height, body weight, BMI, diabetes mellitus, ß-blockers, diuretics, ACE inhibitors/ARBs, eGFR, 25(OH)D, PTH, c-FGF-23.To reduce the risk of potential bias from using different assays for the measurement of each biochemical parameter in different cohorts, we also included a dummy variable for study cohort.Of the available 1920 samples, had to be excluded because of missing values, leaving 1727 for statistical analysis.
We applied the statistical software package SPSS, version 27 (IBM Corp, Armonk, NY, USA) to perform the analyses.

Results
Characteristics of the five studies are presented in Table 2. Briefly, the majority of patients were male and were aged between 50 and years.A substantial percentage of patients had diabetes mellitus and hypertension, whereas less than 10 % of the total number of patients had eGFR values below 30 mL/min/1.73m 2 .Left ventricular ejection fraction and medication use of the study cohorts can be regarded as reflecting the severity of CVD.The concentrations of calciotropic and phosphaturic hormones differed substantially between included studies, and so did the percentage of calciotropic and phosphaturic hormones   Watson statistic is 2.029.Twenty-two values were potential outliers (all studentized residuals > 3) (Supplemental Figure 3).We ran the analysis also without these potential outliers, resulting in a multiple r of 0.569, which explains 32.4 % of the variation in circulating calcitriol.Homogeneity of variance is depicted in Supplemental Figure 3. Bootstrapping did not change results of the multivariable analysis substantially (Supplemental Table 1).
Table 4 illustrates the concentrations of independent biochemical predictors of circulating calcitriol in the LURIC samples below and greater than/equal to the suggested lower reference range of calcitriol (18 pg/mL).All parameters differed significantly between the two categories, with plasma phosphate and FGF-23 values being significantly higher and the other parameters being lower in the category with low calcitriol concentrations than in the category with calcitriol ≥ 18 pg/ mL.Median values of 25(OH)D were in the deficiency range, whereas median c-FGF-23 values were close to the upper limit of its reference range in the category with low circulating calcitriol.Although significantly different between categories, median PTH values were well within the reference range.In the category with calcitriol concentrations < 18 pg/mL and ≥ 18 pg/mL, the vitamin D metabolite profile was low/ suboptimal in 57.0 % and in 28.6 %, respectively (P<0.001).
In step two, we assessed independent predictors of circulating calcitriol in all other samples.Since there was substantial collinearity between body weight and BMI, as indicated by Pearson's correlation coefficient (r=0.825),body weight was not included in the multivariable analysis.All other correlation coefficients between independent variables were < 0.7, indicating no substantial collinearity.All included parameters were significantly associated with circulating calcitriol in the univariate analysis, with the exception of sex, body height, and ßblockers (Table 3).For most continuous parameters, a logtransformation fitted best with circulating calcitriol, whereby negative exponents indicate an inverse association of the parameter with circulating calcitriol.In the multivariable analysis, seven parameters remained significant, among them log-transformed c-FGF-23, logtransformed 25(OH)D, log-transformed eGFR, log-transformed PTH, study cohort, log-transformed BMI, and diabetes mellitus.A total of 42.6 % of the variation in circulating calcitriol was explained by this multivariable model (r=0.653), with log-transformed c-FGF-23 being the strongest predictor (29.0 %).Log-transformed 25(OH)D, logtransformed eGFR, log-transformed PTH, study group, log-transformed BMI, and diabetes mellitus explained 6.2, 4.5, 2.2, 1.8, 0.6, and 0.3 %, respectively, of the variation in calcitriol.The model has no autocorrelation as the value of the Durbin-Watson statistic is 1.940.Normality of residuals is depicted by histogram and P-P-plot (Supplemental Figures 4 and 5).Twenty-one values were potential outliers (all studentized residuals > 3) (Supplemental Figure 6), and we ran the analysis also without these potential outliers.Deletion of the outliers resulted in a multiple r of 0.679, which explains 46.1 % of the variation in circulating calcitriol.Log-transformed c-FGF-23, log-transformed 25 (OHD), log-transformed PTH, study cohort, log-transformed BMI, logtransformed BMI, and diabetes mellitus explained 31.9, 6.3, 2.4, 2.4, 1.8,1.7,and 0.7 %, respectively, of the variation in calcitriol.Bootstrapping did not change results of the multivariable analysis substantially (Supplemental Table 2).Fig. 2 illustrated in a scatterplot the association between measured calcitriol concentrations and calcitriol concentrations predicted by the linear regression analyses.The regression lines were almost identical for both the LURIC subset and the other cohorts, but the LURIC subset covered a smaller range of calcitriol concentrations than the other cohorts.

Discussion
The present investigation in patients with CVD has several major findings.First, c-FGF-23 and 25(OH)D were important determinants of circulating calcitriol.Second, the predictive power of c-FGF-23 was more pronounced in the overall analysis than in the LURIC subset.Third, in the LURIC samples median 25(OH)D concentrations were in the deficiency range and c-FGF-23 concentrations were at the upper end of its reference range in the calcitriol category with concentrations below 18 pg/mL, indicating that these calcitriol concentrations may not reflect 'normal' values.Fourth, the predictive power of PTH for circulating calcitriol was low and PTH showed only a modest inverse association with calcitriol.In samples with calcitriol concentrations below 18 pg/ mL, median PTH levels were well within its reference range.
Overall, predictive power of the multivariable regression analysis after exclusion of potential outliers was r=0.569 in the LURIC subset and 0.679 in other samples.In the field of scientific research even an effect size of r=0.5 has been considered large [30].It is also noteworthy that calcitriol is a difficult analyte.Low concordance and substantial bias between methods has been demonstrated [31].We used four different assays for measuring calcitriol in the overall analysis and also different assays to measure 25(OH)D, PTH, and c-FGF-23.In addition, GFR has only been estimated by established calculations [29].All these uncertainties may have influenced the predictive power.However, in the overall analysis multiple r was still better than in the LURIC subset, despite the restriction to one method per parameter and the inclusion of more potentially predictive parameter in the analysis of the LURIC subset.Study cohort contributed only 0.5 % to the predictive power of estimating circulating calcitriol in the analysis including different cohorts.Altogether, results indicate that the pathophysiology of CVD was more important than the potential bias by data imprecision due to the use of different tests.
The broad range of CVD severity in the analysis of different cohorts, including samples from patients with end-stage heart failure, may not only explain the better predictive power of the overall samples compared to the LURIC subset, but also the fact that log-transformed c-FGF-23 was more predictive in the overall analysis than in the LURIC subset.c-FGF-23 is expressed by cardiomyocytes under conditions of stress [32].The high percentage of low calcitriol concentrations in heart transplant candidates and VAD patients (Table 2) can reliably be explained by their higher c-FGF-23 values, as FGF-23 inhibits 1α-hydroxylase [33], the enzyme that catalyzes the reaction from 25(OH)D to calcitriol.
The failing heart is an engine out of fuel with a reduced cardiac phosphate utilization [34].The inhibition of calcitriol synthesis leads to reduced intestinal phosphate absorption [14] and FGF-23 increases renal phosphate excretion, two mechanisms that can contribute to a maintained phosphate homeostasis in case of reduced phosphate utilization.
Log-transformed 25(OH)D was a strong positive predictor of circulating calcitriol, at least in the subset of LURIC samples.Circulating (OH)D reflects the bodies' vitamin D reservoir.Vitamin D supplementation, and thus an increase in circulating 25(OH)D, can significantly increase circulating calcitriol [35].This has also been shown in the two included vitamin D supplementation studies, where daily vitamin D supplements of 2800 and 4000 IU resulted in a median increase in calcitriol of 5 pg/mL and 7.5 pg/mL, respectively [22,36].The significantly higher percentage of a low/suboptimal vitamin D metabolite profile in the samples with calcitriol concentrations < 18 pg/mL than in samples with and calcitriol concentration ≥ 18 pg/mL further supports the assumption that low vitamin D status contributes to low calcitriol concentrations.
Regarding kidney function, there is an early and progressive disturbance in mineral metabolism and elevation in FGF-23 together with a decline in circulating calcitriol with decreasing eGFR values [37,38].Although eGFR was an independent predictor of circulating calcitriol in our study, the association between eGFR and circulating calcitriol was only modest.Results are in line with the relatively low prevalence of patients with chronic kidney disease stage IV or higher in our samples.
Surprisingly, PTH was only a weak predictor of circulating calcitriol.In addition, PTH was at the lower end of its reference range in those LURIC samples with calcitriol concentrations below 18 pg/mL.The effect of PTH on the cardiovascular system is complex: elevated as well as low PTH levels have been associated with an increased risk of mortality and cardiovascular events in patients with CVD [17][18][19].Similar to FGF-23, PTH has phosphaturic properties [14].The low PTH levels in those LURIC samples with calcitriol concentrations < 18 pg/mL is, thus, in line with the elevated plasma phosphate levels in this group.Since the suppressive effect of vitamin D on PTH is well-known [39], caution is necessary regarding vitamin D supplementation in patients with CVD.The suppressive effect of vitamin D on PTH might, at least in part, contribute to elevated phosphate levels and its adverse effects on vascular calcification [32].
The reference range for calcitriol is not well established.The range has often been based on the statistical approach of a Gaussian distribution of subjects [40], and each assay manufacturer provides its own reference range (Table 2).According to our results of deficient circulating 25(OH)D and elevated c-FGF-23 in the LURIC samples below the proposed reference range of calcitriol, it would be more appropriate to use a functional approach for defining adequate circulating calcitriol concentrations in the future.
Our study has both strengths and limitations.Strengths are the large number of samples, the broad range of CVD severity of included patients, and the insertion of various parameters known to be associated with circulating calcitriol concentrations.One limitation is the use of different methods to assess calcitriol, 25(OH)D, PTH, and c-FGF-23.Additional limitations are that the patients were on multiple therapeutic drugs that may interfere with the physiological regulation of calcium/ phosphate metabolism, although some of these medications were considered in our analysis.The present data from hospital samples with CVD may not be representative for samples selected from the community.Females are underrepresented in this study and results are restricted to the European population of Caucasian ethnicity.
In summary, c-FGF-23 and 25(OH)D are important determinants of circulating calcitriol in CVD patients.The relative importance of these two parameters may vary according the CVD severity.Surprisingly, PTH was only a weak predictor of circulating calcitriol.Altogether, several questions regarding calcitriol regulation still remain.Therefore, future studies should focus on the importance of different and additional determinants of regulating circulating calcitriol in different types and severities of CVD.In addition, the predictive power of circulating calcitriol for clinical outcome in CVD and its pathophysiological basis should be investigated in more detail.

Declaration of Competing Interest
None.

Fig. 1 .
Fig. 1.Circulating calcitriol distribution across different calcitriol categories in the Ludwigshafen Risk and Cardiovascular Health study subset and different other cohorts.The Ludwigshafen Risk and Cardiovascular Health study subset (white bars) comprises of 2183 patients and the other cohorts (black bars) comprise of 1727 patients.

Fig. 2 .
Fig. 2. Predicted compared to measured increase in the Ludwigshafen Risk and Cardiovascular Health study and other cohorts.The Ludwigshafen Risk and Cardiovascular Health study subset (black circles; regression equation: y=0.69+0.97*x,R 2 = 0.324) comprises of 2161 patients and the other cohorts (white circles; regression equation: y=0.34+0.97*x,R 2 = 0.461) comprise of 1706 patients after exclusion of potential outliers.

Table 1
Analyses of biochemical parameters.
A. Zittermann et al. with low values (25[OH]D, calcitriol) or high values (PTH, FGF-23).The distribution of calcitriol concentrations is depicted in Fig. 1, separated by samples of the LURIC subset 1 with LC-MS/MS measurements and all other samples.In a first step, we assessed independent predictors of circulating calcitriol in the LURIC subset 1 with LC-MS/MS measurements.BMI, use of ß-blockers, and calcium corr were not significantly associated with circulating calcitriol in the univariable analysis (Table3).Log-transformed eGFR, log-transformed bilirubin, log-transformed c-FGF-23, log-transformed plasma phosphate, linearly-transformed PTH, log-transformed body height, quadratically-transformed CRP, and quadratically-transformed age explained 6.6, 1.8, 1.5, 1.9, 0.9, 0.6, 0.3, and 0.2 %, respectively, of the variation in calcitriol.Normality of residuals was proven by distribution and P-P-plots (Supplemental Figures.1 and 2).The model has no auto-correlation as the value of the Durbin-

Table 2
Characteristics of included studies.
samples; c the assay specific reference range is used, as presented in Table1

Table 3
Evolution of determinants of circulating calcitriol in patients with CVD.