The impact of the assay for measuring albumin on corrected (‘adjusted’) calcium concentrations

Abstract

Background. The K/DOQI guidelines recommend the use of albumin-corrected calcium (Ca), phosphate, parathyroid hormone (PTH) and calcium-phosphate product as therapeutic targets. The two most common assays for measuring albumin yield discordant results in uraemic patients, the Bromcresol purple (BCP) method providing lower albumin values than the Bromcresol green one (BCG). The aim of this study was to assess the impact of the assay on corrected Ca and, thus, on reaching recommended K/DOQI targets for corrected Ca and CaxP product.

Methods. We measured plasma albumin (both by BCG and BCP), total Ca and phosphate in all our chronic hemodialysis (HD) patients. Total Ca was corrected (“adjusted”) for albumin level by a formula proposed by the K/DOQI.

Results. 89 patients were included, aged 71.2  ±  11.5 years, on chronic hemodialysis for 29 (1-362) months. Albumin level was 3.78 ± 0.24 g/dL by BCG and 3.12 ± 0.27 by BCP (  p < 0.0001). Based on BCG albumin levels there were 12 cases of “hypocalcaemia” (<8.6 mg/dL), 3 cases of “hypercalcaemia” (>10 mg/dL) and 74 cases with “normal” Ca. The corresponding albumin levels were 3.9  ±  0.2; 3.1 ± 0.6 and 3.8 ± 0.2 g/dL, respectively. According to BCP albumin levels, only one patient was labelled as hypocalcaemia , 21 as hypercalcaemia et 67 as “normal” adjusted Ca (albumin 3.1; 3 ± 0.3 and 3.2 ± 0.3 g/dL, respectively). Depending on the use of BCG or BCP, a discrepancy was thus observed in 29 cases (32.6%): 18 cases were classified as hypercalcaemia when albumin was measured by BCP but were considered normal using BCG, whereas 11 cases classified as hypocalcaemia with BCG had normal adjusted Ca with BCP. Concerning CaxP product, 7 discrepancies were detected.

Discussion and conclusion. The choice of either BCG or BCP has a major impact on albumin-adjusted Ca and thus on reaching K/DOQI targets for Ca and CaxP product. Clinicians should take this fact into account for the interpretation of laboratory values and the prescription of drugs related to mineral metabolism and dialysate calcium concentration. The type of assay used for the measurement of albumin should also be recorded and its impact taken into account (or corrected) in multicentric studies and registries.

albumin, bromcresol green, bromcresol purple, corrected calcium, haemodialysis

Introduction

About 40% of serum calcium is bound to albumin, and only free and ionized calcium is biologically active in extracellular fluids. However, the routine measurement of ionized calcium is impractical taking into account severe preanalytical considerations (samples drawn anaerobically, impact of pH, of temperature and of any delay in processing…); as a consequence, the reproducibility of this parameter in clinical practice is poor. The K/DOQI guidelines therefore recommend the use of albumin-corrected (adjusted) total calcium for routine clinical practice, and the K/DOQI target for calcium actually relates to albumin-adjusted calcium. In accordance with the recommendations of the American Association of Clinical Chemistry, we use the wording adjusted rather than corrected, throughout the paper [ 1 ]. The K/DOQI guidelines recommend two conversion equations to perform the albumin adjustment: Orrel's and Payne's formulas [ 2 ].

Most clinical laboratories measure albumin in plasma or serum samples by automated dye-binding methods, which rely on a shift in the absorption spectrum of dyes such as bromcresol green (BCG) or purple (BCP) upon binding to albumin. Immunochemical methods, especially automated immunoturbidimetry and immunonephelometry, are more accurate and precise than colorimetric methods, but are generally time consuming, more expensive and as a consequence, rarely used in clinical laboratories. In current clinical practice, albumin is most frequently measured by BCG or BCP. These assays yield somewhat different results both in normal and chronic kidney disease (CKD) subjects [ 3,4 ]. Therefore, Appendix I of the K/DOQI guidelines for nutrition in chronic renal failure recommends the use of BCG, if available in the reference laboratory [ 5 ].

Surprisingly, the impact of the methodology used for measuring albumin on the adjustment of calcium levels and, thus, on reaching K/DOQI targets for adjusted calcium, has not been studied. We hypothesized that the classification of patients for the K/DOQI calcium target would be markedly different depending on the use of BCG or BCP. We thus simply measured the levels of plasma albumin using both BCG and BCP, total calcium and Ca×P product in our in-centre chronic haemodialysis patients, and thus assessed the impact of the albumin assay on the achievement of K/DOQI targets for calcium and Ca×P product.

Subjects and methods

We included all patients under maintenance haemodialysis in our hospital-HD unit. They all were dialyzed three times a week for 3.5–5 h per session using hollow-fibre Superflux polysulfone dialyzers [FX 50–FX 100 series from Fresenius (Bad Homburg, Germany)]. The standard blood flow was 360 mL/min, and bicarbonate dialysate flow was 500 mL/min.

Blood samples were obtained from non-fasting patients before the first haemodialysis session of the week. In all patients, total plasma calcium, phosphate and albumin were measured. Samples for measurement of total calcium and phosphate were collected in heparin tubes (20 U heparin/mL blood). Total calcium and phosphate were measured in the Laboratory of Clinical Chemistry in the Cliniques Universitaires at St-Luc on a Beckman Coulter Unicel D×C 800 analyser (Fullerton, CA, USA), with reagents from Beckman-Coulter, by indirect potentiometry and unreduced phosphomolybdate complex/UV, respectively, with interassay coefficient of variation (CV) ranging from 1.2 to 1.4% and 3.4 to 4.1%, respectively. Calibration was performed once a day for calcium and once every 2 days for phosphate. The laboratory reference range for total calcium was 8.6–10 mg/dL.

Albumin was measured in two plasma samples drawn consecutively in each patient, in heparinized tubes. One of these samples was analysed in the above-mentioned laboratory by BCP VIS photometry using a Beckman Coulter Unicel D×C 800 analyser (Fullerton, CA, USA), with reagents from Beckman Coulter, displaying interassay CVs ranging from 2.5 to 3.8%. The laboratory reference range for albumin was 3.5–5.2 g/dL. The second sample was used to measure albumin in the Laboratory of Medical Chemistry in Erasme Hospital by BCG VIS photometry on a Roche Modular P analyser (Basle, Switzerland) with reagents from Roche Diagnostics. The interassay CV ranged from 1.7 to 2.4%, with a local reference range of 3.4–4.8 g/dL. Calibrations were performed at least at all reagent lot change (approximately once every 2 weeks).

Total calcium was adjusted for plasma albumin level using one of the two formulas (Payne’s) proposed by the K/DOQI guidelines: corrected calcium (mg/dL) = total calcium (mg/dL) + 0.8 × [4 − albumin (g/dL)]. According to the K/DOQI recommendations, we defined hypocalcaemia and hypercalcaemia as values of albumin-adjusted calcium below or above the normal range of total calcium according to the local laboratory (8.6–10 mg/dL in this case). The Ca×P product was considered within a target when it was <55 mg ² /dL ² .

Data are presented as mean ± SD, median and range (if the distribution is not Gaussian) or percentage, as appropriate. The comparisons were performed using the paired student t -test. The statistical significance level was set at P < 0.05. On the basis of adjusted calcium level, albumin being measured both by BCG and BCP, each patient was classified into one of three groups: hypercalcaemia, hypocalcaemia or normocalcaemia. Agreements and discrepancies in the classification of patients depending on the method used to measure albumin were expressed as percentages. The degree of agreement between continuous variables was estimated according to Bland and Altman [ 6 ].

Results

Ninety-one patients were on chronic HD in our unit. In two of them, the sample for albumin measurement by BCG was missing. Overall, blood samples from 89 patients (48 males) were available for analysis. As shown in Table 1 , patients were aged 71.2 ± 11.5 years and were on chronic haemodialysis for 29 (1–362) months. Thirty-eight patients (42.7%) were diabetic. Forty-six (51.7%) were under calcium carbonate as phosphate binder (dosage: 1.9 ± 0.9 g/day). Twenty-four patients (27%) were treated by sevelamer hydrochloride (dosage: 3800 ± 1785 mg/day), and 10 patients (11.2%) by lanthanum carbonate (dosage: 1472 ± 690 mg/day). Ten patients (11.24%) were under calcitriol (2.1 ± 1.1 μg/week), and 10 patients under cinacalcet (dosage: 33 ± 9.5 mg/day). Sixty, 3 and 26 patients (67.4, 3.4 and 29.2%, respectively) were dialyzed with a calcium level in the dialysates of 1.75, 1.5 and 1.25 mmol/L, respectively.

Table 1

Characteristics of included patients

n = 89
Males (%)	48 (53.9)
Age (years)	71.2 ± 11.5
HD vintage (months)	29 (1–362)
Diabetes (%)	38 (42.7)
Use of phosphate binders
Calcium carbonate (%)	46 (51.7)
Sevelamer hydrochloride (%)	24 (27)
Lanthanum carbonate (%)	10 (11.2)
Use of calcitriol (%)	10 (11.2)
Use of cinacalcet (%)	10 (11.2)
Calcium level in the dialysate
1.75 mmol/L (%)	60 (67.4)
1.5 mmol/L (%)	3 (3.4)
1.25 mmol/L (%)	26 (29.2)
Non-corrected Ca level (mg/dL)	8.9 ± 0.37
Phosphorus level (mg/dL)	4.5 ± 1.2
Non-corrected Ca×P (mg 2 /dL 2 )	40 ± 11.1
Intact PTH level (pg/mL)	319 ± 256
CRP (mg/dL)	1.8 ± 1.9
Haemoglobin (g/dL)	11.2 ± 1.1
CO 2 (mmol/L)	21.3 ± 2.1

Overall, the total calcium level was 8.9 ± 0.37 mg/dL and the phosphate level 4.49 ± 1.2 mg/dL. The albumin level measured by BCG was 3.78 ± 0.24 g/dL, and 3.12 ± 0.27 g/ dL by BCP ( P < 0.0001). The mean difference in serum albumin between both methods was thus 0.66 ± 0.1 g/dL and the median difference 0.65 (0.43–0.94) g/dL. The mean total calcium adjusted for albumin measured by BCG and BCP was 9.08 ± 0.37 mg/dL and 9.62 ± 0.4 mg/dL, respectively ( P < 0.001), with a mean difference of 0.53  ±  0.08 mg/dL and a median difference of 0.52 (0.34– 0.79) mg/dL. Bland and Altman analysis showed a good agreement between BCG and BCP for albumin and albumin-adjusted calcium levels (Figure 1 ), although a systematic bias was observed. As shown in Table 2 , when total calcium was adjusted for albumin measured by BCG, 74 patients (83.1%) had a calcium level within the K/DOQI targets, 12 patients (13.5%) had hypocalcaemia and 3 patients (3.4%) had hypercalcaemia. The corresponding BCG-albumin levels were 3.8 ± 0.2 g/dL, 3.9 ± 0.2 mg/dL and 3.1 ± 0.6 mg/dL, respectively. In contrast, using albumin measured by BCP, 67 patients (75.3%) were normocalcaemic, 1 patient (1.1%) was hypocalcaemic and 21 patients (23.6%) were hypercalcaemic. Their albumin levels were 3.2 ± 0.3 g/dL, 3.1 g/dL and 3 ± 0.3 g/dL, respectively. Depending on the assay used for measuring albumin, a discrepancy was therefore observed in 29 cases (32.6%): 18 cases were classified as hypercalcaemia when albumin was measured by BCP but were considered normal using BCG, whereas 11 cases classified as hypocalcaemia with BCG had normal adjusted calcium with BCP.

Fig. 1

Bland and Altman analysis for albumin (top) and adjusted calcium (bottom) using BCG and BCP.

Table 2

Discrepancies on reaching K/DOQI targets

	Calcium adjusted for albumin by BCG n (%)	Calcium adjusted for albumin by BCP n (%)
Normocalcaemia	74 (83.1)	67 (75.3)
Hypocalcaemia	12 (13.5)	1 (1.1)
Hypercalcaemia	3 (3.4)	21 (23.6)
Ca×P >55 mg 2 /dL 2	12 (13.5)	19 (21.4)
Discrepancies
Calcaemia	29 (32.6)
Ca×P	7 (7.9)

Overall, 44 patients (49.5%) had phosphate levels within the K/DOQI targets, 27 patients (30.3%) had hypophosphataemia and 18 patients (20.2%) had hyperphosphataemia. Using unadjusted total calcium, the mean Ca×P product was 40 ± 11.1 mg ² /dL ² . The mean Ca×P product using a calcium level adjusted for albumin measured by BCG and BCP was 40.8 ± 11.4 mg ² /dL ² and 43.2 ± 12 mg ² /dL ² , respectively ( P < 0.001), with a mean difference of 2.4 ± 0.78 mg ² /dL ² , and a median difference of 2.3 (0.45–6.3) mg ² /dL ² . When BCG was used, 12 patients (13.5%) had a Ca×P product above the K/DOQI target, whereas 19 patients (21.4%) were above the target when albumin was measured by BCP. Thus, seven discrepancies were observed in the achievement of the Ca×P product target (Table 1 ). Using the other formula (Orrel’s) proposed by the K/DOQI guidelines to adjust the total calcium level for the albumin level [corrected calcium (mg/dL) = total calcium (mg/dL) + 0.0704 × [34 − albumin (g/dL)]] did not change the main results: the difference between mean total calcium and mean Ca×P product, both adjusted for albumin measured by either BCG or BCP, remained significant (8.64 ± 0.4 mg/dL and 9.1 ± 0.4 mg/dl, respectively, P < 0.001; 40 ± 11.1 mg ² /dL ² and 40.9 ± 11.4 mg ² /dL ² , respectively, P < 0.001). The classification of patients regarding KDOQI targets still showed substantial differences depending on the use of either BCG or BCP: a discrepancy in the achievement of calcaemia and Ca×P targets was observed in 28 (31.5%) and 2 (2.3%) cases, respectively.

Discussion

Depending on the method used for measuring albumin (BCG or BCP), our results show a discrepancy in the classification of haemodialysis patients for the albumin-adjusted calcium level in about one-third of the cases. Since the routine measurement of ionized calcium (considered as the gold standard calcium level) is difficult to perform, total calcium adjusted for albumin level has been recommended as a surrogate marker by the K/DOQI guidelines [ 2 ]. Recently, Kidney Disease: Improving Global Outcomes (KDIGO) confirmed that measuring ionized calcium is the preferred method for evaluating serum or plasma calcium, but that given the difficulties in sample processing and associated cost, its clinical use may not be widely implemented [ 7 ]. If total calcium is used instead, it should be adjusted for albumin in the case of hypoalbuminaemia, although KDIGO acknowledges that ‘there is currently a lack of standardization of the formulas used to determine adjusted calcium’. The impact of the method used for the measurement of albumin on adjusted calcium level is not even mentioned.

As the most specific method for albumin measurement (immunonephelometry) is expensive, most laboratories currently use dye-binding methods (BCG or BCP) for measuring albumin. In a recent annual report of the U.S. Renal Data System, among the forms indicating the laboratory method used for measuring albumin (42% of all forms), 74% used BCG and 26% BCP [ 8 ]. BCG and BCP yield different values, and these differences have long been recognized. BCG is known to underestimate albumin in the high normal range, and overestimate albumin below the normal range [ 3 ]. BCP agrees very closely with the gold standard of nephelometry because of the absence of cross-reactivity with non-albumin proteins. Nevertheless, in uraemic patients BCP underestimates albumin levels, especially in chronic haemodialysis patients, probably because uraemic toxins inhibit the binding of BCP to albumin, but not that of BCG [ 9 ]. Joseph et al . [ 4 ] compared albumin measured by BCG and BCP against the gold standard of nephelometry in 23 peritoneal dialysis (PD) and 53 HD patients. BCG correlated well with nephelometry, both in PD and HD patients, with a slight (4%) overestimation by BCG in HD patients. In contrast, BCP provided figures lower than those obtained by immunonephelometry by ∼10% in PD and 19% in HD patients. Beyer et al . also observed albumin underestimation by BCP (7.9%) compared to nephelometry in HD, but not in PD and in CKD patients not on dialysis [ 10 ]. Emphasizing the relevance of the discrepancies for the albumin levels with the most widely used methods, Clase et al . proposed the following interconversion formula, obtained in 535 patients with CKD (HD, PD, renal transplant and CKD patients) and without CKD, applicable even when serum albumin is <3 g/dL: albumin (BCG) = 5.5 + albumin (BCP), in g/L [ 11 ]. Although this work was published in 2001, it was not mentioned by the K/DOQI guidelines (2003), which simply recommend that the BCG method should be requested, if available [ 5 ]. A number of studies have investigated the correlation between ionized calcium, non-adjusted total calcium and albumin-adjusted calcium calculated by different equations, as well as their impact on the classification according to calcium level [ 12–15 ]. But none of these studies took into account the impact of the assay used for measuring albumin.

The aim of our work was neither to identify the gold standard method for measuring albumin in clinical practice, nor to propose the best adjustment (correction) formula, nor to study the correlation between adjusted, total and ionized calcium. We simply show that the choice of one or the other of the two commonly used methods for albumin measurement has a major impact on adjusted calcium and thus on the classification of the patients according to K/DOQI albumin-adjusted calcium target. In the management of individual patients, this may lead to inappropriate decisions regarding the use of calcium-containing phosphate binders, vitamin D compounds, cinacalcet and dialysate calcium concentration. In epidemiological studies or registries, the lack of mention of the method used for albumin measurement may also result in substantial misinterpretation or misclassification. A conversion formula for adjusted calcium depending on the use of BCG or BCP should ideally be developed in a large cohort of CKD patients.

Surprisingly, the type of assay used for albumin measurement was not mentioned in many recent publications. In a Pubmed research, among the 15 studies found using ‘calcium’, ‘albumin’ and ‘haemodialysis’ as keywords, and published from January 2006 to June 2008 in five major nephrological journals [ 14–28 ], only five (33.3%) indicated the method used for albumin measurement [ 14–18 ]. Similarly, in the annual report of the U.S. Renal Data System, only 42% of the forms indicated the laboratory method used for measuring albumin in 2005 [ 8 ]. The risk of substantial errors further increases when results of different registries are compared or mixed in systematic reviews or meta-analyses.

One limitation of our study should be acknowledged: we only included a relatively small number of HD patients. Thus larger studies, both in HD and other populations with kidney disease such as PD, kidney graft recipients and CKD patients, should be performed.

In conclusion, the choice of either BCG or BCP has a major impact on albumin-adjusted Ca and thus on reaching K/DOQI targets for Ca and Ca×P product. Clinicians should take this fact into account for the routine interpretation of laboratory values and the prescription of several drugs and dialysate calcium concentration. The type of assay used for the measurement of albumin should also be recorded and its impact taken into account (or corrected) in multicentric studies and registries.

Conflict of interest statement . None declared.

© Oxford University Press 2009