Misclassification of Calcium Status Based on Albumin-Adjusted Calcium: Studies in a Tertiary Hospital Setting

Abstract

BACKGROUND

Clinical laboratories measure total calcium and adjust for albumin concentrations to predict calcium status. We compared total and adjusted calcium (Adj-Ca) with ionized calcium (Ca²⁺) for correct assignment of calcium status. The effect of restriction of Adj-Ca reporting in patients with hypoalbuminemia was determined on the basis of frequency of misclassifications.

METHODS

Extraction of laboratory results was performed for 24 months. Adj-Ca was calculated from a modified Payne formula. A further prospective data set for 6 months was collected after stopping reporting of Adj-Ca for patients with an albumin <3.0 g/dL. The agreement between Ca²⁺ and Adj-Ca or total Ca was assessed with Cohen's kappa statistic.

RESULTS

In 5553 hospitalized patients, 13604 paired Ca²⁺ results were analyzed retrospectively. Prospective collection in 1113 paired samples was from 450 patients. Adj-Ca was a poor predictor of calcium status compared to the Ca²⁺ reference standard in both data sets (agreement 56.9% in the first, 65.6% in the second data set). Renal failure and low albumin concentrations were associated with worse agreement between Adj-Ca and Ca²⁺. Restriction of reporting of Adj-Ca to albumin concentrations >3.0g/dL improved correct classification of calcium status from 65.6% to 77.6% (P < 0.0001). Total Ca performed better than Adj-Ca for low albumin (<3.0g/dL) and performed similarly in samples with albumin >3.0g/dL.

CONCLUSIONS

Adj-Ca is unreliable for the classification of calcium status in hospital patients when compared to Ca²⁺. Adj-Ca overestimates calcium for patients with renal impairment and albumin concentrations <3.0g/dL. Restriction of reporting Adj-Ca for albumin below 3.0 g/dL reduces the number of misclassified patients.

The measurement of calcium in blood is one of the most commonly requested laboratory tests. Calcium concentrations are tightly controlled, and abnormalities of calcium concentration can have widespread effects on neurological, renal, gastrointestinal, and bone function (1). Accurate laboratory assessment of patient calcium status is essential in several areas of clinical medicine including the diagnosis of primary hyperparathyroidism and the treatment and monitoring of chronic kidney disease (2, 3).

Calcium circulates in the blood in 3 forms. Approximately 50% circulates as free Ca²⁺ (ionized calcium) with the remainder circulating bound to plasma proteins (40% bound predominantly to albumin) or complexed with anions (around 10% complexed with phosphate, bicarbonate, or lactate) (4, 5). It is the Ca²⁺ fraction, which is biologically active and under homeostatic control through the action of the calcium-sensing receptor. The calcium-sensing receptor maintains stable Ca²⁺ concentrations through regulation of parathyroid hormone secretion by the parathyroid glands and through parathyroid hormone–dependent and parathyroid hormone–independent mechanisms in the kidney and bone (6).

Despite Ca²⁺ being the physiologically important measurand, clinical laboratories routinely measure total calcium. Dye binding methods such as o-cresolphthalein or arsenazo (III) are most frequently used in routine laboratories (7). The alternative of direct measurement of Ca²⁺ is complicated by problems due to preanalytical requirements and manual handling precluding full automation (2). pH changes due to delayed separation and exposure to ambient air seem to be the most important preanalytical concerns (8). The preanalytical requirements can be surmounted by prompt separation of serum from red cells and the use of a sealed container until analysis (2).

In a large number of different patient populations there are drawbacks to measuring total calcium concentrations. In patients with hypoalbuminemia, total calcium results appear low despite normal Ca²⁺ owing to the substantially lower fraction of calcium bound to albumin (9). This low result has led to several formulas to “adjust” the total calcium for albumin concentration (adjusted calcium, Adj-Ca) (9–11). Several studies have questioned the ability of Adj-Ca to correctly assign patient calcium status in certain groups, particularly in patients with chronic kidney disease (10, 12, 13), critical illness (14, 15), and primary hyperparathyroidism (16). In addition, changes in the binding affinity constant of albumin for calcium as the albumin concentration decreases limit the utility of Adj-Ca in patients with hypoalbuminemia (2, 17). Common practice does not reflect these limitations, and requesting and reporting of Adj-Ca is widespread. Some laboratories restrict the reporting of albumin-adjusted calcium for patients with low albumin concentrations or add an interpretive comment stating that Adj-Ca may be misleading in certain clinical conditions (e.g., renal failure, acid–base disorders). We adopted a practice of restricting reporting of Adj-Ca for patients with an albumin of <3.0 g/dL and recommending Ca²⁺ measurement via an interpretive result comment in these patients in late 2016.

We investigated the ability of Adj-Ca to correctly assign calcium status as compared with Ca²⁺ as the reference standard and assessed in which patient groups calcium status estimation by Adj-Ca may be misrepresented. Classifications based on total unadjusted calcium were also assessed. An additional aim was to determine the effect of the restriction of Adj-Ca reporting in patients with hypoalbuminemia (<3.0 g/dL) on the number of patient results released with misclassified calcium status.

Materials and Methods

Alfred Pathology services 3 hospital campuses, including a >500-bed tertiary hospital that houses a large emergency and trauma center, a 45-bed intensive care unit, heart/lung/kidney and bone marrow transplant units, and other specialty medical and surgical services. This study was approved by the Alfred Health ethics committee (Project no. 642/17) and was performed in accordance with the Declaration of Helsinki.

DATA EXTRACTION

We obtained 2 separate data sets. The first retrospectively covered the 24 months before restricted reporting of Adj-Ca for patients with an albumin of <3.0 g/dL and a prospective data set covering a 6-month period following the practice change.

For the first data set, results for total calcium, albumin, creatinine in plasma, and pH and Ca²⁺ in whole blood samples collected within 30 min of each other were extracted from the laboratory information system (Cerner Pathnet Millenium). Patients younger than 18 years and patients with a pH outside the reference interval of 7.35–7.45 were excluded to reduce any bias introduced into Ca²⁺ measurements. From 5553 individual patients, 13604 paired results were available for analysis.

A second data set was extracted with results for total calcium, albumin, creatinine in plasma, and Ca²⁺ in a separate serum tube collected on the same blood draw. These data were collected prospectively for 6 months following the laboratory restriction of Adj-Ca reporting. From 450 individual patients, 1112 paired results were available for analysis.

LABORATORY METHODS

Calcium, albumin, and creatinine were measured in plasma samples on an Abbott Architect ci16200 system for both data sets. Calcium was measured by the arsenazo (III) method (analytical CV, 1%). Albumin was measured by the bromcresol purple method, and creatinine was measured by the alkaline picrate kinetic method (analyzers and reagents from Abbott). The Chronic Kidney Disease Epidemiology Collaboration equation was used to calculate estimated glomerular filtration rate (eGFR) from serum creatinine (18). For the first data set, pH and Ca²⁺ were measured in venous or arterial whole blood taken in a blood gas syringe on Siemens Rapidlab 1260 blood gas analyzers. For the second data set, Ca²⁺ was measured in serum in tubes filled with no headspace, centrifuged, and sealed until measurement on Siemens Rapidpoint 500 blood gas analyzers. Ca²⁺ concentrations reported were not corrected for pH.

Albumin-adjusted total calcium was calculated from a modified Payne formula (7), with the slope determined by an inhouse regression for our bromcresol purple albumin method against total calcium.

$$\mathrm{Adjusted}{\text{calcium}}_{\mathrm{SI}\mathrm{units}}\left(\text{mmol}/\text{L}\right)=\mathrm{total}\mathrm{calcium}\left(\text{mmol}/\text{L}\right)+0.018\times \left[40-\mathrm{albumin}\left(\text{g}/\text{L}\right)\right]$$

$$\mathrm{Adjusted}{\text{calcium}}_{\mathrm{conventional}\mathrm{units}}\left(\text{mg}/\text{dL}\right)=\mathrm{total}\mathrm{calcium}\left(\text{mg}/\text{dL}\right)+0.72\times \left[4-\mathrm{albumin}\left(\text{g}/\text{dL}\right)\right]$$

CLASSIFICATION OF CALCIUM STATUS

Ca²⁺ measurements were used as reference standard for the classification of calcium status. The reference interval for Ca²⁺ in our laboratory is 4.44–5.12 mg/dL (SI units, 1.11–1.28 mmol/L); this interval was derived previously in our laboratory from 77 healthy individuals. Patients were classified as hypo-, normo-, or hypercalcemic. This classification was compared with classifications based on paired Adj-Ca by use of a reference interval of 8.4–10.4 mg/dL (SI units, 2.1–2.6 mmol/L, AACB harmonized reference interval) (19).

To further investigate the effect of albumin on classification of calcium status by Adj-Ca, we plotted [Adj-Ca minus Ca²⁺] against albumin concentrations for both data sets.

STATISTICAL ANALYSES

Descriptive statistics were performed in Microsoft Excel (version 14). The agreement between Ca²⁺ albumin-adjusted calcium was assessed with Cohen's kappa statistic by Analyse-it for Microsoft Excel (version 2.20, Analyse-it Software, Ltd.). χ² tests were performed for the difference between proportions of correctly classified, released, and restricted results. For visualization, data were imported into the R statistical computing environment (20). Figures were plotted with the R package ggplot2 (21).

Results

PATIENT CHARACTERISTICS

Patient characteristics for both data sets are shown in Table 1. For the first data set, samples were from 3612 male (median age, 61 years; range, 18–97) and 1941 female (median age, 61 years; range, 18–98) patients. For the 5553 individual patients, there were 13604 paired Ca²⁺ and Adj-Ca measurements. We excluded 12810 (48%) Ca²⁺ measurements with pH outside the reference interval (pH, 7.35–7.45). The majority (63%) of measurements were performed in a critical care (intensive care unit or emergency department) setting (n = 8472).

Table 1.

Patient characteristics.^a

	24-month data (before reporting restriction)	6-month data (after reporting restriction)
Individual patients, n	5553	450
Men, n (%)	3612 (65)	231 (52)
Age, years, median (range)	61 (18–97)	68 (18–92)
Women, n (%)	1941 (35)	219 (48)
Age, years, median (range)	61 (18–98)	71 (19–99)
Paired measurements, n	13604	1112
Patient location, n (%)
EDb or ICU	8472 (62)	16 (1.4)
Ward: hospital	4036 (30)	498 (45)
Ward: subacute care	936 (7)	175 (15.6)
Renal or dialysis	160 (1)	423 (38)
Ionized calcium, mg/dL, median (IQR)	4.48 (4.32–4.68)	4.68 (4.4–4.96)
Total calcium, mg/dl, median (IQR)	8.40 (8.00–8.92)	8.88 (8.24–9.48)
Adjusted calcium, mg/dL, median (IQR)	9.48 (9.00–9.92)	9.80 (9.36–10.44)
Albumin, g/dL, median (IQR)	2.6 (2.1–3.1)	2.6 (2.2–3.0)
eGFR, n (%)
>90 mL/min/m2	5288 (39)	193 (17)
60–90 mL/min/m2	3955 (29)	252 (23)
30–59 mL/min/m2	2856 (21)	179 (16)
15–29 mL/min/m2	1080 (8)	101 (9)
<15 mL/min/m2	425 (3)	388 (35)

a

Conversion factors: calcium and Ca²⁺ conventional unit (mg/dL) to SI unit (mmol/L) = 0.25; albumin conventional unit (g/dL) to SI unit (g/L) = 10.

b

ED, emergency department; ICU, intensive care unit; IQR, interquartile range.

The second data set samples were from 450 individual patients, 231 men (median age, 68 years; range 18–92) and 219 women (median age, 71 years; range, 19–99). For the 450 individual patients, there were 1113 paired Ca²⁺ and Adj-Ca measurements. In the second data set, most measurements were from ward patients (45%) in the hospital or in subacute care (15.6%). Thirty-eight percent of measurements were taken from patients on the renal and dialysis wards, likely reflecting a practice change in response to the change in Adj-Ca reporting.

CLASSIFICATION OF CALCIUM STATUS

For paired measurements in the first data set, Adj-Ca was a poor predictor of calcium status compared to the Ca²⁺ reference standard, with an overall agreement of 0.56 and a κ of 0.10 (95% CI, 0.09–0.12). In the second data set, calcium status based on Adj-Ca showed better agreement with calcium status based on Ca²⁺ but was still a poor predictor (observed agreement = 0.66; κ statistic = 0.38; 95% CI, 0.33–0.43). Total calcium, without adjustment, performed better than Adj-Ca at classifying calcium status in both data sets, with an observed agreement of 0.74 (κ = 0.49) for the first data set and 0.74 (κ = 0.55) for the second data set. Adj-Ca was more accurate than total calcium in the first data set for classifying hypercalcemia (71% vs 30%) and normocalcemia (88% vs 70%), but worse at correctly classifying hypocalcemia (11% vs 80%). Findings were similar in the second data set, except that adjusted and unadjusted total calcium were similar for correctly classifying normocalcemia. Tables 2 and 3 show the obtained classifications.

Table 2.

Classification of calcium status for all patients (24-month data before reporting restriction).

Classification by ionized Ca	Classification by Albumin-Adjusted Ca, n (% of total)a			Total
	Hypercalcemia	Normocalcemia	Hypocalcemia
Hypercalcemia	187 (1.4)	76 (0.6)	0	263 (2.0)
Normocalcemia	895 (6.6)	6916 (50.8)	48 (0.4)	7859 (57.8)
Hypocalcemia	153 (1.1)	4695 (34.5)	634 (4.6)	5482 (40.2)
Total	1235 (9.1)	11687 (85.9)	682 (5.0)	13604

Classification by Ionized Ca	Classification by Total Ca, n (% of total)b			Total
	Hypercalcemia	Normocalcemia	Hypocalcemia
Hypercalcemia	81 (0.6)	181 (13.3)	1 (0)	263 (2.0)
Normocalcemia	69 (0.5)	5536 (40.7)	2254 (16.6)	7859 (57.8)
Hypocalcemia	9 (0.1)	1072 (7.9)	4401 (32.4)	5482 (40.2)
Total	159 (1.2)	6789 (49.9)	6656 (48.9)	13604

a

Observed agreement = 0.56, κ = 0.10 (95% CI, 0.09–0.12).

b

Observed agreement = 0.74, κ = 0.49 (95% CI, 0.47–0.50).

Table 3.

Classification of calcium status for all patients (6-month data after reporting restriction).

Classification by ionized Ca	Classification by Albumin-Adjusted Ca, n (% of total)a			Total
	Hypercalcemia	Normocalcemia	Hypocalcemia
Hypercalcemia	171 (15.3)	8 (0.7)	0	179 (16.0)
Normocalcemia	120 (10.8)	508 (45.7)	2 (0.2)	630 (56.7)
Hypocalcemia	1 (0.1)	251 (22.6)	51 (4.6)	303 (27.3)
Total	292 (26.2)	767 (69.0)	53 (4.8)	1112

Classification by ionized Ca	Classification by Total Ca, n (% of total)b			Total
	Hypercalcemia	Normocalcemia	Hypocalcemia
Hypercalcemia	97 (8.7)	82 (7.4)	0 (0)	179 (16.0)
Normocalcemia	3 (0.3)	514 (46.2)	113 (10.2)	630 (56.7)
Hypocalcemia	0 (0)	81 (7.3)	222 (20.0)	303 (27.3)
Total	100 (9.0)	677 (59.9)	335 (30.2)	1112

a

Observed agreement = 0.66, κ = 0.38 (95% CI, 0.33–0.43).

b

Observed agreement = 0.74, κ = 0.55 (95% CI, 0.51–0.60).

In both data sets, Adj-Ca concentrations “over-correct,” misclassifying calcium status as “hypercalcemia” with a normal measured Ca²⁺ (6.5% and 11% of all measurements in data sets 1 and 2, respectively) and “normocalcemia” when Ca²⁺ was low (35% and 23% of all measurements in data sets 1 and 2, respectively).

We explored the effect of albumin concentration and renal function by stratifying paired measurements for patients with and without renal impairment for albumin concentrations. Agreement between Ca²⁺ and Adj-Ca varied markedly with albumin concentrations. For patients with an eGFR ≥60 mL/min/1.73m² and albumin >3.5 g/dL, albumin-adjusted calcium showed good agreement with Ca²⁺ [observed agreement = 0.85, κ = 0.48 (95% CI, 0.41–0.54) for data set 1; observed agreement = 0.93, κ = 0.84 (95% CI, 0.66–1.00) for data set 2]. With lower albumin concentrations and renal impairment, the ability of Adj-Ca to correctly classify calcium status decreased. The observed agreement became worse below albumin concentrations of 3.0 g/dL for patients with and without renal impairment as shown in Tables 1 and 2 in the Data Supplement that accompanies the online version of this article at http://www.clinchem.org/content/vol64/issue12.

For a correction equation assuming a linear relationship in calcium binding to albumin, such as the modified Payne formula used here (Eq. 1), a constant difference between Adj-Ca and Ca²⁺ across all albumin concentrations would be expected. Figs. 1 and 2 show the mean ±2 standard deviations for the difference between Adj-Ca and Ca²⁺ measurements for different albumin concentrations. An inflection upwards is seen below albumin of 3.0 g/dL, with the difference between Adj-Ca and Ca²⁺ above this albumin concentration being relatively constant. This indicates that the linear relationship proposed in Eq. 1 changes inflection and cannot be confidently used below albumin concentrations of approximately 3.0 g/dL.

Albumin-adjusted calcium minus ionized calcium for different albumin levels (24 months of retrospective data).

Fig. 1.

Albumin-adjusted calcium minus ionized calcium for different albumin levels (6 months of prospective data).

Fig. 2.

EFFECT OF RESTRICTED REPORTING OF Adj-Ca FOR PATIENTS WITH ALBUMIN <3.0 g/dL

With the 6 months of data following the reporting restriction, the agreement between calcium status classification based on Adj-Ca and Ca²⁺ was compared for the group in which Adj-Ca was reported (albumin ≥3.0 g/dL) and the group in which it was restricted (albumin <3.0g/dL).

For the released results, 77% were classified correctly (253 of 326), compared with 61% (478 of 786) of restricted results. The difference in proportions was statistically significant (χ² test, P < 0.0001). For the reported Adj-Ca results, Table 4 shows that 41 of 59 (69.5%) of the hypercalcemic results by Adj-Ca were correct; 187 out of 240 (78%) of the normocalcemic results were correct, and 25 out of 27 (92.5%) the hypocalcemic results were correct. Table 5 shows that 55.8% of the hypercalcemic results by Adj-Ca were correct; 61% of the normocalcemic and 100% of the hypocalcemic results were correct for restricted results.

Table 4.

Classification of calcium status for released results (albumin ≥3 g/dL).

Classification by ionized Ca	Classification by Albumin-Adjusted Ca, n (% of total)a			Total
	Hypercalcemia	Normocalcemia	Hypocalcemia
Hypercalcemia	41 (12.5)	1 (0.3)	0 (0)	42 (12.9)
Normocalcemia	18 (5.5)	187 (57.4)	2 (0.6)	207 (63.5)
Hypocalcemia	0 (0)	52 (16.0)	25 (7.7)	77 (23.6)
Total	59 (18.1)	240 (73.6)	27 (8.2)	326

Classification by ionized Ca	Classification by Total Ca, n (% of total)b			Total
	Hypercalcemia	Normocalcemia	Hypocalcemia
Hypercalcemia	29 (8.9)	13 (4.0)	0 (0)	42 (12.9)
Normocalcemia	2 (0.6)	195 (59.8)	10 (3.1)	207 (63.5)
Hypocalcemia	0 (0)	30 (9.2)	47 (14.4)	77 (23.6)
Total	59 (9.5)	238 (73.0)	27 (17.5)	326

a

Observed agreement = 0.77, κ = 0.54 (95% CI, 0.45–0.63).

b

Observed agreement = 0.83, κ = 0.65 (95% CI, 0.57–0.73).

Table 5.

Classification of calcium status for restricted results (albumin <3 g/dL).

Classification by ionized Ca	Classification by Albumin-Adjusted Ca, n (% of total)a			Total
	Hypercalcemia	Normocalcemia	Hypocalcemia
Hypercalcemia	130 (16.5)	7 (0.9)	0 (0)	137 (17.4)
Normocalcemia	102 (12.9)	321 (40.8)	0 (0)	423 (53.8)
Hypocalcemia	1 (0.1)	198 (25.2)	27 (3.4)	226 (28.8)
Total	233 (29.6)	526 (66.9)	27 (3.4)	786

Classification by ionized Ca	Classification by Total Ca, n (% of total)b			Total
	Hypercalcemia	Normocalcemia	Hypocalcemia
Hypercalcemia	68 (8.7)	69 (8.8)	0 (0)	137 (17.4)
Normocalcemia	1 (0.1)	319 (40.6)	103 (13.1)	423 (53.8)
Hypocalcemia	0 (0)	51 (6.5)	175 (22.3)	226 (28.8)
Total	70 (8.8)	439 (55.9)	278 (35.4)	786

a

Observed agreement = 0.61, κ = 0.32 (95% CI, 0.27–0.38).

b

Observed agreement = 0.72, κ = 0.51 (95% CI, 0.46–0.57).

Classifications are also shown based on total calcium. The difference between classifications for patients with an albumin ≥3.0 g/dL was not significant [83% vs 77% for total calcium and Adj-Ca, respectively (P = 0.11)]. For patients with an albumin below 3.0 g/dL, total calcium performed better than Adj-Ca [72% vs 61% for total calcium and Adj-Ca, respectively (P < 0.0001)].

Discussion

In this study we show that the classification of calcium status based on Adj-Ca can be misleading, mainly by overestimation of calcium concentrations. This overestimation results in the misclassification of actual normal Ca²⁺ as hypercalcemia and actual low Ca²⁺ as normocalcemia. For patients with albumin concentrations above 3.0 g/dL and normal renal function, the magnitude of this distortion is minimal, with good agreement between statuses defined by Adj-Ca and Ca²⁺. The effect is more pronounced in patients with renal impairment and decreasing albumin concentrations. The effect of different patient cohorts is reflected in the differences between the first (predominantly acute care) and second (predominantly chronic renal failure) patient data sets.

The poor ability of Adj-Ca to classify calcium status in patients with renal impairment is well described (3, 10, 12, 22). In a cohort of hemodialysis patients, Adj-Ca underestimated hypocalcemia and overestimated hypercalcemia (3); further, Adj-Ca showed a tendency to underestimate hypocalcemia (22). In patients with stage 3–5 chronic kidney disease who are not on hemodialysis, Adj-Ca also underestimates hypocalcemia and overestimates hypercalcemia (12). Our findings are consistent with these studies. In the 24 months data and the 6 months data, there was worse agreement between Adj-Ca and Ca²⁺ status for patients with an eGFR <60 mL/min/1.73m² independent of albumin. A large number of patients from the renal and dialysis ward contributed to the prospective patient cohort, with 58% of the paired measurements taken from patients with an eGFR <15 mL/min/1.73m². For these patients there was poor observed agreement between Adj-Ca and Ca²⁺ assigned status (observed agreement = 0.57, κ = 0.27; 95% CI, 0.21–0.33) with overestimation of hypercalcemia and underestimation of hypocalcemia with Adj-Ca. As mentioned by Goransson et al., the false classification of hypercalcemia with Adj-Ca in these patients may lead to inappropriate withdrawal of vitamin D therapy or adjustment of calcium concentration in the dialysate fluid (3). The misclassification of hypercalcemia by Adj-Ca in this population also has implications for the management of chronic kidney disease mineral and bone disorder. Current guidelines recommend the avoidance of hypercalcemia in adults and maintenance of serum calcium within a targeted range (23).

A striking finding of this study was the decrease in the accuracy of Adj-Ca to classify calcium status with decreasing albumin concentrations for all patients. It has been reported that there are changes in the binding affinity constant of albumin for calcium at low albumin concentrations (17, 24). Besarab et al. found that, in a series of equilibrium dialysis studies investigating free calcium at different albumin concentrations, there was a relatively fixed affinity constant for albumin concentrations between 3.0 g/dL and 5.0 g/dL (17) but a “radical” alteration in calcium–albumin binding for concentrations below 3.0 g/dL, at which a unique affinity constant could not be determined (17). They cautioned against using albumin-adjustment formulas in patients with hypoalbuminemia (17). There are few studies that have investigated this relationship in patients. In a population of 534 patients it was shown that the binding affinity of albumin to calcium increases at low albumin concentrations and that Adj-Ca overestimated calcium for patients in their lowest albumin group (<3.7 g/dL) (24). Additionally, in a recent study of 691 patients with chronic kidney disease, for patients with albumin concentrations <3.0 g/dL, the overestimation of calcium by Adj-Ca was increased compared to normoalbuminemic patients (12). In contrast, Slomp et al. showed that the poor reliability of Adj-Ca to assign calcium status in critical care patients was independent of albumin concentration (14). Their patient population, however, was small (n = 36), and all patients had albumin concentrations below 3.0 g/dL, so the effect of albumin may not have been apparent (14). In contrast, we analyzed 2 large data sets, both retrospectively and prospectively comparing ionized with Adj-Ca. When we plotted Adj-Ca minus Ca²⁺ against albumin concentrations for both data sets, nonlinearity was observed, with the difference in Adj-Ca and Ca²⁺ rising below albumin concentrations of 3.0 g/dL (Figs. 1 and 2). Our study reinforces the early work of Besarab and Pedersen and confirms in a larger patient population that at albumin concentrations below 3.0 g/dL, Adj-Ca becomes unreliable at assigning calcium status. It can be speculated that a 2-spline correction equation with a cutpoint at an albumin concentration of 3.0 g/dL may have better performance at classifying calcium status across all albumin concentrations.

With regard to the laboratory's decision to limit reporting for Adj-Ca in patients with albumin <3.0 g/dL, the data described here are reassuring. For the released results, significantly more results were correctly classified by Adj-Ca than for the restricted results (77% vs 61%, P < 0.0001). This result is due to a reduction in misclassifications of “hypercalcemia when normocalcemic” and “normocalcemia when hypocalcemic.” The classification based on total calcium is considerably more accurate than classification based on Adj-Ca in patients with low albumin. The main misclassification in this group is false hypocalcemia (13.1%, see Table 5) as compared to 38.1% with albumin adjustment. This result is consistent with findings from other groups in which adjustment equations were not shown to outperform total calcium, particularly in patients with renal failure and in a critical care setting (10, 25).

We did not find that Adj-Ca was better than total calcium for classifying patients with albumin above 3.0 g/dL, likely owing to the high proportion of patients with an eGFR <60 mL/min/1.73m² in our cohort. Table 2 in the online Data Supplement shows that agreement for Adj-Ca improves when patients with an eGFR <60 mL/min/1.73 m² are excluded. For patients with an albumin <3.0 g/dL, total calcium may provide a better indication of calcium status when a Ca²⁺ concentration cannot be obtained.

The reduction in released misclassified results occurs at the expense of the restriction of the correctly classified results (a total of 478 results from 1113 requests). We educated clinicians regarding the unreliability of Adj-Ca at low albumin concentrations and noted a large proportion (38%) of measurements in the prospective cohort taken for patients on the renal and dialysis wards. This finding probably reflects a practice change in response to the change in Adj-Ca reporting. In addition, total calcium results are reported, which may provide a better indication of calcium status when a Ca²⁺ concentration cannot be obtained.

A limitation of our study was the exclusion of a number of measurements when the pH was outside the reference interval. A substantial proportion of the measurements were taken in a critical care setting. Further, the second data set included a substantial proportion of measurements taken from patients with advanced renal failure. Care should be taken when generalizing our conclusions to other patient groups, particularly in an ambulatory setting. However, because of the large numbers of measurements we have assessed, and the similar findings regarding the effect of albumin and renal function across 2 quite different patient settings, we feel that the conclusions are widely applicable in a tertiary hospital setting. Because low albumin concentrations are frequently observed in the acutely ill patient population and because most physicians are not aware of this limitation (see also calculators in UpToDate written by experts in the field of abnormal calcium metabolism (26)), this limitation should be communicated to the clinical community. We only assess one calcium adjustment formula in this study (a modified Payne formula). Several correction equations are in use, and their application to this data set may provide different conclusions. The Payne formula is one of the more widespread calcium adjustment formulas used by clinical laboratories and in online calculators, so our findings have wide applicability (26). For assessment of the performance of other equations, we direct the reader to studies by Dickerson and by Lian and Åsberg (25, 27). The bromcresol purple albumin method used by our laboratory may underestimate albumin in uremic patients and those on hemodialysis and therefore lead to an overestimation in Adj-Ca (28). Our conclusions in patients may have less applicability to laboratories using bromcresol green albumin methods.

We note the improved agreement between Adj-Ca and Ca²⁺ for the 6 months data set following the restriction of Adj-Ca reporting than for the 24 months data set (observed agreement, 0.66 vs 0.56; κ, 0.38 vs 0.10).

We believe that selection bias is the most likely contributor to the poorer agreement found for the initial patient cohort because the proportion of measurements performed in a critical care setting (62% for the retrospective data vs 1.4% for the prospective data) differed greatly. The poor predictive capacity of Adj-Ca to assign calcium status in critical care patients has been shown by several groups (14, 15). This poor predictive capacity may be a result of several factors specific to critical care patients that disrupt calcium binding to albumin, including pH abnormalities, calcium binding by infused heparin, metabolic abnormalities causing raised free fatty acids that may compete with calcium binding sites on albumin, and complexation of Ca²⁺ by increased quantities of anions such as lactate (14). We excluded patients with pH abnormalities, but the findings in the data set 24 months before the reporting restriction should be interpreted with these factors in mind. Overall, for the first patient group, even with a considerable proportion of measurements from a critical care setting, in patients with normal albumin concentrations and renal function, Adj-Ca can be applied confidently to assign calcium status.

Conclusions

In this study we have shown that Adj-Ca, using a commonly applied formula, is unreliable for the classification of calcium status in particular patient groups when compared to Ca²⁺. In patients with normal albumin concentrations and renal function, Adj-Ca can be applied confidently to assign calcium status. However, for patients with renal impairment and albumin concentrations below 3.0g/dL, Adj-Ca overestimates calcium status, causing normocalcemic patients to be classified as hypercalcemic and hypocalcemic patients to be classified as normocalcemic. This overestimation may be in part due to the linearity of the correction equation breaking down below albumin concentrations of 3.0 g/dL. In these patients, total calcium has substantially better agreement with calcium status based on Ca²⁺. A laboratory restriction of reporting Adj-Ca for albumin below 3.0 g/dL in a tertiary hospital setting reduces the number of released Adj-Ca results that misclassify patient calcium status.

Author Contributions:All authors confirmed they have contributed to the intellectual content of this paper and have met the following 4 requirements: (a) significant contributions to the conception and design, acquisition of data, or analysis and interpretation of data; (b) drafting or revising the article for intellectual content; (c) final approval of the published article; and (d) agreement to be accountable for all aspects of the article thus ensuring that questions related to the accuracy or integrity of any part of the article are appropriately investigated and resolved.

J.D. Smith, statistical analysis; H.G. Schneider, administrative support.

Authors' Disclosures or Potential Conflicts of Interest:No authors declared any potential conflicts of interest.

Role of Sponsor: No sponsor was declared.