Linear regression

Surrogate endpoint

Statistics

Calcium

Creatinine

Receiver operating characteristic

Retrospective cohort study

Mathematics

Internal medicine

Medicine

Observational study

Renal function

Albumin

Hypercalcaemia

Hypocalcaemia

BMJ Open

<h3>Objectives</h3> Albumin-adjusted total calcium is often used as a surrogate marker for free calcium to evaluate hypocalcaemia or hypercalcaemia. Many adjustment formulas based on simple linear regression models have been published, and continue to be used in spite of questionable diagnostic accuracy. In the hope of finding a more pure albumin effect on total calcium, we used multiple linear regression models to adjust for other relevant variables. The regression coefficients of albumin were used to construct local adjustment formulas, and we tested whether the diagnostic accuracy was improved compared with previously published formulas and unadjusted calcium. <h3>Design</h3> A retrospective hospital laboratory data study. <h3>Data sources</h3> The local hospital laboratory data system. <h3>Setting</h3> Norway, 2006–2015. <h3>Participants</h3> 6549 patients above 2 years of age, where free calcium standardised at pH 7.40, total calcium, creatinine, albumin and phosphate had been analysed in a single blood draw, including hospitalised patients and patients from outpatient clinics and general practice. <h3>Main outcome measures</h3> Diagnostic accuracy by Harrell’s c and receiver operating characteristic curve analysis, using free calcium standardised at pH 7.40 as a gold standard, in subgroups with estimated glomerular filtration rate (eGFR) ≥60 or &lt;60 mL/min/1.73 m<sup>2</sup>. <h3>Results</h3> In the subgroup with eGFR &lt;60 mL/min/1.73 m<sup>2</sup>, the Harrell’s c of unadjusted total calcium (0.801) was significantly larger than those of the local formulas (0.790, p=0.002) and the best formula taken from literature (0.791, p=0.004). In the subgroup with eGFR ≥60 mL/min/1.73 m<sup>2</sup>, no significant differences were found between these three formulas. <h3>Conclusions</h3> Our study shows that the diagnostic accuracy of <i>unadjusted</i> total calcium is superior to several commonly used adjustment formulas, and we suggest that the use of such formulas should be abandoned in clinical practice. If the clinician does not trust total calcium to reflect the calcium status of the patient, free calcium should be measured.

Should total calcium be adjusted for albumin? A retrospective observational study of laboratory data from central Norway