Limited diagnostic utility of albumin-corrected calcium in the intensive care unit: A prospective comparison with ionised calcium

Abstract

Background Albumin-adjusted calcium is widely used despite uncertainty about its accuracy in critically ill adults. We prospectively evaluated the diagnostic performance of total calcium and of multiple published albumin-correction formulae against ionised calcium, the physiological reference standard, in an intensive care unit (ICU) cohort. Methods In this prospective observational cohort study, ionised calcium (iCa), total calcium, albumin, venous pH, creatinine, estimated glomerular filtration rate (eGFR), C-reactive protein (CRP), procalcitonin, magnesium, phosphate and vitamin D were measured simultaneously. The primary outcome was the diagnostic accuracy (area under the ROC curve [AUC]; sensitivity/specificity) of total calcium and of corrected calcium for detecting hypocalcaemia, defined a priori by the reference standard as iCa < 1.12 mmol/L; the pre-specified contrast was the difference between the two correlated AUCs. Twelve published measures were compared (total calcium; Payne; simplified Payne; Orrell; Berry; James; Jain; Thode; and the Pekar and Antonio estimators). Agreement, formal subgroup interaction tests, an explanatory multivariable model with collinearity diagnostics, and an exploratory internally-validated ICU equation were assessed. Results In total, 250 patients were enrolled (median age 70 years; 147 men, 58.8%). Hypocalcaemia prevalence was 58.0% (n = 145). Payne-corrected calcium correctly identified only 15.2% of truly hypocalcaemic patients (Cohen κ = 0.08), versus 62.8% for total calcium (κ = 0.40). No simple albumin-correction formula significantly outperformed unadjusted total calcium (AUC 0.69–0.75; all DeLong p > 0.05 vs total calcium AUC 0.74). Only the four-variable Pekar estimator achieved a higher AUC (0.78; p = 0.005). In the explanatory model, only total calcium, pH, and sex independently predicted iCa (all variance inflation factors < 1.8); albumin, phosphate and vitamin D were not significant. An exploratory ICU-specific equation reached an optimism-corrected R² of only 0.17. Conclusions Albumin-corrected calcium, including all established formulae tested, provides no diagnostic benefit over total calcium in the ICU and frequently masks true hypocalcaemia. Direct ionised-calcium measurement should be prioritised when accurate calcium status affects management. These are diagnostic-accuracy findings; their clinical-outcome implications require outcome-anchored study.