This study consisted of two parts. First a model was developed to calculate calcium excretion by CVVH and, second, this model was used to retrospectively study calcium balance in CVVH patients during a 7-years period. The development and validation of a model to calculate calcium excretion by CVVH was performed as a retrospective cohort study in all consecutive patients, 18 years or older, treated with CVVH and citrate anticoagulation in the ICU of the Leiden University Medical Center (LUMC) from January 1st, 2021 until December 31st, 2021. The LUMC has a 24-bed mixed medical and surgical ICU.

The retrospective calculation of calcium losses and calcium balance in ICU patients on citrate anticoagulated CVVH was performed in all consecutive patients, 18 years or older, treated with CVVH and citrate anticoagulation in the ICU of the Leiden University Medical Center (LUMC) from January 1st, 2014 until July 1st, 2021.

As part of our local CVVH protocol, during 2021, total calcium was measured in an ultrafiltrate sample taken at 6:00 a.m. (EST tube, prod. no: 362725, Becton Dickinson, Rutherford, N.J., USA) on a Cobas 8000 Modular c702 system (Roche Diagnostics; Mannheim, Germany). For this study, we used measurements of total calcium and free calcium determined in the blood at the same time. Total calcium was analyzed in serum (3.5 mL SST(TM) II PET tube with clot activator (silica) and separating gel, prod. no: 367957, Becton Dickinson) using the Cobas 8000 Modular c702 system. Ionized calcium was analyzed in heparin-balanced blood (blood gas syringes RAPIDLyte, prod. no: 00925045, Siemens Healthcare, Sudbury, UK) by an ion-selective electrode method using the RAPIDPoint 500 Blood Gas System (Siemens Healthcare, Erlangen, Germany). Before analysis on the Cobas 8000 systems, samples were centrifuged (RCF 3000g, 8 min., 22 °C). CVVH parameters such as substitution flow, blood flow, fluid withdrawal by CVVH, calcium supplementation, and citrate dose were registered every hour in the Patient Data Management System.

Citrate CVVH was performed using Prismaflex® or Prismax® equipment and either the HF1400 hemofilter (polyarylethersulfone (PAES) membrane, 1.4 m2) or ST150 hemofilter (AN 69 ST membrane, 1.5 m2) (Baxter international inc. Deerfield IL, USA). Citrate was administered pre-filter as Prismocitrate 18/0®, containing 18 mmol of citrate per liter as well as sodium (140 meq/L) and chloride (86 meq/L). At initiation of CVVH, citrate was started at a rate of 2.4 mmol/liter blood concentration in the circuit (8 ml prismocitrate 18/0 per ml/h blood flow). The rate of citrate administration could be lowered according to our local protocol if the ratio of free calcium to total calcium decreased, and citrate dose could be increased in case of frequent clotting of the circuit. As substitution fluids we used Biphozyl® (Baxter Holding B.V, Utrecht, the Netherlands) containing no calcium, Phoxilium® (Baxter Holding B.V, Utrecht, the Netherlands) containing 1.25 mmol/l calcium or Prismasol® (Baxter Holding B.V, Utrecht, the Netherlands) containing 1.75 mmol calcium per liter. Additionally all patients also received calcium supplementation given as magnesium (24 mmol/l)-calcium (54 mmol/l) chloride solution intravenously, titrated on a target ionized calcium concentration between 0.9 and 1.1 mmol/l (from January 1st, 2014 until December 31st 2020) or between 1.15 and 1.30 mmol/l (since January 1st, 2021). Free calcium was measured four times daily.

According to our protocols, patients received Nutrison Protein Plus enteral feeding (Nutricia, Zoetermeer, the Netherlands), 1 ml/kg body weight/h, containing 22 mmol calcium per liter) or 1 ml/kg/h parenteral feeding (SmofKabiven, Fresenius Kabi, Huis ter Heide, the Netherlands) containing 2.53 mmol calcium per liter.

Statistical analyses were made in SPSS version 25.0. For the development of the equation to calculate calcium loss by CVVH ultrafiltrate, the population of patients treated with CVVH in 2021 was randomly divided into a development and a validation set, both containing all measurements of a subset of 50% of patients. A mixed linear model was constructed with CVVH calcium excretion as a dependent variable, ultrafiltrate volume times total calcium concentration in plasma (UF*[Ca]total,plasma), blood flow, dose of citrate, weight, ratio of pre-dilution to post-dilution and used filter (HF1400 or ST150) as fixed effect variables, and patient as a random effect variable. Fixed effect variables were kept in the model if they added to the performance with a p value < 0.10.

The performance of the built model was validated in the independent validation set and reported as the mean of the absolute errors (differences between measured and predicted values). Differences between means in subgroups by blood flow, pre-dilution/post-dilution ratio and plasma ionized calcium were analyzed by one-way Anova. In the same validation population, we validated two models that were previously published. One model by Yu and coauthors: Calcium excretion by CVVH (mmol/h) = (0.0006938*total_ultrafiltrate(ml/min) + 0.7983)*calcium(mmol/l)*(60/1000)*(21.42 + 0.35*total_ultrafiltrate(ml/min)) [7], and one model by Zheng and coauthors: Calcium excretion by CVVH (mmol/h) = ((0.0006938*total_ultrafiltrate(ml/min) + 0.7983) * blood flow (L/h)) * (1 − Hematocrit) * ultrafiltrate (L/h))/(blood flow (l/h) + pre-dilution fluid (l/h)) [8].

For the retrospective analysis of calcium balance during CVVH in ICU patients admitted from 2014 until July 2021, CVVH settings were retrieved on an hourly basis from the Patient Data Management System (Metavision, iMD soft, Tel Aviv, Israel). For calculation of the calcium excretion in the CVVH ultrafiltrate, we applied the equation that we developed and validated in the first part of this study. Administered calcium was calculated as the sum of calcium in the substitution fluid and calcium given intravenously. CVVH calcium balance was calculated as administered calcium minus the calcium excretion in ultrafiltrate without taking into consideration the amount of calcium lost in urine and stools and calcium intake via feeding.