Abstract
Background: Hypocalcemia is very common in critically ill patients. While the effect of ionized calcium (iCa) on outcome is not well understood, manipulation of iCa in critically ill patients is a common practice. We analyzed all-cause mortality and several secondary outcomes in patients with acute kidney injury (AKI) by categories of serum iCa among participants in the Acute Renal Failure Trial Network (ATN) Study. Methods: This is a post hoc secondary analysis of the ATN Study which was not preplanned in the original trial. Risk of mortality and renal recovery by categories of iCa were compared using multiple fixed and adjusted time-varying Cox regression models. Multiple linear regression models were used to explore the impact of baseline iCa on days free from ICU and hospital. Results: A total of 685 patients were included in the analysis. Mean age was 60 (SD = 15) years. There were 502 male patients (73.3%). Sixty-day all-cause mortality was 57.0%, 54.8%, and 54.4%, in patients with an iCa <1, 1–1.14, and ≥1.15 mmol/L, respectively (p = 0.87). Mean of days free from ICU or hospital in all patients and the 28-day renal recovery in survivors to Day 28 were not significantly different by categories of iCa. The hazard for death in a fully adjusted time-varying Cox regression survival model was 1.7 (95% CI: 1.3–2.4) comparing iCa <1 to iCa ≥ 1.15 mmol/L. No outcome was different for levels of iCa > 1 mmol/L. Conclusion: Severe hypocalcemia with iCa < 1 mmol/L independently predicted mortality in patients with AKI needing renal replacement therapy.
Introduction
The serum ionized calcium (iCa) is directly associated with smooth muscle and left ventricular contractility.Citation1–3 These observations have been confirmed in chronic stable hemodialysis patientsCitation4,Citation5 with a mechanistic underpinningCitation6 that has been used to justify targeting higher serum iCa levels in critically ill patients with multi-organ failure. Particularly in patients receiving continuous renal replacement therapy, regional citrate anticoagulation allows targeting a desired serum iCa level through changes in the calcium and citrate infusion rates. However, the theoretical concept of achieving a better hemodynamic status in critically ill patients receiving renal replacement therapy by targeting a high-normal range of iCa is primarily an extrapolation of observations in chronic stable dialysis patients. In fact there are no randomized controlled clinical trials that show a survival benefit by targeting a specific level of serum iCa in critically ill patients with acute kidney injury (AKI). On the other hand, there could even be harm associated with excessive utilization of calcium supplementation by targeting high-normal or high levels of serum iCa in AKI including cardiac arrhythmias, metastatic calcification, nephrolithiasis, and renal tubular acidosis.Citation7–12 This study is aimed to investigate the association between different levels of serum iCa concentrations with several clinical outcomes including 60-day all-cause mortality (primary outcome), 28-day renal recovery, days free from the ICU, days free from the hospital and mean arterial pressure (MAP), in critically ill ICU patients with AKI needing renal replacement therapy.
Methods
This study is a post hoc secondary analysis of the Acute Renal Failure Trial Network (ATN) Study. The primary study results have been previously published.Citation13,Citation14 The ATN Study was a multicenter, randomized, clinical trial aimed at comparing the outcomes of different intensities of renal replacement therapy in critically ill patients with AKI due to acute tubular necrosis (ATN), conducted in 27 centers across the United States. Eligible patients were 18 years of age or older, critically ill with AKI consistent with ATN needing renal replacement therapy and accompanied by sepsis or failure of at least one non-renal organ system. After screening, enrolled eligible patients were randomly assigned to two different intensities of renal replacement therapy using a centralized computer-generated adaptive randomization scheme. Assigned interventions were delivered for up to 28 days after randomization or until renal recovery, discharge from the ICU, withdrawal of care, or death. Patients were followed for up to 60 days to ascertain the primary end point of all-cause mortality.
After obtaining local institutional review board approval, we requested and received the study dataset from the NIDDK central repository. All data files were reviewed and all patients with valid measurements of serum iCa at baseline were included. Ionized calcium was categorized to less than 1 mmol/L (severe hypocalcemia), 1 to 1.14 mmol/L (mild hypocalcemia), and 1.15 mmol/L or more according to existing classification.Citation15 The same cut-points were applied to categorize iCa in subsequent days, as well as for the classification of the time-averaged iCa during ICU stay. Pressor support was defined as use of any type of pressor (epinephrine, norepinephrine, phenylephrine, dopamine, dobutamine, vasopressin) at any point throughout the study from baseline to post-randomization days of 1 to 14, 21, and 28. Renal recovery among survivors to day 28 was defined as being off dialysis by Day 28. Days free from the ICU/hospital was defined as the number of days from ICU/hospital discharge until 60 days after randomization or death, whichever occurred first. MAP was calculated from systolic and diastolic blood pressures throughout the study from baseline to days 1 to 14, 21 and 28 after randomization. In this cohort 1124 patients were randomized. Five patients were excluded due to erroneous levels of iCa at baseline and 434 others were excluded due to unmeasured serum iCa at baseline. The final analysis included 685 patients.
Statistical analysis
Mean ± standard deviation or counts and percentages were used to describe the distribution of continuous and categorical variables, respectively. Median values and interquartile range were used when the distribution of variables was skewed. The Chi-square test was used to compare categorical variables across categories of iCa. Analysis of variance was used to compare the mean of continuous variables across the categories of iCa. Bonferroni correction with post hoc analysis was used to identify statistically significant differences among iCa groups, correcting for multiple measurements. We took two different approaches to assess mortality and renal recovery outcomes. First, Cox regression survival models were applied to test the prognostic value of baseline as well as time-varying iCa on 60-day mortality and 28-day renal recovery. As iCa was measured at irregular intervals and with various frequencies, the last measured iCa was used and carried forward to next measurement or to outcome, in the construct of time-varying survival models. The covariates in the fully adjusted survival models included age, gender, race, baseline components of ATN study predictive risk model for 60-day mortality,Citation16 and time varying covariates during study including number of pressor agents per day, system Sequential Organ Failure Assessment (SOFA) scores, MAP, and mechanical ventilation. In an alternative approach, time-averaged iCa during ICU stay for each patient was calculated assuming a linear trend between subsequent measurements, weighted by the number of days between the observations. For example, an iCa of 1.2 mmol/L followed by an iCa of 1.3 in 2 days, yields a value of 1.25 mmol/L weighted by two days. The sums of such weighted values were then divided by the number of days between the first and last measured iCa during ICU stay to result in the time-averaged iCa for a given individual during the study period. Then, different subgroups of the time-averaged iCa during ICU stay were used to compare their mortality odds ratio versus reference category (time-averaged iCa ≥ 1.15 mmol/L), using multiple logistic regression models. To study the effect of the change in iCa on mortality over time by change from below 1 to ≥1 mmol/L, in addition to testing in Cox models, we categorized the patients in to four subgroups: (A) patients whose serum iCa remained 1 mmol/L or above throughout the study; (B) patients with an initial iCa less than 1 mmol/L but with subsequent levels of ≥1mmol/L; (C) patients with an initial iCa ≥ 1 mmol/L but with subsequent level <1 mmol/L; and (D) patients whose serum iCa remained <1 mmol/L throughout the study. Then multiple logistic regression models were used to explore the mortality odds ratio of subgroups versus group A as the reference category.
To test the independent effect of iCa on days-free from the ICU and hospital, multiple linear regression models with multivariable adjustments were applied. A mixed model with iCa as a repeated measure was applied to show its variability throughout the study and to test its independent effect on MAP. The baseline components of ATN Study predictive risk model for 60-day mortality were imputed by application of highly interrelated variables using multiple linear regression models. The survival models were calculated with and without imputation. p Values of less than 0.01 for differences in the baseline variables are declared. Analyses were performed using SAS version 9.2, and SPSS version 20 (Chicago, IL).
Results
Among the 685 patients entered into this study, the mean age was 60 years (SD = 15) and 502 were male (73.3%). The median number of measurements of iCa after randomization was four with an interquartile range of 2–9. As shown in , there was a graded increase in proportion of post-surgical AKI from the lowest to the highest category of baseline iCa (p < 0.001). Nephrotoxic ATN had the lowest proportion in the first category as compared to other categories (p = 0.009). There was also a graded decrease in severity of cardiovascular SOFA score from the lowest to the highest baseline iCa (p = 0.002). Other variables were not significantly different among the categories of baseline iCa. shows the management of renal replacement therapy using different modalities of dialysis by categories of iCa. The delivered KT/V with IHD and the effluent flow with CVVHDF were not significantly different across the categories of iCa. shows the patterns of change in level of iCa during the ICU stay by baseline categories of iCa. Overall, mean of iCa in patients with the baseline iCa < 1 mmol/L increased to more than 1.05 mmol/L after 3 days in survivors to Day 3 and beyond.
Figure 1. Change of mean ionized calcium ± standard error during course of stay in ICU by categories of baseline ionized calcium (mmol/L). Number of measurements per category per day is shown in the table.
Table 1. Comparison of baseline characteristics in patients with acute kidney injury by level of baseline ionized serum calcium, using analysis of variance with Bonferroni post hoc analysis for multiple comparisons.
Table 2. Management of renal replacement therapy by levels of ionized serum calcium.
shows that there were no significant differences in the outcomes of 60-day mortality, days free from ICU or hospital in all patients as well as in 28-day renal recovery among survivors to Day 28 by categories of baseline iCa. , panel A (Supplementary Table 2) shows the Cox regression survival modeled hazard ratio (HR) of 60-day mortality by categories of baseline iCa as well as iCa as a time varying variable, with increasing degrees of adjustment, using iCa ≥ 1.15 mmol/L as the reference group (HR = 1). Accordingly, when categories of iCa at baseline were compared, there was no difference in the risk of mortality across the categories (Model 1). However, when iCa was entered as a time-varying variable either in an unadjusted manner (Model 2) or in partially to fully adjusted models (Models 3 and 4), iCa less than 1 mmol/L as compared to level ≥ 1.15 was an independent predictor of 60-day all-cause mortality. Similar approach did not show any difference in outcome of renal recovery by 28 day (panel B, Models 1–4, and Supplementary Table 2). Similarly, days free from ICU and Hospital were not associated with categories of serum iCa according to multiple linear regression models (Supplementary Table 3). Application of a mixed model revealed no association between MAP and the level of iCa during ICU stay (Supplementary Table 4). shows the adjusted HR of mortality in the lowest versus the highest category of iCa by clinical subgroups. A low iCa was associated with increased mortality risk in all subgroups although statistical significance was not reached for age ≤60, black and other races, baseline total SOFA score <15, Cleveland ICU Acute Renal Failure score <13, age adjusted Charlson score <4, and baseline albumin <2.3 g/dL.
Figure 2. Hazard ratio (95% confidence interval) of mortality and renal recovery outcomes: Model 1: Unadjusted, using categories of baseline ionized calcium, Model 2: Unadjusted, using ionized calcium as a time-varying variable, Model 3: Model 2, plus gender, race, and baseline elements of ATN study predictive risk model for 60-day mortality, Model 4: Model 3, plus “mechanical ventilation, mean arterial pressure, number of pressor agents per day, and SOFA system (coagulation, cardiovascular, liver, central nervous system, respiratory) during ICU stay” all as time-varying covariates.
Figure 3. Hazard ratio of all-cause mortality in ionized calcium <1 mmol/L compared to ionized calcium ≥1.15 mmol/L by subgroups of patient’s characteristics and baseline study variables, according to fully adjusted model (model 4).
Table 3. Comparison of primary and secondary outcomes by base-line ionized calcium using analysis of variance with Tukey post hoc correction for multiple comparisons; values are count (percentage) or mean ± standard error.
In an alternative approach baseline characteristics of the patients were compared by categories of time-averaged iCa during ICU stay (Supplementary Table 1) which showed no significant differences in distribution of baseline variables as compared to . Using multiple logistic regression analysis, a time-averaged iCa of less than 1 mmol/L, and not iCa of 1 to 1.14 mmol/L throughout stay in ICU as compared to iCa of 1.15 or above was associated with higher odds of 60-day mortality in different models from unadjusted to fully adjusted (Supplementary Table 5). Similar table shows that the 28-day renal recovery has not been different by levels of the time-averaged iCa during ICU stay. Supplementary Figure 1 shows the change in the level and direction of iCa during the ICU stay by subgroups of patients defined by iCa trend. Supplementary Table 6 shows that patients who continued to have an iCa below 1 mmol/L during ICU stay had a significantly higher odds of 60-day mortality as compared to those who had iCa above 1 mmol/L, in several unadjusted to fully adjusted models. Similarly those with a low iCa (<1 mmol/L) who achieved iCa > 1 mmol/L in subsequent days had a mortality similar to patient with iCa > 1 mmol/L at all times.
Discussion
In this study, the 60-day all-cause mortality, 28-day renal recovery, days free from ICU and hospital were not significantly different by the levels of baseline iCa. MAP during ICU stay was not dependent on absolute values of iCa, and multivariable models did not show any relationship between iCa with renal recovery, days free from hospital or ICU. However, survival models using iCa as a time-varying variable suggest that serum iCa < 1 mmol/L compared to iCa ≥ 1.15 mmol/L independently predicts short-term mortality. This is consistent with results from logistic regression models which have used categories of a time-averaged iCa during ICU stay.
Chernow et al.Citation17 reported a study of 259 ICU patients, finding that those with hypocalcemia had a higher mortality, a longer duration of stay in the ICU, and higher likelihoods of sepsis, renal failure and blood transfusion as compared to the normocalcemic group. Desai et al.Citation18 studied 108 critically ill patients admitted to the ICU and found mortality rates of 44% in patients with hypocalcemia compared to 17% in patients without hypocalcemia. In a study of 199 patients, Zivin et al.Citation19 demonstrated that iCa was inversely correlated with the APACHE score and, therefore, suggested that higher mortality observed with hypocalcemia might be a reflection of severity of disease. In a larger study from the Helsinki University Hospital, Hastbacka et al reported the hypocalcemia rate of up to 85% in 993 critically ill patients defined as iCa less than 1.16 mmol/L.Citation20 Although the investigators showed a higher risk of mortality among patients with hypocalcemia, hypocalcemia did not appear to be an independent risk factor for mortality after multivariable adjustment for severity of disease at baseline as measured by the APACHE score. Their analysis, however, was limited to use of a single baseline measurement of iCa per patient and therefore did not take into account the impact of changes in serum iCa over time or the duration of time at each level of iCa. In a cohort of 7024 critically ill patients from four medical centers, Egi et al.Citation15 used a time weighted approach to account for the level of iCa during critical care and found a graded inverse relationship between mortality and serum iCa even after multivariable adjustments. This approach has allowed the investigators to capture worse outcome attributable to changes of iCa over time. In our study, other than use of time-varying Cox survival models, we also calculated a time-averaged iCa during ICU stay and compared risk of mortality across its categories using logistic regression models, and found similar higher mortality risk with iCa < 1 mmol (Supplementary Table 5). Similar results from the two approaches suggest that the latter has been able to capture the higher mortality attributed to change of iCa over time, a phenomenon which was captured similarly with Cox models. Our findings are in agreement with previous observationsCitation15,Citation17–22 and can be considered as reconciling the seemingly contradictory results of the studies by Hastbacka and Egi. Although the mortality was not different by baseline level of iCa after adjusting for severity of disease, the time-varying analysis accounting for change of iCa throughout ICU stay unmasked its prognostic value to predict mortality. The reason that a time-varying Cox model and not the fixed model has unmasked the prognostic value of iCa is that a significant number of patients with a baseline iCa of above 1 mmol/L who subsequently died, dropped their iCa to a level of <1 mmol/L in later days prior to their death. On the other hand, significant number of other patients with iCa < 1 mmol/L at baseline who eventually survived, improved their iCa to a level of ≥1 mmol/L in subsequent days, a phenomenon which could be unmasked by application of a time-varying approach taking into account change of iCa in subsequent days. Consistent with these findings, in an alternative approach, patients who continuously had iCa < 1 mmol/L throughout ICU stay had a significantly higher mortality odds ratio (OR) as compared to patients who kept their iCa ≥ 1 mmol/L at all time, while those severely hypocalcemic patients at baseline with increased iCa to ≥ 1mmo/L in subsequent days did not have different mortality OR as compared to those with iCa ≥ 1 mmol/L at all times (Supplementary Table 6).
Hypocalcemia is a common electrolyte abnormality in critically ill patients and in ICU settings.Citation17,Citation19,Citation20 Although the etiology of hypocalcemia is unclear in the majority of critically ill patients, proposed mechanisms include the effect of sepsis, impairment of parathyroid hormone secretion, resistance of end-organs to the function of parathyroid hormone, alterations in calcium-sensing receptor gene transcription via proinflammatory cytokines, and intracellular accumulation of calcium.Citation21–26 Other less frequent causes of hypocalcemia in this patient population include acute pancreatitis, massive transfusions and vitamin D deficiency. Calcium has numerous crucial roles in the cellular function of almost every system. It is part of signal transduction pathways, contributes to cell membrane potential and nerve conduction, triggers muscle contraction, is a cofactor of many enzymes and acts as the second messenger of many hormones that regulate metabolism and gene expression.Citation3,Citation27
Although, it is not yet clear, whether or not the association between hypocalcemia with higher mortality reflects a cause and effect relationship, the increase from a low iCa along with improvement in severity of disease, and weakening of mortality hazard ratio by further adjustment for surrogates of severity (, Panel A, model 4 compared to model 3) suggest that hypocalcemia itself may be a surrogate for disease severity rather than a therapeutic target. Carlon et al.Citation28 showed no significant improvement in patient hemodynamic variables with calcium supplementation in sepsis and trauma settings. Several model systems have also shown worse outcomes with calcium supplementation, and alleviation of cellular injury with lowering of the extracellular calcium concentration.Citation29–33 A systematic review of clinical trials reveals that no randomized controlled clinical trial has ever evaluated the effect of calcium supplementation on mortality, multiple organ dysfunction, length of stay in the ICU or hospital.Citation34 In spite of that, given the crucial role of calcium in metabolism and cellular function, recommendations have been made in favor of supplementation to normalize the serum calcium in critically ill patients.Citation19,Citation21 However, there is absolutely no evidence to suggest that within physiologic range any level of serum iCa is associated with a better outcome. Given the lack of clinical trial evidence for a beneficial effect of calcium supplementation on outcomes within physiologic and normal limits, the intensity of calcium supplementation to target a pre-defined high-normal or high level of iCa remains speculative.
Subgroup analysis revealed a consistent pattern of higher mortality risk in the lowest category of iCa as compared to the highest category for most subgroups. The subgroup exceptions were likely attributable to insufficient statistical power (for black race and other races) as well as to the potential impact of additional comorbidities acting as competing risk factors for death even at higher levels of iCa in younger patients and in those with less severe illness such as the subgroups with total SOFA < 15, CCFARF < 13, and Charlson < 4.
This study has several strengths. This is the second largest multicenter study of the relationship between iCa and short term mortality and the analyses of main outcomes had sufficient power. It also has highly accurate ascertainment of outcomes due to the clinical trial origins of the study.Citation13 Collection of severity of illness by multiple scoring systems as well as comorbid conditions at baseline provided a unique opportunity to test the effect of iCa after extensive adjustment for the most important prognostic variables. This study also has several important limitations. Although this is a secondary analysis of a well-designed clinical trial, it was not primarily designed to assess the effect of different levels of iCa during the ICU stay on the outcomes. As a consequence, serum iCa was not measured uniformly or consistently for everybody. Comparison of patients with and without iCa measurement shows that the former group was sicker overall and had higher mortality compared to the latter group (55.7% vs. 47.7%; Supplementary Table 7). Thus, differences in patterns of measurement between patients may have introduced ascertainment bias, characterized by a higher likelihood of measuring iCa in patients with more severe illness with a magnitude corresponding to 7% higher mortality in the measurement group, and a direction toward the sicker patients. However, as the median number of iCa measurements was identical in all categories of baseline iCa (n = 4, p = 0.767), the patients’ characteristics and outcomes were not significantly different by these categories, and as the differences in risk of mortality manifested with follow up, we infer that the association of iCa with mortality has not been impacted by ascertainment bias, and although the iCa measurements were not protocol based, evidence for their randomness argues in favor of unbiased estimations and therefore the derived conclusions. There were also up to 15% missing in a few of the baseline covariates posing a significant attrition in sample size in multivariable models without imputation. However, as the results of survival models with and without imputation were not significantly different, we chose to apply the dataset after imputation. Another limitation is presumption of constancy of last measurement for subsequent unmeasured iCa with the “last observation carried-forward” method in Cox survival models which may not be true in every individual case and may have overestimated the effect of hypocalcemia in some patients who have had a rapid correction. In another approach, application of time-weighted averaged values between subsequent measures was an alternative approach which has eliminated the effect of constancy of the pre-missing-value measurements. As the results from the two approaches were similar, we infer that on average the method of handling unmeasured iCa between subsequent measurements have not impacted its relationship with mortality. This study is also unable to assess outcomes in the hypercalcemic range because of the low prevalence of hypercalcemia. This study is neither designed nor aimed to identify the impact of calcium supplementation on outcome. Therefore, it cannot be inferred from the data how much calcium supplementation was used for each patient. We believe that the safety and efficacy of calcium supplementation on hard outcomes are best determined by randomized controlled clinical trials, but even in such trials serum iCa can be and should be used as a reasonable monitoring index to separate arms of the trials targeting different ranges of serum iCa using different intensities of calcium supplementation. In this context this study which aimed testing the effect of different ranges of serum iCa on the defined outcomes has successfully reached to its aims, irrespective of how the level was achieved.
Conclusion
Cox survival models as well as analysis of time-averaged iCa with use of logistic regression models suggested that hypocalcemia with iCa < 1.0 mmol/L was independent predictor of higher mortality in critically ill patients with acute renal failure in need of renal replacement therapy in the studied cohort. There is no evidence to suggest that above this level, different values of iCa are necessarily associated with better outcome including better survival, shorter ICU or hospital stay, better preservation of blood pressure independent of pressor support, or a higher rate of renal recovery. Randomized controlled clinical trials are needed to confirm the independent beneficial effect of calcium supplementation as well as to determine the optimal serum iCa target in critically ill patients.
Key messages
Outcomes of 60-day all-cause mortality, renal recovery, days free from ICU, days free from hospital and MAP during ICU stay in mild hypocalcemia (iCa of 1.0 to 1.14 mmol/L) were not different from normocalcemia (iCa ≥ 1.15 mmol/L) in critically ill patients with AKI needing renal replacement therapy.
Outcomes of renal recovery, days free from ICU, days free from hospital, and MAP during ICU stay in severe hypocalcemia (iCa < 1.0 mmol/L) were not different from normocalcemia (iCa ≥ 1.15 mmol/L) in critically ill patients with AKI needing renal replacement therapy.
Severe hypocalcemia (iCa < 1.0 mmol/L) appeared an independent predictor of 60-day all-cause mortality as compared to normocalcemia in critically ill patients with AKI needing renal replacement therapy.
Hypocalcemia may be a surrogate marker for severity of illness and not necessarily a therapeutic target.
Randomized controlled clinical trials are needed to determine the optimum level of serum iCa and optimal supplementation in critically ill patients.
Declaration of interest
ATN study was supported by the Cooperative Studies Program of the Department of Veterans Affairs Office of Research and Development (VA) and the NIDDK. F.A. is supported by grant 5T32DK7378-32.
Supplementary Material
Download PDF (238.1 KB)Acknowledgements
The ATN study was conducted by the ATN Investigators. The data from the ATN study were supplied by the NIDDK Central Repositories. This manuscript was not prepared in collaboration with Investigators of the ATN study and does not necessarily reflect the opinions or views of the ATN study, the VA, the NIDDK Central Repositories, or the NIDDK. The authors also would like to thank Mallika Kommareddi for her remarks.
References
- Hill-Eubanks DC, Werner ME, Heppner TJ, Nelson MT. Calcium signaling in smooth muscle. Cold Spring Harb Perspect Biol. 2011;3:a004549
- Lang RM, Fellner SK, Neumann A, Bushinsky DA, Borow KM. Left ventricular contractility varies directly with blood ionized calcium. Ann Intern Med. 1988;108:524–529
- Smith JB. Calcium homeostasis in smooth muscle cells. New Horiz. 1996;4:2–18
- Maynard JC, Cruz C, Kleerekoper M, Levin NW. Blood pressure response to changes in serum ionized calcium during hemodialysis. Ann Int Med. 1986;104:358–361
- Henrich WL, Hunt JM, Nixon JV. Increased ionized calcium and left ventricular contractility during hemodialysis. N Engl J Med. 1984;310:19–23
- Kramer W, Wizemann V, Thormann J, Bechthold A, Schutterle G, Lasch HG. Mechanisms of altered myocardial contractility during hemodialysis: importance of changes in the ionized calcium to plasma potassium ratio. Klin Wochenschr. 1985;63:272–278
- Chaudhuri TK, Chaudhuri Tu H. Calcium phosphate colloid: possible nidus for metastatic calcification in hypercalcemia. JAMA. 1972;222:1061
- Hartman TE, Muller NL, Primack SL, et al. Metastatic pulmonary calcification in patients with hypercalcemia: findings on chest radiographs and CT scans. AJR Am J Roentgenol. 1994;162:799–802
- Kerr DN. Hypercalcemia and metastatic calcification. Cardiovasc Res. 1997;36:293–297
- Yang H, Curinga G, Giachelli CM. Elevated extracellular calcium levels induce smooth muscle cell matrix mineralization in vitro. Kidney Int. 2004;66:2293–2299
- Caruana RJ, Buckalew VM Jr. The syndrome of distal (type 1) renal tubular acidosis. Clinical and laboratory findings in 58 cases. Medicine (Baltimore). 1988;67:84–99
- Kiewiet RM, Ponssen HH, Janssens EN, Fels PW. Ventricular fibrillation in hypercalcaemic crisis due to primary hyperparathyroidism. Neth J Med. 2004;62:94–96
- Palevsky PM, Zhang JH, O'Connor TZ, et al. Intensity of renal support in critically ill patients with acute kidney injury. N Engl J Med. 2008;359:7–20
- Palevsky PM, O'Connor T, Zhang JH, Star RA, Smith MW. Design of the VA/NIH Acute Renal Failure Trial Network (ATN) Study: intensive versus conventional renal support in acute renal failure. Clin Trials. 2005;2:423–435
- Egi M, Kim I, Nichol A, et al. Ionized calcium concentration and outcome in critical illness. Crit Care Med. 2011;39:314–321
- Demirjian S, Chertow GM, Zhang JH, et al. Model to predict mortality in critically ill adults with acute kidney injury. Clin J Am Soc Nephrol. 2011;6:2114–2120
- Chernow B, Zaloga G, McFadden E, et al. Hypocalcemia in critically ill patients. Crit Care Med. 1982;10:848–851
- Desai TK, Carlson RW, Geheb MA. Prevalence and clinical implications of hypocalcemia in acutely ill patients in a medical intensive care setting. Am J Med. 1988;84:209–214
- Zivin JR, Gooley T, Zager RA, Ryan MJ. Hypocalcemia: a pervasive metabolic abnormality in the critically ill. Am J Kidney Dis. 2001;37:689–698
- Hastbacka J, Pettila V. Prevalence and predictive value of ionized hypocalcemia among critically ill patients. Acta Anaesthesiol Scand. 2003;47:1264–1269
- Burchard KW, Gann DS, Colliton J, Forster J. Ionized calcium, parathormone, and mortality in critically ill surgical patients. Ann Surg. 1990;212:543–549
- Carlstedt F, Lind L, Rastad J, Stjernstrom H, Wide L, Ljunghall S. Parathyroid hormone and ionized calcium levels are related to the severity of illness and survival in critically ill patients. Eur J Clin Invest. 1998;28:898–903
- Aderka D, Schwartz D, Dan M, Levo Y. Bacteremic hypocalcemia. A comparison between the calcium levels of bacteremic and nonbacteremic patients with infection. Arch Int Med. 1987;147:232–236
- Taylor B, Sibbald WJ, Edmonds MW, Holliday RL, Williams C. Ionized hypocalcemia in critically ill patients with sepsis. Can J Surg. 1978;21:429–433
- Zaloga GP, Chernow B. The multifactorial basis for hypocalcemia during sepsis. Studies of the parathyroid hormone-vitamin D axis. Ann Int Med. 1987;107:36–41
- Song SK, Karl IE, Ackerman JJ, Hotchkiss RS. Increased intracellular Ca2+: a critical link in the pathophysiology of sepsis? Proc Natl Acad Sci USA. 1993;90:3933–3937
- Kelly A, Levine MA. Hypocalcemia in the critically ill patient. J Intensive Care Med. 2013;28:166--177
- Carlon GC, Howland WS, Kahn RC, Schweizer O. Calcium chloride administration in normocalcemic critically ill patients. Crit Care Med. 1980;8:209–212
- Liu W, Matsumori A. Calcium channel blockers and modulation of innate immunity. Curr Opin Infect Dis. 2011;24:254–258
- Carlstedt F, Eriksson M, Kiiski R, Larsson A, Lind L. Hypocalcemia during porcine endotoxemic shock: effects of calcium administration. Crit Care Med. 2000;28:2909–2914
- Matsuda S. Protective effects of calcium antagonist (nitrendipine) on calcium ionophore A23187-induced liver cell injury. Bull Tokyo Med Dent Univ. 1991;38:35–44
- Steinhorn DM, Sweeney MF, Layman LK. Pharmacodynamic response to ionized calcium during acute sepsis. Crit Care Med. 1990;18:851–857
- Li GY, Fan B, Zheng YC. Calcium overload is a critical step in programmed necrosis of ARPE-19 cells induced by high-concentration HO. Biomed Environ Sci. 2010;23:371–7
- Forsythe RM, Wessel CB, Billiar TR, Angus DC, Rosengart MR. Parenteral calcium for intensive care unit patients. Cochrane Database Syst Rev. 2008;(4):CD006163
Supplementary material available online Supplementary Tables 1–7 Supplementary Figure 1