https://orcid.org/0000-0002-1442-4694
Abstract
Low circulating total calcium or albumin-adjusted calcium has been associated with higher mortality, especially in hospital settings; however, these measures tend to misclassify patients with derangements in calcium homeostasis.
As the association of the biologically active ionized calcium with mortality is poorly elucidated, we tested the hypothesis that low plasma ionized calcium is associated with higher risk of all-cause and cause-specific mortality in the general population.
We included 106 768 individuals from the Copenhagen General Population Study. Information on all-cause and cause-specific mortality was from registries and risks were calculated using Cox regression and competing-risks regression by the STATA command stcompet.
During a median follow-up period of 9.2 years, 11 269 individuals died. Each 0.1 mmol/L lower plasma ionized calcium below the median of 1.21 mmol/L was associated with a multivariable adjusted hazard ratio of 1.23 (95% CI, 1.10-1.38) for all-cause mortality. Corresponding hazard ratios for cancer and other mortality were 1.29 (1.06-1.57) and 1.24 (1.01-1.53), respectively. In contrast, for cardiovascular mortality, only high plasma ionized calcium was associated with mortality with a hazard ratio of 1.17 (1.02-1.35) per 0.1 mmol/L higher plasma ionized calcium above the median. We found no interactions between plasma ionized calcium and preexisting cardiovascular or renal disease on all-cause mortality.
In the general population, low plasma ionized calcium was associated with increased all-cause, cancer, and other mortality, while high levels were associated with increased cardiovascular mortality.
Calcium is the most abundant mineral in the human body and an important structural component of bone; however, calcium also plays essential roles in coagulation, intracellular signaling, muscle contraction, and in nervous system function. Hypocalcemia is common in the emergency and intensive care unit setting and has been found to correlate with the severity of illness and mortality in these patient groups (1, 2). However, the association between plasma calcium and mortality in the general population is not clear, as some studies have found high calcium to be associated with higher mortality, whereas other studies have found a U-shaped association between plasma calcium and mortality using either total or albumin-adjusted calcium (3, 4).
Plasma calcium can be measured as the biologically active ionized (free) calcium or as total calcium (free plus protein-bound calcium) and is often reported as albumin-adjusted calcium (total calcium adjusted for albumin concentration) (5). Although ionized calcium is the most physiologically relevant measure of calcium homeostasis, in clinical practice, total calcium is most often used to assess calcium status, as total calcium assays are built into most automated clinical laboratory instruments and consequently are more easily available. However, total calcium and albumin-adjusted calcium may tend to misclassify patients with derangements in calcium homeostasis when using ionized calcium as the gold standard (6, 7). These misclassifications of calcium status could lead to spurious associations between calcium and, for example, mortality; thus, the use of ionized calcium instead of other calcium measures, may be more informative in defining the specific role of calcium in mortality. In Denmark, although the ionized calcium measurement is more laborious, it has traditionally been offered at all hospital laboratories in Denmark and used preferentially to total calcium to assess calcium status. Consequently, and in contrast to routine practice in many other countries, albumin-adjusted calcium measures are not commonly used. In this large cohort from the Copenhagen general population, we uniquely chose to measure ionized calcium over total calcium, as ionized calcium is considered the gold standard of calcium measurements and is widely used in Denmark.
We tested the hypothesis that plasma ionized calcium is associated with risk of all-cause, cardiovascular, cancer, and other mortality in the general population. To do so, we examined 106 768 individuals from the Copenhagen General Population Study with plasma ionized calcium available, of whom 11 269 died during follow-up.
Methods
Study Population
We used the Copenhagen General Population Study, initiated in 2003 with ongoing enrollment (8, 9). Briefly, individuals aged 20 to 100 years were invited randomly from the Danish Civil Registration System to complete a questionnaire that was reviewed together with an examiner at the day of attendance, undergo a physical examination and give blood for plasma measurements. The participation rate was 43%. We included 106 768 individuals of Danish descent who had plasma ionized calcium measurements available. The study was approved by Herlev and Gentofte Hospital and a Danish Ethical Committee, conducted according to the Declaration of Helsinki, and written informed consent was obtained from all participants.
Ionized Calcium Measurements
Ionized calcium was measured in serum using a Konelab autoanalyzer (Konelab Prime 60i, Thermo Scientific, Helsinki, Finland) and an ion selective membrane electrodes (ISE) module as previously described (10). The total measuring range was 0.5 to 4.0 mmol/L and the day-to-day coefficient of variation was 2%. A pH-corrected ionized calcium value was automatically reported; the adjusted value was calculated with the equation: Ca2+ at pH 7.4 = Ca2+sample × 10-0.24(7.4 − pHsample) which is valid between pH 7.2 and 7.6 (11). This correction is reasonable when the patient’s blood pH can be assumed to be close to 7.4 as in our sample with individuals from the general population, thus accounting for pre-analytic factors causing smaller pH alterations such as time uncapped during automated analyses. All samples were collected Monday through Thursday and measured consecutively on fresh serum within a few hours of sampling, depending on the laboratory capacity; all laboratory personnel were blinded to the case status of the participants.
Endpoints
We obtained the date of death or emigration (n = 453) from the Danish Civil Registration System. The national Danish Causes of Death Registry records were used to identify causes of death. We used underlying causes of death to classify deaths as due to cardiovascular disease (ICD-10 I00-I99) and cancer (ICD-10 C00-C97), as well as other causes (individuals with a classified cause of death which was not I- or C-diagnoses). As The national Danish Causes of Death Registry lags the Danish Civil Registration System by approximately 1 year, the most recent deaths could not be classified by cause although death per se was registered (n = 2910), and these cases were included in all-cause mortality but not in cause-specific mortality.
Covariates
All measured and self-reported covariates were ascertained on the day of attendance. The following lifestyle factors were self-reported: information on smoking, alcohol intake, physical activity, income, use of dietary supplements, antihypertensive treatment, and use of diuretics. Body mass index (BMI) was calculated using measured weight divided by measured height squared. Systolic and diastolic blood pressure were measured. Plasma total cholesterol, high-density lipoprotein (HDL) cholesterol, albumin, and creatinine were measured using Konelab autoanalyzer (Konelab Prime 60i, Thermo Scientific, Helsinki, Finland) and standard hospital assays and the estimated glomerular filtration rate (eGFR) was calculated using the 2009 Chronic Kidney Disease Epidemiology Collaboration (CKD-EPI) creatinine equation (12). For creatinine, an isotope dilution mass spectrometry calibrated Jaffe method was used. Non-HDL cholesterol was calculated as total cholesterol minus HDL cholesterol. Plasma 25-hydroxyvitamin D was measured using the DiaSorin Liaison 25-hydroxyvitamin D TOTAL assay (Stillwater, MN, USA) on a subgroup of individuals recruited in 2004-2005 (in which case samples were stored at −80 °C until time of measurement) or collected in 2009-2011 in which case measurements were done on fresh samples (data available on ̴ 23 000 individuals) (13).
Information on diagnosis of ischemic heart disease (ICD-8: 410 to 414, and ICD-10: I20 to I25) from 1977 until December 2018 was obtained from the national Danish Patient Registry, as was information on inflammatory bowel disease (Crohn’s disease ICD-8: 563.01, ICD-10: DK50; ulcerative colitis: ICD-8: 563.91, 563.99, ICD-10 DK51), and osteoporotic fractures (ICD-8: 723.2, 805.2, 805.6, 807.0, 808.00, 808.01, 812.0, 812.1, 812.2, 812.3, 813.4, 813.5, 820, 821.2, 821.3, 823, 824; ICD-10: DM48.4, DM48.5, DM84.3, DS22.0, DS22.1, DS22.3, DS32.0, DS32.1, DS32.5, DS32.7, DS32.8, DS42.2, DS42.3, DS52.2, DS52.5, DS52.6, DS72.0, DS72.1, DS72.2, DS72.4, DS72.8, DS72.9, DS82.1, DS82.2, DS82.3, DS82.4, DS82.5, DS82.6). Information on diagnosis of invasive cancers from 1943 until December 2016 was obtained from the Danish Cancer Registry, which identifies all cancer cases in Denmark (14). The cancer variable included all cancers except for nonmelanoma skin cancer. Gastrointestinal cancer used in sensitivity analyses included esophagus, stomach, colon, and pancreatic cancer. Baseline diabetes was defined as self-reported diabetes of any type, a hospital diagnosis of diabetes before the examination (ICD-8: 249 –250; ICD-10: E10, E11, E13, E14), nonfasting plasma glucose > 11 mmol/L at examination, or use of antidiabetic medication. Information on thyroid and parathyroid operations prior to examination date was obtained from the Danish Patient Registry.
Statistical Analyses
We used STATA v.16.1. Plasma ionized calcium was included both as a categorical variable and as a continuous variable. As to the former, we divided plasma ionized calcium into 3 groups corresponding to the 2.5th and 97.5th percentiles of the dataset adjusted for month and year of measurement to account for any drift in the analysis over the years. Although arbitrary, it is a common practice to define the reference interval as the central 95%-interval, and we hereby created a study-specific reference interval in which all individuals were recruited similarly, and venipuncture samples were taken under similar circumstances and time of day (15). The data were 99% complete and missing values for covariates were imputed according to age and sex for multivariable adjustment to obtain a complete data set; however, if individuals with any missing data were excluded, results were similar to those presented.
First, to examine the association between plasma ionized calcium and all-cause and cause-specific mortality, we estimated cumulative incidences with follow-up as underlying time scale using Kaplan-Meier estimator for all-cause mortality and, for cause-specific mortality, competing-risks survival regression using the STATA command stcompet, which accounts for the competing risk of death by other causes or emigration (16, 17).
Second, we performed multivariable adjusted, restricted cubic spline Cox regression for graphical representation using plasma ionized calcium and 3 knots, as well as univariate kernel density estimation for density graphs. The knots were placed at the 10th, 50th, and 90th percentiles, based on Harrell’s recommended percentiles (18). We used multivariable adjusted Cox proportional hazards regression models with entry at examination and age as the time scale to estimate hazard ratios with 95% CIs. Those emigrating or dying during follow-up were censored at their emigration or death dates, respectively. Multivariable adjustment was for covariates chosen a priori that potentially could confound the association between plasma ionized calcium and the endpoints: month and year of measurement, sex, BMI, physical activity during leisure time, hypertension at baseline, use of dietary supplements, smoking status, income, alcohol intake, estimated glomerular filtration rate (eGFR), plasma albumin, plasma non-HDL cholesterol, cardiovascular disease at baseline (not cardiovascular mortality), cancer at baseline (not cancer mortality). Test for proportionality of hazards over time was performed using graphical methods and residuals; no violations were observed.
Finally, we stratified data on the median plasma ionized calcium value and obtained multivariable adjusted estimates from Cox regression below and above the median per 0.1 mmol/L lower and higher plasma ionized calcium, respectively. For all-cause mortality we performed sensitivity analyses stratified on possible confounders using similar models and tested for interactions using a likelihood ratio test by introducing a 2-factor interaction term in a model also including both factors. The included confounders for stratification were sex, age, BMI, plasma non-HDL cholesterol, plasma albumin, eGFR, cancer at baseline, cardiovascular disease at baseline, alcohol, smoking, income, hypertension, use of diuretics, use of dietary supplements, and diabetes. Furthermore, in sensitivity analyses, we additionally adjusted models for gastrointestinal cancer, inflammatory bowel disease, and osteoporotic fractures at baseline, as these covariates potentially could confound or mediate the association between hypocalcemia and mortality through interaction with vitamin D uptake or parathyroid hormone (PTH), respectively.
Results
A total of 106 768 individuals with plasma ionized calcium measurements available were included in the analyses with a median follow-up of 9.2 years (range, 0.003-15 years; interquartile range, 6.6-11.9) and 11 269 individuals died during follow-up. The median and interquartile range for plasma ionized calcium was 1.11 mmol/L (1.10-1.12 mmol/L) for hypocalcemia (below 2.5th percentile), 1.21 mmol/L (1.18-1.24 mmol/L) for individuals within the reference interval (2.5th to 97.5th percentiles) and 1.36 mmol/L (1.34-1.40 mmol/L) for hypercalcemia (above 97.5th percentile) (Table 1). A total of 2562 out of 109 331 individuals had no plasma ionized calcium measurement available and were not included in this study; missing calcium measurements were distributed over the years with the highest numbers in the years 2011 and 2012 (566 and 523 missing values, respectively, out of approximately 10 000 participants per year) (Supplemental Figure 1 (19)). At most, only a small change in plasma ionized calcium was seen from year to year of recruitment from 2003 to 2014 (Supplemental Figure 2 (19)); part of the difference from year to year could be due to different parts of the greater Copenhagen area being examined in the different years, for example, as the age-distribution may differ slightly in different parts of the greater Copenhagen area.
The association of baseline characteristics with plasma ionized calcium
| Plasma ionized calcium | |||
|---|---|---|---|
| Hypocalcemia < 2.5th percentile | Reference range 2.5th-97.5th percentiles | Hypercalcemia > 97.5th percentile | |
| Plasma ionized calcium, mmol/L | 1.11 (1.10-1.12)) | 1.21 (1.18-1.24) | 1.36 (1.34-1.40) |
| Men, n (%) | 1385 (54%) | 45 841 (45%) | 786 (29%) |
| Age, years | 56 (46-68) | 58 (48-67) | 64 (55-71) |
| Alcohol consumption, g/week | 84 (36-168) | 96 (48-180) | 108 (48-180) |
| Ever smokers, n (%) | 1397 (55%) | 59 101 (58%) | 1645 (62%) |
| High income, n (%) | 1167 (46%) | 43 743 (43%) | 875 (33%) |
| Body mass index, kg/m2 | 25.7 (23.3-28.8) | 25.6 (23.2-28.4) | 25.7 (23.2-28.7) |
| Hypertension, n (%) | 1540 (60%) | 60 846 (60%) | 1822 (68%) |
| Use of diuretics, n (%) | 277 (11%) | 6911 (7%) | 266 (10%) |
| Use of vitamin supplements, n (%) | 1128 (44%) | 49 020 (48%) | 1489 (56%) |
| Plasma non-HDL cholesterol, mmol/L | 3.8 (3.0-4.5) | 3.9 (3.2-4.7) | 4.0 (3.3-4.8) |
| Plasma albumin, µmol/L | 583 (550-617) | 589 (556-625) | 582 (549-616) |
| Renal impairment at baselinea, n (%) | 310 (12%) | 9663 (10%) | 402 (15%) |
| Diabetes at baseline, n (%) | 142 (6%) | 4243 (4%) | 140 (5%) |
| Cancer diagnosis prior to baseline, n (%) | 199 (7.8%) | 6833 (6.7%) | 282 (11%) |
| Cardiovascular disease prior to baseline, n (%) | 175 (6.8%) | 5797 (5.7%) | 172 (6.5%) |
| Vitamin D insufficiencyb, n (%) | 204 (41%) | 10 932 (45%) | 234 (43%) |
| Any thyroid/parathyroid operation prior to baseline, n (%) | 46 (1.8%) | 1226 (1.2%) | 223 (8.4%) |
| Plasma ionized calcium | |||
|---|---|---|---|
| Hypocalcemia < 2.5th percentile | Reference range 2.5th-97.5th percentiles | Hypercalcemia > 97.5th percentile | |
| Plasma ionized calcium, mmol/L | 1.11 (1.10-1.12)) | 1.21 (1.18-1.24) | 1.36 (1.34-1.40) |
| Men, n (%) | 1385 (54%) | 45 841 (45%) | 786 (29%) |
| Age, years | 56 (46-68) | 58 (48-67) | 64 (55-71) |
| Alcohol consumption, g/week | 84 (36-168) | 96 (48-180) | 108 (48-180) |
| Ever smokers, n (%) | 1397 (55%) | 59 101 (58%) | 1645 (62%) |
| High income, n (%) | 1167 (46%) | 43 743 (43%) | 875 (33%) |
| Body mass index, kg/m2 | 25.7 (23.3-28.8) | 25.6 (23.2-28.4) | 25.7 (23.2-28.7) |
| Hypertension, n (%) | 1540 (60%) | 60 846 (60%) | 1822 (68%) |
| Use of diuretics, n (%) | 277 (11%) | 6911 (7%) | 266 (10%) |
| Use of vitamin supplements, n (%) | 1128 (44%) | 49 020 (48%) | 1489 (56%) |
| Plasma non-HDL cholesterol, mmol/L | 3.8 (3.0-4.5) | 3.9 (3.2-4.7) | 4.0 (3.3-4.8) |
| Plasma albumin, µmol/L | 583 (550-617) | 589 (556-625) | 582 (549-616) |
| Renal impairment at baselinea, n (%) | 310 (12%) | 9663 (10%) | 402 (15%) |
| Diabetes at baseline, n (%) | 142 (6%) | 4243 (4%) | 140 (5%) |
| Cancer diagnosis prior to baseline, n (%) | 199 (7.8%) | 6833 (6.7%) | 282 (11%) |
| Cardiovascular disease prior to baseline, n (%) | 175 (6.8%) | 5797 (5.7%) | 172 (6.5%) |
| Vitamin D insufficiencyb, n (%) | 204 (41%) | 10 932 (45%) | 234 (43%) |
| Any thyroid/parathyroid operation prior to baseline, n (%) | 46 (1.8%) | 1226 (1.2%) | 223 (8.4%) |
Continuous variables are shown as median and interquartile range, and categorial variables as number and percent. Vitamin D insufficiency was defined as plasma vitamin D<50 nmol/L.
aRenal impairment defined as estimated glomerular filtration rate (eGFR) < 60 mL/min/1.73m2.
bMeasured in a subgroup of participants.
The association of baseline characteristics with plasma ionized calcium
| Plasma ionized calcium | |||
|---|---|---|---|
| Hypocalcemia < 2.5th percentile | Reference range 2.5th-97.5th percentiles | Hypercalcemia > 97.5th percentile | |
| Plasma ionized calcium, mmol/L | 1.11 (1.10-1.12)) | 1.21 (1.18-1.24) | 1.36 (1.34-1.40) |
| Men, n (%) | 1385 (54%) | 45 841 (45%) | 786 (29%) |
| Age, years | 56 (46-68) | 58 (48-67) | 64 (55-71) |
| Alcohol consumption, g/week | 84 (36-168) | 96 (48-180) | 108 (48-180) |
| Ever smokers, n (%) | 1397 (55%) | 59 101 (58%) | 1645 (62%) |
| High income, n (%) | 1167 (46%) | 43 743 (43%) | 875 (33%) |
| Body mass index, kg/m2 | 25.7 (23.3-28.8) | 25.6 (23.2-28.4) | 25.7 (23.2-28.7) |
| Hypertension, n (%) | 1540 (60%) | 60 846 (60%) | 1822 (68%) |
| Use of diuretics, n (%) | 277 (11%) | 6911 (7%) | 266 (10%) |
| Use of vitamin supplements, n (%) | 1128 (44%) | 49 020 (48%) | 1489 (56%) |
| Plasma non-HDL cholesterol, mmol/L | 3.8 (3.0-4.5) | 3.9 (3.2-4.7) | 4.0 (3.3-4.8) |
| Plasma albumin, µmol/L | 583 (550-617) | 589 (556-625) | 582 (549-616) |
| Renal impairment at baselinea, n (%) | 310 (12%) | 9663 (10%) | 402 (15%) |
| Diabetes at baseline, n (%) | 142 (6%) | 4243 (4%) | 140 (5%) |
| Cancer diagnosis prior to baseline, n (%) | 199 (7.8%) | 6833 (6.7%) | 282 (11%) |
| Cardiovascular disease prior to baseline, n (%) | 175 (6.8%) | 5797 (5.7%) | 172 (6.5%) |
| Vitamin D insufficiencyb, n (%) | 204 (41%) | 10 932 (45%) | 234 (43%) |
| Any thyroid/parathyroid operation prior to baseline, n (%) | 46 (1.8%) | 1226 (1.2%) | 223 (8.4%) |
| Plasma ionized calcium | |||
|---|---|---|---|
| Hypocalcemia < 2.5th percentile | Reference range 2.5th-97.5th percentiles | Hypercalcemia > 97.5th percentile | |
| Plasma ionized calcium, mmol/L | 1.11 (1.10-1.12)) | 1.21 (1.18-1.24) | 1.36 (1.34-1.40) |
| Men, n (%) | 1385 (54%) | 45 841 (45%) | 786 (29%) |
| Age, years | 56 (46-68) | 58 (48-67) | 64 (55-71) |
| Alcohol consumption, g/week | 84 (36-168) | 96 (48-180) | 108 (48-180) |
| Ever smokers, n (%) | 1397 (55%) | 59 101 (58%) | 1645 (62%) |
| High income, n (%) | 1167 (46%) | 43 743 (43%) | 875 (33%) |
| Body mass index, kg/m2 | 25.7 (23.3-28.8) | 25.6 (23.2-28.4) | 25.7 (23.2-28.7) |
| Hypertension, n (%) | 1540 (60%) | 60 846 (60%) | 1822 (68%) |
| Use of diuretics, n (%) | 277 (11%) | 6911 (7%) | 266 (10%) |
| Use of vitamin supplements, n (%) | 1128 (44%) | 49 020 (48%) | 1489 (56%) |
| Plasma non-HDL cholesterol, mmol/L | 3.8 (3.0-4.5) | 3.9 (3.2-4.7) | 4.0 (3.3-4.8) |
| Plasma albumin, µmol/L | 583 (550-617) | 589 (556-625) | 582 (549-616) |
| Renal impairment at baselinea, n (%) | 310 (12%) | 9663 (10%) | 402 (15%) |
| Diabetes at baseline, n (%) | 142 (6%) | 4243 (4%) | 140 (5%) |
| Cancer diagnosis prior to baseline, n (%) | 199 (7.8%) | 6833 (6.7%) | 282 (11%) |
| Cardiovascular disease prior to baseline, n (%) | 175 (6.8%) | 5797 (5.7%) | 172 (6.5%) |
| Vitamin D insufficiencyb, n (%) | 204 (41%) | 10 932 (45%) | 234 (43%) |
| Any thyroid/parathyroid operation prior to baseline, n (%) | 46 (1.8%) | 1226 (1.2%) | 223 (8.4%) |
Continuous variables are shown as median and interquartile range, and categorial variables as number and percent. Vitamin D insufficiency was defined as plasma vitamin D<50 nmol/L.
aRenal impairment defined as estimated glomerular filtration rate (eGFR) < 60 mL/min/1.73m2.
bMeasured in a subgroup of participants.
Compared with individuals with plasma ionized calcium within the reference interval, individuals with hypocalcemia were more often men (54% vs 45%), were younger (median age 56 vs 58 years), and more frequently used diuretics (11% vs 7%) (Table 1). In contrast, individuals with low plasma vitamin D concentration (plasma vitamin D below 50 nmol/L) were evenly distributed among the 3 calcium groups (41%, 45%, and 43% for hypo-, normo-, and hypercalcemia, P from chi2 test = 0.14). Of the individuals with hypercalcemia, 8.4% had thyroid or parathyroid operations performed prior to the examination date, whereas this was only the case for 1.8% and 1.2% of individuals with hypocalcemia and plasma ionized calcium within the reference interval, respectively.
All-Cause and Cause-Specific Mortality
Individuals with hypocalcemia displayed a higher incidence of all-cause mortality compared with individuals with plasma ionized calcium within the reference interval or with hypercalcemia (Fig. 1).
Cumulative incidence of all-cause and cause-specific mortality as a function of plasma ionized calcium. All-cause mortality is from Kaplan-Meier survival analyses, while cardiovascular-, cancer- and other mortality are from competing risk regression by STATA command stcompet (method of Aalen-Johansen).
We found an approximately U-shaped association between levels of plasma ionized calcium and risk of all-cause mortality with a significant increased risk below the median (Fig. 2, upper panel). The multivariable adjusted hazard ratios for all-cause mortality were 1.18 (95% CI, 1.05-1.32) and 1.00 (0.89-1.12) for hypocalcemia and hypercalcemia, respectively (Fig. 2, lower panel). The association was similar for cancer and other mortality, whereas for cardiovascular mortality, a significant increased risk was observed for higher plasma ionized calcium only (Fig. 2).
Risk of all-cause and cause-specific mortality as a function of plasma ionized calcium. Upper panel: spline curves of hazard ratios from Cox regression (solid thick black lines) with 95% CIs (dotted black lines) were adjusted for month and year of measurement, sex, BMI, physical activity during leisure time, hypertension at baseline, use of supplements, smoking status, income, alcohol intake, estimated glomerular filtration rate, plasma albumin, plasma non-HDL cholesterol, cardiovascular disease at baseline (not cardiovascular mortality), cancer at baseline (not cancer mortality). When the lower 95% confidence limit no longer overlap the reference value of 1.0 (solid thin black line at y = 1.0), the hazard ratio is statistically significant. The gray bell shows population distribution of plasma ionized calcium in percent. Lower panel: hazard ratios are calculated using Cox regression. Multivariable I: adjusted for the above-mentioned confounders. Multivariable II: as multivariable I but also adjusted for gastrointestinal cancer, inflammatory bowel disease, and osteoporotic fractures at baseline. Please notice that the numbers of cancer-, cardiovascular-, and other mortality do not add up to the total number of deaths, as 2910 individuals have no cause of death at time of analysis (as this register does not update as frequently as the other registers). Abbreviation: HR, hazard ratio.
Below the median, each 0.1 mmol/L lower plasma ionized calcium was associated with a multivariable adjusted hazard ratio of 1.23 (95% CI, 1.10-1.38) for all-cause mortality; for cancer- and other mortality the corresponding hazard ratios were 1.29 (1.06-1.57) and 1.24 (1.01-1.53) (Fig. 3). For cardiovascular mortality, each 0.1 mmol/L lower plasma ionized calcium below the median was associated with a hazard ratio of 1.19 (0.91-1.56) while each 0.1 mmol/L higher plasma ionized calcium above the median conferred a hazard ratio of 1.17 (1.02-1.35).
Risk of all-cause and cause-specific mortality per 0.1 mmol/L lower or higher plasma ionized calcium. Hazard ratios were calculated using Cox regression and adjusted for month and year of measurement, sex, BMI, physical activity during leisure time, hypertension at baseline, use of supplements, smoking status, income, alcohol intake, estimated glomerular filtration rate, plasma albumin, plasma non-HDL cholesterol, cardiovascular disease at baseline (not cardiovascular mortality), cancer at baseline (not cancer mortality).
Mortality: Sensitivity Analyses
In order to assess if plasma ionized calcium interacts with known cardiovascular risk factors on risk of all-cause mortality, we first stratified analyses on the median plasma ionized calcium and further examined the association between plasma ionized calcium and mortality in strata of different cardiovascular confounders (Fig. 4 and Supplemental Figure 3 (19)). Overall, the estimates were similar in the subgroups with increasing and decreasing plasma ionized calcium above and below the median including in strata according to previously studied risk factors such as ischemic heart disease and renal impairment. Also, when stratifying on other relevant confounders, no significant interactions were found. As a further sensitivity analysis, we adjusted our models for prior parathyroid or thyroid operation and the estimates were similar to those presented (data not shown). We also adjusted models for gastrointestinal cancer, inflammatory bowel disease, and osteoporotic fractures at baseline and found similar hazard ratios as in models not including these potential confounders. For hypocalcemia, hazard ratios after additional adjustment for these potential confounders were 1.18 (1.05-1.32) for all-cause mortality, 1.14 (0.87-1.50) for cardiovascular mortality, 1.23 (1.00-1.52) for cancer mortality and 1.13 (0.90-1.41) for other mortality compared to individuals with plasma ionized calcium within the reference range (Fig. 2, lower panel; Multivariable II).
Risk of all-cause mortality per 0.1 mmol/L lower or higher plasma ionized calcium, stratified by possible confounding factors. Hazard ratios were calculated using Cox regression and adjusted for month and year of measurement, sex, BMI, physical activity during leisure time, hypertension at baseline, use of supplements, smoking status, income, alcohol intake, estimated glomerular filtration rate, plasma albumin, plasma non-HDL cholesterol, cardiovascular disease at baseline (not cardiovascular mortality), cancer at baseline (not cancer mortality). Abbreviations: BMI, body mass index; eGFR, estimated glomerular filtration rate. *P for interaction between stratifying variable in categorical or continuous form and calcium categories on all-cause mortality, using a likelihood ratio test. The interaction term for the continuous variables were in continuous form (and not dichotomized) in order to have maximal power to detect effect modification (interaction terms).
Discussion
Our study shows that low plasma ionized calcium is associated with increased all-cause mortality, cancer mortality, and other mortality in the general population. Furthermore, we found that high plasma ionized calcium was associated with higher risk of cardiovascular mortality. This study is the first to assess plasma ionized calcium in relation to mortality in a large general population study, providing novel insights. Specifically, our study expands the knowledge on this field by using plasma ionized calcium as a risk factor for mortality rather than total calcium or albumin-adjusted calcium, either of which may lead to misclassification of calcium status especially in those with calcium derangements.
The mechanisms behind these observed associations are at present unclear; however, some mechanisms may be proposed. Extracellular calcium has been shown to affect cardiac contractility and electrical conduction, and hypocalcemia specifically has been implicated in QT prolongation and mechanical cardiac dysfunction, which could possibly increase mortality in individuals with other diseases (20, 21). Another possibility is that hypocalcemia triggers an increase in PTH secretion to restore calcium homeostasis, and that high PTH in itself has harmful effects on health such as an effect on cardiomyocytes or mineralization abnormalities (22, 23). The observed association between low plasma ionized calcium and higher cancer mortality could be mediated by effects on the calcium-sensing receptor, involved in cellular homeostasis and functions in both normal and neoplastic tissues. The calcium-sensing receptor has been involved in tissue proliferation, differentiation, apoptosis, and possibly osteolytic bone metastases which may contribute to pathogenesis and mortality of various cancers (24). As for hypercalcemia and cardiovascular mortality, we have previously reported that high plasma ionized calcium is associated with higher risk of cardiovascular disease in the general population (10). A possible mechanism behind higher cardiovascular mortality is that high plasma calcium may directly accelerate the atherosclerotic process by promoting differentiation of vascular smooth muscle cells into an osteogenic phenotype with local mineralization and atherosclerosis (25-27).
Findings from previous studies of the association of plasma calcium measures with mortality including from 1400 to 31 000 participants have been mostly positive although the shape of the association varies. In hospital settings, several studies have found an association of low total calcium, albumin-adjusted calcium, or ionized calcium with in-hospital mortality spanning from intensive care to nonacute admissions (1, 28). However, some studies have found a U-shaped association with all-cause mortality in similar hospital settings (2, 29, 30). To the best of our knowledge, only one other relatively large study has assessed the association between plasma ionized calcium and mortality; however, that study found no association between ionized calcium and mortality among 5147 Danish in-hospital patients followed for a median of 2.7 years (31). The association of low total calcium or albumin-adjusted calcium with mortality in the general population remains unclear as several conflicting findings have been reported, including null findings, associations with high calcium only, or U-shaped associations (3, 32-34). Nevertheless, in a meta-analysis from 2016 including 8 studies and approximately 256 000 participants, higher plasma total calcium was associated with mortality; hypocalcemia was not addressed in that study (4). These results are somewhat in contrast to our findings and the discrepancy may be caused by differences in calcium measures, disparities in included confounders, as well as differences in the study populations. Alternatively, although our population is large, perhaps an even larger sample size is needed to assess the association between high ionized calcium and all-cause mortality. As reviewed above, a large number of studies have been carried out previously on plasma total or albumin-corrected calcium on risk of mortality; however, our study expands the knowledge on this field by assessing the association with plasma ionized calcium, hereby avoiding misclassifications or confounding that may appear when total calcium or albumin-adjusted calcium are used to assess calcium status. In contrast to ionized calcium, total calcium or albumin-corrected calcium may not represent the actual calcium status, as total calcium is sensitive to change in concentrations of plasma proteins and small anions which again may be affected by chronic disease. However, we believe that our study adds credibility to these previous findings.
In addition to the use of ionized calcium, important strengths of our study were the large number of potential confounders adjusted for and the unselected population included. Some limitations to our study should, however, be noted. First, we only had one baseline measurement of plasma ionized calcium, and the latter may not reflect calcium concentrations of an individual over a longer time span. It is possible that using just a single measurement may lead to underestimation of risk associations in this long-term follow-up study. Furthermore, as our chemical analyzer capacity was not unlimited, we prioritized the measurement of ionized calcium over total calcium in the study cohort, as we believed it to be the better measure of calcium status. Unfortunately, and for this reason, plasma total calcium and phosphate measurements are not available, which would have been interesting to include in order to assess the full calcium homeostasis as well as to compare the different calcium measures in our population. Also, we could not directly account for the hypo- or hyperparathyroidism that may have been associated with the observed calcium disturbances, as we did not have PTH measurements available. It is possible that hypocalcemia is associated with mortality through hypoparathyroidism, most often seen following neck surgery (35); however, adjusting analyses for head and neck surgery did not alter the results. In addition, it is possible that other chronic illnesses such as chronic renal disease confounds the association between low ionized calcium and mortality; however, we did not find an interaction with renal function on all-cause mortality, and all analyses were adjusted for renal function in the attempt to remove this confounder. As to other chronic illnesses, including cancer, we adjusted for extensive self-reported and register-based information on the included individuals and confounding by unrecorded severe illnesses seems unlikely. Another potential confounder is the use of steroids, as this may interfere with the absorption of calcium and hereby decrease calcium levels; unfortunately, we were not able to adjust for steroid use in our study. Although an association between steroid-requiring morbidity and mortality cannot be excluded, a direct association between steroid use and calcium absorption has not been shown (36). Finally, several included potential confounders such as physical activity, smoking status, and alcohol intake are based on self-reported information, which may be subject to recall bias or recall errors that may bias or attenuate the results of the study.
The results of our study do not have immediate clinical implications, as causality between low ionized calcium and mortality remains to be shown. However, the finding that the association still exists when assessing ionized calcium in a general population could suggest that previous findings in hospital patients based on total calcium or albumin-corrected calcium are not solely due to misclassification, confounding, or reverse causation, which lends credence to the assumption that calcium could be mechanistically implicated in processes leading to death, both in acute cases and many years later. Future studies are needed to replicate our findings, and to specifically delve into classification of calcium disorders by comparing total or albumin-corrected calcium with ionized calcium, in order to investigate whether the misclassification issues seen with total or albumin-corrected calcium in patient subgroups (6, 7) also manifests in the general population and with mortality as an outcome. Finally, as observational associations are prone to confounding and reverse causation, genetic variants robustly associated with plasma ionized calcium need to be identified and Mendelian randomization studies of genetically high or low ionized calcium and mortality are warranted.
In conclusion, in the general population, low plasma ionized calcium was associated with increased all-cause, cancer-, and other mortality, while high levels were associated with increased cardiovascular mortality. Future large population studies should examine back-to-back ionized vs albumin-corrected total calcium, in order to address (1) the correlation between ionized and albumin-corrected total calcium in a general population setting; (2) the prevalence of discordance in the calcium status between ionized vs albumin-corrected total levels according to high, normal, and low levels of each index; and (3) evaluate the mortality risk of each category compared to the one with normal ionized calcium and normal albumin-corrected total calcium.
Abbreviations
- BMI
body mass index
- eGFR
estimated glomerular filtration rate
- HDL
high-density lipoprotein
- PTH
parathyroid hormone
Financial Support
This work was supported by Copenhagen University Hospital—Rigshospitalet and Copenhagen University Hospital—Herlev and Gentofte Hospital.
Disclosures
The authors have nothing to disclose.
Data Availability
Some or all datasets generated during and/or analyzed during the current study are not publicly available but are available from the corresponding author on reasonable request.
