Study overview

This report is a secondary analysis of data collected in a previous study to assess the impact of specific pre-analytic variables on hemoglobin and zinc assessment in blood [7,10]. The study was conducted completely on the campus of the Children’s Hospital Oakland Research Institute (CHORI), now part of the University of California San Francisco, from 8/26/2019–1/29/2020. The CHORI Institutional Review Board approved this study (2019−055). All participants provided written consent. Sixty adults were recruited from Oakland, California, USA. Inclusion criteria included being 18 years or older, generally healthy with no chronic illness or blood disorders. Exclusion criteria included not meeting inclusion criteria or being pregnant due to known complexities in nutrient homeostasis. Participants were asked to avoid mineral and other supplements for 24 hours before blood draw and provide their age, gender, and race/ethnicity simply as markers of participant diversity. All demographic and anthropometric values were previously published [10]. We utilized the Checklist for Reporting Stability Studies (CRESS) framework for comparability with other studies on the assessment of minerals from blood [11].

Sample processing

Participants provided non-fasting blood samples beginning with fingerstick(s) followed by conventional phlebotomy [10]. Variables commonly associated with blood draw and processing were minimized, including use of the same clinical site, clinical procedures, phlebotomist, blood draw procedures, and sample handling procedures [7,10]. Sequential blood collections were used to generate venous plasma, venous serum, and capillary plasma samples for each participant. Commercial blood collection tubes (BCT) were prepared for each participant, including plasma tubes (BD Vacutainer product# 368381 containing dipotassium ethylenediaminetetraacetic acid (K2EDTA)), serum tubes (BD Vacutainer product# 368380 containing clot activator additive), and capillary tubes (BD Microtainer product# 365974 containing K2EDTA). The anticoagulant K2EDTA is capable of binding to many minerals within the plasma, but this is not a concern for our instrumentation that can quantify mineral levels regardless of binding partners. Sample volumes, processing times, and detailed laboratory steps were previously published [7,10]. The blood fractions were aliquoted into metal-free polypropylene tubes and stored at −70°C in a monitored ultralow freezer until ready for batch analysis.

Analysis of mineral content

Mineral content was determined by inductively coupled plasma optical emission spectrometry (ICP-OES). The ICP-OES methods and operating conditions were previously described and independently evaluated [7,12]. In brief, 100 µl samples of plasma and serum were transferred to metal-free 15 ml conical tubes (Perfector Scientific), digested in 70% OmniTrace nitric acid (VWR) for 12–16 hours, and then diluted to 5% nitric acid with OmniTrace water (VWR). To test for mineral contamination, the BCTs and plasticware were rinsed with 5% nitric acid and transferred to conical tubes for analysis. All samples were vortexed 10–30 sec, centrifuged at 4,000xg for 10 min, and then analyzed on a 5100 Synchronous Vertical Dual View ICP-OES (Agilent Technologies). Method details, detection limits, and wavelength are provided in S1 Table. The ICP-OES was calibrated using National Institute of Standards and Technology (NIST) traceable elemental standards and validated using Seronorm Trace Element Levels 1 and 2 standard reference materials (Sero AS) as external quality controls (S1 Table). Additionally, mean composite elemental values across 6 run days were within the reference range targets provided for Seronorm Trace Element Levels 1 and 2. Interassay precision of mineral content was assessed by testing 30 pooled plasma and 30 pooled serum samples generated from excess blood obtained from this study (S1 Table). For the pooled plasma, the CV was found to be 1.7% for calcium, 1.8% for copper, 2.2% for iron, and 1.8% for magnesium. For the pooled serum, the CV was found to be 3.8% for calcium, 2.7% for copper, 5.0% for iron, and 3.8% for magnesium.

Data analysis

Sample size was based on power calculations for differences in circulating zinc content as previously described [7]. All graphing and statistical testing were conducted using Prism 10 (GraphPad Software, Inc). Outlier analysis was conducted using the GraphPad ROUT algorithm with Q = 0.1% [13]. Data normality was assessed with the D’Agostino-Pearson omnibus K2 normality test. For comparisons involving blood draw site or matrix, significance was determined using two-tailed paired t tests with additional testing using Pearson correlation and Bland-Altman analysis. For all analyses, statistical significance was assigned at p < 0.05.

Assessment of multiple minerals in blood samples

We previously used ICP-OES to determine how zinc measurements was influenced by several pre-analytical variables, but the instrument also records data for other minerals thus allowing for this secondary analysis [7]. BCTs used for venous blood collection were originally selected for certification in measuring zinc concentrations in blood, but they were also certified for other trace elements including calcium (<150 µg/L), copper (<5 µg/L), iron (<25 µg/L), and magnesium (<40 µg/L), according to manufacturer’s data. To confirm this, we assessed several empty tubes from each lot of BCTs for background mineral content. Calcium, copper, iron, and magnesium were below detection in most of the empty tubes, whereas phosphorous, potassium, silicon, sodium, and sulfur were found to be moderately to heavily contaminated within the empty BCTs (S1 Table). Therefore, we limited our analysis to calcium, copper, iron, and magnesium within study blood samples. BCTs used for capillary blood collection had limited trace element certification, with limits for only lead at <1 ng/tube according to manufacturer’s data. Therefore, unused tubes from the same lot were tested by ICP-OES beforehand and were found to have undetectable levels of calcium, copper, iron, and magnesium.

Effect of blood draw site on mineral measurement

Calcium, copper, iron, and magnesium concentrations were quantified in blood drawn from capillary and venous sites from all study participants (Table 1 and Fig 1). All datasets passed normality tests, so parametric statistics were used.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Mineral concentrations differ due to preanalytical variables.

https://doi.org/10.1371/journal.pone.0338582.t001

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Comparison of mineral values from different blood draw site.

The plasma concentrations of calcium (A), copper (B), iron (C), and magnesium (D), are shown from sequential blood draws taken from capillary and venous draw sites. The histograms in the first column show the distribution of mineral concentrations for capillary (red) and venous (blue) samples. The graphs in the second column show the correlation of the mineral concentrations (mean ± SEM) for both draw sites compared to the line of identity (solid black line) and clinical references ranges (green shading) where available [14]. The linear correlation (solid red line) and 95% confidence band (dashed red line) of the mineral concentrations are also shown. Only copper had established reference ranges in venous plasma, while no reference ranges for minerals in capillary plasma were available. The graphs in the third column show the corresponding Bland-Altman assessment for both draw sites showing the line of identity (solid black line), mean bias (solid red line), and 95% confidence interval (dashed red line).

https://doi.org/10.1371/journal.pone.0338582.g001

For calcium, the capillary samples had 3 sample identified as outliers while venous samples had 2 samples identified as outliers (with 0 overlap), resulting in a total of 55 matched samples. The mean calcium concentration was 93.82 ± 0.68 mg/L for capillary and 88.70 ± 0.45 mg/L for venous, with the difference being statistically significant (p < 0.0001) using a two-tailed paired t-test. This indicated that capillary calcium concentration was elevated by 5.12 mg/L (6%) in comparison to venous blood values with a sequential blood draw from the same participants. Bland-Altman analysis revealed a similar bias of 5.13 mg/L for capillary compared to venous calcium concentrations.

For copper, the capillary samples had 1 sample identified as an outlier while venous samples had 1 sample identified as an outlier (with 1 overlap), resulting in a total of 59 matched samples. The mean copper concentration was 0.866 ± 0.024 mg/L for capillary and 0.836 ± 0.024 mg/L for venous, with the difference being statistically significant (p < 0.0001) using a two-tailed paired t-test. This indicated that capillary copper concentration was elevated by 0.031 mg/L (4%) in comparison to venous blood values with a sequential blood draw from the same participants. Bland-Altman analysis revealed a similar bias of 0.031 mg/L for capillary compared to venous copper concentrations.

For iron, the capillary samples had 3 samples identified as outliers while venous samples had 0 samples identified as outliers, resulting in a total of 57 matched samples. The mean iron concentration was 1.282 ± 0.075 mg/L for capillary and 1.159 ± 0.053 mg/L for venous, with the difference being statistically significant (p = 0.046) using a two-tailed paired t-test. This indicated that capillary iron concentration was elevated by 0.123 mg/L (11%) in comparison to venous blood values with a sequential blood draw from the same participants. Bland-Altman analysis revealed a similar bias of 0.123 mg/L for capillary compared to venous iron concentrations.

For magnesium, neither the capillary nor venous samples had outliers, resulting in a total of 60 matched samples. The mean magnesium concentration was 17.82 ± 0.18 mg/L for capillary and 17.39 ± 0.17 mg/L for venous, with the difference being statistically significant (p < 0.0001) using a two-tailed paired t-test. This indicated that capillary magnesium concentration was elevated by 0.43 mg/L (2%) in comparison to venous blood values with a sequential blood draw from the same participants. Bland-Altman analysis revealed a similar bias of 0.43 mg/L for capillary compared to venous magnesium concentrations.

Effect of blood matrix on mineral measurement

Calcium, copper, iron, and magnesium concentrations were also determined within the same venous blood samples processed into plasma and serum from all study participants (Table 1 and Fig 2). All datasets passed normality tests, so parametric statistics were used.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Comparison of mineral values from different blood matrices.

The venous concentrations of calcium (A), copper (B), iron (C), and magnesium (D), are shown from the same blood draws but processed to plasma and serum matrices. The histograms in the first column shows the distribution of mineral concentrations for plasma (lavender) and serum (red) samples. The graphs in the second column show the correlation of the mineral concentrations (mean ± SEM) for both blood matrices compared to the line of identity (solid black line) and clinical references ranges (green shading) where available [14]. The linear correlation (solid red line) and 95% confidence band (dashed red line) of the mineral concentrations are also shown. All found minerals had established reference ranges in venous serum, while only copper had established reference ranges in venous plasma. The graphs in the third column show the corresponding Bland-Altman assessment for both blood matrices showing the line of identity (solid black line), mean bias (solid red line), and 95% confidence interval (dashed red line).

https://doi.org/10.1371/journal.pone.0338582.g002

For calcium, the plasma samples had 3 samples identified as outliers while serum samples had 3 samples identified as outliers and 1 missing value (with 3 overlap), resulting in a total of 56 matched samples. The mean calcium concentration was 89.23 ± 0.50 mg/L for plasma and 91.03 ± 0.42 mg/L for serum, with the difference being statistically significant (p = 0.0001) using a two-tailed paired t-test. This indicated that serum calcium concentration was elevated by 1.80 mg/L (2%) in comparison to plasma values with a sequential blood draw from the same participants. Bland-Altman analysis revealed a similar bias of 1.80 mg/L for serum compared to plasma calcium concentrations.

For copper, the plasma samples had 1 sample identified as an outlier while serum samples had 1 sample identified as an outlier and 1 missing value (with 1 overlap), resulting in a total of 58 matched samples. The mean copper concentration was 0.835 ± 0.024 mg/L for plasma and 0.875 ± 0.026 mg/L for serum, with the difference being statistically significant (p < 0.0001) using a two-tailed paired t-test. This indicated that serum copper concentration was elevated by 0.040 mg/L (5%) in comparison to plasma values with a sequential blood draw from the same participants. Bland-Altman analysis revealed a similar bias of 0.040 mg/L for serum compared to plasma copper concentrations.

For iron, the plasma samples had 1 sample identified as an outlier while serum samples had 0 samples identified as outliers and 1 missing value (with 1 overlap), resulting in a total of 59 matched samples. The mean iron concentration was 1.185 ± 0.052 mg/L for plasma and 1.112 ± 0.050 mg/L for serum, with the difference being statistically significant (p = 0.007) using a two-tailed paired t-test. This indicated that plasma iron concentration was elevated by 0.073 mg/L (7%) in comparison to serum values with a sequential blood draw from the same participants. Bland-Altman analysis revealed a similar bias of 0.073 mg/L for plasma compared to serum iron concentrations.

For magnesium, the plasma samples had 1 sample identified as an outlier while serum samples had 0 samples identified as outliers and 1 missing value (with 1 overlap), resulting in a total of 59 matched samples. The mean magnesium concentration was 17.40 ± 0.18 mg/L for plasma and 17.71 ± 0.18 mg/L for serum, with the difference being statistically significant (p < 0.0001) using a two-tailed paired t-test. This indicated that serum magnesium concentration was elevated by 0.310 mg/L (2%) in comparison to plasma values with a sequential blood draw from the same participants. Bland-Altman analysis had a bias of 0.308 mg/L for serum compared to plasma magnesium concentrations.

Calcium

Calcium is the most abundant mineral in the human body and best known as the key constituent in bones and teeth, but also is essential for muscle contraction, vascular function, neuronal activity, and signal transduction. Calcium adequacy is a public health concern in the USA as approximately 42% of the population remains below the estimated average requirement (EAR) for daily calcium intake [2]. Normal levels for circulating calcium are between 82–102 mg/L for most adults, although those values are specified for only serum – not plasma [14]. Circulating calcium levels are under tight homeostatic control and do not typically reflect whole-body calcium status [2]. Moreover, no clinical or biochemical test has been established which allows for direct determination of an individual’s calcium nutritional status, so the search for biochemical indicators of calcium status remains an active area of research. Measuring calcium levels in the blood may not be useful for determining nutritional status, but blood calcium levels are known to change under certain physiological or pathological states. Conditions that are associated with increased circulating calcium include hyperparathyroidism, Paget’s disease, and certain malignancies [3,4,20,21]. Conditions that are associated with decreased circulating calcium include chronic kidney disease, hypoparathyroidism, leprosy, hepatic cirrhosis, and vitamin D deficiency [3,4,22,23]. Therefore, understanding the impact of pre-analytic variables on calcium assessment in blood has practical clinical value.

The impact of blood draw location on calcium levels has been previously evaluated. Early studies tested the concentrations of numerous blood constituents between capillary and venous samples, finding no significant differences in calcium levels [24,25]. Later groups reported that calcium levels were not different between the blood draw locations or slightly higher in venous compared to capillary serum but judged to not be clinically relevant [26–29]. In contrast, one recent study reported a small but significantly higher level of calcium in capillary compared to venous plasma from the same participants [30]. We also found that calcium was significantly higher in capillary compared to venous plasma. In the Falch, Kupke et al, Doeleman et al, and Parikh et al studies, the trace metal status of the BCTs was not indicated and no background testing was provided. In the Ansari et al study, the trace metal status of the BCTs was not indicated and the BCTs were processed at different times (4 vs. 72 hours) which complicates the comparison of calcium levels. In all studies, calcium levels were estimated with a multimodal clinical analyzer instead of an instrument optimized for mineral measurement.

The impact of blood sample matrix on calcium levels was previously explored in a few papers. Three studies have reported that calcium measurements were modestly elevated in plasma compared to serum [31–33]. However, we found that calcium was significantly higher in serum compared to plasma. In the Herbert study, few details of the BCTs types were provided other than sodium citrate was stated as the anticoagulant for plasma samples. In the Lester and Varghese study, lithium heparin was stated as the anticoagulant for the plasma samples, but no information was given about the serum tubes. In the Linder et al study, EDTA was stated as the anticoagulant for plasma samples so likely the same as the K2-EDTA we used in our study, but there was no information given for the serum tubes. None of the older studies reported certification or testing of BCTs for background calcium levels.

Copper

Copper is a trace mineral with essential roles in energy production, connective tissue maintenance, oxidative stress defense, signal transduction, and neuronal function. The amount of copper needed for good health is modest and typically adequate in the average Western diet [34]. Normal levels for circulating copper are given as between 0.7–1.4 mg/L for plasma and 0.56–1.69 mg/L for serum in most adults [14]. Copper levels in the blood are influenced by changes in inflammation and other minerals, but circulating copper is still considered reflective of whole body copper status [35]. Beyond the use as a nutritional marker, copper levels in the blood are also altered during specific clinical conditions. Conditions that are associated with increased circulating copper include rheumatoid arthritis, anemia, Hodgkin’s lymphoma, and pregnancy [36–39]. Conditions that are associated with decreased circulating copper include Wilson’s disease, Menkes’ syndrome, and Alzheimer’s disease [40–43]. Therefore, it is important to know the impact of pre-analytic variables on copper assessment.

The impact of blood draw location on copper levels was explored in at least one study. Rodríguez-Saldaña et al tested several the levels of several minerals in blood taken from different blood sites, but found no significant differences in copper content [26,44]. In contrast, our findings showed that copper was significantly higher in capillary compared to venous blood. Rodríguez-Saldaña et al used trace metal-certified K2-EDTA tubes for venous samples, but there was no description of certification or background testing of their capillary tubes.

The impact of blood sample matrix on copper levels was previously investigated by three teams. In an older study, Smith et al reported higher copper values in serum over plasma, but the difference was not statistically significant [45]. Two more recent studies provided copper levels in serum and plasma, but neither study was designed to directly test for this so the statistical differences in copper levels is unclear [46,47]. Additionally, neither study reported use of certified tubes nor tested for background copper, though the Toro-Raman et al study did state the BCTs were ‘metal free.” Our results showed that copper was significantly higher in serum compared to plasma.

Iron

Iron is the most abundant trace metal in the body and is essential for oxygen transport, energy production, immune function, oxidative stress defense, and neuronal function. Iron adequacy is a public health concern in the USA, particularly for women of reproductive age [34]. While only 1% of the population is listed as below the EAR for daily iron intake, the true prevalence of iron-deficiency anemia in the USA may be closer to 14–19% [48,49]. Normal levels for iron in the blood are given as between 0.5–1.75 mg/L for adults, although those values are specifically listed for only serum – not plasma [14]. Most of the iron in the blood is bound to the protein transferrin, so the iron saturation of transferrin is a more robust indicator of iron status than blood iron content alone. Furthermore, the circulating levels of iron are under homeostatic control and easily overridden by inflammatory signals, as the sequestration of circulating iron is a central protective mechanism during infection [34]. Measuring total iron levels in the blood may not be optimal for determining nutritional status, but iron levels in the blood are known to change under certain physiological or pathological states. Conditions that are associated with increased circulating iron include acute kidney disease, acute leukemia, and several hemoglobinopathies [50–52]. Conditions that are associated with decreased circulating iron include chronic kidney disease, hypothyroidism, and infection [53–55]. Therefore, understanding the impact of pre-analytic variables on iron assessment in blood could hold practical clinical value.

The impact of blood draw location on iron levels was explored in two studies. Rodríguez-Saldaña et al and Doeleman et al tested the levels of several minerals in blood taken from different blood sites, but observed no significant difference in iron content [26,44]. In contrast, our findings show that iron measurements were significantly elevated in capillary blood compared to venous blood. Questions about the Rodríguez-Saldaña et al and Doeleman et al study are mentioned above.

The impact of blood sample matrix on iron levels was also previously reported. Two teams reported that iron levels were higher in serum than plasma [46,56]. However, our findings showed that iron measurements were significantly elevated in plasma compared to serum. In the Kasperek et al study, heparin was stated as the anticoagulant for plasma samples, but no information was given for the serum tubes. The study did report the background iron levels in the plasma tubes, but not the serum tubes. This study also stated that the iron levels being higher in serum were ‘well established’ but we would dispute that claim. Neither study reported certification or background testing for iron levels in their capillary tubes.

Magnesium

Magnesium is an essential mineral that cofactor in hundreds of enzymes involved in energy production, bone health, nucleic acid synthesis, protein synthesis, ion transport, oxidative stress defense, and signal transduction. Magnesium adequacy is a public health concern in the USA as approximately 51% of the population remains below the EAR for daily magnesium intake [2]. Normal levels for magnesium in the blood are given as between 16–26 mg/L for adults, although those values are specifically listed for only serum – not plasma [14]. Circulating magnesium levels are under tight homeostatic control and do not typically reflect the whole-body magnesium status [2]. Moreover, no clinical or biochemical test has been established which allows for direct determination of an individual’s magnesium nutritional status; as with calcium, the search for a biochemical indicator of magnesium status remains an active area of research. Measuring magnesium levels in the blood may not be useful for determining nutritional status, but blood magnesium levels are known to change under certain physiological or pathological states. Conditions that are associated with increased circulating magnesium include hypothyroidism, Addison’s disease, and chronic kidney disease [3,57,58]. Conditions that are associated with decreased circulating magnesium include hyperthyroidism, acute pancreatitis, chronic alcoholism, and sometimes pregnancy [59–63]. Therefore, understanding the impact of pre-analytic variables on magnesium assessment in blood could hold practical clinical value.

The impact of blood draw location on magnesium levels was also explored in at least three previous studies. Matthiesen et al reported a greater ionized magnesium concentration in venous blood compared to capillary blood, but this is distinct from total magnesium levels and thus not comparable to our analysis [64]. Other groups reported that magnesium levels were not different between the blood draw locations or slightly higher in capillary compared to venous plasma from the same participants [26,30]. We also found that magnesium levels were elevated in capillary blood compared to venous. In the Matthiesen et al study, the trace metal status of the BCTs was not indicated and no background testing was provided. Questions about the Doeleman et al and Ansari et al study are mentioned above.

The impact of blood sample matrix on magnesium levels was also explored in at least one previous paper. Linder and colleagues reported that magnesium values were higher in plasma than in serum [32]. However, our findings were that magnesium levels were elevated in serum compared to plasma. In the Linder et al study, EDTA was stated as the anticoagulant for plasma samples so likely same as K2-EDTA as used in our study, but there was no information on serum tubes. None of these older studies reported certification or background testing of tubes for magnesium levels.