Ionized Magnesium in Serum and Ultrafiltrate: pH and Bicarbonate Effect on Measurements with the AVL 988-4 Electrolyte Analyzer

¹ Lab. Biochim. Clin. Ematol., Ospedale Niguarda Ca'Granda, I20162 Milano, Italy;
² Electronics Design Center, Case Western Reserve Univ., Cleveland, OH 44106-7200;
^a author for correspondence: fax 39-2-64442901, e-mail marta.melotti{at}galactica.it

In blood, free ("ionized") magnesium (Mg²) is in equilibrium with complexed (protein-bound, and organic or inorganic complexed) species. This equilibrium is influenced by pH and by protein and ligand concentration, both in vivo and in vitro. The phenomenon is well known for calcium; pH influences the serum concentration of free calcium ion (S-cCa²) through the competition between hydrogen ions and calcium ions toward binding sites of protein. Fogh-Andersen (1) proposed the relationship dpcCa²/dpH, correcting the measured S-cCa² to the standard pH of 7.4, that, although controversial, is widely used in commercial analyzers (2)(3).

Recently, synthetic neutral carriers for the determination of free magnesium ion concentration (cMg²) have been developed (4)(5)(6)(7), the selectivity of which toward calcium allows the chemometric correction of the free magnesium measurement in the presence of pathophysiological calcium concentrations (8)(9). Magnesium bound to protein represents 30–35% of the total magnesium ion concentration in serum (6)(7)(8)(10), and, as with calcium, an empirical correction of measured serum free magnesium ion concentration (S-cMg²) to a standard pH can be proposed. The slope of the regression log S-cMg² vs pH is ~-0.1 (6)(11), half that for S-cCa². In the S-cMg² range of 0.31–0.76 mmol/L, we found an average value of -0.117 (8), which was reproducible during the life-span of the tested electrodes.

By contrast, Ising et al. (12), evaluating the performance of the Kone Microlyte Magnesium assay (magnesium ionophore ETH 5220), found a dependence of the dpcMg²/dpH relation on the electrode life-time. According to these authors, the built-in value of -0.07 was incorrect, and could produce error in the normalized value up to 10%.

We performed the present study on the AVL 988-4 electrolyte analyzer (AVL Medical Instruments) (8), which uses a highly purified ETH 7025 Mg² ionophore, with a typical slope of 13–15 mV/decade in the presence of 1.25 mmol/L calcium background (4). According to the authors of that report, "with a slope of 15 mV/decade of magnesium in presence of calcium the electrode yielded excellent results on all types of specimen tested" (4).

We measured both ultrafiltrable (UF)¹ and serum free calcium and magnesium and found that, at a measured ultrafiltrate pH of ~8.3, the average ultrafiltrate values were, respectively, 40% and 29% lower than the serum values at actual pH. A difference was expected, and its magnitude prompted us to further investigations.

In the ultrafiltration procedure used, the pH in serum changes, causing a change in S-cCa² and S-cMg² at the interface serum/ultrafilter membrane. Although AVL states that the formula for the correction of S-cCa² is valid only in the pH range of 7.2–7.6, we forced the correction up to pH 8.3. In fact, in several serum samples it was possible to graphically extrapolate the linear trend of the dpcCa²/dpH relation up to pH 8.3, with an error not greater than -10%, as compared with the experimental data. Similarly, for NOVA analyzers, which use the same calcium ionophore as AVL (ETH 1001), the instruction manuals (13) report that, with some limitation, their formulas can be used for pH 6.9–8.0. Hence, we recalculated in 90 samples, having actual pHs from 7.21 to 7.53, the S-cCa² at pH 8.3 (the average pH of the ultrafiltrates) by the formula:

(1)

The mean measured UF-cCa² was 0.75 mmol/L, and the mean S-cCa²_{pH 8.3} recalculated from S-cCa²_{actual pH} was 0.77 mmol/L. For S-cMg² also, we noted a nearly linear behavior of the in vitro dpcMg²/dpH relation in the pH range of 7.2–8.5, probably because of the lower value of the slope. Hence, an analogous correction algorithm was applied to the S-cMg² of the same samples by using the appropriate slope value of 0.117:

(2)

The calculated values were significantly different from the measured cMg² in ultrafiltrate (UF-cMg²_measured), the mean value of which was 0.51 mmol/L, vs a mean recalculated S-cMg²_{pH 8.3} of 0.61 mmol/L (P <0.05, paired Student's t-test). The variables of the linear regression S-cMg²_{pH 8.3} vs UF-cMg²_measured were: slope = 0.927, intercept = 0.094 mmol/L, and correlation coefficient = 0.982.

We described elsewhere several other experiments devoted to the explanation of the different behavior of the magnesium ion with respect to the calcium ion (8). In one of these, based on many reports published previously about the ion binding by bicarbonate (14)(15)(16), we tested aqueous magnesium chloride solutions with physiological phosphate buffer concentration and ionic strength with increasing bicarbonate concentration: the decrease of magnesium, as measured by the electrode, was -0.005 mmol/L per 1 mmol/L of added bicarbonate. We worked out an extended correction formula that took into account the effects of both pH and total carbon dioxide concentration (cTCO₂):

(3)

and used this to recalculate the values of the 68 samples for which cTCO₂ data in ultrafiltrate were available.

Although the paired t-test inidicated a significant difference (P <0.05) between the UF-cMg²_measured (mean value 0.51 mmol/L) and the cTCO₂-effect-corrected S-cMg²_{pH 8.3} (mean value, 0.46 mmol/L), the variables of the linear regression S-cMg²_{pH 8.3} vs UF-cMg²_measured were: slope = 0.991, intercept = -0.043 mmol/L, and correlation coefficient = 0.982. In Fig. 1 , the difference between the two kinds of calculation of S-cMg²_{pH 8.3} is patent.

View larger version (23K): [in this window] [in a new window]   Figure 1. Correlation between ultrafiltrate free magnesium values of 68 samples and, for respective sera, the measured values of free magnesium (+) and the calculated values according to Eq. 2 (------) and Eq. 3 (*). Statistical variables for the measured and calculated (Eq. 2 and Eq. 3 ) results are, respectively: intercept, 0.11, 0.094, and -0.04 mmol/L; slope, 1.206, 0.927, and 0.991; correlation coefficient, 0.982, 0.982, and 0.982.

These last results are in good agreement with those obtained by McGuigan et al. (17), who used the same ultrafiltration procedure we used but a different mathematical approach. They explained the relevant difference between expected and measured anion-complexed magnesium in ultrafiltrate by a pH-dependent complexing of magnesium to UF anions. They found a linear correlation between UF-cMg² and pH, but when the pH of ultrafiltrate was changed with HCl or NaOH, the correlation was not linear. Furthermore, these authors affirmed that by empirically correcting the actual measured values in the ultrafiltrate with this curve, the values came very close to but were not exactly the same as the expected S-cMg²_{pH 7.4}. The reason for this small difference (about 0.04 mmol/L) is not clear yet.

Another interesting, and perhaps connected, phenomenon was seen by us (8) and by Altura et al. (18). We removed carbon dioxide from three pooled sera, which we successively equilibrated back to low pH by tonometry with gaseous CO₂. An additional loss of CO₂ up to alkaline pH did not change the measured S-cMg² (8). Altura et al. (18) repeatedly froze and thawed serum, measured S-cMg² and S-cCa² after each cycle on a different instrument (NOVA STAT Profile 8; NOVA Biomedical), and found no change of S-cMg² as pH was increasing from 7.45 to 7.80. Nonetheless, in both experiments the change of S-cCa² with pH was as expected, with a slope equal to -0.23 in our experiments and a S-cCa² decrease of -0.04 mmol/L per 0.1 pH unit in the Altura experiment.

Taking into account the TCO₂ concentration, we were able to correct the apparent higher degree of binding of magnesium ion than of calcium ion, but we probably overcorrected. As the medium becomes more and more alkaline, the magnesium ion may be sequestered in a chemical species (not necessarily magnesium bicarbonate) and not detected by the electrode in the same way as the free magnesium ion. Acidification of the medium with gaseous CO₂ cannot free the sequestered magnesium ion. The easiest explanation is that at high pH or low temperature, an aliquot of magnesium ion is definitively sequestered in the form of magnesium hydroxide, and hence is not sensed by the ionophore.

However, according to Czaban et al. (16), because of the change of ion environment, the change of activity coefficient and liquid junction potential error could modify the ion data as NaHCO₃ is substituted for NaCl. This phenomenon, described for sodium, could be also extended to magnesium.

Despite the extensive mathematical processing by both the AVL analyzer and us, the reported data demonstrate, at least with this kind of ionophore, that magnesium ion solutions do not exhibit the same behavior as calcium ion solutions, as far as pH changes in the medium are concerned. For other reasons, the data by Ising et al. (12) confirm these findings.

Acknowledgments

We are grateful to Angelo Manzoni (Instrumentation Laboratory) for careful review of the manuscript and helpful discussions.

Footnotes

^{1 1} The ultrafiltration procedure has been performed with the Microcon 10 (Amicon; molecular-mass cutoff, 10 000 Da), containing 500 µL of serum, centrifuged at 10 000g for 45 min at 30 °C without any special anerobic protection, and recovering 300 µL of ultrafiltrate.

References

1. Fogh-Andersen N. Ionized calcium analyzer with built-in pH correction. Clin Chem 1981;27:1264-1267. [Abstract/Free Full Text]
2. Graham GA. Measured ionized calcium (Ca⁺⁺) or calculated Ca⁺⁺ at a pH of 7.4. In: Burrit M, Cormier AD, Mass AHJ, Moran RF, O'Connel KM, eds. Methodology and clinical applications of ion-selective electrodes. Rochester, MN: Davies Printing, 1988:113–20..
3. Sachs C, Chaneac M, Rabouine P, Kindermans C, Dechaux M. Anomalies in pH 7.4 correction in ionized calcium analysers. Ann Clin Biochem 1989;26:542-546.
4. Sachs C, Ritter CH, Puaud AC, Gahramani M, Kindermans C, Spichiger UE, Marsoner HJ. Measurement of ionized magnesium with ion selective electrode. Aizawa M Kwva K eds. Methodology and clinical applications of blood gases, pH, electrolytes and sensor application 1992:177-184 International Federation of Clinical Chemistry Copenhagen. .
5. Altura BT, Shirey TL, Young CC, Hiti J, Dell'Orfano K, Handwerker SM, Altura BM. A new method for the rapid determination of ionized Mg⁺⁺ in whole blood, serum and plasma. Methods Find Exp Clin Pharmacol 1992;14:297-304. [Web of Science][Medline] [Order article via Infotrieve]
6. Maj-Zurawska M, Hulanicki A, Drygienniec D, Pertkiewicz M, Krokowski M, Zebrowski A, Lewenstam A. Ionized and total magnesium level in blood serum and plasma of healthy and ill adults. Electroanalysis 1993;5:713-717.
7. van Ingen HE, Huijgen HJ, Kok WT, Sanders GTB. Analytical evaluation of Kone Microlyte determination of ionized magnesium. Clin Chem 1994;40:52-55. [Abstract/Free Full Text]
8. Zoppi F, De Gasperi A, Guagnellini E, Marocchi A, Mineo E, Pazzucconi F, et al. Measurement of ionized magnesium with AVL 988/4 electrolyte analyzer: preliminary analytical and clinical results. Scand J Clin Lab Invest 1996;56(Suppl 224):259-274. [Web of Science][Medline] [Order article via Infotrieve]
9. Rehak NN, Cecco SA, Niemela JE, Hristova EN, Elin RJ. Linearity and stability of the AVL and Nova magnesium and calcium ion-selective electrodes. Clin Chem 1996;42:880-887. [Abstract/Free Full Text]
10. Hristova EN, Cecco SA, Niemela JE, Rehak NN, Elin RJ. Analyzer-dependent differences in results for ionized calcium, ionized magnesium, sodium, and pH. Clin Chem 1995;41:1649-1653. [Abstract/Free Full Text]
11. Dianela F, Okorodudu AO. A method for the measurement of ionized magnesium. Lab Med 1995;26:70-74.
12. Ising H, Bertschat F, Günter T, Jeremias E, Jeremias A. Measurement of free magnesium in blood, serum and plasma with an ion-selective electrode. Eur J Clin Chem Clin Biochem 1995;33:365-371. [Web of Science][Medline] [Order article via Infotrieve]
13. NOVA 8 ionized calcium/pH analyzer and STAT Profile instruction manuals. Waltham, MA: NOVA Biomedical, 1983;2.4–2.5 and 1989:8.10–8.11..
14. Coleman RL, Young CC. Evidence for formation of bicarbonate complexes with Na⁺ and K⁺ under physiological conditions [Letter]. Clin Chem 1981;27:1938-1939. [Free Full Text]
15. Czaban JD, Cormier AD, Legg KD. The apparent suppression of Na/K data obtained with ion-selective electrodes is due to junction potential and activity coefficient effect, not to bicarbonate binding [Letter], and Coleman RL, Young CC [Reply]. Clin Chem 1982;28:1703-1706. [Free Full Text]
16. Czaban JD, Cormier AD, Legg KD. Establishing the direct-potentiometric "normal" range for Na/K: residual liquid-junction potential and activity coefficient effects. Clin Chem 1982;28:1936-1945. [Abstract/Free Full Text]
17. McGuigan JAS, Gerber DM, Weiss M, Lüthi D. Measurement of ionized magnesium concentration in blood of normal individuals and in patients in intensive care with the AVL 988-4 magnesium analyzer. Satellite Symposium IV: Magnesium ion analysis. 5th European Magnesium Congress, Vienna: June 15–18, 1995..
18. Altura BT, Shirey TL, Young CC, Dell'Orfano K, Hiti J, Welsh R, et al. Characterization of a new ion selective electrode for ionized magnesium in whole blood, serum, plasma and aqueous samples. Scand J Clin Lab Invest 1994;54(Suppl 217):21-36.