Clinical laboratory evaluation of the Orion SS-20 ionized calcium analyser

for the gas stream in the distillation unit. Coagulation of the beer samples, especially when running a large number of samples, can arise in the distillation unit. Coagulation interferes with the smooth flow of sample through the distillation unit and gives erratic results. Coagulation was prevented by mixing the sample with a M potassium hydroxide solution (Figure 2) before pumping it into the distillation unit. The pH of the beer sample then varied between 6 and 7.


Introduction
About 50e/0 of the total plasma calcium exists as free calcium ions (ionized calcium, Ca +), and this is generally considered to be the physiologically active fraction. There is evidence to suggest that its direct measurement is of greater value than total calcium in various disorders [1,2]. Until recently techniques for measurement of ionised calcium (spectrophotometric [3][4][5] and potentiometric [6][7][8][9][10][11][12]) have been too unreliable or time-consuming for routine use in most laboratories, pH and temperature affect the binding of calcium to protein [4], and the necessity to control these variables has led to further analytical complications.
With the introduction of the Orion model SS-20 ('Space Stat '-20) this situation could change. A recent advertisement [13] for the analyser stated: 'Operation is simple: the sample is injected, a button is pushed and the result read on the digital display. Now this important blood parameter can be measured throughout the hospital the central laboratory, in satellite laboratories, in intensive care, in pediatrics. Ionized calcium can now join the list of routine clinical tests Indeed, five recent publications have essentially supported this claim [14][15][16][17][18], but the authors' experience has not been so favourable and some shortcomings of the instrument have been found in this evaluation that have not been previously recognized. These findings are factors which should be considered by potential users of the instrument.

Materials and methods
The SS-20 analyser It is not intended to give a detailed description of the analyser, this is available from the manufacturer: Orion Research Inc., 380 Putnam Ave., Cambridge, Mass. 02139, U.S.A. The instrument is basically designed to measure Ca / at 37C in an anaerobic sample of whole blood, plasma or serum. This is performed in three minute automated cycle in which the sample is pumped past a liquid ion-exchange type of calcium electrode (the 'sensor').
The reference electrode is Ag/AgC1 through which a slow flow of 2 mol/1 KC1 saturated with Ag + is pumped, this then meets the sample stream at a liquid junction. A standard (1.00 mmol/1 Ca" /) is pumped past the sensor following the sample. From the difference in potentials, and using a 'slope' factor calibrated in each run using aqueous standards, the instrument automatically calculates and displays the Ca 2/ result. The reagents and standards used were as those supplied by the manufacturer, and the analyser was used in accordance with the manufacturer's instruction manual. Sensor potentials were monitored on a chart recorder throughout the study.
Ionized calcium: spectrophotometric method The method of Varghese [5] was used: This involved preliminary anaerobic ultrafiltration at 37C, followed by dual-wave-length spectrophotometry of tetramethyl murexide added to the ultrafiltrate.

Dialysable calcium method
The method of Ferreira and Bold [19]  Key: The only previous investigation of day-to-day precision of the SS-20 using serum samples [16] demonstrated worse precision than within-day: day-to-day coefficient of variation (CV) ranged from 1.4-3.9 times the corresponding within-day values (the results obtained from this study were 1.7-2.8 times within-day).
In addition a considerably better day-to-day precision was reported using aqueous solution. The authors' results show similar large differences between within-day and day-to-day precision, but a worse overall precision. The within-day precision data obtained agreed with that reported by Fuchs et al, [14]; however theseauthors found no difference in precision using whole blood or plasma samples, whereas the whole blood measurements in this study are considerably less precise. Variable response with sensor age The reason for the poor day-to-day precision is apparent from Figure 1: the sensor apparently varies from day-to-day in its response to serum samples, relative to its response to aqueous standards. These variations are statistically significant: the mean apparent Ca / of four normal subjects on day 4 was lower than day 2 (p (0.01), then fell further on day 5 (p (0.01) and rose again on day 6 (p(0.001), coincident with sensor replacement. (p values calculated using Student's unpaired t-test). Similar patterns have been observed on other occasions using different sensors; generally results on serum or plasma samples fall (relative to aqueous standards) with advancing sensor age. (Rotation of the sensor ( Figure 1) probably 'rejuvenates' it by exposing a fresh area of membrane to the sample).
It was also observed that this variable sensor response did not occur with dialysed serum. A serum pool was dialysed in Visking tubing against a large volume of dialysate containing 140 mmol/1 NaC1, 5 mmol/1 KC1 and 0.7 mmol/1 MgC12; adjusted to pH 7.4 with conc. HC1 and aliquots of different Ca / (approx. 0.5-2.0 mmol/1) were prepared by the addition of appropriate volumes of 100 mmol/1 CaC12. Results   showed significant day-to-day variability. The low control serum used in this study which had been dialysed did not show as much day-to-day variability as the other controls. Serum standardization To minimize day-to-day errors due to the phenomenon described above, a 'serum standardization' procedure was tried and evaluated. Since there is no reference method for Ca determination, results on control sera given by a new sensor on day 6 were arbitrarily chosen as being 'correct' and results on other days were related to them (Figure 2), forming a series of 'calibration lines'. From these lines, results on other samples could be 'corrected' to their expected values on day 6. This resulted in more consistent day-to-day results (Figure 3), also evidenced by (a) the narrowing of a normal reference interval that was established over several days (Group A, Table II), to a similar range that was established on a single day (Group B, Table  II); (b) by the narrowing of apparent intraindividual variation (Table II); and (c) by improved correlation with the spectrophotometric method. (Schwartz has also advocated the use of serum standards with the 'Electrion' system 11 ]). 96 Reference values (Table II) Group A comprised 20 healthy, ambulant laboratory staff (10 male, 10 female) aged 18-50 years. Specimens were collected and analysed on 7 separate days over the course of 5 weeks. One individual from the group was sampled on 9 days over 5 weeks.
Group B comprised 15 healthy, ambulant university students (8 male, 7 female) aged 18-22 years. All samples were collected and analysed on a single day. Both groups were non-fasting, collected mid-morning. None of the distributions differed significantly from Gaussian (Kolmogorov-Smirnov test, p )0.05). No sex difference was observed (Mann-Witney test, p 0.05).
Some of the apparent difference in [H between whole blood and plasma is probably on artefact, since red cells effect the liquid junction of the pH electrode system [20,21]. The    9"/ partly explained by the same day-to-day variability in sensor response that is described here.

Comparison with the spectrophotometric method
There was good correlation between the methods ( Figure 4); the SS-20 results were a mean of 0.021 mmol/1 lower (significant, Wilcoxon signed rank test, p 0.001). If uncorrected SS-20 results were used the correlation coefficient between the methods was worse, at 0.935.

Interferences
Red cells. Results on whole blood were a mean of 0.18 mmol/1 higher than on plasma samples (  [4,12]). The effect is dependent on the haematocrit ( Figure 5). Both Madsen and Olgard [16], and Larsson and Ohman [22] have found that results are higher using whole blood, through neither report so large a difference as reported in this work. Fuchs et al [14] found no difference. There is also a report of higher results on whole blood using a previous Orion electrode [8].
The phenomenon may be explained by changes in the residual liquid junction potential caused by red cells [20,22,23]; also possibly the fact that pH of whole blood is less likely to change in the SS-20 tubing (see below under 'pH stability').

Protein
The effect of protein on the SS-20 was tested by preparing samples with varying protein content, but with essentially the same Ca2+. Each control serum was separated into three fractions using anaerobic ultra-filtration at 37C [5]. The protein-free ultrafiltrate, the native serum, and the protein-rich retentate were then analysed in quadruplicate by the SS-20. The Ca + in each fraction would be expected to be similar since cation retention during ultrafiltration (the Donnan effect) would be roughly counter-balanced by protein retention, causing relative volume change [3,4]. Dialysable  This effect may be explained on the basis of changes in the residual liquid junction potential, and also by volume exclusion effects of colloids. It has not been reported previously. The 'interference' noted by Madsen and Olgard [16] of albumin added to blood, serum or aqueous standard, is simply a reflection of the calcium binding properties of albumin, and not a direct effect of albumin on the analytical method.

Heparin
Other authors have noted that heparin lowers Ca 2+ because of complex formation [10,17] and possibly changes in the residual liquid junction potential [22]. The authors results support the former, since changes in dialysable calcium parallel those of Ca + (Figure 6). At the concentration used in this study the effect is less than 2%, in agreement with Madsen and Olgard [16].

Miscellaneous interference
Interference from Na+, K+, Mg + +, H + and Tris buffer was assessed by incorporating each in 1.00 mmol/1 CaC12 in at least 4 different concentrations. Interference was calculated from regression coefficients of apparent Ca + against concentration of interfering substance (Table III). (The correlation coefficients were significant (p<0.001) except in the case of K+).
Sodium Effect: A slight negative interference found for sodium conflicts with the report of a ma'rked positive effect [16].
Previous Orion electrodes have been noted to show either positive [10] or negligible [6] sodium error.
Magnesium Effect: Results found here are similar to those of Madsen and Olgard [16] but results using previous Orion electrodes were confusing, both positive [6] and negative [7,10] errors being noted.
pH Effect: There has not been a previous study of the effect of pH on SS-20 results. (The experiment of Madsen and Olgard [16] altering the pH of serum simply demonstrates the pH effect of calcium binding to protein, not direct analytical interference). The effect found is somewhat similar to that reported with a previous Orion electrode [9]. Both pH and magnesium interference appear to arise by altering the time course of the sensor response ( Figure 7). It is certainly advisable to use buffered standards: there is a slight

pH stability
The SS-20 is designed to convey samples anaerobically to the sensor in order to avoid pH rise due to loss of CO2 that would in turn cause increased calcium binding to protein, and consequent lowering of Ca +. It was found that anaerobically collected whole but there was no change in pH when the experiment was repeated using polypropylene tubing of similar dimensions. This has a significant effect on results (Table IV), and is probably due to the fact that silicone rubber is permeable to CO2 (in fact it has been used for specifically that purpose, in membranes of CO2 electrodes [24]). The smaller pH change in whole blood is probably because of its greater buffering capacity; this could also partly explain why results on whole blood are higher than plasma, since the pH of plasma would tend to rise more rapidly. Indeed the difference becomes less when polypropylene tubing is substituted. The variability of the reported difference in results between whole blood and plasma, both in this study and others [14,16,22] could be due to variation in the time the sample is exposed to the SS-20 tubing before analysis; i.e. variations in the speed of sample injection and how promptly the analyser is started. The true extent of pH change in the tubing may have been underestimated in this experiment, since the rise in Ca + values after fitting polypropylene tubing was out of proportion to the observed improvement in pH stability. (Approximately 0.05 mmol/l increase in Ca / is expected per 0. pH decrease [4,12]).

Linearity
There is some deviation from linearity at low Ca 2 / (Figure 8).
Examinations of chart recordings of sensor potentials showed that non-linearity could be ascribed to the sensor, not the electronic signal processing. 'Recovery' of calcium added to or removed from serum was also linear (Figure 9).

Error in first sample
The value measured for the first sample analysed tended to be falsely low by as much as 15%. The error was greatest if more than 10 minutes had elapsed since the last sample, being about 2% low at 5 minutes, with no detectable error at 2 minutes. It was noted using both aqueous standards and plasma samples; from chart recordings it appeared to be due to a slow sensor response. Refers to removable tubing between the injection port and sensor, namely the sample inlet tubing and holding coil. Refers to change in pH measured between samples collected (using blood gas capillary collection tubes) from the injection syringe and the 'overflow outlet' of the analyser during an analytical cycle. (i.e. The rise in pH after transversing the analyser tubing for approximately 45 seconds). *** A single sample was collected from a normal subject, and part of it anaerobically separated. All Ca + measurements were in triplicate.