Skip to content
BY 4.0 license Open Access Published by De Gruyter May 5, 2022

(In)direct chloride ISE measurements, room for improvement

  • Jenny E. Kootstra-Ros ORCID logo EMAIL logo , Eline A.E. van der Hagen , Marith van Schrojenstein Lantman ORCID logo , Marc Thelen , Miranda van Berkel and On Behalf of the SKML General Clinical Chemistry Group

To the Editor,

The calculation of the ‘anion gap’ (AG) is common practice in the diagnostic workup of patients with a metabolic acidosis. The AG is mostly calculated by subtracting the anions chloride (Cl) and bicarbonate (HCO3 ) from the cation sodium and aids in differentiating between two types of metabolic acidosis: normal AG acidosis, where a decreased HCO3 concentration is compensated by increased Cl, and high AG acidosis, where the concentration of other anions than HCO3 or Cl is increased [1]. For disease management, calculation of the AG requires accurate measurement of each of its components.

For the determination of Cl multiple state-of-the-art techniques are available. The current reference method to determine Cl is the inductively coupled plasma-isotope dilution mass spectrometry (ICP-IDMS) method [2, 3]. In daily routine practice, Cl is measured with an ion selective electrode (ISE) using a perm-selective membrane to ensure selective transport of Cl. This is either a direct ISE (predominantly in whole blood) or indirect ISE (in plasma/serum).

In the past decades the accuracy of these indirect Cl electrodes has been questioned. Results from the Dutch External Quality Assurance (SKML) programs have shown a systematic negative bias of particularly Roche ISE platforms compared to the ICP-IDMS method [4]. Other reports confirm a non-selectivity of Roche Cl ISE membranes towards HCO3 [5], [6], [7]. A HCO3 concentration-dependent positive bias was observed for Cl measured on the C6000 ISE module [6]. Fortunately, a subsequent call for optimisation of the Roche ISE electrode has led to the launch of an improved Roche ISE calibrator procedure in 2020, where the amount of HCO3 in the calibrator was lowered to more physiological concentrations (Roche customer communication, April 2020).

To re-assess the performance of the Roche Cl electrode compared to other manufacturers we evaluated the accuracy of Cl measurement using EQA data of the first halves of 2019 and 2021 of the SKML schemes for Blood Gas Parameters and Clinical Chemistry in Blood. Samples with the lowest and highest Cl were included, with an additional ‘spy sample’ derived from pooled unprocessed serum. The Cl in blood gas analysis shows variation in bias between different platforms and concentrations of Cl and is most concordant between Siemens and Radiometer (Figure 1B and D). For the Clinical Chemistry scheme, most EQA samples measured on Abbott Architect and Siemens Advia/Dimension show a bias within the tolerance limits of 1.1% (based on the combined uncertainty of the target value of 1.0% and maximally allowed desirable bias of 0.4%) as measured by the reference method ICP-IDMS which is consistently over time (Figure 1A and C). Beckman AU performs suboptimally both in 2021 and in 2019, but the accuracy as assessed by calculating the deviation from the target value, of Cl measurement on the DXC platform has improved in 2021 (Figure 1A and C). The Cl concentrations in EQA samples from 2019 measured by the Roche Cobas ISE platforms display the highest bias compared to the reference value (Figure 1A, left panel mean bias of −5.6% and −4.5% for the C6000 and C8000 platform respectively). Notably, for the ’spy sample’ this negative bias is improved in 2021 for all Roche modules, although biases are still observed for the regular EQA samples with low and high Cl concentration (Figure 1C, mean bias of −5.0% for the C6000 and −3.6% for both the C8000 and pro platforms). One important difference between EQA samples and the ’spy sample’ is the HCO3 concentration, which is ‘normal’ in the latter, and decreased in regular EQA samples due to handling processes. Remarkably, almost 50% of all users of Roche systems apply correction factors to their Cl concentration (38 of 112 in 2019 and 52 of 112 users in 2021; slope on average 1.03 (range 1.00–1.05) and intercept on average 2.2 in 2019 and 2.0 in 2021 (range 0.0–3.0). Although correction improves accuracy of Cl concentration in the ‘spy sample’, the HCO3 -associated negative bias for Cl in the regular EQA samples persists (Figure 1 A, C inset). Taken together, measurement of Cl by direct and indirect methods lacks harmonization and exceeds the desirable bias based on biological variation [8]. Moreover, optimisation in Roche Cl calibration procedures only shows improvement when HCO3 levels are within normal ranges.

Figure 1: 
EQA results (SKML) for Cl− in 2019 and 2021 per instrument type.
EQA results were obtained from the SKML EQA program of years 2019 Q1/Q2 and 2021 Q1/Q2. For each sample in the ‘Clinical Chemistry in blood’ the participants’ results were stratified to analyser type. For blood gas parameters each concentration was measured in both rounds, for Clinical Chemistry each concentration was measured once, except for the ‘spy sample’ which was measured both in Q1 and Q2. Results deviating >3SD from the mean were excluded. Mean bias (%) vs. Reference Method (RM) target values for low (90.3 mmol/L, closed box), high (128.3 mmol/L, open box) and ‘spy sample’ (108.0 mmol/L, grey box), for 2019 (A) and for low (89.3, closed box), high (126.0, open box) and ‘spy sample’ (106.1 mmol/L, grey box), for 2021 (C). Mean bias (%) vs. ALTM for BGA for low (79.1 mmol/L, closed box) and high (118.2 mmol/L, open box) concentrations, for 2019 (B) and for low (77.8 mmol/L, closed box) and high (120.8 mmol/L, open box) concentrations, for 2021 (D). All data are presented as mean bias ± SEM, depicted n-numbers are the number of individual results. Dotted lines represent the combined uncertainty (1.1%) of the target value (1.0%) and maximally allowed desirable bias (0.4%) for A and C and combined uncertainty (0.4%) as derived from interlaboratory SEM (0.2%) and maximally allowed bias (0.4%) for B and D. Data were processed using R Studio and Graphpad Prism 6. Maximally allowed desirable bias is based on a CVi of 1.1% and CVg of 1.3% [8, 11].
Figure 1:

EQA results (SKML) for Cl in 2019 and 2021 per instrument type.

EQA results were obtained from the SKML EQA program of years 2019 Q1/Q2 and 2021 Q1/Q2. For each sample in the ‘Clinical Chemistry in blood’ the participants’ results were stratified to analyser type. For blood gas parameters each concentration was measured in both rounds, for Clinical Chemistry each concentration was measured once, except for the ‘spy sample’ which was measured both in Q1 and Q2. Results deviating >3SD from the mean were excluded. Mean bias (%) vs. Reference Method (RM) target values for low (90.3 mmol/L, closed box), high (128.3 mmol/L, open box) and ‘spy sample’ (108.0 mmol/L, grey box), for 2019 (A) and for low (89.3, closed box), high (126.0, open box) and ‘spy sample’ (106.1 mmol/L, grey box), for 2021 (C). Mean bias (%) vs. ALTM for BGA for low (79.1 mmol/L, closed box) and high (118.2 mmol/L, open box) concentrations, for 2019 (B) and for low (77.8 mmol/L, closed box) and high (120.8 mmol/L, open box) concentrations, for 2021 (D). All data are presented as mean bias ± SEM, depicted n-numbers are the number of individual results. Dotted lines represent the combined uncertainty (1.1%) of the target value (1.0%) and maximally allowed desirable bias (0.4%) for A and C and combined uncertainty (0.4%) as derived from interlaboratory SEM (0.2%) and maximally allowed bias (0.4%) for B and D. Data were processed using R Studio and Graphpad Prism 6. Maximally allowed desirable bias is based on a CVi of 1.1% and CVg of 1.3% [8, 11].

To evaluate the accuracy of Cl measurements in patient samples, an interlaboratory comparison was performed using pooled plasma samples with either a low (11 mmol/L), normal (22 mmol/L) or high (32 mmol/L) HCO3 concentration. Samples were measured on four different routine chemistry platforms (RC) and blood gas analysers (BGA) and compared to the Cl concentrations determined with the ICP-IDMS reference method (Instand, Düsseldorf, Germany). All RC and BGA analysers tested show a HCO3 -dependent bias for Cl from the reference method ranging from −3.8% for the low HCO3 concentrations to +4.8% for samples with high HCO3 -concentrations (Figure 2). The underestimation of Cl is most prominent in low HCO3 samples on the Cobas 8000 platform (mean bias −3.8%). On the other hand, BGA analysis overestimates Cl compared to the reference method in high HCO3 samples (mean bias +4.8% for RAPIDPoint and ABL, Figure 2). However, the relevance of this overestimation is limited for the calculation of the AG in these high HCO3 samples, since Cl concentrations are of particular importance in metabolic acidosis, characterised by a low HCO3 concentration.

Figure 2: 
Cl− deviation from reference laboratory (%) for samples with high, normal and low HCO3
− levels.
LiHep plasma of anonymised residual patient samples with a low, normal and high HCO3
− concentration were individually frozen, pooled into the three sample levels, and frozen until analysis. Samples were measured on four different RC platforms: Cobas 8000, Cobas Pro (both Roche, Mannheim, Germany), Advia XPT1 (Siemens, Erlangen, Germany), and BGA RAPIDPoint 500-I (Siemens) and ABL 90Flex(Plus) (Radiometer, Kopenhagen, Denmark). The reference measurement was performed with the ICP-IDMS reference method (Instand, Düsseldorf, Germany). Uref: expanded uncertainty of the reference method (k=2).
Figure 2:

Cl deviation from reference laboratory (%) for samples with high, normal and low HCO3 levels.

LiHep plasma of anonymised residual patient samples with a low, normal and high HCO3 concentration were individually frozen, pooled into the three sample levels, and frozen until analysis. Samples were measured on four different RC platforms: Cobas 8000, Cobas Pro (both Roche, Mannheim, Germany), Advia XPT1 (Siemens, Erlangen, Germany), and BGA RAPIDPoint 500-I (Siemens) and ABL 90Flex(Plus) (Radiometer, Kopenhagen, Denmark). The reference measurement was performed with the ICP-IDMS reference method (Instand, Düsseldorf, Germany). Uref: expanded uncertainty of the reference method (k=2).

We evaluated the potential effects of the observed biases in Figure 2 on the clinical decision-making associated with the diagnosis of a metabolic acidosis by in-silico modulation of patient results with high AG acidosis. Therefore, we obtained BGA Cl, sodium, and HCO3 patient results (RAPIDPoint) from the laboratory information system of a tertiary centre, including samples with a pH of <7.35 and HCO3 <20 mmol/L. We (re-)calculated the AG in three ways: (1) using the ‘true’ Cl value (reference-Cl--AG), (2) using the BGA-Cl- value of a RAPIDPoint analyser (BGA-Cl--AG), and (3) using the Cl value as obtained by an RC Cobas 8000 (RC-Cl--AG). For correction, the bias of the most appropriate bicarbonate pool per patient was used (e.g. patient bicarbonate levels of 16 mmol/L and lower were corrected with the bias found in the 11 mmol/L pool, whereas samples above this threshold were corrected with the 22 mmol/L pools) as shown in Figure 2 (−3.8% and −3.4% for RC and +1.5% and + 1.9% for BGA). Subsequently, the proportion of patients above the cut-off value of 12 mmol/L were assessed. The dataset consisted of 2274 patient samples, of which 56% would have a high AG acidosis when using the reference-Cl--AG. With the BGA-Cl--AG method 43% of the patients would have an AG above the cut-off, which is within the uncertainty range of the reference-Cl--AG considering the measurement uncertainty of 1.7 mmol/L [9, 10]. Notably, when AG was calculated with the RC-Cl--AG as much as 87% of the results would be above the cut-off value of 12 mmol/L. These results indicate a similar referral rate for BGA and reference-based assessment of Cl for the determination of the AG, but the RC-Cl--AG overestimates the AG due to a falsely decreased Cl value.

Correction of the RC-Cl--AG with the average reported slope and intercept as mentioned above resulted in 46% of results above the cut-off value. These results indicate that the average used correction factor is sufficient in achieving positive rates similar to reference- and BGA-derived Cl values. However, the general use of a correction factor could negatively affect patient populations with a high HCO3 concentration, due to the HCO3 -dependent nature of the interference.

Overall, our study shows that despite efforts to improve accuracy of Cl measurement by indirect ISE on Roche platforms, a HCO3 -dependent negative bias still exists. Furthermore, our results indicate that BGA also suffer from a HCO3 -interference albeit only present at the higher HCO3 concentrations.

With this study, we aim to create awareness of the clinically relevant bias – both negative and positive – that is observed in state-of-the-art ISE Cl measurements. Standardisation of Cl measurement is technically feasible since the traceability chain to a higher order measurement system with reference material is available. Nevertheless, methods subject to concentration-dependent interference due to non-selectivity cannot be harmonized or corrected. This study shows that about 50% of all Roche platform users apply a factor to compensate for the negative bias. However, proper correction would require sample-specific, interferent concentration-dependent correction factors.

We advise not to use the different methods interchangeably, since the interpretation of the AG will introduce diagnostic error. Also, we dispute the introduction of a factor to correct for this, since the magnitude of interference is not constant. Hence, we strongly urge the manufacturers to improve ISE Cl measurements and diminish non-selectivity to HCO3 , as technical feasibility has been proven by others.


Corresponding author: Jenny E. Kootstra-Ros, Department of Laboratory Medicine, University Medical Center Groningen, Groningen, The Netherlands, Phone: +31 50 3610530, E-mail:

Members of the SKML General Clinical Chemistry Group: R.J.A.C. de Beer, Laboratory of Medical Diagnostics, Rivierenland Hospital, Tiel, The Netherlands; D. L. Bakkeren, Department of Clinical Chemistry, Máxima Medical Center, Veldhoven, The Netherlands; M. van Berkel, Department of Laboratory Medicine, Radboud University Hospital, Nijmegen, The Netherlands; E.A.E. van der Hagen, MCA Laboratory, Queen Beatrix Hospital, Winterswijk, The Netherlands; A.A. van der Horst, Laboratory of Clinical Chemistry and Hematology, Jeroen Bosch Hospital, Den Bosch, The Netherlands; J.E. Kootstra-Ros, Department of Laboratory Medicine, University Medical Center Groningen, Groningen, The Netherlands; C. Ramakers, Department of Clinical Chemistry, Erasmus MC, Rotterdam, The Netherlands; M. van Schrojenstein Lantman, Department of Laboratory Medicine, Radboud University Hospital, Nijmegen, The Netherlands and SKML, Nijmegen, The Netherlands; S. Waaijers, Department of Laboratory Medicine, Radboudumc, Nijmegen, The Netherlands.

Jenny Kootstra-Ros and Eline van der Hagen contributed equally to this work.
  1. Research funding: None declared.

  2. Author contributions: All authors have accepted responsibility for the entire content of this manuscript and approved its submission.

  3. Competing interests: Authors state no conflict of interest.

  4. Informed consent: Not applicable.

  5. Ethical approval: Not applicable.

References

1. Gabow, PA, Kaehny, WD, Fennessey, PV, Goodman, SI, Gross, PA, Schrier, RW. Diagnostic importance of an increased serum anion gap. N Engl J Med 1980;303:854–8. https://doi.org/10.1056/nejm198010093031505.Search in Google Scholar

2. Kramer, U, Kress, M, Reinauer, H, Spannagl, M, Kaiser, P. Candidate reference measurement procedures for chloride, potassium, sodium, calcium, magnesium, and lithium by inductively coupled plasma (isotope dilution) sector field mass spectrometry (ICP-(ID) SFMS) in serum. Clin Lab 2013;59:1017–29. https://doi.org/10.7754/clin.lab.2012.120902.Search in Google Scholar PubMed

3. Joint Committee for Traceability in Laboratory Medicine. Database of higher-order reference materials, measurement methods/procedures and services. Available from: https://www.bipm.org/jctlm/ [Accessed 7 Dec 2021].Search in Google Scholar

4. Geerts, N, Wlazlo, N, Scharnhorst, V. Between analyser differences in chloride measurements and thus anion gap cause different interpretations of the acid-base balance. Clin Chem Lab Med 2016;54:e81. https://doi.org/10.1515/cclm-2015-0477.Search in Google Scholar PubMed

5. Schiemsky, T, Brandt, I. Bicarbonate interference on cobas 6000 c501 chloride ion-selective electrodes. Clin Chem Lab Med 2018;56:e214–5. https://doi.org/10.1515/cclm-2018-0004.Search in Google Scholar PubMed

6. Dimeski, G, Badrick, T, John, AS. Ion selective electrodes (ISEs) and interferences-a review. Clin Chim Acta Int J Clin Chem 2010;411:309–17. https://doi.org/10.1016/j.cca.2009.12.005.Search in Google Scholar PubMed

7. Bihari, S, Galluccio, S. Discrepancy in chloride measurement with decreasing bicarbonate concentrations. Clin Lab 2019;65:190121. https://doi.org/10.7754/Clin.Lab.2019.190121.Search in Google Scholar PubMed

8. Aarsand, AK, Fernandez-Calle, P, Webster, C, Coskun, A, Gonzales-Lao, E, Diaz-Garzon, J, et al.. The EFLM biological variation database. Available from: https://biologicalvariation.eu/ [Accessed 1 Dec 2021].Search in Google Scholar

9. International Organization for Standardization. Medical laboratories — Practical guidance for the estimation of measurement uncertainty. ISO/TS 20914:2019; 2019.Search in Google Scholar

10. Rigo-Bonnin, R, Canalias, F. Measurement uncertainty estimation for derived biological quantities. Clin Chem Lab Med 2021;59:e1–7. https://doi.org/10.1515/cclm-2020-1003.Search in Google Scholar PubMed

11. Fraser, CG, Petersen, PH, Libeer, J-C, Ricos, C. Proposals for setting generally applicable quality goals solely based on biology. Ann Clin Biochem 1997;34:8–12. https://doi.org/10.1177/000456329703400103.Search in Google Scholar PubMed

Received: 2022-03-09
Revised: 2022-04-06
Accepted: 2022-04-13
Published Online: 2022-05-05
Published in Print: 2022-06-27

© 2022 Jenny E. Kootstra-Ros et al., published by De Gruyter, Berlin/Boston

This work is licensed under the Creative Commons Attribution 4.0 International License.

Downloaded on 28.3.2024 from https://www.degruyter.com/document/doi/10.1515/cclm-2022-0220/html
Scroll to top button