Discrepancies Between Point of Care and Central Laboratory Sodium and Potassium Measurements in ICU: Analytical Biases and Physician Awareness

Abstract

OBJECTIVES:

Blood electrolyte panels, especially sodium ion (Na⁺) and potassium ion (K⁺), are fundamental and routine tests. In critical care settings, these measurements are typically performed by central laboratories using indirect potentiometry (IP) or via point-of-care blood gas analyzers using direct potentiometry (DP). Discrepancies between the two methods exist, and clinicians often have limited knowledge regarding the strengths and weaknesses of each technique. We wanted to quantify these observations.

DESIGN:

We compared prospectively Na⁺ and K⁺ measurements obtained from the same blood draw using both IP and DP (including analyses of correlation, agreement, and discrepancies). We also conducted a brief survey among physicians to assess their knowledge regarding IP and DP.

SETTINGS:

A tertiary hospital.

PATIENTS:

A total of 501 paired measurements were prospectively collected from an ICU population.

INTERVENTIONS:

None.

MEASUREMENTS AND MAIN RESULTS:

We assessed the degree of agreement between the two methods. We also examined the impact of proteinemia and hemolysis on Na⁺ and K⁺ values, respectively. Only 31.1% of the 103 responding physicians were aware of the analytical bias in Na⁺ measurements obtained by IP, and 45.6% considered K⁺ measurements from DP to be as reliable or more reliable than those obtained by IP. The agreement between the two methods was moderate (Lin’s concordance correlation coefficient: 0.90 for Na⁺ and 0.93 for K⁺). The 95% limit of agreement for Na⁺ was particularly large (10.48 mmol/L). The divergence between the two methods, defined as a discrepancy in classification (within, below, or above the normal range), occurred in approximately 10% of cases, for both Na⁺ and K⁺.

CONCLUSIONS:

IP and DP were not interchangeable, each exhibiting distinct strengths and limitations. Enhancing physician awareness of the differences between these methods could improve the quality of care.

KEY POINTS.

Question: In ICU patients, how well do sodium ion (Na⁺) and potassium ion (K⁺) measurements obtained by indirect potentiometry (IP, central laboratory) agree with those obtained by direct potentiometry (DP, point-of-care blood gas analyzers), and what are the main causes of discrepancies?

Findings: In this observational study of 501 paired samples from 144 ICU patients, IP and DP showed strong correlation but only moderate agreement (Lin’s concordance correlation coefficient, 0.90 for Na⁺ and 0.93 for K⁺), with clinically relevant discrepancies in 12.1% of Na⁺ and 10.1% of K⁺ measurements.

Meaning: In critically ill patients, IP and DP are not interchangeable; clinicians should preferentially use DP for sodium measurement.

Over the years, technological advances have allowed automated devices to move beyond central laboratories and take on a more prominent role in clinical wards, particularly in operating rooms, emergency departments (EDs), and ICUs.

Among the most widely used point-of-care (POC) devices in the hospital setting are blood gas analyzers (BGA), which have undergone significant improvements over time. These developments have expanded their functionality to measure crucial additional analytes in critical situations, including lactate, sodium ion (Na⁺), potassium ion (K⁺), calcium ion, and hemoglobin, alongside core parameters (blood gas and pH).

Because POC instruments are often operated by personnel without the technical qualifications of laboratory staff—and because rapid results are required—the analytical methods used in POC testing differ from those employed in central laboratories. As a result, these methods are frequently (though not always justifiably) viewed as less robust and are therefore not considered reference methods.

For Na⁺ and K⁺ determination, notable differences exist between the “traditional” (routine) method used in the central laboratory (typically involving heavy, high-throughput analyzers) and those used in POC devices. In the routine laboratory method, Na⁺ and K⁺ are measured using indirect potentiometry (IP) with ion-selective electrodes (ISEs) on plasma or serum, following a dilution step, after a preliminary centrifugation. In contrast, POC devices also use ISE, but apply direct potentiometry (DP) on whole blood, without prior dilution or centrifugation (1). Several issues are then associated with each method, as summarized in Figure 1:

Figure 1.

Summary of key differences—advantages and limitations—between direct potentiometry (DP) and indirect potentiometry (IP). Cost estimates are based on our specific setting (see Methods); electrolyte measurements using our point-of-care (POC) blood gas analyzer are inseparable from gas analyses, as both are performed together mandatorily. Cl^– = chloride ion, HCO₃⁻ = bicarbonate ion, K⁺ = potassium ion, Na⁺ = sodium ion.

• Due to the dilution performed with IP, an analytical bias may occur in case of significant hyperproteinemia (or lipidemia) and hypoproteinemia, which can lead to an under- or over-estimation of Na⁺ concentration, respectively (pseudohyponatremia or pseudohypernatremia) (2), linked to the “electrolyte exclusion effect” (3). As no dilution is performed when using DP on whole blood, the Na⁺ measurement is virtually not impacted by an excess (or a deficit) in total protein/lipid in the sample (2);

• On the other hand, since the whole blood is used as substrate in POC BGA (DP), it is impossible to check the sample for hemolysis (by examining the aspect of the plasma or serum), unless an additional device or a last-generation BGA is used. This could have consequences when dealing with K⁺ measurement: in case of undetected significant pre-analytical hemolysis, the kalemia will be overestimated (pseudohyperkalemia) (4).

Clinical consequences are well documented:

• Erroneous diagnosis and inappropriate treatments can occur in cases of pseudohyponatremia caused by hyperlipidemia or hyperproteinemia when using the reference method (IP). This is particularly problematic if this interference is overlooked, sometimes leading to fatal outcomes (5, 6);

• Similarly, in the case of a hemolyzed sample and blind reliance on POC results, there is a risk of overtreating nonexistent hyperkalemia or failing to detect significant hypokalemia that appears artificially normal due to unrecognized in vitro hemolysis (7).

Aside from pre-analytical errors that play a major role in involuntary “falsifying” biological reality (8–11), the above-mentioned brief examples underscore the relative complexity hidden behind these “simple” routine and frequent assays. Additionally, the scarcity of research in this field (only six studies and two review articles published in critical care journals over the past 25 years, of 61 publications retrieved; see Table SM-5 [https://links.lww.com/CCM/H891] Bibliography study), the infrequent quantitative analysis of the causes of discrepancies between the two methods (IP and DP), and the fact that the dilution factor performed in IP is analyzer‑dependent (12) (and therefore varies from one study to another) all further justify the need for the present study.

In this work, we investigated the correlation between Na⁺ and K⁺ measurements obtained by IP and DP (501 pairs of measures from the same blood draw), and analyzed the agreement between the two methods. We then investigated cases with discordant values to identify potential analytical causes, focusing on hemolysis and marked hypoproteinemia or hyperproteinemia. We also conducted a brief survey among physicians who use POC electrolyte panels in their routine practice, to get a quick overview of their current knowledge regarding the strengths and weaknesses of each method (Supplementary Materials, Tables SM-1 and SM-2 in the Survey section, https://links.lww.com/CCM/H891).

METHODS

Patients

From June 2024 to October 2024, biological results from patients admitted to a 30-bed ICU of a tertiary hospital (Hôpital NOVO, Pontoise, France) who underwent both a blood gas analysis and a plasma electrolyte panel from the same blood draw as part of routine care were included in the study.

Blood was collected either from a peripheral vein/artery or through an arterial catheter, with at least one heparinized tube for plasma electrolyte panel (IP) and one syringe for BGA with electrolyte panel (DP). Blood draws from central or peripheral catheters were not performed, as they are strictly prohibited in our ward. Hematology patients with highly blastic leukemia (> 40 G/L) were excluded from the study, as this condition can cause high spontaneous levels of hemolysis—resulting in pseudohyperkalemia with both IP and DP—if prompt analysis is not performed.

This study was performed under the status of the “internal research” framework and did not require ethical approval or informed consent, in accordance with French regulations.

Bioassays

Electrolyte panel (plasma sample) by IP was performed on a high-throughput analyzer (Alinity Ci, Abbott, IL) after centrifugation of heparinized blood (BD Vacutainer, Lithium Heparin 4 mL tubes; Becton Dickinson, Franklin Lakes, NJ). On this analyzer, the dilution factor applied to the plasma before ISE measurement is 1:24. Electrolyte panel by DP was performed on an ABL90 FLEX BGA (Radiometer, Copenhagen, Denmark) on heparinized whole blood (safePICO Self-fill syringe; Radiometer). Hemolysis level was assessed through the H-index provided by the Alinity analyzer (see Supplementary Materials, https://links.lww.com/CCM/H891).

Cutoffs Used in This Study: Definition and Justifications

See Supplementary Materials (https://links.lww.com/CCM/H891).

Statistical Analysis and Cutoffs

The data was checked for normality and expressed accordingly: in the case of normal distribution, results are expressed as mean ± sd; if not, as median with interquartile range with 25th and 75th percentiles (IQR_25–75). Categorical variables were analyzed using Fisher exact test to compare proportions in subpopulations. Bland and Altman (13) and Lin (14) approaches were used for agreement analysis (15).

Statistical analyses were performed using GraphPad Prism, Version 7.03 (GraphPad Software, San Diego, CA). A p value of less than 0.05 was considered statistically significant. Lin’s concordance correlation coefficient (Lin’s CCC) was calculated using the Real Statistics Resource Pack software (Release 8.9.1, copyright 2013–2023) created by Charles Zaiontz (Real Statistics, Trento, Italy; www.real-statistics.com) on Microsoft Excel (Microsoft, Redmond, WA).

RESULTS

Included Population and Dataset

The results emanating from a total of 501 blood collection procedures corresponding to 144 patients were included in the final dataset, corresponding to an average of 3.5 blood sample collections per patient. Demographic data and medical conditions of the population are summarized in Table 1. Most sampling procedures were performed through an arterial catheter (60.9%), reducing the risk of hemolysis associated with difficult venous or arterial access.

TABLE 1.

Main Characteristics of the Population and Origin of Blood Samples

Parameters	n (%)	Mean (Minimum–Maximum Values) ± sd	Median (IQR)
Number of patients	144
Sex ratio (male/female)	85/59 (59/41)
Age		59 (17–84) ±16.2	61 (51–72)
Type of critical care stay
ICU	115 (79.9)
Intermediate care unit	29 (20.1)
Type of admission
Medical	121 (84.0)
Scheduled surgical	10 (7.0)
Unscheduled surgical	13 (9.0)
Simplified Acute Physiology Score II		43 (8–111) ±17	41 (31.7–53.2)
Number of organ failures		1.4 (0–6) ±1.2	1
Organ failure at admission
Circulatory (need for catecholamine)	44 (30.6)
Respiratory (Pao2/Fio2 < 200)	56 (38.9)
Renal (Kidney Disease: Improving Global Outcomes 3)	43 (29.9)
Cardiac (ejection fraction < 30%)	5 (3,5)
Hematologic (platelets < 100 G/L)	11 (7.6)
Liver (factor V < 50%)	7 (4.9)
Neurologic (Glasgow Coma Scale < 7)	24 (16.7)
Metabolic (lactate > 3 mmol/L)	24 (16.7)
Mechanical ventilation	64 (44.4)
Blood samples	501
Blood obtained from an arterial catheter	305 (60.9)
Arterial blood (with and without catheter)	468 (93.4)
Venous blood	33 (6.6)

IQR = interquartile range.

Data completeness reached 98.6–100%, depending on the parameter, as summarized in Table SM-3 (https://links.lww.com/CCM/H891).

Distribution of Na⁺ and K⁺ Concentration According to Direct or Indirect Potentiometry and Correlation

The distributions of these values are shown in Figure 2. Na⁺ values differed significantly between the two methods, although the difference was not clinically relevant (median [IQR] values were 138 mmol/L [133–141 mmol/L] for Na⁺_DP and 137 mmol/L [133–140 mmol/L] for Na⁺_IP; p < 0.0001; Wilcoxon matched pairs test). For K⁺, a difference was also observed between methods despite nearly identical overall distributions (4.0 mmol/L [3.6–4.4 mmol/L] for K⁺_DP and 4.0 mmol/L [3.6–4.5 mmol/L] for K⁺_IP; p < 0.0001).

Figure 2.

Distribution of sodium ion (Na⁺) and potassium ion (K⁺) values according to each method. The points represent individual values, violin plots illustrate the probability density of these values, and box plots represent the 25th percentile, median, and 75th percentile while boundaries of the whiskers are based on the 1.5 interquartile range value above or below these percentiles.

Correlation between the two methods was strong for both electrolytes: the Spearman correlation coefficient was 0.936 (95% CI: 0.924–0.946) for Na⁺ and 0.944 (95% CI: 0.933–0.953) for K⁺ (Fig. SM-1, https://links.lww.com/CCM/H891).

Agreement Between the Two Methods

A total of 501 paired Na⁺ assays and 498 paired K⁺ assays were obtained by IP and DP. Bland and Altman’s analysis showed a moderate concordance for Na⁺ determination (Table 2) with a very large 95% limit of agreement (width of the interval: 10.48 mmol/L, i.e., ≈100% of the normality range width for this ion). Four points were outside the limits of agreement (Fig. 3). Lin’s CCC was 0.900 (95% CI, 0.882–0.915), consistent with the previous analysis, suggesting a moderate concordance between IP and DP.

TABLE 2.

Median and Distribution Indices of Sodium and Potassium Ions Measurements by Direct and Indirect Potentiometry, With Agreement Analysis Between the Two Methods

Parameters	DP	IP	p
Sodium ion
Median (IQR25–75), mmol/L	138 (133–141)	137 (133–140)	a
Minimum and maximum values, mmol/L	102–154	104–150
Agreement analysis
Bland-Altman (DP–IP)
Mean bias, mmol/L	0.85
Limit of agreement (95%), mmol/L	–4.38 to 6.10
Width of the agreement interval, mmol/L	10.48
Measures outside the agreement interval (n)	4
Lin’s CCC (95% CI)	0.90 (0.88–0.91)
Potassium ion
Median (IQR25–75), mmol/L	4.0 (3.6–4.4)	4.0 (3.6–4.5)	a
Minimum and maximum values, mmol/L	2.5–7.6	2.4–10.0
Agreement analysis
Bland-Altman (IP–DP)
Mean bias, mmol/L	0.07
Limit of agreement (95%), mmol/L	–0.43 to 0.58
Width of the agreement interval, mmol/L	1.01
Measures outside the agreement interval (n)	21
Lin’s CCC (95% CI)	0.92 (0.91–0.93)

DP = direct potentiometry (on whole blood, blood gas analyzer), IP = indirect potentiometry (plasma electrolyte panel), IQR_25–75 = interquartile range with 25th and 75th percentiles, Lin’s CCC = Lin’s concordance correlation coefficient.

^ap < 0.0001 (Wilcoxon matched paired test).

Figure 3.

Bland-Altman plots for sodium ion (Na⁺) and potassium ion (K⁺) assays with limits of agreement of the two methods. The dashed line represents the mean bias, and plain lines represent the 95% limits of agreement. Red dots correspond to measures outside this interval of agreement. DP = direct potentiometry, IP = indirect potentiometry.

For K⁺, the agreement was slightly better (Table 2) with a narrower interval (1.00 mmol/L, i.e., ≈66% of the normality range width) but with a total of 21 points outside the defined 95% limits of agreement. Lin’s CCC was 0.928 (95% CI, 0.916–0.939), indicating also a moderate concordance.

Situations of Disagreement Between the Two Methods Were Not Rare, With an Impact of Proteinemia and Hemolysis

To determine discrepancies between IP and DP, we performed two distinct analyses:

• Method A: We first focused on situations where the absolute difference between the two methods was greater than or equal to 4 mmol/L for Na⁺ and greater than or equal to 0.4 mmol/L for K⁺ (Supplementary Materials, https://links.lww.com/CCM/H891). This analysis is more metrology-oriented than clinically relevant, as a couple of values could be within the normal range (e.g., 135–145 mmol/L for Na⁺) despite having an unacceptable difference (e.g., an 8 mmol/L difference if Na⁺ = 136 by IP and 144 by DP).

• Method B: In a clinically oriented perspective, we analyzed abnormal Na⁺ values on DP (considered the most reliable method for this ion) and abnormal K⁺ values obtained on plasma by IP (considered the reference method for this ion). We compared them to their corresponding values obtained by IP and DP, respectively. Abnormal values were defined as Na⁺ greater than 145 or less than 135 and as K⁺ greater than 5.0 or less than 3.5 mmol/L.

According to method A, a total of 35 couples of measures presented an absolute difference greater than or equal to 4 mmol/L regarding Na⁺ (6.9%), ranging from 4 to 35 mmol/L. This number reached 44 (8.8%) when analyzing K⁺ with the predefined cutoff (≥ 0.4 mmol/L), with absolute differences ranging from 0.4 to 2.6 mmol/L.

According to method B, as summarized in Table 3, 61 situations of divergence (12.1%) were identified where Na⁺ values were either normal with DP but reported as abnormal with IP (n = 24) or abnormal with DP but reported as normal with IP (n = 37). We calculated the positive and negative predictive values (PPV and NPV) for Na⁺ obtained by IP, with DP as reference method: for the diagnosis of hyponatremia, PPV was 84.7% and NPV was 96.8%. For the diagnosis of hypernatremia, PPV was excellent reaching at 100% and NPV was 94.7% (see contingency Table SM-4, https://links.lww.com/CCM/H891), ascribable to the high prevalence of hypoproteinemia in this population (see below).

TABLE 3.

Analysis of Disagreement Between Direct and Indirect Potentiometry With Respect to Established Standards (Normal, Increased, or Decreased Concentrations of the Ion Under Consideration)

Na+	Measures by DP (Regarded As the Reference Method)	Corresponding Measures Obtained by IP
	n (%)	n (%, Relative to the Reference Method)
Total available samples	501 (100)	Normal	Na+ < 135	Na+ > 145
With normal values	322 (64.3)	298 (92.5)	24 (7.5)	0
With abnormal values	179 (35.7)
Na+< 135	144 (80.4)	11 (7.6)	133 (92.4)	0
Na+ > 145	35 (19.6)	26 (74.3)	0	9 (25.7)
Situations of disagreement	61 (12.2)

K+	Measures by IP (Regarded As the Reference Method)	Corresponding Measures Obtained by DP
	n (%)	n (%, Relative to the Reference Method)
Total available samples	497 (100)	Normal	K+ < 3.5	K+ > 5.0
With normal values	366 (73.6)	346 (94.5)	18 (5.0)	2 (0.5)
With abnormal values	131 (26.4)
K+< 3.5	83 (63.4)	15 (18.1)	68 (81.9)	0
K+ > 5.0	48 (36.6)	15 (31.2)	0	33 (68.8)
Situations of disagreement	50 (10.1)

DP = direct potentiometry, IP = indirect potentiometry, K⁺ = potassium ion, Na⁺ = sodium ion.

When focusing on K⁺, divergence was present in 50 situations (10.1%) where K⁺ values were either normal with IP but reported as abnormal with DP (n = 20) or abnormal with IP but reported as normal with DP (n = 30). We did not calculate PPV and NPV, as hemolysis was not assessed on samples submitted to DP (only in those submitted to IP, thanks to H-index).

When analyzing the impact of proteinemia on Na⁺ measurements using IP, we confirmed the significant bias between the two methods, with a difference depending on proteinemia (Fig. SM-2, https://links.lww.com/CCM/H891): the higher the protein concentration, the greater the difference observed between DP and IP (Na⁺_DP–Na⁺_IP; p < 0.0001; analysis of variance), confirming that increase in proteinemia was associated with an underestimation of Na⁺ by IP. Conversely, a decrease in proteinemia was associated with an overestimated Na⁺ by IP, with a strong practical impact as hypoproteinemia (total protein < 60 g/L) was frequent in our ICU setting, concerning 52.7% of the samples. Severe hypoproteinemia (defined as total protein < 45 g/L) reached 4.6% (23 samples).

Regarding the impact of hemolysis on K⁺ values, 33 samples of 501 (6.5%) had a H-index greater than or equal to 0.5. In 31 cases, K⁺ was available both in DP and IP: 87% of these paired measures showed a higher value of K⁺ when assayed with IP compared with DP (on these 31 hemolyzed samples, mean difference between K⁺_IP and K⁺_DP: 0.44 mmol/L, minimum –0.20, maximum 2.60; p < 0.0001) suggesting a greater impact of hemolysis on samples submitted to IP than in syringes submitted to DP. In addition, when focusing on situations where K⁺_DP was greater than K⁺_IP (with a difference ≥ 0.3 mmol/L), 25 samples of 501 were concerned (5%) with a mean difference of 0.42 mmol/L (minimum 0.3, maximum 1.1), but all these samples were negative for hemolysis (H-index < 0.5).

When comparing venous (n = 33) to arterial samples (n = 468), we did not observe any difference in terms of hemolysis (median H-index: 0.03 [0.01–0.09] and 0.04 [0.01–0.10], respectively; p = 0.81) or in terms of discrepancy on K⁺ measurements (median ΔK⁺_DP–IP: 0 [–0.1 to 0.0] and –0.1 [–0.2 to 0], respectively; p = 0.33).

DISCUSSION

Our study primarily identifies three key points: 1) in the critical care setting, the two methods demonstrated an imperfect concordance, with discrepancies between IP and DP that could be clinically significant—potentially leading to errors in therapeutic decision-making; 2) the high prevalence of hypoproteinemia in this ICU population (52.7%) reduced the PPV of IP method (84.7%) when considering the diagnosis of hyponatremia; and 3) most physicians were unaware of the analytical strengths and limitations of each method, as well as of common sources of interference for Na⁺ measurement with IP, and the risk of pseudohyperkalemia with DP (Supplementary Materials, https://links.lww.com/CCM/H891).

Regarding hemolysis and K⁺ measurement, 6.5% of our samples had an H-index greater than 0.5, a rate consistent with data available in the literature (reported to range up to 18.1% in arterial blood gas samples and 8.8% in venous samples [16–19]). Interestingly, in situations where hemolysis was present, we consistently observed that K⁺ was higher with IP than with DP, suggesting prohemolytic factors that are more often present when blood samples were processed for IP as compared with DP. This observation could be explained by two factors: first, the delay between blood collection and analysis is longer with IP (due to the pre-analytical requirement of a centrifugation step and because electrolyte panels are not routinely processed in vital emergency, even if ICU samples are prioritized). Second, the mechanical impact of the centrifugation step, known to promote hemolysis (20), is absent when dealing with DP, performed on whole blood (by definition). Today, there is no routine procedure to detect hemolysis on whole blood samples, even though it could be considered a good laboratory practice and is included in International Organization for Standardization 15189:2022 stating that specified requirements of an examination method may include interfering substances (21). Regarding Na⁺ discrepancies between IP and DP in the particular setting of ICU (where prevalence of hypoproteinemia is high), our results confirm the limited data available on this topic (22–27). The quantitative impact observed was within the same order of magnitude, except for the study by Banerjee and Mehrotra (28), who observed an unusual and debatable mean bias of 9 mmol/L for Na⁺ between IP and DP. In the study by Langelaan et al (25), Na⁺ measured by IP and DP was compared in blood samples from ICU patients: pseudohypernatremia occurred in 32%. The link between IP/DP discrepancy for Na⁺ measurements and plasma water content (which depends on lipids and proteins concentration) was pointed out, leading the authors to propose a post-analytical correction factor. This reduced the prevalence of pseudohypernatremia to 19% in their population. In this regard, Goldwasser et al (24) propose a correction of Na⁺ obtained by IP with proteinemia: applied to our setting, this simple correction works well in situations of hypoproteinemia, but failed to fully compensate for the most severe hyperproteinemia cases observed in our work (e.g., highest proteinemia 111 g/L, corresponding Na⁺_IP: 132 mmol/L and Na⁺_DP: 144 mmol/L, corrected Na⁺ according to Goldwasser et al [24]: 135 mmol/L).

Another factor complicates physicians’ decision-making: an authority bias. In our team and in follow-up discussions with some of the survey respondents, when an abnormal Na⁺ value is provided by a POC device, it could be falsely considered unreliable: the value of Na⁺ provided by the central laboratory (IP) is paradoxically judged more reliable. While the superiority of IP for Na⁺ measurement is factually erroneous, especially in the case of severe hypoproteinemia or hyperproteinemia or hyperlipidemia, this “authority bias” among physicians (because the result emanates from the “authorized” laboratory) can have severe consequences. As a concrete example, one patient in this study was admitted to the ED for headaches, asthenia, and blurred vision with dyspnea (night shift admission). Na⁺ was 123 mmol/L (IP) and this hyponatremia was considered responsible for the symptoms: a limitation of water intake was started with a supposed diagnosis of syndrome of inappropriate antidiuretic hormone secretion. The proteinemia was not seen at this time (delayed result due to control) and symptoms worsened. Finally, the patient was referred to ICU 36 hours later for hyperviscosity syndrome related to hyperproteinemia (123 g/L) due to Waldenstrom macroglobulinemia (immunoglobulin M: 76 g/L). Na⁺ was controlled by using DP and found to be within the normal range; however, the ICU physician initially questioned the result and requested a repeat test. The patient was subsequently given aggressive hydration and underwent plasma exchange therapy, which resolved all clinical symptoms.

Our study has some limitations. As this work was purely observational, lipid concentrations were not systematically collected, precluding a precise assessment of potential interference on Na⁺ measurements by IP. Nonetheless, L-index was rarely high, suggesting that even though patients were sedated with propofol (a lipid emulsion) or sometimes received parenteral feeding, this factor did not play a major role in Na⁺ reliability by IP in our clinical setting.

Second, the fact that a patient could contribute multiple times to the study by providing several pairs of blood samples (an average of 3.5 pairs per patient) may have introduced a bias of overrepresentation.

Third, another limitation is the assumption we made regarding hemolysis: when hemolysis was present in the plasma (IP), we considered its coexistence in whole blood (DP). We made this assumption because we were not technically able to assess the hemolysis level on the whole blood sample, submitted to DP. While results showed that, paradoxically, K⁺ values obtained using DP were lower than those retrieved using IP when hemolysis was present (suggesting a lower level of hemolysis in whole blood samples submitted to DP, see above), we cannot rule out and objectify situations where hemolysis was present in whole blood samples and absent in plasma samples.

Finally, due to software limitations and timestamp inaccuracy, we were unable to reliably analyze the time elapsed between sample collection at the patient’s bedside, receipt by the laboratory, the start of centrifugation, and the actual analysis. Although unlikely, we may have encountered inconsistent delays, which could explain the higher hemolysis in samples subjected to IP.

CONCLUSIONS

In conclusion, the key pragmatic question for critical care physicians becomes: Which method should I use in daily practice? A tempting yet oversimplified answer might be to rely on DP via POC testing for Na⁺, and IP through central laboratory analysis for K⁺. However, several important factors must be taken into account—including the significantly higher cost of POC testing (up to ten times that of IP in our setting) and the urgency related to turnaround time—along with the following considerations. Although IP remains the most commonly used technique for measuring Na⁺ and K⁺ due to its low cost, high throughput, and integrated hemolysis detection capabilities (29), it is not regarded as the reference analytical method for Na⁺. More than 2 decades ago, the International Federation of Clinical Chemistry and Laboratory Medicine recommended DP as the more accurate and clinically appropriate method, especially in patients with abnormal lipid or protein concentrations (30). In line with prior recommendations from various authors (23, 25, 31), our findings support the preferential use of DP (whole blood) for Na⁺ measurement in critically ill patients.

However, to achieve its full potential and provide trustworthy K⁺ measures, the POC electrolyte panel by DP must solve the problem of pre-analytical hemolysis detection. Recent progress is notable in this way: to replace fastidious pre-centrifugation and plasma examination for hemolysis (not applicable in the POC setting), hemolysis detection can now be performed by using a dedicated, semiautomated, and user-friendly device, from whole blood, before performing the BGA-associated electrolyte panel (19). Even better, this task can be performed in-line with a new, disruptive, and fully integrated technology, as demonstrated by Balasubramanian et al (32), using a next-generation BGA. In this device, plasma is locally separated from blood cells using acoustic forces in an acoustofluidic flow cell, allowing photometric assessment of hemolysis in this spatially delimited window, without the need for blood centrifugation.

While POC testing undeniably provides self-sufficiency from the central laboratory, this independence must be accompanied by a thorough understanding of its limitations and potential biases: we encourage all physicians using POC electrolyte panels to remain aware of the respective strengths and weaknesses of DP for Na⁺ and K⁺ measurements.