Comparison of point‐of‐care and central laboratory analyzers for blood gas and lactate measurements

Background Blood gas analysis and blood lactate measurement have important roles in patient management. Point‐of‐care (POC) testing simplifies and provides rapid blood gas and lactate measurements. This study aimed to compare pH, pCO2, pO2, and lactate measurements between a POC device and a benchtop blood gas analyzer typically used in a hospital central laboratory, and to evaluate the inter‐device variability of the POC device. Methods A cross‐sectional study was conducted with a sample size of 100. Each sample was measured for pH, pCO2, pO2, and lactate using a Nova pHOx plus L® benchtop blood gas analyzer in the central laboratory and an i‐STAT® handheld POC device. The results of both devices were compared using Pearson or Spearman correlation coefficients and Bland‐Altman tests. Testing of the inter‐device variability was done by using three different i‐STAT® devices, and the results were compared statistically. Results Strong correlations were observed for all test results. In Bland‐Altman analysis, ≥95% of the results were within the limits of agreement, with the exception of lactate, which had only 93%. The results that were beyond the limits were primarily lactate levels >8 mmol/L. Biases between the benchtop analyzer and the i‐STAT® were not clinically significant, except pH. No significant inter‐device variability was observed between the i‐STAT® analyzers. Conclusion This comparison study of pH, pCO2, pO2, and lactate measurements between Nova pHOx plus L® and i‐STAT® analyzers showed good agreement. However, lactate measurement results >8 mmol/L on the i‐STAT® analyzer should be interpreted with caution.

and lactate measurements required 125 μL specimen for the Nova pHOx plus L ® and 95 μL for the i-STAT ® analyzer. 3,4 Specimens were analyzed on the i-STAT ® analyzer using CG4+ disposable cartridges.
The analytes being studied for blood gas were pH, pCO 2 , and pO 2 because other blood gas parameters (eg, total CO 2 , HCO 3 , Base Excess, and sO 2 ) are calculated based on the three aforementioned parameters on the i-STAT ® analyzer. 3

Within-run and between-day precision tests for i-STAT ® and
Nova pHOx plus L ® analyzers were measured using three levels of control solutions. Within-run tests were performed 10 times for each level of control solution, while between-day tests were performed for 10 consecutive days. The results were then compared with precision data stated by the manufacturer. The inter-device variability of the i-STAT ® was evaluated with ANOVA or Kruskal-Wallis tests.
The correlation between the results from the i-STAT ® POC device and the Nova pHOx plus L ® analyzer was determined using Pearson and Spearman correlation coefficients. In addition, Bland-Altman analysis was used to measure the level of agreement between the i-STAT ® and the Nova pHOx plus L ® based on the values of bias and limits of agreement. Bland-Altman plots were made by plotting the differences between analyzers against the calculated means. The Passing-Bablok regression test was performed to obtain the pH, pCO 2 , pO 2 , and lactate prediction equations to see whether there were constant or proportional differences in pH, pCO 2 , pO 2 , and lactate test results between the handheld and benchtop analyzers. 6 To assess whether the analytical differences between the i-STAT ® and Nova pHOx plus L ® may lead to clinically significant discrepancies, the values of bias between the analyzers were compared to acceptable analytical performance based on the 1992 Clinical Laboratory Improvement Amendments (CLIA) proficiency testing criteria target value for pH ± 0.04 pH units, for pCO 2 ± 5 mm Hg, and for pO 2 ± 3 SD. 7 An acceptable analytical performance for lactate was based on the 2014 Royal College of Pathologists of Australasia (RCPA) Allowable Limits of Performance, with a target value of ±1.0 mmol/L for lactate ≤10.0 mmol/L. 8 Discrepancies between the values obtained by the i-STAT ® and the Nova pHOx plus L ® were not considered clinically significant if they were less than the target value criteria of acceptable analytical performance.

| RE SULTS
Results of the within-run and between-day imprecision study of the i-STAT ® and Nova pHOx plus L ® analyzers are summarized in Table 1.
The precision profile of each analyzer for pH, pCO 2 , pO 2 , and lactate derived from this study showed comparable results to the claims by each manufacturer. In general, the precision profile of the i-STAT ® and the Nova pHOx plus L ® was considered equivalent.
Inter-device variability tests demonstrated no significant variability between the three i-STAT ® analyzers used in this study. There was no P-value below 0.05 in the inter-device variability tests, meaning no significant differences between the test results of the three analyzers were detected. The P-value of the inter-device variability tests is shown in Table 2.
Strong correlation coefficients (r) between the test results of the i-STAT ® and the Nova pHOx plus L ® were observed in the comparison study-pH (r = 0.893), pCO 2 (r = 0.843), pO 2 (r = 0.983), and lactate (r = 0.986); P < 0.001. The results of the Passing-Bablok regression test performed between the i-STAT ® and the Nova pHOx plus L ® are shown in Table 3. There was no constant nor proportional bias observed between the analyzers for pH, while there was a constant bias for pCO 2 , a proportional bias for lactate, and both for pO 2 .
Bland-Altman analysis of pH results displayed the average bias for the i-STAT ® analyzer of 0.049 pH units lower than the Nova pHOx plus L ® with the 95% limits of agreement ranging from −0.027 to 0.124. The average bias of pCO 2 for the i-STAT ® was 0.28 mm Hg lower than the Nova pHOx plus L ® with 95% limits of agreement ranging from −14.16 to 14.71. The average bias of pO 2 was 2.16 mm Hg higher on the i-STAT ® compared with the Nova pHOx plus L ® with 95% limits of agreement ranging from −17.69 to 13.36.
The average bias of lactate for the i-STAT ® was 0.55 mmol/L lower than the Nova pHOx plus L ® with 95% limits of agreement ranging from −0.79 to 1.89. Bland-Altman plots showed that 96% of the pH differences were within the limits of agreement, with 95% for both pCO 2 and pO 2 differences, but only 93% for lactate differences.
Seven out of eight lactate differences outside the limits of agreement were higher than 8 mmol/L. Bland-Altman plots obtained in this study are shown in Figures 1-4.
The mean biases of pCO 2 , pO 2 , and lactate measurements between the i-STAT ® and the Nova pHOx plus L ® analyzers were not considered clinically important according to acceptable analytical performance based on the 1992 CLIA Proficiency Testing criteria and the 2014 RCPA Allowable Limits of Performance target value. 7,8 The mean biases obtained in this study were 0.28 mm Hg for pCO 2 (target value ± 5 mm Hg), −2.16 mm Hg for pO 2 (target value ± 3 SD, SD: 7.76 mm Hg), and 0.55 mmol/L for lactate pO 2 (target value ± 3 SD), and 0.55 mmol/L (target value was ± 1.0 mmol/L). The bias of pH measurement was 0.049, which was slightly higher than the target value of 0.04.

F I G U R E 4 Bland-Altman plot for lactate
The within-run tests on the i-STAT ® analyzer demonstrated a coefficient of variation (CV) <5% for blood gas and lactate when control materials were run 10 times covering three level of concen-

trations. Between-day precision tests of the i-STAT ® analyzer in this
study demonstrated similar results. The precision test CVs acquired in this study were higher than in the study by Leino et al (2011); the CV of which was <3% for blood gas and lactate. 11 Karon et al (2007) reported that the CV of lactate precision tests using the i-STAT ® analyzer was 3%. 12 However, the CV of the i-STAT ® analyzer obtained in this study was less than the manufacturer CV of blood gases and lactate, except for level 1 pO 2 between-day test. 3 Precision test CVs of the Nova pHOx plus L ® in this study displayed similar results, with CV < 5% for blood gases and lactate.
Inter-device variability test of the i-STAT ® analyzer displayed no Bland-Altman plots showed that 95% of the differences were between the range of the limits of agreement for pH, pCO 2 , and pO 2 , while for lactate it was only 93% of the differences were between limits of agreement. Discrepancies between the i-STAT ® and the Nova pHOx plus L ® were more significant for lactate values of more than 8 mmol/L ( Figure 4). Lactate values measured on the i-STAT ® analyzer were lower than on the Nova pHOx plus L ® . Karon et al (2007) had previously compared lactate measurements between the i-STAT ® and core laboratory analyzer (Vitros 250 ® ). 12 In that study, discrepancies were more significant at values >6 mmol/L, with i-STAT ® results being lower than those of the Vitros 250 ® . 12 Leino et al (2011) reported that i-STAT ® analyzers gave lower lactate values compared with core laboratory plasma analyzers, especially with high lactate concentrations. 11 The same tendency was observed by Ismail et al (2015) in their study comparing the i-STAT ® , GEM Premier 4000 ® , and OMNI S ® analyzers. 14 Therefore, clinicians should be informed that caution must be used when comparing high lactate values between the i-STAT ® device and central laboratory analyzers.

Clinically significant bias was observed for pH values measured
with the i-STAT ® analyzer since they were slightly higher than the target value of acceptable analytical performance. In spite of this, the majority of the results were within acceptable limits. However, any pH value that was near the upper or lower reference limit should be interpreted with caution. 7 Information regarding patient clinical condition, medical history, and treatments should be determined whenever possible.
In conclusion, the differences in blood gas (pH, pCO 2 , pO 2 ) and lactate values among the i-STAT ® POC analyzer and benchtop blood gas laboratory analyzer are negligible. However, medical facilities that use both the i-STAT ® analyzer and central laboratory benchtop blood gas analyzers should inform clinicians that caution must be used when comparing high lactate values between point-of-care and benchtop blood gas laboratory measurement results. There was no inter-device variability observed between different i-STAT ® devices, thus measurement results from multiple i-STAT ® analyzers can be employed interchangeably in daily practice.

ACKNOWLEDGMENT
The study had received ethical approval from the hospital's ethical committee.