Inaccurate haemoglobin measurement by blood gas analyzer may lead to severe adverse clinical consequences

Agreement of Potassium, Sodium, Glucose, and Hemoglobin Measured by Blood Gas Analyzer With Dry Chemistry Analyzer and Complete Blood Count Analyzer: A Two-Center Retrospective Analysis

Background

Blood gas analyzers (BGAs) and dry biochemistry analyzers for potassium and sodium are based on direct electrode methods, and both involve glucose oxidase for glucose detection. However, data are lacking regarding whether the results of the two assay systems can be used interchangeably. In addition, there remains controversy over the consistency between BGA-measured hemoglobin and complete blood count analyzer data. Here, we compared the consistency of sodium, potassium, glucose, and hemoglobin levels measured by BGA and dry chemistry and complete blood count analyzers.

Methods

Data from two teaching hospitals, the Zhejiang Provincial People's Hospital (ZRY) and the Qianfoshan Hospital (QY), were retrospectively analyzed based on dry biochemistry and complete blood count analyzer results as the reference system (X) and BGA as the experimental system (Y). Plasma was used for biochemical analysis at the ZRY Hospital, and serum at the QY Hospital. Paired data from the respective hospitals were evaluated for consistency, and biases between methods were assessed by simple correlation, Passing–Bablok regression, and Bland–Altman analyses.

Results

The correlations of potassium, sodium, glucose, and hemoglobin measured by BGA and dry biochemistry and complete blood count analyzers were high, at 0.9573, 0.8898, 0.9849, and 0.9883 for the ZRY Hospital and 0.9198, 0.8591, 0.9764, and 0.8666, respectively, for the QY Hospital. The results of Passing to Bablok regression analysis showed that the predicted biases at each medical decision level were within clinically acceptable levels for potassium, sodium, glucose, and hemoglobin at the ZRY Hospital. Only the predicted bias of glucose was below the clinically acceptable medical decision levels at the QY Hospital, while potassium, sodium, and hemoglobin were not. Compared with the reference system, the mean bias for BGA measurements at the ZRY Hospital was −0.08 mmol/L (95% confidence interval [CI] −0.091 to −0.069) for potassium, 1.2 mmol/L (95% CI 1.06 to 1.42) for sodium, 0.20 mmol/L (95% CI 0.167 to 0.228) for glucose, and −2.8 g/L for hemoglobin (95% CI −3.14 to −2.49). The mean bias for potassium, sodium, glucose, and hemoglobin at the QY Hospital were −0.46 mmol/L (95% CI −0.475 to −0.452), 3.7 mmol/L (95% CI 3.57 to 3.85), −0.36 mmol/L (95% CI −0.433 to −0.291), and −8.7 g/L (95% CI −9.40 to −8.05), respectively.

Conclusion

BGA can be used interchangeably with plasma electrolyte results from dry biochemistry analyzers but does not show sufficient consistency with serum electrolyte results from dry biochemistry analyzers to allow data interchangeability. Good consistency was observed between BGA and plasma or serum glucose results from dry biochemistry analyzers. However, BGA-measured hemoglobin and hematocrit assay results should be treated with caution.

Accuracy and Reliability of Whole Blood Bilirubin Measurements Using a Roche Blood Gas Analyzer for Neonatal Hyperbilirubinemia Screening and Risk Stratification

BACKGROUND

Accurate bilirubin measurements are essential for appropriate management of neonatal hyperbilirubinemia. This study aimed to evaluate the accuracy and reliability of whole blood bilirubin measurements obtained using a Roche blood gas analyzer (Roche TBiL), with total serum bilirubin (TSB) measurements determined by the Ortho VITROS 4600 chemistry system (Ortho TSB) serving as a reference.

MATERIALS AND METHODS

Medical records of hospitalized neonates that underwent simultaneous Roche TBiL and Ortho TSB measurements were reviewed for eligibility selection and data collection. The correlations and differences between two sets of results were determined using Passing-Bablok regression analysis and a Bland-Altman plot, respectively. For eligible newborns, the risk of developing severe hyperbilirubinemia was assessed using the Bhutani nomogram. Weighted kappa analysis was used to evaluate the agreement between risk prediction by the two methods.

RESULTS

We obtained 618 paired Roche TBiL and Ortho TSB results from 309 neonates. Roche TBiL and Ortho TSB measurements showed a good correlation (r = 0.923; 95% CI: 0.905-0.938). Passing-Bablok regression analysis yielded the following equation: Roche TBiL = 0.794 × Ortho TSB + 1.255 mg/dL, with a slope of 0.794 (95% CI: 0.763-0.825) and intercept of 1.255 (95% CI: 1.042-1.417). The average difference between the two methods was 0.1 ± 1.448 mg/dL. A total of 207 neonates were eligible for evaluation of the agreement between the risk-grading methods. Although kappa analysis showed good agreement between the methods, with a weighted kappa of 0.681 (95% CI: 0.610-0.751) across all populations, the values for approximately half of the neonates at intermediate and high risk of hyperbilirubinemia (33/72) were underestimated by Roche TBiL.

CONCLUSION

Our results indicate that Roche TBiL and Ortho TSB measurements in the neonatal population are not consistent. As a point-of-care and trace blood assay, Roche blood gas bilirubin measurements can facilitate primary screening of neonatal hyperbilirubinemia, but it seems to lack accuracy regarding risk stratification, particularly for high-risk newborn individuals.