Figures 2 and 3 show the Bland–Altman diagrams for ▵PCO2, and ▵PCO2/C(a-cv)O2 ratio, respectively. Results for repeatability of calculated variables are presented in Table 3. Although the variability of ▵PCO2 expressed as the CVw and LSC was high, the 95% range, and also the limits of agreement (SDD) for this calculated variable were small (Figure 2). Indeed, the 95% range for ▵PCO2 was 1.4 mm Hg, that is, for a patient whose ▵PCO2 was measured as 6 mm Hg, the patient's “true” value lies between 4.6 mm Hg (≈5) and 7.4 mm Hg (≈7). This suggests that the repeatability of ▵PCO2 was good. However, the 95% range for the ▵PCO2/C(a-cv)O2 ratio was 0.4 mm Hg/mL. Thus, if a patient has a ▵PCO2/C(a-cv)O2 of 1.4 mm Hg/mL, the ▵PCO2/C(a-cv)O2 could be as low as 1 mm Hg/mL or as high as 1.8 mm Hg/mL. This indicates that the inherent variability of this ratio was important. There were no correlations between SDdiff and the averages of the first and second measurements for all measured and calculated variables (see Table S3, supplementary content, http://links.lww.com/MD/A164). Additional results on the repeatability of the measured and calculated variables in patients with septic shock are shown in Table S4 (supplementary content, http://links.lww.com/MD/A165).
The most important findings of our study are as follows: the repeatability of SaO2, ScvO2, PCO2, and pH was very good, whereas the inherent natural variability of PaO2 and PaO2/FiO2 was considerable; ▵PCO2 was found to be repeatable, whereas the repeatability of ▵PCO2/C(a-cv)O2 was poorer.
All BG analytes assayed on point-of-care testing analyzers are measured with uncertainty due to random variation. Laboratory analytes are subject to 3 main sources of variations: preanalytical, analytical, and biological. Therefore, the results of these analytes change with time and the numbers we obtain are not constant. Preanalytical, analytical, and biological variations are included in the within-subject variation, which results from random fluctuation around an individual homeostatic set point.18,19 The magnitude of this variability must be known by the physicians to correctly interpret the results. The repeatability is the degree to which repeated measurements of the same parameter are similar when carried out under the same conditions of measurements.20 In this study, we chose to evaluate the repeatability by taking 2 blood samples immediately one after the other in stable patients to avoid a real modification in the patient's condition. Furthermore, preanalytical and analytical factors were the only 2 sources of variation that we could act on to minimize the within-subject variation and studied the repeatability of each variable reliably. In our study, we used the same preheparinized BG syringe with the same volume and preparation conditions for the 2 blood samples. Also, blood discard volumes of >3 times the deadspace of catheters were withdrawn to avoid the risk of diluting the samples with flush solution.21 Also, special care was taken to dislodge any air bubbles by gently tapping the sides of the samplers. Moreover, the 2 different samples were analyzed immediately after being taken. Therefore, taking all these precautions, we believe that the preanalytical errors were minimized in this study. Similarly, analytical errors were lessened as BG analyzer was calibrated several times a day. Furthermore, GEM Premier 3000 (Instrumentation Laboratory Co) has an active automated performance quality control program termed Intelligent Quality Management, which continually controls the entire process of sample analysis, enables instantaneous error detection, and performs corrective actions for error elimination.
Repeatability measurements could be reported by multiple statistical methods. In this work, we used the Bland–Altman method and the CVw. The CVw is the ratio of within-subject SD to the mean value of the 2 measurements. Thus, the interpretation of the CVw is as a percentage. However, this figure implies that there is a systematic error even in data sets in which no such bias exists (like our data) because the percentage value (eg, 11.7%) of the lowest measurement in the data set is much smaller than 11.7% of the highest measurement. In this study, we found no correlation between mean versus difference for all measured and calculated variables, which confirms the lack of any systematic bias (Table S2, http://links.lww.com/MD/A163). Therefore, the repeatability expressed in absolute units (SDD) was not related to the size of the measurements, and it is a preferable measure for use in daily practice as compared with the CVw and the derived LSC.
The variability that we found in PaO2 was substantial but lesser to that previously reported.1–3 Indeed, our results indicate that if a patient has a reported PaO2 of 70 mm Hg, the PaO2 could be as low as 64 mm Hg or as high as 76 mm Hg due only to inherent variability. Also, in an individual patient, a PaO2 change can be considered significant only if the change between the measurements exceeds 9 mm Hg (Figure 1). The spontaneous variability expressed as mean CVs found in the previous studies ranged from 4.6% to 6.1% for PaO2 and from 3% to 4.7% for PaCO2.1–3 Conversely, in our study, the repeatability was very good for PaCO2 (Table 2). Differences between our and those studies may stem from several factors. First, in the previous studies,1–3 the variability of arterial BGs was evaluated over a time period of about 1 hour. The occurrence of minor alterations in the patient's condition (microatelectasis, mismatch between pulmonary ventilation and perfusion, changes in respiratory rate and oxygen consumption), even in apparently stable patients, cannot completely be excluded during this time. Thus, these changes could have contributed to the increased variability of PaO2 and PaCO2 measurements reported in those studies. Second, our study included a greater number of patients with various types of pathologies; therefore, it was more representative of ICU patients than the previous studies,1–3 among which the largest sample size was only of 29 patients. We cannot explain why we found an important within-subject variability for PaO2. However, our study was designed to evaluate the magnitude of the variability of BG parameters and was not designed to determine the causes of the variability. Moreover, we found that the variability in PaO2 was not reflected in the variability of SaO2 (Table 2). The sigmoidal shape of the oxygen–hemoglobin dissociation curve can explain the very low variability in SaO2. Indeed, at the PaO2 values seen in our study, the dissociation curve is relatively flat, which means that SaO2 does not change significantly, even with large variations in the PaO2. Furthermore, even in patients with SaO2 values no longer on the flat portion of the oxyhemoglobin curve (SaO2 < 94%), the variability of SaO2 remained low, (SDD≈2%) whereas the variability of PaO2 was still significant (SDD≈7 mm Hg) (Table S1, http://links.lww.com/MD/A162). The same results were found in patients ventilated with high PEEP and high FiO2 (Table S2, http://links.lww.com/MD/A163). Thus, in clinical practice, it is probably better to rely on SaO2 values rather than PaO2 when assessing the oxygenation status of patients. However, SaO2 that is provided by GEM Premier 3000 (Instrumentation Laboratory Co) is a calculated and not a measured parameter. Thus, our results regarding SaO2 might not be applicable to other devices that directly measure SaO2 (co-oximeter). These devices, based on spectrophotometric principles, utilize numerous wavelengths of light to measure the concentrations of oxy-hemoglobin saturation directly. Thus, the within-subject variability of a measured SaO2 will depend on the intraindividual variability of parameters that influence the oxy-hemoglobin dissociation curve such as pH, temperature, and 2,3 DPG. In our study, the within-subject variability of pH was minuscule as shown in Table 2. Furthermore, there is no reason to believe that the temperature of the patient has changed over the sampling period. It is for these reasons that we do not think that the within-subject variability of measured SaO2 would be greater than the calculated SaO2. However, further studies are needed to validate this hypothesis.
It is of interest to note that the repeatability of ScvO2 was found to be good and was most likely due to the low variability in PcvO2 (Table 2). Indeed, the 95% range for an individual true value of ScvO2 was small (±2%). For a patient whose ScvO2 was measured as 70%, the “true” value lies between 68% and 72%. Therefore, the reliance that we can place on reading a value of ScvO2 is good. This finding is of clinical importance since implementation of ScvO2 as a resuscitation goal was associated with decreased mortality in septic shock patients.4,22
To our knowledge, this is the first study investigating the spontaneous variability of ScvO2, ▵PCO2, and ▵PCO2/C(a-cv)O2 ratio in critically ill patients. We found a good repeatability of ▵PCO2 related to the small within-subject variability of the parameters on which it depends. It is well known that when a reported result is derived from >1 actual measurement through their addition, subtraction, multiplication, or division, the uncertainty of the final result can be calculated by summing the uncertainty components of the contributing measurements.18 Therefore, the more variables in the formula, the more uncertainty exists in the result. This may explain why the inherent variability of ▵PCO2/C(a-cv)O2 ratio was found to be significant (Table 3), even though the within-subject variability of the parameters on which it depends were small. Monnet et al11 suggested that considering the ▵PCO2/C(a-cv)O2 ratio as a surrogate of the respiratory quotient, this ratio could be used as a marker of global anaerobic metabolism in critically ill patients. They found that a ▵PCO2/C(a-cv)O2 ratio at baseline ≥1.8 mm Hg/mL predicted with high sensitivity and specificity an increase of oxygen consumption in patients whose oxygen delivery increased after fluid administration. However, according to our results, this threshold value could be as low as 1.4 mm Hg/mL or as high as 2.2 mm Hg/mL (Table 3). This ratio is not trustworthy and therefore probably not reliable for use at the bedside.
How significant is an observed change between 2 successive measurements? This must be a common problem in the clinical interpretation of BG monitoring. Changes in results are often interpreted against empirical criteria. For example, the difference may be considered significant when the test result has doubled or tripled. Changes in results are caused by within-subject variation as well as by deterioration or improvement of the patient's condition. The magnitude of the “critical difference” between results—that is, a change that must occur before significance can be claimed—may be calculated as the SDD or LSC. This means that in an individual patient, in 95% of cases, a ▵PCO2 change value >2 mm Hg or 32.4% is a true ▵PCO2 change and is not due to natural variation.
Our study has several limitations. First, it is a single-center study. Our findings might not apply to other populations. However, our ICU admits a variety of medical and surgical patients, and our population is likely to be representative of other general ICU populations. Second, our results depend on the point-of-care BG analyzer used (GEM Premier 3000; Instrumentation Laboratory Co) and might not be applicable to other devices. Nevertheless, the analytical performance of this device is considered well comparable with most established central laboratory BG analyzers.23,24 Interestingly, the SDdiff of duplicate analyses of whole blood samples obtained from the operating rooms, the ICU, or the emergency department showed no differences between the GEM Premier 3000 and reference central laboratory instruments.25 The GEM Premier 3000 is also the device that is used in our central laboratory hospital for BG analysis. Furthermore, this analyzer is used by hundreds of hospitals in Western countries, making our findings highly relevant to medical practice in many institutions. According to the manufacturer, over 12,000 point-of-care machines just like ours are currently in use all over the world. Therefore, we believe that the principles established by our observations are likely to be generalizable.
The short-term repeatability of BGs was good except for PaO2 even in stable ICU patients not undergoing therapeutic intervention. Significant spontaneous variation in the ▵PCO2/C(a-cv)O2 ratio occurred over short-time intervals, whereas the repeatability of ▵PCO2 was good. The clinician should take into account this inherent variability as well as the clinical spectrum of a patient and make informed clinical decisions based on them.
The authors thank the nursing staff of the ICU. Without their participation, this work would not have been possible.
1. Thorson SH, Marini JJ, Pierson DJ, et al. Variability of arterial blood gas values in stable patients in the ICU. Chest
2. Hess D, Agarwal NN. Variability of blood gases, pulse oximeter saturation, and end-tidal carbon dioxide pressure in stable, mechanically ventilated trauma patients. J Clin Monit
3. Sasse SA, Chen PA, Mahutte CK. Variability of arterial blood gas values over time in stable medical ICU patients. Chest
4. Rivers E, Nguyen B, Havstad S, et al. Early goal-directed therapy in the treatment of severe sepsis and septic shock. N Engl J Med
5. Dellinger RP, Levy MM, Rhodes A, et al. Surviving sepsis campaign: international guidelines for management of severe sepsis and septic shock: 2012. Crit Care Med
6. Vallée F, Vallet B, Mathe O, et al. Central venous-to-arterial carbon dioxide difference: an additional target for goal-directed therapy in septic shock? Intensive Care Med
7. van Beest PA, Lont MC, Holman ND, et al. Central venous-arterial PCO2
difference as a tool in resuscitation of septic patients. Intensive Care Med
8. Mallat J, Benzidi Y, Salleron J, et al. Time course of central venous-to-arterial carbon dioxide tension difference in septic shock patients receiving incremental doses of dobutamine. Intensive Care Med
9. Du W, Liu DW, Wang XT, et al. Combining central venous-to-arterial partial pressure of carbon dioxide difference and central venous oxygen saturation to guide resuscitation in septic shock. J Crit Care
10. Mallat J, Pepy F, Lemyze M, et al. Central venous-to-arterial carbon dioxide partial pressure difference in early resuscitation from septic shock: a prospective observational study. Eur J Anaesthesiol
11. Monnet X, Julien F, Ait-Hamou N, et al. Lactate and venoarterial carbon dioxide difference/arterial-venous oxygen difference ratio, but not central venous oxygen saturation, predict increase in oxygen consumption in fluid responders. Crit Care Med
12. Bland JM, Altman DG. Statistical methods for assessing agreement between two methods of clinical measurement. Lancet
13. Chinn S. Statistics in respiratory medicine. 2. Repeatability and method comparison. Thorax
14. Lodder MC, Lems WF, Ader HJ, et al. Reproducibility of bone mineral density measurement in daily practice. Ann Rheum Dis
15. Glüer CC. Monitoring skeletal changes by radiological techniques. J Bone Miner Res
16. Bland JM, Altman DG. Measurement error. Br Med J
18. White GH, Farrance I. Uncertainty of measurement in quantitative medical testing: a laboratory implementation guide. Clin Biochem Rev
19. Houillier P, Coste J, Froissart M. How many measurements to make a decision? Clin J Am Soc Nephrol
20. British Standard Institution. Precision of test methods 1: guide for the determination of repeatability and reproducibility for a standard test method, BS 5497-1. London: British Standard Institution; 1979.
21. Rickard CM, Couchman BA, Schmidt SJ, et al. A discard volume of twice the deadspace ensures clinically accurate arterial blood gases and electrolytes and prevents unnecessary blood loss. Crit Care Med
22. Nguyen HB, Corbett SW, Steele R, et al. Implementation of a bundle of quality indicators for the early management of severe sepsis and septic shock is associated with decreased mortality. Crit Care Med
23. Bénéteau-Burnat B, Bocque MC, Lorin A, et al. Evaluation of the blood gas analyzer Gem PREMIER 3000. Clin Chem Lab Med
24. Steinfelder-Visscher J, Weerwind PW, Teerenstra S, et al. Reliability of point-of-care hematocrit, blood gas, electrolyte, lactate and glucose measurement during cardiopulmonary bypass. Perfusion
25. Jacobs E, Ancy JJ, Smith M. Multi-site performance evaluation of pH, blood gas, electrolyte, glucose, and lactate determinations with the GEM Premier 3000 critical care analyzer. Point of Care
Supplemental Digital Content
Copyright © 2015 Wolters Kluwer Health, Inc. All rights reserved.