Bland-Altman Difference Research Articles

Comparison of white blood cell differential percentages determined by the in‐house LaserCyte hematology analyzer and a manual method

The LaserCyte hematology analyzer (IDEXX Laboratories, Chalfont St. Peter, Bucks, UK) is the first in-house laser-based single channel flow cytometer designed specifically for veterinary practice. The instrument provides a full hematologic analysis including a 5-part WBC differential (LC-diff%). We are unaware of published studies comparing LC-diff% results to those determined by other methods used in practice. To compare LC-diff% results to those obtained by a manual differential cell count (M-diff%). Eighty-six venous blood samples from 44 dogs and 42 cats were collected into EDTA tubes at the Forest Veterinary Centre (Epping, UK). Samples were analyzed using the LaserCyte within 1 hour of collection. Unstained blood smears were then posted to Langford Veterinary Diagnostics, University of Bristol, and stained with modified Wright's stain. One hundred-cell manual differential counts were performed by 2 technicians and the mean percentage was calculated for each cell type. Data (LC-diff% vs M-diff%) were analyzed using Wilcoxon signed rank tests, Deming regression, and Bland-Altman difference plots. Significant differences between methods were found for neutrophil and monocyte percentages in samples from dogs and cats and for eosinophil percentage in samples from cats. Correlations (r) (canine/feline) were .55/.72 for neutrophils, .76/.69 for lymphocytes, .05/.29 for monocytes and .60/.82 for eosinophils. Agreement between LC-diff% and Mdiff% results was poor in samples from both species. Bland-Altman plots revealed outliers in samples with atypical WBCs (1 cat), leukocytosis (2 dogs, 9 cats), and leukopenia (16 dogs, 11 cats). The LaserCyte generated error flags in 28 of 86 (32.6%) samples, included 7 with leukopenia, 8 with lymphopenia, 7 with leukocytosis, 1 with anemia, and 1 with erythrocytosis. When results from these 28 samples were excluded, correlations from the remaining nonflagged results (canine/feline) were .63/.65 for neutrophils, .67/.65 for lymphocytes, .11/.33 for monocytes, and .63/.82 for eosinophils. Although use of a 100-cell (vs 200-cell) M-diff% may be a limitation of our study, good correlation between WBC differentials obtained using the LaserCyte and the manual method was achieved only for feline eosinophils.

The reactivity of paired plasma and serum samples are comparable in the anticardiolipin and anti‐β2‐glycoprotein‐1 ELISAs. Reply to a rebuttal

We thank Wong and Favaloro for suggesting an alternate method for displaying and analyzing our data concerning the reactivity of paired plasma and serum samples in the anticardiolipin (ACL) and anti‐β2‐glycoprotein‐1 (anti‐β2GP1) ELISAs [1Wong R.C. Favaloro E.J. The reactivity of paired plasma and serum samples are comparable in the anticardiolipin and anti‐β2‐glycoprotein‐1 ELISAs: a rebuttal.J Thromb Haemost. 2006; 4: 1434-5Abstract Full Text Full Text PDF PubMed Scopus (4) Google Scholar]. The ELISA data from our original report have now been analyzed using Bland–Altman difference plots (Fig. 1A–D) [2Lewis D.A. Pound M.L. Ortel T.L. The reactivity of paired plasma and serum samples are comparable in the anticardiolipin and anti‐β2‐glycoprotein‐1 ELISAs.J Thromb Haemost. 2006; 4: 265-7Abstract Full Text Full Text PDF PubMed Scopus (12) Google Scholar, 3Bland J.M. Altman D.G. Statistical methods for assessing agreement between two methods of clinical measurement.Lancet. 1986; 1: 307-10Abstract PubMed Scopus (39517) Google Scholar]. The mean difference of the antibody levels in the two pairs of sample types (value detected in plasma sample minus value detected in serum sample) and 95% limits of agreement for these values are shown as dashed lines on the graphs and are listed in the figure legend. The mean differences for both sample pairs in both ELISAs are relatively small (−0.91–1.89 GPL units) and vary in a random manner, indicating that the sample pairs behave similarly in each of the ELISAs. The absolute values of 95% limits of agreement for the samples pairs are within 18 GPL units for the ACL ELISA and 30 anti‐β2GP1 units for the anti‐β2GP1 ELISA, indicating that the differences we found in the antibody values determined using the different sample types are not clinically relevant. As expected, the mean differences between the sample types in both ELISAs were greater at higher antibody levels. If antibody levels above 100 GPL or 100 anti‐β2GP1 units are excluded from the analysis, the absolute values of 95% limits of agreement are < 12 GPL or 12 anti‐β2GP1 units for both sample pairs in both assays (data not shown). Wong and Favaloro also raised concerns about dilution artifacts in plasma samples. In our original letter the antibody levels determined using citrated plasma samples were corrected for dilution of samples (collection of blood using ethylenediamine‐tetraacetic acid (EDTA) as an anticoagulant does not result in dilution of the sample). We do note, however, that the mean differences in antibody levels and the 95% limits of agreement for the two sets of sample pairs for both ELISAs are greater for citrated plasma and serum pairs compared with the EDTA plasma and serum pairs (Fig. 1A–D). It is unlikely that this difference reflects only a dilution artifact as the results vary in a random manner. Analysis of our data using different plots supports our original conclusion that the reactivity of plasma and serum is comparable in the ACL and anti‐β2GP1 ELISAs developed in our laboratory. These data would suggest that it is valid to compare antigen levels in datasets containing a combination of citrated plasma, EDTA plasma and serum samples. Our interest in this comparison was to confirm the ability to identify ACL and anti‐β2GP1 antibodies in different plasma and serum samples for future research studies. For clinical laboratory testing we would agree with adhering to the recommended laboratory practices and not mixing sample preparations. This work was supported by a cooperative agreement (U18DD00014) with the Hematologic Diseases Branch, Centers for Disease Control and Prevention (T. L. Ortel), and grant (U54‐HL077878) from the National Institutes of Health (T. L. Ortel).

Determination of optimal atrioventricular delay for cardiac resynchronization therapy using acute non-invasive blood pressure

In this study, we apply non-invasive blood pressure (BP) monitoring, by continuous finger photoplethysmography (Finometer), to detect directly haemodynamic responses during adjustment of the atrioventricular (AV) delay of cardiac resynchronization therapy (CRT), at different heart rates. Twelve patients were studied with six re-attending for reproducibility assessment. At each AV delay, systolic BP relative to a reference AV delay of 120 ms (SBPrel) was calculated. We found that at higher heart rates, altering the AV delay had a more pronounced effect on BP (average range of SBPrel=17.4 mmHg) compared with resting rates (average range of SBPrel=6.5 mmHg), P<0.0001. Secondly, peak AV delay differed between patients (minimum 120 ms, maximum 200 ms). Thirdly, small changes in AV delay had significant BP effects: programming AV delay 40 ms below the peak AV delay reduced SBPrel by 4.9 mmHg (P<0.003); having it 40 ms above the peak decreased SBPrel by 4.4 mmHg (P<0.0005). Finally, the peak AV delay is highly reproducible both on the same day and at 3 months (Bland-Altman difference: 3+/-8 ms). Continuous non-invasive arterial pressure monitoring demonstrates that even small changes in AV delay from its haemodynamic peak value have a significant effect on BP. This peak varies between individuals, is highly reproducible, and is more pronounced at higher heart rates than resting rates.

Analytical performance and clinical efficacy for cardiovascular risk estimation of an Olympus immunoturbidimetric high-sensitivity C-reactive protein assay

Increased C-reactive protein (CRP) concentration within the reference interval (<10.0 mg/L) is a strong predictor of cardiovascular disease (CVD) in apparently healthy adults. Cutoff points for use of CRP in estimating CVD risk are <1, 1-3 and >3 mg/L for low, average and high relative risk, respectively. For measuring CRP concentrations to assess cardiovascular risk, high-sensitivity CRP (hsCRP) assays have been developed. The aim of this study was to evaluate the analytical performance and clinical efficacy for cardiovascular risk estimation of the Olympus immunoturbidimetric latex CRP assay (sensitive application). The comparative method used was the CardioPhase* hsCRP assay, approved by the Food and Drug Administration for use in CVD risk assessment. The imprecision of the Olympus hsCRP assay in the concentration range 0.2-10.0 mg/L was 0.38-8.16% within runs and 3.75-9.63% between runs. For method comparison studies, 194 fresh serum samples were selected to cover the interval 0.15-10.0 mg/L CRP. Comparison of the Dade Behring and Olympus methods was performed using weighted Deming regression analysis (slope 0.99 mg/L, intercept 0.002 mg/L, S(y,x)=0.02 mg/L, r=0.992) and a Bland-Altman relative difference plot (mean difference -0.002%, SD=0.040%). The agreement between the Dade Behring and Olympus methods for relative risk class assignments was 95.4%. Statistical analysis of the agreement between the two methods for each relative risk class showed that the differences between the methods were not statistically significant (p>0.10). Although previous reports found poor performance of the Olympus CRP tests for use in cardiovascular and peripheral vascular risk estimation, our study proved good analytical performance and clinical efficacy of the Olympus hsCRP assay for this use.

Analysis of protein S-100B in serum: a methodological study

Dysfunction and damage of the human central nervous system can be detected with biochemical markers, and protein S-100B is the best-established such marker. The aim of this study was to evaluate whether the protein is stable during long-term storage, to establish reference values for the new Elecsys S100 test and to compare this new method with the Liaison Sangtec 100 test. We analysed blood samples from 118 blood donors and 196 patients with subarachnoid haemorrhage or head injury. The long-term stability of S-100B in frozen serum samples was evaluated with repeated analysis in 1997 and 2003 using an immunoradiometric assay. Method comparison between the Liaison Sangtec 100 and Elecsys S100 tests was performed using Bland-Altman difference plots. Serum concentrations increased significantly during long-term storage (mean difference 0.15 microg/L; +/-2 SD, 0.55 microg/L). Serum measurements using the Elecsys S100 method in 118 healthy blood donors showed S-100B levels between 0.02 and 0.08 microg/L (mean 0.05). The 95th percentile was 0.07 microg/L. The Liaison Sangtec 100 test usually measured higher concentrations than the Elecsys S100 method, and the difference between the two methods increased with increasing concentrations. The mean difference between the methods was 0.14 microg/L (+/-2 SD, 0.39 microg/L). Protein S-100B is not stable during long-term storage and the two analytical methods are not interchangeable.

Determination of total bilirubin in whole blood from neonates: results from a French multicenter study

Jaundice is frequent in neonates and can cause severe complications, especially in premature neonates, particularly the risk of developing acute bilirubin encephalopathy. Our purpose was to verify if determination of total bilirubin (TBIL) in whole blood on an ABL 735 blood gas analyzer with a spectrophotometer module could provide an analytical alternative to chemical methods of TBIL measurement. Our multicenter comparative study involved four hospital laboratories. We studied the repeatability and reproducibility of ABL 735 TBIL measurements in two control sera of medium (N1, 58.1 micromol/L) and high (N2, 275.3 micromol/L) TBIL levels. The same study was simultaneously conducted on four chemistry instruments (two LX 20, one Integra 800 and one Hitachi 917) using four Jendrassik-Grof derived methods. At one site, repeatability was performed with two adult whole-blood samples containing low and high TBIL levels (55.1 and 312.6 micromol/L). Repeatability tests provided coefficients of variation (CVs) between 0.67% and 1.86% on the ABL 735 system, vs. 0.35% and 1.96% for the chemistry instruments. Reproducibility tests for the same control sera resulted in CVs between 1.01% and 3.55% for the ABL 735 and between 0.52% and 3.65% for the chemistry instruments. Recovery for the N1 and N2 control sera was 102-120%. A correlation study of TBIL determination in whole blood vs. plasma was conducted on 473 neonatal blood samples. Correlation coefficients between whole blood and plasma TBIL ranged from 0.969 to 0.994. Passing-Bablok equations were y=1.17x+9.7 [site 1 (IP)], y=1.01x+5.6 [site 2 (JPB, MR)], and y=1.00x-20 [site 3 (AS)]. Only 10% of the results fell outside the 10% range in the bias-corrected Bland-Altman difference plot for the ABL 735 method compared to traditional laboratory methods. The ABL 735 instrument is reliable for measuring TBIL in 70-microL whole blood samples from neonates. Thus, this method might allow significant blood savings in preterm neonates. Correlation with the reference method for plasma or sera must be established to ensure good follow-up of patients.

Evaluation of fluorescence hs-CRP immunoassay for point-of-care testing

Background C-reactive protein (CRP) is one of acute phase respondents that has been used to monitor infection and inflammation episodes. Recent studies have shown that high-sensitivity C-reactive protein (hs-CRP) is a potential risk predictor for future atherosclerosis and cardiovascular diseases (CVD). Methods We previously developed a fluorescence-based immunochromatographic method for measuring hs-CRP concentrations ( i-CHROMA™ hs-CRP assay) in blood. Whole blood was mixed with detector buffer, and then loaded onto a test cartridge. After 10 min of incubation, the test cartridge was inserted and scanned for acquisition of fluorescence intensity in a laser fluorescence reader ( i-CHROMA™ reader). The fluorescence intensity was microprocessed and converted into the concentration of CRP in blood. The test result of 150 samples by the i-CHROMA™ hs-CRP assay method was compared and evaluated with those by TBA 200FR turbidimetry and BN II nephelometry method. The Deming regression and the Bland-Altman difference plot analysis were used for comparison of hs-CRP test result. Results The i-CHROMA™ hs-CRP assay system exhibited a good linearity with in the whole measuring range ( R = 0.997). The imprecision of intra- and the inter-assay CVs (coefficient of variation) of assay system were CVs< 3% and < 5% in the range of 0.5–20 mg/l, respectively. The i-CHROMA™ hs-CRP assay method correlated well with TBA 200FR turbidimetry and BN II nephelometry assay method ( R = 0.988, N = 143 and R = 0.989, N = 143). Conclusion The i-CHROMA™ hs-CRP assay system is comparable to those of other well-known fully automated hs-CRP assay and is suitable for point-of-care testing (POCT) in detection and quantification of hs-CRP.

Evaluation of a glucose meter for determining the glycemic responses of foods

Background Glucose meters are convenient for measuring postprandial glycemic responses. However, their performance for this purpose has not been evaluated. Methods Glucose responses of 7 potato meals were measured using the One Touch Ultra® (OTU) glucose meter and a reference method (Yellow Springs Instruments Glucose Analyzer, YSI) and the incremental areas under the curves (AUC) and glycemic index (GI) values compared. Results Mean AUC OTU was greater than AUC YSI (192 ± 8 vs. 175 ± 7 mmol×min/l, p = 0.001), but GI OTU tended to be less than GI YSI (69 ± 3 vs. 74 ± 3, p = 0.052). Bland–Altman difference plots showed wide 95% limits of agreement for AUC (−84 to 119 mmol×min/l) and GI (−21 to 26) values of individual subjects, and for the mean GI values of the 7 foods (−11 to 21). Total and error variance of AUC and GI values were greater for OTU than YSI, and food means differed significantly for YSI ( p < 0.01) but not OTU ( p = 0.11). Conclusion AUC and GI values determined by OTU are more variable and do not agree well with those obtained by YSI. Thus, the OTU is not recommended for determining AUC or GI in normal subjects. This conclusion does not necessarily apply to other glucose meters whose performance should be evaluated.

Poor Agreement among Prothrombin Time International Normalized Ratio Methods: Comparison of Seven Commercial Reagents

Prothrombin time (PT) has long been the most popular test for monitoring oral anticoagulation therapy. The International Normalized Ratio (INR) was introduced to overcome the problem of marked variation in PT results among laboratories and the various recommendations for patient care. According to this principle, all reagents should be calibrated to give identical results and the same patient care globally. This is necessary for monitoring of single patients and for application of the results of anticoagulation trials and guidelines to clinical practice. We took blood samples from 150 patients for whom oral anticoagulation had been prescribed. Plasmas were separated and PTs determined by use of seven commercial reagents and four calibrator sets. The differences in results were assessed by plotting, for each possible pair of methods, the differences in INR values for each sample against the mean INR value (Bland-Altman difference plots). Mean results differed significantly (P <0.001) for 17 of 21 possible paired comparisons of methods. Only two pairs of methods produced very similar results when assessed for problems of substantial differences in INR values; a significant, systematic increase in the difference with INR; and a significant systematic increase in the variation in difference with increasing INR values. The agreement among several (and perhaps most) commercial INR methods is poor. The failure of current calibration strategies may severely compromise both the monitoring of individual patients and the application of oral anticoagulation guidelines and trial results to clinical practice.

Haemoglobin A1c: Evaluation of three point of care analysers for use in a paediatric diabetes clinic

The measurement of haemoglobin A1c(HbA1c) by high-performance liquid chromatography (HPLC) is generally deemed unsuitable for point of care testing (POCT) due to its complexity and extended turnaround times (TAT). The aim of this project was to evaluate two new HPLC instruments, the Bio-Rad D10 and the Primus PDQ, as POCT instruments compared with Bayer's DCA2000 HbA1c immunoassay analyser in our paediatric diabetes clinic. A total of 228 samples were analysed, of which 160 analyses were performed in our paediatric diabetes clinic. HbA1c results were compared by the Passing-Bablok agreement test, the Bland-Altman difference analysis, within- and between-run imprecision, and TAT. : The agreement test and difference analysis showed a correlation of r2 = 0.96 and a mean HbA1c difference of < 0.1% between the three analysers. The PDQ and the D10 achieved the within-run target coefficient of variation (CV) of < 2% at an HbA1c of 7.5%. Between-run imprecision at an HbA1c of 10.8% produced CV of 3.5%, 2.4% and 1.6% for the D10, DCA2000 and PDQ, respectively. TAT studies confirmed that the PDQ was substantially faster than the DCA2000 and D10. The PDQ had the shortest TAT, afforded random access and exhibited acceptable imprecision, and hence is the preferred instrument for our POCT environment.

An enzymatic clinical chemistry laboratory experiment incorporating an introduction to mathematical method comparison techniques

An enzymatic laboratory experiment based on the analysis of serum is described that is suitable for students of clinical chemistry. The experiment incorporates an introduction to mathematical method-comparison techniques in which three different clinical glucose analysis methods are compared using linear regression and Bland-Altman difference plots. The experiment may also be undertaken using simple glucose solutions by general biochemistry students.

CREATININE CLEARANCE AFTER KIDNEY TRANSPLANTATION - WHICH METHOD DO WE TRUST?

O219 A discussion is ongoing about an optimal method of glomerular filtration rate (GFR) estimation after kidney transplantation. Creatinine clearance (CrCl) as a measure of GFR is compromised not only outside the normal GFR range, but also by inaccurate urinary collection and by inaccuracy in creatinine measurement. Accepted biochemical difference ± 10% connotes that patients with creatinine clearance 60 ml/min fall in fact in range 53-67 ml/min, additionally, some laboratories are working with a difference that is much higher. Aim: to compare measured CrCl and predicted GFR (Cocroft- Gault) with Nankivell formula as a reference modality. Patients and methods: 120 single measurement in 20 patients divided in two groups: 3-6 months and more than 1 year after Tx. Bland-Altman difference plot was used for statistical analysis. Results: Variability of differences between methods was not constant across the range of measurement. During the period 3-6 months after Tx: measured CrCl was lower than values predicted by Nankivell form in 78% of cases (mean difference -0,18 ml/s, SD 0,23). Cocroft form overestimated the Nankivell GFR by 0,07 ml/s mean, SD 0,13, however a positive bias 0,17 ml/s, SD 0,07 was present in 50%, negative bias -0,06 ml/s, SD 0,09 in 33% of patiens. Different results were seen after 1 year. With GFR up to 1 ml/s, difference between measured CrCl and Nankivell GFR was negligible, however with GFR more than 1 ml/s the measured CrCl values were significantly higher than the Nankivell estimated GFR values (0,17 ml/s mean, SD 0,24). Cocroft form overestimated Nankivell GFR in 0,11 ml/s mean, SD 0,14 in full range of GFR. Conclusion: Measured CrCl as well as predicted GFR Cocroft-Gault and Nankivell formula can not be used interchangeably for longitudinal follow up of kidney transplant recipients. Another parameters, less charged by biochemical determination, sampling or calculation errors would be an advantage.

A rapid, simple measurement of human albumin in whole blood using a fluorescence immunoassay (I)

Background: Human serum albumin (HSA) is the most abundant plasma protein and plays key a role in metabolism. The variation in albumin concentration provides valuable information related to metabolic diseases and diagnostic application. Methods: We constructed two assay systems to quantify the albumin concentration. The immunoassay used a fluorescence (FL) dye to detect albumin in samples and employed the conventional chromatography as a separation system. The assay system consists of an anti-HSA-mAb or an HSA immobilized test strip in a disposable cartridge, a fluorescence-labeled detector buffer and a laser-fluorescence scanner. We mixed the sample with detector, loaded it onto a cartridge, incubated it for 10 min and measured the concentration of albumin in a laser-fluorescence scanner. We examined the comparability of assay with an automated BCG dye binding method using a Hitachi 747 biochemical analyzer. Results: The correlation of coefficient between A T/ A C as converted from the relative fluorescence units (RFU) and albumin concentration displayed reasonable reliability in both the competition and the inhibition assay systems ( r=0.998). Using the Bland–Altman difference plot analysis, we observed an acceptable agreement between two methods, the fluorescence immunochromatography assay (FL-ICA) and the automated BCG dye-binding method of a Hitachi biochemical analyzer, over the clinical relevant range of HSA concentrations. The coefficient of variation (CV) of within- and between-run variation in the immunoassay system was <8% and the recovery fell within 5% in each control sample. In addition to its reliable analytical performance, the assay with whole blood can be completed in 12 min using a one-step operation without any pretreatment. Conclusion: The developed immunoassay system using fluorescence dye and lateral-flow chromatography is a simple, fast and reliable method for quantifying the albumin concentration in whole blood.

Plasma fibrinogen measurement in the horse: comparison of Millar's technique with a chronometric technique and theQBC -Vet Autoreader™

Plasma fibrinogen is widely used in horse practice as an unspecific positive marker of inflammatory diseases; it is also lowered in disseminated intravascular coagulation. Three fibrinogen measurement methods—Millar's heat-denaturation in a microhaematocrit tube, automated reader for heat-denaturation, and chronometric measurement of clot formation after addition of excess thrombin—were compared by means of Passing-Bablock's regression and Bland-Altman difference plots, in blood plasma of 30 clinically healthy and 57 diseased horses. Correlations between the three techniques were excellent (r>0·92). The two heat-denaturation techniques correlated very closely up to 6gl−1, above which the results obtained by Millar's technique started to fall below those obtained by the automatic reader. There was proportional bias between Millar's technique and the chronometric technique, with the latter producing results some 30% lower, indicating that reference intervals and decision limits should be adapted accordingly.

Measurement of Homocysteine and Other Aminothiols in Plasma: Advantages of Using Tris(2-carboxyethyl)phosphine as Reductant Compared with Tri-n-butylphosphine

Aminothiols have been implicated in the pathogenesis of arteriosclerosis, and reliable methods are needed to determine their concentrations in body fluids. We present a comparison of two analytical methods and focus on the reduction of low-molecular weight and protein-mixed disulfides of homocysteine, cysteine, cysteinyl-glycine, and glutathione. The plasma total aminothiol profile was determined by HPLC with fluorescence detection after derivatization with ammonium 7-fluorobenzo-2-oxa-1,3-diazole-4-sulfonate. Disulfides and protein-bound aminothiols were reduced by either tri-n-butylphosphine (the TBP method) or tris(2-carboxyethyl)phosphine (the TCEP method); the effects of temperature, time of reduction, and concentration of reductants were evaluated. The intraassay imprecision (CV) was <3% for all aminothiols using both methods. The interassay CVs for total cysteine (tCys), total cysteinyl-glycine (tCys-Gly), and total homocysteine (tHcy) were <4% and <8% for the TCEP and TBP methods, respectively, whereas for total glutathione (tGSH) the interassay CV was >12% for both methods. Deming regression and Bland-Altman difference plots showed positive biases for total aminothiol concentrations determined by the TCEP method relative to the TBP method. The mean proportional biases were 65%, 27%, 6%, and 60% for tCys, tCys-Gly, tHcy, and tGSH, respectively. The calculated concentrations of total aminothiols by the TCEP method were less influenced by changes in temperature and concentration of reducing agent or by calibrator matrix. The agreement between the TCEP and TBP methods was considerably lower for the determination of tCys, tCys-Gly, and tGSH than for tHcy. For total-aminothiol determination, the TCEP method yields better reproducibility and is more robust than the TBP method.

Sending Blood Gas Specimens Through Pressurized Transport Tube Systems Exaggerates the Error in Oxygen Tension Measurements Created by the Presence of Air Bubbles

Sending Blood Gas Specimens Through Pressurized Transport Tube Systems Exaggerates the Error in Oxygen Tension Measurements Created by the Presence of Air Bubbles

Sending blood gas specimens through pressurized transport tube systems exaggerates the error in oxygen tension measurements created by the presence of air bubbles.

Sending blood gas specimens through pressurized transport tube systems exaggerates the error in oxygen tension measurements created by the presence of air bubbles.