The assessment of sample quality is an essential requirement of a total quality management
system in laboratory medicine,[1 ] as well as for routine and specialized coagulation testing.[2 ] Sample inspection, aimed to identify common sources of interference (i.e., hemolysis,
turbidity, and icterus), has been performed for decades by means of visual assessment.
Many limitations have been highlighted in this strategy, including high inaccuracy
and large interobserver variability, along with challenges related to systematic recording
of interference data either in the laboratory information system or in laboratory
reports.[3 ] The new generation of analyzers for clinical chemistry, immunochemistry, and hemostasis
testing is increasingly being equipped with the so-called serum or plasma indices,
which have improved the ability to estimate the interference of hemolysis (H-index),
icterus (I-index), and turbidity (L-index) by using multichromatic wavelength readings
and formulas with adjustments for compensating potential spectral overlaps.[4 ] The widespread implementation of this approach may considerably increase quality
and safety of laboratory testing, since routine use of reagent-specific thresholds
of interference based on validated clinical acceptability criteria will enable suppressing
data obtained analyzing potentially unsuitable specimens. Therefore, this study was
planned to validate the plasma indices on the novel Siemens Atellica COAG 360 System
(Siemens Healthcare Diagnostics Products GmbH).
The Atellica COAG 360 System is an analyzer consolidating five different analytical
technologies (clotting, either optical or opto-mechanical, chromogenic, immunologic,
high-sensitivity luminescence-based [LOCI], and platelet aggregation testing) on a
single testing platform. Plasma quality assessment is performed by quantification
of plasma indices (“HIL”) on a 4-channel photometer using simultaneous multiple wavelengths
scanning (from 365 to 645 nm). Measurements are performed in triplicate for each index,
preceded by a blank measurement with purified water. The total sample volume used
for HIL assessment is 50 μL and the measurement is completed within 16 seconds.
Our local validation of plasma indices on Atellica COAG 360 entailed the assessment
of limit of blank (LOB), limit of detection (LOD), linearity, imprecision (within
run, between-run, and total), result comparison with standard reference techniques,
as well assessment of interference limits for five tests with nine different reagents
and optical clotting methodology ([Table 1 ]). The entire evaluation was based on routine sodium citrate plasma samples collected
in primary evacuated blood tubes (Vacutest Kima, Kima). The study was approved by
the local Institutional Review Board (University Hospital of Verona, Verona, Italy;
SOPAV2, protocol number: 971CESC).
Table 1
Reagent specifications, allowable total error and interference limits of the tests
used in this study
Test
Name
Reagent specifications
ATE
Interference limit
H-index
I-index
L-index
PT
Innovin
Recombinant human tissue factor and synthetic phospholipids
±5.3%
>8
>4
NI6
PT
Thromborel
Human placental thromboplastin
±5.3%
>5
NI5
>3
APTT
Actin FS
Ellagic acid and purified soy phosphatides
±4.5%
>3
>4
NI6
APTT
Actin FSL
Ellagic acid and purified soy and rabbit brain phosphatides
±4.5%
>5
>5
>4
APTT
Actin
Ellagic acid and liquid rabbit brain cephalin
±4.5%
>5
>3
NI6
APTT
Pathromptin LS
Silicon dioxide and vegetable phospholipids
±4.5%
>8
>5
NI6
Fibrinogen
Fibrinogen VR
Thrombin reagent (Clauss method)
±13.6%
NI9
>2
NI6
Antithrombin
Innovance AT
Chromogenic anti-activated factor X (FXa) assay
±8.3%
NI9
>4
>5
D-dimer
Innovance DD
Particle-enhanced immunoturbidimetric assay
±28%
NI9
>3
NI6
Abbreviations: APTT, activated partial thromboplastin time; ATE, allowable total error;
NI5, no interference up to plasma index of 5; NI6, no interference up to plasma index
of 6; NI9, no interference up to plasma index of 9; PT, prothrombin time.
The calculation of both LOB and LOD was performed as earlier recommended:[5 ] [LOB] = [mean value] + 1.645 × [standard deviation, SD)] of 20 replicates of sample
buffer; [LOD] = [LOB] + 1.645 × [SD] of 20 replicates of a routine citrate plasma
sample with the lowest measurable value of each index (i.e., 1 for all indices). The
linearity was tested using serial dilutions at fixed ratios (1:9; 2:8; 3:7; 4:6; 5:5;
6:4; 7:3, 8:2; 9:1) of a pool of 20 routine citrate plasma samples displaying a high
value of each index (H-index, 8; I-index, 6; L-index, 8) diluted with another pool
of 20 routine plasma samples displaying a low value of each index (i.e., 1 for all
plasma indices). All these dilutions were measured in duplicate and the linearity
was finally assessed as Pearson's correlation coefficient.
The intraassay, interassay, and total imprecision was calculated using three different
pools of 20 citrate plasmas. The pools were specifically selected to obtain low (i.e.,
1 for all indices), intermediate (i.e., between 3 and 4), and high (i.e., between
5 and 8) values for each respective plasma index. The imprecision was measured in
20 consecutive runs for the intra-assay study and 10 runs in 10 working days for the
interassay study, while total imprecision was estimated using the formula of Krouwer
and Rabinowitz.[6 ] Results were finally reported as coefficient of variation (CV%). The interference
limit was finally defined as the value of each plasma index above which the test result
displayed a bias greater than the desirable specification for allowable total error,
as summarized in [Table 1 ].[7 ] This estimation was performed by serially diluting pools of 20 routine plasma citrate
samples displaying high plasma indices values (9 for H-index, 5 for I-index, and 6
for L-index, respectively) with pools of 20 routine plasma citrate samples displaying
the lowest measurable value of each index (i.e., 1 for all plasma indices). Method
comparison was performed using 118 routine citrate plasma samples referred to the
local laboratory for routine hemostasis testing. Plasma hemoglobin (i.e., reflecting
hemolysis) was estimated as H-index on Roche Cobas c501 (Roche Diagnostics AG), as
previously described.[8 ] The performance of this technique was proven to be optimally correlated with the
reference cyanmethemoglobin assay.[8 ] The concentration of triglycerides (i.e., reflecting sample turbidity) and total
bilirubin (i.e., mirroring icterus) in plasma was also measured on Roche Cobas c501,
with the respective reference techniques. The statistical analysis was performed using
Analyze-it (Analyze-it Software Ltd.).
The LOB and LOD was 1 for all plasma indices on Atellica COAG 360. The intra-assay
imprecision was 0% for all indices, the interassay imprecision was 0% for both H-index
and I-index, while it was comprised between 0 and 9.7% for L-index. The total imprecision
was therefore 0% for both H-index and I-index, whereas it was between 0 and 9.7% for
L-index (0% for L-index of 1, 9.7% for L-index of 3, and 8.3% for L-index of 8). The
linearity was 0.993 for H-index, 0.980 for I-index, and 0.978 for L-index, respectively.
According to linearity studies, we estimated that each increment of 1 index unit roughly
corresponded to 0.28 g/L of hemoglobin for H-index, 56 μmol/L of total bilirubin for
I-index, and 3.4 mmol/L of triglycerides for L-index, respectively. These values closely
matched those provided by the manufacturer for hemoglobin (0.37 versus 0.28 g/L) and
bilirubin (74 versus 55 μmol/L), whereas a larger difference was found for triglycerides
(1.4 versus 3.4 mmol/L).
The limit of interference for all reagents tested is shown in [Table 1 ]. Briefly, actin FS was the most hemolysis-sensitive reagent, with clinically significant
variation of values observed with H-index > 3. The other tests displayed intermediate
sensitivity to hemolysis, while fibrinogen, antithrombin, and D-dimer were hemolysis-insensitive
up to H-index of 9. Regarding the I-index, all reagents except Thromborel were found
to be variably sensitive to bilirubin, with fibrinogen values already showing clinically
significant bias with I-index > 2 ([Table 1 ]). Thromborel was the most turbidity-sensitive reagent, since clinically significant
variation of values was observed with L-index > 3. Actin FSL and antithrombin displayed
a clinically significant bias with L-index > 4 and > 5, respectively, while the other
tests were almost lipemia-insensitive up to L-index of 6. Regarding the comparison
of Atellica COAG 360 plasma indices with hemoglobin, total bilirubin and triglycerides
values obtained on Roche Cobas c501, the Pearson's correlations were 0.909 (95% confidence
interval [95% CI], 0.871–0.936) for H-index, 0.922 (95% CI, 0.889–0.945) for I-index,
and 0.964 (95% CI, 0.948–0.975) for L-index, respectively (all p < 0.001) ([Fig. 1 ]). As specifically regards the H-index, Atellica COAG 360 displayed 1.00 sensitivity
and 0.85 specificity compared with Cobas c501 at the upper limit of plasma hemoglobin
concentration found in healthy subjects (i.e., 0.25 g/L).[9 ]
Fig. 1 Comparison of plasma indices measured by Atellica COAG 360 with hemoglobin, total
bilirubin, and triglycerides values obtained on Cobas c501.
Taken together, the results of our study confirm optimal performance of plasma indices
on Atellica COAG 360. The high interassay imprecision of L-index, mainly attributable
to high variability encountered in frozen plasmas displaying L-index > 3, may be considered
virtually insignificant in routine practice, since the threshold for turbidity interference
was found to be > 3 for all reagents tested ([Table 1 ]). We herein conclude that systematic assessment of sample quality through the use
of plasma indices on Atellica COAG 360 is a reliable and viable option for enhancing
the quality in hemostasis testing. As for other coagulation analyzers,[10 ] this study has also allowed estimation of some reagent-specific interference thresholds
([Table 1 ]), which may be applied in routine practice to permit suppressing test results plagued
by clinically unacceptable bias.
