Effect of standardization and normalization on imprecision of calibrated automated thrombography: an international multicentre study
Dr T. Baglin, Department of Haematology, Addenbrooke's NHS Trust, Cambridge CB2 2QQ, UK. E-mail: firstname.lastname@example.org
Calibrated automated thrombography (CAT) enables continuous measurement of thrombin generation (TG). Initial clinical studies using the CAT method showed large variability of normal values, indicating the necessity for a standardized CAT protocol. This international study assessed the intra- and inter-assay imprecision of CAT as well as the inter-centre variability of results in five European centres using locally available reagents and conditions (study 1) and a standardized protocol in which results were normalized (study 2). Samples with and without corn trypsin inhibitor from six healthy volunteers, two haemophilia patients and one protein C deficient patient were assayed. Study 1 confirmed that the use of different sources and concentrations of tissue factor (TF) and different phospholipid (PL) mixtures produced large variability in results. The second study demonstrated that, using the same source and concentration of TF, PL and the same test procedure, this variability could be significantly reduced. Normalization of results improved the inter-centre variability. The benefit of contact factor inhibition prior to TG measurement was confirmed. These results demonstrated that standardization of CAT reduces the variability of results to acceptable limits. Standardization and normalization should be considered in future clinical studies which apply TG testing to clinical decision making.
Thrombin is central to the coagulation process but there is currently no routine laboratory test that can quantitatively measure the thrombin-forming capacity of a plasma sample. Classical clotting tests, such as the activated partial thromboplastin time and prothrombin time assess only time to initiation of clot formation and do not reflect thrombin generation (TG) entirely. Measuring TG by the sub-sampling method was cumbersome and expensive. Over the last 15 years a technique has been developed in which a fluorescent substrate is added to both platelet-poor and platelet-rich plasma samples without defibrination with the course of thrombin formation monitored in real time (Hemker et al, 2002). These technical developments of the TG assay make it potentially applicable to clinical laboratories. Recently, Luddington and Baglin (2004) demonstrated that the clinical measurement of TG by calibrated automated thrombography (CAT) was affected by contact factor inhibition using corn trypsin inhibitor (CTI).
Correlations between the TG test parameters and clinically observed bleeding in patients with haemophilia (Siegemund et al, 2003; Dargaud et al, 2005) and with rare inherited coagulation disorders (Al Dieri et al, 2002) have been published. It has also been shown that CAT is sensitive to hyper-coagulability (Regnault et al, 2004). Measurement of TG is accepted as a research tool but the variety of sources and concentrations of reagents as well as technical constraints limit the potential for clinical use.
The objective of the present international multicentre pilot study was to assess the effect of normalization and standardization on the inter-laboratory variation of CAT as well as intra- and inter-assay imprecision of the test in five centres with proven experience of TG measurement and who have published on the topic of TG using CAT technology.
Materials and methods
Six healthy adult volunteers (three males and three females between 26 years and 53 years of age), an adult patient with protein C deficiency at 55% and a personal history of venous thromboembolism, one adult haemophilia A [Factor (F)VIII = 7 IU/dl] and one haemophilia B (FIX = 4 IU/dl) patient, routinely treated in the Cambridge Addenbrooke's Hospital Haematology Department, were also included after obtaining informed consent.
Peripheral venous blood was collected into S-Monovette® tubes (Sarstedt, Leicester, UK) containing 0·106 mol/l trisodium citrate, loaded or not with CTI 1·45 μmol/l (final concentration in whole blood). Antecubital venipuncture was realized using 18G needle with a light tourniquet. Following a double centrifugation at 3000 g for 15 min at room temperature, platelet poor plasma (PPP) was collected from the upper half volume of plasma supernatant and rapidly frozen at −80°C. The absence of platelets and leucocytes in PPP samples was checked with a Coulter® Gen-STM Hematology Analyser (Beckman Coulter Inc., Fullerton, CA, USA). PPP was always prepared within 30 min of venipuncture. To minimize contact activation, polypropylene tubes and pipette tips were used throughout. All samples were prepared in the same preanalytical conditions in the Department of Haematology, Addenbrooke's NHS Trust, Cambridge, UK. Frozen samples were sent to the four other centres participating in this study (Division of Haematology, National Institute for Biological Standards and Control, Potters Bar, Hertfordshire, UK; Laboratoire d'Hemostase, Hopital Edouard Herriot, Lyon, France; and INSERM U 734, Universite Henri Poincare, Nancy, France; Clinical Hemostaseology, University of Leipzig, Leipzig, Germany).
Design of the study
The present study comprised two parts. In the first part, the inter-laboratory variation of the CAT results was assessed using ‘locally available’ reagents in each centre e.g. tissue factor (TF), phospholipids (PL), buffers and thrombin substrate. In the second part, the inter-centre variation, intra- and inter-assay imprecision of the CAT method were assessed by using a standardized protocol and after normalization of results against the mean of the normal plasma samples.
Calibrated automated measurement of thrombin generation
All centres participating in the study measured TG according to the method described by Hemker et al (2002), in a Fluoroscan Ascent® fluorometer (Thermolab systems OY, Helsinki, Finland) equipped with a dispenser. Fluorescence intensity was detected at wavelengths 390 nm (excitation filter) and 460 nm (emission filter). Briefly, 80 μl of PPP was dispensed into round-bottomed 96 well-microtitre plates. 20 μl of a mixture containing TF and PL was added to the plasma sample. The starting reagent (20 μl per well) contained fluorogenic substrate and CaCl2. A dedicated software program, Thrombinoscope® (Thrombinoscope bv, Maastricht, The Netherlands) enabled the calculation of thrombin activity against the calibrator (Thrombinoscope bv, Maastricht, The Netherlands) and displayed thrombin activity with time.
Thrombin generating capacity was assessed in 18 PPP samples (nine with CTI and nine without CTI) using local reagents and conditions. Each sample was measured using five determinations in each single run over 3 d (three independent measurements) and using separate aliquots on each of the 3 d. Each centre specified the details of reagents used in their study, including respective final reagent concentrations (TF, PL vesicles, working buffer, starting solution buffer, thrombin substrate and calibrator), version of the Thrombinoscope® software and the number and level of experience of the operators. The molar concentration of Innovin® and Thromborel S® (Dade Behring) are not declared by the manufacturer. Therefore, each centre determined the concentrations of TF specified for Innovin® and Thromborel S® using their local method for TF measurement.
Thrombin generating capacity was measured by CAT in separate aliquots of the same 18 samples using the supplied TF, PL, substrate and buffers, as noted above e.g. using five determinations in each single run using two wells for the calibrator and five wells for the samples over 3 d. All reagents and buffers required for TG measurement were centrally prepared in the Department of Haematology, Addenbrooke's NHS Trust, Cambridge, UK and frozen components were shipped to the four other centres participating in this study. Intra- and inter-assay coefficients of variation (CV%) were calculated. Thrombin generation was monitored using only the half microtitre plate for each experiment. Total reading time was 90 min for all experiments and reading frequency was systematically established at every 15 s interval.
CAT reagents used in the standardized protocol
Recombinant human TF, Innovin®, was obtained from Dade Behring (Marburg, Germany) and used at a final concentration of 1·5 pmol/l. TF concentration was determined using the Actichrome® TF activity assay (American Diagnostica Inc., Greenwich, CT, USA). The PL vesicles used at a final concentration of 4 μmol/l, were obtained from Avanti Polar Lipids (AL, USA) and consisted of 20 mol% phosphatidylserine, 20 mol% phosphatidylethanolamine and 60 mol% phosphatidylcholine and were prepared by an extrusion method based upon that of Falls et al (2000). Briefly, PL were combined and the solvent removed by evaporation at 45°C under N2. The lipids were than resuspended in TBS (20 mmol/l Tris, 150 mmol/l NaCl, pH 7·4) and extruded 29 times through a 0·1 μm polycarbonate filter (Glen Creston Ltd, Stanmore, UK). HEPES-buffered saline contained 20 mmol/l HEPES (Sigma Aldrich, Poole, UK), 140 mmol/l NaCl and 5 mg/ml of bovine serum albumin (BSA) (Sigma Aldrich, Poole, UK), pH 7·35. This buffer was stored at −20°C until use. A fresh mixture of fluorogenic substrate and CaCl2 was prepared before each experiment. Fluorogenic substrate, Z-Gly-Gly-Arg-AMC, was obtained from Bachem (Bubendorf, Switzerland). The mixture of 2·5 mmol/l fluorogenic substrate and 0·1 mol/l CaCl2 was prepared using buffer containing 20 mmol/l HEPES and 60 mg/ml of BSA, pH 7·35. The Calibrator with the activity of 600 nmol/l human thrombin was obtained from Thrombinoscope BV (Maastricht, The Netherlands). CTI was purchased from Cambridge Bioscience (Cambridge, UK). Polypropylene round-bottomed Greiner microtitre plates, available in each centre, were used.
Thrombin generation test results as well as the raw data of experiments were collected for statistical analyses. Statistical analysis was performed using the Graph Pad Instat 3·0® software package (San Diego, California, USA). The mean, standard deviation and coefficient of variation (CV %) were calculated for each centre. The inhibitory effect of CTI was evaluated using Wilcoxon test. A P-value of <0·05 was considered statistically significant.
Study 1: thrombin generation measurement using local reagents
A number of differences in local conditions were seen (Table I). Primarily, these involved the source and concentration of TF. In addition, centres 2 and 5 used different PL mixtures, centre 4 an older version of software and centre 5 a lower substrate concentration. In all the centres, the experiments were carried out over 3 d but results were obtained for 2 d only in centres 2 and 3. In centre 2 a problem with the fluorometer caused a loss of data. In centre 3 operator error resulted in a 10 times higher concentration of TF being used on 1 d. The locally-derived reagents exhibited different sensitivities to the hypocoagulable and hypercoagulable samples. This is illustrated in Table II, where endogenous thrombin potential (ETP) values are expressed as a ratio against the mean ETP of the normal group. The reagent combinations used by centres 1 and 5 were particularly sensitive to the hypercoagulable sample. The combination used in centre 4 showed greatest sensitivity to the hypocoagulable samples (Table I and II– study 1). The addition of CTI improved sensitivity to the hypercoagulable sample at the very low-TF concentration used by centre 4 and the reagent combination used by centre 1 (Table I and II– study 1). This finding was not seen in the other centres and had no effect upon the sensitivity to the hypocoagulable samples (Table II– study 1). The variability of results was assessed by the calculation of inter-centre CV values.
Table I. Study 1 – comparison of endogenous thrombin potential (ETP) values from each sample with or without corn trypsin inhibitor (CTI) and inter-centre coefficients of variation (CV) values obtained using locally available reagents and protocols. ETP values shown as mean of n = 15 (standard deviation; SD) for each tested sample.
|Mean N1–N6 (mean ± SD)||1312 (343)||897 (205)||2954 (535)||2909 (462)||889 (152)||803 (150)||1194 (325)||494 (133)||768 (194)||810 (224)||62||83|
|Protein C: 55%||2629 (204)||2175 (184)||2963 (260)||3036 (222)||1130 (36)||1005 (85)||1635 (153)||892 (155)||1602 (179)||1546 (181)||39||51|
|FIX: 4 IU/dl||409 (87)||304 (43)||2202 (251)||2169 (138)||222 (30)||209 (37)||56 (17)||28 (13)||1163 (118)||1163 (121)||109||115|
|FVIII: 7 IU/dl||521 (122)||334 (51)||1695 (84)||1764 (173)||202 (26)||180 (31)||114 (22)||63 (5)||668 (81)||566 (121)||99||118|
|Tissue factor (final concentration)||Thromborel S-Dade Behring (2·5 pmol/l)||Innovin–Dade Behring (5 pmol/l)||Innovin–Dade Behring (1 pmol/l)||Innovin–Dade Behring (0·5 pmol/l)||Innovin–Dade Behring (0·56 nmol/l)|| |
|Phospholipids (final concentration)||PC 60-PS 20-PE 20 mol% (4 μmol/l)||PC20-PS13-PE45-PA22 mol% (4 μmol/l)||PC 60-PS 20-PE 20 mol% (4 μmol/l)||PC 60-PS 20-PE 20 mol% (4 μmol/l)||Pathromtin SL–Dade Behring diluted at 1:2500|| |
|Thrombin substrate (final concentration)||Z-Gly-Gly-Arg-AMC Bachem (2·5 mmol/l)||Z-Gly-Gly-Arg-AMC Bachem (2·5 mmol/l)||Z-Gly-Gly-Arg-AMC Bachem (2·5 mmol/l)||Z-Gly-Gly-Arg-AMC Bachem (2·5 mmol/l)||Z-Gly-Gly-Arg-AMC Bachem (1·96 mmol/l)|| |
|Microtitre plates||Greiner U650201||Greiner U 650101||Greiner U 650101||Greiner U650204||Greiner U650201|| |
Table II. Comparison of the ratio of mean endogenous thrombin potential (ETP) for the pathological samples against the mean ETP derived from the normal group.
| Protein C: 55%||2·00||2·42||1·00||1·04||1·27||1·25||1·36||1·80||2·08||1·90|
| FIX: 4 IU/dl||0·31||0·33||0·74||0·74||0·24||0·26||0·04||0·05||1·51||1·43|
| FVIII: 7 IU/dl||0·39||0·37||0·57||0·60||0·22||0·22||0·09||0·13||0·86||0·69|
| Protein C: 55%||1·47||1·53||1·46||1·52||1·47||1·63||1·40||1·55||1·32||1·47|
| FIX: 4 IU/dl||0·44||0·45||0·52||0·43||0·36||0·38||0·31||0·35||0·5||0·46|
| FVIII: 7 IU/dl||0·33||0·35||0·43||0·39||0·26||0·25||0·31||0·30||0·46||0·41|
Study 2: thrombin generation measurement using a standardized protocol
The only differences between centres were the use of an old version of the software in centre 4 and different operators in each centre. In centre 3, each study day was operated by a different laboratory assistant. The experiments were carried out over 3 d in all centres. Table III summarizes the standardized CAT conditions used by all participating centres. In standardized conditions, ETP results obtained with and without contact factor inhibition were compared. In samples taken in CTI, ETP values were significantly lower (P < 0·001) than in the citrated samples. To assess the effect of normalization, results were normalized against the mean ETP value of the normal plasma samples (N1–N6). Table III shows significantly improved inter-centre CV values in standardized conditions, after normalization. The sensitivity to the pathological samples is shown in Table II– study 2, and the intra-assay CV values for each pathological sample in each centre are shown in Table IV.
Table III. Study 2 – comparison of endogenous thrombin potential (ETP) values from each sample with or without CTI and inter-centre CV values obtained using standardized reagents and protocols and after normalization.
|Mean N1–N6 (mean ± SD)||1835 (213)||1719 (153)||2025 (407)||1853 (486)||958 (256)||829 (271)||1396 (219)||1200 (210)||1966 (237)||1690 (292)||0 (0*)||0 (0*)|
|Protein C: 55%||2702 (111)||2633 (72)||2973 (274)||2835 (213)||1417 (104)||1357 (146)||1959 (67)||1860 (64)||2598 (119)||2426 (115)||4·8 (5·9*)||4·6 (3·2*)|
|FIX: 4 IU/dl||817 (65)||778 (50)||1069 (586)||805 (63)||347 (70)||319 (67)||433 (53)||424 (68)||983 (155)||792 (77)||21·6 (9·3*)||11·5 (4·4*)|
|FVIII: 7 IU/dl||608 (86)||606 (66)||887 (364)||740 (101)||253 (70)||212 (50)||444 (46)||370 (34)||919 (122)||689 (140)||23·2 (17·8*)||18 (7·5*)|
| Tissue factor (final concentration)||Innovin – Dade Behring (1·5 pmol/l)|
| Phospholipids (final concentration)||PC60-PS20-PE20 mol% (4 μmol/l)|
| Thrombin substrate (final concentration)||Z-Gly-Gly-Arg-AMC – Bachem (2·5 mmol/l)|
| Thrombinoscope||V22.214.171.124||V126.96.36.199||V188.8.131.52||V2.106||V184.108.40.206|| |
Table IV. Intra-assay coefficients of variations for the three pathological samples measured in the presence or absence of CTI.
| Protein C: 55%||4·6||2·7||7||4·6||2·9||7·9||4·2||11·9||2·8||2·7|
| FIX: 4 IU/dl||12·9||10·6||9·6||6·5||6·3||9·5||21·6||35||3·5||3|
| FVIII: 7 IU/dl||14·6||4·4||3·5||8·8||4·8||7·9||17·8||7·6||2·5||2·6|
| Protein C: 55%||3·2||2·6||4·4||6·4||3||2·9||1·4||2·7||4·3||4|
| FIX: 4 IU/dl||7·7||5·5||18·8||6·3||5·8||3·5||5·1||2·5||4||8·4|
| FVIII: 7 IU/dl||6·7||9·1||20·4||8·3||7·2||5·5||8·4||4·6||13·3||5·2|
A review of previously published clinical studies using CAT indicated a large variability in results; dependent on reagents and methods, making comparisons between studies difficult (Baglin, 2005). In the first part of this study the variety of sources and concentrations of TF produced a significant variation in results. It has been previously shown that the TG test could detect hypo- (Al Dieri et al, 2002; Dargaud et al, 2005) and hyper-coagulability (Luddington & Baglin, 2004; Regnault et al, 2004). It was not the primary purpose of this study to verify these data but when comparing results from pathological samples obtained with very different TF concentrations it was noticed that sensitivity was TF concentration dependent. As the TF concentration decreased, the discrimination of hypocoagulable samples from normal increased (Table I and II– study 1). Discrimination of the hypercoagulable protein C deficient sample was not as obviously TF dependent.
The combination of CTI and 0·5 pmol/l TF gave the clearest differentiation of the hyper- and hypo-coagulable samples from the normal group but made the assay more sensitive to contact activation. In centre 2 and 5, which used high-TF concentrations, the inhibitory effect of CTI was minimal (Table I and II– study 1). This supports the previous finding that, at TF ≥ 5 pmol/l, TG is less influenced by contact factor activation.
Inter-centre variability was reduced in the second study using a standardized protocol. ETP values obtained in normal controls and patients were in agreement with previously published data in thrombophilic (Luddington & Baglin, 2004) and haemophilia patients (Dargaud et al, 2005) and abnormal samples were correctly detected (Table III and II– study 2). The comparison of ETP results obtained with and without contact factor inhibition showed very significantly lower thrombin generating capacity in the presence of CTI (P < 0·001). This result confirms previously published data by Luddington and Baglin (2004), showing the necessity for using CTI in a low-TF-triggered TG measurement without concern for the effect of interference from in vitro contact activation. Even these carefully prepared plasma samples showed a variable degree of in vitro contact activation (intra-assay CVs 1·4–20·4%) (Table IV– study 2). This was abolished using 1·45 μmol/l CTI with intra-assay CV values falling below 9·1% in pathological samples. This is in agreement with the 3–5% intra-assay CV reported by Hemker et al (2003) and <9·5% reported in the multicentre assessment of Lawrie et al (2003).
The inter-assay variability of CAT with the standardized protocol and CTI 1·45 μmol/l was between 1·2% and 39·9% (data not shown). The highest variance was seen in centre 3 where different, relatively inexperienced, laboratory staff conducted the experiments on each day. Normalization eliminated this variance (Table II– study 2).
Another source of variability with the standardized protocol was the version of Thrombinoscope software used for measurements. Only centre 4 used an older version of the software (Table I and III) and despite the accuracy of their results, there was a variance of results in comparison with centres 1, 2 and 5 which used exactly the same test conditions (Table III). This difference was again eliminated following normalization (Table II– study 2). The manufacturer confirmed that the difference could be explained using version of Thrombinoscope® software because older versions of Thrombinoscope® had different coefficients of correction used for the calculations in comparison with later versions (V3).
Our results showed a significant improvement of inter-centre CV values using normalization and a standardized CAT protocol (4·4–23·2% compared with 39–109% in study 1) (Table I and III). Further reduced imprecision was also seen in the presence of CTI 1·45 μmol/l (4·6–18% compared with 51–118% in study 1) (Table I and III). Comparison of the three centres with exactly the same conditions and software (centres 1, 2 and 5) showed an inter-centre CV value below 7·5% (Table III).
In conclusion, study 1 confirmed a large variability of ETP values between centres using different sources and concentrations of TF and different PL mixtures. Results were not comparable and multicentre clinical studies using CAT would not be achievable. However, study 2 demonstrated that, using the same source and concentration of TF, PL and the same test procedure, this variability could be significantly limited. In addition, under these conditions, contact factor inhibition was shown to improve the intra-assay CV. With a simple normalization procedure the inter-centre variability could be further significantly reduced. The results emphasized that experienced operators and the use of the same version of the software are required to obtain similar results with the same plasma sample. The results also emphasized the requirement for a standardized protocol using standard reagents and a reference plasma for normalization before organizing multicentre clinical trials to assess the potential predictive value of CAT for clinical outcomes before this method is used more widely in clinical laboratories.
None of the authors have any conflict of interest. Specifically, none have received funding from Thermo Lab systems OY, Thrombinoscope Bv, Biodis or Haematologic Technologies Inc.