CLIN. CHEM. 32/10, 1857-1862 (1986)

CLINICAL CHEMISTRY, Vol. 32, No. 10, 1986 1857

Turbidimetric Determination of Prothrombin Time by Clotting in a CentrifugalAnalyzerG. 0. Gogstad,1 K. H. Dahl,1 A. Christophersen,2 and A. Bjerke1

Two thromboplastin reagents (“Thrombotest” and “Normo-test Automated”) were used in evaluation of an automatedmethod for determination of prothrombin time based onturbidimetric measurement of clot formation in a centrifugalanalyzer. We used 60 plasma samples from patients withvarious diseases or being treated with oral anticoagulant and16 normal plasma samples. Prothrombin times were calculat-ed by a computer connected to the analyzer, a reading beingmade at either a certain per cent increase in total absorbanceor a fixed absorbance increase. Both correlated well with themanual method (r = 0.98-0.99). The reading points bestfitting the manually obtained data were estimated by minimiz-ing the residual sum of squares in regression analysesperformed at various absorbance increases. The per centreading was better in this respect. Normotest Automatedcould be nearly perfectly related to the manual method,whereas Thrombotest showed a (negligibly) small deviation.Reproducibility was good within run (CV �3.2%) as well asbetween batch of the reagents, as assessed from variation inINR (CV �4.9%). We conclude that turbidimetry of clotformation may be validly used in automation of the prothrom-bin-time test. The equipment needed and the total time peranalysis are about as for chromogenic substrate methods,but reagent cost is considerably lower.

Addftlonal Keyphrases: coagulation liver function moni-toring therapy with anticoagulants manual method comparedCobas Blo

The prothrombin-time test is widely used for monitoringof oral therapy with anticoagulants and for screening ofliver function. The basic test system involving thromboplas-tin and Ca2� for activation of the extrinsic coagulationcascade, introduced by Quick (1), was later modified step bystep until a combined reagent, insensitive to variations inFactor V and fibrinogen, was introduced by Owren (2).

The test is originally based on measuring the time to form

a visible fibrin clot. Initially, the test was performed by themanual tilt-tube technique, but it soon appeared that auto-mation was desirable, to accommodate the increasing num-ber of tests. Various mechanically based instruments havebeen introduced, but the degree of automation has not beensatisfactory. In later years, investigators have presentedvarious reagents for monitoring prothrombin time, in whichthe clotting reaction is replaced by a chromogenic substrate

for thrombin (3-9). A similar principle has also been pre-

sented for the activated partial thromboplastin time test (9,10). Such methods have been claimed to be the ones of choice

‘Hemostasis Research Section, NYCOMED AS, Oslo, Norway.2lnstitute of Nutritional Research, University of Oslo, Norway.

Address correspondence to: G. 0. G., at NYCOMED AS, Box4220-Torshov, 0401 Oslo 4, Norway.

Received March 14, 1986; accepted July 8, 1986.

for automation of the overall coagulation assays. However,clot formation per se is accompanied by an absorbancechange that is readily measurable spectrophotometrically.This principle has also been investigated for automation onvarious kinds of analyzers (11-14).

In the present study we more closely investigated automa-tion of prothrombin-time testing based on spectrophotomet-ric detection of clot formation.

Materials and Methods

Materials

Instrumentation. We used a Cobas Bio centrifugal analyz-er linked to an IBM Personal Computer through a RS 232 CInterface, and a “Cobpoll handshake” program (NYCOMEDAS, Oslo, Norway).

Reagents. Two combined thromboplastin reagents weretested: “Thrombotest” contains bovine-brain thromboplas-

tin; “Normotest Automated” contains rabbit-brain thrombo-plastin. Both are from NYCOMED AS.

Procedures

Test procedures. Blood from normal persons or patientswas anticoagulated with citrate and plasma was separated

by centrifugation. Each plasma was analyzed in duplicateby the manual tilt-tube technique and in duplicate by ourautomated method, in the centrifugal analyzer. Two sepa-rate runs on the analyzer were performed.

Each vial of Thrombotest was reconstituted with 11 mL ofa 3.2 mmol/L solution of CaC12 and each vial of Normotest

Automated was reconstituted with 10 mL of distilled water.The contents of several vials were pooled before the analyseswere started.

The manual tilt-tube technique for the two reagents wasperformed as follows: An aliquot of each plasma was diluted6:4 with 0.15 moIIL NaC1. We incubated 250 pL of eachreconstituted reagent in test tubes at 37#{176}Cfor 5 mm. ForThrombotest, 50 �L of diluted plasma was added to each testtube, mixed immediately with the reagent, and the stop-

watch was 8tarted. The time to form the clot was recordedafter it was observed on gently tilting the tube in the waterbath. For Normotest Automated the procedure was identicalexcept that 25 �L of diluted plasma was used.

Measurements in the centrifugal analyzer were per-formed according to the same principle. The instrument wasloaded with reconstituted reagent and undiluted plasmasamples. The dilution of samples in the manual method wascompensated for by addition of a smaller volume of plasmaand addition of diluent by the instrument. Instrumentsettings are listed in Table 1.

Computer programs. The system consists of two differentprograms, one polling (foreground) and one computing (nor-mally background) program.

The polling program continually checks the RS232C

TIMEFig. 1. Clotting reaction curveThe lines fitted to the curve describe the bas&lne (lag-phase), the steepest partof thereactioncurve, and the upper plateau of the reaction (absobanoe of dotted

1858 CLINICAL CHEMISTRY, Vol. 32, No. 10, 1986

Table 1. Programs Used In the Cobas BI0 CentrIfugalAnalyzer

Thrombotest Normotest

1. Units own own2. Calculation factor 1.000 1.0003. Standard 1 concn 0 04. Standard 2 concn 0 05. Standard 3 concn 0 06. Limit 0 07. Temperature, ‘C 37.0 37.08. Type of analysis 5 59. Wavelength, nm 360 360

10. Sample vol, �L 30 1511. Diluent vol, �L 20 2012. Reagent vol, �L 250 25013. Incubation time, s 0 014. Start reagent vol, �L 0 015. Time of first reading (s) 1.0 1.016. lime interval (s) 10 1017. Number of readings 30 1518. Blanking mode 119. Printout mode 3.

interface for any incoming characters, If any, it is checkedfor STX (ASCII 2). When this character has entered theRS232C, the subsequent incoming characters are stored in abuffer until ETX (ASCII 3) is achieved. A ifie with a ifie-name made up by a user-defined six-letter code plus a two-

digit counter is created in the A-drive. The informationstored in the buffer is sent to the file and the file-name isstored in a message ifie and appended to a report file. TheCPU is then given to the computing program and results arecomputed and written to the printer before polling is contin-

ued.Computing of clotting time. For each reaction, three lines

(ofthetypey = a + bt)arefittedtothedatabyuseoflinearregression (Figure 1). The first line, y = a0 + b0t is fitted

through the data representing the ground line (lag-phase).The next line is fitted through the data representing thesteepest part of the reaction curve, y = a,,,� + b,,�t and thethird line, y = �nd + bej is fitted through the data

representing the upper plateau of the reaction (absorbanceof clotted plasma).

From a theoretical point of view, clotting time can becomputed according to two basic principles:

Method 1: computing the time to reach a fixed increase inabsorbance (M) from the baseline (“absorbance method”).The total absorbance change occurring during clot forma-

w0z

0

0)

‘C

tion may vary considerably. Consequently, this method maybe hampered by variable relationship between a certain

absorbance change (LA) and physical formation of the clot.From Figure l#{192}it can be seen that the clotting time PTcorresponding to a certain LA can be computed by use of theequation

PT = � - (ao + L4W(b0 - b,,,�) (1)

Method 2: computing the time to reach a certain percent-age of the total absorbance change (“ratio method”). Thismethod may be less dependent on total absorbance change,

but may depend on variations in kinetics of clot formation.From Figure 1, it can be seen that the absorbance change

LA corresponding to a certain relative absorbance increase

can be computed from the equation

LA = %M[a� + bend(ai,..� - aond)/(bend - bmax) - (80 +

bo(am�x - a0)/bo - b�,�)] (2)

and, by using equation 1 and 2 together, the clotting timecan be calculated.

The behavior of these two methods for determination ofprothrombin time formed part of our investigation and isdescribed in the Results section.

Computing the Best i�A and %z�A for a Given Set of

Automated and Manually Observed Clotting Times

For regression analysis of the unknown parameters inequations 1 and 2 we used the regression analysis program

“BMDPAR” (© 1981, Regents of University of California),which has no need of determining analytically the deriva-tive of the equation with respect to the unknown parameters

to be estimated.The following regressions were performed:

1. Regression with only equation 1 used in determiningLA.

2. Regression with equational and 2 used in determining%LA.

The optimum LA and %LA were estimated by minimizingthe residual sum of squares.

Plasma Samples

Samples were collected from a randomized patient materi-al that included patients with liver diseases, thromboembol-ic disorders, bleeding conditions, and patients on long-termstabilized therapy with oral anticoagulant. In addition tothese patients, we collected a group of samples from non-medicated normal persons.

We measured 61 patients’ plasmas and 16 normals withThrombotest, and 56 patients’ plasmas and 16 normals withNormotest.

Calculation of INA

The INR (International Normalized Ratio) is the recom-mended way of expressing the coagulation activity of aplasma (15, 16). Strictly speaking, the INR is valid only forfresh plasmas from patients on stabilized oral anticoagula-tion therapy. However, this way of expressing the coagula-tion activity of a plasma is useful for comparing various lotsof the thromboplastin reagent used. For Thrombotest, theINR is calculated according to the formula INR = (P’rR)�’,in which P’FR is the ratio of prothrombin time for thepatient to prothrombin time for normals, and 181 (Interna-tional Sensitivity Index) is a factor, unique to each lot of thereagent, specified by the manufacturer.

Normotest Automated does not fit the same calculation

WAVELENGTH (nm)

Table 2. CharacteristIcs of the Overall Optimal Automated Reading Points, and the Relationship between theCorresponding Prothrombin Times (x) and the Prothrombin Times Obtained with the Manual Method (y)

Thrombotest Normotest Automated

Absorbance method

y = 1 .04x - 1.830.9870

A = 0.13 ± 0.0048.83

1271

CLINICAL CHEMISTRY, Vol. 32, No. 10, 1986 1859

model and the INR is calculated according to an alternativemethod (17): INR = (PTR)a. 10d, in which a and d arefactors that are unique to each lot of the reagent, and aregiven by the manufacturer.

Results

Absorption Spectra

So we could visualize the absorbance change during theclotting reaction, thromboplastin reagent and plasma weremixed and the reaction was followed over the spectral range300-600 nm in a Hewlett Packard 8451A Diode ArraySpectrophotometer.

Figure 2 shows the absorption spectra arising in thecourse of the reaction with use of Thrombotest, and the insetto Figure 2 shows the corresponding differential spectraobtained by subtracting from each spectrum the spectrummeasured before the reaction started. The reaction is typicalfor a turbidimetric alteration: increasing absorbance changewith shorter wavelengths. There is a general absorbanceminimum at about 380 nm. A somewhat shorter wavelength(360 nm) was chosen for the measurements, as a compro-mise between lowest background absorbance and high ab-sorbance change during clotting. Measurements made withuse of Normotest Automated gave similar results.

Optimization of the Automated Assay

We compared results of the two automated calculationmethods with the manually obtained data sets, using linearregression analysis. The correlation coefficients obtained forvarious percentages of absorbance increase and for variousfixed absorbance increases were similar. The correlationwas about the same with the two calculation procedures, butslightly better for Thrombotest than for Normotest Auto-mated (Table 2).

Fig. 2. Absorption spectra obtained in the course of the clotting reaction

(with use of Thrombotest)The numbers on each curve indicate seconds after mti�ng plasma and thrombo-plastin reagent. Inset: corresponding difFerence spectra obtained by subtractingthe spectrum of the initial reaction mixture from each of the other spectra

We evaluated the results as to what extent each of the two

calculation principles could give results identical with thoseobtained by manual readings. In the ideal situation, the line

of regression between the automated and the manuallyobtained data sets should of course have a slope of 1.0 and asimultaneous intercept at zero.

The best reading levels for the two calculation principlesand the two thromboplastin reagents were approached byusing an iterative data program, searching for the leastresidual sum of squares. The optimal reading levels and thecorresponding data of the regression analyses are presentedin Table 2. The plots are shown in Figure 3A-D. Twoseparate runs on the centrifugal analyzer gave nearly thesame result. It can be seen from the table that results by theratio method corresponded to those by the manual readingmethod more closely than did the absorbance method.Generally, Normotest Automated approached the manualmethod better than Thrombotest did. Indeed, the ratiomethod applied on Normotest gave a nearly perfect regres-sion to the manual data.

Influence of Reading Method on Interpretation ofCoagulation Activity of a Plasma

The prothrombin time for the normal plasma and thesensitivity coefficients (“ISI” for Thrombotest and “a” and“d” for Normotest) are characteristics of each batch of theproducts and the methods applied to them. Thus, the man-ual method is described by a set of these factors for eachbatch. If one uses these factors in the automated methodolo-gy, assuming that this fits the manual behavior of thereagent, a certain variation in the INR calculated for eachplasma may arise. This situation actually occurs, becausethe products in question have supplier-presented valuesthat are based on the manual method. If the user changes toan automated system, the calculation of each plasma’sactivity in terms of INR or per cent would be simplified ifthe methodology followed that of the manual technique.

The consequence for interpretation of the plasma coagula-tion activity by using the automated method and themanual curve was ruled out. The variation in INR for threetheoretical levels of activity was calculated as a function ofvariation in the reading point. The three curves shown ineach diagram of Figure 4 represent the activity levels of anormal plasma (target INR = 1.0), a plasma in the uppertherapeutic level (target INR = 2.8), and a plasma in thelower therapeutic level (target INR = 4.8). The target 1NR-values are shown by arrows in each diagram. It can be seenthat only for Normotest Automated measured with the ratiomethod does there exist a reading point for which all threetarget INR values may be reached simultaneously. All theother diagrams show variations from the target INR at somelevel. For Thrombotest the lower therapeutic level demon-strates an INR of 4.6 at the overall optimal reading point onusing the ratio method, i.e., an error of 0.2 INR unit. The

Linear regression equationCoeff. of correlationOptimal reading point (± SD)Mean square of errorsResidual sum square of errors

Ratio method

y= 1.07x- 6.00.9941

53.5% ± 1.3%21.05

3641

Absorbanco method

y= 1.llx- 8.640.9922

A = 0.31 ± 0.0132.49

5620

Ratio method

y= 1.OOx-0.060.9857

30.7% ± 0.9%9.22

1327

A

200

C’)

-J

z

B /.

.

S

.

C

50 100 150 200

C

250

125

100

75

50

25

125

100

75

50

25

21

200

150

D

Cl)

-J

z

25 50 75 100 125

AUTOMATED (S)

S

.S

S

.

50 100 150 200 250

4

25 50 75 ioo 125

3

z2

4

3

z

AUTOMATED (5) AUTOMATED (S)Fig. 3. Correlation between automated and manual methodsThe automated values are obtained at the calculation levels that gave overall results most Identical to manual values. A. Thrombotest, ratio method; B. Normotest, ratiomethod; C. Thrombotest, absorbance method; D. Normotest, absorbance method

A5 -- B5

1 1

25 50 75 100 25 50 75 100 0.2 0.4 0.1 0:2

% OF TOTAL ABSORBANCE CHANGE ABSORBANCE INCREASE

Fig. 4. Influence on clinical interpretation of automatically obtained prothrombin times In terms of International Normalized Ratios (INR) when thedata for various reading levels are applied to manually observed characteristics of the thromboplastin reagentsA, ratio method; B, absorbance method. Left: Thrombotest; right: Normotest. Three theoretical INR of 4.8 (upper curves), 2.8 (middle cur/es) and 1.0 (lowercurves), respectively, were chosen. The arrows Indicate the level of the target INR: Each INR corresponds to a certain manual prothrombin time for eachthromboplastin. For each automated reading level (horizontal axes) a lInear-regressIon equation between automated and manual values was calculated. Themanual prothrombln times corresponding to each of the INR may be converted to automated prothrombln times, which in tom correspond to INR values on thelNRiprothrombin-time scale. These latter values are plotted on the diagrams and describe the error from target INR at various reading levels

2

1860 CLINICAL CHEMISTRY, Vol. 32, No. 10, 1986

AUTOMATED CS)

error per 10% deviation in per cent absorbance increase(ratio method) is about 0.1 INR unit. For the absorbancemethod applied to Thrombotest the lower therapeutic leveldemonstrates an INR of 4.5 at the overall optimal reading

(I)

-J

z

C/)

-J

z

point. Each 0.1 absorbance-unit deviation corresponds to achange in INR of about 0.15. For Normotest Automated, theINR change is about 0.4 for each 0.1 A deviation from theoptimum as measured with the absorbance method. The

Thromboteit

SD CV,%

Plasma

CLINICAL CHEMISTRY, Vol. 32, No. 10, 1986 1861

overall optimal reading point gives an INR of about 4.7.It can be seen from Figure 4A that an optimal reading

point for the therapeutic range as measured with Thrombo-test corresponds to a reading at 60-70% absorbance in-crease.

Reproducibility

A set of plasmas with various coagulation activities waschosen for the study of reproducibility. Table 3 shows theresults. The CV was in the range 0.5-3.2% for Thrombotestand 0.7-2.9% for Normotest Automated when the ratiomethod was used. Repetition of this series of readingsshowed a similar CV for each plasma. Thus this variationseems to be a property of each plasma, and also apparentlyis independent of the total absorbance increase.

We measured the same set of plasmas, using five different

batches of Thrombotest and four batches of NormotestAutomated. The batches were different with respect toclotting times and sensitivity and covered the normal varia-tion in available batches of the products. Each batch wastested in a separate run. To make the results comparable,we converted all the prothrombin times to INR, using thecharacteristic normal prothrombin time and descriptivefactors “ISI” (Thrombotest) and “a” and “d” (NormotestAutomated) for each batch. The results are shown in Table4. The CV was 0-3.7% for Thrombotest and 0-4.9% forNormotest Automated. Because the batches were run sepa-rately, the results also reflect variation between runs.

Discussion

This study is based on the change in absorbance thataccompanies the change of soluble fibrinogen to fibrinpolymers as a final step in plasma coagulation. Hence, if thecourse of this change may be related to the coagulation

Table 3. Reproducibility of Automatically ObtainedProthrombin Times (s) within Run, Calculated by the

Ratio Method a

Normotest Automated

SD CV,%

36.2 1.2 3.2 25.5 0.3 1.342.3 0.2 0.5 24.6 0.5 1.965.4 0.4 0.7 45.5 0.6 1.477.1 1.1 1.4 56.1 1.3 2.387.1 0.5 0.6 56.3 0.5 0.8

100.0 2.9 2.9 56.9 1.6 2.8105.0 2.7 2.5 67.9 2.0 2.9136.9 1.0 0.7 87.2 0.6 0.7

Values are for eight plasma samples, n = 12 for each.

Table 4. ReproducibIlity of INR Determination betweenBatch of “Thrombotest” and “Normotest Automated,”

with Use of the Ratio Method of Calculation and theAutomated Method of Reading.

Thrombotest (n = 5) Normotest Automated (n = 4)

no. SD CV, % Range SD CV, % Range

1 1.01 0.04 3.7 0.96-1.05 1.06 0.01 0.8 1.05-1.072 1.11 0.02 2.1 1.07-1.13 1.20 0.06 4.9 1.17-1.293 1.83 0.03 1.7 1.80-1.88 2.30 0.05 2.2 2.2-2.34 2.26 0.06 2.3 2.3-2.4 2.65 0.10 3.8 2.6-2.85 2.76 0.06 2.4 2.7-2.8 3.28 0.05 1.5 3.2-3.36 2.80 0 0 - 3.20 0.08 2.6 3.1-3.37 2.98 0.08 2.8 2.9-3.0 3.43 0.05 1.5 3.4-3.58 4.72 0.05 1.0 4.7-4.8 5.20 0 0 -

activity of the plasmas, fibrin formation serves as a naturalvariable for use in the spectrophotometric measurement of

prothrombin time.Two basic principles for calculating prothrombin time

from absorbance change during clot formation have beeninvestigated in this study. The “ratio method” takes intoaccount possible variations in total absorbance change andis based on measurement at a certain per cent of totalabsorbance increase. The “absorbance method” measures acertain fixed absorbance change, assuming that this reflectsa fixed amount of fibrin formed in each plasma. Ourresults show that results by the two methods correlate wellwith the manual method with either of the two thrombo-

plastin reagents investigated. The correlation coefficientchanges negligibly with the levels of reading, which makethe considerations in the following discussion independentof this factor.

The approximation of results by the automated methodsto those by the manual techniques is important, to simplifythe general applicability of the automated reading. Thethromboplastin reagents we investigated are both presentedwith batch-wise characteristic factors obtained by the man-ual tilt-tube method. The manual technique has been cho-sen in these calibrations in order to avoid the influence ofvariations in mechanically based instruments. The manualmethod is also the method used for calibration of theinternational reference thromboplastins (15, 16). Thus, ifresults by the automated method correspond satisfactorilywith those by the manual technique, the coagulation activi-ty of each plasma is easily calculated from the characteristicdata supplied by the manufacturer for each batch of thesereagents.

Generally, the ratio method correlates with the manual

method better than does the absorbance method. The reasonfor this may be that the variation in total absorbancechange influences the results if a fixed absorbance is used.In fact, a variation in total absorbance change of as much as0.4-1.0 A was seen for Thrombotest and 0.3-0.7 A forNormotest Automated. These variations were obviously inpart due to lipemic plasmas, but other factors also areprobably involved. The ratio method thus appears to obviatethe problem of lipemic plasmas and probably also otherartifacts reported for turbidimetric methods (18). Because

the thromboplastin reagents both contain an excess offibrinogen, the variations could not be due to variation inthe amount of this factor in the plasmas. The optimalreading points were also different for the two thromboplas-tin reagents. The reason for this is not readily explained,and if the automated method is used with other reagents, aoptimization similar to the one performed in this studyappears to be necessary. The optimal reading point of 53.5%for Thrombotest is very close to the reading point used inanother study (13). These investigators found that readingat 50% absorbance increase gave reliab� results, but theydid not perform any optimization studies. The regression

equation when reading in a mechanically based clottingtimer (Amelung) was also less favorable than the optimalequation of our study.

Whereas Normotest Automated may be fitted nearlyperfectly to the manual method, Thrombotest shows v�ria-tion depending on the level of coagulation activity. Optimi-zation with respect to the therapeutic range would make65% of the absorbance increase a more reliable readingpoint than 53.5%. However, in spite of these variations,their practical influence on interpretation of the results is

1862 CLINICAL CHEMISTRY, Vol. 32, No. 10, 1986

small. An error of 0.2 in INR appears to be less critical ontaking the therapeutic range of INR = 2.8-4.8 into consider-ation. Furthermore, the variation in the determination ofINR between laboratories, under well-controlled conditions,has been reported with a CV of 11-13%. The variation forThrombotest is about 4% at the lower therapeutic level.With respect to interpretation of the results, the automatedmethod generally appears to be quite flexible. Variations of10% from the optimal absorbance change give INR changesof 0.1 and 0.2 at the lower therapeutic level for Thrombotestand Normotest, respectively. This must be considered lesscritical for the clinical interpretation of the results.

The within-run reproducibility of the method was good, asis that between batch of thromboplastin reagents. Theplasma samples we used also covered a great variety ofdiseases without any systematic influence on the behavior ofthe automated vs the manual method. The results thusunderline the general applicability of the automated meth-od.

The study as a whole demonstrates that by carefuloptimization of the reading method the prothrombin timeassay may be automated on a spectrophotometric basis. Inprinciple, this gives the clotting method access to all kinds ofspectrophotometrically based analyzers.

The use of fibrinogen as substrate for thrombin hascertain theoretical advantages over chromogenic substrates.First, fibrinogen is part of the total reaction to be monitored,and may influence the result by modulating the clottingfactors. An example of this is the immobilization of throm-bin on the fibrin polymer (19), which may alter the reactionkinetics. Second, chromogenic substrates may inhibit cer-tain coagulation factors (20), which makes the chromogenictype of assay sensitive to certain factor variations in a less-reliable way. Third, accumulated clinical experience basedon the clotting-type reactions may be continuously fullyutilized. Finally, both principles of automation requireadvanced calculation and hence a computer and a dataprogram to serve the analyzer. In either case, the time spentwill be more or less the same based on the need forpreparation of plasmas, loading the machine, and analysistime. The main difference will be a more economical assay

that obviates the cost of chromogenic substrates. By carefulinstrument construction the present method also opens theway for automated centrifugal analyzers, which may mea-sure whole-blood samples.

References

1. Quick A,J. The prothrombin in hemophilia and in obstructivejaundice. J Biol Chem 1935;109:LXXffl.2. Owren PA. Thrombotest. A new method for controlling anticoag-ulant therapy. Lancet 1959;ii:754.

I,

3. Becker U, Jering H, Roschlau P. Prothrombin time test with achromogenic substrate as thrombin indicator [Abstract]. Haemoeta-sis 1982;12:73.4. Jonker JJC, Klarenberg RA, Jilek F, Moeller G. Evaluation of anovel automated prothrombin time (PT) test on the base of achromogenic substrate [Abstract]. Haemostasis 1982;12:53.5. Becker U, Barti K. A modified prothrombin time test with achromogemc peptide substrate as a thrombin indicator� Applicationto the Cobas Bio centrifugal analyzer [Abstract]. Thromb Haemo-stasis 1983;50:236.6. Jonker JJC, Klarenberg JA, Jilek F. Experience with twoautomated chromogenic substrate assays in controlling oral antico-agulant therapy [Abstract]. Thromb Haemostasis 1983;50:236.

7. Becker U, Jering H, Bartl K, Jilek F. Automated prothrombin-time test with use of a chromogemc peptide substrate and acentrifugal analyzer. Clin Chem 1984;30:524-8.8. Tripodi A, Mannucci PM. Clinical evaluation of a fully automat-ed chromogemc method for prothrombin time compared with aconventional coagulation method. Clin Chem 1984;30:1392-5.9. Duncan A, Bowie EJW, Owen CA, Fass DN. A clinical evalua-tion of automated chromogemc tests as substitutes for conventionalprothrombin time and activated partial thromboplastin time tests.Clin Chem 1985;31:853-5.10. Yamada K, Meguro T. A new AFI’T assay employing achromogenic substrate and a centrifugal autoanalyzer. Thromb Has1979;15:351-8.11. Becker U, Bartl K, Lill H, Wahlefeld AW. Development of aphotometric assay for activated partial thromboplastin time and itsapplication to the Cobas Bio centrifugal analyzer. Thromb Res1985;40:721-30.12. Andersen I, Thorsen S. The application of the reaction rateanalyzer LKB 8600 as an automated coagulometer [Abstract].Scand J Clin Lab Invest 1977;37:7.13. Janson Th, Syversen 0. Automated determination of Thrombo-test and Normotest using LKB 2086 Mark II kinetic analyzer[Abstract]. Thromb Haemo8tasis 1979;42:224.14. Lund Karlsen R, Visdal A, Stensberg A. An automated methodfor Thrombotest on the “Cobas Bio” centrifugal analyzer. Scand JClin Lab Invest 1982;42:000.15. Christensen NJ. Measurements of some blood clotting factorsusing a centrifugal analyzer (Multistat ifi). Scand J Clin Lab Invest1983;43:521-6.16. WHO Expert Committee on Biological Standardization. 28threport. WHO Tech Rep Series 1977;610:14-5, 45-51.

17. WHO Expert Committee on Biological Standardization. 31streport. WHO Tech Rep Series 1981;658:185-205.

18. Tomenson JA. A statistician’s independent evaluation. In: vanden Besselaar AMHP; Gralnick HR. Lewis SM, eds. Thromboplas-tin calibration and oral anticoagulant control. The Hague: Mar-tinus Nihoff, 1984:87-108.19. Gutman SJ, Didisheim H, Markowitz H. Two new artifacts inautomated coagulation testing. Am J Clin Pathol 1980;73:583-8.20. Fareed J, Measmore HL, Kindel G, Balls JU. Inhibition ofserine proteases by low molecular weight peptides and their deriva-tives. Ann NY Aced Sci 1981;370:765-84.

$