Clin.lob. Haeniat.

1979, I, 75-86

REVIEW

Automated coagulation detection systems*

J . A . K O E P K E , M D & G .G.KLEE, MD, PhD

Department of Pathology, University of Iowa Hospitals and Clinics,
Iowa City, Iowa 52242

Accepted for publication 20 February 1979

Similar to trends noted earlier in chemistry and haematology laboratories, there has
been a continuing increase in the use of automated and semi-automated devices for
coagulation testing. For example, i n 1962, essentially all coagulation tests were done
using manual techniques. By 1969, however, about 40% of prothrombin time testing
was being done using a semi-automated device, most often the Fibrometer (BioQuest).
A recent perusal of the College of American Pathologists (CAP) survey indicates that
more than 80% of the more than 7000 American hospital laboratories now d o
prothrombin times and partial thromboplastin times with a variety of such devices
(Koepke 1977). In fact, 95%, of larger hospital laboratories rely on instruments for
coagulation testing. Similar trends are noted in Europe and Japan. Although the
Fibrometer is still the most popular instrument in the United States, nine other
instruments challenge the Fibrometer (cf. Table I).
Instruments having two levels of operator dependence are marketed. For purposes
of this discussion, a machine will be designated as semi-automated if the endpoint
detector is automated (i.e. non-operator dependent) while the operator is required to
be continually present for sample application. The widely-used Fibrometer is an
example of a semi-automated instrument.
Systems, which after they are loaded, do not require continuous operator attention
will be called automated. Characteristic features include continuous refrigeration
of queued samples, precisely timed and regulated activation and incubation steps,
and digital data output. The Lancer/Sherwood Coagulyzer and the MLA Electra
exemplify such systems.
The use of these devices has been associated with improved precision of coagula-
tion testing. For example, manual prothrombin time precision, expressed as the
coefficient of variation, varies around 5-7% while most automated devices vary
around 4-5%, in the normal range; somewhat higher variation is seen in the elevated
range. Table 2 shows the improving precision of both manual and the most popular
* Presented at the ICSH Symposium on Standardization in Haematology at Ihe XVIT Congress of
the Jnternational Society of Haematology, Paris, 27 July 1978.

0141-9854/79/0500-0075$02.00 8 1979 Blackwell Scientific Publications 75

76 J.A.Koepke and G.G.Klee

Table I. Coagulation instruments--r978*

Device x
Auto-Fi (Dade) 1 .o
Clotek (Hyland) 5.6
Clot Timer (Mechrolab) 1 .o
Coag-a-Mate (Gen. Diag.) 8.0
Coagulation Analyzer ( BioiData) 3 .o
Coagulyzer (Lancer/Sherwood) 6.0
Electra (MLA) 6.6
Fibrometer (BioQuest) 65.0
Prothrombin Timer (Emdeco) I .o
Thrombitron (Clay Adams) < 1.0
All others < 1.0

* Data from CAP Proficiency Testing Programs.

semi-automated prothrombin systems used in the CAP surveys from the last decade.
The choice of an instrument for coagulation testing will take into consideration
initial purchase price, reagent cost, repair service availability and workload. The
instrument under consideration must also be compatible with the reagents being
used (Sabo 1977).

Historical perspective of coagulation instruments

A comprehensive historical review of coagulation instruments is not relevant for a
paper of this length ; however, selected references will illustrate that many of the basic
methodologies for coagulation instrumentation were initially described over 50
years ago. While the current instruments may be more accurate and/or precise, the
basic concepts are very similar to the earlier instruments. More extensive reviews of
early coagulation instruments can be found in the book by Nygaard (1941), and the
articles by Cohen (191I) and von Kaulla (1962).
One of the earliest methods for detecting blood coagulation was based on drawing
a white horse hair through a capillary tube filled with blood. Vierordt (1878) described
this method in which the clot function was detected by the visual appearance of shreds
of fibrin formed on the hair. This basic concept was used by the Dade corporation in
the 1970s in the development of the [email protected] instrument detects the adherence
of a clot to the filaments by pulling the clot past a light beam.
An interesting coagulation detection system using an indirect measurement of
viscosity was the ‘coagulometer’ developed by Brodie & Russell ( I 897). This device
employed a hanging drop of venous blood on a glass slide in which the erythrocytes
were set in motion by a stream of air intermittently blown on the edge of the drop.
The endpoint was determined by microscopic examination at the time when the
viscosity of the specimen increased and the erythrocyte movement ceased. A 1970s
counterpart utilizing a similar technique is the Danish instrument described by
 Automated coagulation detection systems 77

Table 2. Prothrombin time precision*

Normal range Elevated range

Survey ( I 2- I 4s) (20-24s)
year manual automated manual automated

I969 11.0 7.3 8.0 7.0

1970 6.7 5.9 7. I 7.7
1971 7.6 5.7 8.9 8.2
1972 7.5 5.5 8.6 7.4
1973 8.9 6.4 8.6 7.5
I974 7.8 6.1 8.3 7.7
1975 5.4 4. I 8.8 7.4
I976 5.4 4.7 8.2 6.4
1971 5.2 4.4 7-8 7.7
1978 5.6 4.7 7.4 5.7

* Expressed as coefficient of variation.

Dybkaer & Kuster (1972). In this instrument, rhythmic changes in air pressure forces
the coagulation mixture through a constriction in a U-shaped tube. As the blood
clots, the viscosity increases and passage of the fluid through the tube decreases. This
viscosity change is sensed by a pressure change in a mercury manometer.
Another instrument for detecting blood coagulation was developed by Fuld &
Schlesinger (1912), which employed a method similar to that which was used in
H yland’s ClotekB instrument. Fuld’s original ‘Thrombometer’ was based on tilting
a U-shaped tube containing a metal bead. When the blood clot formed, the bead no
longer moved. The Clotekm uses a test vial with a magnetically suspended sphere.
The instrument moves the vial up and down and detects the formation of a clot when
the metal sphere is moved out of the light bean (Figure I).
An apparatus called a ‘coagulometer’ was developed by Cannon & Mendenhall
( r g ~ q )which
, was based on adherence of fibrin strands to a probe. At timed intervals,

Figure I. Coagulation
PERMANENT MAGNETIC detection system using a
SUSPENSION
magnetically suspended
mixing sphere. The tube
PHOTOCELL RECEIVES moves up and down and
LIGHT AND STOPS TIMER
when the clot forms it
pulls the sphere out of
the light beam to stop
BALL TRAPPED I N CLOT’
‘._.I’

BALL REMAINS the timer. Hyland Clo-

MOVES WITH TUBE ’ .
‘ SUSPENDED BETWEEN tek@.(Reproduced with
(DOWNWARD I N THIS EXAMPLE) MAGNETS the permission of Tra-
.
.
-a
TEST TUBE DRIVEN venol Laboratories, Inc.,
UP AND DOWN Deerfield, Illinois.)
78 J.A.Koepke and G.G.Klee

Cross section ot
disposable reaction tray

Electro-optical Electro-optical
sensor @ sensor @

Amplifier @ Amplifier @

Detector Detector
circuit @ circuit @

Figure 2. Dual channel photo-electric clot detector. Samples are placed in two concentric circles
of reaction wells. This tray rotates over a common light source and two independent electro-
optical detector circuits. General Diagnostics Coag-A-Mate.

a copper hook was dipped into and lifted out of the blood until a fibrin thread was
detected. Each time the hook was pulled back, a line was marked on a smoked drum
until the fibrin clot formed. This method resembles the current FibrometeP and the
previous ‘prothrombin timer’ by Emdeco. However, in these modern instruments
electrical conductance of the fibrin stand rather than mechanical resistance is used to
detect the endpoint.
Most of the more successful coagulation instruments presently available are based
on photo-electric measurements of clotting plasma (Figure 2 ) . Some of the early work
-

56.6 s
r
f- Noise from Recorder output
pipettor 2.5 mv/cm

,/First
derivative
0.5 v/cm

Figure 3. Time course of the voltage output of the electro-optical detector circuit using light trans-
mission. The recorder output corresponds to changes in optical density. The clot is detected when
the derivative of this signal goes below the threshold. Lancer/Sherwood.
 Automated coagulation detection systems 79

in this area was done by Kugelmass (1923). His apparatus called a ‘nephelometer’
utilized a light source and a thermopile connected to a galvanometer. At that time,
reliable photoelectric devices were not available. Baldes & Nygaard (1936) developed
a ‘coagulometer’ using an automobile headlight, a 37.5OC water bath, and a photo-
electric cell. This original instrument had to be used in a dark room to avoid stray
light. Further improvements in the construction and electronics made these instru-
ments more stable. More recently, Owen & lsaacson (1967) patented an ‘automatic
prothrombin timer apparatus’ which used the first derivative of the light transmission
signal to detect the clot. Most of the current optical systems utilize the first or second
derivative to avoid the differences in absolute optical density from the plasma of one
patient to the next (Figure 3).
The early investigators realized that there may be importance in the continuous
measurement of the coagulation changes. Koffmann (1910) developed a ‘coagulo-
viscosimeter’ which allowed the continuous monitoring of the clotting of blood and
plasma. Recently it was discovered that the peak value of the first derivative of the
optical density clotting curves is linear with fibrinogen and logarithmic with the other
individual clotting factors. This rate of change of the optical density appears to have
clinical significance (Hougie et al. 1978) and the Bio-Data Corporation has recently
patented an apparatus using these thrombokinetics (Eichelberger et al. 1977).
Although the modern-day instruments are based on the same principles as the
older instruments, considerable advances have occurred. The earliest instruments did
not have constant temperature control during testing or automatic endpoint detec-
tion. There were problems with stability, reproducibility, and differences between
patient samples, especially with optical systems, though the use of derivatives of the
curves has helped this problem. Only the more recent instruments have had refriger-
ated turntables which hold multiple samples and automated provisions for pipetting
reagents and printing results. Even currently, only the Dade Auto-Fi@has provisions
for automatically pipetting the plasma samples. The modern instruments tend to
eliminate more of the manual aspects of the testing in order to increase reliability
and reproducibility.
The perpetuation of many of these early basic methods for measuring blood coag-
ulation in addition to other methods based on impedence or clot elasticity probably
signifies that no one method is clearly superior to the others and each laboratory
must choose a coagulation system which seems to serve the needs of their specific
situation.

Principles of instrumental methods

Thc evolution of devices for clotting endpoint detection parallels the principles of
the manual techniques. The technologist may have used a hook or loop of wire to
stir and test the liquid reaction mixture until a visible fibrin strand can be pulled from
it. Alternately, the reaction tube can be mixed quickly, then tilted back and forth
with a rhythmic wrist action until the liquid begins to gel and visible fibrin strands
appear. The precision and accuracy of these techniques depend upon the technolo-
80 J.A.Koepke and G.G.Klee

Table 3. Fibrin strand detection instruments

SEMI-AUTOMATED
Clotek (Hyland)
Clot Timer (Mechrolab)
Fibrometer (BioQuest)
Prothrombin Timer (Erndeco)
Sonoclot (Sienco)
Thrombitron (Clay Adams)
AUTOMATED
Auto-Fi (Dade)

gist’s experience, dexterity and consistency in detecting the sometimes poorIy visible
endpoint. Additional variables in the manual methods include the waterbath tem-
perature, the time the moist reaction tube is out of the water bath being cooled by
surface evaporation, the calibration of manual pipettes, the cleanliness of glassware,
the accuracy of stopwatches, and the like. The beauty of the instruments being
marketed is that they precisely control most of the above variables with resulting
improved precision of coagulation testing.
Roughly paralleling the wire-loop technique, instruments have been made which
sense the formation of a fibrin strand and record this endpoint. By far the most widely
used instrument of this or any other type is the Fibrometer. In principle, a moving
electroprobe is rotated about a stationary probe in the reaction mixture. When the
newly-formed fibrin strand bridges the gap between these probes, electroconductivity
is established and the time of this event is recorded. Thermal heating blocks and semi-
automated pipetting devices are provided. Although the readout is digital in 0.1
second intervals the 0.5 second cycling time results in one-half second recording
intervals. Since the probes dip into the reaction mixture and remain there through
fibrin formation, carryover can be a problem. Properly used, this instrument gives
acceptable results, well under 5% CV on between day precision studies. Additional
instruments using the fibrin-strand detection principle are listed in Table 3.
The other manual method is the popular tilt-tube procedure. The reaction mixture
is observed for gel formation just prior to the formation of a fibrin web. The coagu-
lation instruments utilizing a photo-optical endpoint detection method may be con-
Table 4. Photo-optical instruments

SEMI-AUTOMATED
Coagulation Profiler (Bio/Data) CP-8
Electra (MLA) 750
I.E.C. Colysgraph (Damon)
AUTOMATED
Coag-A-Mate or Coag-A-Pet (General
Diagnostics), single or dual channel
Coagulyzer (Lancer/Sherwood)
Electra (MLA) 600
 Automated cougulation detection systems 8I

Reagent Dispenser

Figure 4. Coagulation detector util-

i7ing clot adherence to moving
filaments. The cotton filaments are
/ T'm'ng continuously drawn through the
Ended
reaction tray. As the clot forms it is
drawn past a light beam to trigger
the endpoint of coagulation. Dade
Auto-Fi. (Reproduced with permis-
sion of Dade Division of American
Hospital Supply Corp., Miami,
Florida).

sidered descendents of this technique. There is usually an automated delivery of

reagent(s) into previously loaded sample containers. A photosensitive cell observes
the rate of increasing optical density (AOD). At a predetermined point of maximal
increase of AOD the endpoint is detected and recorded. With some of these instru-
ments, the changing optical density is plotted on a graph and additional information
may be derived from analysis of these charts. Table 4 lists the several instruments
which take advantage of the optical density principle.
Several coagulation instruments do not clearly fit into either of thesecategoriesand
will therefore be discussed separately. A completely automated coagulation system
was introduced by the Technicon Corporation several years ago. The testing was
carried out on a continuously moving strip of mylar tape carrying several drops of
plasma to which were added thromboplastin or partial thromboplastin containing a
granular iron oxide suspension. Mixing was accomplished by a rotating magnet
mounted beneath the mylar film, which swirled the iron particles. The endpoint
detector was a light beam which was directed at the mylar film. When a discrete clot
was formed it concentrated or incorporated all of the dispersed iron oxide allowing
the underlying shiny mylar to reflect the light beam into the detector, thus signalling
that coagulation had occurred. The approach which this manufacturer proposed was
quite promising. Unfortunately, it is no longer being pursued.
Within the past 2 years another manufacturer, Dade, has introduced an automated
coagulation instrument based upon a principle again somewhat different from
previously available instrumentation. Their instrument, the Auto-Fi, however,
blends principles from the fibrin-strand electromechanical and the optical detection
systems outlined above but more closely follows the fibrin-strand principle. In
simplest terms the Dade Auto-Fi works as follows; the plasma and reagent are
automatically mixed in a specially-made reaction vessel through which two parallel
82 J.A.Koepke and G.G.Klee

Table 5. lnstrumental differences with common thromboplastin’

Thromboplastin:
Thromboplastin-C Activated Simplastin
Instrument (Dade) (Ortho) (Dade) (General Diagnostics)

FIBRIN STRAND
Prothrombin Timer (Emdeco) - I .05 -
Fibrometer (BioQuest) 1 .c7 0.96 0.99
Clot Timer (Mechrolab) - I .oo -

Clotek (Hyland) I .07 I .03 0.96

OPTICAL DENSITY
Coag-A-Mate
(General Diagnostic) - - - 0.92
Electra (MLA) I .02 1.01 0.9 I 0.88
Coagulation Analyzer
(BioiData) 0.96 0.94 - 0.88
Coagulyzer (Lancer) 0.9.5 0.91 0.90 0.84

*All results presented as mean prothrombin time for all laboratories using indicated system divided
by grand mean prothrombin time (18.17 s).

threads are drawn. When coagulation occurs, the clot adhering to the threads is
drawn past a detector which signals that the reaction is complete (Figure 4). Eval-
uations and performance in the field are quite positive and comparisons with con-
ventional methods have been good.
While instrumentation has resulted in a marked improvement in the precision of
coagulation testing, the recent profusion of types and models of equipment and
reagents has also resulted in some confusion for laboratory workers regarding
interlaboratory precision. lntra- as well as inter-laboratory precision is usually
significantly improved by instruments, provided only single or identical instruments
are being compared and all parties are using the same manufacturer’s reagents
(Koepke 1975). If the reagents are held constant inter-instrumental differences can be
examined. The College of American Pathologists (CAP) survey data can be examined
to ascertain if instrumental biases exist, and if so, to determine their extent (Koepke
1977). Table 5 examines the CAP prothrombin time survey data in this manner
(Koepke et al. 1977). At least 20 laboratories used the same system (method/reagent)
and we therefore feel the mean values are comparable insofar as this study is con-
cerned. From examination of Table 5 it is apparent that significant instrumental
differences in prothrombin time measurements occur. Most obvious are the differ-
ences between fibrin strand and optical density systems; times measured in optical
density machines are estimated to be 6-15”/0 shorter than when done using the fibrin
strand principle.

Proposed performance standards for coagulation instruments

While the cost of purchasing and operating an instrument are important considera-
 Automated rocrgulntion deterlion systeriis 83

Table 6. Precision goals for coagulation instruments*

Testing mode:
Prothrombin Partial thromboplastin Fibrinogen
time (%) t i r e (activated) (X) (79
INTKALABORATORY
within r u n <2 ‘5
between run (same day) <3 (6
between day 14 (7
INTERLABORATORY
regional pools 15 ‘5 18
surveys <6 16 < I0

* Expressed as coefficients of variation in the normal range.

tions in the purchase of any instrument for a medical laboratory, the prime con-
siderations should be directed toward the performance of such instruments. In the
following paragraphs b e have developed a series of performance standards for these
ins tr LI m ents.
I t must be pointed out that these are suggested standards that at this time have no
endorsement by any standard setting groups. However, it is our feeling that tentative
performance standards must be developed so that the laboratorian can have a
benchmark against which to judge the many different instruments being marketed.
Appropriate considerations for workload, as well as cost of the machine, reagents
and supplies must also be made. In the next section of this paper summaries of
variety of data, including performance data, from coagulation instruments which are
currently being used. The reader can then compare the individual machines with
these tentative pcrformance standards.

PRECISION REQUIREMENTS

The pursuit of precision in the clinical laboratory must be tempered by a realization

of the requirements for the clinical care of the patient. In our judgement the precision
requirements for these tests are both reasonable and attainable. Table 6 catalogues
these limits.

SENSITIVITY REQUIREMENTS

The requirements for sensitivity in coagulation testing are predicated priniarily upon
the medical needs of the clinician. In the case of the prothrombin time, anticoagulant
therapy becomes the major consideration while with the partial thromboplastin time
the diagnosis and replacement therapy of patients with haemophilia is the primary
consideration. Finally, sensitivity to low levels of fibrinogen is especially important in
the detection of defibrination syndromes.
84 J.A.Koepke and G.G.KIee

The following factor levels should produce abnormal results : prothrombin time
-12.5% of factors 11, V, VlI or X ; partial thromboplastin time-30% of factor
VlII and 15% of IX; fibrinogen-50 mg/dl.

L l N E A R f T Y O F F A C T O R ASSAYS O R PLASMA D I L U T I O N P R O C E D U R E S

Factor assay and plasma dilution procedures should yield straight line relationships
when the data is plotted on semilogarithmic paper. Failure of linearity indicates
suboptimal experimental conditions, circuIating inhibitors or poor technique.

ACTIVATION
In the partial thromboplastin time mode, activation must be carefully controlled
both in regard to time as well as temperature, while optimal activation times may vary
with the brand of partial thromboplastin being used. They must be consistent. A
variation of less than 2 5 yd from the mean activation time is felt to be reasonable.
The activation temperature should be 37 f I "C.

C A R R Y OVER

While many instruments by their design cannot carry over plasma from one specimen
to the next, in those where this possibility exists such carry over must be I or less of
the saniple volume.

REAGENT A N D A C T I V A T O R C O M P A T I B I L I T Y

Since there are many different thromboplastins and partial thromboplastins available
it is required of the instrument manufacturer that compatibility or incompatibility
of all reagents and activators be stated. If such information is not available it should
be so stated.

VERIFICATION O F INSTRUMENT PERFORMANCE

Jnstrument performance should be verified using both carefully prepared test plasmas

Table 7.Specifications of test plasmas

- ~~

Testing mode :
Prothrombin Partial thromboplastin Fibrinogen
Constituent time time (activated)

Prothrombin Complex,
Factors 11, V, VII or X (><) 12.5-100 > 50 >50
Factor VIII (%) >50 1.5-100 > 50
Factor IX (x) 250 1.5-100 '50
Fibrinogen (mgidl) > 50 mg/dl > 5 0 mgidl 50-700 mgldl
 Automated coagulation detection systems 85

(cf. Table 7) as well as fresh patient specimens. The following data should be obtained :
a Ten replicate analyses done on normal and abnormal patient plasmas in the morn-
ing. Repeat procedure on the same plasma in the afternoon. Refrigerate these plasmas
between procedures. Calculate within run and between run precision.
b Analyse appropriate lyophilized plasmas daily for 20 days. Calculate between day
precision.
c lnterlaboratory precision requires enrollment in regional quality control and pro-
ficiency survey programmes.
d Determination of extent, or lack, of interference by plasma haemoglobin, bilirubin
and/or lipid requires the logging of patient specimen results in which the levels of
these potential interferents are above predetermined levels. (Haemoglobin > 50 mg/dl,
bilirubin> 10 mg/dl, total lipids > 750 mg/dl.)
e Carry over will be determined by appropriate queueing of normal and abnormal
plasmas so that contamination of a markedly deficient plasma with a normal one can
be determined by calculation.

References
BALDESE.J. & NYGAARD K.K. (1936) Determination of the coagulability of the blood plasma by the
photo-electric cell. Mayo Clinic 11, 151
BRODIET.G. & RUSSELL A.D. (1897) The determination of the coagulation time of blood. 21, 403
CANNON W.B. & MENDENHALL W.L. (1914) Factors affecting the coagulation time of blood. 1.
The graphic method of recording coagulation used in these experiments. Anz. J. Physiol. 34,
225
COHENM.D. (1911) The coagulation-time of the blood as affected by various conditions. Arch.
intern. Med. 8, 684
DYBKAER R. & KUSTERH. (1972) Semiautomatic coagulation meter utilizing changes in viscosity.
Scand. J. din. lab. Invest. 29, 265
EICHELBERCER J.W., KENTF.M. & SIKOL M. (1977) Method and apparatus for determining defi-
ciencies in enzymatic reactors particularly clotting factor levels in blood plasmas. United
States Patent Number 4, 047, 890
FULDE. & SCHLESINGER E. (1912) Uber die Gerinnung des Blutes. Klin. Wochenschr. 2, 1323-1327
HOUCIEC., MCPHERSON R.A., BROWNJ.E., LAKIN-THOMAS P.L., MELARAGNO A., ARONSON L. &
BAUGH R.F. (1978) The Passovoy defect. Further characterization of a hereditary hemorrhagic
diathesis. N . Engl. J . Med. 298, 1045-1048
KOFPKE J.A. (1975) Inter-laboratory trials: The quality control survey program of the College of
American Pathologists. In Quality Control in Haernatology, eds S.M. Lewis & J.F. Coster,
pp. 53-67. Academic Press, London
KOEPKE J.A. (1977) Evaluation of materials and methods for coagulation testing. In Quality Control
in Laboratory Medicine, eds J.B. Henry & J.L. Giegel, pp. 157-166. Masson, New York
KOEPKE J.A., GILMER JR P.R., TRIPLETIT D.A. & O'SULLIVAN M.B. (1977) The prediction of pro-
thrombin time system performance using secondary standards. Am. J . Clin. Path. 68, 191-194
KOFFMANN K. (1910)Der Koaguloviskosimeter mit spezieller beruecksichtigung seiner klinischen
verwendbarkeit fuer gerinnungs best immugen des blutes. Klin. Med. 59,415
KUGELMASS I.N. (1923) Contribution physico-chimique au mecanisme de la coagulation du sang.
Arch. Int. Physiol. 21, 139
NYGAARD K.K. (1941) Hemorrhagic Disease. Photoelectric Study of Blood Coagulability. C.V.
Mosby, St Louis
OWENC.A. & ISAACSONJ. (1967) Automatic prothrombin timer apparatus. United States Patent
Number 3,307,392
86 J.A.Koepke and G.C.Klee

SABOM.G. (1979) Instrumentation and reagents in the coagulation laboratory. I n Labovatovy

Evuluation of HemostaAis,ed. D.A. Triplett. American Society of Clinical Pathologists, Chicago,
Illinois
VIERORDT C.H. (1878) Die Gerinnungszeit des Blutes in gesunden und Kranken Zustanden. Avch.
Heil. 19, 193
VON KAULLAK.N. (1962) Quantitative Methods for Recording Blood Coagulation: Theoretical
and Practical Aspects. In Progress in Hematology, Vol. 111, ed. L.M. Tocantins. Grune &
Stratton, New York