1989 Clin Pathol 1993;46:198-203

ICSH recommendations for measurement of

erythrocyte sedimentation rate
International Council for Standardization in Haematology (Expert Panel on Blood
Rheology)

Introduction formed on undiluted blood samples of haema-

Tests of erythrocyte sedimentation provide a tocrit of 0 35 or less under standardised con-
measure of the acute phase response to ditions in a Westergren open ended glass
inflammatory disease' and reflect the sedi- pipette that meets ICSH specifications.'
mentation of red cells in autologous plasma. These undiluted blood samples are anticoag-
The term erythrocyte sedimentation rate ulated with EDTA (dilution less than 1%)
(ESR) is retained because of traditional but not diluted with citrate anticoagulant.
usage, although a single measurement after This method will now be recognised by ICSH
60 minutes is not a rate; in recording the ESR as its reference method. The reference
it is assumed that the reading is made at 60 method, and the standardised method
minutes unless a different time is specified. described below, exist to allow users to pre-
Sedimentation is accelerated by an increase in pare ESR reference material for verification
the plasma concentration of acute phase pro- or quality control purposes in their own labo-
teins of large molecular size but sedimenta- ratories. The results from both methods
tion is also accelerated by anaemia which may should be expressed as ESR = (undiluted) x
or may not be part of inflammatory disease. mm. For comparison with the traditional
The ESR may therefore reflect both the method using diluted blood, a correction for-
hyperproteinaemia and the anaemia of mula' can be applied: diluted blood ESR mm
inflammatory disease and differs from tests, = (undiluted blood ESR m x 0 86) - 12.
such as plasma viscosity, which reflect only
the protein component of the acute phase
response. ICSH standardised method
The increased mobility of patients and the AIM
benefits to laboratories of sharing their expe- The ICSH standardised method should be
rience has led to the need for measurements directly comparable with and traceable to the
between laboratories to be comparable. This ICSH reference method and used as an alter-
can be achieved by using a reference method. native to it for verification or quality control,
ICSH has defined this as an exactly described or both, of working (routine) methods.
technique which, in the opinion of an Expert
Panel, provides sufficiently accurate and pre- BLOOD SAMPLE
cise measurement for it to be used to assess Blood of haematocrit (or PCV) of 0 35 or less
the validity of other such laboratory methods. should be obtained by clean venepuncture
The original ICSH reference method for over a maximum period of 30 seconds and
measuring the ESR2 was based on the without excessive venous stasis. Blood of
methodology of Fahraeus3 and Westergren4 higher haematocrit should not be used
using diluted blood (4 vols blood plus 1 vol because reproducibility of sedimentation may
citrate) in open ended glass tubing of 300 be poor in narrow tubes. A manual or vacu-
mm in length, mounted vertically in a rack or um extraction venepuncture technique can be
stand. Modifications of these specifications, used and the blood should be taken into
in particular the use of undiluted blood, are EDTA anticoagulant (dilution less than 1%)
now recommended as the basis of a new without further dilution in citrate. The ESR
ICSH reference method. should be set up within 4 hours of vene-
Recent developments, including biohazard puncture.
awareness and difficulty in obtaining equip-
ment to perform the reference method, have MIXING OF BLOOD SAMPLE
prompted ICSH to introduce a standardised Mixing of blood with EDTA anticoagulant
Members of Expert Panel: method as an alternative to, and potential (1 5 mg/ml blood) is necessary at the time of
B S Bull (USA), M Caswell replacement for, the reference method. venepuncture, but further mixing immediate-
(UK), E Ernst (Austria),
J M Jou (Spain), A Kallner For working (routine) methods, ICSH ly before the ESR test is set up is critically
(Sweden), J A Koepke now recommends specifications for selected important for reproducibility. For standard
(USA), S M Lewis (UK),
G D 0 Lowe (UK), methods. These are procedures whose reliabil- tubes (10-12 mm x 75 mm containing 5 ml
M W Rampling (UK), ity has been verified against the reference or blood and with an air bubble comprising at
J Stuart (UK) Chairman.
Correspondence to: standardised method and which minimise the ieast 20% of the tube volume) there should
Professor J Stuart, biohazard risk of the test procedure. be a minimum of eight complete inversions
Department of (1800 x 2) with the air bubble travelling from
Haematology, The Medical
School, University of end to end of the tube. Mixing must not
Birmingham, Birmingham ICSH reference method cause haemolysis. Non-standard tubes,
B15 2T7
Accepted for publication In 1988 ICSH proposed, for purposes of particularly when narrower, may require
26 August 1992 intermethod comparability, an ESR per- more than eight inversions and the required
ICSH recommendations for measurements ofESR 199

number should be determined. Mixing ICSH reference method over a range of ESR
should be continued until immediately before values of 15-105 mm. In this comparison
the ESR pipette is filled at the start of the 95% of the differences should be 5 mm or
test. less, with the larger differences associated
with higher ESR values.
SEDIMENTATION PIPETTE SPECIFICATIONS
The pipette (tubes) should be colourless, cir- ICSH selected methods
cular, and of sufficient length to give a 200 AIM
mm sedimentation scale. The scale may be ICSH wishes to promote the use of ESR
marked on the pipette or separately and methodology that minimises biohazard risk; it
should comprise clearly marked lines at 1 mm is especially important to avoid blood spillage
intervals and be numbered from 200 at the or aerosol generation during the test proce-
bottom up to 0 at appropriate intervals. If dure. The purpose of designating working
separate from the pipette, the scale must be (routine) methods as ICSH selected is pri-
part of a pipette holding device that ensures marily to specify criteria that will allow mean-
precise and reproducible alignment of the
pipette and scale. If reading of the pipette is ingful comparison of results between
optico-electronic rather than visual, a marked laboratories. ICSH also considers that such
scale is unnecessary. criteria will encourage development of new
The bore of sedimentation pipettes for the methods of low biohazard risk that give com-
Westergren ESR has traditionally been 2-55 parable results with those of the ICSH refer-
mm + / - 0v 15 mm.5 ICSH now recommends ence method. A manufacturer who wants a
that the pipette diameter should be not less working method to be recognised as ICSH
than 2-55 mm (no upper limit is specified but selected should undertake a study of compa-
the volume of blood required should be min- rability with either the ICSH reference
imised). The bore should be constant (within method or the ICSH standardised method. A
5%) throughout its length and the interior of similar study should be performed by an
the pipette should be circular (difference independent expert. The studies should be
between long and short axes not exceeding documented and readily available for scrutiny
0 1 mm). on request to the manufacturer.
The ESR pipette should be disposable. For verification or quality control of an
Glass or plastic may be used but, if plastic, ICSH selected method in'routine use, com-
should not show adhesive properties towards parability checks against the ICSH standard-
blood cells and should not release plasticisers ised method are recommended.
that affect blood or alter sedimentation. If a BLOOD SAMPLE
mould release agent is used in the manufac- Blood should be obtained by clean venepunc-
turing process, this must similarly not affect ture over a maximum period of 30 seconds
blood or alter sedimentation. and without excessive venous stasis. A manu-
The pipette should be filled with blood to a al or vacuum extraction venepuncture tech-
height of at least 200 mm. Adjustment of the nique can be used. The ESR test should be
blood column or scale should be possible to set up within 4 hours of venepuncture. Blood
allow correction for slight variation in the samples can be stored for more than 4 hours
nominal volume and ensure an initial reading at 4 C, but any such longer period of storage
of zero. During the sedimentation period, and must be validated and the evidence be
during subsequent disposal, the system must available from the manufacturer for scrutiny.
prevent blood spillage or aerosol generation. If certain types of blood sample-from cases
of hyperlipidaemia or hyperbilirubinaemia,
PIPETTE HOLDING DEVICE for example-are unsuitable for testing, this
The pipette should be held vertical (con- should be stated.
firmed by a plumb line or equally effective For blood samples that are diluted at
device), protected from direct sunlight, venepuncture, 4 vols of blood may be taken
draughts and vibration, and kept at a con- directly into 1 vol of sterile sodium citrate
stant temperature (+ / - 1 C) within the anticoagulant-diluent. Vacuum tubes and
range 18-25C during the period of sedimen- tubes containing liquid anticoagulant for this
tation. purpose have a finite storage life which
should be carefully defined by the manufac-
EXPRESSING THE RESULT turer. Storage conditions must also be clearly
This should be recorded as the sedimentation specified. Alternatively, the blood may be
occurring at 60 minutes from the beginning taken first into a primary anticoagulant
of the test and expressed as: ESR (undiluted) (EDTA) that does not cause significant dilu-
= x mm (as for reference method above). tion (<1%) of plasma protein, followed by
dilution in sterile sodium citrate. For the
COMPARABILITY BETWEEN ICSH STANDARDISED above purposes, the concentration of trisodi-
AND ICSH REFERENCE METHODS um citrate dihydrate (Na,C6H507.2H20)
This procedure requires analysis of reference should be within the range 0-10-0-136 mol/l.
material (fresh human blood) on which the This solution should be discarded if it
ESR has been determined by the ICSH refer- becomes turbid; if kept in a reusable contain-
ence method. The standardised method er, particular care must be taken to remove all
should be verified by comparison with the traces of any detergent used for cleaning the
200 ICSH Expert Panel on Blood Rheolog

container. If an alternative anticoagulant, EXPRESSION OF RESULT

diluent (such as saline), or dilution factor is The result should be expressed as: ESR = x
used, comparability of the result with that mm (where 1 mm for diluted blood is equiva-
obtained using the ICSH standardised (or lent to 1 mm for undiluted blood at 60 min-
reference) method should be documented utes according to the ICSH reference or
and readily available from the manufacturer standardised method; see reference method
for scrutiny. for correction formula for lack of dilution in
reference/standardised methods). If an alter-
native method of recording the end point is
MIXING OF BLOOD WITH ANTICOAGULANT- used and the result is expressed as
DILUENT Westergren equivalent units, this must be
Immediately before the ESR test is set up, the clearly stated. If the result is adjusted mathe-
blood sample should be mixed as specified for matically for initial height of the blood col-
the standardised method-at least eight com- umn, haematocrit, ambient temperature, or
plete inversions (1800 x 2) for a 10-12 mm duration of sedimentation, the method of
x 75 mm blood container and more inver- adjustment should be described. The ESR
sions if the interior diameter of the container result for the working method, whether
is smaller. If the working method incorpo- requiring mathematical correction or not,
rates an automatic mixing device, its effec- should therefore be expressed so as to achieve
tiveness should be validated by the comparability with the ICSH reference
manufacturer. method.

SEDIMENTATION PIPETTE SPECIFICATIONS

Pipette (tube) dimensions are not specified VERIFICATION
for ICSH selected methods but comparability Verification of any working (routine) ESR
of the test result with the ICSH standardised method using diluted blood should be per-
(or reference) method must be validated formed against the ICSH standardised
using the protocol described below. In the method at a rate determined by the laborato-
interests of safety, it is preferable that the ry's standard operating procedure and espe-
pipette be disposable but, if reused, special cially when a new batch of tubes or fresh
attention must be paid to cleaning, with stock of citrate is used. Verification against
removal of all contaminants and an appropri- the ICSH standardised method, which is per-
ate verification check made thereafter. The formed on undiluted blood, will detect errors
tube may be made of plastic or glass but, if in the volume or quality of the diluent and in
plastic, should not show adhesive properties the adequacy of mixing the diluent with
towards blood cells and should not release blood in the working method. The ICSH
plasticisers that affect blood or alt&r sedimen- standardised method can therefore also be
tation. Comparability data with the ICSH used for quality control purposes.
standardised (or reference) method should be To obtain blood for the verification exer-
readily available from the manufacturer. cise it may be convenient to select a patient
EDTA sample of haematocrit of 0-35 or less
that contains an adequate residue of blood,
PIPETTE HOLDING DEVICE after all other tests are completed, and has an
A rack or stand should be provided to hold increased ESR value (range 15-105 mm) as
the sedimentation pipettes motionless. If not known from testing or as judged by clinical
held vertical, the angle of incline should be details or the extent of sedimentation after
specified and comparability of the result with the sample has stood undisturbed for 30-60
that of the ICSH standardised (or reference) minutes. The EDTA blood in the tube
method should be validated. The pipettes should then be mixed by at least eight com-
should be protected from direct sunlight, plete inversions. After filling the pipette of the
draughts and vibration, and be maintained at standardised method, another aliquot of
a constant temperature ( 10C) within the blood from the same, or a duplicate, EDTA
range 18-25 C for the duration of the test. If sample should be analysed by the laboratory's
applicable, adjustment of the blood column working method after dilution (4 vols blood
to zero should be possible to correct for slight plus 1 vol citrate).
deviation in the nominal volume and ensure If the blood sample for the working ESR
an initial reading of zero. method was taken by venepuncture directly
into a ESR tube containing citrate, or if such
a dilution was performed in the laboratory,
RECORDING THE END POINT the blood sample for the ICSH standardised
The traditional Westergren method estab- ESR should be taken from a separate EDTA
lished that the end point should be read at 60 sample without dilution. This sample is usu-
minutes. However, some systems allow read- ally available because a routine blood count is
ings at other times, which currently range normally requested in parallel with the ESR.
from 20 to 120 minutes, or at multiple inter- This blood sample should again have an ESR
mediate times (every 5 minutes, for example. value between 15-105 mm and a haematocrit
The clinical usefulness of these alternative of 0-35 or less.
times is yet to be evaluated. At present, ICSH If a manufacturer wishes to demonstrate
recommends that measurement be made at comparability of a new working method with
60 minutes or normalised to 60 minutes. the ICSH standardised (or reference)
ICSH recommendations for measurement of ESR 201

method, the verification exercise may be per- Protocol for evaluation of working ESR
formed using normal blood with added fib- methods against the ICSH reference or
rinogen to achieve an adequate range of ICSH standardised method
increased ESR values (see protocol below). This protocol is based on ICSH recommen-
This verification should be followed by a sim- dations for type testing equipment and appa-
ilar exercise, in collaboration with an external ratus used for haematological analysis,6 ICSH
clinical laboratory, using patients' blood sam- guidelines on selection of laboratory tests for
ples. The results for both exercises should be monitoring the acute phase response,' ICSH
documented and readily available for scrutiny recommendations for measuring the ESR of
on request to the manufacturer. human blood,5 and this document.
Verification of a working method is In this protocol, the ESR equipment which
achieved if 95% of the results obtained fall is the subject of the evaluation will be referred
within the limits shown in the table. to as the "test system". The reference method
is that of ICSH' using Westergren type glass
tubes without anticoagulant diluent. The
REFERENCE VALUES ICSH standardised method is as described in
Reference values should be established locally this document for undiluted blood.
in accordance with the ICSH recommenda-
tion on reference values67 and expressed as
for diluted blood (see selected method PRELIMINARY
above). In view of the progressive rise in ESR This is general information provided by the
with age, separate values should be estab- manufacturer and confirmed when the test
lished for each decade of adult life in men system is installed in the evaluation laborato-
and women. Several other clinical variables ry. It should include the following:
influence the ESR and may therefore affect 1 Brand name and model, manufacturer,
physiological reference values: haemoglobin and distributor.
concentration, medication, menstrual cycle, 2 Suggested local price and cost of main-
pregnancy, and smoking. tenance contract
3 Terms of guarantee
4 Overall dimensions and bench area
MICROMETHODS requirement
Micromethods may be introduced for use in 5 Details of electrical supply and other
children or to reduce the draw volume for necessary services, computer and robotic
adult patients. Documented evidence of com- interface requirements, and requirement for
parability with the ICSH reference or stan- waste disposal
dardised method must be readily available 6 Instruction manual giving principles of
from the manufacturer. operation, degree of automation, data presen-

ESR values (mm) for verification of comparability of working (routine) method with ICSH standardised method
Working Working Working
Standardised Method Standardised Method Standardised Method
Method* Limitst Method* Limitst Method* Limitst
15 3-13 45 18-37 75 40-68
16 4-14 46 18-38 76 40-69
17 4-15 47 19-38 77 41-70
18 4-15 48 20-39 78 42-71
19 5-16 49 20-40 79 43-72
20 5-17 50 21-41 80 44-73
21 6-17 51 22-42 81 45-74
22 6-18 52 22-43 82 45-76
23 6-19 53 23-44 83 46-77
24 7-19 54 24-45 84 47-78
25 7-20 55 24-46 85 48-79
26 8-21 56 25-47 86 49-80
27 8-21 57 26-48 87 50-82
28 9-22 58 26-49 88 51-83
29 9-23 59 27-50 89 52-84
30 10-24 60 28-51 90 53-85
31 10-25 61 29-52 91 53-86
32 11-25 62 29-53 92 54-88
33 11-26 63 30-54 93 55-89
34 12-27 64 31-56 94 56-90
35 12-28 65 32-57 95 57-91
36 13-29 66 32-58 96 58-93
37 13-30 67 33-59 97 59-94
38 14-30 68 34-60 98 60-95
39 14-31 69 35-61 99 61-96
40 15-32 70 35-62 100 62-98
41 15-33 71 36-63 101 63-99
42 16-34 72 37-64 102 64-100
43 17-35 73 38-65 103 65-101
44 17-36 74 39-66 104 66-103
105 67-104
* Standardised method: EDTA anticoagulated but undiluted whole blood of haematourit of 0-35 or less
t Working method: 4 vols EDTA blood plus 1 vol citrate diluent
The values incorporate a correction for dilution of blood by citrate in the working method. Proposed working method valid if
95% of results are within indicated limits.
202 ICSH Expert Panel on Blood Rheology

tation, method used for specimen mixing, 100 samples from patients with a wide variety
volume of specimen, maintenance procedure of diseases and with ESR results distributed
and trouble shooting evenly in the range 15-105 mm. Occasional
7 Certificate of electrical conformity to a blood samples fail to give a clear plasma-ery-
recognised standard (for example IEC throcyte interface after sedimentation; if this
10:10:1: 1990). occurs in either the test system or standard-
ised (reference) method, the pair of values
SAFETY ASSESSMENT should be eliminated from the data set.
1 Microbiological: to test for aerosol or droplet When blood specimens for the test system
contamination during normal operation, one are diluted (4 vols blood plus 1 vol citrate),
of the following two procedures, or an equiv- the undiluted ESR values for the ICSH stan-
alent alternative, should be used: dardised (or reference) method must be cor-
(a) A series of tubes should be filled with a rected for lack of dilution. The ICSH (1988)
suspension of a marker organism (such as formula' can be used for this purpose but ver-
Serratia marcescens) and treated as blood spec- ification of comparability of the test system
imens. Petri dishes containing nutrient agar may be determined with more accuracy over
are placed appropriately in the vicinity of the the range (15-105 mm) of ESR values by
test system which is allowed to operate in its using the table which already incorporates a
usual way. Petri dishes are then incubated correction for dilution. Validation is achieved
and examined for growth of the marker if 95% of the test system results fall within
organism. the limits shown in the table. Use of the table,
(b) A few drops of a fluorescent chemical rather than the formula, is now recommend-
marker are added to blood samples which are ed.
then handled in the usual way for the ESR. Paired results should be plotted on linear
Sheets of white absorbent paper are placed graph paper, with differences of the test sys-
over possible areas of contamination in the tem ESR from the ICSH standardised (or ref-
vicinity of the test system. These are then erence) ESR plotted on the vertical axis and
examined under ultraviolet light for evidence the means of the two methods on the hori-
of droplet contamination. zontal axis.8
2 Mechanical: any potential hazard arising 4 Sensitivity of response to added fibrino-
from the design of the test system and from gen should be determined. A concentrated
any moving parts should be noted. solution of fibrinogen of about 20 g/l is made
3 Waste disposal: any potential hazard by dissolving human fibrinogen (such as Kabi
(microbiological, chemical, or other) should Pharmaceuticals, Sweden) in distilled water.
be assessed. This is dialysed overnight against phosphate
buffered saline (PBS; pH 7*4, normosmotic)
TECHNICAL ASSESSMENT to remove the high salt content. This is the
Before the evaluation is formally started, the stock fibrinogen solution whose fibrinogen
staff who will carry it out should have a pre- concentration should be measured. To each
liminary period of training or familiarisation. of 5 aliquots of 5 ml normal blood is added 1
This may include a pilot study. ml of PBS alone, or mixtures of PBS and
1 Fresh human blood specimens are col- stock fibrinogen, equivalent to adding 0, 5,
lected directly into the specified containers 10, 15 and 20 mg of fibrinogen.
containing anticoagulant diluent, according Calculation of correlation coefficient and
to the manufacturer's instructions. Alterna- slope gives an assessment of linearity of
tively, blood can be collected into EDTA and response and sensitivity.
subsequently diluted according to the manu- 5 If the test system does not demand one
facturer's instructions. Specimens should specified specimen container, a comparison
cover the range of results 15-105 mm with must be made between alternatives, including
about the same number of specimens in each comparison of glass and plastic. Paired results
quartile. Blood for ESR tests should be stored on 20 tests should be analysed as above.
at ambient temperature (18-250C) until test-
ed and the tests should begin within 4 hours EFFICIENCY ASSESSMENT
of collection. If the test system does not 1 The clarity, ease of reference, and com-
incorporate an automatic mixing device, the prehensiveness of the instruction manual
specimens should be mixed as specified for should be evaluated.
the standardised method: at least eight com- 2 Operational timing is established. This is
plete inversions (1800 x 2) for a 10-12 mm based on a study carried out on a batch of at
x 75 mm blood container and more inver- least 10 specimens and extends from speci-
sions if the interior diameter of the container men registration to result printout.
is smaller. 3 The level of training required by the
2 Precision should be based on replicate operator should be assessed.
measurements (10 if possible) of a specimen 4 Reliability and maintenance: a written
from each quartile. The precision of the record is kept of any incidents during the
ICSH standardised (or reference) method period of evaluation, especially noting any
should similarly be determined for compara- "down time" due to failure of the test system.
tive purposes. 5 Cost analysis should include capital cost
3 Comparability between the test system over a nominal 5 year amortisation and cost
and the ICSH standardised (or reference) of annual service/maintenance contract; con-
method should be tested in parallel on at least sumables; and labour costs, taking account of
ICSH Export Panel on Blood Rheology 203

the required seniority of the operator and the of blood or aerosol generation. This stan-
operational timing, as above. Financial com- dardised method may be used for verification
parison with the laboratory's current working or quality control of other ESR methods and,
method should be calculated on the basis of in future, may replace the reference method.
10, 30, 50, 100, and 300 tests per day. 3 ICSH selected methods-ICSH recom-
mends specifications for working methods,
CONCLUSION OF EVALUATION using diluted or undiluted blood, which may
This should take account of the technical and be considered as ICSH selected methods for
functional reliability of the test system, cur- routine use. A protocol is outlined for evalua-
rent laboratory practice, the impact of the test tion of such working methods against the
system on the organisation of the laboratory ICSH reference method or the new ICSH
and staff reaction to its introduction, and standardised method.
resource implications of the cost analysis
described above. Comments on these recommendations are invited and should
be addressed to the ICSH Executive Secretary and to the
Chairman of this panel.

1 International Committee for Standardization in

Haematology (Expert Panel on Blood Rheology).
Summary Guidelines on selection of laboratory tests for monitor-
The Expert Panel on Blood Rheology of the ing the acute phase response. J Clin Pathol 1988;41:
1203-12.
International Council for Standardization in 2 International Committee for Standardization in
Haematology (ICSH) has prepared new rec- Haematology. Reference method for the erythrocyte
sedimentation rate (ESR) test on human blood. Br J
ommendations for measurement of the ery- Haematol 1973;24:671-3.
throcyte sedimentation rate (ESR) under the 3 Fahraeus R. The suspension-stability of the blood. Acta
Med Scand 1921;55: 1-228.
following categories: 4 Westergren A. Studies of the suspension stability of the
1 ICSH reference method-ICSH now blood in pulmonary tuberculosis. Acta Med Scand
192 1;54:247-82.
recognises, as its reference method for the 5 International Committee for Standardization in
ESR, the sedimentation of EDTA-anticoagu- Haematology. Recommendation for measurement of
erythrocyte sedimentation rate of human blood. Am J
lated but undiluted blood in traditional Clin Pathol 1977;68:505-7.
Westergren pipettes that meet ICSH specifi- 6 International Committee for Standardization in
Haematology. Protocol for type testing equipment and
cations. apparatus used for haematological analysis. J Clin Pathol
2 ICSH standardised method-ICSH rec- 1978;31:275-9.
7 International Federation of Clinical Chemistry (IFCC)
ommends specifications for a new standard- and International Committee for Standardization in
ised method for the ESR based on the Haematology (ICSH). Approved recommendation
(1986) on the theory of reference values. J Clin Chem
sedimentation of EDTA-anticoagulated, but Clin Biochem 1987;25:337-42.
undiluted blood in pipettes with a 200 mm 8 Bland JM, Altman DG. Statistical methods for assessing
agreement between two methods of clinical measure-
scale and which are designed to avoid spillage ment. Lancet 1986;1:307-10.