Articles

Early surgery versus initial conservative treatment in

patients with spontaneous supratentorial intracerebral
haematomas in the International Surgical Trial in
Intracerebral Haemorrhage (STICH): a randomised trial
A David Mendelow, Barbara A Gregson, Helen M Fernandes, Gordon D Murray, Graham M Teasdale, D Terence Hope, Abbas Karimi,
M Donald M Shaw, and David H Barer for the STICH investigators*

Lancet 2005; 365: 38797

See Comment page 361
*Investigators listed at the end
of report

Summary
Background Spontaneous supratentorial intracerebral haemorrhage accounts for 20% of all stroke-related sudden
neurological decits, has the highest morbidity and mortality of all stroke, and the role of surgery remains
controversial. We undertook a prospective randomised trial to compare early surgery with initial conservative
treatment for patients with intracerebral haemorrhage.
Methods A parallel-group trial design was used. Early surgery combined haematoma evacuation (within 24 h of
randomisation) with medical treatment. Initial conservative treatment used medical treatment, although later
evacuation was allowed if necessary. We used the eight-point Glasgow outcome scale obtained by postal
questionnaires sent directly to patients at 6 months follow-up as the primary outcome measure. We divided the
patients into good and poor prognosis groups on the basis of their clinical status at randomisation. For the good
prognosis group, a favourable outcome was dened as good recovery or moderate disability on the Glasgow outcome
scale. For the poor prognosis group, a favourable outcome also included the upper level of severe disability. Analysis
was by intention to treat.

Correspondence to:
Prof A David Mendelow, Head of
Department of Neurosurgery,
Newcastle General Hospital,
Newcastle upon Tyne
NE4 6BE, UK
[Link]@[Link]

Findings 1033 patients from 83 centres in 27 countries were randomised to early surgery (503) or initial conservative
treatment (530). At 6 months, 51 patients were lost to follow-up, and 17 were alive with unknown status. Of 468
patients randomised to early surgery, 122 (26%) had a favourable outcome compared with 118 (24%) of 496
randomised to initial conservative treatment (odds ratio 089 [95% CI 066119], p=0414); absolute benet 23%
(32 to 77), relative benet 10% (13 to 33).
Interpretation Patients with spontaneous supratentorial intracerebral haemorrhage in neurosurgical units show no
overall benet from early surgery when compared with initial conservative treatment.

Introduction
Spontaneous supratentorial intracerebral haemorrhage
affects 20 in 100 000 people every year and communitybased studies have indicated a mortality of more than
40%.1 Most survivors are disabled. The role of medical
and surgical treatment continues to be controversial.
Much of this controversy relates to the penumbra of
functionally impaired (but potentially viable) tissue
around the haematoma. Such an ischaemic penumbra is
associated with brain oedema related to the presence of
thrombin.26 Simulated removal of the mass lesion
improves perfusion in the surrounding brain tissue.7,8
However, clinical studies have yielded conicting
results regarding the importance of such a
penumbra.9,10 If a penumbra exists in patients
with spontaneous intracerebral haemorrhage, clot
evacuation could then restore function to the
surrounding brain tissue and improve outcome, but
clinical imaging studies have so far failed to provide
conclusive evidence for or against this theory. Elevated
intracranial pressure and reduced cerebral perfusion
pressure have been associated with poor outcome,
[Link] Vol 365 January 29, 2005

lending support to a possible benet from early

surgical intervention.11
In 1961, McKissock and colleagues12 reported the rst
prospective randomised controlled trial in neurosurgery
and showed that operative treatment was associated with
a worse outcome than conservative treatment for
patients with spontaneous supratentorial intracerebral
haemorrhage. That trial has affected the management of
this disorder for most of the past half century. In 1989,
Auer and co-workers13 reported the opposite result in a
trial of endoscopic removal of haemorrhage in
100 patients. In the same year, this nding was
contradicted by Juvela and colleagues14 (who supported
the view of the McKissock group) but the trial was too
small to detect less than a substantial effect of surgery.
Since these three initial trials, a further six have been
reported and meta-analysis of the rst seven has shown
no rm conclusions regarding the role of operative
treatment.15 A recent trial that used intracavity thrombolysis16 did not suggest a benet from surgery, whereas a
trial of CT-guided mechanical aspiration did,17 but again
in a small number of patients. Many non-randomised
387

Articles

studies, including a large observational study of more

than 7000 patients from Japan,18 have identied
important prognostic criteria in patients with
intracerebral haemorrhage.
Improved surgical techniques, neuroimaging, neuroanaesthesia, and perioperative monitoring and care have
all led to improved outcomes from surgery in many
conditions. Hence, a randomised trial of the management of patients with spontaneous supratentorial
intracerebral haemorrhage was timely. The International
Surgical Trial in Intracerebral Haemorrhage (STICH)
aimed to assess whether a policy of early surgical
evacuation of the haematoma in patients with spontaneous supratentorial intracerebral haemorrhage would
improve outcome, in terms of death and disability,
compared with a policy of initial conservative treatment. Additionally, it aimed to improve denitions of
the indications for early surgery. STICH was designed
as an international, multicentre, parallel-group study.

Methods
Patients
Randomisation commenced in 1995 in Newcastle, UK,
with initial funding from the Stroke Association (UK).
By the beginning of 1998, 11 centres were registered and
further funding was then obtained from the Medical
Research Council (MRC) in the UK for an international,
multicentre trial. By the end of recruitment in February,
2003, 107 centres were registered with the trial and
1033 patients had been recruited. The full trial protocol
was published in 1999.19 Every centre obtained written
ethical approval according to national and local
guidelines before being eligible to join the study, and
recorded consent to participation as required by local
procedures. Patients were eligible for inclusion if they
had CT evidence of a spontaneous supratentorial
intracerebral haemorrhage that had arisen within 72 h
and if the responsible neurosurgeon was uncertain
about the benets of either treatment (the clinical
uncertainty principle). Study guidelines recommended
that eligible patients should have a minimum
haematoma diameter of 2 cm and a Glasgow coma score
of ve or more.
Patients were not eligible if: the haemorrhage was
probably due to an aneurysm or an angiographically
proven arteriovenous malformation; the haemorrhage
was secondary to a tumour or trauma; patients had a
cerebellar haemorrhage or extension of a supratentorial
haemorrhage into the brainstem; patients had severe
pre-existing physical or mental disability or severe
comorbidity that might interfere with the assessment of
outcome; surgery could not be undertaken within 24 h of
randomisation.
Informed consent according to the criteria set by the
local research ethics committee in every centre had to be
obtained in writing before randomisation. If consent
could not be obtained because the patient was in coma,
388

confused, or dysphasic, assent was obtained from

relatives (if this was regarded by the local ethics
committee as an acceptable alternative).

Procedures
We used the 24-h telephone randomisation service
provided by the Clinical Trial Service Unit (CTSU) at the
University of Oxford. The responsible neurosurgeon,
having obtained consent or assent, completed a onepage randomisation form before telephoning the CTSU.
These key baseline data were recorded before treatment
was allocated by the CTSU. A deterministic
minimisation algorithm, based on side of haematoma
and minimum depth from cortical surface, was initially
used to ensure balance between the two groups, within
every country where patients were recruited. However,
the algorithm was later reprogrammed by the CTSU
independently of the trial steering and trial management
committees. As a result, at least 50% of patients were
allocated using simple randomisation alone. This
procedure did not compromise the study, and the
recorded imbalance in overall numbers was entirely
consistent with randomisation when so many strata
were included.
Patients randomised to early surgery had their
haematoma evacuated within 24 h of randomisation by
the method of choice of the responsible neurosurgeon,
combined with the appropriate and best medical
treatment. Patients randomised to initial conservative
treatment had the best medical treatment. Later
evacuation had to be allowed if it became necessary
because of neurological deterioration. Primary outcome
was death or disability using the extended Glasgow
outcome scale 6 months after ictus. Secondary outcomes
included mortality, the Barthel index, and the modied
Rankin scale.
Structured postal questionnaires included questions
required to assess the Glasgow outcome scale,20 Barthel
index, and modied Rankin scale. Questionnaires were
sent directly to the surviving patients or carers for
completion at 6 months as a technique of masking
surgeons to the outcome. They were translated into
German, Czech, Spanish, Hungarian, Polish, Russian,
Ukrainian, Swedish, Dutch, Chinese, Hindi, Greek,
Turkish, Latvian, and Lithuanian. Patients family
practitioners (in the UK) or consultants (outside the UK)
were contacted at 4 months to conrm that the patient
was still alive and to conrm his or her place of
residence.
Questionnaires were sent to patients at 5 months for
completion by the patient, relative, or carer, and a
reminder was sent at 6 months. In a few countries where
the postal system was poor, patients were requested to
attend a follow-up clinic where the questionnaires could
be distributed and obtained. Also in some countries
where literacy or language (or dialect) was problematic,
an independent, masked interviewer questioned the
[Link] Vol 365 January 29, 2005

Articles

patients. Further reminders were sent if needed and

exhaustive inquiries were made with neurosurgeons to
conrm the status of patients who failed to respond,
including whether they were still alive or had changed
their address.
In the UK, resource use data relating to types of
surgery, length of stay, and use of services after
discharge were obtained from questionnaires sent to
patients at 3 and 6 months. Hospital records
supplemented these data. Patients in the UK were also
asked to complete the EuroQol at 6 months to rate their
health outcomes. A case report form was completed at
2 weeks or discharge (depending on which was earlier)
to record place of discharge, operative procedures, when
and why procedures took place, and any adverse events.
These forms were also used to record patients Glasgow
coma scores and Glasgow outcome scales, which were
used by the independent data monitoring and ethics
committee to monitor the progress of the trial. These
forms were then returned to the trial ofce in Newcastle,
together with copies of the randomisation forms and
prerandomisation CT scans. Measurements were made
on the prerandomisation scans following a strict
protocol,21 which could be used to conrm the data
reported on the randomisation forms. Postoperative CT
scans were not requested from centres.
Our trial was undertaken in accordance with MRC
guidelines for good clinical practice in clinical trials.
Monitoring visits were undertaken to the major
recruiting centres to ensure adherence to the protocol.
All completed forms were entered onto passwordprotected databases and extensively checked for errors.
Any data omissions or deviations from protocol were
notied immediately to the centre investigators for
correction. Extensive logic checking ensured that the
data were validated before the unmasking of the results.

Statistical analysis
Calculations for sample size were based on the outcome
distribution expected from published work and from a
prospective sample of 259 patients with intracerebral
haemorrhage admitted to Newcastle General Hospital
between Jan 1, 1994, and July 31, 1996. Therefore, with a
favourable outcome of 40% from initial conservative
treatment, a sample size of 800 would be needed to show
a 10% absolute benet from surgery (two-sided
signicance level of 005) with 80% power. A safety
margin of 25% was built in to allow for protocol
violations and crossovers, making a total sample size of
1000.
We anticipated that problems with compliance would
arise (patients might withdraw consent for the
operation, demand surgery after randomisation, or
refuse to complete follow-up questionnaires) and that
complete outcome information would be obtained for
90% of survivors. Potential crossovers included patients
allocated to initial conservative treatment, who were
[Link] Vol 365 January 29, 2005

deemed by the responsible neurosurgeon to need

surgery within the rst 72 h because their condition
deteriorated, and some patients allocated to early surgery
who deteriorated and died before surgery could take
place or whose surgery was delayed because the
operating theatre was needed by another patient. All
analyses were undertaken with SPSS software or
RevMan Analyses 1.0.2.
Analysis was on an intention-to-treat basis using all
available data. Primary outcome analysis was a simple
categorical frequency comparison using a 2 test for
favourable and unfavourable outcomes at 6 months.
Patients were assigned to one of two groups on the basis
of clinical status at randomisation. Prognosis was
estimated from the following equation, which was
derived from observational studies of non-STICH
patients with spontaneous intracerebral haemorrhage:
Prognostic score=(10admission Glasgow coma
score)age (years)(064volume[mL]). Patients were
divided into good and poor prognosis groups by the
median prognostic score. For patients with a poor
prognosis, a favourable outcome included the good
recovery, moderate disability, and upper severe disability
categories of the extended Glasgow outcome scale,
whereas for those with a good prognosis, favourable
outcome encompassed good recovery and moderate
disability.
This
prognosis-based
outcome
for
International STICH was declared before the results
were unmasked.2224 The notion that a different outcome
threshold is justied for more severely affected patients
is familiar to clinicians who normally judge the quality
of an outcome with respect to the severity of the disorder
at presentation. This prognosis-based methodology has
been proposed by several researchers.25,26
Secondary outcomes included mortality, which was
analysed by a simple categorical frequency comparison
and a survival analysis. Prognosis-based outcome
analysis was also used for both the Barthel index and the
modied Rankin scale. For the good prognosis group, a
Barthel index of 95/100 or higher, or a Rankin score of
two or below, were regarded as favourable outcomes,
whereas for those with a poor prognosis, the equivalent
thresholds were 65 or higher for the Barthel index and
three or below for the Rankin score. For all patients,
death was classied as an unfavourable outcome.
All prespecied subgroup analyses were by intention
to treat and compared the primary outcome across the
subgroups with formal interaction tests. The variables
and prespecied subgroups were: age (65 vs 65
years); haematoma volume (50 mL vs 50 mL);
Glasgow coma score (8 vs 9 to 12 vs 13); lobar vs
basal ganglia/thalamic haematoma, or both; thrombolytic or anticoagulant treatment (any vs none);
severity of neurological decit (normal or weak vs
paralysed arm, normal or weak vs paralysed leg, normal
speech vs dysphasia or aphasia); type of intended
operation (craniotomy vs other). Furthermore, side of
389

Articles

haematoma (left vs right), depth from the cortical

surface (1 cm vs 1 cm), and country were used as
minimisation criteria.
A cost analysis covering a period of up to 6 months
after randomisation was undertaken for UK patients.
The unit costs used (base year 2001) were an amalgam of
local costs from one participating centre (Newcastle) and
nationally published gures in those circumstances
where individual unit costs were not available.27
The protocol for this study was peer reviewed and
accepted by The Lancet; a summary of the protocol was
published on the journals website, and the journal then
made a commitment to peer-review the primary clinical
manuscript.19

Men
Age (years)
Pre-ICH modied Rankin index
0
1
2
3
4
5
Pre-ICH mobility
1 200 m or more outdoors
2 Walk indoors
3 Unable to walk
Time between ictus and
randomisation (h)
Glasgow coma score
58
912
1315
Affected arm
Normal or weak
Paralysed
Not assessable
Affected leg
Normal or weak
Paralysed
Not assessable
Speech
Normal
Dysphasic or aphasic
Not assessable

390

Early surgery
(n=503)

Initial conservative
treatment (n=530)

285 (57%)
62 (5270)

306 (58%)
62 (5371)

387 (79%)
81 (16%)
20 (4%)
4 (1%)
0
0

397 (76%)
92 (18%)
21 (4%)
13 (2%)
2
0

477 (97%)
13 (3%)
1

499 (96%)
13 (3%)
6 (1%)

22 (1036)

20 (1035)

99 (20%)
199 (40%)
205 (41%)

106 (20%)
211 (40%)
213 (40%)

195 (39%)
300 (60%)
8 (2%)

225 (42%)
298 (56%)
7 (1%)

245 (49%)
250 (50%)
8 (2%)

269 (51%)
251 (47%)
10 (2%)

130 (26%)
296 (59%)
77 (15%)

142 (27%)
314 (59%)
74 (14%)

Any anticoagulation or thrombolytic

treatment contributing to ICH
Past medical history*
Hypertension
On antihypertensives
Previous myocardial infarction
Previous stroke
Smoker
Other medical disorders

39 (8%)

55 (10%)

341 (69%)
225 (46%)
28 (6%)
30 (6%)
146 (30%)
132 (27%)

378 (72%)
263 (50%)
44 (8%)
43 (8%)
134 (26%)
143 (27%)

Prognostic score
Good prognosis

270 (41 to 500) 292 (31 to 569)

245 (49%)
271 (51%)

Role of the funding source

The sponsor of the study had no role in study design, data
collection, data analysis, data interpretation, or writing of
the report. The corresponding author had full access to
all the data in the study after unmasking and had nal
responsibility for the decision to submit for publication.

Results
1033 patients from 83 centres in 27 countries were
randomised: 503 to early surgery and 530 to initial
conservative treatment. Details of all patients age, sex,
previous medical history, and Glasgow coma score at
presentation are shown in table 1. The groups were well
matched at baseline. More than half the patients were
men and ages ranged between 19 and 93 years, with a
median of 62 years (IQR 5270). Time from ictus to
randomisation varied from 2 to 72 h, with half being
randomised within 20 h (1036). A fth of patients
presented in coma (ie, Glasgow coma score 8),
whereas two-fths had a score of 13 or above.
Haematoma characteristics at randomisation are shown
in table 2. About two-fths of the haematomas were
lobar and a similar number were located in the basal
ganglia or thalamic regions, with the rest extending
through both sites. Slightly more haematomas were
located on the left side than the right. The volume, using
the Broderick method,28 varied from 4 mL to 210 mL
(median 38 mL [2462]) and the median depth from the
cortical surface was 1 cm (02).
The trial prole is shown in gure 1. Eight patients
were withdrawn from the study after randomisation: one
was recruited by an ineligible centre, one was withdrawn
by the centre, ve withdrew consent, and the centre lost
all data for one. Thus, process data were available for
496 patients randomised to early surgery and 529 to
initial conservative treatment. Another 43 patients were
lost between the 2-week follow-up and 6-month followup. For a further 17 patients, their status at 6 months
was not recorded but they were known to have died after
6 months. These 17 patients were included in the
survival analysis as they were known to have been alive
at follow-up, but were excluded from all other analyses
because their Glasgow outcome scales at 6 months were
Early surgery
(n=503)
Site of haematoma
Lobar
Basal ganglia/thalamic
Both
Not assessable
Left side of haematoma
Haematoma volume (mL)*
Minimum depth from cortical
surface (cm)

Initial conservative
treatment (n=530)

196 (39%)
214 (40%)
210 (42%)
224 (42%)
94 (19%)
90 (17%)
3 (1%)
2
265 (53%)
285 (54%)
40 (2463)
37 (23 60)
10 (0120)
10 (0020)

ICH=intracerebral haemorrhage. Data are number of patients (%) or median (IQR).

*Data missing for between nine and 14 patients in early surgery and between three and
ten cases in initial conservative treatment groups. Good prognosis is a score greater
than 27672.

Data are number (%) or median (IQR). *Volume=lengthwidthheight/2.28

Table 1: Baseline characteristics

Table 2: Haematoma characteristics

[Link] Vol 365 January 29, 2005

Articles

Early surgery
(n=465)
Time between ictus and surgery (h)
Surgery ,12 h from ictus
Time between randomisation and
surgery (h)
Surgery ,12 h from randomisation

339 (73%)

35 (25%)

Surgical method
Craniotomy
Burrhole
Endoscopy
Stereotaxy
Other
Not recorded
Additional neuro procedure
Re-evacuation
External ventricular drain
Intracranial pressure monitoring
Other
Not recorded

346 (75%)
37 (8%)
31 (7%)
34 (7%)
16 (3%)
1
72 (16%)
27 (6%)
18 (4%)
10 (2%)
12 (3%)
5 (1%)

119 (85%)
10 (7%)
7 (5%)
3 (2%)
1 (1%)
0
24 (17%)
8 (6%)
8 (6%)
4 (3%)
4 (3%)
0

64 (14%)

23 (16%)

94 (24%)
157 (40%)
145 (37%)

85 (73%)
25 (22%)
6 (5%)

42 (9%)
113 (24%)
239 (51%)
71 (15%)

3 (2%)
17 (12%)
88 (63%)
32 (23%)

49 (11%)
129 (28%)
216 (46%)
71 (15%)

3 (2%)
22 (16%)
83 (59%)
32 (23%)

102 (22%)
93 (20%)
130 (28%)
140 (30%)

11 (8%)
11 (8%)
37 (26%)
81 (58%)

Status before evacuation

Paralysed and sedated
Glasgow coma score
38
912
1315
Affected arm
Normal
Weak
Paralysed
Not assessable/not recorded
Affected leg
Normal
Weak
Paralysed
Not assessable/not recorded
Speech
Normal
Dysphasic
Aphasic
Not assessable/not recorded

30 (1649)
74 (16%)
5 (212)

Initial conservative
treatment (n=140)
60 (2799)
7 (5%)
31 (1182)

Data are number (%) or median (IQR).

Table 3: Surgery details

unknown. Losses to follow-up balanced between the two

groups. Thus, 965 patients had complete follow-up data
for the primary outcome analysis.
Of 496 assessable patients randomised to early
surgery, 465 (94%) underwent surgery (table 3).
However, in 28 (6%), surgery was undertaken more than
24 h after randomisation. Reasons for patients not
receiving surgery were: condition deteriorated (six
patients), second intracerebral haemorrhage or
extension (four), other major clinical event (three),
improvement (four), refusal by relative (six), confusion
over allocation (two), uncertain haematoma size and
location (two), and reason not recorded (four).
Of 529 assessable patients randomised to initial
conservative treatment, 140 (26%) underwent surgery
after an initial period of observation (table 3). Reasons
for these patients undergoing operations were:
rebleeding (17 patients), neurological deterioration (82),
clinical deterioration (20), no improvement on
[Link] Vol 365 January 29, 2005

1033 patients with intracerbral

haemorrhage randomised

503 allocated early

surgery

530 allocated initial

conservative
treatment

7 lost to
follow-up
496 analysed at
2 weeks

1 lost to
follow-up
529 analysed at
2 weeks

19 lost to
follow-up
477 followed up
at 6 months

24 lost to
follow-up
505 followed up
at 6 months

9 alive but
status
unknown
468 analysed at
6 months
6 early follow-up
327 timely follow-up
135 late follow-up

8 alive but
status
unknown
497 analysed at
6 months
4 early follow-up
347 timely follow-up
146 late follow-up

Figure 1: Trial prole

conservative treatment (four), raised intracranial

pressure (three), oedema (ve), altered consciousness
(two), coma (one), aneurysm (one), not waking after
external ventricular drain (one), family request (one),
and reason not recorded (three). The most frequently
used surgical technique was craniotomy (465 patients
[77%]) and table 3 shows details of the patients status
immediately before surgery.
Comparison of tables 1 and 3 shows that patients in the
initial conservative treatment group who received surgery
had deteriorated substantially from their randomisation
level; in fact 59% had deteriorated by three or more
points on the Glasgow coma score. Additionally, those in
this group who went on to have clot evacuation (n=140)
compared with those who did not (n=389) were more
likely to be men (91 [65%] vs 214 [55%], p=00403), and
have haematomas with a volume greater than 50 mL (81
[58%] vs 107 [27%], p00001), be supercial (102 [73%]
vs 180 [46%], p00001), be lobar (71 [51%] vs 142 [37%],
p00001), and haematomas were more likely on the
right side (76 [54%] vs 169 [43%], p=00274).
With the prognosis-based dichotomy of the extended
Glasgow outcome scale, 122 (26%) patients allocated to
early surgery had a favourable outcome at 6 months,
compared with 118 (24%) allocated to initial conservative
treatment (odds ratio 089 [95% CI 066119],
p=0414). Early surgery had an absolute benet of 23%
and a relative benet of 10% (13 to 33; table 4). The
mortality rate at 6 months for the early surgery group
was 36% compared with 37% for the initial conservative
treatment group (odds ratio 095 [073123], p=0707);
391

Articles

Early surgery
(n=468)
Primary outcome
Favourable
122
Unfavourable
346
Not recorded
Secondary outcomes
Mortality
Alive*
304
Dead
173
Prognosis-based modied Rankin index
Favourable
152
Unfavourable
312
Not recorded
4
Prognosis-based Barthel index
Favourable
124
Unfavourable
341
Not recorded
3

Initial conservative treatment

(n=497)

Absolute
benet (95% CI)

(26%)
(74%)

118 (24%)
378 (76%)
1

23 (32 to 77)

(64%)
(36%)

316 (63%)
189 (37%)

12 (49 to 72)

(33%)
(67%)

137 (28%)
351 (72%)
9

47 (12 to 105)

(27%)
(73%)

110 (23%)
377 (77%)
10

41 (14 to 95)

Data are number (%). *Includes 17 patients who were alive at 6 months but status was unknown.

Table 4: Outcomes at 6 months

early surgery had an absolute benet of 12% and a

relative benet of 2% (8 to 11; table 4). Survival during
the rst 6 months did not signicantly differ between
the two groups (log-rank test, p=0678; gure 2).
With the prognosis-based modied Rankin scale, 152
(33%) in the early surgery group had a favourable
outcome compared with 137 (28%) in the initial
conservative treatment group (p=0116); early surgery
had an absolute benet of 47% and a relative benet of
17% (95% CI 4 to 37; table 4). With the prognosis-based
Barthel index, 124 (27%) in the early surgery group had a
favourable outcome compared with 110 (23%) in the
initial conservative treatment group (p=0144); early
surgery had an absolute benet of 41% and a relative
benet of 18% (6 to 42; table 4).
Results of the subgroup analyses are shown in
gure 3. The only subgroup to show heterogeneity of
10
09

Probability of survival

08
07
06
05
04

Early surgery

03

Initial conservative

02
01
0
0
Numbers
at risk (alive)
Early surgery 477
505
Initial
conservative

30

60

90

120
Days

150

180

210

240

366
380

337
349

321
339

314
329

309
324

304
319

304
316

304
316

Figure 2: Kaplan-Meier survival curves

392

treatment response was depth of haematoma from

cortical surface. A favourable outcome from early
surgery was more likely if the haematoma was 1 cm or
less from the cortical surface (absolute benet 8%;
015); interaction between depth from cortical surface
and treatment was signicant (p=002). A favourable
outcome from early surgery was also more likely if the
intended method of evacuation was craniotomy
(absolute benet 6%; 1 to 12), but interaction between
intended method of surgery and treatment was not
signicant (p=007). Open craniotomy was chosen for
346 (75%) patients given early surgery and was
associated with a non-signicant relative benet of 28%
(3 to 59). A uniformly poor outcome was seen in
patients in coma (Glasgow coma score 8): early surgery
increased the relative risk of poor outcome for comatose
patients by 8% (3 to 20).
Patients in the early surgery group who did not have
surgery tended to have a poor outcome: favourable
outcomes for this group applied to only ve (20%) of
25 patients with outcome assessed (741). 29 (22%) of
130 patients with outcome assessed (1530) in the initial
conservative treatment group who went on to have
surgery afterwards, had a favourable outcome.
Our cost analysis included 77 UK patients (37 early
surgery, 40 initial conservative treatment) for whom
complete details of hospital stay could be obtained. All
early surgery patients in this UK substudy had surgery:
nine (24%) had a favourable outcome; 12 (30%) patients
who received initial conservative treatment had surgery:
eight (20%) had a favourable outcome. Total costs per
patient were calculated from surgery, hospital and longterm stay, and allied service costs (physiotherapy,
occupational therapy, speech therapy, and day hospital)
after discharge.
As expected, surgery costs were higher in the early
surgery group than the initial conservative treatment
group (mean 2250 [SD 922] vs 797 [1091], t test
p00001; 1=US $183) because these patients were
more likely to have surgery. In the initial conservative
treatment group, total hospital stay costs were nonsignicantly higher than in the early surgery group
(15 507 [11 593] vs 18 599 [13 911] t test, p=029),
because of a non-signicantly extended stay in hospital
(609 [556] vs 785 [577] days, t test, p=018; median
34 [IQR 1793] vs 73 [20114]). Allied service costs were
also non-signicantly higher in the initial conservative
treatment group than early surgery (mean 695
[SD 1268] vs 1118 [1620], t test, p=021). Mean total cost
for early surgery was 18 452 (12 123) compared with
20 514 (14 163) for initial conservative treatment (t test,
p=050) during the rst 6 months after randomisation,
but again this difference was not signicant.
CT scans were returned for 958 (93%) patients. Of
these, measurements could not be made on eight (of
which one was aneurysmal and another a subdural
haemorrhage). Additionally, another 12 (two cerebellar,
[Link] Vol 365 January 29, 2005

Articles

Study
or
subcategory

Early
surgery
n/N

Initial conservative
treatment
n/N

Age
65
65

182/262
164/206

204/284
174/212

089 (062129)
085 (052139)

GCS
58
912
1315

80/88
140/187
126/193

83/99
158/196
137/201

193 (078475)
072 (044116)
088 (058134)

Side of haematoma
Left hemisphere
Right hemisphere

186/246
160/222

208/265
170/231

085 (056128)
093 (061140)

Site of haematoma
Lobar
Basal ganglia/thalamus

107/181
236/284

130/194
247/300

071 (047108)
105 (069162)

Haematoma volume
50 mL
50 mL

211/302
135/166

238/323
140/173

083 (058117)
103 (060177)

Depth from cortical surface

1 cm
1 cm

170/257
174/208

192/260
184/234

069 (047101)
139 (086225)

Intended method of evacuation

Craniotomy
Others

238/324
108/144

267/337
111/159

073 (051104)
130 (078215)

Deficit of affected arm

Normal/weak
Paralysed

110/182
231/279

135/206
238/284

080 (053121)
093 (060145)

Deficit of affected leg

Normal/weak
Paralysed

150/229
192/232

169/248
201/239

089 (061130)
091 (056148)

Deficit of speech
Normal
Dysphasia/aphasia
Not assessable

72/124
216/276
58/68

92/136
228/289
58/71

066 (040110)
096 (064144)
130 (053320)

38/46
340/450

051 (017146)
093 (069126)

49/62
66/78
15/19
33/42
20/29
16/25
15/20
33/43
30/43
27/34
35/43
39/58

067 (029154)
066 (028155)
100 (021476)
085 (029252)
113 (036350)
069 (021229)
090 (024343)
236 (073761)
087 (033224)
054 (019153)
050 (018139)
179 (077414)

Any thrombolytic or anticoagulant treatment

24/34
Anticoagulant treatment
322/434
No anticoagulant
Country
UK
Germany
Spain
Poland
Latvia
Lithuania
Russia
Czech Republic
Macedonia
South Africa
India
Other countries with
less than 20 patients

43/60
51/65
15/19
25/33
20/28
11/20
19/26
39/44
24/36
29/43
26/38
44/56

Odds ratio
(fixed)
95% CI

01

Odds ratio
(fixed)
95% CI

02
05 10 20
50 100
Favours early
Favours initial
surgery
conservative treatment

Figure 3: Prespecied subgroup analysis

n=number of unfavourable outcomes using prognosis-based outcome. N=number randomised in group. GCS=Glasgow coma score.

nine with brainstem involvement, and one

glioblastoma multiforme) did not full the inclusion
and exclusion criteria but were included in the
intention-to-treat analysis. There were very high
intraclass correlations (ICC) between measurements
reported on the randomisation form and those made on
[Link] Vol 365 January 29, 2005

the CT scans (volume ICC=073 [95% CI 070077];

depth from cortical surface ICC=081 [074085]).

Discussion
Our ndings show that favourable outcomes (from
prognosis-based indices) in patients with intracerebral
393

Articles

haemorrhage treated with early surgery or initial

conservative treatment do not differ signicantly.
Prespecied subgroup analysis also showed little
difference between the two treatments, except for depth
of the haematoma. Patients with haematomas 1 cm or
less from the cortical surface were more likely to have a
favourable outcome from early surgery than those with
deep haematomas.
Randomised controlled trials in surgical patients with
severe neurological disorders, especially those causing
disturbance of consciousness, are difcult to undertake
and probably will become even more problematic in the
near future because of new legislation such as the
European Clinical Trials Directive. International STICH
has taken 8 years to recruit over 1000 patients in a
rigorous attempt to dene the role of early surgery for
patients with spontaneous supratentorial intracerebral
haemorrhage. Although the number of patients
randomised in International STICH exceeds the total
number of patients in all nine previous randomised
controlled trials, our results still leave much
uncertainty.
Careful consideration was given to outcome
assignment in this trial.24 The traditional split into
favourable and unfavourable outcomes using the vepoint Glasgow outcome scale is useful in trials of
younger patients with diffuse brain injury. However,
patients in this study tended to be older (median
62 years) and all had focal neurological decits. For
young patients recovering from head injury or
subarachnoid haemorrhage, any outcome short of
independence outside the home can be classied as
unfavourable, whereas for an older stroke victim,
independence achieved within the home is worthwhile
compared with dependence on 24-h care.
The eight-point Glasgow outcome scale divides the
severe disability category into those who are dependent
and independent within the home. To achieve an
outcome of moderate disability, patients have to be able
to use public transport and shop independently.20
Although this outcome might be expected for a patient
with a minor bleed, a patient initially in coma (Glasgow
coma score 9) or with a severe hemiplegia would not be
expected to reach that kind of independence. We
therefore set out to differentiate treatment targets for
patients according to their initial prognosis (on the basis
of an independently derived formula). This prognosisbased dichotomy substantially improved the power of the
study to detect a modest treatment effect. However, we
still cannot give a denite answer to the question: can a
policy of early surgical intervention for patients with ICH
be recommended and, if so, under what conditions?
The decision to compare a policy of early surgery with
one of initial conservative treatment, including an option
for delayed surgery if necessary, was dictated by practical
and ethical considerations. Recruitment of centres would
have been exceedingly difcult and probably unethical if
394

the protocol prevented neurosurgeons from operating on

patients in the initial conservative treatment group, even
if patients conditions deteriorated.29,30
In fact, this operative intervention took place in about a
quarter of patients in the initial conservative treatment
group. By contrast, only 6% of patients randomised to
the early surgery group did not undergo operation, and
6% had the operation more than 24 h after
randomisation. Patients in the early surgery group who
did not have surgery tended to have a poor outcome:
patients in the initial conservative treatment group who
later had surgery also did not have good outcomes.
However, whether this difference would have been
substantial enough to show a signicant difference
between the two groups is unlikely.
Potential bias in the trial was reduced by use of
concealed randomisation through the CTSU and with
follow-up by postal questionnaire. Investigators had to
provide details of patients before being told of treatment
group assignments and had to provide prerandomisation
CT scans that were used as conrmation. Postal
questionnaires sent to patients for outcome assessment
ensured that this masking was not biased by investigator
perceptions.
There was little difference in mortality (1%) between
the two demographically well-balanced groups. Stroke
physicians might prefer to consider the other secondary
outcome measures (ie, modied Rankin scales and
Barthel indices), because they are commonly used in
stroke trials. However, the potential absolute benet
from early surgery was only 47% for the Rankin scale
and 41% for the Barthel index even with the more
sensitive prognosis-based outcome assignment.
Previously published screening-log data30 set these
STICH results in context with other non-randomised
intracerebral haemorrhage patients.
The overall results of the trial suggest that surgeons
were justied in their uncertainty. Therefore, the
subgroups that might emerge from such a wide
spectrum of clinically relevant characteristics are
clinically important to consider. These subgroups were
prespecied in the protocol but few differences were
seen except in three groups of patients: rst, in patients
in whom the intracerebral haemorrhage reached to
within 1 cm of the cortical surface. In this subgroup (the
rst of three minimisation criteria), there was a 29%
relative benet for early surgery. Can this be interpreted
as a group of patients who might benet from this
policy? Traditional statistical and mathematical opinion
would say no, because subgroup analysis, even if
prespecied, must reach a level of signicance that is
much greater than in the trial itself.31,32 Correction for
12 pre-specied subgroups would require multiplication
of the p value by 12 and thus the signicance would
disappear. Therefore, statistical opinion would conclude
that evidence is insufcient to recommend either policy,
even in this prespecied subgroup.
[Link] Vol 365 January 29, 2005

Articles

Second, there are good clinical reasons for the

assumption that craniotomy (compared with stereotaxic
aspiration or endoscopy) might do harm in deep-seated
intracerebral haemorrhage, whereas in supercial
haematomas, craniotomy is easy and less complicated.
Most operative procedures in this trial were open
craniotomies, so the benecial result for supercial
haematomas might indicate the advantage of clot
evacuation via craniotomy and the disadvantage of open
craniotomy for any haematomas that have not reached
the cortical surface. Trauma of open craniotomy for deep
haematomas might thus offset any benet. This
assumption can be seen, to some extent, by the superior
results in the craniotomy subgroup. For patients
allocated to early surgery in International STICH, the
type of operation was left to the discretion of the
admitting surgeon. Three-quarters of these patients had
open craniotomy, which was associated with a nonsignicant relative benet of 28%.
Third, perhaps the clearest and most disappointing
result was the uniformly poor outcome in patients
presenting with intracerebral haemorrhage in coma. In
those with initial Glasgow coma scores of eight or below,
nearly all were classied as having unfavourable
outcomes (ie, lower severe disability or worse on the
prognosis-based outcome scale). Early surgery raised the
relative risk of poor outcome for these patients by 8%.
Thus, a relative benet of over 3% from surgery, or an
absolute benet of about 25%, can effectively be
excluded, which means that, for patients in coma,
surgery is probably harmful, and even at the most
optimistic estimate, about 40 operations would be
needed to achieve one more favourable outcome
(probably upper severe disability) in such patients.
The philosophy of subgroup analysis needs to be
considered in the context of a pragmatic surgical trial
such as International STICH. Although post-hoc
analysis can generate signicant differences that are
meaningless,32 the main problem arises if signicant
differences are generated that are meaningful, but are
dependent on how the data are classied or analysed.
This problem is especially the case in surgical trials in
which interventions have great risk: so subgroup
analyses need to be considered carefully. If one accepts
that argument, then surgeons, clinicians, and
statisticians need to set the threshold for post-hoc
subgroup analyses and for the prespecied subgroup
analyses. No post-hoc analyses have yet been undertaken
with these data.
Our economic analysis, although limited to UK
patients, suggests that early surgery is less costly than
initial conservative treatment, but again this difference
was not signicant. This reduced cost is associated with
a shorter hospital stay and use of less costly or fewer
services after discharge into the community. Since this
result is not related to a raised mortality rate, patients
having early surgery could recover more quickly than
[Link] Vol 365 January 29, 2005

those with initial conservative treatment, but a much

more detailed study would be needed to investigate this
difference.
These results need to be interpreted with respect to the
nine previous randomised controlled trials in
supratentorial intracerebral haemorrhage.1214,16,17,3336 If all
ten trials are analysed, no net benet from surgery is
seen. However, although some studies12,34 can be
omitted,15 still no overall benet accrues from early
surgery. In future, operative trials of image-guided,
stereotaxically-assisted evacuation of intracerebral
haemorrhage might be justied, especially for deep basal
ganglia haematomas. With supercial haematomas,
craniotomy might be the preferred method for
neurosurgeons comparing the need for early surgery
with a wait and watch conservative policy.
There is insufcient evidence to justify a general policy
of early operative intervention in patients with
spontaneous supratentorial intracerebral haemorrhage,
compared with one of initial conservative treatment.
Patients with supercial haematomas might benet
from surgery, especially by craniotomy, but this
benecial effect needs to be established. The results of
International STICH should encourage surgeons to
organise such trials.
Contributors
A D Mendelow and the steering committee conceived the study. All
listed collaborators contributed to the trial design and interpretation of
the data. B A Gregson and G D Murray analysed the study.
A D Mendelow and B A Gregson drafted the paper, but all listed
collaborators edited and revised the report.
International STICH
Management team
A D Mendelow (principal investigator), B A Gregson (trial director),
L Stobbart (data manager 200104), A J Pearson (data manager
19992001), J Wilson (data manager 199899), H M Fernandes
(principal investigator), D Barer (principal investigator), A Gholkar
(neuroradiologist), J Barnes (research nurse 200203) T Wooldridge
(research nurse 19982001), M S Siddique (research fellow 199899),
K S M Prasad (research fellow 2003), P S Bhattathiri (research fellow
200304), M Deverill (economist 200103), P Mitchell (200204).
Steering committee
M D M Shaw (chair), A D Mendelow, Abbas Karimi (independent
representative), D Barer, G M Teasdale (principal investigator),
D T Hope (independent representative). Observers: G D Murray
(principal investigator and study statistician), H M Fernandes,
B A Gregson, D Morgan (MRC representative 200304), L Cotterill
(MRC representative 200002), H Jones-Jenkins (MRC representative
199899), L Stobbart, A J Pearson, G Taylor (statistician 200203),
J Wilson.
Data monitoring committee
M Harrison (chair), A Skene, R Ross-Russell, A Strong (200104),
F Iannotti (19992000), R Illingworth (199899).
Centre investigatorscity (in alphabetical order), country (number of
patients recruited)
Adelaide, Australia (2)P Reilly; Allentown, PA, USA (2)D J Chang;
Assam, India (2)N C Borah; Athens, Greece (9)G Stranjalis,
S Koras; Bahia Blanca, Argentina (1)G Troccoli; Bangalore, India
(4)S Kolluri; Barcelona, Spain (10)J Cabiol; Beijing, China (2)
Y Zhao; Belfast, UK (5)D Byrnes; Berlin, Germany (11)
A Unterberg, S Kroppenstedt; Bialystok, Poland (30)J Lewko,
P Szydlik; Bilbao, Spain (20)J Garibi, I Pomposo; Birmingham, UK

395

Articles

(2)J Wasserberg; Bloomington, IL, USA (0)K Kattner: Bologna, Italy

(4)P Limoni, E Pozzati; Brno, Czech Republic (24)M Smrcka,
T Svoboda; Brugge, Belgium (0)G Vanhooren; Bucharest, Romania
(0)A Cristescu; Cambridge, UK (8)J D Pickard, P J Kirkpatrick;
Cape Town, South Africa (13)D G Welsh; Debrecen, Hungary (7)
D Bereczki, S Szabo; Dessau, Germany (16)A Kleindienst; Dresden,
Germany (15)G Schakert, J Koy; Dundee, UK (9)S Eljamel;
Dusseldorf, Germany (0)G Woebker; Edinburgh, UK (4)
I R Whittle; Erlangen, Germany (2)R Fahlbusch; Esp Santo, Brazil
(0)J Valladares; Essen, Germany (2)J Pospiech; Frankfurt, Germany
(0)V Siefert; Gdansk, Poland (11)W Wasilewski; Genova, Italy (0)
P Severi; Giessen, Germany (9)W Deinsberger; Glasgow, UK (3)
K W Lindsay; Granada, Spain (1)M J Katati; Gran Canaria, Spain
(1)J Morera-Molina; Gothenburg, Sweden (2)L Pellettleri; Graz,
Austria (2)H Tritthart, F Unger; Griefswald, Germany (2)
M R Gaab; Haywards Heath, UK (3)J Norris, J Goodden; Heraklion,
Greece (7)A Yannopoulos; Homburg, Germany (18)F Cortbus,
K Schwerdtfeger; Hong Kong, China (13)W S Poon; Houston, TX,
USA (0)A B Valadka; Hull, UK (1)D Crimmins; Istanbul, Turkey
(1)S Bahar, Y Krespi; Jackson, MS, USA (1)R Hunt Bobo; Kaunas,
Lithuania (24)A R Gvazdaitas, R Vilcinis; Koln University, Germany
(9)N Klug, C Wedekind; Koln Merhaim, Germany (5)T Klein;
Kubang Kerian University Hospital, Malaysia (7)J M Abdullah,
A R I Ghani Zamzuri; Leeds, UK (0)P Van Hille; Leuven, Belgium
(9)J Gofn; Linkoping, Sweden (5)J Hillman; London Charing
Cross Hospital, UK (0)N Mendoza, K S ONeill: London Atkinson
Morley Hospital, UK (3)B A Bell; London Queens Square Hospital,
UK (2)N Kitchen; London Royal Free Hospital, UK (1)N Dorward;
Lubeck, Germany (2)G Nowak; Lublin, Poland (0)T Trojanowski;
Lund, Sweden (0)H Saveland; Maastricht, Netherlands (10)
G Blaauw; Magdeburg, Germany (2)R Firsching; Mendoza, Argentina
(9)B Odoriz; Middlesbrough, UK (8)S M Marks; Milwaukee, WI,
USA (1)T Genarelli; Monza, Italy (0)S M Gaini; Moscow Burdenko,
Russia (6)S Eliava; Moscow Institute, Russia (2)M Piradov;
Munster Clemens, Germany (37)K R H von Wild, A Sepehrnia;
Munster University, Germany (35)C Schul, H Wassmann; Murnau,
Germany (0)H Jaksche; Newcastle, UK (66)A D Mendelow,
H M Fernandes, P Mitchell; New Delhi, India (58)V S Mehta,
M Tripathi; Nottingham, UK (1)D T Hope; Novosibirsk, Russia (38)
A Krivoshapkin; Oklahoma, USA (3)C M Loftus; Oxford, UK (4)
R S C Kerr; Perth, Australia (0)G J Hankey; Plzen, Czech Republic
(15)P Vacek; Port Elizabeth, South Africa (65)R Keeley; Prague,
Czech Republic (28)M Mohapl; Prague Homolcne, Czech Republic
(3)V Dbaly; Preston, UK (2)C Davis; Riga, Latvia (58)E Valeinis,
I Aksiks; St Gallen, Swtizerland (0)G Hildebrant; Santander, Spain
(9)A Vazquez-Barquero; Shefeld, UK (1)A Kemeny; Siddhapudur,
India (0)R Murali; Singapore (2)J Thomas; Skopje, Macedonia
(79)K Lozance; Stoke on Trent, UK (0)J Singh; Thiruvanathapuram,
India (21)R N Bhattacharya; Toyoake, Japan (10)S Harada; Uppsala,
Sweden (0)L Persson; Usti nad Labem, Czech Republic (20)
M Sames; Uzhgorod, Ukraine (21)V Smolanka; Valladolid, Spain
(0)C Martin; Verona, Italy (1)L Cristofori; Vilnius, Lithuania (23)
G Sustickas; Warsaw, Poland (43)A Marchel; Wiesbaden, Germany
(5)R Schonmayer; Wurzburg, Germany (0)J Meixensberger.
Conict of interest statement
There is a potential conict of interest in countries where
neurosurgeons undertake fee for service medicine. Some members of
the management team were funded by the MRC or the Stroke
Association to undertake the study. There are no other conicts of
interest.
Acknowledgments
We thank the staff of the CTSU in Oxford for their help in providing the
randomisation service. We would especially like to thank the patients
and their relatives who agreed to participate in the study and all the
medical and nursing staff in the many centres who have supported the
trial. The study was funded by the MRC (UK), the Stroke Association
(UK), and the Northern Brainwave Appeal.
References
1
Dennis M. Outcome after brain haemorrhage. Cerebrovas Dis 2003;
16 (suppl 1): 913.

396

10
11

12

13

14

15

16

17

18

19
20

21

22

23

Yang GY, Betz AL, Hoff JT. The effects of blood or plasma clot on
brain edema in the rat with intracerebral hemorrhage.
Acta Neurochirurgica supplementum 1994; 60: 55557.
Lee KR, Colon GP, Betz AL, Keep RF, Kim S, Hoff JT. Edema from
intracerebral hemorrhage: the role of thrombin. J Neurosurg 1996;
84: 9196.
Lee KR, Kawai N, Kim S, Sagher O, Hoff JT. Mechanisms of edema
formation after intracerebral hemorrhage: effects of thrombin on
cerebral blood ow, blood-brain barrier permeability, and cell
survival in a rat model. J Neurosurg 1997; 86: 27228.
Xi G, Wagner KR, Keep RF, et al. Role of blood clot formation on
early edema development after experimental intracerebral
hemorrhage. Stroke 1998; 29: 258086.
Xi G, Keep RF, Hoff JT. Erythrocytes and delayed brain edema
formation following intracerebral hemorrhage in rats. J Neurosurg
1998; 89: 99196.
Nehls DG, Mendelow AD, Graham DI, Teasdale GM, McCulloch J.
Experimental intracerebral haemorrhage: early removal of a
spontaneous mass lesion improves late outcome. Neurosurgery
1990; 27: 67582.
Kingman TA, Mendelow AD, Graham DI, Teasdale GM.
Experimental intracerebral mass:- time-related effects on local
cerebral blood-ow. J Neurosurg 1987; 67: 73238.
Siddique MS, Fernandes HM, Wooldridge TD, Fenwick JD,
Slomka P, Mendelow AD. Reversible ischemia around
intracerebral hemorrhage: a single-photon emission computerized
tomography study. J Neurosurg 2002; 96: 73641.
Heiss WD. Ischemic penumbra: evidence from functional imaging
in man. J Cereb Blood Flow Metab 2000; 20: 127693.
Fernandes HM, Siddique S, Banister K, et al. Continuous
monitoring of ICP and CPP following ICH and its relationship to
clinical, radiological, and surgical parameters. Acta Neurochirurgica
2000; 76 (suppl): 46366.
McKissock W, Richardson A, Taylor J. Primary intracerebral
haemorrhage: a controlled trial of surgical and conservative
treatment in 180 unselected cases. Lancet 1961; 278: 22126.
Auer LM, Deinsberger W, Niederkorn K, et al. Endoscopic surgery
versus medical treatment for spontaneous intracerebral hematoma:
a randomized study. J Neurosurg 1989; 70: 53035.
Juvela S, Heiskanen O, Poranen A, et al. The treatment of
spontaneous intracerebral hemorrhage: a prospective randomized
trial of surgical and conservative treatment. J Neurosurg 1989; 70:
75558.
Fernandes HM, Gregson B, Siddique S, Mendelow AD. Surgery in
intracerebral hemorrhage: the uncertainty continues. Stroke 2000;
31: 251116.
Teernstra O, Evers S, Lodder J, Leffers P, Franke C, Blaauw G.
Stereotactic treatment of intracerebral hematoma by means of
plasminogen activator: a multicentre randomized controlled trial
(SICHPA). Stroke 2003; 34: 96874.
Hosseini H, Leguerinel C, Hariz M, et al. Stereotactic aspiration of
deep intracerebral hematomas under computed tomographic
control: a multicentric prospective randomised trial. 12th European
Stroke Conference 2003, Valencia, Spain: 57.
Kanaya H, Kuroda K. Intracerebral hematomas. In: Kaufman HH,
ed. Development in neurosurgical approaches to hypertensive
intracerebral haemorrhage in Japan. New York: Raven Press, 1992:
197209.
Anon. Protocol 99PRT/7 International STICH (Surgical Trial in
Intracerebral Haemorrhage). Lancet, 1999.
Wilson JT, Pettigrew LE, Teasdale GM. Structured interviews
for the Glasgow outcome scale and the extended Glasgow
outcome scale: guidelines for their use. J Neurotrauma 1998;
15: 57385.
Bhattathiri PS, Gregson B, Prasad KS, et al. Reliability assessment
of computerized tomography scanning measurements in
intracerebral hematoma. Neurosurg Focus 2003; 15: E6.
Anon. Protocol 99PRT/7 International STICH (Surgical Trial in
Intracerebral Haemorrhage). Lancet, 2003. [Link]
com/info/[Link]?n1=authorinfo&n2=Protocol+reviews&uid=
14545 (July 14, 2003).
Shaw MDM, Mendelow AD, Teasdale GM, Murray GD,
Gregson BA. Alterations to STICH protocol. Lancet 2003; 362:
1244.

[Link] Vol 365 January 29, 2005

Articles

24

25
26

27

28
29
30

31

Mendelow AD, Teasdale GM, Barer D, Fernandes HM,

Murray GD, Gregson BA. Outcome assignment in the
international surgical trial in intracerebral haemorrhage.
Acta Neurochirurgica (Eur J Neurosurg) 2003; 145: 67981.
Berge E, Barer D. Could stroke trials be missing important
treatment effects. Cerebrovasc Dis 2002; 13: 7375.
Murray GD, Barer D, Choi S, et al. Design and analysis of phase III
trials with ordered outcome scalesthe concept of the sliding
dichotomy. J Neurotrauma (in press).
Netten A, Rees T, Harrison G. Unit costs of health and social care.
Canterbury: Personal Social Services Research Unit, University of
Kent at Canterbury, 2001.
Broderick JP, Brott TG, Grotta JC. Intracerebral hemorrhage
volume measurement. Stroke 1994; 25: 1081.
Fernandes HM, Mendelow AD. Spontaneous intracerebral
haemorrhage: a surgical dilemma. Br J Neurosurg 1999; 13: 38994.
Gregson BA, Mendelow AD, on behalf of the STICH Investigators.
International variations in surgical practice for spontaneous
intracerebral haemorrhage. Stroke 2003; 34: 259398.
Altman DG, Matthews JN. Statistics notes. Interaction 1:
heterogeneity of effects. Br Med J 1996; 313: 486.

[Link] Vol 365 January 29, 2005

32

33

34

35

36

ISIS-2 Collaborative Group. Randomised trial of intravenous

streptokinase, oral aspirin, both or neither among 17 187 cases of
suspected acute myocardial infarction: ISIS-2. Lancet 1988; 332:
34960.
Batjer HH, Reisch JS, Allen BC, Plaizier LJ, Su CJ. Failure of
surgery to improve outcome in hypertensive putaminal
hemorrhage: a prospective randomized trial. Arch Neurol 1990; 47:
110306.
Chen X, Yang H, Cheng Z. The comparative study of the total
medical and surgical treatment of hypertensive intracerebral
haemorrhage. Acta Academia Medicinae Shanghai 1992; 19: 23440.
Zuccarello M, Brott T, Derex L, et al. Early surgical treatment for
supratentorial intracerebral haemorrhage: a randomized feasibility
study. Stroke 1999; 30:183339.
Morgenstern LB, Frankowski RF, Shedden P, Pasteur W,
Grotta JC. Surgical treatment for intracerebral hemorrhage
(STICH). A single-center, randomized clinical trial. Neurology 1998;
51: 135963.

397