вход по аккаунту


Controversy in neurology The canadian study on TIA and aspirin. A critique of the canadian TIA study

код для вставкиСкачать
4. Singer HS, Pepple JM.Ramage AL, et al: Gillrs de la Tourette
syndrome: further studies and thoughts. Ann Neurol4:21-25,
5. Van Woert MH, Yip LC, Balis ME: Purine phosphoribosyltransferase in Gilles de la Tourette syndrome. N Engl J Med
296:210-212. 1077
Chronic Subdural
Hematoma Simulating
Transient Ischemic Attacks
Roger S. Williams. M D
The brief report of Robin et a1 [2] in the August, 1978,
issue of the Annals appeared as we completed evaluation of
an identical case, suggesting that this syndrome may not be
A 48-year-old man was in an automobile accident in April,
1978, in which he sustained momentary unconsciousness, a
fracture of the left humerus, and contusion of the left thigh.
Neurological examination was normal during the initial
three-week hospitalization. He embarked on a business
trip to Europe in May, during which time he developed
headaches of increasing frequency and severity. Six weeks
after the accident his right arm fell suddenly to his side at
dinner. Paralysis was abrupt and was associated with tingling paresthesias of the right face and hand. The sensorimotor deficit was maximum for about three minutes
and receded gradually and completely over five minutes.
Symptoms recurred while he was shaving the next morning, during which time he was momentarily unable to speak
but was heard to utter a few unintelligible sounds. His
symptoms resolved within ten minutes but recurred three
times in the next 48 hours, prompting admission to the
Maida Vale Hospital in London (Mr David Thomas). Cranial C T scan revealed a substantial low-density subdural
fluid collection on the left side, which was drained via burr
holes. The headache disappeared promptly, and transient
neurological deficits did not recur until three weeks later,
when sudden tingling paresthesias of the right face and arm
were associated with faintness and resolved gradually over
15 minutes. Momentary paresthesias of the face and arm
recurred daily over the next week, prompting hospital admission upon his return home. Physical examination revealed a blood pressure of 130/80, absence of carotid
bruits, and normal heart sounds. Neurological examination
was remarkable only for a right extensor plantar response.
An electroencephalogram revealed intermittent, irregular
6 to 9 H z activity in the left temporal region. A cranial C T
scan was normal except for bony surgical defects. No abnormalities were noted on a left common carotid arteriogram. The patient refused medications, and his symptoms
disappeared gradually over a period of three weeks. He had
been symptom free for three months at this writing.
The pathogenesis of transient focal sensorimotor deticits
in our patient and those of Robin et al[2] is not known. T h e
abnormalities must relate in some way to cerebral trauma,
and their paroxysmal and intermittent character suggests
the possibility of a focal seizure discharge. However, in a
recent review of 127 cases of somatosensory epilepsy-an
uncommon seizure syndrome--only 11 patients with
paroxysmal paresthesias experienced simultaneous focal
paresis not preceded by tonic-clonic motor activity [ 13.
Cerebral trauma was a common predisposing factor (18%,),
but only 3 cases occurred in association with chronic subdural hematoma, and only 1 patient developed symptoms
prior to surgical drainage.
Although our case and those cited by Robin et al establish that transient focal sensorimotor deficits may occur in
association with and following removal of chronic subdural
hematoma, the symptoms are indistinguishable clinically
from those of transient cerebral ischemia. When such
symptoms occur in association with chronic subdural
hematoma, as identified by cranial scan, and persist for
more than a short interval following its removal, carotid
arteriography may be necessary to exclude the presence of
coexistent vascular disease.
1. Mauguicre F, Courjon J: Somatosensory epilepsy: a review of
127 cases. Brain 101:307-332, 1978
2. Robin JJ. Maxwell JA, Pitkethly DT: Chronic subdural
hematoma simulating transient ischemic attacks. Ann Neurol
4:154, 1978
Controversy in Neurology:
the Canadian Study on TIA
and Aspirin
A Critique of the Canadian TIA Study
John F. Kurtzke, M D
Perhaps the most important recent treatment trial in neurology has been the multicentered Canadian study on transient ischemic attacks (TIA) under the direction of H.J. M.
Barnett [ 11. Under a randomized, double-blind, placebocontrolled regimen, patients were given total daily dosages
of either: aspirin, 1,300 mg; sulfinpyrazone, 800 mg; both
drugs; or neither drug (placebo). The study groups totalled
From the Departments of Neurology and Community Medicine,
Georgetown University School of Medicine, and the Neurology
Service, Veterans Administration Medical Center, Washington,
From the Neurology Service, Massachusetts General Hospital,
Boston. MA 021 14.
Accepted for publication Feb 23, 1979.
Address reprint requests to Dr Kurtzke, Neurology Service, Vcterans Administration Medical Center, 50 Irving St, NW, Washington, D C 20422.
Notes and Letters
585 patients followed for an average of twenty-six months.
Treatment failure was defined on two levels, the first including either first T I A o r a “hard end point,” and the other
the “hard end points” alone. The latter were stroke or
death (all deaths but 1 were vascular). Periods of observation began after one week of therapy and continued
through six months after withdrawal from treatment,
whenever this had occurred during the trial.
In concept, control, and performance, this was an outstanding work; its difficulties are really appreciated only by
those involved in similar endeavors. There were precise
definitions of patients for inclusion and exclusion, efforts to
obtain all suitable prospects, maximal coordination and ongoing control among the centers, and a study as large asand for as long as-seemed practical. T h e able leadership of
Dr Barnett is evident, and the prestigious McMaster University biostatisticians were integrally involved from start
to finish.
Questions as to its conduct could be raised, such as the
observation periods defined and the inclusion of “minor
stroke” patients. But these were valid decisions made in
advance by the group. My own bias would have been to
limit the “hard end points” to stroke alone, whether fatal or
not. To me, the essential question in TIA is: does a treatment prevent stroke? 1 believe a totally different study
should be designed if one is considering TIA as a risk factor
for all atherothrombotic disease throughout the body. But
again, the decision was the authors’, and the data available
permit assessment of stroke alone as an index of treatment
failure. However, since sample size was naturally predicated on their end-points, it then proved to be insufficient
for decision, as we shall see.
The essential conclusion made by the Canadians was that
the use of aspirin “reduced the risk . . . of stroke or death
by 31 percent ( P < 0.05)” [ 11. T h e basis for this conclusion
lies in the top part of the Table. Results are presented as
ratios of observed to expected end-points, so that ratios
under 1.OO represent “success.” The ratio 0.68 is the result
for those who received both drugs (A+,S+); that of 1.21,
those who received only sulhnpyrazone (A-,S+); that of
0.96, those receiving only aspirin (A+,S-); and that of
1.14, those receiving placebo (A-,S-). “Aspirin effect”
was declared to be the sum of the A + column ( A + , S + and
A+,S-) compared to the sum of the A- column. This
provides the cited risk reduction of 3 1‘‘i ( 1 - [A+/A-] =
1 - 0.69 x 100%).
The second part of the Table shows the results when the
end-points are limited to stroke alone, whether fatal or not.
By the same calculations. the risk reduction for the aspirin
group is 25C6, and by a simple chi-square test, this no
l o n g e r differs significantly from expectation at the 5<4
But the basic problem lies in the validity of this method
of assessment, which I contend is not appropriate here. T o
explain this. we need to invoke biostatistics. When we
compare results between two groups by means of some
statistical test and say that the groups differ significantly at
the 5% level, what is meant is that we reject at this probability level the null hypothesis-here, that there is no difference between the groups. What this rejection means is
that, if each group were a random sample of its parent
Annals of Neurology
Vol 5 N o 6 June 1979
“ H a r d End Poinis“ i n C a n a d i a n TIA Study
Expressed as Obsen>edlExpertedRatios
by Treatment. with Risk Reduriion Percentages
Hard End Point
Stroke or death
All stroke
aspirin; S = sulhnpyrazone; RR = risk reduction
”population” or “universe,” there is only 1 chance in 20
that samples differing by this amount or more would be
found if they were in fact random samples of the same
This explanation refers to what is called the a or Type 1
error: the chance o r probability of incorrectly rejecting the
null hypothesis. But there is also a /3 o r Type I 1 error,
which is the probability of incorrectly acrepiing the null
hypothesis of no difference when in fact the groups are noi
samples of the same parent population. This can also be
translated into the “power” ( 1 - p ) of a test. In simple
terms, it is an answer to the question, how much might the
groups differ ( b e drawn from different populations) even
though no “significant” difference is observed? Too often
we jump to the false conclusion that when no difference is
Jemonstrated ( a error) there is no difference in fact. O u r
proper conclusion from the usual Type I testing is that if
the test result does not attain o u r significance level, a difference between the groups is not proved. It does not mean
that we have proved there is no difierence between them.
This distinction is directly relevant to the factorial design
of analysis used by the Canadians. Among the four cells in
the Table, it is valid to add columns or rows only i fth ere is
no ititeruction. To physicians, of course, drug interaction is a
common occurrence. To statisticians it has a special meaning. In the model the Canadians have used, statistical interaction can be defined as the dgference between the effects of aspirin in the presence of sulfinpyrazone and in its
absence. In other words, it is the difference of the differences between rows and columns. If there is no interaction, this difference should be 0. T h e aspirin difference
with sulfinpyrazone is 0.53 (1.21 - 0.68); without the
drug it is 0.18 (1.14 - 0.96). Subtracting 0.18 from 0.53
leaves 0.35. If 0.53 is taken as “maximal” aspirin effect,
then two-thirds of this effect (0.35/0.5 3 ) could be attributed to “interaction.”
The Canadians, of course, tested their data for this possibility and properly stated that there is no significant interaction between the therapies. But this speaks only to the
a or Type 1 error. And when the validity of the analysis
rests entirely on the fact ofno interaction, I d o not believe
that assessment is sufficient.
In the lower part of the Table are a set of observed to
expected ratios labeled “Hypothetical.” These are what
would be found if there were total interaction-in
words, if each drug singly were useless, but in combination
they produced a notable therapeutic effect. In fact, if one
compares this hypothetical set with a set constructed on the
null hypothesis of no interaction, using a total N of 120 ( 3 0
per cell), that null hypothesis would not be rejected ( x p 3=
4.707; p > 0.10). This means that, if this comparison is
appropriate. it would appear impossible to reject even
100%) interaction in rhe Canadian study with the numbers
at hand. Therefore I believe it is not permissible to proceed
as if there were no interaction. To me, the only valid assessment for this study is that made directly among four
regimens: both drugs; each drug; placebo. And here, from
the tests I was able to d o with the data published, I could
find no significant difference among the regimens.
This argument holds as well for the reported difference
in aspirin effect by sex, wherein the Canadians concluded
that “among men, the risk reduction for stroke or death
was 48 percent ( P < 0.005), whereas no significant trend
was observed among women” [ 11. In the analyses by separate regimen, my comparisons between the sexes showed
no significant differences.
My conclusions from this really excellent study are: that
the decision in favor of aspirin is not valid; that any true
drug effect appears to lie with the combination treatment
and not with aspirin; that the differences by sex are not
significant; and that even the apparent benefit of combined
therapy does not attain statistical significance with the tests
I was able to use. Accordingly, to me, the efficacy of these
modalities of treatment in TIA remains unproved.
The Canadian Cooperative Study Group: A randomized trial of
aspirin and sulfinpyrazone in threatened stroke. N Engl J Med
200~53-59, 1978.
H. J. M . Barnett, FRCP(C), M. Gent, BSc, MSc,
D. L. Sackett, MSc, MD, and D. W. Taylor, MA
O n Behalf of the Canadian Cooperative Study Group
We have been gratified by the response of the scientific
community to our publication of the results of the Canadian Cooperative Study. Although dissenting views have
been expressed, it is through proposition and rebuttal of
such alternative explanations that we move closer to the
truth. In responding to Dr Kurtzke, we shall repeat our
conclusions from the trial and summarize the bases for
them. We will identify what we regard as the flaws in Dr
Kurtzke’s alternative explanation and briefly discuss the
other published criticisms.
In the Neu England Journal of Medicine report we concluded: “ I t would be reasonable, o n the basis of these two
studies (American and Canadian], to recommend that men
Address reprint requests to Dr Barnett, Department of Clinical
Neurological Sciences, University Hospital. PO Box 5339, Postal
Station A, London, Ont, Canada N6A 5A5.
with transient cerebral ischemia or minor strokes who are
able to tolerate it be treated with aspirin.” Furthermore,
we stated: “Although a life-table analysis did not show a
statistically significant interaction between aspirin and
sulfinpyrazone, a trend favoring the combination therapy
was observed among men. This finding raises the possibility
that the combination of aspirin and sulfinpyrazone is the
treatment of choice in men, but such a conclusion would
have to be tested by a formal study comparing aspirin with
the combination of aspirin and sulfinpyrazone.”
The bases for these conclusions can be summarized as
1. Two independent randomized clinical trials (the Cana-
dian Cooperative Study and an American trial of aspirin
[ 11) yielded consistent results on the benefit of aspirin in
threatened stroke.
2. Among all study subjects in the Canadian trial, aspirin
produced a risk reduction of 19% (p c 0.05) for continued TIA, stroke, or death and a risk reduction of
3 1%’ ( p c 0.05) for the harder, more important events
of stroke or death. Since no statistically significant interaction between the study drugs was observed, the
orthodox and appropriate statistical analysis was the one
employed in our report.
3. There was a statistically significant interaction between
sex and response to aspirin: women did not benefit, but
in men the risk of stroke or death fell by 48% ( p =S
0.005). Furthermore, the validity of this sex difference
has received additional support both from other trials of
aspirin in thrombbembolism (men, but not women,
showed a beneficial response to aspirin in the prevention of thromboembolism after total hip replacement
[3]) and from laboratory experiments (male, but not
female, rabbics were protected by aspirin from experimentally induced thrombosis [ 4 ] ) .
4 . There was no statistically significant interaction between
the study drugs. Thus, although one cannot conclude
that no interaction occurred, there is certainly no persuasive evidence of drug synergism or antagonism.
Nonetheless, two interesting trends in the data were
noted. First, men receiving the combination tended to
d o better than men taking aspirin alone, and for this
reason we suggested that a subsequent trial to determine whether the trend was real would be of interest.
The second trend was for men taking sulfinpyrazone to
d o worse than men being given placebo. The overall
interaction (which, all agree, was not statistically
significant) resulted from both trends; thus, if men on
sulfinpyrazone had done as well as or better than men
o n placebo, the overall trend toward drug interaction
would have been small indeed!
Dr Kurtzke’s dissenting view rests on three assumptions:
(1) that one can isolate that portion of the aspirin benefit
which is due to drug interaction; ( 2 ) that the trend toward
drug interaction arises solely from the tendency of men
receiving the combination to fare better than men taking
aspirin alone; and (3) that one can restrict analysis of the
Canadian trial to the events of stroke and stroke death, and
can exclude subjects reported to have died from other
Notes and Letters
Без категории
Размер файла
313 Кб
tia, aspirin, stud, neurology, canadian, controversy, critique
Пожаловаться на содержимое документа