вход по аккаунту


Controversy in the interpretation of clinical trials.

код для вставкиСкачать
or cardiac arrhythmias, which might have contributed to
cardiac emboli productive of stroke. I t was a trial that attempted to establish or deny the value of two platelctinhibitory drugs in patients who were suffering from, as
nearly as clinical and angiographic data could define, T I A
and minor stroke related to embolic phenomena of
artery-to-artery origin. The full name of the trial, indeed,
was the Canadian Cooperative Study on Recent Cerebral
Emboli of Arterial Origin.
Surgical critics [9, lo] have been concerned about the
lack of emphasis given to the possible benefit of surgery for
such patients as were entered in this trial, and have asked
questions concerning the severity of angiographically demonstrated lesions related to the outcome. A complete report analyzing angiographic findings relative to the clinical
picture and results of therapy is in preparation. Preliminary
analysis indicates that a majority of the patients were
afflicted with obstructions which enthusiasts of endarterectomy would consider “surgical lesions.” Since carotid
endarterectomy has not yet been accredited in a formal
study satisfying the design requirements of modern scientific method, it was thought that the patients in our study
were not denied “standard therapy” [2, 51. However, the
study was in no way intended to deny or prove the value of
In o u r judgment, the medical community requires validation of all forms of medical therapy and is beginning to
be equally Jemanding of good evidence for proposed surgical therapy. I t is n o longer satisfactory to take short-cuts
to these answers by the doubtful strategy of “using patients
as their own controls,” by utilizing surgical morbidity and
mortality figures alone as a basis for performing an operation, or by comparing surgical results with dissimilar
nonoperaced and unrandomized cases. Randomized clinical
trials are heavy consumers of time, energy, and money, but
in an illness such as threatened stroke with an extremely
varied and unpredictable outcome, alternative avenues to
arrive at decisions about therapeutic advantage are more
likely to lead to uncertainty.
We look forward to continuing discussion on this and
other trials of means to reduce the toll of disability and
untimely death in threatened stroke. In these attempts,
despite some recent commentary to the contrary 161, the
scientific community of neurologists and neurosurgeons
must band together with biostatisticians and methodologists and bring the expertise of all these specialties to
bear on the problem. Otherwise, our critics will recall the
words of George Bernard Shaw, when he stated: “ I t does
happen exceptionally that a practising doctor makes a contribution to science; but it happens much oftener that he
draws disastrous conclusions from his clinical experience
because he has no conception of scientific method, and
believes, like any rustic, that the handling of evidence and
statistics needs no expertness” [8].
1. Fields WS, Lemak NA, Frankowski RF, et al: Controlled trial
of aspirin in cerebral ischemia. Stroke 8:301-316, 1977
2. Fields WS, Maslenikov V, Meyer JS, et al: Joint study of
extracranial arterial occlusion. Progress report of prognosis
following surgery or nonsurgical treatment for transient cere-
bral ischemic attacks and cervical carotid artery lesions.JAMA
211:1993-2003, 1970
Harris WH, Salzman WE, Athanasoulis CA, et al: Aspirin
prophylaxis of venous thromboembolism after total hip replacement. N Engl J Med 297:1246-1249, 1977
Kelton JG, Hirsh J, Carter CJ, et al: Sex differences in t h e
antithrombotic effects of aspirin. Blood 52:1073- 1076, 1978
Kurtzke JF: Formal discussion: Effect of duration of treatment
with anticoagulants o n the prognosis of transient ischemic attacks, in Cerebral Vascular Diseases: Ninth Princeton Conference. New York, Grune & Strarton, 1974, pp 190-193
Millikan CH: Clinical investigation of things or people (guest
editorial). Neurology (Minneap) 281744-745, 1978
Preventing stroke (editorial). Br Med J 2:454-455, 1978
Shaw GB: The Doctor’s Dilemma. New York, Penguin, 1977,
P 27
9. Thompson JE: Correspondence. N Engl J Med 299954.1978
10. Trout HH: Correspondence. N Engl J Med 299:954, 1978
11. Whisnant J: Correspondence. N Engl J Med 299:953, 1978
Controversy in the
Interpretation of Clinical Trials
Prof P. Armitage
A good deal of evidence suggests that attempts to compare the effectiveness o f therapeutic regimens other than
by carefully organized trials with random allocation are
likely to be useless, except in the rare situations in which
large and immediate improvements occur. It does not follow that every randomized trial provides clear evidence
about the relative merits of the treatments under study or
clear guidance for future clinical practice. A trial may have
been poorly organized or executed; or, although impeccable in design and conduct, it may be too small to provide
any useful information.
Increasingly, we seem to read reports of trials conducted
with meticulous care, involving tremendous resources of
money, manpower, and time, that are nevertheless unable
to settle burning issues in an unambiguous way. These situations arise particularly in the long-term treatment of
chronic diseases, in which the rate of occurrence of adverse
events is relatively low and large numbers of patients must
be treated for substantial periods. T h e prevention of stroke
in patients who have suffered at least one transient ischemic attack provides an example. Both the Aspirin in Transient Ischemic Attacks trial [23 and the Canadian Cooperative Study [ 11 were well conceived and well executed. Yet
their findings leave room for alternative interpretations. Dr
Kurtzke proposes reasons for doubting the Canadian authors’ conclusion that aspirin is beneficial in the prevention
of stroke; the Canadians make a spirited reply.
Dr Kurtzke performs a useful service in drawing attention to Type I1 errors in significance tests: a nonsignificant
result may arise even when a true difference exists. This
potential error is widely recognized by clinical invesFrom the Department of Biomathematics, University of Oxford,
Oxford, England.
Address reprint requests to Prof Armitage, Pusey Street, Oxford,
OX 1 252, England.
Notes and Letters
tigators, and a cooperative group will usually try to guard
against the chance of missing important differences when
deciding on the size and length of a study [ 3 ] . However,
Kurttke applies the argument to consideration of the interaction between aspirin and sulfinpyrazone. The Canadian Study Group found no significant evidence of an interaction; but, he argues, an interaction might have gone
undetected. Indeed, the data point somewhat in that direction. The plain fact of the matter is that the data are consistent with a range of possibilities, including the Canadian authors’ hypothesis that aspirin has an effect and
sulhnpyrazone does not, and Kurtzke’s extreme hypothesis
that only the combined treatment (A+,S+ in his table) has
an effect. A further problem is the apparent interaction
with sex: not only does the aspirin effect occur only in men
(for which group it is highly significant), but the interaction
between the two drugs is also more marked for males than
for females.
If alternative interpretations cannot be ruled out by further scrutiny of the data, can other approaches resolve the
issues? I believe three considerations broadly, although not
conclusively, favor the Canadian authors’ interpretation.
First, the simplest explanation is that aspirin is beneficial in
the presence or absence of sulfinpyrazone. (Occam’s razor
might be invoked here.) In addition, other evidence 121
favors aspirin. Finally, the protagonists are using the word
interaction in a strictly statistical sense, to indicate that the
effect of one drug varies with the presence or absence of
the other drug. If each drug is effective alone, it is virtually
certain that an interaction of some sort exists. Absence of
interaction means that effects are additive, and whether this
is so must depend o n the particular response variable used.
I t is unlikely, for any chosen response variable, that effects
are exactly additive, and therefore it is likely that some
degree of interaction exists. This last consideration might
appear to favor Kurtzke’s interpretation. However, his hypothesis in its extreme form implies a special type of interaction: each drug separately is ineffective but both together are effective. This suggests either marked drug
interaction in the pharmacological sense, for which there
is no supporting evidence, or a highly nonlinear dose/
602 Annals of Neurology
Vol 5
No 6 June 1979
response relationship (for example, such that the doses of
each drug used are ineffective, but that double the doses
are highly effective). Either possibility seems implausible,
though by no means inconceivable.
Additional evidence on the effects of aspirin and
sulfinpyrazone in the secondary prevention of stroke would
be welcome. However, it remains to be seen whether further large-scale trials can be mounted. Potential investigators will have to consider whether, to increase the
prospects of obtaining definitive results, they can plan and
execute even larger studies than the Canadian trial. In many
long-term trials, in which results are regularly monitored,
investigators have felt compelled on ethical grounds to stop
at a point where differences are, at most, only mildly
significant. W e should perhaps consider whether the ethical
balance now tilts in the direction of longer continuation of
trials: marginal significance levels, with perhaps inadequate
information about side-effects, are a poor basis for the
choice of future treatment.
Investigators will also need to consider whether in future
trials they can properly include a placebo. The balance of
evidence is in favor of a beneficial effect from aspirin in
males, either alone or in combination with sulfinpyrazone,
and many will agree with the Canadian authors [ 11 that the
next step might be to compare aspirin alone with the combination of aspirin and sulfinpyrazone. Others will feel that
more resources must be spent in a final attempt to clarify
the aspirin effect. Perhaps both strategies can be pursued
by different teams. The effort will be considerable, but the
therapeutic gain may be immense.
1. The Canadian Cooperative Study Group: A randomized trial of
aspirin and sulhnpyrazone in threatened stroke. N Engl J Med
299:53-59, 1978
2. Fields WS, Lemak N A , Frankowski RF, et al: Controlled trial of
aspirin in cerebral ischemia. Stroke 8:301-316, 1977
3. Pocock SJ, Arrnitage P, Galton D A G : The size of cancer clinical trials: an international survey, in Methods and Impact of
Controlled Therapeutic Trials in Cancer (UICC Technical Report 36). Geneva, Union Internationale contre le Cancer,
1978, Part 1, pp 5-25
Без категории
Размер файла
207 Кб
clinical, controversy, trials, interpretation
Пожаловаться на содержимое документа