Sanofi_edit_Layout 1 22/10/2009 15:19 Page 55
Assessing Therapeutic Options and Individualising Treatment According to Patient Needs
Subgroups in Stroke Table 1: Chance Findings in Subgroups
Stroke is a very heterogeneous clinical syndrome. Using the Trial of ORG
10172 in Acute Stroke Treatment (TOAST) classification,
6
approximately Astrological Birth Sign Deaths 2P
15% of stroke cases are cardioembolic, 15–20% are caused by large-
ASA Placebo
artery disease (thromboembolic or haemodynamic), 20% by small-
Libra or Gemini 150 147 NS
All other signs 645 869 <0.000001
vessel disease, 5–10% are ‘other’ (dissection, venous, genetic or
cerebral amyloid arteriopathy) and 40% are undetermined. There are
ASA = acetylsalicylic acid; NS = not significant.
many examples of treatment that affect the subtypes differently. For
example, warfarin is a good treatment option for cardioembolic stroke Figure 1: Effect of Endarterectomy for Asymptomatic
or patients with atrial fibrillation (AF), but is harmful in patients with
Carotid Stenosis on the Risk of Any Stroke and
Operative Death by Sex
small-vessel disease. Even within particular subtypes there is evidence
that treatment should be tailored to individuals. For example, there is
Events/Patients
the question of whether blood pressure should be treated in patients
Subgroup Surgical Medical OR 95% CI
with severe carotid occlusive disease. For the majority of patients with
Males
symptomatic carotid stenosis, the risk of suffering a stroke increases
ACTST 51/1,021 97/1,023 0.50 0.35–0.72
with blood pressure. This relationship holds in patients with only
ACAS 18/544 38/547 0.46 0.26–0.81
unilateral severe carotid stenosis or occlusion, but is reversed in
Total 69/1,565 135/1,570 0.49 0.36–0.66
patients with bilateral >70% carotid stenosis or occlusion, suggesting
Females
that aggressive blood pressure lowering may be harmful in this group.
7
ACTST 31/539 34/537 0.90 0.55–1.49
ACAS 15/281 14/287 1.10 0.52–1.32
More subtle physiological differences can also influence the effects of
Total 46/820 48/824 0.96 0.63–1.45
treatments. For example, there are differences between the sexes in
0 0.5 1 1.5
the pathology of atherosclerotic plaques. Women tend to have smooth
Odds ratio (95% CI)
stenoses with fibrous caps and occasional endothelial erosion,
ACTST = Asymptomatic Carotid Surgery Trial; ACAS = Asymptomatic Carotid
whereas men are more likely to have thin fibrous caps and ruptured
Atherosclerosis Study; OR = odds ratio; CI = confidence interval.
8
Adapted from Rothwell et al., 2004.
plaques. Therefore, the expectation is that carotid endarterectomy
(CEA) for asymptomatic stenoses will be less effective for women then
for men, and the data reflect that. Subgroup analysis of the Absolute Risk
Asymptomatic Carotid Surgery Trial (ACST) and the Asymptomatic It is a commonly held misconception that the overall result of a trial
Carotid Atherosclerosis Study (ACAS) show that benefit of CEA at five- is a good measure of treatment in the average patient. In fact, the
year follow-up is confined to men, with a statistically significant truth is sometimes diametrically opposite: the overall results are
subgroup–treatment effect interaction (see Figure 1).
8
more often driven by the effect of treatment in a small number of
high-risk individuals.
11
The average patient often provides very little
information; they are often relatively low-risk and therefore
contribute few events to the trial outcome. This effect is often seen
There is increasing evidence that the when trial populations are stratified according to their predicted
risks and benefits of treatments used
baseline risk. Such stratification is particularly helpful when the trial
treatment itself has a risk of complications, such that treatment may
in stroke medicine do differ between not be justified in patients who have a low risk of a poor outcome
subgroups and individuals.
without treatment.
By contrast, high-risk patients are most likely to benefit from more
aggressive treatment, and there are several situations in stroke
However, subgroup analyses do not usually show significant medicine in which it is possible to use validated risk scores to identify
heterogeneity. In fact, they often disprove firmly held pre hoc such individuals. For example, the ABCD2 score can be used to
hypotheses that a particular group will benefit more or less than others. identify patients at high risk of stroke after a TIA.
12,13
A validated score is
For example, clinicians expected that CEA would be less effective in available to identify high-risk patients with symptomatic carotid
patients with lacunar stroke than in those with apparently stenosis,
11
and the CHADS score is widely used in patients with AF.
14
In
thromboembolic events. However, a subgroup analysis of data from the primary prevention, the Framingham risk model is widely used to predict
Carotid Endarterectomy Trialists’ Collaboration (CETC) showed a greater the 10-year risk of stroke. When trial results are stratified using such risk
reduction in risk of recurrent stroke in patients randomised after a scores, it is sometimes possible to demonstrate increases in both
lacunar transient ischaemic attack (TIA) or stroke.
9
Therefore, subgroup relative and absolute risk reduction with increasing baseline risk.
11,15
analysis is often useful in preventing clinicians from targeting treatments
too narrowly. Subgroup analysis is perhaps most useful not in identifying Summary and Conclusions
differences in responsiveness to treatment between different There is increasing evidence that the risks and benefits of treatments
pathological or physiological groups, but in identifying more mundane used in stroke medicine do differ between subgroups and individuals.
interactions with factors such as the urgency with which treatment is More effort needs to be made to identify differences between
given or the stage of disease at which it is used. For example, the pre-defined clinically important subgroups and to determine the
benefits of CEA in patients with symptomatic carotid stenosis falls effects of stratification of trial populations by predicted baseline risk
rapidly with delay from the presenting TIA or stroke to surgery.
10
of stroke and/or other important outcomes. n
EUROPEAN NEUROLOGICAL REVIEW 55