Sanofi_edit_Layout 1 22/10/2009 15:19 Page 54
Brain Trauma Stroke
Understanding Risks – Assessing Therapeutic Options and
Individualising Treatment According to Patient Needs
Satellite Symposium – ‘Understanding Risks: Assessing Therapeutic Options and
Individualising Treatment According to Patient Needs’
Sponsored by sanofi-aventis and Bristol-Myers Squibb
Held on 27 May 2009 during the 18th European Stroke Conference in Stockholm.
The following article comprises summaries of the four presentations given at the symposium. The views and opinions expressed,
including some discussion of off-label usage, are those of the authors and not necessarily those of sanofi-aventis or Bristol-Myers Squibb.
Individualised Treatment Considerations
relation to clinical events.
3
However, while the theoretical benefits
in Stroke Prevention
of targeting treatments in this manner may be accepted, reliable
data on likely effects of treatment in subgroups or individuals are
not always available. N-of-1 trials are not possible for most
Peter M Rothwell interventions and so clinicians must extrapolate from grouped data
using either subgroup analysis or risk modelling. Rarely, multiple
Professor of Clinical Neurology, University of Oxford different trials are performed in different groups of patients with
specific indications, as was the case with clopidogrel, in which case
This article serves as an introduction to the problems and potential the overall trial results can be used to explore any heterogeneity
benefits of trying to target stroke treatments at individuals rather than of treatment effect.
simply prescribing a one-size-fits-all regimen. There are a range of
opinions on whether it is possible to genuinely target treatments at Subgroup Analysis
individuals or if it will always be the case that doctors have to rely on Subgroup analysis is often criticised as it can produce chance
group data from large clinical trials. Individualising treatment is not a findings with low reliability. An oft-cited example is from the Second
new concept. In 1320 Henri de Mondeville wrote in his Chirurgie: International Study of Infarct Survival (ISIS-2), detailing the effect of
“Anyone who believes that anything can be suited to everyone is a acetylsalicylic acid (ASA) in acute myocardial infarction (MI) by
great fool, because medicine is practised not on mankind in general, birth sign. It would appear from this analysis that ASA is highly
but on every individual in particular.” However, others have argued effective for people born under most birth signs, but ineffective in
that “…it would be unfortunate if desire for the perfect (i.e. knowledge those born under Libra and Gemini (see Table 1).
4,5
However, this
of exactly who will benefit from treatment) were to become the example is a little unfair. The researchers simply combined the
enemy of the possible (i.e. knowledge of the direction and two (non-adjacent) birth signs that happened to have the least
approximate size of the effects of treatment of wide categories of treatment effect. A formal test of subgroup–treatment effect
patient).”
1
Nevertheless, there have been attempts to try to bring interaction across all 12 signs of the zodiac, which would be the
these extremes of view together to find the middle ground:
2
only appropriate analysis, would show no statistically significant
heterogeneity. However, there are genuine examples of subgroup
“Some argue that when large randomised controlled trials are analyses that were considered to be potentially genuine when
performed, the effects of most medical interventions are relatively first reported but that were shown not to be true. These include
modest, and that very large pragmatic trials with broad entry criteria the findings that:
3
are therefore necessary to have the statistical power even to quantify
these modest overall effects reliably. Others argue that the effects of ASA is ineffective in women;
treatment are often so modest precisely because the trials are antihypertensive treatment is ineffective for primary prevention in
performed in such heterogeneous populations of patients, and that women and in the elderly;
stratification into less heterogeneous clinical subgroups or risk beta-blockers are ineffective after acute MI in the elderly and in
groups is necessary.” patients with inferior MI;
thrombolysis is ineffective more than six hours after acute MI and
There are several reasons why targeting treatments may be in patients with a previous MI; and
necessary. Differences in treatment effects between subgroups or tamoxifen is ineffective in women with breast cancer below 50
individuals may be due to differences in the absolute risk of a poor years of age.
outcome without treatment or to differences in risk of
complications of treatment, differences in underlying pathology However, although some of these subgroup observations did show a
(stroke is a good example of a clinical syndrome with multiple statistically significant subgroup–treatment effect interaction, none
underlying pathologies that do not all respond in the same way to was pre-defined. Post hoc subgroup analysis is analogous to betting
treatment) and differences in severity of disease, natural history of on a horse after the race has finished. If completed correctly with
disease, stage of disease (i.e. symptomatic versus asymptomatic), subgroups defined in advance and with appropriate tests of statistical
and treatment effect may depend on the timing of treatment in significance, subgroup analysis is a reliable tool.
54 © TOUCH BRIEFINGS 2009