| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
(Stroke. 2003;34:2065.)
© 2003 American Heart Association, Inc.
Comments, Opinions, and Reviews |
Epidemiology Division, National Stroke Research Institute, West Heidelberg, Victoria, Australia
Prevention of intracerebral hemorrhage (ICH) is still the most effective method for reducing the impact of this devastating disease. To facilitate better development of prevention programs, we need precise estimates of risk and elucidation of minor risk factors.
In the past, there have been some major difficulties surrounding the identification of risk factors for ICH. The first is that ICH is a relatively uncommon disease, making up between 10% and 15% of all strokes in Western countries.1 Consequently, to ensure adequate cases of ICH, cohort studies must be very large, and patients must be accrued over many years. Alternatively, case-control methodology could be used and, if carefully conducted to reduce bias, may provide accurate results. Moreover, it is only since the CT era that ICH could be studied as a separate entity with diagnostic accuracy. Inclusion of cases with other types of stroke that have differing etiologies would result in a weakening of any associations identified.
To address some of the limitations of small data sets and diversity of risk factors studied, Ariesen and colleagues2 have conducted a meta-analysis of previous work in this area. This is a timely analysis, given the number of publications in recent years. Apart from age and sex, they have confirmed that the most important risk factor for ICH is hypertension.35 Despite different definitions of hypertension in these studies, the relationship between hypertension and ICH is clear. They have also confirmed an association between heavy alcohol consumption and ICH.4,6,7 Further support for alcohol consumption as a risk factor is provided by their summary showing increasing odds ratios with increasing level of alcohol consumption. The summary odds ratios are substantial, with an approximate doubling in the odds ratio of ICH with each decade of life, whereas among men, heavy drinkers, and hypertensive people, the risk of ICH approximately triples.
Although the associations between these 4 factors and ICH are impressive, the summary odds ratios for the case-control studies are subject to some imprecision for 2 main reasons. First, as acknowledged by the authors, only univariate odds ratios were summarized. The use of multivariate analyses would have enabled the isolation of the exposure (eg, hypertension) from the effects of any imbalance in the distribution of potentially confounding factors between cases and controls and would show whether the association is truly independent. Without this adjustment, the odds ratio can be altered in either direction; ie, it can be overestimated or underestimated.8 Second, some of the studies included in the meta-analyses were based on matched samples. Matching eliminates substantial imbalance and therefore controls for confounding on the matched variables. Because the matching was not retained in the analyses of these matched studies, there is a tendency for the odds ratio to be biased toward unity.8 Alternative analytic techniques incorporating the use of odds ratios and standard errors (or confidence intervals)9 would enable the production of summary odds ratios based on both adjusted and, when appropriate, matched analyses, as well as enabling inclusion of studies without published raw data.3,6
Because of the likely imprecision in the odds ratios, particular caution must be taken in interpreting the results of the lesser potential risk factors, ie, smoking, hypercholesterolemia, and diabetes. Because the reported odds ratios are small (in the order of 1.2 to 1.3) and because there has been no adjustment for potentially confounding factors, these findings may be spurious.
In support of their findings, Ariesen et al also summarized relative risks from cohort studies (some were adjusted and others were not). These results were both in agreement with those of the case-control summary odds ratios (hypertension, smoking, and diabetes) and in disagreement with them (alcohol consumption and hypercholesterolemia). The disparate findings for alcohol consumption can be explained by the fact that alcohol consumption was not subcategorized according to level of consumption for the cohort studies. The findings for hypercholesterolemia were not in agreement between the cohort and case-control studies, making interpretation difficult.
In summary, Ariesen and colleagues have made an important contribution to our understanding of risk factors for ICH. Hypertension and heavy alcohol consumption are important and preventable risk factors for ICH. Treatment and management of hypertension remain the most effective prevention strategy for ICH, given both the relatively high prevalence of hypertension within the community and the strong association between hypertension and ICH. In regard to the other potential risk factors (hypercholesterolemia, smoking, and diabetes), the most important message is that if these factors increase the risk of ICH, this risk is modest. At present, there still remains insufficient evidence that these factors are important contributors to the development of ICH.
| References |
|---|
|
|
|---|
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
Stroke Home | Subscriptions | Archives | Feedback | Authors | Help | AHA Journals Home | Search Copyright © 2003 American Heart Association, Inc. All rights reserved. Unauthorized use prohibited. |