Truelsen T, Lindenstrøm E, Boysen G
Department of Neurology, Hvidovre Hospital, Denmark.
Stroke. 1994 Apr;25(4):802-7. doi: 10.1161/01.str.25.4.802.
We wished to test the validity of a stroke probability point system from the Framingham Study for a sample of the population of Copenhagen, Denmark. In the Framingham cohort, the regression model of Cox established the effect on stroke of the following factors: age, systolic blood pressure, the use of antihypertensive therapy, diabetes mellitus, cigarette smoking, prior cardiovascular disease, atrial fibrillation, and left ventricular hypertrophy. Derived from this model, stroke probabilities were computed for each sex based on a point system. The authors claimed that a physician can use this system for individual stroke prediction.
The Copenhagen City Heart Study is a prospective survey of 19,698 women and men aged 20 years or older invited to two cardiovascular examinations at 5-year intervals. The baseline examination included 3015 men and 3501 women aged 55 to 84 years; 474 stroke events occurred during 10 years of follow-up. In both cohorts initial cases of stroke and transient ischemic attack recorded during 10 years of follow-up were used. We used the statistical model from the Framingham Study to establish a corresponding stroke probability point system using data from the Copenhagen City Heart Study population. We then compared the effects of the relevant risk factors, their combinations, and the corresponding stroke probabilities. We also assessed stroke events during 10 years of follow-up in several subgroups of the Copenhagen population with different combinations of risk factors.
For the Copenhagen City Heart Study population some of the risk factors (diabetes mellitus, cigarette smoking, atrial fibrillation, and left ventricular hypertrophy) had regression coefficients different from those of the Framingham Study population. Consequently, the probability of stroke for persons presenting these risk factors and their combinations varied between the two studies. For some other risk factors (age, blood pressure, and cardiovascular disease), no major differences were found. The recorded frequency of stroke events in subgroups of the Copenhagen population was compatible with the estimated probability intervals of stroke from the Copenhagen City Heart Study and with those from the Framingham Study, but these intervals were very large.
The majority of risk factors for stroke identified by the Framingham Study also had a significant effect in the Copenhagen City Heart Study population. The differences found could be due partly to different definitions of these factors used by the two studies. Although estimated stroke probabilities based on point systems from the Copenhagen City Heart Study and the Framingham Study were similar, the points scored in the two systems did not always correspond to the same combination of risk factors. Such systems can be used for estimating stroke probability in a given population, provided that the statistical confidence limits are known and the definitions of risk factors are compatible. However, because of the large statistical uncertainty, a prognostic index should not be applied for individual prediction unless it is used as an indicator of high relative risk associated with the simultaneous presence of several risk factors.
我们希望检验弗雷明汉研究中的卒中概率评分系统对丹麦哥本哈根一部分人群的有效性。在弗雷明汉队列研究中，考克斯回归模型确定了以下因素对卒中的影响：年龄、收缩压、抗高血压治疗的使用、糖尿病、吸烟、既往心血管疾病、心房颤动和左心室肥厚。基于该模型，根据评分系统计算了每种性别的卒中概率。作者声称医生可使用该系统进行个体卒中预测。
哥本哈根市心脏研究是一项对19698名20岁及以上男女进行的前瞻性调查，这些人每隔5年接受两次心血管检查。基线检查纳入了3015名年龄在55至84岁的男性和3501名年龄在55至84岁的女性；在10年的随访期间发生了474例卒中事件。在这两个队列中，均使用了随访10年期间记录的卒中及短暂性脑缺血发作的初始病例。我们使用弗雷明汉研究的统计模型，利用哥本哈根市心脏研究人群的数据建立了相应的卒中概率评分系统。然后我们比较了相关危险因素、其组合以及相应卒中概率的影响。我们还评估了哥本哈根人群中具有不同危险因素组合的几个亚组在10年随访期间的卒中事件。
对于哥本哈根市心脏研究人群，一些危险因素（糖尿病、吸烟、心房颤动和左心室肥厚）的回归系数与弗雷明汉研究人群的不同。因此，在这两项研究中，存在这些危险因素及其组合的人群发生卒中的概率有所不同。对于其他一些危险因素（年龄、血压和心血管疾病），未发现重大差异。哥本哈根人群亚组中记录的卒中事件发生频率与哥本哈根市心脏研究以及弗雷明汉研究估计的卒中概率区间相符，但这些区间非常大。
弗雷明汉研究确定的大多数卒中危险因素在哥本哈根市心脏研究人群中也有显著影响。发现的差异可能部分归因于两项研究对这些因素的不同定义。尽管基于哥本哈根市心脏研究和弗雷明汉研究的评分系统估计的卒中概率相似，但两个系统中的得分并不总是对应相同的危险因素组合。只要知道统计置信限且危险因素的定义一致，此类系统可用于估计特定人群中的卒中概率。然而，由于存在较大的统计不确定性，除非将其用作与几种危险因素同时存在相关的高相对风险的指标，否则预后指数不应应用于个体预测。
