<International Circulation>: at the ESC 2012 with Professor Pocock.  Composite endpoints are widely used in clinical trials at present.  What are the major drawbacks of composite endpoints and how can we improve the usage of these composite endpoints?
Professor Pocock :We need composite endpoints because looking at a single endpoint such as cardiovascular death， we may not have enough power， not enough events to come to reliable conclusions.  So we choose a composite of several events such as cardiac event， myocardial infarction， or stroke to combine the evidence to get enough power to look at treatment differences.  The main problem is that composite endpoints vary in their clinical importance.  Obviously a cardiac death is worse than a non-fatal myocardial infarction for the patient.  So that is the first problem， you are comparing events of unequal importance.  Secondlarythe treatment effects may differ in the different components， so you might have a big effect on myocardial infarction but an opposite effect on stroke and so you must avoid coming to na?ve conclusions when you are looking at a composite.

　　Professor Pocock : Additionally， repeat events with an individual are ignored in a composite analysis because typically all you look at is the time to first event.  So if an individuals has had recurrent hospitalizations for example， these will be ignored in a composite analysis.

　　《国际循环》：当前在临床试验中广泛使用复合终点，复合终点的主要缺陷是什么？如何改进对复合终点的使用？
Pocock博士：我们需要复合终点，这是由于如果观察心血管死亡这样的单一终点，可能不具备足够的统计学效能，没有足够的事件以支持得出可靠的结论。因此，我们选择了几个事件（例如心脏事件、心梗或卒中）的复合终点，把证据联合起来以获得足够的统计学效能来观察治疗之间的差异。主要问题是不同复合终点的临床重要性有所不同。显然，对一位患者来说，心血管死亡比非致死性心梗更糟糕。因此，这就产生了第一个问题，我们比较的事件重要性并不相等。其次，治疗对不同终点的效应可能不一致，治疗可能对心梗的作用更为明显，但是对卒中的作用刚好相反，因此解读复合终点时，必须要避免得出肤浅的结论。另外，复合终点分析中并未包括一个患者重复发生的事件，因为通常我们只是观察至首次发生某事件的时间。因此，比如说某个患者再次住院，这并不会被纳入复合终点的分析当中去。


<International Circulation>: what are some major problems in the design of of randomized clinical trials for cardiovascular diseases?
Professor Pocock: You mean in general?  Well that is a big question. The key issues are always to decide which patients you are studying and exactly what the treatment comparison is going to be and what are the outcome measures you are going to study.  And then to do the trial with sufficient quality; so there are no biases in the issues of comparison， issues with randomization and blinding.  And the fifth issue is we need to make the trial big enough with long enough follow up so you get reliable conclusions.  Now iI could pick away at each of those items， which would I pick?  The biggest problem historically is that trials were not big enough with sufficient follow up， I think that is one of the things.  I think I would stick at that one as the single big statistical item that needs most improvement.

　　《国际循环》：心血管疾病的随机临床试验在设计上有哪些主要问题？
Pocock博士：你是说总体上来讲？这是个大的问题。核心一直是决定要研究哪些患者，要比较哪些治疗，要观察哪些终点指标。然后是试验设计的质量要合乎要求，包括比较时没有偏倚、随机的问题和盲法的问题。还有样本量要足够大、随访时间足够长，以得出可靠的结论。现在我想挑一个方面来讲。哪一个呢？以往的试验最大的问题是样本量不够大，随访时间不够长，我想这是一个问题。我觉得这是最大的统计学问题，也是最需要改进的地方。

<International Circulation>: What is your opinion on studies using surrogate endpoints for cardiovascular diseases and how would you estimate the correlation between surrogate endpoints and hard endpoints?
Professor Pocock: Yes for surrogate endpoints， we used to rely on them much more in trials of cardiovascular diseases.  For instance to show that an antihypertensive agent could be licensed， all you had to show historically that it lowered blood pressure.  You did not have to show that it reduced the risk of stroke or myocardial infarction.  In heart failure drugs used to be improved without looking at the hard endpoints of hospitalization and mortality， so I think we have learned not to trust surrogate endpoints now in deciding whether a new drug treatment or new device should be brought on to the market.  I think we have some stories where the surrogate endpoint has not matched well to the serious clinical endpoints.  I can remember one clinical trial of a drug called Moxonidine， which the surrogate endpoint was norepinephrine level change was staggeringly wonderful but the drug actually caused an increase in mortality when it was tried out in a major trial. So your second part was about how we look at the relationship between the surrogate endpoint and the hard endpoint， you can only really get to that in a trial when you are doing the trial with the hard endpoint.  I suppose you could also study observational data on the link between surrogates and hard endpoints.  Technically it is not difficult to do you just need quality databases in either observational studies， registries， or in trials to provide that.  So it is important to do that.

　　《国际循环》：你如何看待采用了心血管疾病替代终点的研究？你如何估计替代终点和硬终点之间的相关性？
Pocock博士：过去我们在心血管疾病的试验中更多地采用替代终点。例如，想让降压药物获得批准，过去只有证明该药能够降低血压就行了。不需要证明该药能够降低卒中或心梗的风险。心衰治疗药物只要能够改善心衰就行，不用观察药物对住院和死亡这样的硬终点的影响。因此，我想现在我们在决定一个新药或新型器械能否上市时，已经改变为不依赖替代终点了。我想有一些研究提示我们，替代终点并不能够很好地对应临床硬终点。我记得有一个药物叫莫索尼定，该药对替代终点去甲肾上腺素水平的作用相当不错，但实际上一项重要的试验显示该药增加死亡率。你的第二个问题是如何看待替代终点和硬终点之间的关系。只有在观察硬终点的研究中，我们才能了解到替代终点和硬终点之间的关系。另外，也可以通过观察性研究的数据来分析替代终点和硬终点之间的关系。从技术上讲这没什么难的，只需要有高质量的数据库，无论是观察性研究、注册研究还是其他临床试验的数据库。做到这一点很重要。
 

[1]  [2]  下一页