但是在人的试验上,有各种因素会影响达到ideal。首先基因上人作为高级动物,基因复杂度比低等动物要高,from person to person的variation也比较大,做到对照组跟试验组DNA组成的similarity不可能接近到跟老鼠一样。而其他因素,传统意义上的control ,因为cost,时间和sample size等原因,一般只能控制<10种因素,比如:性别一致,种族(race)一致,年龄大体相当(比如,常见标准是within 2-5 yrs之类),BMI属于同一个大体范畴(比如一般划分为太瘦,正常,超重3个非常宽松的档),和对被研究的疾病有重大影响的其他药物使用情况一致等。经常不能被控制的,比如人吃不吃维生素拉,饮食锻炼习惯拉,这些只能指望randomize以后大体一致。所以作为西医basis的对照组假设,只能是“大体”成立。这个“大体”,在好多情况下已经good enough,但不是说it’s the best of all worlds,它离causal inference需要的counterfactual ideal有差距。. .и
我不清楚中医如何work的,但是lack of statistical evidence of a significant treatment effect 并不足以说明中医没有用。It is possible (but dont ask me for a specific likelihood) 中医属于personalized medication 这个范畴。一个可能是,it does not work in a uniform way across the whole population, but rather yields very specific benefit for a highly specific subgroup. If we were to identify what works for whom, it is likely that traditional medicine can complement current advances in western medicine.
对于“中医worked for a specific person that I know of and/or I think highly of"是否可以作为证据说明中医有疗效,我考虑有如下几种可能:1placebo effect(比如被某大神摸一下就觉得自己百病消的人,轮子转转奏是这样的),2原来本身就是误诊,3的确非常有效,但是因为原因不明,无法readily applied to the next person who suffers from apparently similar symptoms. 4,随机出现疗效的case被大量宣传,而数量更多的毫无疗效的 case作为silent evidence被suppressed. (国内大量所谓营养品广告属于这个范畴)。其实Anecdote 和 case study是不可以用来做causal inference的。5,机理被研究的很清楚,可以明确而准确的apply到下一个suitable patient 身上,哪怕下一个病人看起来症状不一样(这个是理想医学状态了,我管这个也叫counterfactual ideal)。
其实现在西医里一个热门方向就是personalized medicine. 因为传统办法tend to ignore the inter-personal difference。比如,有的药物 at best也只对部分人(hypothetically, let's say it works for 10% of the targeted diseased population)有作用,同时对90%吃药的人有相对轻微的副作用。如果能有办法identify which of the 10% will respond to the drug and only give them the medication, a larger percentage of people can avoid taking some harsh medication that will not help them but give them hasty side effects.
好像大家都同意,现在国内的很多中医的确就是骗子。但是在有足够证据dismiss 中医的validity之前,我选择不totally dismiss it. 不知道说清楚没,whew…….跟写GT作文一样。。。。。。。。。。。。