可以通过检验,然后比较其作用的大小

1. If the model assumpions hold, you migh use ANOVA.
2. If the model assumption is violated, you might perform non-parametric tests.
3. MANOVA (Multivariate ANOVA) can test the 5 biomarks together, but it does not compare the 5 biomarks.
4. Maybe you can build a logistic regression, using treatment or control as response, and the 5 biomarks as predictors. After variable selection,you might find which biomark is better.
5. You might check the correlations beween the 5 biomarks.
6. For the prognosis, I don't know what variable you are going to use. If it is the time free of symptoms, you might use COX models.

As I don't know much about your study, I can only say this. Hope more friend join the discussion.

谢谢二位，我想问下ROC曲线的可信区间为0-1，这个是哪里出现问题了呢？应该都是具体到多少，0-1这个不对吧