3-group analysis

Hello, I have little project, and today doubts just brought me down.

I have sample of 25 patients, and 3 types of disease (a person can have more than one disease). Also i have a bunch of data about every person (like blood sugar).
I would like to compare means (lets stick to blood sugar example) of those 3 groups but I don't know how being in different groups at the time will affect it. On top of that Im limited to nonparametrical tests (lack of normality).
Can i use three Mann-Whitney tests, or must I use something like Kruskal-Wallis ? Or maybe something else ?

Thank You for help