Kind of a tough one without fully understanding your process. People often measure one attribute by measuring something different and using a correlation. For example, when you measure temperature with an electrical temperature probe, you are not really measuring temperature, you are measuring the change in current flowing through a metal whose conductivity changes with temperature. Do you have to say you are measuring electrical current? No, because we know that the relationship between current and temperature is extremely good, so we just call it temperature. If you are using a standard, widely accepted way of computing RDS, then I see no problem with calling that the dependent variable. If you are computing a new way of calculating RDS, then I think you need worry about how noise in your model correlation interacts with your experiment. In that case, I would call the amount of glucose after chemical digestion the dependent variable and then do the correlation between that and RDS by using an approved technique. Either that or use the approved technique in the first place to directly measure RDS.
Does that help?