You can select cases to include or exclude those above or below the desired thresholds. For example, exclude cases where variable x is greater than or equal to 300. This would eliminate those with doses at 600 and 800 leaving 200 and 400. Then you can "flip" the data by selecting out those below and analyze those at the 600 and 800 dose. Does that make sense?
Wait are you saying you want the 4 dose measures (i.e. 4 levels) transformed to one IV with only 1 level (200 and 400 as one)? Or did you want to perform a 2x2 using dosage as IV. This is a bit different than what I mentioned above. In that you can simply go to Transform - Visual Binning then break your variable into respective parts. Or you can do a median or mean split the hard way and recode them into categorical variables 1=200 & 2=400. I am sure I my explanation is confusing as hell. I tried . Much easier to perform than to explain.