Non-statistician needs help!

I am not a statistician, just a medical doctor trying to do my own statistics and now I have run into some problems that I cannot solve by myself. I think I need to explain some background first:
  1. Carboplatin (chemotherapy) is dosed using Calvert formula: Dose (mg) = AUC (mg ml-1 min) x [GFR (ml/min) + 25 (ml/min)]
  2. GFR is estimated using the Cockcroft-Gault formula: [(140-age) x (Wt in kg) x (0.85 if female)] / (72 x Cr)
The Cockcroft-Gault formula may overestimate eGFR and to avoid toxicity, GFR is often capped at a certain level, for example 125 ml/min. The aim was then to compare of capped patients had an inferior outcome in terms of residual disease at surgery. We found that capped patients had a significant worries outcome. However, the groups were not completely balanced as patients in the capped group was younger (which is expected since their GFR higher (= better kidney function) and more overweight (also expected as weight is a variable in the Cockcroft-Gault formula used to estimate the GFR).
  • Since capping, the variable of interest, is not independent of either weight/BMI or age, can all be included in the multivariable analysis?
  • Moreover, it is well known that the Cockroft-Gault formula can overestimate the GFR in obese patients, which would mean that obese patients are prescribed higher doses than they should have and they should therefore have a superior outcome. However, when I look at our data I see the opposite, i.e. that obese patients have an inferior outcome both in the capped group and in the non-capped group.
So can anyone advise me how to analyse this when the variables are dependent of each other?
Last edited: