Background: The Cockcroft and Gault (C-G) method of calculating creatinine clearance (CLcr) does not appear to perform well at low concentrations of serum creatinine and the extremes of body sizes.
Aim: To evaluate the performance of a modified C-G equation to predict gentamicin clearance.
Methods: Demographics and gentamicin pharmacokinetic parameters were collected from 999 subjects as part of routine clinical care1. Linear regression analyses were performed to determine the relationships between gentamicin clearance and CLcr as calculated by the original C-G equation2 or via modified versions of the C-G equation. The C-G equation was modified in 2 ways: 1) Total Body Weight (TBW) was replaced with Lean Body Weight (LBW)3 and the sex effect was retained, i.e. 0.85, and 2) TBW was replaced with LBW and the sex effect removed as sex is included in the LBW equation. A serum creatinine round-up method was applied to patients with a serum creatinine concentration less than 0.06 mmol l-1 for all analyses.
To assess whether the relationships observed with the modified C-G equations above held true when building covariate models, an analysis was conducted using the dataset and pharmacokinetic models as described by Matthews et al.,4. In summary, the dataset comprised of 697 adult patients with a median (range) of 3 (1-20) concentrations per patient, resulting in a total of 2567 serum gentamicin concentrations for population pharmacokinetic analysis.
In the previous analysis4, the typical value of clearance (TVCL) was estimated as the sum of renal and non-renal components of clearance i.e. TVCL=CLR + CLNR where CLR=CLR*RF*FSZCL and CLNR=CLNR*FSZCL. Renal Function (RF) was defined as a typical population value adjusted by fractional changes in creatinine production, age and sex i.e. RF=CPR*FSEX*FAGE, and FSZCL is the fractional effect of size on clearance.
Using this previous structural model, we evaluated three different covariate models: 1) the best covariate model by Matthews et al., where FSZCL=(TBW/70)0.75 and the estimated sex effect was 0.82, 2) FSZCL=LBW/70 with a estimated sex effect, and 3) FSZCL=LBW/70 with no sex effect estimated i.e. fixed to 1. The selection of the best covariate model was based on the reduction in objective function.
Results: Linear regression analysis showed that the C-G equation calculated by LBW without the 0.85 sex effect had the best correlation with gentamicin clearance (r2=0.7372, ME=4.23, RMSE=16.1) when compared to the original C-G equation (r2=0.6734, ME=-20.18, RMSE=31.4).
Both modified C-G models which incorporated LBW with or without the sex effect showed superiority over the original covariate model i.e. allometrically scaled TBW and an estimated sex effect (p<0.001). The sex effect when estimated with LBW in the modified C-G model was 0.95, implying that there is only a minimal sex effect when using LBW in the C-G equation. Importantly, there was no statistically significant difference between the two LBW covariate models based on reduction in objective function (p>0.05).
Conclusion: We have shown that CLcr estimated by covariate models which include LBW better describe gentamicin clearance in these datasets. These findings can potentially provide better predictions of gentamicin clearance, thus improving initial dosing estimates for target concentration intervention.
1. Duffull S.B. et al. (2004). Clin Pharmacokinet. 43(15):1167-78
2. Cockcroft D.W. & Gault M.H. (1979). Nephron 16(1): 31-41
3. Janmahasatian S. et al. (2005). Clin Pharmacokinet. 44:1051-65
4. Matthews I. et al. (2004).Br J Clin Pharmacol 58(1):8-19