Background 51Cr EDTA is a radioisotope used to estimate glomerular filtration rate (GFR), particularly in patients receiving renally cleared anticancer agents such as carboplatin. We propose that current methods for determining 51Cr EDTA clearance from plasma radioactivity (counts/minute) result in biased estimates of GFR. The aims of the study were (i) to develop and test a population pharmacokinetic model for 51Cr EDTA disposition, (ii) to compare model-predicted GFR values to those obtained using the conventional method for determining isotopic GFR (slope-intercept method), (iii) to compare model-predicted GFR to creatinine-based eGFR equations, and (iv) to determine if differences in GFR estimates could change dosing decisions for carboplatin in the clinic.
Methods Data from 40 patients who received 7.4 MBq of 51Cr EDTA by intravenous bolus were available for analysis . Four plasma concentrations were measured at approximately 2, 4, 6 and 24 hours after the dose. A population analysis was conducted using NONMEM. Covariates analysed included total body weight, fat-free mass , sex, age and creatinine clearance (CLCr) calculated using Cockcroft-Gault . The final model was evaluated using a non-parametric bootstrap and a visual predictive check. Model predictions of GFR were compared to the conventional slope-intercept method of measuring isotopic GFR, creatinine-clearance from 24h urine data, and eGFR equations (i.e. Cockcroft-Gault, MDRD 4-variable  and CKD-Epi ) using mean prediction error (MPE) and root mean square error (RMSE). In a hypothetical example, a carboplatin dose for a 40 year old male patient from our dataset was calculated from the Calvert equation  using the model-predicted GFR value and compared to doses predicted with other GFR estimates.
Results A total of 159 51Cr EDTA plasma concentrations collected from 40 patients were analysed. A two-compartment pharmacokinetic model provided the best fit to the data. Significant covariates included CLCr on 51Cr EDTA clearance and total weight body on central volume. Relative to the pharmacokinetic model, 24h urine data provided unbiased GFR predictions. The slope-intercept method for isotopic GFR produced positively biased estimates (MPE 15.5 mL/min/1.73m2 [95% CI 8.9, 22.2]). The eGFR equations (i.e. Cockcroft-Gault, MDRD and CKD-Epi) produced negatively biased estimates relative to the pharmacokinetic model (MPE -19.0 mL/min/1.73m2 [95% CI -25.4, -12.7], -20.1 mL/min/1.73m2 [95% CI -27.2, -13.1] and -16.5 mL/min/1.73m2 [95% CI -22.2, -10.1] respectively). A carboplatin dose of 880mg was predicted for a hypothetical patient using the model-predicted GFR value and 24h urine data in the Calvert equation. The isotopic GFR determined from the slope-intercept method over-predicted the dose by 110 mg (+12.5%) and the eGFR equations (i.e. Cockcroft-Gault, MDRD and CKD-Epi) under-predicted the dose by 140mg (-16%), 230mg (-26%) and 160mg (-18%) respectively.
Conclusion A population pharmacokinetic model for 51Cr EDTA disposition was developed and evaluated. Relative to the model, the slope-intercept method for determining isotopic GFR and eGFR equations produced biased estimates of renal function resulting in clinically significant differences in carboplatin dosing.
1. Putt TL et al (2014) Eur J Clin Pharmacol 70(10):1221-1226
2. Janmahasatian S et al (2005) Clin Pharmacokinet 44(10):1051-1065
3. Cockcroft DW, Gault MH (1976) Nephron 16(1):31-41
4. Levey AS et al (1999) Ann Intern Med 130 (6):461-470
5. Levey AS et al (2009) Ann Intern Med 150(9):604-612
6. Calvert AH et al (1989) J Clin Oncol 7(11):1748-1756