To the editor
The SEN recommends estimating the glomerular filtration rate (GFR) by means of the MDRD-4 equation, using the serum creatinine levels (sCr) approximated to 2 decimals if the units are mg/dL.1
We analyzed the impact on the prevalence of chronic renal disease (CRD) if one decimal is used instead of two decimals, as recommended. We calculated the systematic error and the dispersion (normal and absolute difference between the results of the MDRD-4 with the two approaches) and the inter-method variability by means of the relative difference (absolute difference divided by the mean of the GFR using the approaches multiplied by 100). We also analyzed the impact on the prevalence of hidden renal disease (HRD). We collected 8,967 consecutive blood analytical parameters from patients older than 18 years, requested from Primary Health Care. Through a personal code we identified 8,070 subjects (10.3% of the population of the area), with a mean age of 57.4 ± 18.8 years (range 18-107), of which 62.9% were women; and 40.7% were older than 65 years. If a patient had more than one determination done, we selected the lowest sCr value, to avoid the error attributed to acute renal failure.
By using sCr expressed with 2 decimals, we identified 640 people with GFR < 60 mL/min/1.73 m2 (68.9% women), and the prevalence of CRD was 7.3%; whereas when the sCr value was expressed with one decimal, 699 people were identified (69.8% women) and the prevalence was 8%, which means an increment of 9%.
For the total population the mean bias was -0.3 ± 2.8 mL/min/1.73 m2, with a dispersion of 2 ± 2.1 mL/min/ 1.73 m2 and a GFR variability of 0.6 ± 0.6%. This analysis was completed with a Bland-Altman plot to see the correlation between the dispersion and the MDRD-4 equation using 2 decimals. The dispersion increases as the GFR increases (r = 0.128, p < 0.001), even in patients with CRD (r = 0.427, p < 0.001). The dispersion was not affected either by age or sex (data not shown). For the sCr values the dispersion shows a saw-like hyperbolic curve associated to the approach with one decimal of the sCr, similar to that described for the GFR and sCr levels, so that for sCr values < 1.5 mg/dL the decrease of the dispersion is exponential.
The diagnosis of HRD defined by a GFR < 60 mL/min/1.73 m2 and normal sCr values (for women < 1.2 mg/dL and for men < 1.3 mg/dL) was made in 320 patients when two decimals were employed (50%), and in 253 (39.5%), when only one decimal was employed, with a decrease of 26%. Among women, 288 (65.3%) were diagnosed with HRD with 2 decimals and 251 (56.9%) with one decimal, that is a decrease of 15%; while among men, 32 (16.1%) were diagnosed with 2 decimals and 2 (1%) with one decimal, that is a decrease of 1.500%.
Our results are obtained from a not selected population from Primary Health Care, with a prevalence of CRD that is similar to that found in randomized studies performed on the general population2 and is coherent with previous data from our area.3 The study shows that although there is a close relation between the results of the GFR using one or two decimals, the use of one decimal overestimates the prevalence of CRD by 9% and underestimates the prevalence of HRD by 26%. These differences highlight the importance of following the recommendations when performing studies in this field.
