Estimating renal function in children: a new GFR-model based on serum cystatin C and body cell mass.

This PhD thesis is based on four individual studies including 131 children aged 2-14 years with nephro-urologic disorders. The majority (72%) of children had a normal renal function (GFR > 82 ml/min/1.73 square metres), and only 8% had a renal function < 50% of the normal mean value. The present thesis´ main aims were: 1) to develop a more accurate GFR model based on a novel theory of body cell mass (BCM) and cystatin C (CysC); 2) to investigate the diagnostic performance in comparison to other models as well as serum CysC and creatinine; 3) to validate the new models precision and validity. The model´s diagnostic performance was investigated in study I as the ability to detect changes in renal function (total day-to-day variation), and in study IV as the ability to discriminate between normal and reduced function. The model´s precision and validity were indirectly evaluated in study II and III, and in study I accuracy was estimated by comparison to reference GFR. Several prediction models based on CysC or a combination of CysC and serum creatinine have been developed for predicting GFR in children. Despite these efforts to improve GFR estimates, no alternative to exogenous methods has been found and the Schwartz´s formula based on height, creatinine and an empirically derived constant is still recommended for GFR estimation in children. However, the inclusion of BCM as a possible variable in a CysC-based prediction model has not yet been explored. As CysC is produced at a constant rate from all nucleated cells we hypothesize that including BCM in a new prediction model will increase accuracy of the GFR estimate. Study I aimed at deriving the new GFR-prediction model based on the novel theory of CysC and BCM and comparing the performance to previously published models. The BCM-model took the form GFR (mL/min) = 10.2 × (BCM/CysC)E 0.40 × (height × body surface area/Crea)E 0.65. The model predicted 99% within ± 30% of reference GFR, and 67% within ±10%. This was higher than any other model. The present model also had the highest R E2 and the narrowest 95% limits of agreement. If replacing BCM with weight (Weight-model) the results were almost as convincing. The total day-to-day variation of the GFR-estimate (7.7%) was low. The two new models are, however, still not sufficiently accurate to replace exogenous markers when GFR must be determined with high accuracy. Study II aimed at determining biological variation and analytical precision of serum CysC and creatinine. The precision of CysC (1.7%), and creatinine (2.5%) was very good and the day-to-day variation of CysC and creatinine (within-subject variation between two days) also proved very low (6.4% for both analytes). Because of a relatively low ratio between within-subject variation and between-subject variation neither CysC nor creatinine seems qualified to discriminate between normal and reduced renal function, which was also confirmed in study IV. However, the relatively low total day-to-day variation of 6.6% (CysC) and 6.9% (creatinine) indicate that both are suitable for detecting changes in renal function over time. Study III aimed at determining biological variation and analytical precision of BCM and all other parameters given by measurement by bioimpedance spectroscopy (BIS). Depending on parameter the precision was 0.3-0.8% in children ≥ 6 years and 0.5-2.4% in children < 6 years with a statistically significant difference between the two age-groups (p < 0.001). Within-day variation was 1.1-2.8% and between-day variation 2.4-5.7%. The median value of three repeated measurements is recommended in order to avoid incorrect measurements. Study IV aimed at investigating the diagnostic performance of the BCM-model by: 1) Determining cut-off levels for a three-sided diagnostic procedure with the following outcomes: normal renal function, reduced renal function, indeterminable; 2) Calculating the diagnostic probabilities of reduced renal function for the indeterminable results. The lower the number of children in between cut-off levels, the better the diagnostic performance. The BCM-model resulted in the smallest percentage (39%) of indeterminate children in need for further investigation. In conclusion, with the models developed in the present thesis we are able to provide the clinician with both a reasonably accurate estimate of renal function and a probability of reduced renal function. Furthermore, the positive results from study II and III on precision and biological variation indicate that CysC, creatinine and BCM are very stable variables, which is an indirect validation of the BCM-model´s precision and validity. This is also reflected in the relatively low total day-to-day variation of the GFR-estimate.