The equations used to estimate glomerular filtration rate (GFR) based on serum creatinine level are limited by their dependence on muscle mass. Although cystatin C level predicts clinical outcomes better than creatinine level in the general population, its role in estimating GFR in the reference range is unclear. Cystatin C level is not influenced by muscle mass, but by several other non-GFR determinants. We investigated whether regression models using cystatin C level alone or in combination with creatinine level in principle would improve GFR estimation in the general population compared with models using creatinine level alone. Study of diagnostic accuracy. A representative sample (n = 1,621; aged 50-62 years) of the general population in Tromsø, Norway, without coronary heart disease, stroke, diabetes mellitus, or kidney disease. Individuals had participated in the Renal Iohexol Clearance Survey (RENIS-T6), part of the sixth Tromsø Study. Performance of multiple linear and fractional polynomial regression models with plasma creatinine and/or cystatin C levels as independent variables and measured GFR as a dependent variable. Plasma iohexol clearance. Creatinine measured with an enzymatic method. Cystatin C measured with a particle-enhanced turbidimetric immunoassay. In internal validation of models with cystatin C, creatinine, or both levels, percentages of GFR estimates within 10% of measured GFR were 61% (95% CI, 58%-63%), 62% (95% CI, 59%-64%), and 68% (95% CI, 65%-70%), respectively. Models with either cystatin C or creatinine level had very similar precision and ability to detect GFR <90 mL/min/1.73 m(2), whereas models based on both markers performed better. Only middle-aged individuals of European ancestry were investigated. Lack of standardization between cystatin C assays. No external validation of regression models. Models based on cystatin C alone are not superior to those based on creatinine, but models based on both markers can improve GFR estimation in the reference range.
Read full abstract