A novel surface-enhanced Raman scattering (SERS) based approach for the quantitative determination of creatinine in human serum is described. Using isotopically labeled (2-13C, 2,3-15N2) creatinine as internal standard, SERS acquires the character of a ratio method that works similar to the well-established isotope dilution techniques. In conjunction with multivariate data analysis, the method was successfully applied for quantifying creatinine at clinically relevant levels and below. A partial least-squares regression model was generated from a set of 87 calibration spectra covering the full range of mole fractions of neat creatinine. The prediction performance of the model was thereafter validated with independent reference samples giving a standard deviation of less than 2%. Finally, a conditioning procedure to prepare real serum samples for SERS-based creatinine analysis was worked out and validated. Measured serum creatinine concentrations are within 3% of the values obtained from gas chromatography/isotope dilution mass spectrometry on the same serum starting material.