HPLC with various detection methods, including ultraviolet absorbance, is frequently used to separate and measure creatinine and creatine in serum, plasma, and urine. Current HPLC and other analytical methods for the measurement of creatinine, including capillary electrophoresis and gas chromatography–mass spectrometry, have been reviewed recently (1). In 1990, Paroni et al.(2) reported a cation-pairing HPLC method with ultraviolet detection at 236 nm for the measurement of creatinine in serum and urine. This method involves use of cimetidine as an internal standard, treatment of the sample (100 μL of serum or 30-fold dilution of urine) with acetone (400 μL) to precipitate proteins, complete drying of the supernatant (300 μL) after sample centrifugation, reconstitution with the mobile phase, and injection into the HPLC apparatus. Paroni et al. (2) reported a good correlation between their procedure and other methods, including the Jaffe method. We have modified several steps of the HPLC method originally described by Paroni et al. (2). The modifications outlined in detail below have simplified the previous method and made it more easily adapted to automated analysis.
HPLC analyses were performed with a Pharmacia LKB Model 2248 pump and an analytical column [125 × 3 mm (i.d.)] packed with Nucleosil 120-3 C18 from Macherey-Nagel. The mobile phase consisted of water–acetonitrile (95:5 by volume) containing 10 mmol/L of the sodium salt of 1-octanesulfonic acid (Sigma-Aldrich), which served as the cation-pairing agent. The pH of the mobile phase was adjusted to 3.2 with orthophosphoric acid. The flow rate was 1 mL/min. [In the method of Paroni et al. (2), the more lipophilic cation-pairing agent 1-decanesulfonate (at 10 mmol/L) in water–methanol (50:50 by volume) was used.] The Model Spectroflow 783 variable ultraviolet/visible detector (Kratos Analytical) was set at 236 nm for creatinine or 215 nm for creatine and creatinine. Analyses were performed at ambient temperature (22–26 °C). Creatinine and creatine (98% purity) were purchased from Sigma-Aldrich. Stock solutions (20 mmol/L) of creatinine and creatine were prepared in distilled water and diluted appropriately with the mobile phase.
The major modifications were in the sample treatment procedures. In the present study, for quantitative measurements of urinary creatinine, we diluted 10 μL of centrifuged (800g for 5 min) native urine with 990 μL of the mobile phase (final dilution, 1:100 by volume) and injected a 200-μL aliquot. For quantitative measurements of circulating creatinine and creatine, we mixed 100 μL of plasma (anticoagulated with EDTA, citrate, or lithium heparin) or serum with 100 μL of acetonitrile, incubated the mixture for 3 min, centrifuged it at 2400g for 10 min at 5 °C, diluted 50 μL of the supernatant with 950 μL of the mobile phase (final dilution, 1:40 by volume), and injected a 200-μL aliquot into the HPLC instrument. Precipitation of proteins from plasma or serum by acetonitrile was complete, whereas we found that treatment of urine samples (by this procedure or by acetone precipitation) was unnecessary (data not shown). These modifications simplify the method, considerably shorten the time for sample preparation because no drying of samples is required, and make the use of an internal standard such as cimetidine unnecessary if the HPLC method has adequate precision.
Mixtures containing equimolar concentrations of synthetic creatine and creatinine in the range 0–200 μmol/L were used to prepare calibration curves in urine, plasma, or serum. The mean (SD) retention time of creatinine was 7.26 (0.04) min (CV = 0.6%; n = 12) and did not depend on the amount of creatinine injected. By contrast, the retention time of creatine decreased with increasing amounts injected, and the variation in the retention time was relatively large [mean (SD) retention time, 3.36 (0.16) min; CV = 4.8%; n = 12]. The linear regression equations for peak area (y; in arbitrary units) and concentration (μmol/L) of the injected calibrator (x) were: y = 9312x − 9220 (r = 0.998) for creatine at 215 nm; y = 41 494x − 40 668 (r = 0.999) for creatinine at 215 nm; and y = 10 215x − 14 885 (r = 0.998) for creatinine at 236 nm. Thus, the creatinine assay at 215 nm was approximately four times more sensitive than that at 236 nm. Nevertheless, detection at 236 nm allows for more specific analysis of creatinine and is recommended for use.
To estimate the limit of detection of the method for creatinine, we injected 200-μL aliquots of a 500 nmol/L solution in the mobile phase, which was equivalent to 100 pmol of creatinine, and measured the absorbance at 236 nm; the experiment was performed in quadruplicate. The HPLC peak observed at the retention time of creatinine was measured with a signal-to-noise (S/N) ratio of 21:1 and an imprecision (CV) of 5.6%. A urine sample with a basal creatinine concentration of 24.1 mmol/L was serially diluted with the mobile phase in 1:10 (by volume) steps and analyzed by HPLC. Injection of a 200-μL aliquot of a 1:105 dilution of the sample, which corresponded to 48.2 pmol of creatinine, yielded a creatinine peak with a S/N ratio of 8:1. A serum sample (100 μL) with a basal creatinine concentration of 82 μmol/L was treated with acetonitrile (100 μL) and diluted with the mobile phase to a final dilution of 1:200. Injection of a 200-μL aliquot of this dilution, corresponding to 82 pmol of creatinine, yielded a creatinine peak with a S/N ratio of 6:1. These concentrations were comparable to those obtained from analysis of aqueous solutions of synthetic creatinine. Paroni et al. (2) reported a limit of detection (S/N ratio of 3:1) of 0.5 mg/L for plasma creatinine, which corresponds to an injected amount of 22 pmol. Thus, the detection limit of the present method appears to be similar to that of the method of Paroni et al. (2).
Representative chromatograms from the analysis of synthetic, urinary, and circulating creatinine are shown in Fig. 1A⇓ . A 20 mmol/L solution of creatinine in distilled water and a urine sample were each diluted 1:100 with the mobile phase. A 100-μL aliquot of a serum sample was treated with 100 μL of acetonitrile, and a 50-μL aliquot of the supernatant was diluted with the mobile phase (950 μL), giving a final dilution of 1:40. Each 200-μL aliquot was injected, and the absorbance at 236 nm was measured.
We added creatinine to urine and serum samples at relevant concentration ranges. The final dilutions were 1:100 for urine and 1:40 for serum. Linear regression analysis of peak area (y) and concentration (x) of creatinine added (0–10 mmol/L for urine; 0–200 μmol/L for serum) yielded a slope (SD) of 136 (5) and y-intercept of 1151 (27), with r = 0.996, for a urine sample with a basal creatinine concentration of 9.7 mmol/L, and a slope (SD) of 237 (4) and y-intercept (SD) of 25 298 (343), with r = 0.999, for a serum sample with a basal creatinine concentration of 109 μmol/L.
We validated the HPLC method for the quantitative determination of creatinine in plasma and urine by adding creatinine in triplicate to a fresh plasma and to two urine samples with different basal concentrations (Table 1⇓ ). Linear regression analysis between measured (y) and added (x) creatinine concentration yielded a slope (SD) of 1.23 (0.03) and y-intercept of 60.1 (1.1) μmol/L (r = 0.999) for plasma, a slope (SD) of 1.01 (0.03) and y-intercept of 6.34 (0.10) mmol/L (r = 0.998) for urine sample 1, and a slope (SD) of 0.93 (0.03) and y-intercept (SD) of 0.94 (0.01) mmol/L (r = 0.999) for urine sample 2. Within 1 working day, a urine sample was worked up (1:100 dilution) and analyzed six times by HPLC. In this urine sample, the mean (SD) measured creatinine concentration was 1.63 (0.08) mmol/L with an imprecision (CV) of 5%.
We measured creatinine in 24 urine samples from healthy controls and 23 urine samples from 8 patients with end-stage liver disease by the present HPLC method and by an automated technique based on the frequently used Jaffe method. The patients received a wide variety of drugs, including narcotics, muscle relaxants, diuretics, glucocorticoids, cephalosporins, catecholamines, and antimycotics. One urine sample was obtained before a liver transplantation, and two urine samples were obtained by means of a urethal catheter during transplantation between 40 and 60 min and between 60 and 240 min after reperfusion of the graft. Approval for the study protocol was obtained from the local Institutional Review Board for Studies in Humans. Informed consent was obtained from each patient. Urine samples were aliquoted (1 mL) and frozen at −20 °C until being measured in parallel by the present HPLC method (sample dilution, 1:100 by volume; injection of 200-μL aliquots; flow rate, 1 mL/min; detection at 236 nm) and by the alkaline picrate method (sample dilution, 1:5 by volume), which was carried out on a Hitachi automated analyzer (Hitachi Ltd.) by a certified clinical institution. In this study, drugs were not determined in the urine of the patients.
Linear regression analysis between the creatinine concentrations measured by the present HPLC method (y) and those measured by the alkaline picrate method (x) yielded a mean (SD) slope of 1.4 (0.02) and y-intercept of 0.50 (0.14) mmol/L (r = 0.997) for the control samples and a slope of 1.49 (0.06) and y-intercept of −0.15 (0.59) mmol/L (r = 0.981) for the patient samples. The results were also compared by plotting the difference for creatinine measurements (HPLC − Jaffe) vs the mean concentration (Fig. 1B⇓ ) according to the method of Bland and Altman (3). Analysis by the Jaffe method gave urinary creatinine concentrations that were lower than those measured by HPLC in the whole concentration range (0.2–33 mmol/L). Best agreement between the methods was observed at creatinine concentrations <3 mmol/L. The greatest disagreement between the HPLC and Jaffe methods was for analysis of the urine samples from the patients and for urinary creatinine concentrations >10 mmol/L (Fig. 1B⇓ ). In accordance with the statement of the clinical institution at which the urinary creatinine was measured, all urine samples were diluted with distilled water by the same factor: 1:5 (by volume). We believe that the discrepancies noted at higher creatinine concentrations are attributable to interfering compounds in the Jaffe method, in which the extent of interference increased with increasing concentration of the interfering compounds.
We measured creatinine in serum samples from 17 hospitalized patients in duplicate by the present HPLC method with detection at 236 nm. The serum creatinine concentrations were determined with a mean imprecision (SD) of 4.5 (3.6)% and ranged between 58 and 219 μmol/L. In seven of these samples, creatinine was also measured by HPLC analysis with detection at 215 nm. The ratio (SD) of the respective creatinine concentrations measured at the two wavelengths (236 nm/215 nm) was 0.97 (0.09) with a CV of 9.3%. In these analyses, the ratio of the peak areas for creatinine at the two wavelengths (215 nm/236 nm) was 4.4 (0.4). This ratio agrees with that obtained for aqueous solutions of creatinine calibrators (see above). Thus, these findings suggest no interferences from coeluting substances in these serum samples.
The proposed HPLC method is simple, rapid, and suitable for the accurate measurement of creatinine in urine, plasma, and serum samples from healthy and severely diseased humans. It also offers the opportunity to analyze creatine in plasma or serum. The method’s simplicity and streamlined sample treatment make it suitable for automated analysis of creatinine.
- © 2004 The American Association for Clinical Chemistry