A simplified, rapid LC-MS/MS assay for serum and salivary creatinine

Highlights

• •A rapid, precise LC-MS/MS assay for creatinine has been developed.
• •The method is applicable to serum and saliva and uses a small sample volume.
• •Glucose and bilirubin interfere in the colorimetric Jaffe creatinine assay.
• •No interference was observed in the LC-MS/MS method.
• •This assay could be useful for routine clinical use.

Abstract

In routine clinical laboratories, serum creatinine is typically measured on automated analyzers using colorimetric or enzymatic assays, which are both susceptible to interferences that can lead to incorrect measurement. Here, we present a straightforward and rapid LC-MS/MS assay for serum creatinine using methanol extraction, with separation performed using a strong cation exchange column. Results from this newly developed method were compared against those from an automated Abbott Architect kinetic Jaffe method. We also assessed the effect of bilirubin and glucose, as interferants, on both methods.

Our LC-MS/MS assay has a run time of 1.1 min, uses a relatively small sample volume of 10 µL and has a within-batch imprecision of 1.1–1.8% at the concentrations tested, which is within the range necessary for routine clinical use. Recovery from serum samples spiked with creatinine was >95%, and glucose and bilirubin were not found to interfere in the assay. Glucose was observed to significantly interfere in the kinetic Jaffe method, while bilirubin had a variable effect. We also determined that creatinine can be measured from saliva using our method, and that salivary concentrations are, on average, 15% of those in serum. This assay provides an alternative for patient sample analysis where interference is expected in routine creatinine methods.

1. Introduction

Serum creatinine is produced as a metabolic by-product of muscle creatinine, of which 1–2% is converted daily. The amount of creatinine produced is proportional to an individual’s muscle mass and, hence, varies with age and sex [1]. In healthy individuals, creatinine moves through the kidney without modification, consequently it is useful as a marker of kidney health via glomerular filtration rate (GFR). Although the reference ranges for serum creatinine are wide due to inter-individual variation, formulae to calculate an estimated GFR (eGFR) from the serum creatinine concentration, by taking into account age, sex and other factors, are now common. This provides a more accurate assessment of kidney function versus considering serum creatinine concentration alone [2].

In routine clinical laboratories, due to high workload and requirements for rapid turn-around-time, serum creatinine is usually measured on automated analysers using colorimetric or enzymatic assays. In the UK, the majority of laboratories utilise the alkaline picrate (Jaffe) method, where creatinine reacts with picrate under alkaline conditions to form a coloured complex, which is then detected spectrophotometrically. However, this method is susceptible to result variation due to interference from compounds such as bilirubin, ketones, glucose and protein [3]. Although modifications, such as kinetic Jaffe assays and compensation factors have improved performance, interference continues to affect result accuracy. Enzymatic methods rely on the use of enzymes, such as creatinine iminohydrolase or creatininase, with the eventual production of a chromophore or ammonia. These methods are also susceptible to result variation from interferences, such as haemoglobin and bilirubin, however, they are considered to have improved specificity compared to the Jaffe assay [4].

We have adapted an LC-MS/MS assay, previously developed in our laboratory [5], to produce a simplified extraction protocol, shorter run time and reduced number of mobile phases. This assay is rapid, cost-effective and precise; it could be used in situations where patient samples are likely to contain interferants or where improved accuracy is required, for example in clinical trials. This method is also applicable to saliva samples.

2. Methods

2.1. Sample collection and preparation

A stock solution, prepared by the addition of anhydrous creatinine (Sigma-Aldrich, Poole, UK; stated purity ≥ 98%) to phosphate buffered saline (PBS; Sigma-Aldrich, Poole, UK), was used to produce standards with creatinine concentrations between 31.25 µmol/L and 2000 µmol/L. Quality control (QC) samples were prepared in a similar manner, but from a separate stock solution so that standards and QC samples were independent of each other. PBS was chosen as the standard matrix since creatinine-free serum is not available.

Anonymized patient sera were obtained from surplus laboratory samples in our hospital where creatinine had already been measured using the routine method. Pooled surplus serum was used for determining the effect of glucose concentration on measured creatinine; anhydrous glucose (Sigma-Aldrich, Poole, UK) was dissolved in water to form a stock solution, which was then used to spike serum samples with up to 54 mmol/L of glucose (a consistent volume of water was added to all samples). Similarly, surplus serum samples were used to investigate the effect of high concentrations of bilirubin. Bilirubin powder (Sigma-Aldrich, Poole, UK) was dissolved in dimethyl sulfoxide (DMSO; Sigma-Aldrich, Poole, UK) at a concentration of 30 mM, and further diluted in DMSO to produce stock solutions for spiking serum. A range of serum bilirubin concentrations was achieved, using a consistent volume of bilirubin solution for each. External Quality Assessment (EQA) samples were obtained from RIQAS Global EQA/Proficiency Testing Scheme.

Paired fasting serum and saliva samples from 42 patients with end stage renal disease (45% of whom were on dialysis) who were taking part in the BASIC-HHD study [6] were analysed for creatinine using the LC-MS/MS method. This study was reviewed and approved by the Greater Manchester West Health Research Authority National Research Ethics Service, reference number: 12/NW/0170. Patients were subdivided into three groups: A (not on dialysis, n = 23), B (in-centre haemodialysis, n = 12), C (home haemodialysis, n = 7). Saliva was collected by passive drool technique, and patients were given detailed instructions prior to sample collection. Patients were requested to:

• •Avoid dental work for 48 h
• •Abstain from alcohol for 24 h
• •Avoid teeth brushing for 2 h
• •Avoid eating for at least 1 h
• •Rinse their mouth with water not more than 10 min and not more than 15 min prior to sample collection.

To prepare the samples, 500 µL methanol (Honeywell LC-MS quality; VWR, Lutterworth, UK) was added to 10 µL of sample (serum or saliva) or standard/QC along with 10 µL of internal standard (10 mg/L creatinine-d3 in methanol) in a 96-deep-well plate (Porvair Sciences, Wrexham, UK). The plate was heat sealed and the samples were vortexed and centrifuged at 880 × g for 5 min before transfer to the autosampler for analysis.

2.2. Chromatography and mass spectrometry

Using a Waters Acquity I Class UPLC, 2 µL of sample supernatant was injected onto a 4 mm × 3 mm strong cation exchange SecurityGuard column (Phenomonex, Macclesfield, UK) and eluted with 100% mobile phase A as initial conditions (2 mmol/L ammonium acetate, 0.1% formic acid in water) stepping up to 100% mobile phase B (50 mmol/L ammonium acetate, 0.1% formic acid and 10% methanol in water) at 0.4 min, holding for 0.2 min before returning to initial conditions for 0.5 min. Flow rate was 0.6 mL/min and the total run time was 1.1 min. Creatinine was found to elute at 0.75 min under these conditions (Fig. 1).

Fig. 1.

Example chromatograms showing creatinine and creatinine-d3 eluting from the column at approximately 0.75 min post-injection. Ion intensity (%) is on the y-axis.

A Waters Xevo TQD tandem mass-spectrometer (Waters, Manchester, UK) was used with MassLynx 4.1 software and MassLynx TargetLynx™ software for data analysis. Electrospray ionisation in positive mode was used. The capillary was maintained at 0.3 kV, the cone voltage was 26 V, source temperature 150 °C, desolvation temperature 400 °C, with a collision energy of 14 eV. The specific transitions monitored in multiple reaction monitoring mode were m/z 114 > 43.9 (creatinine) and 117 > 46.9 (creatinine-d3). Calibration curves were produced by TargetLynx™ software, using the ratio of analyte peak area to internal standard peak area, to allow calculation of unknown values.

2.3. Method validation

For assessment of within-batch imprecision, QC material at 3 different concentrations and patient serum were used. The coefficient of variation (CV) was then calculated from 10 replicates of each sample. Between-batch imprecision was assessed over a period of several months using QC material only, as creatinine in serum is not stable for this length of time [7]. To determine linearity, 8 standards (consisting of creatinine spiked into PBS) up to 2000 µmol/L were analysed and the response plotted against concentration. The lower limit of quantification (LLOQ) was assessed by determining the concentration at which CV and bias were both<20%. Concentrations of 5 and 10 µmol/L creatinine in PBS were analysed 10 times each in the same batch and the CV and bias calculated.

In order to assess whether ion suppression/enhancement was present, a post-column constant infusion of internal standard at a concentration of 0.25 mg/L was set up. A number of samples were extracted as described in 2.1 (minus internal standard) and injected as in 2.2. The effect of injected samples on the internal standard signal was compared to blank methanol.

Recovery was assessed through analysis of samples spiked with 500 µmol/L of creatinine. Recovery in PBS was compared with that of two different pooled serum samples. In the case of the pooled serum, endogenous creatinine was measured in the unspiked samples and this was accounted for when calculating the % recovery of spiked creatinine.

2.4. Creatinine, glucose and bilirubin analysis using Abbott Architect analysers

An Abbott Architect c16000 analyser with Abbott creatinine reagents (kinetic Jaffe method; Ref 3L81) was used to provide a comparison method for patient serum samples. Abbott glucose reagents (hexokinase enzymatic assay; Ref 3L82-21) were used with the same analyser to measure serum glucose in patient serum and spiked samples. The Abbott bilirubin method used was colorimetric (diazo reagent; Ref 6L45) and the icteric index was also determined on the analyser by spectrophotometry.

2.5. Statistical analysis

Analyse-it software for Microsoft Excel (version 2.3) was used for Bland-Altman plots and Deming regression. The Mann-Whitney U test was used to compare salivary:serum creatinine ratios between patient groups.

3. Results

3.1. Validation

The within batch imprecision of the assay was found to be < 1.9% (Table 1) with QC samples and a patient serum sample covering a range of concentrations. The between batch imprecision over several months was excellent (<3.3% using QC material). This meets the minimum analytical goal for imprecision of < 4.46%, based on biological variation data [8].

Table 1.

Imprecision data for LC-MS/MS creatinine assay.

Creatinine concentration	Within-batch CV, n = 10 (%)	Between-batch CV (%), n = 10
50 µmol/L (QC material)	1.4	3.2
67 µmol/L (patient serum)	1.5	–
300 µmol/L (QC material)	1.8	1.7
600 µmol/L (QC material)	1.1	1.7

The assay was linear to at least 2000 µmol/L with an r² value of 0.999728, which is suitable for the range of results seen in a clinical setting. An example standard curve is shown in Fig. 2. The LLOQ was found to be 10 µmol/L (CV = 4.2% at this level, bias 2.9%, with 10 replicates).

Fig. 2.

Typical standard curve for creatinine, showing response (peak area ratios for creatinine/creatinine-d3) plotted against concentration of creatinine.

Ion suppression was found to be minimal when comparing injection of extracted serum samples and methanol, using a post-column infusion of creatinine-d3 with no appreciable difference seen (data not shown). In addition, recovery was compared between PBS samples and pooled serum spiked with 500 µmol/L creatinine (Table 2); recovery was >95% for both sample types, indicating that ion suppression is not significant for extracted serum samples.

Table 2.

Recovery in PBS and pooled serum, 500 µmol/L creatinine spike.

Sample type	% Recovery
PBS	98.4
PBS	98.5
Pooled serum	95.0
Pooled serum	98.0

3.2. Patient sample comparison

A comparison of 20 patient serum samples using a kinetic alkaline picrate (Jaffe) assay on an Abbott Architect c16000 analyser revealed good correlation with the LC-MS/MS assay (Fig. 3), which would be expected for samples without significant levels of interferants. Deming regression showed small constant and proportional biases for the LC-MS/MS assay, compared to the Jaffe method (Fig. 3A), which were not statistically significant (constant bias −6.79 µmol/L, 95% confidence limits −20.89 to 7.3, proportional bias 11%, 95% confidence limits −7% to 29%). A Bland-Altman plot is shown in Fig. 3B and is suggestive of a small proportional bias; the average bias was + 3.4 µmol/L. Testing of four EQA samples showed an average bias of + 6.88% compared to the all-method mean for the scheme (Table 3).

Fig. 3.

Method comparison between a kinetic alkaline picrate assay (Abbott Architect c16000 analyser) and the LC-MS/MS assay using 20 anonymised surplus serum samples. (A) Deming regression analysis of LC-MS/MS creatinine plotted against the Abbott Jaffe creatinine result. (B) Bland Altman (difference) plot illustrating the agreement between the two methods. Average bias and 95% limits of agreement are shown.

Table 3.

EQA sample results.

EQA sample	All-method mean µmol/L	Abbott result µmol/L	LC-MS/MS result µmol/L	% Bias for LC-MS/MS method
1	136.025	138.3	143.59	5.4
2	371.606	380.1	401.42	7.7
3	122.5	120.2	130.29	6.2
4	125.5	127.7	136.3	8.3

3.3. Degree of interference from glucose and bilirubin

Glucose is known to interfere in Jaffe creatinine methods. To determine the degree of interference observed in the Abbott Architect method and the LC-MS/MS method, a pooled serum sample was spiked with glucose solution to produce five concentrations, including an unspiked sample which had 6.5 mmol/L endogenous glucose present. The spiked samples were then analysed using both creatinine methods, and glucose was also measured to confirm the concentration present. Significant interference was observed in the Abbott Jaffe method, with higher measured creatinine observed at glucose concentrations of 16.9 mmol/L and greater (Fig. 4). At a glucose concentration of 54.3 mmol/L, the creatinine measured by the Jaffe method was 54% higher compared to the unspiked sample. In contrast, creatinine measured by LC-MS/MS was unaffected by high concentrations of glucose, with a maximum increase in measured creatinine of 3.1% at the highest glucose concentration (Fig. 4A).

Fig. 4.

Effect of glucose on measured creatinine. (A) Pooled serum was spiked with glucose to provide a range of concentrations. Glucose was measured using a hexokinase enzymatic assay and creatinine was analysed using both the Abbott Architect Jaffe method and by LC-MS/MS. The percentage change in measured creatinine compared to the unspiked sample was calculated for both creatinine methods. (B) Fifteen patient samples with high concentrations of endogenous glucose were analysed for creatinine by both assays and the % difference between the Abbott Jaffe method and the LC-MS/MS assay is shown for each sample.

In addition to spiked samples, fifteen patient samples with high concentrations of endogenous glucose were analysed using both creatinine methods. In the presence of glucose concentrations ≥ 16.3 mmol/L, higher creatinine results were obtained with the Jaffe method compared to the LC-MS/MS method, except for one sample with a high creatinine concentration of 350 µmol/L (Fig. 4B). Given the minimal bias demonstrated with the initial method comparison study, and taken together with the data from serum samples spiked with glucose, we can infer that this is due to interference from glucose. Although the effect was much more variable than with the controlled spiking study, significant interference was again seen with the Abbott assay, particularly at higher glucose concentrations.

In order to test the effect of high concentrations of bilirubin on measured creatinine, five serum samples with differing creatinine concentrations were chosen, and various concentrations of bilirubin in DMSO were added and compared to DMSO alone. The solubility of bilirubin was a limiting factor and this prevented high concentrations from being reached in some samples. Samples were analysed for creatinine (Jaffe method), bilirubin and icteric index on an Abbott Architect analyser and creatinine was also measured using the LC-MS/MS method. There was a poor correlation between icteric index and measured bilirubin (data not shown), which may be due to solubility issues. The effect of high concentrations of bilirubin on measured creatinine was variable for the Abbott assay, with 3 of the 5 samples (A, C and E) showing a decrease in creatinine as the icteric index increased above 600 (Fig. 5A). The largest decreases in measured creatinine observed were 33%, 25% and 6% for samples A, C and E, respectively. The effect was less obvious when creatinine was plotted against bilirubin concentration (Fig. 5B). For the LC-MS/MS method, no appreciable change in the measured creatinine was noted with high bilirubin concentrations or high icteric indices (Fig. 5C and D).

Fig. 5.

Effect of bilirubin on measured creatinine. Five serum samples were spiked with bilirubin in DMSO or with DMSO alone, and the bilirubin concentration and icteric index were measured on an Abbott Architect analyser. Abbott Jaffe creatinine and LC-MS/MS creatinine values are plotted against icteric index (A, C) or measured bilirubin (B, D).

3.4. Analysis of salivary creatinine

Creatinine can also be measured in saliva [9] and we have tested 42 paired fasting serum and saliva samples from patients with renal impairment using this assay, 19 of which were from patients on dialysis (groups B and C). On average, salivary creatinine concentrations were 15.4% of serum levels although large variations between sample pairs were seen, ranging from 0.04% to 77.7% (Table 4 and Fig. 6). This variation did not appear to be concentration dependent, and differences in salivary:serum creatinine ratios between the patient groups were not statistically significant.

Table 4.

Salivary creatinine concentrations compared to serum. Group A: patients not on dialysis, Group B: in-centre haemodialysis, Group C: home haemodialysis.

Patient Group (n)	Mean serum creatinine	Mean salivary creatinine	Mean salivary % of serum creatinine	Range of % results
A (23)	429.157	59.303	13.3	1.9–39.5
B (12)	849.679	141.328	16.0	5.9–40.8
C (7)	703.966	171.446	21.4	0.4–77.7

Fig. 6.

Measurement of creatinine in saliva. Paired, fasting serum and saliva samples from 42 patients with renal impairment were analysed for creatinine by LC-MS/MS. The ratio of salivary creatinine to serum creatinine is expressed as a percentage.

4. Discussion

This assay represents a simple and extremely rapid means of determining creatinine in serum and saliva using a small sample volume. The imprecision (both within- and between-batch) was low and the linear range and limit of quantification were acceptable for routine clinical use. We observed a positive bias with EQA samples compared to the all method mean; this may reflect a lack of negative interference (for example from bilirubin) in our method. The method comparison data suggested small constant and proportional biases compared to a kinetic Jaffe method, however, these were not statistically significant.

There are a number of other published LC-MS methods for serum creatinine. Stokes et al. [10] developed a candidate reference method and compared the results to the established gas chromatography-isotope dilution-MS assay, however, the run time was significantly longer than the method described here. Similarly, Dodder et al. [11] reported a potential LC-MS reference method for serum creatinine. Liu et al. [12] described an LC-MS/MS method suitable for routine use, however, performance characteristics were not detailed. Harlan et al. [13] utilised turbulent flow LC-MS/MS, with online solid phase extraction followed by separation on a porous graphitic carbon column and MS/MS detection. This methodology requires specific instrumentation and is likely to be more expensive than our assay. Most recently, Ou et al. [14] described an LC-MS/MS method for serum creatinine with a run time of 3 min and with imprecision < 3.84%, using 50 µL of sample and a similar extraction protocol.

The multiple interferences in the Jaffe method have been well-established for many years [12], [15], [16], [17], [18], [19], [20], [21]. We have shown here that glucose is a significant interferent in the Abbott kinetic Jaffe assay, while bilirubin had a variable effect depending on the sample and the creatinine concentration. As the LC-MS/MS assay does not suffer from these interferences, this method may offer benefits for patients where interferences are likely, such as for patients with high glucose levels or jaundice.

LC-MS/MS equipment is costly and is not available in all laboratories, and LC-MS/MS assays generally require a greater level of staff expertise and staff time when compared to automated clinical chemistry analysers. However, the simple and rapid sample preparation for this assay means that extensive training would not be required. In our laboratory we have a wide range of clinical LC-MS/MS assays in routine use, and a number of trained staff. It would, therefore, be feasible to offer this method as and when required for specific patients, however, this would not be the case in laboratories without this type of set up. In terms of reagent and consumable costs it would be cheaper per sample than an enzymatic creatinine assay and comparable to the Jaffe method, although it is not envisaged that this LC-MS/MS method would replace routine automated analysis as the increased turn around time inherent in batched analysis would not be acceptable for the clinical requirements for this test. With increasing automation of sample preparation and LC-MS/MS instruments, in the future, this position may change.

Creatinine is thought to enter saliva via ultrafiltration and variations in salivary flow rate may influence the concentration. The salivary creatinine results presented here generally agree with published studies [9], [22] showing salivary creatinine levels to be 10–15% of serum concentrations, although we observed a wide variation in the relative concentrations. Contamination from food can increase salivary creatinine levels, however, the samples tested in this study were collected from apparently fasting subjects. These data are preliminary and further validation of saliva as a matrix for this assay would be required prior to clinical use. An LC-MS/MS assay for the simultaneous measurement of salivary uric acid and creatinine has recently been published, however this is the only other study to date that has demonstrated measurement of salivary creatinine using this technique [23].

It has been shown elsewhere that patients with renal impairment have higher salivary creatinine levels [9], [24]. In the latter study it was also found that whilst there was a negative correlation between serum and saliva creatinine levels in healthy subjects, there was a significant positive correlation in patients with renal impairment [24]. The authors suggested that this could be due to changes in the permeability of salivary gland cells in patients with chronic kidney disease (CKD) or to the effect of an increased concentration gradient for creatinine between plasma and saliva in those with reduced renal function. It is possible that salivary creatinine may be useful in a clinical setting for CKD patients, particularly as sample collection is non-invasive and the volume required is small, although further studies would be necessary with larger patient groups prior to introduction into routine use.

Declaration of conflicting interests

None.

Funding

The BASIC-HHD study was funded by Baxter Clinical Evidence Council and NIHR-CLAHRC for Greater Manchester, UK.