Assessing the matrix effects of hemolyzed samples in bioanalysis

LC–MS/MS application in bioanalysis is a very powerful technique due to its inherent high sensitivity and selectivity. Matrix effects, first described in the 1990s [1], for LC–MS/MS analyses have been reported for a number of biological matrices, including plasma, whole-blood and serum [2–8]. Matrix effects result from endogenous interferences (e.g., phospholipids) in extracted samples that co-elute with the analyte of interest and thereby compete with analyte ions for gas-phase droplet formation, resulting in ion suppression or ion enhancement. Matrix effects are common and, provided the overall signal suppression or enhancement is consistent between matrix sources, acceptable [9,10]. Inconsistent matrix effects between matrices, can affect a method’s overall accuracy and precision and thereby its quantitative application. In such cases, modification of the method using additional sample clean-up [5], modification of chromatographic conditions, dilution of samples and reduction of matrix volume are typical examples of solutions employed to eliminate the impact of matrix effects. Assessment of matrix effects during method validation for LC–MS-based assays is routinely carried out [9,10].

Hemolysis is a process that occurs during or after blood collection, in which lysis of red blood cells results in the release into plasma of blood pigments such as hemoglobin and bilirubin and elevated salt concentrations such as potassium, which results in a pink to red discoloration of the plasma [11]. In clinical diagnostic research, hemolysis has been reported to interfere with spectrophotometric methods. The amount of analyte interference will depend on the relative degree of hemolysis or the extent of discoloration of the sample (Figure 1) and has been the reason for sample rejection in clinical diagnostic labs. During the conduct of a clinical study for drug pharmacokinetic (PK) analysis, blood samples are drawn at various sampling time points and processed typically into plasma or serum for subsequent LC–MS/MS analysis. While steps to minimize hemolysis are taken at the time of blood draw (e.g., use of direct venipucture as opposed to use of a plastic indwelling catheter, careful handling during blood sample taking, processing and transport) some hemolyzed samples are unavoidable. Rejection of hemolyzed samples from a PK study is not a practice that is encouraged. It can lead to incomplete plasma concentration–time profiles that, particularly around C_max, could potentially impact the outcome of the study. Hemolysis effect is therefore a special type of matrix effect that includes the impact of the contents of lysed red blood cells on analyte quantitation. Hemolysis effect, if found, can have a dramatic impact on analyte quantitation and require modification of the extraction and/or chromatographic techniques to ensure the application of rugged accurate and precise bioanalytical methods to support clinical trials.

Failures encountered

In our laboratory, during the course of method development/validation, hemolysis effect is assessed in plasma that is purposely spiked with hemolyzed whole-blood (2% v/v) to generate a visibly severely hemolyzed sample. The quantitation analysis of an analyte in plasma and hemolyzed plasma at low and high analyte concentrations is then determined. If there is less than a 15% difference in the amount found in plasma compared with hemolyzed plasma, the method is shown to have no hemolysis effect.

Here, we report results for four different assays (Table 1) in which hemolyzed samples were initially found to have an impact on the quantitation of the analytes. In some cases, the effect was so significant that even at high analyte concentrations, close to the upper limit of quantitation, ion suppression in hemolyzed samples yielded undetectable analyte levels. In each case, a modification to the method was required to eliminate/minimize the impact and allow subsequent method validation.

The solution

▪ Analysis of atorvastatin & metabolites in human plasma

Atorvastatin, its metabolites O-hydroxy atorvastatin, P-hydroxy atorvastatin and the internal standard, atorvastatin-d5, are extracted from human plasma (0.20 ml) by liquid–liquid extraction (LLE). An aliquot of the extract is injected into an LC–MS/MS system. The analytes are separated by HPLC using reversed-phase chromatographic conditions, detected by a tandem mass spectrometer.

Analysis of hemolyzed plasma samples revealed a visible interference with a retention time close to that of the internal standard, deuterated atorvastatin (Figure 2). This interference consequently had an impact on the quantitation of atorvastatin in low concentration samples (Table 2). The solution to this problem was a slight adjustment to the organic and aqueous portion of the mobile phase, in order to improve the chromatographic separation of the interference from the internal standard (Figure 3). Under these improved conditions, there was no hemolysis impact on the assay (Table 2).

▪ Analysis of phenylephrine in human plasma

Total phenylephrine (free and conjugated) was determined in human plasma (0.2 ml) by hydrolysis of conjugates using β-glucuronidase, followed by extraction. Phenylephrine and the internal standard were extracted by SPE. An aliquot of the extract was injected into an LC–MS/MS system. The analytes were separated by HPLC using hydrophilic interaction LC conditions and detected using a tandem mass spectrometer.

The bioanalytical method for total phenylephrine in human plasma was originally validated using a close structural analog, differing from phenylephrine by a single hydroxyl group (Figure 4). Phenylephrine and its analog internal standard demonstrated close chemical properties, with more than 85% extraction recoveries and relative retention time (phenylephrine to internal standard) of 0.9. During the assessment of the impact of hemolysis effect, greater than 15% difference was observed in hemolyzed samples at both high and low concentrations compared with plasma samples. This impact can be eliminated simply by dilution of hemolyzed samples 1:1 with plasma to reduce the degree of hemolysis to 1% (Table 3). Application of this method in sample analysis is therefore only valid in nonhemolyzed or marginally hemolyzed plasma. Severely hemolyzed samples are reportable only after dilution to a lesser degree of hemolysis. While this approach affords a simple solution to the problem, it has the disadvantage that severely hemolyzed samples that fall below the limit of quantitation, due to the dilution, will give nonreportable values. Further improvements to the method were made, including a substitution of the internal standard from an analog to a deuterated internal standard, where extraction recoveries were greater than 88% and the relative retention time (phenylephrine to internal standard) was 1.0. With this method modification, the impact from hemolysis effect on phenylephrine was matched by the deuterated internal standard and therefore hemolysis effect, even in severely hemolyzed samples, had no impact on analyte quantitation (Table 3). These findings clearly demonstrate the potential disadvantages of even very close structural analogs as candidates for assay internal standards. Here, a more rugged and reliable method was developed that utilized a deuterated internal standard to compensate for the hemolysis effect.

▪ Analysis of carvedilol & metabolites in human plasma

Carvedilol, its metabolites O-desmethylcarvedilol, 4´-hydroxyphenylcarvedilol and 5´-hydroxyphenylcarvedilol and the internal standard desacetyldiltiazem were extracted from human plasma (0.30 ml) by SPE. The analytes were separated by HPLC using reverse-phase chromatographic conditions and detected using a tandem mass spectrometer.

Using this method, hemolysis effect for the assay for carvedilol and its metabolites in human plasma was assessed. The hemolysis effect was found to be dependant on analyte concentration and had no impact on the parent drug, carvedilol, which was present at concentrations 20-times higher than those of the metabolites. Accurate quantitation of the metabolites was not possible in hemolyzed samples, as the metabolite responses were either severely suppressed or totally undetectable (Table 4, Figure 5). The SPE method was accordingly reviewed and no further refinements to the wash/elution conditions could be applied to resolve the observed hemolysis effect. It was also clearly evident that the final extracts of hemolyzed samples were still discolored after extraction, indicative of the co-extraction of strongly colored blood pigments such as hemoglobin or bilirubin. The extraction method was therefore modified to include a protein precipitation step prior to SPE analysis to remove hemoglobin, and an additional acid wash step was added to the SPE method to facilitate the elimination of any residual ‘heme ions’. With these modifications, the method was subsequently revalidated to verify that there was no hemolysis effect on the accurate quantitation of each of the analytes and the method was suitable for PK application (Table 5).

▪ Analysis of olazapine in human plasma

The olanzapine method was initially developed in human plasma using Waters HLB SPE cartridges, using carefully selected wash and elution solvent conditions (maximum organic wash and minimum organic elution). Despite the use of a deuterated internal standard during the assessment of the hemolysis effect, a huge amount of ion suppression was evident in hemolyzed samples compared with plasma controls, such that olanzapine, and its internal standard olanzapine-d3, were virtually undetectable (Figure 6). An attempt to change the selectivity of the assay and adopt a more systematic extraction approach, by changing the bonded phase of the SPE sorbent to a mixed-mode cation exchange, was tried subsequently, again with the same phenomenon [5]. The combination method of PPT followed by mixed-mode cation exchange SPE, as previously described for carvedilol, was evaluated next. Under these conditions, both drug and internal standard were detectable, but analyte and internal standard responses were still 50% lower in hemolyzed samples compared with plasma controls. The observed suppression was matched in drug and internal standard and the concentrations found in plasma versus hemolyzed plasma were within 5% (Table 6). However, as the signal-to-noise in low concentration samples were considerably lower in hemolyzed samples, accuracy and precision of the method at the low end of the calibration curve was not considered comparable in plasma and hemolyzed samples. In addition, as hemolyzed samples had 50% lower internal standard responses they were consistently flagged as internal standard outliers, as were hemolyzed unknowns during study sample analysis. Accordingly, the extraction methodology was modified, to LLE. Using this approach, the impact from hemolyzed samples was completely eliminated and analyte responses for both drug and internal standard were consistent in plasma and hemolyzed plasma (Figure 7, Table 7).

The final validated assay involved extraction of olanzapine and the internal standard, olanzapine-d3, from 0.50 ml of human plasma, using potassium K₂EDTA as the anticoagulant, by LLE into an organic medium, evaporated under nitrogen, reconstituted in 0.30 ml of mobile phase. An aliquot of this extract was injected into a HPLC system and detected using an EP10+ tandem mass spectrometer.

Practical tips

While assessment of hemolysis effect impact is being reported for LC–MS/MS-based assays [8], there is currently no standardized approach. From the work presented here, our recommendation is a simple approach using a spiked hemolyzed sample to ‘simulate’ a severely hemolyzed sample expected during the conduct of a clinical study. The following procedure is suggested as a practical approach:

• ▪ The effect of drug analysis in plasma in the presence of ruptured erythrocytes (red blood cells) is assessed;

• ▪ Quality control (QC) samples at high and low analyte concentrations, with (hemolyzed) and without (nonhemolyzed) the presence of 2% hemolyzed blood, are analyzed, in replicates of six and compared along with a blank hemolyzed control sample for selectivity assessment (Figure 8);

• ▪ These samples are analyzed along with a duplicate standard curve and QC samples, including blank and blank with internal standard and analyte concentrations measured.

The acceptance criteria are:

• ▪ The concentration of analyte(s) in 2% hemolyzed samples must be within ±15% of the nonhemolyzed reference of the same concentration level for both QC high and QC low;

• ▪ A failed (2% hemolyzed samples) experiment may be repeated using a lesser degree of hemolysis sample (e.g., 1%). Only study samples with the same degree of hemolysis would yield reportable values;

• ▪ The blank control sample should be free of significant interference at the retention time of the analytes.

Future perspective

Understanding the need to assess the impact from matrix effects has been widely studied in routine bioanalysis using LC–MS/MS-based assays. Quantitative assessments of analytes spiked into a number of different individual matrix sources is a standard method validation parameter to ensure method quantitation consistency between matrix sources, as recommended in the 2001 guidance from the US FDA [9] and further emphasized after Crystal City III in 2006 [10]. Hemolysis effect is a special type of matrix effect. Hemolyzed samples, even as evidenced by their visible appearance, can differ dramatically from sample to sample depending on the degree of hemolysis, making them very heterogeneous. We have clearly demonstrated that the impact from hemolysis effect on a bioanalytical assay can range from a marginal impact to an effect that is significant, and the corresponding ion suppression is so marked that the analyte is not detectable at any given concentration. We have shown the approaches taken to resolve any impact from hemolysis effect can range from simple to complex methodologies to ensure accurate analyte quantitation in severely hemolyzed samples at any given concentration over the calibration range of an assay. In bioanalysis, this is essential to ensure universal application of the method to all samples to support preclinical and clinical PK trials.