Comprehensive characterization and resolution of discrepant spectrophotometric bilirubin results in patients on eltrombopag therapy

Timothy H.T. Cheng 1 , Teresa K.C. Tsui 1 , Jeffrey S.S. Kwok 1 , Lydia C.W. Lit 1 , Elaine Y.L. Wong 1 , Richard K.T. Kam 1 , Denis Grote-Koska 2 , Antje Staaden 2 , Hitoshi Okada 3 , Noriko Fuke 3 , Raymond S.M. Wong 4 , Chi-Kong Li 5 , and Michael H.M. Chan 1
  • 1 Department of Chemical Pathology, Prince of Wales Hospital, The Chinese University of Hong Kong, Shatin, Hong Kong
  • 2 Hannover Medical School, Institute of Clinical Chemistry, Hannover, Germany
  • 3 Department of Pediatrics, Kagawa University, Kagawa, Japan
  • 4 Sir YK Pao Centre for Cancer and Department of Medicine and Therapeutics, The Chinese University of Hong Kong, Shatin, Hong Kong
  • 5 Department of Paediatrics, The Chinese University of Hong Kong, Shatin, Hong Kong
Timothy H.T. Cheng
  • Corresponding author
  • Department of Chemical Pathology, Prince of Wales Hospital, The Chinese University of Hong Kong, Shatin, Hong Kong
  • Email
  • Search for other articles:
  • degruyter.comGoogle Scholar
, Teresa K.C. Tsui
  • Department of Chemical Pathology, Prince of Wales Hospital, The Chinese University of Hong Kong, Shatin, Hong Kong
  • Search for other articles:
  • degruyter.comGoogle Scholar
, Jeffrey S.S. Kwok
  • Department of Chemical Pathology, Prince of Wales Hospital, The Chinese University of Hong Kong, Shatin, Hong Kong
  • Search for other articles:
  • degruyter.comGoogle Scholar
, Lydia C.W. Lit
  • Department of Chemical Pathology, Prince of Wales Hospital, The Chinese University of Hong Kong, Shatin, Hong Kong
  • Search for other articles:
  • degruyter.comGoogle Scholar
, Elaine Y.L. Wong
  • Department of Chemical Pathology, Prince of Wales Hospital, The Chinese University of Hong Kong, Shatin, Hong Kong
  • Search for other articles:
  • degruyter.comGoogle Scholar
, Richard K.T. Kam
  • Department of Chemical Pathology, Prince of Wales Hospital, The Chinese University of Hong Kong, Shatin, Hong Kong
  • Search for other articles:
  • degruyter.comGoogle Scholar
, Denis Grote-Koska, Antje Staaden, Hitoshi Okada, Noriko Fuke, Raymond S.M. Wong
  • Sir YK Pao Centre for Cancer and Department of Medicine and Therapeutics, The Chinese University of Hong Kong, Shatin, Hong Kong
  • Search for other articles:
  • degruyter.comGoogle Scholar
, Chi-Kong Li and Michael H.M. Chan
  • Department of Chemical Pathology, Prince of Wales Hospital, The Chinese University of Hong Kong, Shatin, Hong Kong
  • Search for other articles:
  • degruyter.comGoogle Scholar

Abstract

Background

Eltrombopag is a thrombopoietin receptor agonist used for the treatment of thrombocytopenic conditions. It can cause pH-dependent discoloration of plasma/serum. Eltrombopag is potentially hepatotoxic. It can affect the assessment of hyperbilirubinemia because of its (i) absorbance at ~450 nm (bilirubin), (ii) absorbance at ~550 nm (diazo-bilirubin) and (iii) it can cause yellowish discoloration of the eyes at normal circulating bilirubin levels.

Methods

We collected 66 samples from patients on a range of eltrombopag dosages up to 150 mg daily. Bilirubin was measured using multiple routine spectrophotometric analyzers, the Doumas reference method and high-performance liquid chromatography (HPLC). Plasma/serum eltrombopag concentrations were determined using liquid chromatography tandem mass spectrometry (LC-MS/MS). Spike-in and admixture experiments delineated the effects of eltrombopag and its metabolites.

Results

Forty-nine of 52 samples from patients on ≥50 mg daily eltrombopag therapy showed significantly discrepant inter-analyzer total bilirubin results, a difference up to 64 μmol/L (3.7 mg/dL). In one sample, total bilirubin varied from 8 to 65 μmol/L (0.4–3.8 mg/dL) by different routine analyzers, with direct bilirubin ≤4 μmol/L (0.2 mg/dL). There was a positive correlation between total bilirubin difference and plasma eltrombopag concentration (r = 0.679), and spike-in experiments demonstrated that Beckman AU and Doumas reference methods were susceptible to positive interference. HPLC can quantify bilirubin after separating eltrombopag, and results suggest different analyzers are affected to varying degrees by eltrombopag and its metabolites.

Conclusions

Eltrombopag and its metabolites can cause positive interference to the spectrophotometric measurements of total bilirubin. Accurate measurements of total bilirubin may improve our understanding of the prevalence of hyperbilirubinemia in patients on eltrombopag therapy.

Introduction

Eltrombopag is an oral, small-molecule thrombopoietin receptor agonist approved for the treatment of adult [1] and pediatric [2] immune thrombocytopenic purpura (ITP), aplastic anemia [3] and hepatitis C virus-associated-thrombocytopenia [4]. It has also been studied in patients with myelodysplastic syndrome and acute myeloid leukemia [5]. The initial dosing regimen generally starts at 25–50 mg daily, with higher doses (≥75 mg) used for the treatment of aplastic anemia, myelodysplastic syndrome and acute myeloid leukemia.

Hepatobiliary adverse effects have been reported to occur in ~15% of patients on eltrombopag therapy [1]. The US Food and Drug Administration has suggested that discontinuation should be considered if alanine aminotransferase (ALT) rises to 3 times the upper limit of the reference interval [6]. Elevations in aspartate aminotransferase (AST), and bilirubin levels may be supportive evidence of hepatobiliary dysfunction. As continuous eltrombopag therapy is generally required to maintain treatment response, accurate biochemical assessment of liver function is important to inform dose adjustments and discontinuation.

Eltrombopag is a diacid with carboxyl and phenolic functional groups and its color in solution varies from yellow to reddish-brown with increasing pH [7], [8]. It can potentially interfere with the assessment of hyperbilirubinemia because of (i) its absorbance peak at ~450 nm overlapping with bilirubin [9], (ii) absorbance at ~550 nm affecting diazotized bilirubin [8] and (iii) eltrombopag has been reported to cause yellowish discoloration of the eyes in the presence of normal circulating bilirubin levels [10].

There have been multiple small-scale reports of high-dose eltrombopag (200–300 mg daily in adults [7], [8], [11], 75–150 mg daily in children [12]) causing a reddish-brown discoloration of plasma/serum and discrepant total bilirubin results by automated colorimetric assays. For example, two groups have demonstrated that the Beckman Coulter DxC 800 analyzer is susceptible to producing undetectable levels of total bilirubin in patients on high-dose eltrombopag [7], [8]. Discrepant bilirubin results (ranging from <2 to 100 μmol/L) due to eltrombopag therapy appear to be method- and reagent-specific [8].

Pharmacokinetic studies demonstrate that the Cmax after a single dose of 50–75 mg eltrombopag is ~10 mg/L occurring at 2.5–4 h post-dose [13], [14], [15]. Spike-in experiments showed that very high serum concentrations of eltrombopag of up to 500 mg/L [8], [11], [16] are required to significantly alter bilirubin measurements from baseline levels. Only accounting for the eltrombopag parent drug, this suggests that very high oral doses, and thus serum eltrombopag concentrations, will cause discrepant bilirubin results. However, none of the previous studies measured serum eltrombopag concentrations in patient samples with discrepant bilirubin results.

We noted a 79-year-old woman on a stable 50 mg daily eltrombopag for ITP who had her blood taken 6 days apart in two hospitals which utilized the Beckman AU 5800 and Roche cobas 8000 analyzers for routine chemistry testing. All of the biochemical analytes in her liver and renal function tests were very comparable, apart from total bilirubin, which measured 43 μmol/L (2.5 mg/dL) on the Beckman analyzer and 25 μmol/L (1.5 mg/dL) on the Roche analyzer. This patient was not clinically jaundiced.

In view of possible discrepant total bilirubin results in patients on as low as 50 mg daily dose of eltrombopag therapy, we collected 66 plasma samples from 27 patients on a wide range of eltrombopag doses to compare bilirubin measurements. We quantified plasma and serum eltrombopag levels using liquid chromatography tandem mass spectrometry (LC-MS/MS), and correlated concentrations with discrepant bilirubin measurements. The discrepant bilirubin results were further characterized by (i) dose-response spike-in experiments, (ii) admixture experiments with control pooled serum and (iii) bilirubin measurement using the Doumas reference method and high-performance liquid chromatography (HPLC).

Materials and methods

Study subjects and samples

In accordance with Institutional Review Board guidelines, all investigations were conducted for the purpose of accurate bilirubin quantitation for patient care. Plasma and serum samples were collected as part of routine clinical testing from October to December 2018 from the Prince of Wales Hospital, United Christian Hospital and Tseung Kwan O Hospital, Hong Kong. Plasma and serum samples were collected using Greiner Bio-One (Austria) Vacuette Lithium Heparin Sep and Z Serum Sep Clot Activator tubes. Archived samples that were surplus to clinical requirements were retrieved for further analysis. Samples were protected from light and kept at 4 °C for up to 6 days, and then stored at −80 °C. We collected a total of 66 samples from 27 patients on eltrombopag therapy for ITP and aplastic anemia. Sixty-three percent of the patients on eltrombopag were female, and the median age was 58 years (range: 10–88 years). This included two pediatric patients with severe aplastic anemia on 100 and 150 mg eltrombopag daily. The 10 patients under treatment for aplastic anemia were on a higher median eltrombopag dose of 75 mg daily compared to the 17 patients treated for ITP (median 25 mg). At the time of collection, patients were on eltrombopag therapy for a median of 22.6 weeks (IQR 17.0–62.9 weeks). No-eltrombopag control serum from patients not taking eltrombopag were randomly selected from the routine laboratory based on having comparable bilirubin levels.

Routine spectrophotometric measurement of bilirubin

Total and direct bilirubin were assayed on automated analyzers using diazo-based end-point reactions where bilirubin concentration was quantitated based on spectrophotometric measurements. Total and direct bilirubin were measured using the Beckman (USA) AU 5800 (TBILC and DBILC) and Roche (Switzerland) cobas 8000 (BILT3 and BILD2) systems. The Beckman AU 5800 system had a coefficient of variation (CV) of 1.7% and 1.3% at total bilirubin levels of 19 and 77 μmol/L (1.1 and 4.5 mg/dL), respectively. The CVs for direct bilirubin were 3.8% and 1.9% at levels of 5 and 25 μmol/L (0.3 and 1.5 mg/dL), respectively. The Roche cobas system total bilirubin had CVs of 2.7% and 1.7% at 17 and 84 μmol/L (1.0 and 4.9 mg/dL), respectively. The CVs for direct bilirubin were 2.5% and 1.5% for concentrations of 26 and 127 μmol/L (1.5 and 7.4 mg/dL). For a single eltrombopag patient serum sample, bilirubin was also measured using Abbott (USA) Architect c16000, Radiometer (Denmark) ABL835, Vitros (USA) dry slides (diazo-based total bilirubin, mordant-based unconjugated and conjugated bilirubin) and Wako, Fujifilm (Japan) vanadate method (reduction in bilirubin absorbance).

Doumas reference method of total bilirubin measurement and HPLC method of unconjugated bilirubin measurement

The Doumas reference method, based on Jendrassik-Grof, measures total bilirubin based on the spectrophotometric quantitation of azobilirubin at 598 nm after unconjugated bilirubin is released by caffeine, benzoate and acetate in the diazotized sulfanilic acid solution [17]. As the azobilirubin absorption spectra is pH dependent, the addition of alkaline tartrate (the final reagent) increases the pH and changes the absorption peak to near 598 nm making it less sensitive towards certain interferents. The overall absorbance was corrected by a blank value (sample without the indicator reaction). The measurement range of the accredited reference measurement procedure was 5–525 μmol/L (0.3–30.7 mg/dL) with inter-assay CVs of 0.5%–1.4% [18]. Measurements were done in an officially accredited calibration laboratory of the DGKL-RfB (German Association of Clinical Chemistry, Reference Institute for Bioanalytics).

Serum unconjugated bilirubin was measured using HPLC as previously described [19], [20]. LC was performed on a C18 column using a 30-min gradient and water-soluble molecules eluted first. Unconjugated bilirubin was detected in the form of (ZZ)-bilirubin (4Z,15Z-bilirubin, major component), which eluted at ~22.5 min and (ZE)-bilirubin (4E,15Z-bilirubin, main bilirubin photoisomer), which eluted at ~12.5 min. Bilirubin concentrations were determined based on the absorbance at 455 nm using the molar absorption coefficient with a CV of 1.4%.

LC-MS/MS quantitation of plasma and serum eltrombopag concentration

Plasma and serum eltrombopag concentration was determined by LC-MS/MS based on validated analytical methods [13], [14]. Eltrombopag and internal standard (eltrombopag-13C4) were sourced from Toronto Research Chemicals (Canada). After protein precipitation by methanol, a 5-min gradient was run on a C18 column, and eltrombopag eluted at 2.2 min and was detected using MRM channels of 443.23>183.06 m/z. Linear range for eltrombopag in plasma and serum was 0.05–30 mg/L (validated for three times dilution) with total imprecision <7.2%.

Dose-response spike-in experiments and admixture with control pooled serum

Interference studies including spike-in experiments and admixture experiments were done based on the Clinical and Laboratory Standards Institute (CLSI) guidelines EP07Ed3E Interference Testing in Clinical Chemistry 3rd Ed [21] and EP37Ed1E Supplementary Tables for Interference Testing in Clinical Chemistry – 1st Ed [22].

Statistical analyses were performed using R (https://www.R-project.org/).

Results

Dose-dependent inter-analyzer discrepancy in total bilirubin between Roche cobas and Beckman AU systems

Visual inspection of the plasma samples from patients on eltrombopag therapy showed no discoloration in patients on ≤25 mg daily dose. A slight reddish discoloration could be observed in some samples starting at 50 mg, and a deep reddish-brown discoloration could be seen in samples from patients on ≥100 mg daily (Supplementary Figure 1). Total and direct bilirubin were measured in a total of 66 plasma samples from 27 patients on eltrombopag therapy, and the inter-analyzer concordance was compared with 40 no-eltrombopag controls with total bilirubin levels ranging from 0 to 100 μmol/L (0–5.8 mg/dL).

In the no-eltrombopag controls, there was a strong correlation (Pearson’s r>0.99) between the Roche cobas and Beckman AU systems for both total bilirubin and direct bilirubin (Figure 1A and Supplementary Figure 2). For patients on eltrombopag therapy, the Beckman AU system produced higher total bilirubin results in a dose-dependent manner such that patients on a higher eltrombopag oral dose showed a larger discrepancy between the two analyzers. The Bland-Altman plot of total bilirubin difference (Beckman – Roche) against average total bilirubin concentrations showed that 49 of the 52 samples from patients on ≥50 mg daily eltrombopag had >3 SD (5.70 μmol/L|0.33 mg/dL) positive deviation from the mean difference defined by the no-eltrombopag controls (Figure 1B). There was no significant difference between patients on eltrombopag and no-eltrombopag controls for direct bilirubin measured on the two analyzers (Supplementary Figure 2, Mann-Whitney U-test, Beckman/Roche for eltrombopag patients vs. no-eltrombopag controls, p=0.44). The direct bilirubin was <16 μmol/L (<1 mg/dL) in all eltrombopag patient samples and direct bilirubin accounted for a median of 7% of total bilirubin. Thus, if the raised total bilirubin measured by Beckman represented a true hyperbilirubinemia, it would be predominantly contributed by indirect/unconjugated bilirubin.

Figure 1:
Figure 1:

Comparison between Roche cobas and Beckman AU total bilirubin.

(A) Correlation between Roche cobas and Beckman AU total bilirubin. The index line (y=x) is the dashed red line. 95% CIs are shaded in gray. For no-eltrombopag controls, Beckman=1.00 (95% CI 0.97–1.02)×Roche+1.38 (95% CI 0.73–1.93 μmol/L), r=0.997. For patients on eltrombopag therapy, Beckman=9.59 (95% CI 6.58–15.00)×Roche–61.00 (95% CI −115.70 to −31.89 μmol/L), r=0.296. (B) Bland-Altman (difference) plot for Beckman-Roche total bilirubin concentration. The horizontal dashed line represents the mean difference in no-eltrombopag controls. The semi-dashed and dotted horizontal lines represent two and three standard deviations from the mean difference (+1.0 μmol/L|0.060 mg/dL).

Citation: Clinical Chemistry and Laboratory Medicine (CCLM) 58, 10; 10.1515/cclm-2019-0684

Correlation between plasma eltrombopag concentration and discrepant total bilirubin results

Plasma eltrombopag concentrations were determined by LC-MS/MS for 59 of the 66 samples. There was a positive correlation between plasma eltrombopag concentration and total bilirubin difference (r=0.679, p=3.3×109, Figure 2A). For patients on 50–75 mg daily, plasma eltrombopag concentration ranged from <0.05 to 17.4 mg/L and the total bilirubin difference ranged from 0 to 23 μmol/L (0–1.3 mg/dL). For patients on 100–150 mg daily, plasma eltrombopag concentration ranged from 6.6 to 45.3 mg/L and the total bilirubin difference ranged from 12 to 64 μmol/L (0.7–3.7 mg/dL). We found that there was a positive correlation between plasma eltrombopag concentration and Beckman total bilirubin concentration (r=0.679, p=3.5×109). Plasma eltrombopag concentration and Roche total bilirubin concentrations were not correlated (p=0.43, Figure 2B). This may suggest that the positive correlation between plasma eltrombopag concentration and Beckman total bilirubin concentrations was driving the positive correlation observed between plasma eltrombopag concentration and the Beckman – Roche difference. Thus, the eltrombopag dose-dependent rise in Beckman total bilirubin could either be due to positive interference, or the rise in a (unconjugated) bilirubin species that is detectable by the Beckman assay and not the Roche assay.

Figure 2:
Figure 2:

Plasma eltrombopag and bilirubin concentration.

(A) Correlation between eltrombopag concentration and total bilirubin difference (Beckman-Roche). Linear regression (ordinary least squares) showed a significant positive correlation r=0.679, p=3.3×109. 95% CI is shaded in gray. y=1.29 (95% CI 1.02–1.86) x+10.85 (95% CI 5.44–16.02 μmol/L). (B) Correlation between eltrombopag concentration and total bilirubin concentration. There was a significant positive correlation between eltrombopag concentration and total bilirubin measured by Beckman (y=1.34 x [95% CI 1.07–1.94]+21.75 [95% CI 15.34–27.15 μmol/L], r=0.679, p=3.5×109), but not total bilirubin measured by Roche (p=0.43). Gray vertical dotted lines connect the corresponding Beckman (red) and Roche (blue) total bilirubin measurement for each sample, such that the length of the dotted line represents the (Beckman-Roche) difference in part A of the figure. 95% CIs are shaded in gray.

Citation: Clinical Chemistry and Laboratory Medicine (CCLM) 58, 10; 10.1515/cclm-2019-0684

For direct bilirubin, there was no correlation between plasma eltrombopag concentration and Roche and Beckman measurements, or the difference (p>0.28, Supplementary Figure 3).

Eltrombopag dose-response spike-in experiments

We conducted dose-response spike-in experiments with five levels of eltrombopag concentration in order to assess how increasing concentrations of eltrombopag (parent drug without metabolites) affect the magnitude of discrepant total bilirubin results. Level 1 consisted of serum pooled from no-eltrombopag controls with undetectable eltrombopag levels by LC-MS/MS. This served as the base pool serum with total bilirubin and direct bilirubin concentrations of 27 and 11 μmol/L (1.6 mg/dL and 0.7 mg/dL), which approximates the bilirubin levels we see in the eltrombopag patients. An 8 g/L interferent stock solution was made from pure eltrombopag dissolved in DMSO, and increasing amounts were spiked-in to aliquots of the base pool serum to reach final concentrations as follows:

  • Level 1: no-eltrombopag base pool serum
  • Level 2: 10 mg/L, ~Cmax after a single 50–75 mg oral dose
  • Level 3: 50 mg/L, approximating maximal concentration (45.3 mg/L) observed in our case series
  • Level 4: 150 mg/L
  • Level 5: 300 mg/L, CLSI recommended maximal eltrombopag test concentration, 3 times the highest drug concentration under therapeutic treatment (100 mg/L).

To minimize dilution of the sample matrix, for the maximal Level 5 (300 mg/L eltrombopag), the interferent stock solution contributed only 3.8% of the total volume. We also prepared a base pool serum with 4% DMSO to verify that the solvent did not cause significant differences in bilirubin measurement to the base pool serum (difference <2 μmol/L|<0.1 mg/dL, p>0.53).

Total and direct bilirubin levels were measured in triplicate using the Roche cobas and Beckman AU systems, and total bilirubin was also measured using the Doumas reference method. In parallel to Levels 1–5 in the spike-in experiment, we measured the pooled serum (ETP patient A, serum eltrombopag concentration=11.3 mg/L) from a patient with aplastic anemia on 100 mg daily eltrombopag who persistently had a total bilirubin of 60–80 μmol/L (3.5–4.7 mg/dL) by Beckman and 10–20 μmol/L (0.6–1.2 mg/dL) by Roche analyzers. The patient’s eyes had a tinge of yellowish discoloration. In comparison to the spike-in samples, this patient sample likely contained eltrombopag metabolites as well as the parent drug.

Visual inspection of the serum samples showed that the ETP patient A serum (LC-MS/MS: 11.3 mg/L) had a darker reddish-brown color compared to the Level 3 spiked with 50 mg/L eltrombopag (Figure 3) suggesting that metabolites contributed to the color. Roche total bilirubin values remained stable from Level 1–5 (average of triplicate measurements 26–27 μmol/L, 1.5–1.6 mg/dL) whereas both the Beckman and Doumas methods showed a monotonic increasing sequence with increasing spiked eltrombopag concentrations (Figure 3). The CVs for all triplicate measurements were <3.1%. At an eltrombopag concentration of 300 mg/L, the Beckman total bilirubin showed a 37% increase from the baseline (10 μmol/L|0.6 mg/dL), and the Doumas total bilirubin showed a 284% increase from the baseline (71 μmol/L|4.2 mg/dL). The absorbance spectrum for the end-point of the diazo reaction for the Doumas method demonstrates how eltrombopag contributes to the positive interference at the measurement wavelength of 598 nm (Supplementary Figure 4). The reaction monitoring for Roche and Beckman systems was unremarkable.

Figure 3:
Figure 3:

Eltrombopag dose-response spike-in experiment.

The appearance of the serum samples is displayed above, including serum from a patient on 100 mg daily eltrombopag therapy (ETP patient A). Total bilirubin concentration measured in triplicate by Roche (blue), Beckman (red) and Doumas reference method (green) in dose-response spike-in experiment with Levels 1–5 eltrombopag concentration of up to 300 mg/L. 95% CIs of triplicate measurements are marked with error bars, and the average of three measurements are marked as black crosses. 95% CIs of the linear regression (ordinary least squares) lines are shaded in grey. Beckman (y=0.033 [95% CI 0.032 – 0.037] x+27.90 (95% CI 27.61–28.27 μmol/L), r=0.994, p=6.4×10−14) and Doumas (y=0.24 [95% CI 0.23–0.28] x+26.77 [95% CI 25.22–28.29 μmol/L], r=0.995, p=3.00×10−14).

Citation: Clinical Chemistry and Laboratory Medicine (CCLM) 58, 10; 10.1515/cclm-2019-0684

For the eltrombopag patient serum with 11.3 mg/L eltrombopag, the Doumas method produced a total bilirubin of 15 μmol/L (0.9 mg/dL) despite the possibility of positive interference, while the Beckman result total bilirubin result was elevated at 65 μmol/L (3.8 mg/dL). Interpreted in conjunction with the visual appearance of the serum sample and correlation between plasma eltrombopag and discrepant results (Figure 2), this suggests that eltrombopag metabolites may contribute to total bilirubin interference and that eltrombopag (parent) and its metabolites may contribute to a variable extent to positive interference in different total bilirubin assays. The direct bilirubin concentrations measured by Roche and Beckman were not affected by up to 300 mg/L eltrombopag (difference 1 μmol/L|<0.1 mg/dL, Supplementary Figure 5).

Bilirubin measurement of eltrombopag patient serum with HPLC and other routine analyzers

Since the Doumas reference method is susceptible to positive interference by eltrombopag, we sought to quantify the serum bilirubin concentration in eltrombopag patient serum samples using HPLC separation. As Beckman bilirubin results pointed to a predominant indirect hyperbilirubinemia, we employed a 30-min HPLC protocol for unconjugated bilirubin derived by molar absorbance. HPLC was able to separate eltrombopag and its metabolites in the measurement of unconjugated bilirubin (Supplementary Figure 6).

Duplicate analysis of the Level 1–5 (eltrombopag 0–300 mg/L) serum of the spike-in experiment showed that HPLC consistently quantitated the unconjugated bilirubin levels at 15–16 μmol/L (0.8–0.9 mg/dL).

We measured HPLC unconjugated bilirubin, Roche direct bilirubin and Roche total bilirubin for 30 no-eltrombopag controls and 27 eltrombopag (0.3–20.7 mg/L) patient samples. There was a strong positive correlation between total bilirubin quantitated as (i) HPLC unconjugated bilirubin+Roche direct bilirubin, and (ii) Roche total bilirubin in controls (r=0.995, Figure 4A). Similarly, there was a strong correlation for eltrombopag patients regardless of the oral dose, and the Bland-Altman plot showed that the total bilirubin differences observed in eltrombopag patients were within 3 SDs of the mean (as defined by the controls, Figure 4B). This suggests that the Roche total and direct bilirubin were less susceptible to interference by eltrombopag and its metabolites and is able to produce bilirubin results that are compatible with HPLC. In contrast, Beckman bilirubin measurements were not compatible with HPLC in patients on eltrombopag therapy (Supplementary Figure 7).

Figure 4:
Figure 4:

Relationship between HPLC and Roche total and direct bilirubin.

(A) Correlation between total bilirubin (quantitated by HPLC unconjugated bilirubin+Roche direct bilirubin) and Roche total bilirubin. Total bilirubin measurements for 30 no-eltrombopag controls are in gray (y=0.94 [95% CI 0.89–0.97] x–0.20 [95% CI −0.81 to 0.44 μmol/L], R=0.995). The index line (y=x) is the dashed red line. 95% CI is shaded in gray. (B) Bland-Altman (difference) plot for Roche total bilirubin – total bilirubin (HPLC unconjugated bilirubin+Roche direct bilirubin). The semi-dashed and dotted horizontal lines represent two and three standard deviations from the mean difference (−2.0 μmol/L|0.12 mg/dL), based on no-eltrombopag controls.

Citation: Clinical Chemistry and Laboratory Medicine (CCLM) 58, 10; 10.1515/cclm-2019-0684

Table 1 shows the bilirubin measurements (average of two to three measurements) for the eltrombopag patient serum (ETP patient A) using (i) diazo-based spectrophotometric methods: Roche cobas c8000, Abbott Architect c16000, Beckman AU 5800, Doumas reference method, (ii) dry chemistry: Vitros diazo-based total bilirubin, mordant-based unconjugated and conjugated bilirubin, (iii) direct spectrophotometric: Radiometer ABL835, (iv) vanadate-based reduction in bilirubin absorbance (Wako, Fujifilm) and HPLC unconjugated bilirubin. Replicate measurements differed by less than 2 μmol/L (0.12 mg/dL). These results suggest that some analyzers are susceptible to interference to varying degrees.

Table 1:

Bilirubin measurement using different platforms.

AnalyzerMethodTotal bilirubin (TB) μmol/L, mg/dLDirect bilirubin (DB) μmol/L, mg/dLCalculated (TB-DB) indirect bilirubin μmol/L, mg/dLConjugated bilirubin μmol/L, mg/dLUnconjugated bilirubin μmol/L, mg/dL
Wako (Fujifilm)Vanadate oxidation; DB and TB 450/546a nm, R1 pH 2.9, R2 pH 7.0 (both assays). Traceable to SRM 916a

Refs. 23695, 23295, 413-23891, 416-22801
8 (0.4)4 (0.2)4 (0.2)
Roche cobas c80003,5-Dichlorophenyl diazonium salt; DB 546/800a nm R1 pH 1.9, R2 pH 1.3, TB 546/600a nm, R1 pH 1.0, R2 not stated.

Standardized against Doumas reference method.

Refs. 05589061, 05795419
11 (0.6)4 (0.2)7 (0.4)
Abbott Architect c160002,4-Dichloroaniline diazonium; DB 548/660a nm, TB 548/604a nm, R1 contains sulfamic acid (DB) and HCl (TB), pH not stated

Refs. 8G63, 6L45
12 (0.7)3 (0.2)9 (0.5)
Vitros TBIL slide4-(N-carboxymethylsulfonyl) benzenediazonium hexafluorophosphate, 520/460a nm, pH 3.0, traceable to SRM 916a

Ref. 815 9931
30 (1.8)
Vitros BuBc slideMordant-based; 400, 460 nm, pH 8.0

Ref. 838 3051, 161 2365
0 (0)14 (0.8)
Radiometer ABL835Direct spectrophotometric measurement of undiluted whole blood at multiple wavelengths (478–672 nm)61 (3.6)
Beckman AU 58003,5-Dichlorophenyldiazonium tetrafluoroborate;

DB 570/660a nm pH <1.0, TB 540/660a nm pH 4.2, traceable to SRM 916a, separate sample blank channel

Refs. OSR6111, OSR6112
65 (3.8)3 (0.1)62 (3.7)
Doumas reference methodDiazonium, 598 nm, with separate sample blank tube, pH ~6 for diazotization, pH ~13 after the addition of tartrate15 (0.9)
HPLC bilirubinHPLC; 30 min gradient, molar absorbance 455 nm12 (0.7)

Billirubin was measured for a single sample from ETP patient A on 100 mg eltrombopag therapy, serum eltrombopag concentration was 11.3 mg/L. The average of two to three replicates is displayed. aSecondary wavelengths for dichromatic measurements. Roche TB and DB measurements are compatible with HPLC bilirubin (see Figure 4).

Assessment of linearity by admixture of eltrombopag patient serum and no-eltrombopag control serum

Patient samples can contain both the parent drug and its metabolites, and spike-in experiments can only assess the former of the two components. HPLC bilirubin is not available to most routine laboratories when faced with discrepant results or the suspicion of interference.

We assessed if a nine-point five-replicate admixture experiment with a no-eltrombopag control pool could be used to look for the loss-of-linearity in an analyzer affected by positive interference. We again pooled serum from a patient on 100 mg daily eltrombopag (ETP patient A, serum eltrombopag concentration=17.8 mg/L, Roche total bilirubin of 11 μmol/L (0.7 mg/dL) and Beckman total bilirubin of 59 μmol/L (3.5 mg/dL)). The eltrombopag patient pool was admixed with no-eltromobpag control serum (Roche total bilirubin 10 μmol/L|0.6 mg/dL, Beckman 12 μmol/L|0.7 mg/dL) with five replicates for each level (CVs <3.3% for replicate measurements). For both the Roche and Beckman analyzers, the observed results are comparable to the expected results based on admixture proportions. There is no obvious loss-of-linearity for the Beckman system affected by positive interference (Figure 5). This suggests that the assay linearity is preserved (when the proportion of eltrombopag and its metabolites are kept constant), and there is no threshold interferent concentration where the discrepancy is disproportionately more severe. The preserved linearity may also suggest that even low eltrombopag (and metabolites) concentrations may be causing a small degree of positive interference, even if the degree of discrepancy is not significantly different compared with the biological or analytical variation.

Figure 5:
Figure 5:

Admixture experiment to assess linearity.

Serum from a patient on eltrombopag therapy (serum eltrombopag=17.8 mg/L) with discrepant total bilirubin results: Roche total bilirubin of 11 μmol/L (0.7 mg/dL) and Beckman total bilirubin of 59 μmol/L (3.5 mg/dL). Admixture with no-eltrombopag control serum (Roche total bilirubin 10 μmol/L | 0.6 mg/dL, Beckman 12 μmol/L|0.7 mg/dL) shows no obvious loss of linearity for both the Roche and Beckman assays, despite positive interference in the Beckman system.

Citation: Clinical Chemistry and Laboratory Medicine (CCLM) 58, 10; 10.1515/cclm-2019-0684

Discussion

Eltrombopag represents an intriguing challenge to clinical chemists because it is a drug with hepatotoxic effects that can affect the biochemical measurement and clinical assessment of hyperbilirubinemia by multiple mechanisms. Patients on eltrombopag therapy have differing total bilirubin results when measured by direct, vanadate-based or diazo-based spectrophotometric methods. As previously reported [10], we also had a patient with yellowish discoloration of the eyes (ETP patient A, see Table 1) when circulating bilirubin levels were normal (by HPLC). Thus, clinical jaundice may be unreliable in the assessment of hyperbilirubinemia in patients on eltrombopag.

Our results demonstrate that discoloration of plasma can be seen in some samples starting at 50 mg daily dose in patients on eltrombopag therapy. Simultaneous analysis using two diazo-based spectrophotometric total bilirubin measurements on routine analyzers revealed that 49 of the 52 samples from patients on 50–150 mg daily eltrombopag showed significant inter-analyzer discrepancy with elevated total bilirubin on the Beckman AU 5800 analyzaer. This is important because 25–50 mg daily is the starting dose for ITP, with higher doses required for other conditions such as aplastic anemia. The quantitation of direct bilirubin appears not to be affected as there was no significant inter-analyzer differences between no-eltrombopag controls and patients on eltrombopag.

Plasma eltrombopag concentrations in these patient samples were less than 50 mg/L, and the correlation between eltrombopag concentration and bilirubin difference was only 0.679. Dose-response spike-in experiments showed that much larger concentrations of eltrombopag (parent) were required to significantly affect bilirubin measurement, and thus providing evidence for the first time that eltrombopag metabolites likely play a role in bilirubin interference. This renders previous eltrombopag spike-in studies of limited value because significant interference can be present at much lower plasma/serum eltrombopag concentrations. Mono-oxygenation (M1) and acyl glucuronide (M2) metabolites of eltrombopag can be detected in the systemic circulation after eltrombopag intake [23], and further investigations are required to delineate their effects on bilirubin measurement. The proportion of M2 in plasma rises up to 48 h after a single dose [23], and thus the concentration of metabolites may be increased in patients on chronic therapy. Variable interferent effects may be caused by differing population pharmacokinetics [15] and variable concentrations of eltrombopag and its metabolites.

The Doumas reference method is susceptible to positive interference by eltrombopag, and thus cannot be relied upon for the accurate quantitation of total bilirubin in patients on eltrombopag therapy. Chromatographic separation of eltrombopag from bilirubin demonstrated that the Roche cobas 8000 analytical system is less susceptible to interference compared to some of the other analyzers employing spectrophotometric methods. This may be related to differences in measurement parameters such as reagent composition, pH, and measurement wavelengths.

Eltrombopag can cause negative interference to Beckman DxC [7], [8], and varying levels of positive interference for total bilirubin by direct spectrophotometric methods and diazo-based spectrophotometric methods. Furthermore, our current understanding of hyperbilirubinemia and hepatotoxicity in patients on eltrombopag therapy can be dependent on the degree of interference in different analyzers, and accurate quantification of total bilirubin in eltrombopag patients may improve our understanding of its prevalence.

Acknowledgments

The authors would like to thank Danny K.W. Wong, Shui-Wah Ng, Giselle Dongji Li, Michael Peake and Kay Weng Choy.

Author contributions: All the authors have accepted responsibility for the entire content of this submitted manuscript and approved submission.

Research funding: None declared.

Employment or leadership: None declared.

Honorarium: None declared.

Competing interests: The funding organization(s) played no role in the study design; in the collection, analysis, and interpretation of data; in the writing of the report; or in the decision to submit the report for publication.

References

  • 1.

    Wong RS, Saleh MN, Khelif A, Salama A, Portella MS, Burgess P, et al. Safety and efficacy of long-term treatment of chronic/persistent ITP with eltrombopag: final results of the EXTEND study. Blood 2017;130:2527–36.

    • Crossref
    • PubMed
    • Export Citation
  • 2.

    Grainger JD, Blanchette VS, Grotzinger KM, Roy A, Bussel JB. Health-related quality of life in children with chronic immune thrombocytopenia treated with eltrombopag in the PETIT study. Br J Haematol 2019;185:102–6.

    • Crossref
    • PubMed
    • Export Citation
  • 3.

    Townsley DM, Scheinberg P, Winkler T, Desmond R, Dumitriu B, Rios O, et al. Eltrombopag added to standard immunosuppression for aplastic anemia. N Engl J Med 2017;376:1540–50.

    • Crossref
    • PubMed
    • Export Citation
  • 4.

    Dusheiko G, Afdhal NH, Giannini E, Chen P-J, Han K-H, Kamel YM, et al. Final results of open-label treatment with eltrombopag during ENABLE 1: a study of eltrombopag as an adjunct for antiviral treatment of hepatitis C virus associated with thrombocytopenia. Blood 2011;118:2232.

    • Crossref
    • Export Citation
  • 5.

    Mittelman M, Platzbecker U, Afanasyev B, Grosicki S, Wong RS, Anagnostopoulos A, et al. Eltrombopag for advanced myelodysplastic syndromes or acute myeloid leukaemia and severe thrombocytopenia (ASPIRE): a randomised, placebo-controlled, phase 2 trial. Lancet Haematol 2018;5:e34–43.

    • Crossref
    • PubMed
    • Export Citation
  • 7.

    Cardamone D, Milone MC, Glaser L, Frey NV, Kricka LJ. Eltrombopag and serum of a different hue. Arch Pathol Lab Med 2013;137:1175.

    • Crossref
    • PubMed
    • Export Citation
  • 8.

    Choy KW, Wijeratne N, Doery JC. Eltrombopag: liver toxicity, kidney injury or assay interference? Pathology 2016;48:754–6.

    • PubMed
    • Export Citation
  • 9.

    Meloun M, Pilařová L, Pekárek T, Javůrek M. Overlapping pK a of the multiprotic hemostyptic eltrombopag using UV-Vis multiwavelength spectroscopy and potentiometry. J Solution Chem 2017;46:2014–37.

    • Crossref
    • Export Citation
  • 10.

    Rodgers GM, Kurtti AL, Gilreath JA. Are eltrombopag plasma and skin hyperpigmentation related? The eyes have it. Am J Hematol 2019;94:394–5.

    • Crossref
    • PubMed
    • Export Citation
  • 11.

    Gounden V, Zhao Z. Eltrombopag interference in routine chemistry testing. Ann Clin Biochem 2016;53:611–4.

    • Crossref
    • PubMed
    • Export Citation
  • 12.

    Kutasz E, Barbieri A, D’Isa G, Toscano A, Chilelli C, Jurado V, et al. Eltrombopag interferes with the measurement of plasma total bilirubin in pediatric patients in an automated colorimetric method. Clin Chem Lab Med 2019;57:e111–3.

    • Crossref
    • Export Citation
  • 13.

    Williams DD, Peng B, Bailey CK, Wire MB, Deng Y, Park JW, et al. Effects of food and antacids on the pharmacokinetics of eltrombopag in healthy adult subjects: two single-dose, open-label, randomized-sequence, crossover studies. Clin Ther 2009;31:764–76.

    • Crossref
    • PubMed
    • Export Citation
  • 14.

    Maddela R, Gajula R, Pilli NR, Siddiraju S, Maddela S, Makula A. Liquid chromatography-tandem mass spectrometric assay for eltrombopag in 50 μL of human plasma: a pharmacokinetic study. J Pharm Biomed Anal 2014;98:68–73.

    • Crossref
    • Export Citation
  • 15.

    Gibiansky E, Zhang J, Williams D, Wang Z, Ouellet D. Population pharmacokinetics of eltrombopag in healthy subjects and patients with chronic idiopathic thrombocytopenic purpura. J Clin Pharmacol 2011;51:842–56.

    • Crossref
    • PubMed
    • Export Citation
  • 16.

    Adams DF, Sellers TS. Eltrombopag: investigation of previously reported serum discoloration and negative interference on a total bilirubin diazo method. Arch Pathol Lab Med 2016;140:391–2.

    • Crossref
    • PubMed
    • Export Citation
  • 17.

    Doumas BT, Kwok-Cheung PP, Perry BW, Jendrzejczak B, McComb RB, Schaffer R, et al. Candidate reference method for determination of total bilirubin in serum: development and validation. Clin Chem 1985;31:1779–89.

    • Crossref
    • PubMed
    • Export Citation
  • 18.

    Klauke R, Kytzia H-J, Weber F, Grote-Koska D, Brand K, Schumann G. Reference measurement procedure for total bilirubin in serum re-evaluated and measurement uncertainty determined. Clin Chim Acta 2018;481:115–20.

    • Crossref
    • PubMed
    • Export Citation
  • 19.

    Itoh S, Isobe K, Onishi S. Accurate and sensitive high-performance liquid chromatographic method for geometrical and structural photoisomers of bilirubin IX alpha using the relative molar absorptivity values. J Chromatogr A 1999;848:169–77.

    • Crossref
    • PubMed
    • Export Citation
  • 20.

    Okada H, Kawada K, Itoh S, Ozaki M, Kakutani I, Arai T, et al. Effects of bilirubin photoisomers on the measurement of direct bilirubin by the bilirubin oxidase method. Ann Clin Biochem 2018;55:276–80.

    • Crossref
    • PubMed
    • Export Citation
  • 21.

    CLSI. Interference testing in clinical chemistry, 3rd ed. CLSI guideline EP07. Wayne, PA: Clinical and Laboratory Standards Institute, 2018.

  • 22.

    CLSI. Supplemental tables for interference testing in clinical chemistry, 1st ed. CLSI supplement EP37. Wayne, PA: Clinical and Laboratory Standards Institute, 2018.

  • 23.

    Deng Y, Madatian A, Wire MB, Bowen C, Park JW, Williams D, et al. Metabolism and disposition of eltrombopag, an oral, nonpeptide thrombopoietin receptor agonist, in healthy human subjects. Drug Metab Dispos 2011;39:1734–46.

    • Crossref
    • PubMed
    • Export Citation

Footnotes

Supplementary Material

The online version of this article offers supplementary material (https://doi.org/10.1515/cclm-2019-0684).

If the inline PDF is not rendering correctly, you can download the PDF file here.

  • 1.

    Wong RS, Saleh MN, Khelif A, Salama A, Portella MS, Burgess P, et al. Safety and efficacy of long-term treatment of chronic/persistent ITP with eltrombopag: final results of the EXTEND study. Blood 2017;130:2527–36.

    • Crossref
    • PubMed
    • Export Citation
  • 2.

    Grainger JD, Blanchette VS, Grotzinger KM, Roy A, Bussel JB. Health-related quality of life in children with chronic immune thrombocytopenia treated with eltrombopag in the PETIT study. Br J Haematol 2019;185:102–6.

    • Crossref
    • PubMed
    • Export Citation
  • 3.

    Townsley DM, Scheinberg P, Winkler T, Desmond R, Dumitriu B, Rios O, et al. Eltrombopag added to standard immunosuppression for aplastic anemia. N Engl J Med 2017;376:1540–50.

    • Crossref
    • PubMed
    • Export Citation
  • 4.

    Dusheiko G, Afdhal NH, Giannini E, Chen P-J, Han K-H, Kamel YM, et al. Final results of open-label treatment with eltrombopag during ENABLE 1: a study of eltrombopag as an adjunct for antiviral treatment of hepatitis C virus associated with thrombocytopenia. Blood 2011;118:2232.

    • Crossref
    • Export Citation
  • 5.

    Mittelman M, Platzbecker U, Afanasyev B, Grosicki S, Wong RS, Anagnostopoulos A, et al. Eltrombopag for advanced myelodysplastic syndromes or acute myeloid leukaemia and severe thrombocytopenia (ASPIRE): a randomised, placebo-controlled, phase 2 trial. Lancet Haematol 2018;5:e34–43.

    • Crossref
    • PubMed
    • Export Citation
  • 7.

    Cardamone D, Milone MC, Glaser L, Frey NV, Kricka LJ. Eltrombopag and serum of a different hue. Arch Pathol Lab Med 2013;137:1175.

    • Crossref
    • PubMed
    • Export Citation
  • 8.

    Choy KW, Wijeratne N, Doery JC. Eltrombopag: liver toxicity, kidney injury or assay interference? Pathology 2016;48:754–6.

    • PubMed
    • Export Citation
  • 9.

    Meloun M, Pilařová L, Pekárek T, Javůrek M. Overlapping pK a of the multiprotic hemostyptic eltrombopag using UV-Vis multiwavelength spectroscopy and potentiometry. J Solution Chem 2017;46:2014–37.

    • Crossref
    • Export Citation
  • 10.

    Rodgers GM, Kurtti AL, Gilreath JA. Are eltrombopag plasma and skin hyperpigmentation related? The eyes have it. Am J Hematol 2019;94:394–5.

    • Crossref
    • PubMed
    • Export Citation
  • 11.

    Gounden V, Zhao Z. Eltrombopag interference in routine chemistry testing. Ann Clin Biochem 2016;53:611–4.

    • Crossref
    • PubMed
    • Export Citation
  • 12.

    Kutasz E, Barbieri A, D’Isa G, Toscano A, Chilelli C, Jurado V, et al. Eltrombopag interferes with the measurement of plasma total bilirubin in pediatric patients in an automated colorimetric method. Clin Chem Lab Med 2019;57:e111–3.

    • Crossref
    • Export Citation
  • 13.

    Williams DD, Peng B, Bailey CK, Wire MB, Deng Y, Park JW, et al. Effects of food and antacids on the pharmacokinetics of eltrombopag in healthy adult subjects: two single-dose, open-label, randomized-sequence, crossover studies. Clin Ther 2009;31:764–76.

    • Crossref
    • PubMed
    • Export Citation
  • 14.

    Maddela R, Gajula R, Pilli NR, Siddiraju S, Maddela S, Makula A. Liquid chromatography-tandem mass spectrometric assay for eltrombopag in 50 μL of human plasma: a pharmacokinetic study. J Pharm Biomed Anal 2014;98:68–73.

    • Crossref
    • Export Citation
  • 15.

    Gibiansky E, Zhang J, Williams D, Wang Z, Ouellet D. Population pharmacokinetics of eltrombopag in healthy subjects and patients with chronic idiopathic thrombocytopenic purpura. J Clin Pharmacol 2011;51:842–56.

    • Crossref
    • PubMed
    • Export Citation
  • 16.

    Adams DF, Sellers TS. Eltrombopag: investigation of previously reported serum discoloration and negative interference on a total bilirubin diazo method. Arch Pathol Lab Med 2016;140:391–2.

    • Crossref
    • PubMed
    • Export Citation
  • 17.

    Doumas BT, Kwok-Cheung PP, Perry BW, Jendrzejczak B, McComb RB, Schaffer R, et al. Candidate reference method for determination of total bilirubin in serum: development and validation. Clin Chem 1985;31:1779–89.

    • Crossref
    • PubMed
    • Export Citation
  • 18.

    Klauke R, Kytzia H-J, Weber F, Grote-Koska D, Brand K, Schumann G. Reference measurement procedure for total bilirubin in serum re-evaluated and measurement uncertainty determined. Clin Chim Acta 2018;481:115–20.

    • Crossref
    • PubMed
    • Export Citation
  • 19.

    Itoh S, Isobe K, Onishi S. Accurate and sensitive high-performance liquid chromatographic method for geometrical and structural photoisomers of bilirubin IX alpha using the relative molar absorptivity values. J Chromatogr A 1999;848:169–77.

    • Crossref
    • PubMed
    • Export Citation
  • 20.

    Okada H, Kawada K, Itoh S, Ozaki M, Kakutani I, Arai T, et al. Effects of bilirubin photoisomers on the measurement of direct bilirubin by the bilirubin oxidase method. Ann Clin Biochem 2018;55:276–80.

    • Crossref
    • PubMed
    • Export Citation
  • 21.

    CLSI. Interference testing in clinical chemistry, 3rd ed. CLSI guideline EP07. Wayne, PA: Clinical and Laboratory Standards Institute, 2018.

  • 22.

    CLSI. Supplemental tables for interference testing in clinical chemistry, 1st ed. CLSI supplement EP37. Wayne, PA: Clinical and Laboratory Standards Institute, 2018.

  • 23.

    Deng Y, Madatian A, Wire MB, Bowen C, Park JW, Williams D, et al. Metabolism and disposition of eltrombopag, an oral, nonpeptide thrombopoietin receptor agonist, in healthy human subjects. Drug Metab Dispos 2011;39:1734–46.

    • Crossref
    • PubMed
    • Export Citation
OPEN ACCESS

Journal + Issues

Search

  • View in gallery

    Comparison between Roche cobas and Beckman AU total bilirubin.

    (A) Correlation between Roche cobas and Beckman AU total bilirubin. The index line (y=x) is the dashed red line. 95% CIs are shaded in gray. For no-eltrombopag controls, Beckman=1.00 (95% CI 0.97–1.02)×Roche+1.38 (95% CI 0.73–1.93 μmol/L), r=0.997. For patients on eltrombopag therapy, Beckman=9.59 (95% CI 6.58–15.00)×Roche–61.00 (95% CI −115.70 to −31.89 μmol/L), r=0.296. (B) Bland-Altman (difference) plot for Beckman-Roche total bilirubin concentration. The horizontal dashed line represents the mean difference in no-eltrombopag controls. The semi-dashed and dotted horizontal lines represent two and three standard deviations from the mean difference (+1.0 μmol/L|0.060 mg/dL).

  • View in gallery

    Plasma eltrombopag and bilirubin concentration.

    (A) Correlation between eltrombopag concentration and total bilirubin difference (Beckman-Roche). Linear regression (ordinary least squares) showed a significant positive correlation r=0.679, p=3.3×109. 95% CI is shaded in gray. y=1.29 (95% CI 1.02–1.86) x+10.85 (95% CI 5.44–16.02 μmol/L). (B) Correlation between eltrombopag concentration and total bilirubin concentration. There was a significant positive correlation between eltrombopag concentration and total bilirubin measured by Beckman (y=1.34 x [95% CI 1.07–1.94]+21.75 [95% CI 15.34–27.15 μmol/L], r=0.679, p=3.5×109), but not total bilirubin measured by Roche (p=0.43). Gray vertical dotted lines connect the corresponding Beckman (red) and Roche (blue) total bilirubin measurement for each sample, such that the length of the dotted line represents the (Beckman-Roche) difference in part A of the figure. 95% CIs are shaded in gray.

  • View in gallery

    Eltrombopag dose-response spike-in experiment.

    The appearance of the serum samples is displayed above, including serum from a patient on 100 mg daily eltrombopag therapy (ETP patient A). Total bilirubin concentration measured in triplicate by Roche (blue), Beckman (red) and Doumas reference method (green) in dose-response spike-in experiment with Levels 1–5 eltrombopag concentration of up to 300 mg/L. 95% CIs of triplicate measurements are marked with error bars, and the average of three measurements are marked as black crosses. 95% CIs of the linear regression (ordinary least squares) lines are shaded in grey. Beckman (y=0.033 [95% CI 0.032 – 0.037] x+27.90 (95% CI 27.61–28.27 μmol/L), r=0.994, p=6.4×10−14) and Doumas (y=0.24 [95% CI 0.23–0.28] x+26.77 [95% CI 25.22–28.29 μmol/L], r=0.995, p=3.00×10−14).

  • View in gallery

    Relationship between HPLC and Roche total and direct bilirubin.

    (A) Correlation between total bilirubin (quantitated by HPLC unconjugated bilirubin+Roche direct bilirubin) and Roche total bilirubin. Total bilirubin measurements for 30 no-eltrombopag controls are in gray (y=0.94 [95% CI 0.89–0.97] x–0.20 [95% CI −0.81 to 0.44 μmol/L], R=0.995). The index line (y=x) is the dashed red line. 95% CI is shaded in gray. (B) Bland-Altman (difference) plot for Roche total bilirubin – total bilirubin (HPLC unconjugated bilirubin+Roche direct bilirubin). The semi-dashed and dotted horizontal lines represent two and three standard deviations from the mean difference (−2.0 μmol/L|0.12 mg/dL), based on no-eltrombopag controls.

  • View in gallery

    Admixture experiment to assess linearity.

    Serum from a patient on eltrombopag therapy (serum eltrombopag=17.8 mg/L) with discrepant total bilirubin results: Roche total bilirubin of 11 μmol/L (0.7 mg/dL) and Beckman total bilirubin of 59 μmol/L (3.5 mg/dL). Admixture with no-eltrombopag control serum (Roche total bilirubin 10 μmol/L | 0.6 mg/dL, Beckman 12 μmol/L|0.7 mg/dL) shows no obvious loss of linearity for both the Roche and Beckman assays, despite positive interference in the Beckman system.