| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Drug Monitoring and Toxicology |
1 Immunosuppressive Drug Monitoring, Institute of Liver Studies, King’s College Hospital, London, UK.
2 Institute of Liver Studies, Guy’s King’s and St. Thomas’ School of Medicine, King’s College London, London, UK.
aAddress correspondence to this author at: Immunosuppressive Drug Monitoring, Institute of Liver Studies, King’s College Hospital, Denmark Hill, London SE5 9RS, UK. Fax 44-20-7346-3760; e-mail nigel.brown{at}kingsch.nhs.uk.
 |
Abstract |
|---|
 Top Abstract IntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Methods: We collected 1156 blood samples from 277 adult and 121 pediatric recipients of liver, renal, and bone marrow grafts or hepatocyte or pancreatic islet cell implants. TRL was measured in whole blood by MEIA and LC-MS/MS, and hematologic and biochemical data were collected when available.
Results: LC-MS/MS was significantly more precise (P <0.02) than the MEIA with increased sensitivity. The MEIA had a median difference of 16.2% vs LC-MS/MS overall, and this was significantly affected by patient cohort (P <0.001). The difference was greater in adult or pediatric liver graft recipients while they were inpatients rather than outpatients (31.8% and 14.0% vs 7.5% and 6.5%, respectively). The difference was also greater in bone marrow than kidney graft recipients (32.8% vs 15.8%, respectively). Multiple linear regression analysis showed significant inverse relationships of this difference with hematocrit (packed cell volume) and plasma albumin (P <0.001) in the total cohort and a positive relationship with plasma bilirubin in a subgroup of pediatric liver graft recipients.
Conclusions: Patients with a low packed cell volume and plasma albumin are likely to show artificially high concentrations of TRL when measured by MEIA. The increased risk of underimmunosuppression must be considered should doses be reduced to lower these seemingly high TRL concentrations.
|
Introduction |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Therapeutic monitoring of TRL by microparticle enzyme immunoassay (MEIA) has been the mainstay for monitoring of this agent for the last decade. However, the MEIA has a relatively high limit of detection (LOD) (4) and shows particularly poor precision at lower TRL concentrations with a lower limit of quantification (LLOQ) of 3.1 µg/L (5). This has led to a report (6) suggesting caution in the interpretation of TRL concentrations <9 µg/L. Certainly, accurate and precise measurement below 5.0 µg/L is problematic, and the prevalence of low concentrations in stable liver graft recipients has increased because of wider use of TRL-sparing regimens (e.g., to 38.6% of our results in adult liver outpatients in the last 12 months). Recent developments in liquid chromatography–tandem mass spectrometry (LC-MS/MS) technology may address this need, delivering assays for TRL with rapid turnaround times and improvements in accuracy, precision, and sensitivity in the lower therapeutic range.
Several reports in the literature have described LC-MS/MS methods for measurement of TRL in whole blood and compared them with MEIA (7)(8)(9). These studies showed a 10–18% difference for MEIA values with respect to LC-MS/MS, with no examination of the possible causes of this difference. The present study compares an established LC-MS/MS method (9) with MEIA in a large number of blood samples from a diverse range of patients. It attempts to explain the differences between the two methods in relation to the hematocrit, which has been noted to adversely affect the extraction of TRL (for MEIA) from whole blood (10)(11) and markers of hepatic and renal function (thought to modulate TRL clearance).
|
Materials and Methods |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
-glutamyltranspeptidase, phosphate, and creatinine results were collected because these have either known or hypothesized relationships to TRL elimination or binding in blood. The numbers of patients and samples in each group, the details of the two phases in which the comparisons were performed, and the summary statistics on the routine hematologic and biochemical data are given in Tables 1
and 2
, respectively, of the Data Supplement that accompanies the online version of this article at http://www.clinchem.org/content/vol51/issue3/.
|
|
analytical methods
The MEIA was performed on an IMx analyzer according to the manufacturer’s instructions (Tacrolimus MEIA-II; Abbott Diagnostics). The LC-MS/MS assay was performed with a Waters 2795 HPLC system and a Micromass Quattro Micro mass spectrometer (Waters Ltd.) according to the method of Keevil et al. (9). Calibrators were prepared from blank, calibrator, and quality-control pools of whole blood supplied by ChromSystems to provide an eight-point calibration curve with concentrations of 0, 0.68, 3.4, 6.7, 13.4, 18.0, 27.0, and 36.0 µg/L. The full curve was run daily and used for all subsequent analyses. Quality-control samples were provided by Recipe and Abbott Diagnostics. In addition, samples from the International Proficiency Testing Scheme for TRL (PT) (12) were analyzed and the MEIA TRL calibrators were also analyzed by LC-MS/MS to further assess assay performance. The LOD of the assays were calculated as 3 SD of 10 replicates of the zero calibrator, and the LLOQ was calculated as the lowest concentration at which the interassay variability and accuracy of the assay were 
20%.
Patient samples (derived from the sources shown in Table 1 of the online Data Supplement) were analyzed by MEIA and LC-MS/MS within 3 days of each other. To evaluate whether analytical error underlay a marked discrepancy in assay results, we selected several patient samples at random and reanalyzed them by MEIA (n = 92) and LC-MS/MS (n = 142) for separate determinations of interassay variance. Pooled patient samples used to assess intra- and interassay precision were prepared by collecting samples with TRL concentrations (MEIA) of <4.0, 4.0–8.0, 8.0–12.0, and 12.0–20.0 µg/L to produce total volumes of 30 mL each. The concentration of TRL in each pool was confirmed by MEIA.
statistics
Data were analyzed by use of Analyze-It for Excel (Ver. 1.48; Smart Software) for method comparisons and by SPSS (Ver. 11.5; SPSS Inc) for comparison of data. Of the 1156 samples analyzed by both methods, 58 (5.0%) had TRL concentrations <2.0 µg/L (the cutoff value routinely used in this laboratory) when measured by MEIA and were excluded from the statistical analyses.
|
Results |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Analysis of TRL-enriched samples issued by the PT scheme gave a mean (SD) recovery of 104.5 (8.6)% by LC-MS/MS, but the MEIA showed significantly higher recovery (P = 0.001) of 135.3 (16.4)%. The MEIA-II calibrators assayed by LC-MS/MS gave a median (range) recovery of 102 (95–105)%. Analysis of PT scheme patient pooled samples showed no significant difference between the methods for this group of samples. None of the results for the samples was outside ±2 SD of the respective group mean values calculated by the PT scheme organizers.
Interassay precision data obtained with quality-control and pooled patient samples (Table 1
) showed higher precision with the LC-MS/MS. Again, duplicate analyses showed that the LC-MS/MS was significantly more precise (P <0.001, Mann–Whitney) with a variance of 4.1(0–38)% vs 13 (0–58)%.
differences between meia and lc-ms/ms assays
For the complete sample cohort (n = 1098), MEIA showed a median (range) difference of 16.2% (–110.3% to 200.0%) relative to LC-MS/MS. Analysis of the data by Passing and Bablok regression gave a slope of 1.134 (95% confidence interval, 1.104–1.167) and an intercept of 0.291 (0.067–0.448) µg/L.
The difference between methods varied among the patient groups (P <0.001, one-way ANOVA) as shown in Table 2
. Post hoc analysis (Tukey) showed a smaller difference for the ALO group than the ALI and BM groups (P <0.001), a smaller difference for the PLO than the BM or PLI cohorts (P <0.001 and 0.02, respectively), and a smaller difference for the renal transplant than the ALI and BM groups (P <0.001). Difference plots for adult liver, pediatric liver, and renal recipients and for bone marrow graft/hematology patients are shown in panels A through D, respectively, of Fig. 1
. These data suggest an effect of the time post transplantation and the organ engrafted on the difference between methods.
|
effect of hematologic and biochemical variables on all samples
Multiple stepwise linear regression with intermethod difference as the dependent variable gave an optimal relationship of PCV, albumin, creatinine, and phosphate as predictor variables (P <0.001, <0.001, <0.039, and <0.042, respectively) with the adjusted r2 being 0.343 (P <0.001). Bivariate analysis showed that the difference between methods correlated inversely with the PCV (Spearman r = –0.53; P <0.001), and stratification of the PCV results (Fig. 2
) showed an increasingly large difference at a PCV of 
0.360. Similarly, for plasma albumin concentrations, the difference correlated inversely with plasma albumin (Spearman r = –0.46; P <0.001), increasing even within the reference interval (Fig. 3
).
|
|
There was a significant correlation between the intermethod difference and creatinine in both the adult (r = 0.11; P = 0.015; n = 551) and pediatric groups (r = 0.146; P = 0.05; n = 185), which were subdivided to control for the effect of body mass on serum creatinine. However, in the adults the slope of the correlation line approached 0, whereas in the pediatric group, 10 of the 11 samples with a high creatinine (>100 µmol/L) came from one patient. There was no significant bivariate correlation between intermethod difference and plasma phosphate (Spearman r = –0.03; P = 0.5). These data, coupled with the significance of the effects in the regression analysis, suggest weak effects overall of creatinine and phosphate on intermethod differences.
effect of hematologic and biochemical variables in each patient group
Biochemical and hematologic data (Table 2 of the online Data Supplement) differed significantly between the discrete patient cohorts (P <0.001, one-way ANOVA for all variables), justifying separate analysis of the interassay differences in each by multiple linear regression. The negative effect of PCV on interassay differences was maintained in the ALO (P <0.001; n = 161), ALI (P <0.001; n = 206), and PLO (P = 0.017; n = 53) groups. The negative effect of albumin on interassay differences was maintained in the ALI (P = 0.008) and in the miscellaneous (P = 0.018; n = 11) groups. In the PLI group, there was a positive correlation between plasma bilirubin and interassay differences (P <0.001; n = 128), although 20 of these data points were from two patients with bilirubin concentrations >250 µmol/L. Therefore, the relationship between interassay differences and increases in bilirubin in the pediatric cohort should be interpreted cautiously. There were no significant effects of any serologic marker in the BM, renal transplant, or pharmacokinetic profile sample cohorts.
|
Discussion |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Low PCV values are known to increase the concentration of TRL measured by MEIA, possibly by affecting the extraction of TRL from blood cells (10) or by a cross-reaction of non-TRL components in the extracted sample and the anti-tacrolimus antibody used in the MEIA (11)(15). Additional explanations may be an increasing hepatic extraction of TRL from blood with decreasing PCV (16) or with low concentrations of TRL-binding proteins in blood (e.g., 
1-acid glycoprotein and albumin). However, albumin in buffer had no effect on hepatic TRL extraction in studies of perfused rabbit livers (16). The net effect of this proposed increase in extraction efficiency will be an increase in TRL metabolism, with likely increases in metabolite concentrations in blood, particularly in cholestasis. This is suggested by the effect of high bilirubin concentrations on the interassay difference in the PLI group and may be masked in the ALI group by the use of artificial liver support, which decreases the concentrations of bilirubin (and other compounds) in plasma. TRL is tightly bound to blood cells and plasma proteins (17), and if the PCV in particular is decreased (16), the unbound fraction of TRL will increase. The likely consequences will be a marked effect on extraction efficiency, both in vivo (affecting metabolism) and in vitro (affecting assay extraction efficiency), and a disproportionate increase in TRL and its metabolites.
A relationship between TRL metabolites and the difference between MEIA and LC-MS/MS results has been proposed in heart and lung (7)(9) and liver and kidney recipients (8)(13)(14). TRL has at least 15 metabolites (18), but only 1 of these, M2 (31-O-demethyl tacrolimus), shows substantial immunosuppressive activity (19)(20) comparable to the parent compound. The MEIA has been shown to cross-react with TRL metabolites M2, M3, and M5 at cross-reactivities of 54–67% of that of TRL (Abbott MEIA II assay insert). Although there are few studies of TRL metabolites in blood in a wide range of patients in the immediate posttransplantation or long-term stable populations (8)(21)(22)(23)(24), there does appear to be an increase in metabolite concentrations during periods of impaired liver function (25) and after 2 weeks post transplantation (21). The latter effect is confirmed by an increase in interassay differences from 6% to 40% between days 6 and 10 post transplantation, with no corresponding marked change in PCV or albumin concentration observed in one patient in the ALI group with intensive posttransplantation monitoring.
All of these data are consistent with the smaller interassay difference observed in the outpatient groups characterized by relatively higher PCVs and plasma albumin concentrations (Table 2 of the online Data Supplement). There may be an additional contribution from a decrease in the total amount of CYP3A-inducing medication taken by these patients (e.g., steroids) compared with the inpatient population. An alternative explanation might be ion suppression of TRL or the internal standard in the LC-MS/MS assay. Such suppression may be greatest in patients with poor hepatic or renal function because of the concentrations of potentially interfering compounds. However, our own, albeit limited, assessment of ion suppression and the known inaccuracy of the MEIA with enriched samples suggest this is not the case, as do the parallel observations of comparable differences in at least five other studies using dissimilar LC-MS/MS methodologies (7)(8)(9)(13)(14).
Further resolution of the basis of these differences between LC-MS/MS and MEIA results will require more detailed studies on the effect of the PCV and plasma bilirubin and albumin concentrations on TRL metabolite concentrations, extraction efficiency, and ion suppression. The effects of these variables on interassay differences for patients receiving TRL for indications other than liver transplantation also require investigation, given the low percentage of samples in the BM group with biochemical and hematologic data and the narrow range of PCV values in the renal group (Table 2 of the online Data Supplement).
A final consideration must be differences between the magnitude of the difference in this and other published series, reported as 10–18% (7)(8)(9)(13)(14), compared with the patient samples circulated in the Tacrolimus International Proficiency Testing Scheme (median 6%). A possible explanation is that metabolite concentrations in the stored pooled samples (median difference, 6.9% and 5.6% in liver and renal pools, respectively; n 
10) are lower than those found in fresh samples from corresponding patients (median of 7.5% and 15.8%, respectively, in this study) because of instability. Another possibility is that the PT pools are prepared largely from outpatient samples and therefore have near-normal PCV and albumin values. The additional possibility, that MEIA routinely overestimates TRL concentrations, is suggested from consistent overestimates by the assay of TRL concentrations in enriched PT samples, and we have confirmed this with parallel observations using independent quality-control samples. However, analysis by LC-MS/MS of the MEIA calibrators supplied by Abbott gave recoveries between 95% and 105%; thus, the complexity of these findings requires further clarification.
We found the LC-MS/MS assay to be significantly more precise than the MEIA over the clinically relevant concentration range, with CVs ranging from 4.7% to 15% for the LC-MS/MS compared with 9.1–28% for the MEIA (Table 1
). This confirms the findings of Ghoshal and Soldin (6), which led them to suggest that TRL concentrations <9.0 µg/L should be treated with caution. In our series of 989 samples monitored routinely (median MEIA TRL concentration = 6.6 µg/L; interquartile range, 4.2–9.3 µg/L), 72.7% had concentrations below this value, and this high proportion emphasizes the reason for concern in relation to current therapeutic and management practices.
In summary, we have reported that the MEIA-II for TRL gives results that are a median of 16.2% higher than those obtained by LC-MS/MS, confirming several reports in the literature. We have been able to demonstrate that this difference is related to the PCV and plasma albumin, with decreases in both variables increasing the difference between the assays. In patients with normal or near-normal PCV and plasma albumin concentrations within the reference interval, the customary small difference between the results (<7%) is unlikely to affect clinical management. However, patients with marked decreases in PCV and/or plasma albumin will show proportionately lower TRL concentrations when samples are measured by LC-MS/MS, and this may also apply during severe cholestatic episodes. The precise cause of this larger difference is unclear at present, but the increased values seen with the MEIA could lead to the false impression of adequate immunosuppression in patients with a low PCV or albumin (attributable, for example, to a poorly functioning or rejected liver graft or graft-vs-host disease in bone marrow transplant patients). The introduction of LC-MS/MS to this laboratory for measurement of TRL has led to a marked improvement in assay performance. It has enabled the delivery of TRL concentrations, seemingly without interference from metabolites and with increased sensitivity and precision, in a timely manner. An audit of effective therapeutic ranges for TRL in the transplant populations is now required to evaluate the clinical impact of this improved analytical technique.
|
Footnotes |
|---|
|
References |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
The following articles in journals at HighWire Press have cited this article:
|
|
 |
|
  R. Borrows, G. Chusney, M. Loucaidou, A. James, S. Goel, S. Borrows, J. Van Tromp, T. Cairns, M. Griffith, N. Hakim, et al. Analysis of Factors Influencing Tacrolimus Levels and Immunoassay Bias in Renal Transplantation J. Clin. Pharmacol., August 1, 2007; 47(8): 1035 - 1042. [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
) and the difference for the outpatient group by the solid line (with individual points as 
). The median differences between methods for the in- and outpatients are 31.7% and 7.7%, respectively, in adults and 14% and 6.1%, respectively, in the children. In C and D, the median difference is indicated by the dashed line and is 32.7% and 16.0%, respectively.
