Multicenter Laboratory Comparison of Iohexol Measurement

Abstract

Background:

Iohexol is utilized for measurement of kidney glomerular filtration rate (GFR). Until recently, there have not been available proficiency standards to assist in calibrating a laboratory’s results. In view of a shift in calibration at the University of Rochester Medical Center (URMC) laboratory, serving as Central Biochemistry Laboratory for the CKiD study, we performed a multi-centered laboratory comparison.

Methods:

Two batches of 30 fortified sera and patient samples from serum or heparinized plasma were sent for duplicate analysis to URMC, University of Minnesota (UMN), Mayo Clinic, and University of Lund. Five proficiency testing materials from Equalis AB were also provided. Iohexol calibration was performed using dilutions of Omnipaque™ 300 and concentrations measured by HPLC or LC-MS/MS (Mayo).

Results:

UMN and Lund agreed well. URMC calibration was 11–13% lower, and Mayo was 4–8% lower for fortified samples. URMC corrected calibration was 3–8% higher for these samples. When measured values were adjusted for the results of the Equalis samples, all laboratories agreed within 1–2% on all iohexol concentrations.

Conclusions:

For 12 URMC calibrator lots from 11/ 2006 to 3/ 2016, the factor quantifying the underestimation of measured to true iohexol concentration was 0.89. If each concentration were divided by 0.89, the calculated GFRs would be reduced by 10–11%. GFR results for CKiD were adjusted for this shift in calibration. Regular examination of iohexol proficiency testing materials, free exchange of samples among laboratories, and standardized dilution of the stock iohexol for calibration would help to bring more universal agreement to this assay.

Introduction

The NIH NIDDK-funded Chronic Kidney Disease in Children Study (CKiD) was designed to measure glomerular filtration rate (GFR) at regular intervals in pediatric subjects with chronic kidney disease (CKD): years 1 and 2, and then every two years. GFR was estimated using newly generated GFR estimating equations (1) during the same visits and at visits at which there was no GFR measured. In this way renal progression could be carefully monitored for each of the subjects. At the beginning of the CKiD study, its Central Biochemistry Laboratory (CBL) at the University of Rochester Medical Center (URMC) was asked to choose an optimal method for measuring GFR in children. Because of the value of repeated measurements, it was felt that the GFR marker should be non-radioactive. In addition, because many of the children were expected to have urological and bladder dysfunctional issues related to their CKD, it was recommended that a plasma disappearance marker be used instead of urinary clearance of the exogenous filtration agent. Inulin, the original GFR gold standard, has been difficult or impossible to obtain in the USA, is expensive when available, and difficult to measure accurately. The CBL chose to utilize iohexol, a widely used, non-ionic, very safe radio-contrast material also known as Omnipaque™, which was supplied to CKiD by GE Healthcare. This exogenous glomerular filtration agent provides an accurate and reproducible measurement of GFR determined from the iohexol dose divided by the area under the plasma disappearance curve (2–4). To make a GFR measurement, a known amount of iohexol is administered as a single bolus intravenously and blood is collected for serum or plasma iohexol concentration measurement at specific time intervals over the ensuing 24 hours. For patient and staff convenience reasons, the time for blood collections is often performed over 2 to 8 hours after the initial dosing, Longer times are required for patients with very low GFR and therefore very slow serum or plasma clearance of the iohexol. The rate of iohexol elimination is proportional to the GFR, which is calculated using formulas that require input of the initial dose concentration (5–9) . Thus, if a calibration discrepancy between the analytical method and the dosing protocol exists, this bias will be directly reflected in the calculated GFR measurement. To summarize this effect clinically, lower values would artificially elevate the calculated GFR and higher values would artificially decrease the calculated GFR.

Most laboratories measuring iohexol use dilutions of Omnipaque™ 300, which contains 647 g/L of iohexol equivalent to 300 g/L of organic iodine, to calibrate their measurement procedures. In late 2014 the URMC Laboratory undertook a revision of their iohexol measurement procedure adding iopamadol as an internal recovery standard. When iohexol from alternate sources (Fluka and European Pharmacopoeia) was obtained and prepared at specific concentrations, it became apparent that the currently assigned calibrator values underestimated the targeted iohexol concentrations. When the calibrators were restored to their theoretical assigned values, the immediate effect was that the iohexol measured concentrations rose about 12% compared to previously measured concentrations. This caused the measured GFRs to be about 10–11% lower than previous determinations. Because of this significant reduction in GFR of CKiD subjects, we organized an inter-laboratory comparison study with other established laboratories for native patient samples and targeted iohexol samples. The results of this study would allow us to validate the new calibrator values and to discern if previous calibrations needed adjustment based on the new data. We also examined if an external proficiency testing program might serve to bring the lab assays into better alignment across laboratories making GFR measurements using iohexol as an exogenous filtration agent in the US and Europe.

Methods

Laboratories and Samples:

Laboratories were selected for the comparison study based on their history of success and accuracy in measuring iohexol. Most of the important early publications concerning the use of iohexol came from the University of Lund, Sweden (Lund, in the Figures and Tables) (10, 11) where iohexol was first developed as a radiocontrast agent and subsequently for making GFR measurements. An effort was made to also include a facility utilizing mass spectrometry, as most labs currently utilize high performance liquid chromatography (HPLC). The Mayo Clinic (Mayo) was recruited for this purpose. After sending out a survey to clinical laboratories in the USA, we found that only the University of Minnesota Medical Center (UMN) was regularly measuring iohexol concentrations for both clinical and research purposes (12).

Two batches of samples were prepared, coded, and sent at ambient temperature to each laboratory for analysis. Phase 1 comprised 30 samples including 24 fortified samples for which iohexol was spiked into human sera, human heparinized plasma, sheep sera, and phosphate buffered saline (PBS). There were also 6 pooled sera from patients who had been administered iohexol for measuring GFR collected previously and stored frozen (−20° C) for less than 3 months. Phase 2 comprised 35 samples, including 12 sera and 12 simultaneously obtained plasma samples from a total of 3 research study subjects plus 6 pooled sera and plasma from these same subjects. These patient samples were stored for less than a month at −20°C. For further analysis, the fortified samples were grouped into low (~15 mg/L), medium (~75 mg/L), and high (~200 mg/L) concentrations. In addition, there were 5 external quality assurance samples from Equalis AB, a not-for-profit external quality assessment provider based in Sweden. The Equalis samples were pooled patient sera to which iohexol was added to achieve desired concentrations in the range anticipated from human GFR measurement studies. These samples had an assigned (target) value based on the amount of iohexol added plus a consensus mean ±SD value from over 30 different laboratories, primarily from Europe during 2015. The Equalis samples were stored at −80°C and aliquots distributed to the laboratories at ambient temperature.

The labs in our study measured each batch of samples on two separate days. Results were sent electronically to a biostatistician at URMC, who was otherwise not involved in the preparation of test samples or iohexol measurements, and who performed the analyses and prepared the graphs.

Measurement Procedures:

Table 1 shows the laboratories and a comparison of measurement procedures used; three laboratories (Lund, UMN, and URMC) used HPLC with the URMC laboratory making measurements with two different sets of calibrator values (preSep2014 and postSep2014). Mayo Clinic used LC-MS/MS (13). Initial measurement procedures in samples were developed using high performance liquid chromatography (HPLC) (10, 11, 14, 15). Each laboratory performing HPLC utilized the area of the larger of two iohexol peaks occurring in the chromatogram (2, 10, 12). The two peaks appear to come from hindered rotation of the anilide N-acetyl group due to bulky iodine atoms attached to the central benzene ring of the iohexol(16). These two peaks are essentially “rotational isomers” that interchange very slowly at room temperature in aqueous solutions, but rapidly at temperatures above 100°C. Because Omnipaque™ solutions are autoclaved for sterilization, the ratio of the large peak to small peak rapidly equilibrates to approximately 74% and 26%, respectively, and remains at this ratio upon cooling; there is no evidence that Omnipaque™ varies from lot-to-lot. Because there are several chiral carbons in iohexol, there are likely in addition multiple optical isomers. However, most HPLC columns, at least those with non-chiral stationary media, cannot resolve true optical isomers despite the fact that all the columns used by the laboratories performing HPLC measurement procedures can easily resolve the “rotational isomers.” The chromatographic column employed in the Mayo LC-MS/MS measurement procedure does not resolve these “rotational isomer” peaks found in aqueous iohexol solutions (see Table 1). Column chromatography (C18 and/or C8) was performed by all labs, and detection was monitored at 245 or 254 nm absorbance for HPLC and the m/z 821.8–602.8 transition was used by LC-MS/MS. Internal standards were used to assure recovery during sample pretreatment by all participating laboratories except the UMN, specifically Isopaque Cysto-100 for Lund, iopamidol for URMC, and d3-iohexol for Mayo’s LC-MS/MS.

Table 1:

Summary of Iohexol Methodology

Laboratory	Methodological Approach	Specimen	Stock Calibrator	Working Calibrators	Protein Removal Technique	Chromatography Column	Detection	Internal Standard	Peak(s) Quantification
Lund	HPLC	Serum	Omnipaque™ 300 dilute 100-fold in water	Stock calibrator diluted into human serum	50uL sample with 200uL 0.33 mol/L percholric acid	C18 column, citrate/acetonitrile/methanol mobile phase	Abs 245 nm	Isopaque Cysto −100	Ratio of second iohexol peak height to internal standard peak height compared to external calibrators run in the same batch
U of Rochester HPLC Lab	HPLC	Serum	Omnipaque™-300 diluted to 10 g/L in phosphate buffer	Stock calibrator diluted into sheep serum	50uL sample with 200uL 5% percholric acid	C8 column, 20 mM, pH 2.5 phosphate buffer and 4% acetonitrile isocratic	Abs 254 nm	Iopamidol	Ratio of second iohexol peak area to internal standard peak area compared to external calibrators run in the same batch
U of MN	HPLC	Heparinized Plasma	Omnipaque™-300 diluted to 10 g/L in phosphate buffer	Stock calibrator diluted into 0.9% aqueous sodium chloride	Ultrafiltration with 30 kDa MCWO filter	C18 column, sodium acetate buffer w/tetrabutylammonium phosphate mobile phase	Abs 254 nm	None	Height of second iohexol peak compared to peak height of standard from external calibrators run in the same batch
Mayo Clinic	LC-MS/MS	Heparinized Plasma	Omnipaque™-300 diluted to 10 g/L in phosphate buffer	Stock calibrator diluted into reagent-grade water	Acetonitrile supernate	C18 column, formic acid/methanol mobile phase	m/z 821.8-602.8 transition	d3-iohexol	Both isomeric peaks are merged and quantified in one integration event.

All the iohexol procedures employ a protein removal step (see Table 1) either via acid precipitation (Lund, URMC) or 30 kDa molecular weight cut off ultrafiltration (UMN) for HPLC and acetonitrile precipitation for LC-MS/MS (Mayo). The Lund and URMC laboratories normally use sera and the UMN and Mayo use heparinized plasma samples. For this study both sample types were measured by all laboratories. A high concentration stock calibrator was prepared by volumetric dilution of Omnipaque™ 300 into water or PBS. The assigned value of the calibrator for HPLC and MS techniques corresponded to the total amount of iohexol (including both rotational isomers). Working calibrators were prepared by diluting this stock into human serum, sheep serum, 0.9% saline, or water. In the HPLC techniques, the second, larger iohexol peak height or area was compared to an internal standard height or area for quantification against external calibrators run in the same batch of samples (10). For mass spectrometry both iohexol isomers were quantified in one integration event. Calibration to total iohexol concentration provides for uniform quantitative results across all methods.

At URMC, sera were stored refrigerated for less than 4 weeks and analyzed in batches comprised of 40–50 samples; the multiple serum and plasma samples were analyzed on the same HPLC run. Each analytical run consisted of calibration standards of known concentration, positive and negative controls and blanks, repeat injections of previously run patient samples as controls, and patient study samples. Calibration standards were prepared by spiking human or sheep serum with a stock solution of Omnipaque™ 300; the latter was prepared by diluting volumes of Omnipaque™ 300 into water. At URMC, levels of 25, 50, 100, 250, 500, and 1000 mg/L were prepared as calibrators in sheep serum and stored frozen. Quality control samples were similarly prepared in sheep serum at target values of 12, 60, and 600 mg/L. New quality controls were validated by running as unknowns with current calibrators and controls; if the measured values were within 20% of the target values, then the new quality control material lot would be put into use and the measured values assigned to the material. For the calibrators, the new lot was run as unknowns with current calibrators. If the results were within 15% of the target values then the new calibrator lot was put into use with the measured values assigned as calibrator set points. Prior to 2014 the URMC laboratory had only one source of iohexol (Ominipaque™ 300) from which both calibrators and controls were prepared.

Throughout the history of the assay at URMC, each calibrator batch was validated with respect to the one immediately preceding it. Assigning the new calibrator values based on the values measured using the previous calibrator lot was performed to ensure that there is minimal variation in the measured values going from the old to the new lot of calibrators. The quality control measured values for iohexol remained unchanged in transitioning from one lot of calibrators to the next. Thus, values assigned to the new lot of calibrators were verified by measuring the quality control samples whose batches were replenished on a different schedule than for the calibrators. Supplemental Table 1 shows the history of assigned values for each calibrator from 2003 to 2014. It is noteworthy that in introducing the calibrator prep of 11/19/2006, the assigned values decreased 11% across the range, and that the assigned values remained close to these levels for the subsequent 8 calibrator preparations from 2006 to 2014. Thus, there was a persistent discrepancy between the expected calibrator values based on the amount of Omnipaque™ added and the assigned calibrator values that was not identified until alternative preparations of iohexol were examined.

Statistics:

Statistical analyses were performed using SAS 9.4 (SAS Institute Inc., Cary, NC). Descriptive statistics such as mean, median, and standard deviation were used to summarize the samples by sample source (fortified vs pooled patient), level (low, medium and high) and by each lab. The intraclass correlation coefficient (ICC) was calculated to assess the differences between runs at different times within each site. The medians of patient samples were compared among the laboratories; the actual expected sample concentrations were compared with laboratory measurements of the spiked preparations. The numbers of samples that were more than 8% different from the target values were compared among laboratories using Fisher’s Exact Test. The 8% is a quality goal of Equalis from which 83% of results were within ± 8% of the consensus values of the spiked samples used for proficiency testing. Bland Altman difference plots (17) were used to visualize and assess agreement among measurements of spiked samples.

Results

Phase 1:

The intraclass correlation coefficient between two runs exceeded 0.993 for each site, showing excellent reproducibility. Figure 1 shows the difference in the measured and target values for 24 samples fortified with iohexol. The mean differences were <4% for all laboratories except URMC (preSep2014) which was 11% below the target values. An absolute difference >8% occurred in 1 sample for Lund, 0 for UMN, 1 for Mayo and 4 for the postSep2014 calibration at URMC. Examination of the clustered concentrations showed for URMC preSep2104 calibration at low (~15 mg/L) concentrations a mean difference of −5.7% with 1 of 8 results exceeding 8% difference from theoretical compared with 3 of 8 results exceeding 8% for postSep2014 calibration; at medium (~75 mg/L) concentrations the mean difference was −13.0% with 7 of 8 results exceeding 8% difference compared with 1 of 8 for postSep2014 calibration; and at high (~200 mg/L) concentrations the mean difference was −14.0% with 8 of 8 samples exceeding 8% difference compared with 0 of 8 for postSep2014 calibration.

1.

Phase 1 comparison of measured vs. target values for iohexol in fortified samples. Plot symbols for the various matrices are as listed below the figure. The solid line is the mean difference for all samples and the dashed lines denote the 95 percentile limits of agreement for the mean values.

With regard to the 6 pooled patient samples, Lund, UMN and Mayo had median concentrations of 91.4, 88.8, and 90.9 mg/L, respectively (Figure 2). URMC with preSep2014 calibration had a lower median of 79.4 mg/L. The revised URMC postSep2014 calibration had a median of 93.1 mg/L, much closer to that of Lund, UMN, and Mayo.

2.

Median iohexol concentrations in 6 pooled patient samples included in Phase 1.

Phase 2:

Figure 3A shows the median values for 12 individual patient samples (overall concentrations from 25 to 300 mg/L) and 3 pooled samples (from the same concentration range). Median values were similar for serum or plasma from individual patient samples for Lund and UMN with values of 119.3 to 122.4 mg/L, and for URMC preSep2014 calibration and Mayo but with lower values of 107.4 to 109.3 mg/L. URMC postSep2014 calibration had different values from either group with median values of 127.9 and 128.2 for serum and plasma, respectively. In the 12 paired comparisons performed by each site, there was no significant difference between plasma and serum iohexol concentrations, with the mean of serum values averaging 1–2 mg/L greater than plasma means (data not shown). The 3 pooled samples demonstrated a similar pattern except the pooled plasma values were systematically lower than the pooled serum values. The same rank order of values was observed at each of the 4 time points for sample collection for each laboratory (Supplemental Figure 1A).

3.

Phase 2 comparison of the median concentrations of 12 patient (light shade) and 3 pooled (dark shade) samples containing iohexol: A) left panel is plasma and right panel is serum samples. B) Same measurements shown in part A corrected using a factor derived from results for the Equalis EQA samples vs. the Equalis target values.

The relationship among the median values for patient samples was also observed for the 5 Equalis EQA samples (Figure 4). Mean values for Lund and UMN were close to the target values, whereas URMC (preSep2014) and Mayo were 7–8% lower and URMC (postSep2014) values were 8% higher than the target values.

4.

Difference between laboratory measured and target values for 5 Equalis EQA samples. The solid line is the mean value for the 5 samples and the dashed lines denote the 95 percentile limits of agreement for the mean values.

In order to reconcile the differences among the laboratories, we examined a regression of each lab’s measured Equalis samples values vs. the target values and also vs. the EQA participants’ consensus values (Supplemental Table 2). The consensus values were based on assays by approximately 30 laboratories with CV <4%. The consensus values were within 2 SDs of the target values based on mass, and differed by less than 2% for 4 of 5 samples and by 4% for one sample. Because the intercepts of these regression values were not significantly different from zero, it was possible to harmonize each measured value for patient and pooled samples from each laboratory by dividing by the observed slopes. Harmonization of results to the Equalis target values brought the median values for patient serum and plasma samples into close agreement with median values 117.0–120.4 mg/L (difference 2.9%) for all laboratories (Figure 3B). Similarly the median values for the pooled samples were aligned (differences were 2.9% for serum and 4.4% for plasma) with the pooled plasma values remaining lower than the pooled serum values. None of the laboratories had any patient or pooled sample results that exceeded 8% of the mean laboratory values after correction based on the Equalis EQA values.

Using data from 12 lots of calibrators used at URMC over the time span from November 2003 to September 2014, the CKiD Data Coordinating Center (led by A. Muñoz), examined the regression data of all the measured concentrations of iohexol corresponding to target (i.e. fixed and known) concentrations of 25 to 1000 mg/L iohexol. The intercepts of these calibrator corrections were insignificant and the slopes of the regressions averaged 0.89. Figure 5A shows the measured concentrations plotted against the target calibrator concentrations; the line goes through the origin and has a slope of 0.885. Residual analysis showed no bias and increasing dispersion with increasing values of the target concentrations (Figure 5B). Mixed effect models to incorporate increasing dispersion and treating lot as a random effect showed no need for an effect of intercept and confirmed a factor of 0.89 for the slope. Thus, the CKiD iohexol concentrations were shown to be 89% of what they should have been since late 2006. The iohexol concentrations since late 2006 were therefore divided by 0.89 to yield higher values for all studies performed since November 2006 and up to September 2014, when recalibrated methods were implemented in the laboratory. For CKiD as well as other studies using the URMC laboratory, after increasing the concentrations by 1/0.89=1.12, the measured GFR would be reduced by approximately 11%; when other equations were used to calculate GFR solely from later time points using the Brochner-Mortensen (5, 6, 16) or Ng (7) approximations, the decrement in GFR would be slightly smaller.

5.

Calibration of CKiD iohexol concentrations after November 2006. A) Measured original concentrations (Y axis) vs. true concentrations determined from nominal concentrations of calibrators (X axis), showing a slope of 0.89 with no significant intercept. B) Residuals as a function of true concentrations, showing no significant differences from zero when 0.89 x true concentrations is subtracted from the measured iohexol concentrations.

Discussion

This study has shown that in the absence of an external reference material, iohexol concentrations differed by more than 10% from one lab to another. Such differences in concentrations will cause roughly similar proportional differences in measured GFRs, but in the opposite direction from the iohexol concentration measurement bias. While each lab utilized dilutions of GE Healthcare’s Omnipaque™ 300 to prepare calibrators, such calibrators were likely to depend on technical performance (measurement and pipetting) and possibly matrices into which the Omnipaque™ was diluted. The variability between lots of Omnipaque™ appears to be less than 2% based on Equalis proficiency testing results and GE Healthcare’s lot-specific certificates of analysis. Thus, lot-to-lot variation in the concentration of iohexol in Omnipaque™ 300 does not appear to have contributed significantly to the observed inter-laboratory variability. All laboratories in our study used volumetric dilutions of Omnipaque™ 300 in water or buffer to prepare stock solutions to then further dilute into human or animal sera for preparing working calibrators. The matrix did not seem to cause systematic problems with accuracy because there were minimal differences between plasma and serum values obtained from the same subjects at the same time.

The URMC laboratory prepared calibrators and controls from the same lot of Omnipaque™ 300, so some error in preparation may not have been readily detected. Indeed, this appears to have happened in November 2006 when an abrupt shift in assigned values was noted (see Supplemental Table 1). At that time there was no available trueness control reference material to assess the accuracy of a laboratory’s iohexol concentration measurements. When the Equalis EQA program’s control material became available in 2014, it appeared at URMC that there was a consistent discrepancy in measured versus actual iohexol concentrations. This discrepancy could not be explained by insufficient recovery of the aqueous phase of the samples or by changes in the lot of Omnipaque™ 300.

Despite these more recent calibration issues, the GFR estimating formula relating 0.413 x height (cm)/Scr (mg/dL) to iohexol GFR, published by CKiD in 2009 (1) utilized most of the measured GFRs from before the shift in 2006 and has been validated by other laboratories subsequently (18–21).

In order to improve calibration, we recommend a method for preparing calibrators from Omnipaque™ 300 stock. Because the material in the vial is viscous and easily adsorbed to pipette tips, it would be common to underestimate the apparent concentration following dilution. Therefore, weighing the Omnipaque™ 300 is recommended rather than assuming accurate volume delivery from pipettes in order to give a more accurate amount of Omnipaque™ 300 actually transferred to the stock calibrator solution used to prepare the working calibrators. The density of Omnipaque™ 300 at 20°C is 1.345 g/mL and at 25°C is 1.342 g/mL. The stock solution (i.e. 10 g/L) is mixed thoroughly (e.g, with a magnetic stirrer) to obtain a homogenous solution. Serum or plasma can then be spiked with this stock solution to desired target concentrations again preferably using a gravimetric dilution technique. All working calibrators should be mixed thoroughly to obtain homogeneous solutions. We believe this method of preparing calibrators will allow for more consistent calibrations among different laboratories.

It should be noted that if a laboratory utilizes multiple blood sampling including early time collections, accuracy in the range of 200 mg/L is required. However if the laboratory is only utilizing blood sampling in the renal phase (beyond 2 hours after iohexol infusion), then most concentrations encountered will be less than 100 mg/L, in the range of the medium and low concentrations as defined in this report. Some laboratories dilute specimens above 100 mg/L and this dilution process may add to the measurement uncertainty.

The performance of the URMC and Mayo laboratories was improved when their measurement results were adjusted based on the Equalis proficiency testing materials. In general, the laboratories from Lund and UMN had less bias and better reproducibility, at least over the 6 m of our study. Lund subscribes to the Equalis EQA program (22), and the UMN started participating in the Equalis program in late 2015. In addition, these two laboratories have exchanged samples in the past to assure consistency between them. Participation in an EQA/proficiency testing program appears to be the most efficient way to assess and maintain the accuracy of a laboratory’s iohexol measurement. When measuring GFR by plasma disappearance of any molecule, including iohexol, internal and external quality assessment including the use of blind replicates of previously tested patient samples and participation in an EQA/proficiency testing program is particularly important. Neither serum- nor plasma-based certified reference materials are not currently available for iohexol; however the Equalis materials can be ordered from Europe for regular delivery to US laboratories in order to assure consistent iohexol measurement results,

Impact Statement:

Iohexol is an exogenous marker commonly administered for the GFR measurements in both clinical and research study-based renal function testing. Measured GFR in iohexol-based protocols is typically obtained through the monitoring of iohexol plasma disappearance. This manuscript highlights the potential ramifications of analytical bias in plasma concentration measurements in the measured GFR. Given the lack of calibration reference materials for iohexol and the impact analytical bias imposes on a GFR measurements, this paper recommends stringent calibration practice, extensive quality control measures along with inter-laboratory specimen exchange or external proficiency testing to assure accurate quantitation of iohexol in patient samples.