Evaluation of a candidate reference measurement procedure for serum free testosterone based on ultrafiltration and isotope dilution--gas chromatography--mass spectrometry.
Considerable controversy has arisen about the analytical validity, in particular, of single-step nonextraction FTe assays (15-17). A major argument has been that the design of some assays insufficiently accounts for the law of mass action (3,18). Although this problem has been documented in validation studies against a so-called gold standard, such as an analytical method with a valid physicochemical basis [see, for example, Refs. (14,19-21)], these assays have continued to be marketed and used by the endocrine laboratory community. Recent European legislation, as decreed in the EC Directive on In Vitro Diagnostic Medical Devices (98/79/EC) (22) and supporting ISO/CEN standard (23), has provided a new strong incentive to bring analytical validation to the forefront. Indeed, since December 2003, a new commercial assay can obtain the CE label only after the diagnostic manufacturer has extensively documented the validation data. For well-defined analytes, this validation is to be done by split-sample measurements with a SI-traceable reference measurement procedure (RMP) [Note: in current metrologic terminology, the term gold standard is replaced by RMP (23)]. For steroid hormones such as Te, it is state of the art to use isotope dilution-mass spectrometry (ID-MS) as the RMP (24). This is because the theoretical measurement principle guarantees accurate, specific, and matrix-independent measurement and because ID-MS can be calibrated directly with gravimetrically weighed-in primary calibrator material. However, for FTe, ID-MS must be combined with a technique capable of separating free from bound hormone without equilibrium disturbance. In this respect, there is general agreement that ED and/or UF, at least in expert hands, are the best procedures (3).
From this perspective, the aim of our study was to develop such a candidate RMP. We opted for OF to separate free from bound Te and coupled it to ID-gas chromatography-MS (UF-ID-GC /MS) for direct measurement of Te in the ultrafiltrate. As part of the validation process, we performed the proposed measurement procedure in parallel with an indirect ED (iED) and SyD measurement procedure, as established in the laboratories of two of the authors, on a panel of 38 male sera.
Materials and Methods
Te standard material (with a purity >99.5% according to the manufacturer's certificate) was purchased from Sigma-Aldrich. Isotope dilution was performed with [3,41-[sup.13][C.sub.2]]-Te from Cambridge Isotope Laboratories. Working solutions of both unlabeled and labeled Te (concentration, ~2.60 nmol/L) were diluted on a gravimetric basis from ethanolic stock solutions (~1040 nmol/L, pro analysis; Merck). The quality of the other solvents was gradient grade (methanol; Romil), ultrapure (water; 18.2 [omega]; produced with an Elga Elgastat Maxima Analytical water purification system), pro analysis (formic acid; Fluka), or dried (cyclohexane; Merck). The HEPES buffer (52.75 mmol/L), pH 7.4, contained 5.265 g/L NaCl, 0.224 g/L K[H.sub.2]P[O.sub.4], 0.275 g/L MgS[O.sub.4] x 7 [H.sub.2]O, 12.570 g/L HEPES, 0.3 g/L urea, 0.275 g/L Ca[Cl.sub.2] x 2 [H.sub.2]O, 0.9 g/L NaOH, and 0.520 g/L Na[N.sub.3] (all from Sigma-Aldrich). For UF, Millipore Centricon YM-30 and YM-10 OF devices (cutoffs of 30 and 10 kDa, respectively; membrane of regenerated cellulose) were used. In addition, three different OF device types from Vivascience were investigated: the Vivaspin 2 variant with membranes of regenerated cellulose, polyethylenesulfone, or cellulose acetate, all with a 10-kDa cutoff. The membranes in the Centricon devices were horizontally placed, whereas those in the Vivaspin devices were vertically mounted. Solid-phase extraction was done with Waters Oasis HLB 1-mL (30 mg) extraction cartridges. The heptafluorobutyric anhydride derivatization reagent was from Macherey and Nagel. For internal accuracy control, a lyophilized serum sample [with a TTe target value of 19.40 nmol/L, as determined by an IDGC/MS RMP (25)] was purchased from the German Society of Clinical Chemistry and Laboratory Medicine.
Serum was ultrafiltered in a thermostatically controlled, fixed-angle centrifuge (Sorvall RT7 plus with a SL-50RT rotor). The solid-phase extraction cartridges were mounted on a JT Baker SPE-12G Column Processor[R] device. For HPLC, a SpectraSystem P1000 pump (Thermo-Separation Products) with a variable-wavelength ultraviolet detector (Philips) was used. The HPLC was equipped with a Hypersil BDS [C.sub.18] column [150 x 2.1 mm (i.d.); 5 [micro]m bead size; Alltech]. GC-MS was performed with a Finnigan MAT Incos XL mass spectrometer combined with a 5890 Series II GC (Hewlett Packard). The GC fused-silica column was a DB-1, i.e., methylsilicone type [20 m x 0.18 mm (i.d.); 0.4 [micro]m film thickness; J&W Scientific]. The GC was equipped with a programmable temperature vaporizing injection system (Gerstel GmbH). Densities were measured with a Mettler Toledo DA-110M density meter.
UF-ID-GC/MS. To 1 mL of each serum sample, directly pipetted into a Centricon YM-30 OF device, 1.4 mL of HEPES buffer was added. All volumetric steps that would contribute to method accuracy were gravimetrically controlled. During equilibration of the mixture at 37[degrees]C for a minimum of 30 min, the centrifuge was preheated by rotating at 33558 for 1 h until a temperature of 37 [+ or -] 1[degrees]C was reached. The samples were then immediately centrifuged at 24028 for 16 min, which yielded, on average, 800 [micro]L of ultrafiltrate. To the ultrafiltrate, [3,4-[sup.13][C.sub.2]]-Te was added in an amount equivalent to the ultrafiltered endogenous Te to obtain a 1:1 isotope ratio on MS determination. This required that before starting the analysis, the FTe concentrations in the samples had been roughly determined. The mixture was again allowed to equilibrate for 1 h. Meanwhile, the Oasis HLB cartridges were conditioned by wetting consecutively with 1 mL of methanol and water. hninediately after the conditioning step, the ultrafiltrate, acidified with 40 [micro]L of 50 mL/L formic acid, was applied and passed through the cartridge at a flow rate of ~0.4 mL/min. The cartridge was washed with, respectively, l mL of water and 1 mL of a methanol water mixture (60:40 by volume). Te and its [3,4-[sup.13][C.sub.2]]analog were subsequently eluted with 1 mL of a methanol-water mixture (90:10 by volume). The elutee was evaporated to dryness at 50[degrees]C under [N.sub.2]. The residue was redissolved in 50 [micro]L of the HPLC mobile phase and injected on the HPLC column. Isocratic chromatography was performed with a mixture of methanol-water (60:40 by volume) at a flow rate of 0.2 mL/min. The fraction containing Te and [3,4-[sup.13][C.sub.2]]-Te (eluting between 8.50 and 11.50 min) was collected and evaporated at 50[degrees]C under [N.sub.2]. The evaporated residue was redissolved in 20 [micro]L of cyclohexane, and 20 [micro]L of heptafluorobutyric anhydride was then added and allowed to react for 30 min at 70[degrees]C. After evaporation of the reaction mixture at room temperature under [N.sub.2], ~12[micro]L of cyclohexane was added, from which 6 [micro]L (0.12 pmol) was injected in the solvent vent mode (70[degrees]C). After 0.3 min, the split was closed, and the temperature of the injector increased to 280[degrees]C at a rate of 12[degrees]C/s. After 2 min at 280[degrees]C, the injector was cooled to the initial temperature. The helium column head pressure was 12 psi. The GC oven was programmed for the following temperature gradient: hold at 150[degrees]C for 1.3 min, then increase to 300[degrees]C (ramp rate, 20[degrees]C/min) and hold for 3 min. The column was then heated at 325[degrees]C for 8.5 min. Both the derivatived Te and [3,4-[sup.13][C.sub.2]]-Te eluted at 10.45 min. For MS detection in the selected-ion monitoring mode (SIM), the molecular ions of the diheptafluorobutyryl derivatives of both unlabeled and labeled Te were monitored at m/z 680 and 682, respectively. Dwell times were chosen to obtain at least 15 cycles under the peaks.
For calibration, equivalent amounts of Te and [3,41-[sup.13][C.sub.2]]-Te (~0.35 pmol of each) were sampled on a gravimetric basis to obtain a 1:1 isotope ratio. The mixtures were evaporated to dryness and derivatived as described for the processed serum samples. The ID-MS one-point calibration protocol is described in detail elsewhere (26). iED. iED was performed in Membra-Cel[TM] MD 10-14 tubing as described previously (14, 27). Briefly, 125 [micro]L of plasma containing a trace amount of [1,2,6,7-[sup.3]H]-Te, diluted 1:5 with saline, was dialyzed under continuous shaking for 15 h at 37[degrees]C against 5 mL of saline. After equilibrium was reached, 0.5 mL each of diluted plasma and the dialysate was added to separate vials containing 10 mL of liquid scintillation mixture and counted. The free fraction was calculated according to Slaunwhite and Sandberg (28). The purity of the tracer was checked every 4 months, and if necessary, the tracer was purified by paper chromatography.
SyD. SyD was performed in a Dianorm dialysis apparatus with Visking dialysis membrane, as described elsewhere (11). Briefly, 500 [micro]L of each sample was incubated for 30 min at 37[degrees]C with a trace amount of paper-chromatography-purified [1,2,6,7-[sup.3]H]-Te. After incubation, 180 [micro]L of the incubate was pipetted into one compartment of a temperature-equilibrated dialysis cell, and the other compartment was filled with an equal volume of the same serum sample without tracer. Dialysis was performed for 2.5 h at 37[degrees]C. Subsequently, 100 [micro]L of the samples in each compartment was counted by liquid scintillation, and the free fraction was calculated as described previously. Calibration of the dialysis system was performed by assessment of membrane permeability with use of a calibrator serum (normal plasma with 100 nmol/L Te added) in which the FTe fraction had been established by direct ED-ID-GC /M5.
Measurement of the TTe concentration was performed according to our previously published candidate RMP (29). In short, to a certain volume of serum, representing 2.6 nmol of Te, was added an equivalent amount of [3,4-[sup.13][C.sub.2]]-Te, and the mixture was allowed to equilibrate. The sample was then extracted with dichloromethane. After centrifugation, the organic phase was transferred to another vial for evaporation to dryness under [N.sub.2]. The residue was fractionated sequentially by Sephadex[TM] LH-20 chromatography and HPLC. The final collected fraction was evaporated to dryness under [N.sub.2], derivatived, and analyzed by GC-MS as described for FTe.
The measurement procedure was performed according to preset analytical specifications: a maximum total CV ([CV.sub.T]) of 2.0% and maximum systematic deviation from the target of a certified reference material of 0.9% (30).
UF-ID-GC/MS was compared with iED and SyD by split-sample measurement of 38 serum samples. The UF-ID-GC/MS measurement protocol consisted of analyzing each sample in two independent runs, which means that in each run, separate calibration and sample pretreatment were used. In total, nine measurement series were needed to process all samples twice. For iED and SyD, duplicate measurements were performed, but in the same run. This protocol gave a total of two measurement series for iED and six for SyD. For calculation of the FTe concentration from the fractions determined by iED and SyD, the TTe concentrations as obtained by the IDGC/MS candidate RMP were used.
The sera used for the method-comparison study were obtained from single blood donations from 38 apparently healthy males. Sera were purchased from DiaServe Laboratories GmbH. Serum was isolated by centrifugation 1 h after collection. The sera were stored at 4[degrees]C for 3 days, filtered, and fractionated into 1-mL portions in Eppendorf vials. The aliquots were frozen and sent on dry ice to the laboratory in Ghent. The sera were also kept on dry ice during transport from Ghent to the other participating laboratories. Receipt of the samples in the frozen form was confirmed. Until analysis, in each laboratory specimens were stored at -20[degrees]C. All samples were characterized for SHBG (with the SHBG immunoradiometric assay from Orion Diagnostica) and albumin (with N antiserum to human albumin; Dade Behring Inc.).
IMPRECISION AND INTERNAL ACCURACY CONTROL DURING THE METHOD COMPARISON
To control the between-run reproducibility of the OF procedure during the split-sample measurements, internal quality control (IQC) was performed. It consisted of repeated duplicate analysis of two serum samples with FTe concentrations of 138.7 and 431.7 pmol/L, respectively. Because the number of samples analyzed per analytical run surpassed the capacity of the centrifuge (two centrifugal runs were needed for the OF process), we split the duplicates of each serum between the two centrifugal runs. This IQC protocol was performed in a total of 20 runs, including the 9 series needed for completing the method comparison to calculate the [CV.sub.T] and the between- and within-day CVs ([C.sub.dd] and [C.sub.wr], respectively) according to the NCCLS EP-5 protocol (31). For internal control of the ID-GC/MS measurements after UF, a lyophilized IQC sample gravimetrically diluted with 9 g/L NaCl to a concentration of 350.7 pmol/L, was analyzed in duplicate in each run. These results were evaluated in terms of the mean deviation from the target value and the [CV.sub.T]. For the measurements using iED and SyD, the standard operating procedures of the respective laboratories were followed, which included in-house IQC samples.
STATISTICS AND GRAPHICS USED IN THE METHOD COMPARISON STUDY
The results of the method-comparison study are represented graphically as scatter plots (iED or SyD on the x axis, UF-ID-GC/MS candidate RMP on the y axis). Deming regression analysis was performed for the different method pairs, using CBstat software (Ver. 4.3.2; from K.
Linnet, Risskov, Denmark). The 95% prediction interval was calculated as 1.96 times the CVs of the method pairs ([CV.sub.mp]). The latter were calculated with the following formula:
[CV.sub.mp] = [[square root of ([CV.sub.A])/[square root of 2].sup.2] + ([CV.sub.A])/[square root of 2].sup.2]]
In this equation, [CV.sub.A] represents the within-run imprecision of either the iED or SyD procedure and was calculated from the duplicate measurements of the serum samples according to the formula:
[CV.sub.wr] = [square root of [summation][d.sup.2]/2n]/x x 100,
where d represents the difference between the duplicates, n is the number of duplicates, and [bar.x] is the mean of the serum concentrations. The [CV.sub.B] in the equation for [CV.sub.mp] represents the imprecision for the UF-ID-GC/MS candidate RMP and was calculated using the difference between the duplicates performed in two analytical runs.
Results and Discussion
UF-ID-GC/MS CANDIDATE RMP
For the development of our UF-ID-GC/MS candidate RMP, we started from the TTe equivalent (29). Because GC/MS measurement of Te can be performed with a limit of detection of 2 pmol/L (signal-to-noise ratio of 3), we considered this procedure appropriate for quantification of FTe in sera from apparently healthy males (FTe concentration range observed in this study, 106-640 pmol/ L). However, to achieve specifications appropriate for a RMP, the signal-to-noise ratio at which quantification is performed is important. We set the limit at ~25. Whereas for serum FTe concentrations of 100 pmol/L and processing of 0.8 mL of ultrafiltrate, we achieved a ratio of ~50-70, concentrations of 15-20 pmol/L necessitated processing of 1.6 mL of ultrafiltrate to reach our limit, but with increasing imprecision. Because working with a higher volume of ultrafiltrate was not considered practicable, we estimated the limit of quantification of our current procedure at 15-20 pmol/L. The only variant we introduced was substitution of the liquid-liquid extraction and Sephadex chromatography by solid-phase extraction with Oasis HLB cartridges. The reconstructed ion chromatograms after solid-phase extraction and HPLC were as clean and interference-free as obtained with our traditional procedure (29). Nevertheless, we investigated possible interference by other steroid hormones (Table 1). As indicated in Table 1, interference by some steroids could immediately be excluded on theoretical grounds because the relative molecular weights of the possible heptafluorobutyryl derivatives were lower than the m/z values monitored for Te and [13[C.sub.2]]-Te. The other steroid hormones were derivatitived under exactly the same conditions as optimized for Te and subjected to GC/MS analysis in both the scan (20 ng of derivatitived substance injected) and SIM mode (2 ng, except for epitestosterone, for which only 150 pg was injected). As expected, epitestosterone derivatized in exactly the same way. However, the derivative of the Te isomer did not interfere because its relative retention time was 0.93. For (3[beta],17([beta])-5[alpha]-androstan-3,17-diol and (3[beta],17([beta])-androst-5-ene-3,17-diol, we observed from the registered spectra that they also yielded a diheptafluorobutyryl derivative with molecular ions at m/z 684 and 682, respectively. The derivatives eluted in our GC program with relative retention times of 1.02 and 1.00, respectively. Although on this basis the derivative of (3[beta],17([beta])-5[alpha]-androstan-3,17-diol did not interfere, we confirmed this fact by performing SIM at m/z 682 and 680. With respect to the derivative of (3[beta],17[beta])-androst-5-ene-3,17-diol, we observed that, because of the suboptimal derivatization conditions (they were optimized for Te), the signal abundance in SIM for 2 ng was very low in comparison with that for Te. For that reason, we injected decreasing amounts of the derivative and found that at <500 pg the signal-to-noise ratio was already <3. On this basis, we assume that interference after processing of a serum sample containing physiologic concentrations of (3[beta],17([beta])-androst-5-ene-3,17-diol would be very unlikely (32).
IQC of the ID-GC/MS measurement procedure during 20 measurement series showed a mean deviation from the target of -0.5% and an associated [CV.sub.T] of 1.1%, a [C.sub.dd] of 0.4%, and a [CV.sub.wr] of 1.0% (Table 2). From these data we concluded that our slightly modified RMP demonstrated adequate trueness and precision and that no further validation of the ID-GC/MS measurement procedure was necessary.
For separation of TTe from protein-bound Te, we selected OF on the basis of the speed and ease of performance. Devices with five different membrane types and/or size cutoffs were evaluated (Table 3). Selected criteria were the degree of protein leakage, determined with the Coomassie[R] Plus Protein assay from Pierce in the ultrafiltrate from 1 mL of serum (n = 8), and adsorption of Te, estimated from recovery experiments after OF of a 9 g/L NaCl solution to which 433 pmol/L Te had been added (n = 5) (33). We defined the limit for protein leakage on the basis of the error the leakage would induce in the FTe measurements, with 1% error considered allowable. Assuming a total protein concentration of 70 g/L and a mean FTe concentration of 2.6% (34), we allowed a maximum of 0.025% protein leakage. As can be seen from Table 3, all tested ultrafiltration devices fulfilled this protein leakage criterion.
With respect to the adsorption studies, the results showed best performance by the Centricon YM-30 device, which was in perfect agreement with the recommendations on OF for free thyroid hormones (33, 35). With respect to the 2% adsorption of Te observed from the saline solution with this device (Table 3), we tested whether the adsorption took place by contact with the serum or with the ultrafiltrate. In the first case, this would not influence the FTe measurement results because of the thermodynamic equilibrium between TTe and FTe. These investigations were done by adding tritiated Te to undiluted and 1:1.5-diluted serum. (Although small dilutions reduce the TTe concentration, they have little effect on the FTe concentration.) This experiment showed a fourfold higher adsorption from undiluted serum, whereas less ultrafiltrate was generated (550 [micro]L) compared with the diluted sample (700 [micro]L). From this we concluded that adsorption of Te indeed was attributable to contact with the serum. Although it was not possible to monitor each OF device for protein leakage or adsorption, we verified by use of random controls with repeated analysis of two human sera in duplicate whether there was sufficient evidence for the consistency of the OF process in terms of absence of protein leakage and adsorption (see discussion of IQC results below). These IQC measurements also allowed us to study the lot-to-lot variation of the OF devices, as has been described for free thyroxine (33).
To avoid disturbing the equilibrium between bound and free Te, we paid attention to the temperature and pH, which are known to influence OF (4). However, the risk of equilibrium disturbance also depends on the duration of the OF process, which in turn depends on the viscosity of the serum and thus on the dilution and the OF yield. The Donnan effect (36, 37), which also depends on the initial sample volume and pH, should be kept as low as possible. From this perspective, we performed several experiments to optimize the OF process. With respect to temperature, OF should be done at 37[degrees]C; however, even with a thermostatically controlled centrifuge, it is necessary to ensure that this temperature is reached from the beginning of the centrifugal process onward because it is in the initial phase that the yield of ultrafiltrate is the greatest. We observed that we had to preheat the empty centrifuge by rotating the rotor at 33558 for ~1 h to obtain a temperature of 37 [+ or -] 1[degrees]C. Under these conditions, we confirmed the temperature-related difference (4, 8, 9); there was an increase in FTe concentration of ~2.8% for every 1[degrees]C that the temperature was closer to 37[degrees]C (P <1.32 x [10.sup.-10]). Because of the effect of pH, it was necessary to ensure that before UF, the serum was at a physiologic pH of 7.4. We used HEPES buffer for this purpose, either by adding 50 [micro]L of a concentrated buffer solution (185 g/L HEPES, 8.8 g/L NaOH) to 1.0 mL of serum before OF or by diluting the serum with the 52.64 mmol/L HEPES buffer as described earlier. The latter way of adjusting the pH was considered best because of our experience indicating that diluted serum needs shorter OF times with a higher OF yield. To test the maximum allowable dilution without alteration of the FTe concentration (9, 38, 39), we compared three different dilutions with undiluted serum. As can be seen from Table 4 (values under "dilution of serum with buffer"), there was a significant difference, amounting to 13.7%, only after addition of 9 mL of buffer (dilution 1:10). This dilution effect agrees with predictions that can be made on the basis of the law of mass action. We therefore decided to perform all further experiments and the final OF procedure using the 1:2.4 dilution with the 52.75 mmol/L HEPES buffer. With respect to the Donnan effect, we investigated the relationship between the volume of sample subjected to the OF process and the yield of ultrafiltrate. As can be seen from Table 4, on processing of 1.0, 1.7, and 2.4 mL of diluted (1:2.4) serum and generating an ultrafiltrate yield of ~33%, we found no significant difference in FTe concentration. The data in Table 4 indicate that a yield of ~17% is too low, whereas a yield of ~33% is sufficient. On the basis of these experiments, we opted to use a sample volume of 2.4 mL (1 mL of serum + 1.4 mL of buffer) for UF, which would generate ~0.8 mL of ultrafiltrate and would give results above the limit of detection of our ID-GC/MS procedure.
With respect to the precision of the UF-ID-GC/MS candidate RMP, the IQC results according to the NCCLS EP5-A protocol (Table 2) demonstrate an acceptable precision in terms of [CV.sub.wr], [C.sub.dd], and [CV.sub.T]. These CVs also include the variability of the OF process and, thus, give sufficient evidence for its reproducibility and the absence of lot-to-lot variation in the OF devices (33). For the four different lots used to date, we have not observed any unacceptable variation. However, because the latter is a risk (33), we recommend validating the performance of each new lot before use with at least two IQC samples with known target concentrations. For comparison, Table 2 shows the [CV.sub.T] values for iED and SyD. The data are slightly in favor of UF-ID-GC/M5. The same applies for the CV values calculated from the duplicate results of the serum samples, i.e., 3.3% vs 5.6% and 7.6% for iED and SyD, respectively. The mean [CV.sub.T] as calculated for UF-IDGC/MS (3.7%) was used to calculate the acceptance criterion during the method-comparison study, namely, to predict the maximum difference between duplicates determined on two separate occasions, i.e., (1.96 x [square root of 2] x [CV.sub.T]), or 10.3%. The largest difference we observed was 6.8%; therefore, no UF-ID-GC/MS measurements for any of the 38 serum samples had to be repeated.
To validate the newly developed UF-ID-GC/MS candidate RMP, we performed split-sample measurement of 38 sera against iED and SyD. Because iED and SyD are indirect and measure the FTe fractions (%), the corresponding concentration was calculated by multiplication of the TTe concentration. The laboratories in this study determine TTe concentrations with a RIA. However, in view of the fact that in this study the procedures for estimating the FTe fraction were of interest, rather than differences in TTe assays, as discussed recently in the literature (40, 41), we used the TTe concentrations as determined with our SI-traceable TTe candidate RMP (29). These concentrations ranged from 7.6 to 47.2 nmol / L, with a mean of 16.7 nmol / L.
Scatter plots of the FTe results obtained by OF-IDGC/MS vs those obtained by iED and SyD are shown in Fig. 1. The plots also include the line of equality, the Deming regression line, and the 95% prediction interval. The corresponding outcomes of the Deming regression analysis are summarized in Table 5. The serum FTe concentrations measured by UF-ID-GC/MS ranged from 106 to 640 pmol/L, with a mean of 299.4 pmol/L, and corresponding FTe fractions of 1.1-2.8% (mean, 1.8%) of the TTe. For iED and SyD, the FTe fractions ranged from 1.3% to 3.3% (mean, 2.2%) and from 1.2% to 4.4% (mean, 1.8%), respectively, giving FTe concentrations of 128-660 pmol/L (mean, 360.4 pmol/L) and 122-750 pmol/L (mean, 301.9 pmol/L), respectively.
As shown in Table 5, UF-ID-GC/MS correlates fairly well with both iED and SyD. However, there is a proportional difference between OF-ID-GC /MS and iED or SyD, with a negative intercept for iED and a positive intercept for SyD, both significantly different from zero. From these data, one can infer that iED and SyD also correlate well, but with a constant systematic difference. A comparison of the [S.sub.y|X values shows that they are highest for the comparison of UF-ID-GC/MS with iED. Although most of the spread around the regression line can be explained by the [CV.sub.T] (most data are within the 95% prediction limits), these observations are presumptive that iED might be most susceptible to individual sample matrix-related effects. Further visual inspection of all scatter plots shows that in Fig. 1B, the 95% prediction interval best includes the line of equality with almost symmetric location of the data pairs along that line. This indicates that our OF-IDGC/MS candidate RMP agrees best with SyD. However, there are more data pairs outside the prediction limits than would be statistically expected, which could be the result of some degree of sample-related effects in one of the measurement procedures.
[FIGURE 1 OMITTED]
The method comparisons show that in the absence of an absolute gold standard, it is not possible to conclude which of the three procedures gives the "true" FTe concentration. Therefore, it might be appropriate here to consider the degree of independence of the three procedures, which is indeed a hallmark of a SI-traceable RMP. With respect to our candidate UF-ID-GC/MS RMP, we showed that direct measurement of the FTe concentrations in ultrafiltrates is feasible. This contrasts with iED and SyD, which cannot be claimed as RMPs because of the indirect measurement principle. The problem with UF, however, is the uncertainty as to whether the measured concentration is the same as the FTe concentration in the original sample. Technical problems such as protein leakage, unequal distribution of FTe resulting from the Donnan effect, and failure to keep the temperature at 37[degrees]C can be ruled out for UF, as shown from the IQC data. A potential problem lies in the fact that during UF, unfilterable constituents of the serum become concentrated, including those that may possess positive or negative cooperative effects toward the interaction between Te and its binding proteins. The finding that moderately diluting serum with buffer has virtually no effect on the measured FTe demonstrates that, if present, such effects are probably quantitatively unimportant. The nature of the dilution buffer might be crucial in this respect. If we consider the measurement principle of iED, where the tracer that is present at equilibrium in the dialysate is determined relative to the total amount added to the sample, the main effect on accuracy is tracer impurity. However, this can be minimized by dilution, which at the same time will reduce the Donnan effect to negligible proportions. Dilution must be taken into account when estimating the FTe fraction for undiluted serum because this quantity must be multiplied by an independently measured TTe concentration to obtain an estimate of the FTe concentration. Theoretically, the particular proportions of Te, SHBG, and albumin and their affinities in serum allow for an accurate result when optimal algorithms are used to correct even for high dilutions. Nevertheless, the same restrictions as those mentioned above for OF with regard to the absence of undialyzable components affecting Te-protein interaction and buffer composition apply. Lastly, SyD is not affected by undialyzed substances and the Donnan effect discussed above for OF and ED because the same sample is present on both sides of the membrane and no physical separation of free and bound Te takes place. Therefore, the FTe concentration in the original sample is not altered in SyD. The effect of tracer impurities is also so small that dilution is not necessary. However, to translate the dialysis rate into the free fraction, SyD needs to be calibrated by either direct OF or ED. This is probably the biggest hurdle to the trueness of SyD measurements because an error in the assignment of the free fraction for the calibrator serum propagates as a systematic error in all results. Therefore, to evaluate which of the three procedures most closely approximates the ideal, one has to balance the uncertainties inherent to each of them. In UF-ID-GC/M5, the only uncertainty remaining is independence of dilution as an indicator of the absence of effects of unfilterable substances. Although demonstrated for the present candidate RMP on 38 sera from apparently healthy males, this needs extension to different kind of samples. The uncertain factor in iED is the extent to which the necessary correction for dilution follows the theoretical model for a larger variety of samples. For SyD, although the above uncertainties do not exist, the uncertainty lies in the calibration. However, in this study, the validity of the calibration process was ensured assigning the value by use of a generally accepted RMP (ID-GC/MS) (24).
In summary, the present study shows that we were able to develop a state-of-the-art candidate RMP based on OF in combination with ID-GC /MS for direct measurement of FTe in sera from males. When we challenged the OF-ID-GC/MS candidate RMP in a method-comparison study with 38 male sera against two indirect "gold standard" methods, UF-ID-GC/MS showed good correlation and acceptable agreement between results. Nevertheless, small systematic differences and sample-related effects were observed. From these observations and theoretical considerations about the independence of the measurement procedures used, it is clear that even the best available and scrupulously performed measurement procedures have technical and fundamental limitations and that, consequently, the scientific community will have to accept that there will remain a degree of arbitrariness about the best way to measure free hormone concentrations. It will be up to authoritative organizations such as the IFCC (43) and the Joint Committee on Traceability in Laboratory Medicine (44) to decide on which measurement procedure should be used by the in vitro diagnostics industry in establishing metrologic traceability of routine assays for serum free hormones. Last but not least, the next challenge for our laboratory will be optimization of our candidate RMP to achieve sufficient sensitivity for measurement of FTe concentrations in females with androgen deficiencies.
We gratefully acknowledge financial support by the Fund for Scientific Research-Flanders G.0038.01. References
(1.) Anderson DC. Sex-hormone-binding globulin [Review]. Clin Endocrinol 1974;3:69-96.
(2.) Mendel CM. The free hormone hypothesis: a physiologically based mathematical model [Review]. Endocr Rev 1989;10:232-74.
(3.) Ekins R. Measurement of free hormones in blood. Endocr Rev 1990;11:5-46.
(4.) Obmifiski Z. Changes in the free (unbound) fraction of testosterone in serum in vitro as affected by pH and temperature. Exp Clin Endocrinol Diabetes 1998;106:85-8.
(5.) Kley HK, Bartmann E, Kruskemper HL. A simple and rapid method to measure non-protein-bound fractions of cortisol, testosterone and oestradiol by equilibrium dialysis: comparison with centrifugal filtration. Acta Endocrinol 1977;85:202-19.
(6.) Mathor MB, Wajchenberg BL. A simplified method for the calculation of unbound or free testosterone by equilibrium dialysis of undiluted plasma. Clin Chim Acta 1985;145:213-8.
(7.) Umstot ES, Baxter JE, Andersen RN. A theoretically sound and practicable equilibrium dialysis method for measuring percentage of free testosterone. J Steroid Biochem 1985;22:639-48.
(8.) Hammond GL, Nikser JA, Jones LA, Suteri PK. Estimation of the percentage of free steroid in undiluted serum by centrifugal ultrafiltration-dialysis. J Biol Chem 1980;255:5023-6.
(9.) Vlahos I, MacMahon W, Sgoutas D, Bowers W, Thompson J, Trawick W. An improved ultrafiltration method for determining free testosterone in serum. Clin Chem 1982;28:2286-91.
(10.) Adachi K, Yasuda K, Fuwa Y, Goshima E, Yamakita N, Miura K. Measurement of plasma free steroids by direct radioimmunoassay of ultrafiltrate in association with the monitoring of free components with [14C]glucose. Clin Chim Acta 1991;200:13-22.
(11.) Swinkels LM JW, Ross HA, Benraad TJ. A symmetric dialysis method for the determination of free testosterone in human plasma. Clin Chim Acta 1978;165:341-9.
(12.) Willcox DL, McColm SC, Arthur PG, Yovich JL. The application of rate dialysis to the determination of free steroids in plasma. Anal Biochem 1983;135:304-11.
(13.) Sodergard R, Backstrom T, Shanbhag V, Carstensen H. Calculation of free and bound fractions of testosterone and estradiol-17R to human plasma proteins at body temperature. J Steroid Biochem 1982;16:801-10.
(14.) Vermeulen A, Verdonck L, Kaufman JM. A critical evaluation of simple methods for the estimation of free testosterone in serum. J Clin Endocrinol Metab 1999;84:3666-72.
(15.) Rosner W. Errors in the measurement of plasma free testosterone [Letter]. J Clin Endocrinol Metab 1997;82:2014-5.
(16.) Winters SJ, Kelley DE, Goodpaster B. The analog free testosterone assay: are the results in men clinically useful? Clin Chem 1998;44:2178-82.
(17.) Rosner W. An extraordinarily inaccurate assay for free testosterone is still with us [Letter]. J Clin Endocrinol Metab 2001:86; 2903.
(18.) Rinaldi S, Geay A, Dechaud H, Biessy C, Zeleniuch-Jacquotte A, Akhmedkhanov A, et al. Validity of free testosterone and free estradiol determinations in serum samples from postmenopausal women by theoretical calculations. Cancer Epidemiol Biomark Prev 2002;11:1065-71.
(19.) Gruschke A, Kuhl H. Validity of radioimmunological methods for determining free testosterone in serum. Fertil Steril 2001;76: 576-82.
(20.) Morley JE, Patrick P, Perry HM. Evaluation of assays available to measure free testosterone. Metabolism 2002;51:554-9.
(21.) Miller KK, Rosner W, Lee H, Hier J, Sesmilo G, Schoenfeld D, et al. Measurement of free testosterone in normal women and women with androgen deficiency: comparison of methods. J Clin Endocrinol Metab 2004;89:525-33.
(22.) Directive 98/79/EC of the European Parliament and of the Council of 27 October 1998 on in vitro diagnostic medical devices. Off J Eur Communities 1998 (Dec 7):L331/1-L331/37.
(23.) In vitro diagnostic medical devices, measurement of quantities in biological samples. Metrological traceability of values assigned to calibrators and control materials, EN/ISO 17511. Brussels, Belgium: European Committee for Standardization.
(24.) Stockl D, Franzini C, Kratochvila J, Middle J, Ricos C, Siekmann L, et al. Analytical specifications of reference methods. Compilation and critical discussion [Review]. European EQA-organizers Working Group B. Eur J Clin Chem Clin Biochem 1996;34:319-37.
(25.) Siekmann L. Determination of steroid hormones by the use of isotope dilution-mass spectrometry: a definitive method in clinical chemistry. J Steroid Biochem 1979;11:117-23.
(26.) Thienpont LM, Van Nieuwenhove B, Stockl D, De Leenheer A. Calibration for isotope dilution mass spectrometry-description of an alternative to the bracketing procedure. J Mass Spectrom 1996;31:1119-25.
(27.) Vermeulen A, Stoica T, Verdonck L. The apparent free testosterone concentration, an index of androgenicity. J Clin Endocr 1971;33:759-67.
(28.) Slaunwhite WR, Sandberg AA. Transcortin: a corticosteroid-binding protein of plasma. J Clin Invest 1959;38:384-91.
(29.) Thienpont LM, De Brabandere VI, Stockl D, De Leenheer AP. Use of cyclodextrins for prepurification of progesterone and testosterone from human serum prior to determination with isotope dilution gas chromatography/mass spectrometry. Anal Chem 1994;66: 4116-9.
(30.) Thienpont LM, Franzini C, Kratochvila J, Middle J, Ricos C, Siekmann L, et al. Analytical quality specifications for reference methods and operating specifications for networks of reference laboratories. Recommendations of the European EQA-Organizers Working Group B. Eur J Clin Chem Clin Biochem 1995;33:94957.
(31.) National Committee for Clinical Laboratory Standards. Evaluation of precision performance of clinical chemistry device; approved guideline. NCCLS document EP5-A. Wayne PA: NCCLS, 1999.
(32.) Tagawa N, Takano T, Fukata S, Kuma K, Tada H, Izumi Y, et al. Serum concentration of androstenediol and androstenediol sulfate in patients with hyperthyroidism and hypothyroidism. Endocr J 2001;48:345-54.
(33.) Holm SS, Andreasen L, Hansen SH, Faber J, Staun-Olsen P. Influence of adsorption and deproteination on potential free thyroxine reference methods. Clin Chem 2002;48:108-14.
(34.) Whitley RJ, Meikle AW, Watts NB. Endocrinology part VII: gonadal steroids. In: Burns CA, Ashwood ER. Tietz textbook of clinical chemistry, 2nd ed. Philadelphia, PA: WB Saunders Company, 1994:1843-86.
(35.) Holm SS, Hansen SH, Faber J, Staun-Olsen P. Reference methods for the measurement of free thyroid hormones: evaluation of potential reference methods for free thyroxine. Clin Biochem 2004;37:85-93.
(36.) Thode J, Fogh-Andersen N, Siggaard-Andersen M, Siggaard-Andersen O. Donnan effect or protein interference in ionised calcium measurements? Ann Clin Biochem 1983;20:271-3.
(37.) Fogh-Andersen N, Bjerrum PJ, Siggaard-Andersen 0. Ionic binding, net charge, and Donnan effect of human serum albumin as a function of pH. Clin Chem 1993;39:48-52.
(38.) Moll GWM, Rosenfield RL. Some limitations of using equilibrium dialysis to study human serum albumin-testosterone interaction. J Clin Endocrinol Metab 1978;46:501-3.
(39.) Sinha-Hikim I, Arver S, Beall G, Shen R, Guerrero M, Sattler F, et al. The use of a sensitive equilibrium dialysis method for the measurement of free testosterone levels in healthy, cycling women and in human immunodeficiency virus-infected women. J Clin Endocrinol Metab 1998;83:1312-8.
(40.) Taieb J, Mathian B, Millot F, Patricot MC, Mathieu E, Queyrel N, et al. Testosterone measured by 10 immunoassays and by isotope-dilution gas chromatography-mass spectrometry in sera from 116 men, women and children. Clin Chem 2003;49:1381-95.
(41.) Wang C, Catlin DH, Demers LM, Starcevic B, Swerdloff RS. Measurement of total serum testosterone in adult men: comparison of current laboratory methods versus liquid chromatographytandem mass spectrometry. J Clin Endocrinol Metab 2004;89: 534-43.
(42.) Swinkels LMJW, Van Hoof HJC, Ross HA, Smals AGH, Benraad TJ. Concentrations of salivary testosterone and plasma total, non-sex-hormone-binding globulin-bound, and free testosterone in normal and hirsute women during administration of dexamethasone/synthetic corticotropin. Clin Chem 1991;37:180-5.
(43.) International Federation for Clinical Chemistry and Laboratory Medicine. www.ifcc.org (accessed September 15, 2004).
(44.) Bureau International des Poids et Mesures. Joint Committee on Traceability in Laboratory Medicine. www.bipm.org/en/committees/jc/jctlm (accessed September 15, 2004).
KATLEEN VAN UYTFANGHE,  DIETMAR STOCKL,  JEAN M. KAUFMAN,  TOM FIERS,  H. ALEC ROSS,  ANDRE P. DE LEENHEER,  and LINDA M. THIENPONT  *
 Laboratory for Analytical Chemistry and a Laboratory  Medical Biochemistry and Clinical Analysis, Faculty of Pharmaceutical Sciences, Ghent University, Gent, Belgium.
 Laboratory for Hormonology and Department of Endocrinology, Ghent University Hospital, Gent, Belgium.
 Department of Chemical Endocrinology, University Medical Centre, Nijmegen, The Netherlands.
(5) Nonstandard abbreviafions: FTe and TTe, free and total testosterone, respectively; SHBG, sex-hormone-binding globulin; (i)ED, (indirect) equilibrium dialysis; UF, ultrafiltration; SyD, symmetric dialysis; RMP, reference measurement procedure; ID-GC/MS, isotope dilution-gas chromatography-mass spectrometry; SIM, selected-ion monitoring; and IQC, internal quality control.
* Address correspondence to this author at: Laboratory for Analytical Chemistry, Faculty of Pharmaceutical Sciences, Ghent University, Harelbekestraat 72, B-9000 Gent, Belgium. Fax 32-9-264-8198; e-mail linda.thienpont@ ugent.be.
Received May 17, 2004; accepted August 17, 2004.
Previously published online at DOI: 10.1373/clinchem.2004.037358
Table 1. Steroid hormones tested for possible interference on SIM at m/z 680 and 682 (molecular ions of the diheptafluorobutyryl derivatives of Te and [sup.13][C.sub.2]-Te) [C.sub.2]]-Te). Steroid hormone (commonly used name) Interference (a) Estrogens 3-Hydroxyestra-1,3,5 (10)-trien-17-one (estrone) 1 (17[beta])-Estra-1,3,5 (10)-triene-3,17-diol (estradiol-17[beta]) 1 Progestins Pregn-4-ene-3,20-dione (progesterone) 1 Corticosteroids (11 [beta])-11,17,21-Trihydroxypregn- 4-ene-3, 20-dione (cortisol; hydrocortisone) 2 (11 [beta])-11,21-Dihydroxypregn- 4-ene-3,20-dione (corticosterone) 2 21-Hydroxypregn-4-ene-3,20-dione (deoxycorticosterone) 2 Androgens 17[alpha]-Hydroxyandrost-4-en-3-one (epitestosterone) 3 (r = 0.93) (b) (17[beta]-17-Hydroxyestr-4-en-3-one (19-nortestosterone; nandrolone) 1 (17[beta]-17-Hydroxy-17-methylandrost- 4-en-3-one (17-methyltestosterone) 2 (5[alpha]17)-17-Hydroxyandrostan-3-one (stanolone; dihydrotestosterone) 1 5[alpha]-Androstane-3,17-dione 1 Androst-4-ene-3,17-dione 1 (3[alpha], 5[alpha])-3-Hydroxy- androstan-17-one (androsterone) 1 (3[alpha], 5[alpha])-3-Hydroxy-androstan- 17-one (epiandrosterone) 1 3[beta],3-Hydroxyandrost-5-en-17-one (dehydroisoandrosterone; prasterone) 1 (3[beta],17[beta])-5[alpha]-Androstan-3,17-diol 4 (r = 1.02) (3[beta],17[beta]) Androst-5-ene-3,17-diol 5 (r = 1.00) 17[alpha]-Hydroxypregn-4-en-20-yn-3-one (ethisterone) 2 (a) Codes for interference: (1), Not possible on theoretical grounds. (2), No molecular ion or fragment ions visible in the mass spectrum in the m/z range 676-686; no SIM signal at m/z 680 or 682. (3), Although the spectrum showed that the molecule derivatizes in the same way as testosterone (m/z of the molecular ion at m/z 680), interference can be excluded on the basis of the relative retention time. (4), Forms a diheptafluorobutyryl derivative with molecular ion at m/z 684; however, with a relative retention time of 1.02 min and no detectable SIM signal at m/z 682 and 680. (5), Reacts to the diheptafluorobutyryl derivative with molecular ion at m/z 682. (b) Relative retention time in comparison with testosterone. Table 2. Precision data for the UF-ID-GC/MS candidate RMP in comparison with iED and SyD. Measurement Concentration, [CV.sub.wr], [CV.sub.dd], procedure pmol/L (a)% (b)% Candidate RMP 138.7 3 3.1 (including UF) 431.7 2.2 2.5 ID-GC/MS 349 1 0.4 iED SyD CV from duplicate measurements, Measurement [CV.sub.T], % procedure (c) % Candidate RMP 4.3 3.3 (including UF) 3.3 ID-GC/MS 1.1 iED 7.8 (14) 5.6 SyD 5.0 (42) 7.6 (a) Within-run CV. (b) Between-day CV. (c) Total CV. References in which the CVT for the iED and SyD methods are reported are given. Table 3. Characteristics of the tested UF devices in terms of protein leakage and adsorption. Protein leakage Adsorption UF device % CV, % % CV, % Vivaspin 0.005 25 Very high polyethylene- (a) and sulfone extremely 10 variable Vivaspin 0.006 32 Very high regenerated and cellulose extremely 10 variable Vivaspin 0.006 10 Very high cellulose and acetate 10 extremely variable Centricon 0.010 8 6 0.3 YM-10 Centricon 0.014 10 2 1.5 YM-30 (a) Range, 66-99%. Table 4. Optimization of the UF process in terms of dilution of the serum, sample volume on the membrane, and generated volume of ultrafiltrate. Ultrafiltration conditions Difference CV, (a)% P vs (a), % Dilution of serum with buffer (a) 1 mL + 0.05 mL 2.4 (b) 1 mL + 1.4 mL 0.3 1.6 < 0.3588 (c) 1 mL + 3 mL 1.3 2.9 < 0.1203 (d) 1 mL + 9 mL 13.7 4.1 < 2.14 x [10.sup.-9] Initial sample volume with an ultrafiltration yield of 33% (a) 1 mL 1.4 (b) 1.7 mL 0.5 1.1 < 0.2457 (c) 2.4 mL 0.4 1.4 < 0.2931 Generated volume of ultrafiltrate (yield from 2.4-mL sample) (a) 0.8 mL (33%) 1.4 (b) 0.4 mL (17%) 1.4 1.3 < 0.0128 (c) 1.1 mL (46%) 0.2 1.7 < 0.3823 (a) Experiments performed in quadruplicate with three different sera. Table 5. Summary of the Deming regression analysis data for the method-comparison study. Method pair Slope (SE) Intercept (SE), pmol/L UF-ID-GC/MS vs IED 0.9767 (0.085) 52.61 (27.62) UF-ID-GC/MS vs SyD 0.9208 (0.078) 21.43 (20.11) Method pair Correlation [S.sub.y|x], coefficient (a) pmol/L UF-ID-GC/MS vs IED 0.9447 (P < 0.001) 41.97 UF-ID-GC/MS vs SyD 0.9701 (P < 0.001) 30.7 Method pair [CV.sub.mp], (b)% UF-ID-GC/MS vs IED 5.8 UF-ID-GC/MS vs SyD 4.6 (a) Standard deviation of y on x. (b) CV for the method pair.
|Printer friendly Cite/link Email Feedback|
|Title Annotation:||Endocrinology and Metabolism|
|Author:||Van Uytfanghe, Katleen; Stockl, Dietmar; Kaufman, Jean M.; Fiers, Tom; Ross, H. Alec; De Leenheer, A|
|Date:||Nov 1, 2004|
|Previous Article:||Lithium monitoring by reverse iontophoresis in vivo.|
|Next Article:||Validation of a new automated renin assay.|