Description and validation of an equilibrium dialysis ID-LC-MS/MS candidate reference measurement procedure for free thyroxine in human serum

Calibration

For single-point calibration [14], Thyroxine (3, 3′, 5, 5′-tetraiodo-L-thyronine; T4) certified reference material IRMM-468 (VWR International, NL) with a purity of 98.6 ± 0.70% (expanded uncertainty, k=2) was used [15] and concentrations of calibration solutions were adjusted for impurity. As internal standard, [¹³C₆]-T4 (ISO-Sciences; King of Prussia; art. nr. 5031) was used.

Three independently prepared standard stock solutions were diluted to a working solution. All volumetric steps were gravimetrically performed (Sartorius SE2-OCE). To prepare T4 standard stock solution, 15 mg T4 standard was weighted, transferred to a glass vial (Perkin Elmer) and dissolved in methanol + 0.125% (v/v) NH₄OH (Biosolve; Acros) with sonification and stored overnight at −30 °C. Intermediate T4 stock solutions A (80 µg/g), B (2.5 µg/g), C (60 ng/g) and working solution (1.1 ng/g) were prepared in methanol. [¹³C₆]-T4 internal standard stock solution (25 µg/g) was used to prepare solution A (2.63 µg/g), solution B (8.0 ng/g) and working solution (0.2 ng/g) in methanol.

A 3-iodo-L-tyrosine (MIT; Sigma, art. nr. I8250) solution was used to prevent adsorption of T4 to vials and added to T4 standard working solutions, internal standard solution B and working solution at a concentration of 5,000 times higher than the T4 concentration.

Equilibrium dialysis

ED was performed by the defined conventions as described [13]. They required biochemical composition of dialysis buffer resembling the ionic environment of serum; samples to be buffered to a pH of 7.4 ± 0.03 (at 37 °C) during dialysis; thermostatic control to be maintained during dialysis at 37.0 ± 0.5 °C; the use of an identical volume of dialysand/dialysate compartments and these compartment should be separated by a membrane of regenerated cellulose and adequate cut-off [13].

Frozen samples were thawed at room temperature and 1 mL serum was buffered to pH of 7.4 ± 0.03 at 37 °C by adding 1/10 (v/v) concentrated HEPES (0.776 M (N-(2-hydroxyethyl) piperapiperazine-N9-(2-ethylsulfonic acid)) (VWR International, NL) and 0.22 mol/L NaOH (sodium hydroxide; Merck, art. nr. 1.06498)). Samples were dialyzed against HEPES buffer (52.74 nM HEPES, 22.5 mmol/L NaOH, 91.6 mmol/L NaCl (Merck), 1.65 mmol/L KH₂PO₄, 2.68 KCl (Merck), 1.12 mmol/L MgSO₄.7H₂O (Merck), 5.0 mmol/L urea (Merck), 1.90 mmol/L CaCl₂.2H₂O (Merck), 8 mmol/L NaN₃ (Merck)) set at pH 7.4 ± 0.03 at 37 °C with 10 M NaOH and when necessary adjusted just before use.

Regenerated cellulose dialysis membranes (Dianorm, ref 10.14, Harvard Apparatus, art. nr. 74–2100) with a diameter of 63 mm and 5 kDa cut-off value were pretreated with deionized water and dialysis buffer at 37 °C. 1.0 mL dialysis cells (PTFE) were used and consisted of two identical halves between which the dialysis membrane was placed. 1.0 mL pH-adjusted serum was injected in one half of the dialysis cell and an equal amount of dialysis buffer was injected in the other half. Assembled dialysis cells were then placed in the Dianorm^® Equilibrium dialyser in a water bath at 37 ± 0.5 °C for 4 h and continually rotated. An appropriate amount of [¹³C₆]-T4 was added to a borosilicate vial. The required amount of [¹³C₆]-T4 was estimated using an IA (FT4 on a E801 random access analyzer, Roche diagnostics) or based on the previous single measurement result with the cRMP to achieve ratios of T4/[¹³C₆]-T4 close to 1:1. After dialysis, dialysate was emptied from the cells, added to the borosilicate vial with [¹³C₆]-T4 and equilibrated for at least 1 h before proceeding with LLE [16]. Whereas the JCTLM listed RMP uses a solid phase extraction step to purify the dialysate for LC-MS/MS measurement of T4 [13], the cRMP described here uses a liquid-liquid extraction (LLE) step [16]. In brief, 65 μL of 35% formic acid solution (CHO₂H; Biosolve; cat. nr. 000 69 141 A8) was added to the sample followed by three wash steps with 900 μL C₆H₁₂ (cyclohexane; Merck; cat. nr. 1.02817.1000). After an additional 35 μL of 35% formic acid (v/v) solution T4 and [¹³C₆]-T4 were extracted twice in 1 mL CH₃CO₂C₂H₅ (ethyl acetate; VWR International, NL; cat. nr. 83621.290). After evaporation on a nitrogen flow at 40 °C (Techne Dryblock DB-3D) extracts were reconstituted in 1 mL of ethyl acetate and dried twice to reduce residual formic acid, subsequently reconstituted into 80 μL reconstitution solution (25% acetonitrile (CH₃CN), 0.02% (v/v) formic acid) and transferred to an injection vial. Every sample was processed and measured at least in duplicate in independent experiments.

Isotope-dilution liquid chromatography tandem mass-spectrometry (ID-LC-MS/MS)

ID-LC-MS/MS analyses of T4 were performed using a Waters Acquity 2D UPLC system coupled to a Waters Xevo TQ-XS mass spectrometer (Milford, MA, USA). Mobile phases A and B consisted of 0.02% formic acid in water and 0.02% formic acid in acetonitrile (Biosolve; cat. nr. 000 1204 101), respectively. Gradient program and mass spectrometer settings were described in Table 1. Acquisition and data processing were performed using Waters MassLynx Software. Calculations were made in Excel (Supplementary Table 1).

Evaluation of dialysis conventions of accurate pH, temperature and buffer composition

HEPES buffers were prepared as specified within maximum allowed deviation of 1.5% (w/w) (HEPES, NaCl, KH₂PO₄, KCl, urea, CaCl₂.2H₂O, NaN₃), 5% (w/w) NaOH or 1% (w/w) MgSO₄.7H₂O and pH was set at 7.4 ± 0.03 at 37 °C. pH stability during ED in dialysis buffer and serum was tested (n=4) at 15 min and after 4 h of dialysis (pH-meter; Metrohm 744, Applikon) at 37 °C. Continuous temperature measurements were performed using a temperature logger (Tinytag Plus 2 TGP-4020; Gemini data loggers) and probe (PB5001-1M5 10K NTC). Accuracy of the temperature reading was verified with a second calibrated thermometer (Fluke 53 II, FLUKE USA).

Specificity and sensitivity

Potential interferences on T4 quantitation were evaluated in each individual serum sample and compared to (matrix free) calibrators by assessment of peak shape, ratio of T4 qualifier to quantifier of mass transitions (Q1/Q2) of both T4 and [¹³C₆]-T4, peak width at half its height (W1/2) and retention times (RT). An accurate T4 RT compared to the RT of the T4 calibrator and compared to the sample [¹³C₆]-T4 RT, a normal Gaussian peak shape and less than 5% deviation of Q1/Q2 compared to the mean calibrator Q1/Q2 was used to confirm absence of interferences in serum samples. The limit of detection (LOD) was calculated in three independent experiments by measuring blank dialysis buffer and dialysis buffer to which 1.31 pmol/L T4 was added, each in triplicate (n=9). LOD was calculated as follows: average T4 quantification in the blank + 3* standard deviation (SD) of the 1.31 pmol/L samples. The lower limit of quantitation (LLOQ) was determined by serial measurement of a serum pool, composed of sera from patients with severe hypothyroidism, at a concentration with an expected 20% CV. Ion suppression was assessed by continuous infusion of labeled standard. The abundance was compared between a serum dialyzed matrix and mobile phase at the RT of T4.

Linearity

Linearity of the measurement range was evaluated following principles as described in CLSI document EP6-A [17]. In short, a dialyzed serum pool of patients with severe hypothyroidism and hyperthyroidism were used to prepare a series of dialysates with varying concentrations of T4 with known dilution ratios relative to another. T4 in these samples was measured after LLE by ID-LC-MS/MS in quadruplets and a-linearity was assessed [18].

Evaluation of accuracy and precision

Accuracy of the LLE–ID-LC-MS/MS procedure was evaluated using serum-based T4 certified reference material (CRM) (Certified Reference control sample, Thyroxine 103.7 nmol/L RfB DGKL, Germany, D-K-15117-01-00; 40122) diluted in dialysis buffer with MIT. Precision was evaluated by analyzing serum pools with levels at the lower and upper reference limits and at clinical levels of strong hypo- and hyperthyroidism. Controls were aliquoted and kept at −80 °C until analysis. FT4 was measured in duplicate and for up to 20 separate experiments over 15 months. Imprecision was expressed as % coefficient of variance (CV), where both within-run CV (CV_wr) and total CV (CV_T) were calculated based on CLSI document EP5.

Accuracy of the cRMP was evaluated by an interlaboratory comparison with the IFCC endorsed FT4 cRMP at Ref4U, Laboratory of Toxicology, Faculty of Pharmaceutical Sciences, Ghent University, Ghent, Belgium; (JCTLM DB identification number: C8RMP1) [13], at the National Center for Environmental Health, Centers for Disease Control and Prevention, Atlanta, GA, USA (CDC) and at the Reference Material Institute for Clinical Chemistry Standards, Yokohama, Kanagawa, Japan (ReCCS). For this comparison, serum samples in euthyroid state (n=20) were collected (Solomon Park Research Laboratories, Kirkland WA) with informed consent, serum was centrifuged and aliquoted and distributed to participating centers. Samples were analyzed in 3–6 independent sample preparations using one replicate per run and a Deming regression analysis was performed to compare our FT4 cRMP with the mean of four laboratories and a bias was calculated (Bland-Altman plot). In addition, imprecision of these samples was calculated.

Measurement uncertainty was assessed at three levels. Type A uncertainty was obtained from imprecision of repeated measurements in serum pools; Type B uncertainty was estimated from uncertainties in purity of primary reference material, inaccuracy in weighing of standards, measurement of serum dialysate density, sample-related effect and unattributed variances.

Robustness of the FT4 candidate cRMP

To demonstrate robustness of the cRMP an experiment in which serum was diluted with HEPES buffer was performed. Four sera were serially diluted up to 8-fold (excluding the 1:1 dilution in the dialysis procedure) and FT4 was measured. Results were calculated as percentages compared to undiluted sera.

Specificity and sensitivity

Sample and (matrix free) calibrator peak characteristics for T4 and [¹³C₆]-T4 from four independent experiments were summarized in Table 2. No interferences based on peak characteristics were observed in samples. Ion suppression up to 30% was found (data not shown). The use of an carbon-13 isotope labeled T4 internal standard ([¹³C₆]-T4) resulted in an identical chromatographic behavior compared to the unlabeled standard. Both the use of [¹³C₆]-T4 and an adjusted ratio of T4/IS at 1:1 minimizes potential inaccuracy due to ion suppression.

Linearity was evaluated in serum dialysates over a range of 1.0–113 pmol/L and a non-significant a-linearity (p=0.54) was observed. Furthermore, limit of detection (LOD) and limit of quantification (LOQ) were assessed and estimated at 0.39 and 1.39 pmol/L (CV 23.9%), respectively.

Accuracy and imprecision

Table 3 shows the imprecision of serum pools and T4 CRM. The mean CV of the individual 20 serum samples in the euthyroid range used for the method comparison was 4.1%. The relative Measurement Uncertainty was assessed at three levels. Type B uncertainty was minimized by using primary IRMM-468 standards and using gravimetric measurements. For the recommended measurement protocol of n=3 (singlicate measurement of a sample on three independent occasions), the relative expanded uncertainty was estimated at ≤7.6% (Supplementary Table 2), which is comparable to the JCTLM listed RMP [13].

Accuracy of the FT4 candidate cRMP

The accuracy of the candidate cRMP was evaluated with a method comparison of serum samples from 20 volunteers and measured by the cRMP described here (laboratory at Radboudumc) as well as the other participating centers. The Radboudumc cRMP has a slope of 1.07 (1.02–1.13 95% CI), an intercept of −0.835 (−1.9 to 0.3 95% CI), and a correlation coefficient (r²) of 0.98 (Figure 1A) with an average bias of 2.9% (1.35–4.39 95% CI, p=0.0009) compared to the overall mean (Figure 1B). Results from all participating laboratories (UGhent, CDC, ReCCS, Radboudumc) compared to the overall mean are displayed in Figure 2.

Figure 1:

Accuracy of the FT4 candidate cRMP. (A) Deming regression analysis; on the y-axis the FT4 concentration measured at the Radboudumc and on the x-axis the mean FT4 concentration measured at the University of Ghent, CDC ReCCS and Radboudumc (overall mean). (B) Bland-Altman plot; on the y-axis the relative difference (mean FT4 Radboudumc- FT4 overall mean/FT4 overall mean; %) and on the x-axis the mean FT4 concentration measured at Ghent University, CDC, ReCCS and Radboudumc (overall mean).

Figure 2:

FT4 scatterplot of all reference network laboratories. Scatter plot representing the bias of the reference network laboratories (UGhent, CDC, ReCCS, Radboudumc) compared to the overall mean of four participating laboratories. On the x-axis the overall mean was shown and on the y-axis the % difference per laboratory from the overall mean was depicted.

Robustness of the FT4 candidate cRMP

Robustness of the FT4 candidate cRMP in sera with different thyroid hormone binding capacities was investigated by measuring serum FT4 in a dilution series. Based upon concentrations of T4 and binding proteins and their dissociations constants, FT4 in euthyroid sera remains constant upon dilution with an inert buffer [19]. A dilution of sera up to 16-fold in sera with FT4 concentrations in the range of 16.5–35.7 pmol/L displayed an average deviation of 1.9% demonstrating its robustness in samples with low T4 binding protein concentrations.