ORIGINAL Annals of Nuclear Medicine Vol. 1, No. 1, 15-22, 1987 Measurements of serum-free thyroid hormone concentrations by ultrafiltration a comparison with equilibrium dialysis and mathematical calculation Norimichi KONNO,* Kohji HAGIWARA,** Hideo TAGUCHI,** Shigeki MURAKAMI** and Shizuko TAGUCHI *Department of Internal Medicine, **Department of Radiology, Department of Central Laboratory, Hokkaido Central Hospital for Social Health Insurance, Nakanoshima, Sapporo, 062, Japan An ultrafiltration method (UF) for measuring free thyroxine (FT4) and free triiodothyronine (FT3) using the Diaflow YM membrane (Centricon-10) is described. The results are compared with those by equilibrium dialysis (ED) and also by mathematical calculations derived from T4, T3, and binding protein concentrations. The precision with the UF method was excellent. The normal ranges of FT4 and FT3 by the three methods are all comparable. There was a high degree of correlation of FT4 or FT3 results by UF with those by ED and by calculation (r=0.940-0.974, n=161, P<0.001). FT4 and FT3 by all methods agreed well for hyperthyroidism, hypothyroidism, and for patients with low T4-binding globulin. The mean FT3 in pregnancy was lower than the normal value for all methods, and FT4 concentrations by UF and calculation also decreased in late pregnancy. The mean FT4 by UF and ED in low T3 syndrome were significantly higher than in the normal controls, while the calculated FT4 was lower. The FT3 in low T3 syndrome distributed normal to subnormal in all methods. These results indicate that a) the UF method is a reliable reference method for measuring FT4 and FT3 concentrations; b) the UF results agree well with those by ED and also with theoretically derived values in subjects with thyroid diseases and TBG abnormalities; c) for patients with low T3 syndrome, the FT4 results obtained by UF and ED are similarly discrepant from the calculated results, implying the existence of binding inhibitor(s) which affect both UF and ED measurements. Key words : Free thyroid hormones, Ultrafiltration, Equilibrium dialysis, Calculation, Diagnostic aid, Thyroid function INTRODUCTION CIRCULATlNG THYROID HORMONES, thyroxine (T4) and triiodothyronine (T3), are bound to the binding proteins, T4-binding globulin (TBG), T4-binding pre-albumin (TBPA), and albumin, and only very small fractions of T4 and T3 are in the unbound (free) form in human serum.1 Because free T4(FT4) and Received April 28, 1987; revision accepted April 28,1987. For reprints contact : Norimichi Konno, Department of Internal Medicine, Hokkaido Central Hospital for Social Health Insurance, Nakanoshima, Sapporo, 062, JAPAN. free T3(FT3) are generally not affected by variations in T4-binding proteins,2,3 FT4 and FT3 concentrations reflect the thyroid function more closely than the total concentrations of T4 and T3. The most widely used method for estimation of FT4 and FT3 is equilibrium dialysis (ED),4,5 but this method is time-consuming and not convenient for clinical use. To overcome this, a number of radio-immunoassays (RIA) based on different principles have been developed for estimates of FT4 and FT3, but results with these may be influenced by changes in serum proteins and by the effect of nonthyroidal illnesses (NTI).6,7 Ultrafiltration techniques have been developed in recent years for use with dialysis tubing or micro-partition systems8-13 and this technique is an alternative to ED and RIA. In the present study, we evaluate the ultrafiltration technique using Amicon YM Diaflow membrane for FT4 and FT3 measurements and compare the results with those by ED and mathematical calculations. MATERIALS AND METHODS Patients Six groups of patients were studied: (I) 45 normal subjects who visited the hospital for routine health examinations, ages 38-68 years ; (II) 17 hyperthyroid patients with supranormal T4 and T3 ; (III) 10 hypothyroid patients with subnormal T4 and (or) T3 and supranormal thyrotropin; (IV) 5 patients with low TBG; (V) 40 pregnant women ; (VI) 44 patients with NTI, including 9 patients with acute hapatitis associated with supranormal TBG, and 35 patients with low T3 syndrome. The etiological diagnoses of the low T3 syndrome include malignant neoplasma (22 cases), cerebral vascular disorders (four cases), rheumatoid arthritis (three cases), renal failure (two cases), cardiac failure (two cases), pulmonary failure (one case), and hepatic failure (one case). None of these patients were treated with heparin or aspirin. All serum samples were stored at -20deg.C until assayed. Reagents and Apparatus Tracer. T4 labeled with 125I (T4 tracer) and T3 labeled with 125I (T3 tracer), both with a listed specific radioactivity of more than 1.2 mCi/ug dissolved in 75% ethanol, were obtained every four weeks from Amersham International plc. On receipt, the tracers were evaporated under nitrogen at 37deg.C, and human serum albumin (Sigma Chemical Company, T4 and T3 concentrations in the albumin were 0.08 ug/g, and 5.5ng/g, respectively by RIA) in 0.063 M phosphate buffer (pH=7.4) in 0.05% saline was added to yield final concentrations of 1.0 to 1 .2% of albumin and incubated for 20 min at 37deg.C in darkness. The mixture was dialyzed overnight at 4deg.C against 2.0 L of phosphate buffer to remove labeled contaminants of the tracers. This purification procedure was performed every two weeks. The paper chromatographic analysis (butanol : acetic acid : water, 78 : 5 : 17 v/v) showed that the dialyzed specimens of T4 and T3 tracers contained less than 1.O% iodide. The tracer concentration for the assay was approximately 20 uCi/ml, or 16.7 ug/L for both T4 and T3 . Apparatus. For the ultrafiltration, a Centricon centrifugal microconcentrator (Centricon-10, Amicon Corp., Lexington, Mass, USA) with a Diaflow YM-10 membrane and membrane support base, an O-ring, and a filtercup was used.This membrane has a relative molecular mass (Mr) cut-off of 10,000 daltons. Procedures Ultrafiltration. We added 50 ul of T4 tracer or 25 ul of T3 tracer to 100ul of serum diluted with 650 ul (for FT4 assay) or 675 ul (for FT3 assay) of 0.15 M phosphate buffer (pH=7.4) in 0.05% saline. After mixing, the diluted sera were kept in the test tubes for 10 min at room temperature. A 700 ul portion of each serum sample was transferred to the Centricon-10 device, and incubated for a further 20 min at 37deg.C. It was centrifuged at 1,OOO xg for 40 min at 37deg.C. Operation instructions recommend the use of a fixed angle rotor for centrifugation, however a conventional swinging-bucket rotor suitable for obtaining a sufficient volume of filtrate was used. A 300 ul portion of filtrate was added to 1.7 ml of phosphate buffer and 1 ml of the carrier T4 and T3 solution. The radioactive T4 and T3 were separated by magnesium chloride precipitation techniques as described elsewhere.5,14 We determined the total radioactivity by counting a 24 ul aliquot of the original diluted serum samples. The recovery of T4 and T3 by MgCl2 precipitation were 95% and 83%, respectively. The free fractions of iodothyronines were calculated as follows : where the 8 in the denominator is the dilution factor. For the FT3 assay, the results were corrected for the 83% yield during the MgCl2 precipitation. FT4 and FT3 concentrations were expressed as the products of total T4 (or T3) and %FT4 (or %FT3). By this procedure, 20 single samples can be assayed in less than three hours. Equilibrium Dialysis. FT4 and FT3 in the 80-fold diluted sera were also measured by equilibrium dialysis with the tracer method and MgCl2 precipitation techniques as described elsewhere5,14. Mathematical Calculation. The calculation of free thyroid hormone concentrations was made from the following equation, according to Lecureuil et al15 : where, F is the molar concentration of FT4 or FT3 CTBG, CTBPA, CAlb are the molar concentrations of TBG, TBPA and albumin, and n is the number of binding sites of albumin. We selected values for the binding constants (K) that were within the wide ranges quoted by different authors (Table 1)16-21. The program was written for an M-20 (Olivetti, Japan) personal computer. Total T4, T3, TBG, TBPA and albumin. Total T4, T3, and TBG were measured by RIA.14,22 Serum TBPA and albumin were measured by immunotur-bidimetry.23 Albumin concentration in the ultrafiltrate was measured by sensitive RIA for urine analysis (Albumin RIA kit, Diagnostic Products Corporation, USA). Precision of Assay Interassay CV (n=11) were: T4 1.8%; T3 4.4%; TBG 7.9%; TBPA 4.7%; albumin 6.9%; %FT4 by ED 11.0%; %FT3 by ED 12.5 % for normal control sera. Statistical Analysis Analysis was by least-square regression, and groups were compared by Student's t-test. RESULTS In preliminary experiments, we used the albumin RIA kit for urine analysis (approximate sensitivity 0.1 mg/L) to detect protein leakage into the ultrafiltrates. Leakage was 0.00158+-0.00036% (mean+-S.D.) (n =4) albumin from normal sera during ultrafiltration at 37deg.C for 40 min. This suggests that the protein leakage through the membrane does not cause appreciable error in the FT4 and FT3 measurements. During centrifugation at 1,OOO x g, progressive increase in filtrate volume was seen from 10 through 60 min. The %FT4 and %FT3 for each centrifugation period were measured with 8-fold diluted sera, and all values were nearly constant (Fig. 1). The binding of thyroid hormones to the Centricon-10 devices was examined. An 800 ul volume of 8-fold diluted sera enriched with tracers was centrifuged at 37deg.C for 40 min at 1,OOO x g, and radioactivity bound to the device was counted after three times washing with water. The binding of T4 was 0.18+-0.014% (n=3) and for T3 it was 0.39+-0.09% (n=3) of total. The effect of temperature during centrifugation on %FT4 and %FT3 was also examined, and there were progressive increases in both fractions from 4deg.C to 37deg.C. The deiodination of T4 and T3 tracers during the 20 min incubation at 37deg.C was examined by paper chromatographic analysis using 8-fold diluted sera. With T4 tracer containing 0.5% iodide, or T3 containing 0.2% iodide added to sera, the iodide contamination of T4 tracer after 20 min of incubation was 0.63+-0.12% (n=5) and for T3 tracer, 0.48+-0.1% (n=5). The effect of serum dilution of %FT4 and %FT3 is shown (Fig. 2). The dilution of sera from normal subjects, hyperthyroidism and pregnancy patients showed an initial decline (approximately 30%) up to 8-fold dilution. Further dilution had little effect on both %FT4 and %FT3. In the serum from low T3 syndrome subjects, there was a progressive decline of %FT4 (48.3%) and %FT3 (35.3%) at 8-fold dilution and they declined further thereafter. The intra-assay precision (CV) of the UF method was 2.2% for %FT4 (n=5), and 3.1% for %FT3 (n=5), and the inter-assay CV was 6.8% for %FT4 (n=10) and 12.6% for %FT3 (n=10) for samples in the normal range. The mean values (+-S.D.) for measured and calculated free thyroid hormones, together with the values for the various parameters we used in our computation, are shown in Table 2, and the individual values in low TBG, pregnancy, and NTI are also shown in Fig. 3 and Fig. 4. The normal ranges of FT4 were 0.95 x 10-11 to 2.4 x 10-11 mol/L for UF, 10-11_2.1xlO-11 mol/L for ED, and 0.95 x 10-11_2.2 x 10-11 mol/L for calculated FT4. The normal ranges of FT3 concentrations were 2.1 x 10-12_4.6 x l0-12 mol/L for UF, 2.1 x 10-12_5.0 x 10-12 mol/L for ED, and 2.4 x 10-12 _5.0 x 10-12 mol/ L for calculated FT3. The measured and calculated normal ranges for FT4 and FT3 were not significantly different. Compared with the values for normal controls, both FT4 and FT3 by these methods were significantly higher in hyperthyroidism and lower in hypothyroidism. Both mean values for FT4 and FT3 by these methods were similar to the corresponding normal controls in the low TBG group. The mean FT4 concentrations of the pregnant women in the 2nd and 3rd trimester were significantly lower than the mean value for normal controls, whether calculated or measured by UF, but not by ED although the mean FT4 by ED was lower than for normal controls. The FT3 values in the pregnant women in all trimesters were significantly lower than in the normal controls with all the methods. The individual values for FT4 and FT3 showed that both measured and calculated FT4 m pregnancy distributed essentially within the normal ranges, while the FT3 values were subnormal in 5 of 39 for UF, and in 7 of 39 for ED. The mean FT4 values in NTI with high TBG were normal in all methods employed, but the FT3 values in this group were significantly lower than the normal values by calculation and measured with UF but not by ED. The measured FT4 concentration for low T3 syndrome was significantly higher than that of normal controls, whereas the calculated value was significantly lower. Individual FT4 also showed that the measured FT4 values distributed from subnormal to supranormal, in contrast to the normal-to-subnormal distribution of the calculated FT4 values in NTI. The FT3 concentrations, both measured and calculated, distributed from normal to subnormal in NTI, and the mean values in the low T3 syndrome were significantly lower than the corresponding normal values. Table 3 summarizes the regression analysis of the results obtained by the three methods for FT4 and FT3 measurements. When all data were analyzed together, the best correlations were between UF and ED methods for FT4 and FT3 (r=0.974, and 0.972, respectively). The calculated FT4 or FT3 has a similar correlation with the UF and ED methods. When each subgroup was subjected to separate statistical analysis, there was poor correlation between UF and ED for FT3 m normal subjects and in pregnant women, and between ED and calculations for FT3 in normal subjects and in pregnant women. The calculated FT4 and FT3 values showed good correlations with both ED and UF methods for all groups analyzed. DISCUSSION Determination of free thyroid hormone concentrations using the Amicon ultrafiltration device with a Diaflow YM membrane offers several advantages. This method is rapid, reproducible, and technical]y simple. A minimal serum albumin leak will not produce appreciable errors in the measurements. The %FT4 and %FT3 values are unchanged throughout the centrifugation. In addition, binding of tracers to the device and deiodination of tracers are minimal during the procedures. These points must be added to the evaluations in previous reports.11-13 The normal ranges for FT4 an FT3 concentrations by the UF method in the present study were somewhat lower than previously shown with undiluted sera.9,10 This is presumably due to the effect of serum dilution, as the data showed serum dilution to produce a lowering of free thyroid hormone fractions, as previously shown in theoretical and experimental studies.17,24-26 There are reports where the free thyroid hormone concentrations by the UF method were compared with the ED method,11-13 however no reports have included a comparison with mathematical calculations. The mathematical model for the interaction between thyroid hormones and their transport proteins in sera has been proposed by several investigators.15,17,27,28 The calculated values for FT4 and FT3 varied considerably, possibly because of different values for affinity constants of the binding proteins employed in the calculations.27,28 The model and calculation method by Lecureuil et al,15 which we used here, is simpler than that by Prince and Ramsden.17 The affinity constants used here can still be considered acceptable, as the measured and calculated normal free thyroid hormone concentrations were quite similar. The values for both methods also correlated well with each other in all the combined data, and also in the various groups analyzed separately. The calculated and measured free thyroid hormone concentrations agree well with each other and are predictable in hyperthyroidism, hypothyroidism, and in patients with low TBG levels. The FT4 concentrations decrease as pregnancy progresses in the calculated values as well as in the UF measured values but not in the ED-measured FT4. The mean FT3 values are also lower than the normal values in pregnancy for all methods. The reason for the discrepancy with the FT4 results in late pregnancy as measured by UF and ED is unknown, but large variation in the ED method may be involved because the interassay CV for ED is larger than that for UF. Although the reduced FT4 and FT3 levels in late pregnancy were established theoretically,29 there is disagreement with previous investigations on the UF measured FT4 and FT3 values,3 Our data agree with Shannon et al,12 Weeke et al,10 and Lee et al,30 but not with Faber et al,9 Sophianopoulos et al,11 and Wang et al.13 The FT4 and FT3 concentrations by the present UF method agree well with the calculated values, and it is apparent that this method may be useful in evaluating the thyroid functional status with altered TBG concentrations. We found discrepancies between measured and calculated concentrations of FT4 but not for FT3 in low T3 syndrome. The FT4 as measured by UF and ED were both significantly higher than the normal values, while the calculated values were conversely lower. Similar discrepancies were reported by Brown-Grant et al. using a different model for FT4 calculations and the ED method.27 One of the reasons for this discrepancy may be due to the presence of compound(s) that interfere with hormone-protein interaction.31 The progressive decline of FT4 and FT3 by serum dilution in the low T3 syndrome are in accord with the findings by Nelson et al,26 and Weeke et al,10 and may be interpreted to mean that the serum from this syndrome may contain such an inhibitor(s). Chopra et al have shown that free fatty acid is a likely candidate for the inhibitor, and this inhibitor affects the T4 binding stronger than the T3 binding.31 Although the role of free fatty acid has been discounted recently,32 our agreement between calculated and measured FT3 in the low T3 syndrome, unlike the FT4, may be consistent with Chopra et al. Our data also indicate that the inhibitor may be equally effective with ED and UF systems. However, we cannot exclude other possibilities such as an alteration of binding characteristics of TBG,33 or the presence of a different type of TBG in the sera of low T3 syndrome.34 The principles of the UF method have been well established,35 and we describe a method for measurements of FT4 and FT3 which is a reliable and useful reference method for evaluation of thyroid functional status in clinical and experimental studies. REFERENCES 1. Robbins J, Rall JE: Proteins associated with the thyroid hormones. Physiol Rev 40 : 415-89, 1960 2. Ekins R, Edwards P, Newman B : The role of binding proteins in hormone delivery, in Free Hormones in Blood, pp. 3-43, Albertini A and Ekins RP (eds.), Elsevier Biomedical Press, Amsterdam, New York, Oxford, 1982 3. Wilke TJ : Estimation of free thyroid hormone concentrations in the clinical laboratory. Clin Chem 32: 585-592, 1986 4. Ingbar SH, Braverman LE, Dawber NA, et al : A new method for measuring the free thyroid hormone in human serum and an analysis of the factors that influence its concentration. J Clin Invest 44 : 1679-1689, 1965 5. Sterling K, Brenner MA : Free thyroxine in human serum : Simplified measurement with the aid of magnesium precipitation. J Clin Invest 45 : 153-163, 1966 6. Helenius T, Liewendahl K : Improved dialysis method for free thyroxine in serum compared with five commercial radioimmunoassays in nonthyroidal illness and subjects with abnormal concentrations of thyroxin-binding globulin. Clin Chem 29: 816-822, 1983 7. Csako G, Zweig MH, Benson C, et al: On the albumin-dependence of measurements of free thyroxine, II. Patients with nonthyroidal illness. Clin Chem 33 : 87-92, 1987 8. Thorson SC, Wilkins GE, Schaffrin M, et al: Estimation of serum-free thyroxine concentration by ultrafiltration. J Lab Clin Med 80: 145-154, 1972 9. Faber J. Rogowski P, Kirkegaard C, et al : Serum free T4, T3, rT3, 3,3'-diiodothyronine and 3',5'-diiodothyronine measured by ultrafiltration. Acta Endocrinol 107: 357-365, 1984 10. Weeke J, Boye N, Orskov H : Ultrafiltration method for direct radioimmunoassay measurement of free thyroxine and free triiodothyronine in serum. Scand J Clin Lab Invest 46: 381-389, 1986 11. Sophianopoulos J, Jerkunica I, Lee CN, et al : An improved ultrafiltration method for free thyroxine and triiodothyronine in serum. Clin Chem 26 : 159-162, 1980 12. Shannon N, Woolf PD : Determination of free thyroxine in serum by ultrafiltration : Validation of a method and preliminary results. Clin Chem 30: 1770-1773, 1984 13. Wang YS. Hershman JM, Pekary AE: Improved ultrafiltration method for simultaneous measurement of free thyroxine and free triiodothyronine in serum. Clin Chem 31 : 517-522, 1985 14. Konno N, Nakazato T. Hagiwara K, et al : A comparison of measurements of serum free T3 concentration by equilibrium dialysis, free T3 index, and T3 : TBG ratio in thyroidal and nonthyroidal illnesses. Acta Endocrinol 103 :501-508, 1983 15. Lecureuil M, Crouzat-Reynes G, Besnard JC, et al : Application du dosage radioimmunologique de la "thyroxine-binding globulin" a la determination de la concentration de thyroxine plasmatique libre. CR Acad Sc (Serie D) 286: 81-83, 1978 16. Green AM, Marshall JS. Pensky J, et al : Thyroxine-binding globulin: Characterization of the binding site with fluorescent dye as probe. Science 175: 1378-1380, 1972 17. Prince HP, Ramsden DB: A new theoretical description of the binding of the thyroid hormones by serum proteins. Clin Endocrinol 7 : 307-324, 1977 18. Cheng SY, Pages RA, Saroff HA, et al : Analysis of thyroid hormone binding to human serum prealbumin by 8-anilino-1-naphthalene sulfonate fluorescence. Biochemistry 16: 3707-3713, 1977 19. Tabachnick M : Thyroxine-protein interaction. IV. Thermodynamics values for the association of T4 with human serum albumin. J Biol Chem 242 : 1646-1650, 1967 20. Steiner RF, Roth J, Robbins J : Binding of thyroxine by serum albumin as measured by fluorescence quenching. J Biol Chem 241 : 560-567, 1966 21. Robbins J, Rall JE: The iodine-containing hormones, in Hormones in Blood, pp. 575-688, Gray CH, James VHT (eds.), Academic Press, London, 1979 22. Konno N, Morikawa K : Seasonal variation of serum thyrotropin concentration and thyrotropin response to thyrotropin-releasing hormone in patients with primary hypothyroidism on constant replacement dosage of thyroxine. J Clin Endocrinol Metab 54 : 1118-1124, 1982 23. Taguchi S, Konno N, Taguchi H : Immunoturbidimetric assay of serum albumin and thyroxine-binding prealbumin concentrations. Clin Endocrinol (Japanese) 34: 817-822, 1986 24. Kamikubo K. Komaki T, Nakamura S, et al : Theoretical consideration of the effects of dilution on estimates of free thyroid hormones in serum. Clin Chem 30: 634-636, 1984 25. Geiseler D, Ritter M : Effect of sample dilution on measurements of free (unbound) hormones. Clin Chem 30: 28-32, 1984 26. Nelson JC, Weiss RM: The effect of serum dilution on free thyroxine (T4) concentration in the low T4 syndrome of nonthyroidal illness. J Clin Endocrinol Metab 61 : 239-246, 1985 27. Brown-Grant K, Brennan RD. Yates FE : Simulation of the thyroid hormone-binding protein interactions in human plasma. J Clin Endocrinol Metab 30: 733-751 1970