Four Sensitive Thyrotropin Assays Critically ... - Clinical Chemistry

CLIN. CHEM. 35/8, 1785-1789 (1989)

Four Sensitive Thyrotropin Assays Critically Evaluated and Compared Demetrios S. Sgoutas,’ Earl G. Barton,2 Margareta Hammarstrom,1 Philomena M. Pete,2 and Sandra A. Sgoutas’ We compared four sensitive procedures for thyrotropin (TSH)-Corning’s Magic-Ute, ElectroNucleonics’ Delfia, Baxter’s Stratus, Nichols’ Allegro-for their ability to cornpletely discriminate TSH concentrations in sera in euthyroidism, hyperthyroidism, and hypothyroidism. We evaluated the analytical and clinical performance of these procedures according to previously published criteria. All procedures we exarnined fulfilled the criterion stipulating 0.4 milli-int. unit/L, and all four procedures clearly distinguished hypothyroid from euthyroid subjects. In a hyperthyroid-euthyroid comparison, three of the procedures, the Magic-Lite, Delfia, and Allegro, differentiated the two with 97% accuracy, the Stratus procedure with only 90% accuracy. The procedures appeared to differ even more in the measurement of TSH in serum of patients undergoing suppressive treatment with thyroid hormones and in hospitalized nonthyroidally ill patients. The observed differences among procedures were thought to be related in part to a matrix serum effect, which is accentuated in samples from hospitalized patients. Historically, measurement ofthyrotropin (TSH) in serum has been generally accepted as the most valuable biochemical test for diagnosis of primary hypothyroidism.3 It has also been used for diagnosis of secondary (pituitary) and tertiary (hypothalamic) hypothyroidism. The recent availability of monoclonal antibodies has allowed the introduction of imniunometric assays with improved sensitivity, which reportedly enabled laboratories to measure suppressive influences on TSH secretion so precisely as to allow discrimination between low euthyroid and hyperthyroid populations. Several methods are now available for measuring TSH based on radiometric (1, 2), chemoluminometric (3), fluorometric (4), radial partition iminunometric (5, 6), and enzyme-amplified immunoassays (7, 8). However, results generated by these assays have shown that some but not all methods allow use of the TSH concentration in sera to completely discriminate between euthyroid and hyperthyreid patients. Furthermore, monitoring TSH in patients during suppression therapy with thyroid hormone (12), and other studies conducted in hospitalized populations, show remarkable differences among results from various methods (13-15). Here we have evaluated and compared the performance of four highly sensitive procedures relative to re-

Departmenta of Pathology & Laboratory Medicine, ‘Emory University, Atlanta, GA 30322, and 2Veterans Administration Medical Center, Decatur, GA. 3Nonstandard abbreviations:TSH, thyrotropin (thyroid-stimulating hormone);T4, thyroxin; Fr4, free thyroxin; T3, triiodothyronine; TBG, thyroxin-binding globulin; and IRP, International Reference Preparation. Received April 10, 1989; acceptedJune 2, 1989.

cently proposed criteria for sensitive TSH assays (16-18). Our aim was to provide information about these procedures in regard to their usefulness for measuring very low TSH concentration and in diagnosis of thyroid disorders. Materials and Methods Blood Specimens and Subject Groups The patients’ samples were taken from specimens received in the last six months. Clinical data were obtained from medical records or from the attending physician. Blood was sampled in the morning between 0800 and 1000 hours, and the serum was separated and stored at -40 #{176}C until analysis. TSH was measured in five groups ofpatients. Subjects in these groups had previously been classified on the basis of both clinical data and results of measurements of total triiodothyronine (T3), total thyroxin (T4), free thyroxin (Fr4), and thyrotropin (TSH) in serum. Group A consisted of48 subjects with no clinical signs of thyroid dysfunction and with normal results for biochemical assessment, including assays for T4, T3, and FT. Most of the subjects were blood-bank donors and hospital employees who were undergoing physical examination. Group B consisted of 28 samples from hyperthyroid patients who were chosen on the basis of biochemical proffles ofabove-normal T4, above-normal T3, normal TBG, and increased VF4. The diagnosis of hyperthyroidism was clinically confirmed in addition to biochemical signs other than TSH measurements. Some patients in this group, given the thyroliberin-stimulation test, showed no respouse, which can occur in patients with ophthalmic Grave’s disease or nodular goiter. Group C consisted of 14 patients with untreated primary hypothyroidism. Their concentrations of T4 and VF, in serum were below normal, and their TSH values were high (>8 milli-int. unita/L). Group D consisted of 18 patients who were receiving chronic treatment with thyroid hormone to suppress nontoxic goiter or thyroid cancer. Group E consisted of 36 hospitalized patients, some of them in an intensive-care unit with critical non-thyroidal illnesses for whom routine thyroid-function tests had been ordered. TSH Assays We evaluated the following commercial immunometric procedures involving monoclonal antibodies for human TSH: (I) Magic-Lite (Ciba-Corning Diagnostics Corp., Medfield, MA), (II) Delfia (ElectroNucleonics, Fairfield, NJ), (III) Stratus (Dade Division, Baxter Inc., Miami, FL), and (IV) Allegro (Nichols Institute, San Juan, Capistrano, CA). The Magic-Lite method is a two-site immunochemoluminescence assay involving two mouse monoclonal antibodies to TSH. An acridinium ester is used as chemiluminescence label, the final separation step is accomplished magnetically, and the chemiluminescence ofthe separated particles is measured in an automated luminometer. A detailed description of the method was given previously (3). CLINICAL CHEMISTRY, Vol. 35, No. 8, 1989 1785

reacting peptides. Human serum freed ofT4 and T3 was from Chemicon International (Los Angeles, CA) and drug-free serum from Bio-Rad Laboratories (Anaheim, CA).

Table 1. TSH Assay Sensitivity (Intra-Assay Analysis of 10 Replicates) TSH, mull-mt. unlt/L

Sensitivity Analyticala Reportedb Actuatc

Magic-Lit. 0.06 0.04 0.10

Detfis

Stratus

0.06 0.04 0.12

0.12

Other Methods

Allegro 0.10

0.08 0.22

0.04 0.12 Analysis of the zero standard of each kit, mean + 2 SDs. b Statistical valuesreportedby the manufacturer. Analysisof serumwith suppressed I

C

TSH, mean

+

2 SOs.

The nomsotopic Delfia is a dissociation-enhanced lanthanide fluoroimmunoassay. It incorporates a europium label and solid-phase separation (microtiter trays). A detailed description of the method, including description of the Arcus (Model 1230) time-resolved fluorometer, appeared previously (4). The Stratus radial partition fluorometric immunoassay, performed on the Stratus analyzer, was also described in detail elsewhere (5, 6). All pipetting, washes, incubation, and data reduction are performed automatically by the

We measured F’F4 concentrations in serum with an analog-based technique (Amersham International, Bucks., U.K.), total T4 by using Abbott’s TDx methodology (Abbott Diagnostics, North Chicago, IL), and total triiodothyronine cF3) by using Magic T3 immunoradiometric assay (Ciba Corning Diagnostics). Occasionally, and during thyroliberin testing, we measured TSH by a conventional RIA (Hybritech Inc., San Diego, CA). The between-batch precision ofthe VF4 assay was 5.2% at 16.5 pmolJL and 6% at 36 pmol/L; that of T3 was 5.1% at 1 nmolJL and 43% at 2 nmoIJL; and that ofTSH was 4.8% at 1.5 milli-int. unitlL. Reference values were 9-26 pmoIJL for FF,, 1.0-2.5 nmol/L for T3, and 0.2-4./milli-int. units/L for TSH. Statistical methods. Regression analysis, and Student’s t-test were carried out with statistical programs for the Apple-McIntosh computer. Results

instrument.

Table 1 shows the sensitivities

The Allegro procedure is an immunoradiometric assay involving three monoclonal antibodies with specificity for three different sites of the PSI! molecule. One of the monoclonal antibodies is labeled with 125!, and the other two are coupled to biotin. The addition ofan avidin-coated plastic bead allows specific and quantitative binding and separation of the bound hormone. The procedure has been previously deathbed in detail (9). We counted radioactivity with a Crystal H (Packard Instrument Co., Downers Grove, IL), which reduces the data for immunoradiometric assays on-line by using a spline function with an automatic smoothing factor. Procedures I, III, and IV involve calibrators that are standardized with the World Health Organization’s 2nd International Reference Preparation for Human TSH (IRP 80/588), whereas the calibrators in procedure II are standardized with the 1st Standard (IEP 68/38). All assays were performed as recommended by the manufacturer. However, because we desired to measure very low TSH concentrations but the dose-response curve was not linear between the zero-point and the lowest standard point, we added an extra low standard to all procedures. We prolow standards by serially diluting the lowest standard provided with the kit by the zero standard ofthe corresponding kit. We did not test for hormone cross-reactivities, but data supplied by the manufacturers demonstrated no significant cross-reaction with a wide variety of possible cross-

ofthe

four procedures.

We

assessedanalytical sensitivity (19, 20) by replicate (n = 10) measurements of the zero standard supplied with each procedure and derived a mean and standard deviation (SD) from which we calculated analytical sensitivity by interpolating the mean plus 2 SDs from the standard curve. The analytical sensitivities for procedures I, II, III, and IV were 0.06, 0.06, 0.12, and 0.1 milli-int. unit/L, respectively. These values were obtained with intra-assay CVs ranging from 10% to 12%. Assaying pooled sera from thyrotoxic patients who did not respond to thyroliberin stimulation yielded a more meaningful estimate of actual sensitivity: values were 0.1, 0.12, 0.22, and 0.12 milli-int. unitlL for procedures I, II, III, and IV, respectively, with interassay CVs ofl2% to 15% (21). In this study any TSH value of less than the actual sensitivity of a procedure was considered undetectable for that procedure. Following the example of other investigators, we make the distinction between analytical and actual sensitivity because human serum as matrix is different from the zero standard of commercial kits. In fact, the actual sensitivity with an interassay CV of 12% to 15% is closer to the “functional” methodologic sensitivity limit, as previously defined (16, 1 7, 20, 21). Table 2 shows the within- and between-assay precision of analyses for pooled patients’ sera containing a low concentration of TSH (0.2, 0.5, and 1.0 milli-int. unitfL) and of “Lyphocheck” controls (1.6 and 9.1 milli-int. units/L; BioRad Labs). Each pooled serum and commercial control was

Ta ble 2. Prec isbn of the Four TS H Assays CV, % TSH cones, mull-mt unlts/L

0.2 0.5 1.0 1.6 9.1

Maglc-Ute

Detfia

Allegro

Stratus

Within

Between

WIthIn

Between

WithIn

Within

Between

8.8 5.5

10.2 6.2

9.2 6.5

10.9 7.4

7.5 4.1

9.8 5.8

10.3 8.5

12.1 9.3

5.1

5.8

5.8

6.2

4.0

5.5

6.1

7.2

3.1

4.2

4.6

4.8

3.3

4.2

5.8

6.2

2.5

3.1

2.4

2.9

3.3

3.9

2.6 3.4 a Pools and their assays aredescribedin the text. 1786 CLINICAL CHEMISTRY, Vol. 35, No. 8, 1989

Between

Table 3. Means, SDs, Normal Range, and Cutoff Values with the Four Assays Untransformed TSH, mull-mt. units/I. Procedure

Mean

SD

Magic-Ute Delfia

2.09 1.88

1.36

Ln-transformed TSH

Normalrang. 0.45-5.4 0.40-5.3

1.15 1.62

Mean 0.57 0.46 0.85 0.64

Stratus 2.75 0.45-6.0 Allegro 2.20 1.19 0.46-5.4 a The corresponding antilog (es) to 3 naturallog SD below themeannaturallogvalueforthe healthy population.

assayed 10 times with each procedure for the calculation of the within-assay CV and seven times in duplicate over a period of 10 days for the calculation of the between-assays Cv. Table 3 shows the TSH values for 48 euthyroid subjects measured by the four procedures. Procedure II had the lowest and procedure III the highest TSH values. The normal ranges were practically identical and all had a nongaussian distribution. Statistical analysis of the natural-log-transformed values gave a normal distribution with calculated cutoffvalues for normal of0.31, 0.28, 0.45, and 0.36 for procedures I, II, III, and IV, respectively. Serum pools with TSH concentration near the lower limit of normal values (0.4 milli-int. unit/L) gave C of 10-12% and the mean lower limit value (log-transformed) minus 2.6 SD gave values of 0.31, 0.26, 0.44, and 0.36 milhi-int. unit/L for procedures I, II, III, and IV, respectively. These values were comparable with the cutoff values (Table 3) and represent the expression of variability at the low normal limit (16). The correlation among results by all four methods was assessed by analysis by all methods of 88 patients’ samples covering a range ofresults from 0.4 to 45 milli-int. unitsfL. Regression of these samples yielded the following: MagicLite = (0.921 x Delfia) + 0.328, r = 0.973; Magic-Lite = (0.805 x Stratus) + 0.120, r = 0.879; Magic-Lite = (1.07 x Allegro) + 0.12, r = 0.923; Delfia = (1.29 x Stratus) + 0.530, r = 0.963; Delfia = (1.08 x Allegro) 0.1, r = 0.910; and Stratus = (1.057 x Allegro) + 0.12, r = 0.946. Figure 1 shows results for patients in groups A, B, and C. For group A, and in procedure I, 20 patients’ samples had undetectable TSH concentrations; seven patients had detectable but below normal TSH, and one sample was in the normal range. In procedure II, TSH in 19 samples was undetectable, detectable but below normal in eight, and in one sample was in the normal range. Procedure III gave -

DELFIA

I Ill II

I

I 0 0

Normal

co -

cmo

a I 0

I D U

MAGIC-IJTE

.1 1 10 100 TSH Ievels(mllll-lnt.unlts/L) Fig. 1 TSH values for hypothyroid,hyperthyroid,and normal subjects, as measuredby all four procedures Eachvalueis the meanof duplicatedeterminatIons

0.45 0.36

00

DELF1A

.01

.

-

)

Hypathyroiod

MAGIC-UTE

0.28

I

-

STRATUS

Hypedhyrod

0.31

TSH concentrations in 12 samples, detectable but below normal TSH in 13, and three were in the normal range. With procedure IV, 20 samples had undetectable TSH concentrations, seven were detectable but below normal, and one was in the normal range. The same sample appeared to be unexpectedly normal in all four procedures. Figure 1 also shows that all hypothyroid patients had TSH concentrations above the high reference limit (>7.0 milliint. unitsfL) and none in the normal range. Figure 2 shows the distribution of TSH concentration among patients taking thyroid hormones for suppression therapy (group D). In procedure I, 15 of 18 samples had undetectable TSH concentrations; the other three were detectable but below the normal range. In procedure II, 16 had TSH concentrations that were undetectable and two were detectable but below normal. In procedure HI, 12 were undetectable; the remaining six were detectable and clearly in the normal range. In procedure IV, 16 were undetectable and two were detectable. With a few exceptions, the same samples had undetectable TSH concentrations by all procedures. Figure 3 shows the distribution of TSH concentrations among patients with major nonthyroid illness (group E). In procedure I, eight of 36 patients had undetectable TSH concentrations, eight had detectable TSH, and 20 were in the normal range. The respective distribution of TSH values in procedure H was 10, 8, and 18; in procedure III, 5, 11, and 20; and in procedure IV, 6, 12, and 18. When serum TSH concentrations were grouped according to serum T4 and P3 concentrations [subgroup I (n = 22) low serum T3, normal serum ‘F4;subgroup H (n 14) low serum T3 and low serum TJ, the differences were not statistically significant. For subgroup I, the mean serum TSH in the Delfia assay was 1.4 (SD 1.4) milli-int. unitsfL, and for subgroup H, it was 1.5 (SD 1.6) milli-int. unitsfL. Most patients in this group did not undergo thyroliberin testing.

) 0 Of)

Cutoff valuea

undetectable

ALLEGRO

liiiIllI I

SD 0.58 0.58 0.55 0.55

.01

.1

TREATED

10

TSHlevels (mIIIInL unft. /I)

Fig. 2. TSH results from treated patients compared with thosefrom

controls Each point

is the averageof twodeterminations CLINICAL CHEMISTRY, Vol. 35, No. 8, 1989 1787

cx

0 cocDDttInEmtlr1ttfDcDO I

AU1#{176}RO

)

0

0 0 arix,

STRATUS

ID OQ cD 0

ccacoco cr.rzD tWcX1OltEtXD

SICK EUT)IYROZI

DEI.RA

0

0

OCOtDLXIDO

MAWC.IJTE

I

.0t

ISH I.,.Ii

IS

lmKII-InI. nII,dL)

Fig. 3. TSH concentrationsin nonthyroidallyill patientscompared with those in controls Eachpointis the averageof two determinations Finally, to evaluate the matrix effect, we tested horse, rabbit, bovine, and mouse sera; serum free of T3 and T4; and commercial drug-free serum. The results, in apparent human TSH miuhi-int. unitsfL, were respectively: MagicLite 0.11, 0.33, 0.04, 0.09, 0.16, and 0.05; Delfia 0.01, 001, 0.01, 0.05, 0, and 0.02; Stratus 0.04, 3.12, 0.28, 2.76, 1.34, and 1.85; and Allegro 0.21, 0.03, 0.02, 0.30, 0.05, and 0.1. Discussion We now discuss our results relative to these criteria as recently described by Klee and Hay (16) for analytical performance and clinical usefulness of sensitive TSH assays. The most important criterion with regard to analytical performance requires that the assay gives values for serum TSH that do not overlap between the limits ofvariability of the lower value for normal and the assay sensitivity limit (16, 1 7). Results in Tables 1 and 3 clearly show no overlap between lower normal values and assay sensitivities. This suggests that all four procedures meet the first requirement, that they can justifiably be called “sensitive,” and that they merit further clinical trials with hyperthyroid patients. Another criterion refers to the clinical usefulness of the assay, such that at least 95% of hyperthyroid patients should have values below the normal range. This criterion initially required that 95% ofhyperthyroid patients should give undetectable TSH values. However, as the technology improved, undetectable values became detectable; consequently, this criterion has been changed to allow a decision level not based on assay detection limit. Hence, the American Thyroid Association’s Nomenclature Committee has defined a sensitive TSH assay as one in which serum from 95% ofthe clinically hyperthyroid patients gives results >3 SD below the mean value (alter logarithmic transformation) for serum from normal subjects (22). Our results indicate that in procedures I, II, III, and IV, 96%, 97%, 90%, and 97%, respectively, of hyperthyroid patients gave TSH concentrations >3 SD below the mean normal value (after logarithmic transformation). Therefore, all procedures except procedure III met this requirement; procedure III also gave more TSH values overlapping with the normal range. Carayon et al. (10), evaluating the ability of five sensitive TSH kits to distinguish euthyroid subjects from hyperthyroid patients, reported undetectable TSH concentrations in euthyroid subjects and detectable TSH concentrations in 1.3% to 8.4% of the hyperthyroid subjects. Similarly, Rodriguez-Espinosa et al. (11) reported that four 1788 CLINICAL CHEMISTRY, Vol. 35, No. 8, 1989

ofthe six kits they evaluated distinguished clearly between hyperthyroid and euthyroid subjects. In contrast, Hershman et al. (12) found that all five commercial immunometric kits they evaluated clearly distinguished hyperthyroid from euthyroid patients. A third criterion requires that at least 95% of clinically euthyroid subjects should have basal TSH concentrations within the established normal range. All four procedures met this requirement. Two other criteria require that at least 95% of patients who have a normal TSH should respond to thyroliberin testing and that at least 95% of patients who have an absent response to thyroliberin should have basal TSH values below the lower normal limits. TSH responses to thyroliberin testing were not studied here because appropriate documentation has been presented in the medical literature for all four procedures (4-9). In addition to the comparison of the four assays by the proposed criteria by Klee and Hay (16), we also included a group of patients undergoing suppressive treatment with thyroid hormone. These patients had either undetectable or low TSH concentrations by all four procedures. The majority of these patients had high FT, values, and any who were tested showed no response to thyrohiberin. Four patients, however, had detectable TSH concentrations by procedures I and IV and normal values by procedure III, suggesting some method dependency of the results. Spencer (13) noted that some patients on Synthroid (T4) suppression for thyroid carcinoma may display a measurable thyroliberin response when basal TSH concentrations are undetectable. Klee and Hay, on the other hand, using a different assay, suggested that this response is rarely seen (16). Nevertheless, and despite method dependency, sensitive TSH becomes the test of choice for monitoring thyroid therapy (22). Another possible application of any sensitive TSH method is in the evaluation of thyroid function in hospitalized patients with nonthyroidal illness. Several previous studies have addressed the problem in detail (26-28). To further study the clinical usefulness of the procedures under evaluation, we analyzed sera from 36 hospitalized patients. The wide range of TSH values and the lack of correlation of TSH with T4 and T3 concentrations cast doubts on the clinical usefulness of sensitive TSH procedures for differentiating sick euthyroid patients from those with true thyroid disorders. Possible explanations for this include individual abnormalities of the hypothalamic-pituitary-thyroid axis, concomitant drug therapy such as dopamine or glucocorticoids (23-28), or methodological problems associated with assays. For patients with nonthyroidal illness, procedures I, II, and IV showed a higher number of equivocal undetectable TSH results than did procedure III. This difference clearly suggests a loss of adequate specificity of one or more of the procedures under study. Some of the factors limiting the specificity of a procedure are specificity of antibodies or matrix effects (23). Other investigators have extensively studied specificity of antibodies and explained discrepancies in TSH concentrations on the basis ofantibody differences (23). Matrix effects have received little attention, perhaps because of the technical difficulties involved (23, 29, 30). In our evaluation, tests of some animal and other sera showed that all procedures were affected to a certain extent (with the Stratus procedure showing the greatest effect). This might explain how

the same sample could give undetectable TSH concentrations in procedures I, H, and IV and detectable or even normal in the Stratus procedure. We believe that further investigation is warranted to evaluate more fully the matrix effect on these and other sensitive assays. References 1. Seth J, Kellett HA, Caldwell G, et al. A sensitiveimmunoradiometric assay for serum thyroid stimulating hormone: a replacement assay for the thyrotrophin releasing hormonetest? Br Med J 1984;289:1334-6.

2. Toft AD. Use of sensitive immunoradiometric assay for thyrotropin in clinical practice. Mayo Clin Proc 1984;63:1035-.42. 3. BounaudMP, BounaudJY, Bouin-Pineau MH, et al. Chemiluminescence ilnmunoassay of thyrotropin with acridinium-eaterlabeled antibody evaluated and compared with two other immunoassays. Clin Chem 1987;33:2096-100. 4. Kaihola HL, hjala K, Viikari J, et al. Determination of thyrotropin in ser.irn by time-resolved fluoroiminunoassay evaluated. Clin Chem 1985;31:1706-9. 5. Rugg JA, Flea CW, Dawson SR, et al. Automated quantification of thyrotropin by radial partition immunoassay. Clin Chem

1988;34:118-22. 6. Giegel JL, Brotherton MM, Cromn P, et al. Radial partition immunoassay. Clin Chem 1982;28:1894-8. 7. Clark PMS, Price CP. Enzyme amplified imniunoassays. A new ultrasensitive assay of thyrotropin evaluated. Clin Chem 1986;32:88-92.

8. Pekary AE, Turner LF, Hershman JM. New immunoenzymatic assay for human thyrotropin compared with two radioimmunoassays. Clin Chem 1986;32:511-4. 9. Wood WG, Wailer D, Hantke V. An evaluation of six solidphase thyrotropin (TSH) kits. J Clin Chem Clin Biochem 1985;23:461-71. 10. Carayon P, Martino E, Bartalena L, et al. Clinical usefulness and limitations of serum thyrotropin measurement by “ultrasensitive” methods. Horm Res 1987;26:105-17. 11. Rodriguez-Espinosa J, Mora-Brugues J, Ordonez-Llanos J, et al. Technical and clinical performance of six sensitive immunoradiometric assays of thyrotropin in serum. Clin Chem 1987;33:1439-45. 12. Hershman JM, Pekary EA, Smith VP, et al. Evaluation of five high-sensitive American thyrotropin assays. Mayo Clinic Proc 1988;63:1133-9. 13. SpencerCA. Clinical utility and cost-effectivenessof sensitive thyrotropin assaysin ambulatory and hospitalized patients. Mayo

Clin Proc 1988;63:1214. 14. Spencer C, Elgen A, Shen D, et al. Specificity of sensitive

assays of thyrotropin (TSH) used to screen for thyroid disease in hospitalized patients. Chin Chem 1987;33:1391-6. 15. Bayer MF, Macoviak JA, McDougall hR. Diagnostic performance of sensitive measurement of serum thyrotropin during severe non-thyroid illness: their role in diagnosis of hyperthyroidiam. Clix Chem 1988;33:2178-84. 16. Klee GG, Hay ID. Assessmentof sensitivity of thyrotropin assays for an expanded role in thyroid function testing: proposed criteria for analytiCal performance and clinical utility. J Cliii Endocrinol Metab 1988;64:461-71. 17. Bayer MF. Performance criteria for appropriate characterizetion of(”highly sensitive”) thyrotropin assays [Letter]. Clin Chem 1987;33:630-1. 18. RossMD. New sensitive immunoradiometric assays for thyrotropin. Ann Intern Med 1986;104:718-20. 19. Rodbard D. Statistical estimation of the minimal detectable concentration (“sensitivity”) ofradioligand assays. Annal Biochem

1978;90:1-12. 20. Weeks I, Sturgess M, Siddle K, et al. A high sensitivity immunochemiluminometric assay for human thyrotropin. Clin Endocrinol (Oil) 1984;20:489-95. 21. COmmittee Ofl Nomenclature of the American Thyroid Association. Revisednomenclaturefor tests of thyroid hormonesand thyroid-related proteins in serum [Letterl. J Cm Endocrinol Metab 1987;64:461-71. 22. Ross DS. Subclinical hyperthyroidism: possible danger of overzealous thyroxine replacement therapy. Mayo Cliii Proc 1988;63:1223-9.

23. Musto JD, Pizzolante JM, Chesarone VP. A comment on thyrotropun measurement and evaluation [Opinion]. Cm Chem 1984;30:329-30. 24. Heinen E, Hermann J, Konigshausen T, et al. Secondary hypothyroidism in severe nonthyroidal illness. Horm Metab Res 1981;13:284-8. 25. Ratnaike S, Goodwin M, Deam D. Anomalous thyrotropun values. Clin Chem 1987;33:1213-4. 26. Vierhapper H, Laggner A, Waldhausl W, et al. Impaired

secretion of TSH in critically ill patients with low T4 syndrome. Acts Endocrunol (Copenhagen)1982;101:542-9. 27. Wehmann RE, Gregerman RI, Burns WH, et al. Suppression of thyrotropin in the low thyroxine state of severe nonthyroidal illness. N Engi J Med 1985;312:546-62. 28. Gow SM, Elder A, Caldwell G, et al. An improvedapproach to thyroid function testing in patients with non-thyroid illness.Clin Chins Acts 1986;158:49-58. 29. Garrett PE. The matrix effect. Part I [Letter]. Clin Immunol 1987;10:9. 30. Garrett PE. The matrix effect. Part II [Letter]. Clin Immunol 1987;10:67.

CLINICAL CHEMISTRY, Vol. 35, No. 8, 1989 1789