CLIN.CHEM. 33/8, 1439-1445 (1987)

CLINICALCHEMISTRY, Vol.33,No.8,1987 1439

Technical and Clinical Performance of Six Sensitive Immunoradiometric Assays of Thyrotropinin SerumJ. Rodriguez-Espinosa,J. Mora-Brugues,J. Ordoftez-Llanos,M. Cortes-RIus,J. M. Queraito-Companye, and F. Gonzaiez-Sastre

We evaluated the analytical and clinical performance of sixcommercial immunoradiometric assay kits for thyrotropin(TSH) in serum in 218 subjects, with and without thyroiddysfunction. Detection limits of the six kits were lower (from0.07 to 0.25 milli-int. unit/L) than that of conventional TSHRiA (0.7 milli-int. unitlL). Precision was adequate over a widerange of concentrations, although interassay CVs at verylow concentrations were good for only two kits (7.3% and13.1%). Results by all the kits correlated to about the samedegree with the TSH AlA (r = 0.92 to 0.98). All showed apositive correlation between both the basal and post-thyroli-berm (TRH)-stimulation values for TSH (r = 0.78 to 0.88),and all showed similar euthyroid reference ranges for basalconcentrations of TSH. With four of the six kits we couldclearly distinguish most of the hyperthyroid patients fromhealthy euthyroids; however, basal and post-TRH TSH val-ues were not sufficientto discriminate among groups ofpatients with different grades of thyroid hyperfunction.

Additional Keyphrases: initial screening of thyroid function

thyroliberin-stimulation test radioimmunoassay compared

effect of nonthyroidal illnesses measuring degrees of thyroid

impairment “kit”methods

In clinical practice, radioimmunoassay (RIA) of basalconcentrations of thyrotropin (TSH) in serum has beenextremely useful in the recognition and follow-up of cases ofprimary hypothyroidism.’ Most commercially availableTSH RIA reagent kits can measure TSH with adequateaccuracy and reproducibility in the above-normal range.However, these conventional TSH RIAs, now routinely usedin most clinical laboratories, are too insensitive and impre-cise to evaluate the low TSH concentration in serum ofhyperthyroid patients. They are thus inadequate to distin-guish hyperthyroid from euthyroid subjects when basal (i.e.,before treatment) serum samples are assayed.

Attempts to make these assays sufficiently sensitive andprecise for accurately measuring subnormal TSH concentra-tions have involved changing some assay conditions. Al-though such changes have led to the development of sensi-tive RIA methods, these are generally unacceptable forroutine clinical laboratory use, because they usually requirelong incubation times and are relatively laborious (1,2). Forthesereasons,stimulationof TSH production by administer-ing thyroliberin (TRH) has often been used as a majordiagnostic tool when hyperthyroidism is suspected and the

Servei de Bioquimica, Hospital de la Santa Creu i Sant Pau,Avgda. S. Antoni M. Claret 167, 08025 Barcelona, Spain.

‘Nonstandard abbreviations: H, thyrotropin (thyroid stimu-lating hormone); TRH, thyroliberin (thyrotropin-releasing hor-mone); maL4, immunoradiometric assay; FF4, free thyroxin; T3,triiodothyromne; and NT!, nonthyroidal illness.

Received March 17,1987;accepted May 12, 1987.

results of measuring thyroid hormone concentrations areequivocal (3). More recently, however, immunoradiometricassay (IRMA) kits for serum TSH have become commerciallyavailable. These kits are sensitive enough to distinguisheuthyroid subjects from hyperthyroid patients, and appearto be potentially useful as the initial screening test forthyroid disease, so that theTRH testmay be redundant (4-13).

We have evaluated six of these kits. Here we discuss theiranalytical and clinical performance in various groups ofsubjects, with and without thyroid diseases.

Materials and MethodsTSH Assays

We evaluated the following commercial RIA and IRMA

(kits 2-7) procedures for TSH: kit 1, double-antibody RIA“NHS-TSff’ (Diagnostic Products Corp., Los Angeles, CA);kit 2, TSH MaiaClone kit (Serono Diagnostic Ltd., Woking,Surrey, U.K.); kit 3, Tandem-R TSH (Hybritech Europe,S.A., Liege, Belgium); kit 4, TSH RIA 100 (Pharmacia AB,Uppsala, Sweden);kit 5, TSH-uu (Euro-Diagnostics By,Apeldoorn, The Netherlands); kit 6, Sucrosep TSH IRMA(Boots-Ceiltech Diagnostics Ltd., Berkshire, U.K.); and kit7, RIA-gnost hTSH (Behringwerke AG, Marburg, F.RG.).Table 1 lists the characteristics of the various kits,

including the calibration materials (International ReferencePreparations) used by the manufacturers. All assays wereperformed exactly as stated in the kit instructions, exceptforkits 4 and 5, where we prolongedthe incubation periodsto 18 h to obtain more complete binding. Also, for the TSHassays by kit 1 we prolonged both the first and secondincubation periods to 24 and 5 h, respectively (14). Allassays were performedby two operators, usingthreediffer-ent lots ofkitsforeach method.

We counted radioactivity with a Model 1271 RIA GammaCounter (LKB Wallac, Thrku, Finland), which reduced thedata forIRMA assays on-lineby usinga spline function withautomaticsmoothing factor. Results with the RIA kit werecalculated from a logit-log representation of the calibrationcurve.

Measurements ofThyroidHormones

We measured the concentrations offreethyroxin (FF4)and totaltriiodothyronine (T3) in serum by using commer-cialRIA procedures from Diagnostic Products Corp. (Coat-A-Count FF4 RIA kit and T3 double-antibody RIA kit,respectively). The between-and within-run CVs forserumFF4 were 5.3% and 4.0% at 4.5 pmol/L, and 5.0% and 3.5%at 25.7 pmol/L, respectively. The between- and within-runCVs for serum T3 were 4.0% and 2.0% at 1.5 nmol/L, and5.0% and 2.5% at 4.1 nmol/L, respectively.

Patients

Blood specimens were collected from 218 adults (bothsexes) attending our hospital between January and June

KIt 1 KIt 2 Kit3 KIt 4 Kit 5 KIt6 KIt 7

Calibrator 68/38 80/558 68/38 80/558 68/38 68/38 80/558Assay method RIA, double-antibody IRMA 2-site IRMA 2-site IRMA 2-site IRMA 2-site IRMA 2-site IRMA

Phase Liquid Solid(particle)

Solid(bead)

Liquid Liquid Solid(particle)

Solid(tube)

Antibodies Polyclonal Monoclonal Monoclonal Polyclonal Polyclonal Monoclonal MonoclonalSamplevol, L 200 200 200 100 200 100 200Pipeting steps 4 5 5 4 4 7 4Wash steps 0 1 2 0 0 2 1Reagents ready for use No Yes Yes Yes No Partly YesAssaytime,h 30 2-3 3-4 18 18 2_3c 23d

Incubationtemp Ambient Ambient 37#{176}C Ambient Ambient Ambient Ambient

Standardsn 7 7 6 7 7 8Range, milli-int. units/L 0-60 0-50 0-100 0-50 0-60 0-250Matrix(serum) Human Human Human Human Human BovineSummarized from manufacturers inserts, except for the incubation times we usedwith kits 1, 4, and5. b Involvessimultaneousreactionof sample with three

monoclonal antibodies; separation is achieved by adding antibodies to fluorescein isothiocyanate linked to magnetizable particles. cOrbital agitator. ‘Horizontalshaker.

70-50

Human

Table 1. MaIn Characteristics of the Seven Kits Used for Thyrotropin Assays a

1440 CLINICALCHEMISTRY, Vol. 33, No. 8, 1987

1986.Blood was sampled inthemorning between 08.00 and10.00 hours, and the serum was separated and stored at-20 #{176}Cuntil analysis (within three months). The subjectswere divided into the following groups:

Group 1: These 77 euthyroid subjects consisted of 57women and 20 men, ages 17 to72 years (mean 38.2, SD 15.3years). None had a clinical history of thyroid disease, andnone was using any medication known to affect results ofthe thyroid-function tests.

Group 2: Thirty-one untreated patients, 25 women andsix men, ages 19 to 78 years (mean 49,SD 14.8years), withsymptoms and signs of thyrotoxicosis due to Graves’ disease(n = 20) or to toxic nodular goiter (n = 11).

Group 3: Thirteen untreated patients,12 women and oneman, ages 34 to 78 years (mean 57, SD 12.8 years), withautonomously functioning nodular goiter, without clinicallydetectable manifestations of thyrotoxicosis.

Group 4: Thirty patients, 28 women and two men, ages 22to 60 years (mean 42, SD 11.5 years), who were takinglevothyroxin (100-225 pg/day) orally to suppress TSH secre-tion, as chronic treatment fornontoxicgoiteror thyroidcancer.

Group 5: Twenty-four patients, 22 women and two men,ages 17 to 72 years (mean 50, SD 14.6 years), with subclini-cal hypothyroidism. Six were taking antithyroid drugs, 10with primary hypothyroidism were taking low doses oflevothyroxin, and eight had had subtotal thyroidectomy.These patients were classified as having subclinical hypo-thyroidism on the basis of normal concentrations of FF4 inserum and an above-normal TSH response 30 mm afterTRH administration, with basal TSH values near the upperlimit of the reference interval (see below).

Group 6: Forty-threeeuthyroid patients, 23 women and20 men, ages 38 to 84 years (mean 63, SD 10.3 years), withvarious severe nonthyroidal illnesses (NTI), being treated inthe IntensiveCare Unit. The primary clinical diagnoseswere bacterial sepsis (n = 3), liver disease with ascitis (n =

7), malignancy (n = 8), hematopoietic cancer (n = 2),aplastic anemia (n = 1), uncontrolled diabetes mellitus (n =

4), congestive heart failure (n = 3), respiratory failure (n =

5), acute pancreatitis (n = 5), and chronic renal failure (n =

5). None of these patients was receiving therapeutic agentsknown to alter thyroid function, and none had a clinical

history ofthyroiddisease. Allhad abnormally low concen-trationsofT3 in serum.

All subjects in these groups had previously been classifiedas euthyroid or as having thyroid hyper- or hypofunction onthe basis of both clinical examination and results of mea-surements in serum ofFF4, T3, basal TSH (RIA, kit 1), andthe TSH response 30 mm after intravenous administrationof 400 ig of TRH. (The NTI patients did not undergo TRHtests.) TSH responses to TEll administration were consid-ered as subnormal or as exaggerated when results of theRIA method (kit 1) 30 mm after ‘FRH were <3.0 or >30.0milli-int. units/L, respectively. Normal reference valueswere based on results for 175 healthy euthyroid subjectsstudied in our laboratorywith the same commercial kit.Basal and TSH responses to TRH as measured by TSH-mMAassays were not used to determine the diagnostic categoriesof the subjects.

Results

Table 2 lists thedetectionlimitsand within-and between-run precision (CV) for all seven methods. The detectionlimits were calculated from the precision relative to the zerostandard (+2 SD and -2 SD for IRMA and RIA methods,respectively)(15), being 0.25milli-int.unitlLforkits2,3,and 5;and 0.7,0.10,0.09, and 0.07 forkits1, 4, 6, and 7,respectively.The within- and between-run precision analyses were

both carried out by assaying replicate samples of a specimenof pooled patients’ sera containing a very low concentrationofTSH, and control sera with low, intermediate, and highconcentrationsof TSH (“Lyphocheck,”lotno. 07300; Bio-Red Labs., Anaheim, CA). CVs were lowerforallthe IRMAkits than for the RIA procedure. Five ofthe six IRMA kitsshowed good within-run CVs (<7%) at low, intermediate,and high concentrationsofTSH. Kits 4,5, and 7 showed thelowestwithin-run CV (<10%) at very low TSH concentra-tions.Between-run CVs were good (<10%) forallthe IRMAkits at intermediateand high concentrations.Kits 4,5,6,and 7 gave the lowestbetween-run CV (<10%) at a lowconcentration, whereas kits 2 and 3 gave CVs of 11.7% and14.1%, respectively, at this same TSH concentration. Theonly procedure that showed a between-run CV <10% (7.3%)at a very low concentration of TSH was kit 7. Between-run

Methody

Kit7Method x Kit2 Kit3 KIt 4 KIt5 KIt 6

Kit 1 (n =

Meana12) 0.7 Kit 1

r 0.967CV, % Slope 0.86

Kit2 (n = 12) Intercept 0.39Mean Kit2CV, % r 0.993

Kit 3 (n =

Mean12) Slope

Intercept0.84

-0.01CV,% Kit3

Kit4 (n= 12) r 0.991Mean Slope 1.11CV, % Intercept -0.03

Kit5(n=12)Mean

Kit4r 0.993

CV, % Slope 1.06Kit6 (n = 12) Intercept -0.06

Mean Kit5CV, % r 0.993

Kit7(n =

Mean12) Slope

Intercept1.13

-0.35CV,% Kit6

rSlopeInterceptWe assayed77 serum samples with eachkit.

0.9550.79

-0.03

CLINICAL CHEMISTRY, Vol. 33, No. 8, 1987 1441

Table 2. Detection Limit and Precision (CV, %) of theSeven TSH Assay Kits

WIthIn-run CV, % Between-run CV, %DetectIon

Method IImIt I II III IV I II III IV

0.25

0.25

0.10

0.25

0.09

0.07

- 1.5 11.5 32 - 1.4 14.539- 9.0 7.0 6 - 15.0 13.5 11

0.5 1.4 15.0 30 0.5 1.9 13.43411.8 5.6 1.0 3 17.0 11.7 6.1 5

0.4 1.1 10.7 28 0.4 1.2 10.3 2616.0 11.9 2.3 3 17.6 14.1 6.1 4

0.4 1.5 10.3 27 0.4 1.7 11.5 286.6 6.2 2.9 3 16.0 7.0 6.8 5

0.4 1.6 13.0 35 0.6 1.8 14.1 369.9 3.3 4.9 6 19.7 9.4 7.7 7

0.6 1.5 10.3 28 0.6 1.5 10.8 2810.9 4.8 3.4 6 13.1 7.7 5.2 9

0.5 1.5 9.8 25 0.5 1.4 9.6256.0 1.3 3.5 3 7.3 4.9 5.8 4

miIIi-int.units/L. I, II, III, and IV: control sara with very low. low, intermedi-ate, andhighconcentrationsof TSH, respectively.

CVs for the remaining kits were 13.1% (kit 6), 16.0% (kit 4),17.0% (kit 2), 17.6% (kit 3), and 19.7% (kit 5).

All IRMA kits showed good linearity fora patient’ssamplecontaining 100 milli-int. units ofTSH per liter,when serialdilutionswith TSH-free serum were assayed.

We did not testforcross reactivity ofotherglycoproteinsor pituitary hormones in the nu,IA-TSH assays, but themanufacturers’ data (package inserts) indicate no signifi-cant cross reaction with a wide variety of possiblecross-reacting peptides.

To assess correlation studies with patients’ samples fromall diagnostic groups, we used Deming-debiased regression(16) to compare results ofiRMA procedures with those of theRIA technique (Table 3). TSH values <0.7 and >50 milli-int. unitsfL were excluded from the regression analysiswhen IRMA results were compared with RIA results. AllIRMA methods had approximately the same degree of corre-lation with the RIA procedure (r = 0.92 to 0.98, n 77).Table 3 also shows that all combinations between IRMA

methods showed similar correlations (r = 0.95 to 0.99, n =

77).

ClinicalStudies

Figure 1 shows basal TSH concentrations and theirresponses 30 mm after TEll administration as measuredwith the different kits in the serum of individual subjects,grouped according to their thyroid status. Table 4 showsserum FF4 and T3 concentrations in these same groups ofsubjects.

Regression analysis between basal and post-TEll TSHvalues obtained by all the IRMA kits showed a positive andsignificant correlation (P <0.01) between both TSH mea-surements (r = 0.845 for kit 2, r = 0.863 for kit 3, r = 0.868for kit 4, r = 0.780 for kit 5, r = 0.826 forkit6,and r = 0.886for kit 7), irrespective of clinical diagnosis.

Euthyroid subjects (groups I and 6). Of the healthyeuthyroid subjects, 43% showed undetectable (<0.7 milli-flit.unitfL) basal TSH concentrations as measured by the

Table 3. Regression Analysis (Deming Debiased) ofTSH Results Obtained by Various Techniques

0.9880.781.19

0.9780.760.42

0.9940.77

-0.04

0.9620.770.54

0.9690.770.65

0.991 0.9860.78 0.770.08 0.22

0.9891.020.12

0.9900.980.28

0.9840.980.15

0.9241.100.63

0.9511.27

-0.73

0.9651.41

-0.006

0.9531.39

-0.21

0.9581.48

-0.66

RIA procedure. The highestbasal serum TSH value foundby this method was 3.0 milli-int. unitsfL.

All thebasal serum TSH concentrations in healthy euthy-roid subjects were detectable by the nu procedures. Thelowest basal TSH value observed in this group of subjectswas obtained by kit 6 (0.23 milli-int.unitlL), being �0.25milli-int. unitlL with the remaining five mr kits. No basalTSH value >3.8 milli-int. units/L was found in any of the sixIRMA kits. According to one-way analysis of variance, signif-icant differences (P = 0.01) for basal TSH concentrationswere observed between the six IRMA kits. Paired Student’s t-tests showed no significant differences between kits 3, 4, 5,and 7. Basal serum TSH concentrations for kits 2 and 6 werenot significantly different, but both were significantly high-er (P <0.01) than values obtained with the other four kits.

The range of TSH responses at 30-mm post-TRH wassimilar for kit 3 (4.1 to 20.7 milli-int. units/L), kit 4 (4.1 to19.1 milli-int. units/L), kit 5 (3.7 to 19.3 milli-int. unitsfL),and kit 7 (4.4 to 20.7 milli-int. unitsfL), but wider for kit 1(3.5 to 27.6 milli-int. unitsfL), kit 2 (5.5 to 25.1 milli-int.unitWL), and kit 6 (3.2 to 25.5 milli-int. units/L).

Euthyroid patients with NTI showed a greater range ofbasal serum TSH concentrations than did healthy euthyroidsubjects, but not significantly different according to theMann-Whitney statistic (17). When measurements of se-rum TSH concentrations from all IRMA kits were groupedand analyzed according to serum FF4 values (subgroup I, n= 21, low serum T3, normal serum FF4; subgroup II, n = 22,low serum T3, low serum FF4), no statistically significantdifferences were found. Abnormally low TSH concentrationswere found in both subgroups of patients with NTI, al-though subnormal TSH concentrations were more commonin patients from subgroup II than in those from subgroup I(Table5).

Hyperthyroids (groups 2,3, and 4). Most patients in thesethree groups showed basal TSH concentrations below the

Sd.oI 3] *00 Saul 30.30 Saul 30.10 80.01 30.10 Saul 30.10 Saul

CR012 I 3032 2 0532 3 303*2 4 .P 5 .P 8

Kb, 2

10.30

0.30

2.50

3.30

0.30

5.30

3

3.30

0.50

0.23

C23

‘2:

0.30

3*2.5O

3.30

0.50

0.24

00.25

30.30

25.30

0.30

.8

12.50

.30

is

25

80.01 80.1, Saol 30.], Saul 30.], Saul 30*3, Saul 30 .ln Saul

I P 2 P 3 .P 4 003.2 5 .P 8

63, 6

Saul 30.1, Saul 30.10 Saul 30.1, Saul 30.10 Saul 33.3, Saul

COUP I 2 P 3 05352 5 COOP 5 33003.0 8

Basal 30.1, Saul 30.In Saul 33.3, Saul 30.3, Saul 30.1, Saul

0032 I P 2 0032 3 CR512 0251.2 5 00332 0

1442 CLINICAL CHEMISTRY, Vol. 33, No.8,1987

Bo.*I 30 0usd 30 lo 0u.oI 30 .1 0oI 3] .Iu 8usd 3D ‘flu Saul

043P 0032 2 030123 3032 4 2 5 COUP 8

13, 3

,

I!...

.

.... ..

;

:.

-----

....

:

.

:. ..

:#{149}.

Basal 30 .lu Saul 30.00 Bo.ul 30 .lu Saul 30 sIn Saul 30 Saul

05512 I 00512 2 330.2 3 0032 4 P 5 2 B

633 5

r

B’

:A

Kb II Kbr 7

Fig. 1. Basal and 30-mm post-TRH values(ordinate, milli-int.units/L) for thyrotropin (TSH) as measured with six IRMA kits in healthyeuthyroidsubjects (group 1), overtly hyperthyroid patients(group2), patientswithautonomously functioning nodular goiter (group 3), patients receivinglevothyroxin suppressive therapy (group 4), subclinicalhypothyroidpatients (group 5), and patients withseverenonthyroidalillnesses(group6)Brokenlines indicatethenormalreferenceintervalsfor the respectiveassays.The numberof subiectsis giveninparentheses.KIt2:Seiono;kit 3: 1-tybiltech; kit 4:Pharmacia;kit 5: Euro-Diagnostics;kit 6: Boots-Celltech;kit 7: Behringwerke.‘mIU, milli-Int.units

lowest reference intervalof allnu assays. All patientsfrom thesegroups showed subnormal TSH concentrationswhen sera were analyzedwith kit4.Kit 7 also showed goodperformance in differentiating these three groups fromhealthy euthyroid subjects; only basal TSH valuesin twopatients from the group on levothyroxin suppressive thera-py overlapped. Overlapped basal TSH concentrations be-tween patients from these three groups and those fromhealthy euthyroids were more frequent with kit 2 (three ofthe 13 patients from group 3, and two of the 30 from group4), and kit 6 (two of the 13 patients from group 3, and 10 ofthe 30 from group 4). The nu kitsthat showed the most

overlapping among healthy euthyroid subjects and patientswith thyroid hyperfunction were kit 3 (12 of the 31 fromgroup 2, eightofthe 13 from group 3, and 16 of the 30 fromgroup 4), and kit5(10 from group 2, nine from group 3, and20 from group 4).None of the clinically hyperthyroid patients (group 2)

produced measurable TSH response to TRH stimulationwhen assayed with kits 1 (RIA) and 2.The remaining mkits showed detectable, but subnormal, TSH responses asfollows(Figure1):14/31casesinkit3 (range <0.25 to 0.71milli-int. unitlL); 23/31 in kit 4 (range <0.1 to 0.33 milli-int.unitlL); 19/3 1 in kit 5 (range <0.25 to 1.02 milli-int. unit/L);

Table 4. EstImates for FT4 and T3 ConcentratIons InSera from Various Subject Groups

Mean * SD(and range)

Group ClInIcal conditIon n Fr4, pmol/L

1 Healthy euthyroid 77 17.7 ±3.0(12.0-25.0)

2 Overtly hyperthyroid 31 56.8 ±24.8(26.0-119.0)

3 Subclinical hyperthyroid 13 18.3 ±3.0(nodular goiter) (14.0-25.0)

4 Thyroxin-treated 30 27.7 ±4.6(suppressivedoses) (18.4-37.0)

5 Subclinical hypothyroid 24 15.7 ±3.1(11.4-20.7)

6 Nonthyroidalillness 43 13.0±5.1(1.6-23.0)

13, nmol/L

1.8±0.33(1.2-3.2)5.5 ±2.04(3.6-13.0)2.2 ±0.44(1.9-3.2)

1.6 ±0.33

(1.2-2.4)

(<0.31-1.1)

Table 5. Euthyrold PatIents wIth Nonthyroidal Illness Who Showed Abnormally Low Concentrations for SerumThyrotropln as Measured by Some or All of the IRMA KIts

TSH, mllll-Int. unlts/L

DIagnosIs Age Sex KIt 2 KIt 3 KIt4 KIt 5 Kft 6 KIt 7F4,

pmol/L13,

nmol/L

Bacterial sepsis 69 M 0.6 0.5 0.3 1.0 0.6 0.4 15.4 0.92Diabetesmellitus 57 F 0.5 0.25 0.38 1.1 0.5 0.4 21.3 1.17Malignancy 69 M 0.4 <0.25 <0.10 1.2 0.21 0.24 18.3 0.80Liver disease 64 F 0.7 0.25 0.6 1.4 0.8 0.6 21.0 1.07Respiratory failure 43 F <0.25 <0.25 0.15 0.8 0.5 0.4 20.0 0.98Bacterial sepsis 66 F 0.5 0.6 0.5 0.9 0.7 0.21 8.0 0.55Respiratoryfailure 55 F <0.25 <0.25 <0.10 1.2 0.14 0.08 1.6 0.31Heartfailure 59 M 0.4 0.6 <0.10 0.5 0.21 0.16 8.3 0.31Respiratoryfailure 60 F 0.8 <0.25 0.5 1.3 1.2 0.7 6.7 0.9Aplasticanemia 60 F <0.25 <0.25 <0.10 <0.25 0.10 0.10 10.3 0.67Bacterialsepsis 64 F <0.25 <0.25 0.12 <0.25 <0.09 0.08 6.7 0.78Heartfailure 62 M <0.25 0.25 0.26 1.0 0.33 0.18 10.6 0.50Referenceintervals: 0.29-3.6 0.29-3.8 0.42-2.9 0.45-3.0 0.23-3.8 0.25-3.3 12-25 1.2-3.3

Valuesin italicsindicateabnormally low concentrationsof serumTSH.

CLINICALCHEMISTRY, Vol. 33, No. 8, 1987 1443

4/31 in kit 6 (range <0.09 to 0.21 milli-int. urntfL); and 13/31 in kit 7 (range <0.07 to 0.11 milli-int. unitfL). DetectableTSH responses to TRH were observed more frequently inpatients from groups 3 and 4, these responses clearlyoverlappingbetween both groups when assayed with all sixIRMA kits (Figure 1). However, in spite of the finding ofdetectable TSH in response to TEl! stimulation, none ofthese kits showed TSH responses overlapping those ob-tained in the group of healthy euthyroid subjects.

Subclinical hypothyroids (group 5). All patients classifiedashaving subclinical hypothyroidism according to the exag-gerated TSH responses to TRH stimulation tests, as man-sured with the conventional RIA procedure, also showedabnormally high TSH responses when these were deter-mined by allsix isit& kits. In this group ofpatients,basalserum TSH concentrations as measured by RIA fell into thereference ranges in 41.6% of patients, while with the IRMA

procedures these basal values were normal in 54% (kit 3),33% (kit 6), 25% (kits 2 and 5), 21% (kit 7), and 12.5%(kit 4).

DIscussion

The advantages over conventional RIA procedures ofmeasuring serum TSH by the nua methods we studied areclearly demonstrated by our results. The high degree of bothanalytical sensitivity and precision, together with greaterpracticability and speed, makes these assays superior toroutinely used conventional TSH RIA.

Sensitivity ofthe TSH nms evaluated here was 2.8- to12-fold greater than that of our routine TSH RIA, being<0.5 milli-int. unit/L for all of them, which isthe detectionlimit recommended fora clinically suitable TSH assay (2).The variability of the IRMA methods showed that all the kitsstudied are adequately precise over a wide range of concen-trations, and that all have better within- and between-runprecision than does theRIA technique. Nevertheless, whenwe compared precision at very low TSH concentrationsbetween IRMA kits, only two (kit 6 and kit 7) showed anacceptable between-run CV(13.1% and 7.3%, respectively),while the CVs for the remaining kits were >15%. One of themain points emerging from this study was the excellentsensitivity of kit 7, which also proved to be the most simple,rapid, and precise.

All nu kits gave similar euthyroid reference intervals,although there were differences between standard prepara-tions for calibration. No kit showed basal TSH values>4.0milli-int. unitsfL, consistent with an upper limit of normalof 3.0 or 4.0 milli-int. units/L as recently proposed (18). Thelowest limits of the normal reference interval ranged from0.23 (kit 6) to 0.45 (kitS) milli-int. unitfL. The low referencelimits for kit.s 2, 3, and 7 agree with results recentlyreported by others using the same kits (5,8, 11), but lowervaluesineuthyroidsubjects have been found with kit 6(11,19). Regarding kit 4, we observedthe same low referencelimit as that given in the kitprotocol,but this was not sowith kit 5.

Our results confirming the value of the serum TSH asdetermined by some IRMA methods in discriminating be-tween healthy euthyroid and overtlyhyperthyroid subjectsare in agreement with previous reports (5-13,19). However,notallthen kitsstudied here performed equally well indifferentiating these two groups of subjects. Basal serumTSH estimates with kits 2,4,6, and 7 clearly distinguishedhyperthyroid patients from healthy euthyroid subjects, butkits 3 and 5 showedconsiderableoverlapping between thetwo groups of subjects. With kit 4, in which polyclonalantibodies are used, thyroid hyperfunction was identified inall patientsin the three groups by detecting subnormalvalues for TSH. This was confirmed by the finding ofsubnormal orabsent PSI! responses to PHI! administrationin all these patients. Kit 7 also showed excellent perform-ance in differentiating healthy euthyroid subjects frompatients with different grades of thyroid hyperfunction,

1444 CLINICALCHEMISTRY, Vol. 33, No. 8, 1987

because basal TSH values in only two patients from thegroup on levothyroxin suppressive therapy clearly over-lapped. Kit 6 gave good results with respect to analyticalprecision, but basal TSH concentrations for the groups ofpatients with lesser grades of thyroid hyperfunction consid-erably overlapped with those for healthy euthyroid subjects.Furthermore, this kit was technically themost complicated.

Our results indicate that most patients with differentgrades of thyroid hyperfunction (groups 2, 3, and 4) can bedistinguished from healthyeuthyroids (group 1) when basalTSH concentrations are measured with kits 2,4, 6, and 7.However, neitherbasal nor TSH responses to TRH weresufficient to distinguish among the three groups ofpatientswith different grades of thyroid hyperfl.mction. Many pa-tients from these groups showed subnormal but detectableincrements in TSH in response to TEl! administration, thusindicating absence of total TSH suppression. Moreover, thefinding ofan undetectable basal TSH values did notallowusto predict the absenceofT$H response to TRH in many ofthese patients, particularly those patients with nodulargoiter and those on suppressive therapy who showed thyroidhormone concentrations within the normal reference inter-val. The biological significance of such small TSH responsesto TRH remains to be established, and our findings andthose of othersappear to indicatethat measurements ofbasal concentration of TSH in serum may be insufficient toprove completesuppressionofTSH in thesecases (20-22),and that use of TRH tests may be convenient in some ofthese patientswhen IRMA kits are used forTSH measure-ments. Interestingly, some patients with a normal basalTSH concentration showed an exaggerated response to TRH(subclinical hypothyroidism), confirming that theTRH testmay be an important tool in the diagnosis of lesser grades ofhypothyroidism (23,24).The disagreementobservedbetween TSH concentrations

as measured with the differentkitsinaliquotsofthe sameserum specimens couldbe explained, at least in part, by thedifferences in standard preparations for calibration,thenature and composition of the TSH-free matrix, and theexperimental conditions and procedures.

A common problem in clinical practice is the patient withNTI who is suspected ofbeing hyperthyroid and forwhomresults of routine thyroid-function tests are often equivocalor difficult to interpret. To be able to demonstrate whichbasal TSH concentrations fallwithinreferenceintervalsinsuch cases would be ofgreatclinicalvalue,becauseitwouldhelp to resolvethis problem by obviatingotherdiagnosticprocedures (19). However, the finding of subnormal orundetectable TSH concentration in theserum ofsome ofourpatients with NTI indicates a loss of diagnostic specificity,as subnormal basal values for TSH obtained with thesekitsin this group of patients are similar to those found in ourpatients with thyroid hyperfunction. This observation isparticularly relevant from both a clinical and theoreticalpoint of view, because it indicates that measurement only ofbasal TSH concentrations in these patients may not sufficeto detect those with thyroid hyperfunction.

It has been postulated that the low thyroid hormoneconcentrations observed in patients with severe NTI some-times reflect a central hypothyroidism caused by reducedpituitary secretion of TSH, representing adaptation to se-vere systemic illness (25, 26). Serial TSH determinationswith an unusually sensitive TSH RIA have recently beenreported in a prospective study of patients who were under-going bone-marrow transplants and showed an illness-

associated decrease in circulating TSH (27). Although fol-low-up studies were not carried out in our group of patientswith NTI, the fact that most of those with subnormal TSHconcentrations also had low values for serum FF4 and T3could suggest that suppression of TSH secretion contributesto this subnormal thyroid hormone concentration in severe-ly ill patients. However, the serum TSH concentration wasnot subnormal in most of our patients with severe NT!, andsome of our patients with NTI and normal serum FF4 alsohad subnormal serum TSH values.

We conclude that the major advantage of someof the mr&kits we studied is their ability to distinguish healthyeuthyroid from hyperthyroid subjects. The use of such kitsmay be a powerful tool in the diagnosis of thyroid disease,both by excluding thyrotroxicosis and by identifying pa-tients needing further studies. Moreover, data from thisstudy suggest that by using certain of these kits to measurebasal TSH concentrations, TRH tests should rarely benecessary to confirm a diagnosis of hyperthyroidism. Thehighly significant correlation found between the basal valuefor TSH and the TSH response to TRH would support this.Nevertheless, it is necessary to take into account that asubnormal basalTSH value does not necessarilyindicateoverthyperthyroidism.Basal undetectableor subnormalTSH concentrationscan also be found in other clinicalconditionsinthe presenceofnormal,high borderline(thy-roidsuppressivetherapy,autonomouslyfunctioningthyroidnodules, or apparent remission after antithyroid therapy),or abnormally low (nonthyroidal illness) concentrations ofthyroid hormone. In these cases it would at least be neces-sary to estimate FF4 as wellas to measure basal TSH, andin cases where complete suppression of TSH secretion mustbe proved, TRH tests are convenient. Lastly, if a patient issuspected of having subclinical hypothyroidism, the TRHtest should also be used when the value for basal TSH isnormal.

We gratefully acknowledge the Spanish representatives of Ser-ono, Hybritech, Pharmacia, Euro-Diagnostics, Boots-Ceiltech, andHoechst (Behringwerke) for supplying the kits usedin this study.

References1. Wehmann RE, Rubinstein HA, Nisula BC. A sensitive, conve-nient radioimmunoassay procedurewhich demonstrates thatserumh-TSH is suppressed below the normal range in thyrotoxic patients.Endocr Res Conimun 1979;6:249-55.2. Wide L, Dahlberg PA. Quality requirements of basal s-TSHassays in predicting an s-TSH responseto TRH. Scand J Clin LabInvest 1980;40(Suppl 155):l01-l0.3. Zilva JF, Pannall PR. Thyroid function: TSH. In: Clinicalchemistry in diagnosis and treatment, 4th ed. Chicago: Year BookMedical Publishers, Inc., 1984:176-92.4. Alexander WD, Kerr DJ. First-line test of thyroid function[Letter]. Lancet 1984;ii:647.5. Allen KR, WatsonD. Thyrotropinasthe initial screening test forthyroid disease [Letter]. Clin Chem 1984;30:502-3.6. Seth J, Kellett HA, Caidwell C, et al. A sensitive iminunoradio-metric assay for serum thyroid stimulating hormone: a replacementfor the thyrotropin releasing hormone test? Br Med J1984;289:1334-6.7. John R, JonesK. An automatedimmunoradiometricassay forhuman thyrotropin. Clin Chem 1984;30:1396-8.8. Bayer MK, Kriss JP, McDougall JR. Clinical experience withsensitive thyrotropin measurements: diagnostic and therapeuticimplications. J NuclMed 1985;26:1248-56.9. Maler JS,Manotti SE, Knee GR, Goodman DBP, Strauss JF ifi.Identification of hyperthyroid patients by means of a sensitiveassay for thyrotropin.Clin Chem 1985;31:642-4.

CLINICAL CHEMISTRY, Vol. 33, No. 8,1987 1445

10. Rodriguez-EspinosaJ, Mora-BruguesJ, Lopez-Calull C, Cortes-Rius M. La determinacion basal deTSH per analisis inniunorradio-metrico (mMA) puede hacer innecesario eluse de Ia prueba deestimulacionconTRH en el diagnostico de los estadoshipertiroideos[Letter]. Med Clin (Barcelona) 1985;85:255-6.11. WoodBG, Waller D, Hantke U. An evaluation ofsix solid-phasethyrotropin (TSH) kits. J Clin ChemClin Biochem 1985;23:461-71.12. Clark PMS, Price CP. Enzyme-amplified immunoassays: a newultrasensitive assay of thyrotropin evaluated. Clin Chem1986;32:88-92.13. Martino B, Bambini G, Bartelena L, et al. Human serumthyrotropin measurement by ultrasensitive immunoradiometricassay as a first-line test in the evaluationof thyroid function.ClinEndocrinol (Oxford) 1986;24:141-8.14. Kubasik NP, Same DO, Brodows RG, Sine HE. Assay ofthyrotropin in hyperthyroidism [Letter]. Clin Chem 1983;29:1688.15. Rodbard D. Statistical estimation of the minimal detectableconcentration (“sensitivity”) for radioligand assays. Anal Biochem1978;90:1-12.

16. Cornbleet PJ, Gochman N. Incorrect least-squares regressioncoefficients in method-comparison analysis. Clin Chem1979;25:432-8.17. SiegelS. Estadistica no parametrica aplicada a las ciencias dela conducta. Mexico City:Editorial Trillas, S. A., 1982:143-55.18. Durham AP. The upper limit ofnormal for thyrotropin is 3 or 4milli-int. unitsfL. Clin Chem 1985;31:296-8.19. BassetF,Eastman CJ, Ma G,Maberly GF, Smith HC. Diagnos-tic value of thyrotropin concentrations in serum as measured by a

CLIN. CHEM. 33/8, 1445-1447 (1987)

sensitive immunoradiometric assay. Clin Chem 1986;32:461-4.20. Kirkegaard C, Bregengard C. Is the TRH test necessary?[Letter]. Lancet 1985;i:222.21. Rodriguez-EspinosaJ, Ordoflez-Llanos J, Mora-Brugues J, La-pez-Calull C, Cortes-Rius M. Sensitive immunoradiometric assay(mzs) for serum thyrotropin may be insufficient to make thethyrotropin-releasing factor teat unnecessary [Letter]. Clin Chem1985;31:1081-2.22. Spencer CA, Lai-Rosenfeld AO, Guttler RB, et al. Thyrotropinsecretion in thyrotoxic and thyroxine-treated patients: assessmentby a sensitive immunoenzymometric assay. J Clin EndocrinolMetab 1986;63:349-55.

23. Evered DC, Ormston BJ, Smith PA, Hall R, Bird T. Grades ofhypothyroidism. Br Med J 1973;i:657-62.24. Bastenie PA, Bonnyns M, Vanhaelst L Grades of subclinicalhypothyroidism in asymptomatic autoimmune thyroiditis revealedby the thyrotropin-releasing hormone test. J ClinEndocrinol Metab1980;51:163-6.25. Heinen E, Herrmann J, Konigahausen T, Kruskempfer H.Secondary hypothyroidism in severe nonthyroidal illness. HormMetab Res 1981;13:284-8.26. Vierhapper H, Laggner A, Waldhausl W, Grubeck-LoebensteinB, Kleinberger G. Impaired secretion of TSH in critically illpatients with low T4 syndrome. Acts Endocrinol (Copenhagen)1982;101:542-9.27. Wehmann RE, Gregerman RI, Burns WH, Saral R, Santos OW.Suppression of thyrotropin in the low-thyroxine state of severenonthyroidalillness.N Engl J Med 1985;312:546-52.

Thyroid Function in Hospitalized Patients: Effect of Illness and Serum Albumin ConcentrationsRajesh K. Desal,1 IshwarlalJIalal,’JohnM. Altchlson,2andSeptimusM.Joubert’

We assessed the clinicalutility of measuring thyrotropin(TSH) inserum by immunoradiometry and of measuring totalthyroxin (TT4), totaltriiodothyronine (113), free thyroxin(FT4), and free triiodothyronine (FT3). We used a group of110 healthy volunteers, 45 ill hypoalbuminemic patients, and42 iii normoalbuminemicpatients.In addition,the free thyrox-in index (FTI) and 114:thyroxin-bindingglobulin(TBG) ratiowere alsocalculated.The hypoalbuminemicgroup had sig-nificantlylowerFT4, FT3, 114, 113, and FTI concentrations,butonlyFT3 and 113 were significantlylowerintheillnormo-albuminemic group as compared withcontrols.We foundsignificantcorrelationbetween FT4 and albumin(r = 0.372,P <0.001) and FT3 and albumin (r = 0.465,P <0.001).TSHconcentrationswere undetectableintwo of45 hypoalbumin-emic patients,significantlyhigherinthe rest. The T14/TBGratiowas theonlyparameterofthyroidfunctionthatremainedunchanged inthe illpatients.

Thyroid status, particularlyin sick patients, may prove tobe difficultto assess (1-8). This problem isaggravatedbythe fact that although thereisa wide spectrum ofthyroid-

S. A. Medical Research Council,Preclinical DiagnosticChemis-try Research Unit; and Departments of’ Chemical Pathology and2Surgeiy, University of Natal Medical School, P0 Box 17039,Congella 4013,R.S.A.

Received October 20,1986;acceptedApril 30, 1987.

function tests available, there is no consensus on the rela-tivevalueofthesetests(9,10). Recent developments haveallowed direct measurement of the free (unbound) fraction ofthyroid hormones in serum with relative ease (11). Howev-er, it has been suggested that the labeled analogs involvedin these methods interact with binding sites on circulatingalbumin, which could result in spuriously low values for freethyroid hormone concentrations (12). One manufacturer(Amersham) has modified the free thyroid hormone assaywith an “albumin blocker”; another (DPC) claims theiranalog has negligible albumin binding properties.

In view of the fact that the vast majority of hospitalizedpatients for whom thyroid-function tests are requested havean intercurrent illness and, in our hospital, many also havehypoalbuminemia, we decidedtoassess the relative meritsof the assays for free and total thyroid hormone in illpatients with and without hypoalbuminemia.

Methods and Patients

The followingwere measuredby radioimmunoassay: freethyroxin3 (FF4) and free triiodothyronine (FF3) (“Amerlex-M,” Amersham, U.K.; and “Coat-a-Count,”DPC, LosAnge-les, CA); total thyroxin (TF4), total triiodothyronine (‘FF3),

Nonstandard abbreviations: FF4 (‘FF4), free (and total) thyrox-in; FF3 (‘FF3), free (and total) triiodothyronine; TBG, thyroxin-binding globulin; TSH, thyrotropin (thyroid-stimulating hormone);DPC, Diagnostic Products Corp.; FF1, free thyroxin index.