Survey
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
CLIN. CHEM. 33/8, 1391-1396 (1987) Specificity of Sensitive Assays of Thyrotropin (TSH) Used to Screen for Thyroid Disease in Hospitalized Patients Carols Spencer,AndreaElgen,Dennis Shen, Marysla Duds, StephanIe Quails, Steven Weiss, and John Nicoloft Thyrotropin (TSH) concentrations were measured in 1580 hospitalized patients and 109 normal persons. Using the mean ±3 SD limits of the log values for the controls (0.356.7 milli-int. units/L), the proportion of abnormal TSH results in the hospitalized patients was 17.2%. TSH was undetectable (<0.1 milli-int. unit/L) in 3.1% of patients, suggesting hyperthyroidism, and high (>20 milli-int. units/L) in 1.6%, suggesting hypothyroidism. On follow-up of 329 patients, 62% with abnormal TSH (<0.35 or >6.7 milli-int. units/L) and 38% with normal TSH concentrations, only 24% of those with undetectable TSH had thyroid disease: 36% of them were being treated with glucocorticoids and 40% had nonthyroidal illness (NTi). Although half the patients with TSH >20 millimt. units/L had thyroid disease, 45% of patients had high TSH values associated with NT1. TSH concentrations usually returned towards normal when patients’ therapy with glucocorticoids was discontinued or they recovered from NTI. TSH test sensitivity appeared good when the mean ±3 SD limits of the reference population were used, i.e., no cases of hyper- or hypothyroidism, as identified by free thyroxin index (FT4I), were missed. However, TSH test specificity was inferior to that of the FT4I test (90.7% vs 92.3%), although specificity could be improved to 97.0% if the wider TSH reference limits of 0.1 to 20 milli-int. units/L were used-limits considered pathological if applied to outpatients. Evidently, different reference intervals for TSH are needed for hospitalized and nonhospitalized patients. We conclude that a “sensitive TSH assay” is not a cost-effective thyroid screening test for hospitalized patients as compared with the FT4I. AdditIonal Keyphrases: thyroid status screening effect of nonthyroidal illness, glucocorticoids Recently, sensitive immunometric methods have been developed for quantifying thyrotropin (TSH), for use in routine clinical laboratories (1,2).’ These new methods offer substantially improved sensitivity over conventional clinical TSH assays (3-7), and can completely distinguish between normal TSH concentrations and the low values that typify thyrotoxicosis (8-16). TSH values in thyrotoxic patients characteristically fall below the detection limit of 0.1 milli-int. unitlL with these new methods. Consequently, the new TSH assays are being widely promoted as first-line screening tests of thyroid function on the basis of data Clinical Research Center, University of Southern 2025 Zonal Avenue, Los Angeles, CA 90033. Abstract presented at the American Federation California, for Clinical Research, Western Regional Meeting, February 5, 1987. 1 Nonstandard abbreviations: TSH, thyrotropin (thyroid-stimu- lating hormone); NT!, nonthyroidal illness(es); VI’41, free thyroxin index (total T4 x T3 uptake ratio); T4, total thyroxin; T3, total triiodothyronine; T3UR, T3 uptake ratio; FT’31,free T3 index (T3 X T3UR) Received February 9, 1987; accepted May 22, 1987. showing good sensitivity and specificity for diagnosing both hyper- and hypothyroidism (4, 6, 10, 11, 15-19). Although there have been some studies of TSH specificity in patients with nonthyroidal illnesses (NT!) (20,21), data on the value of these new sensitive tests for diagnosing thyroid disease in unselected hospitalized patient populations are few. The recent report of Dubuis and Burger (22), in which 3% of 269 hospitalized patients had undetectable TSH values that were not due to hyperthyroidism, suggests that the specificity of these sensitive TSH tests may be impaired as a result of NT! or in the presence of complicating drug therapy(s). We undertook the study reported here to further examine the specificity and clinical utility of TSH measurement for evaluating thyroid function in hospitalized patients. MaterIals and Methods Normal Subjects We obtained sera at 08:00-10:00 h from 109 fasting normal subjects with no clinical, historical, or biochemical evidence of thyroid disease. Fifty-two subjects were apparently healthy physicians and laboratory volunteers from the LAC/USC Medical Center and 57 were employees of the Nichols Institute, San Juan Capistrano, CA. The mean age was 32.9 (SD 9.1) y, the ratio of males to females was 1/1.5. Subjects with positive titers for antimicrosomal antibody or antithyroglobulin antibody, suggestive of autoinunune thyroid disease (23), were excluded. The log-transformed TSH values of the samples followed the expected gaussian distribution, having a mean of 1.56 milli-int. units/L (1 SD limits 0.96 to 2.54,2 SD limits 0.59 to 4.12,3 SD limits 0.35 to 6.72 milli-int. unitsfL). Hospitalized Patients Sera remaining after morning (06:00-11:00 h) SMAC panel tests (admission requests) for 1580 patients hospitalized at the acute medical facility of the LACIUSC Medical Center (outpatients were excluded) were selected on the basis of sufficient volume for the thyroid and TSH tests. The mean age of these patients was 43.7 (SD 17.4) y, the male/female ratio 1/0.6. TSH values were determined within 24 h and were grouped with reference to the mean, ± 1 SD, ±2 SD and ±3 SD values of the log normal distribution of the normal subjects shown below in Figure 1. Group A (n = 49) had undetectable (<0.1 milli-int. unit/L) TSH. Group B (n = 137) had detectable but subnormal (<-3 SD, 0.1 to 0.35 milli-int. unitiL) values. Group C (n = 139) had values between the -2 SD and -3 SD normal limits (0.36 to 0.59 milli-int. unitiL). Group D (n = 1061) had values within the normal reference interval (±2 SD normal limits, 0.60 to 4.1 milli-int. units/L). Group E (n = 107) had values between the +2 and +3 51) normal limits (4.2 to 6.7 milli-int. unitsfL). Group F (n = 61) had values exceeding the +3 SD normal limit but <20 milli-int. units/L (6.8 to 20.0 milli-int. units/L). Group G (n = 26) had high values suggestive of hypothyroidism (>20 milli-int. unitsIL). CLINICALCHEMISTRY, Vol. 33, No. 8, 1987 1391 We performed clinical and biochemical follow-up examination(s), measuring total thyroxin (T4), total triiodothyronine (T3), T3 uptake ratio (T3UR), free T4 index, free T3 index, and antimicrosomal antibody in a subgroup of 329 patients. Of these, 172(52%) had abnormal TSH values (i.e., outside 3 SD from the mean): 45 of the 49 from Group A, 70/137 from Group B, 35/61 from Group F, and 22/26 from Group G; and 157 (48%) were patients with normal values for TSH: 54 of 139 from Group C, 70/1061 from Group D, and 33/107 from Group E, the members in these latter groups being selected at random. A complete drug history, including exposure to iodide, was taken at each examination. On the basis of follow-up studies we classified these patients as either (a) having a thyroid problem; (b) receiving drugs known to affect TSH secretion, such as glucocorticoids (24) or dopamine (25); or (c) having solely NT! (26). The major nonthyroidal diagnoses were trauma (9.1%), renal disease (8.8%), liver disease (7.0%), pulmonary disease (5.8%), cardiac disease (5.5%), central nervous system lesions (5.5%), carcinoma (3.6%), gastrointestinal problems (3.3%), and infection (2.7%). Twenty-nine patients were classified as having thyroid disease, on the basis of both the clinical examination and an abnormal FT4I (<4.5 or >12.0) or a high FF31 (>200) value, whether at the initial evaluation or on subsequent restudy after recovery and discharge from the hospital. Patients who were receiving therapy with thyroid hormone or who had a history of thyroid surgery or radioiodide thyroid treatment were also classified as having thyroid disease, as were two patients being treated with a combination of thyroxin and glucocorticoid. Most (n = 45) of the drug-treated patients were receiving glucocorticoids; however, four patients receiving dopamine were also included in this classification. In all, 251 patients without evidence of thyroid disease and not receiving glucocorticoid or dopamine therapy were classified as having NTI. We also measured retrospectively T4 in the remaining 1251 patients not evaluated as part of the 329-patient subgroup. We then measured T3UR and calculated FF4! for the 12.4% of the patients’ sera that had abnormal T4 values (<45 or >120 j.zgfL). In 7.8% of these patients, FT4I values were also abnormal (<45 or >120). When possible, these patients were evaluated by recall, or otherwise by chart review, to exclude the presence of thyroid disease, and these data were used in calculating test specificity and sensitivity Data Reduction Student’s unpaired-t analysis of group data was made relative to the NTI patients with normal TSH (Group D). When analyte concentrations were undetectable, we used the value of the detection limit for the statistical analyses. Results Figure 1 illustrates the distribution of TSH values obtained for the normal subjects and the hospitalized patients. Clearly, the range of concentrations for the hospitalized patients was wider than that for the normal reference population, with only 67% of the patients’ values falling within the 2 SD limits of the reference values, and only 84% falling within the 3 SD limits. Some (3.9%) of the patients had borderline high TSH values (Group F) and 1.6% had clearly above-normal values suggestive of hypothyroidism (Group G). In contrast, 8.6% of the patients had subnormal but detectable TSH values (Group B), and 3.1% of patients (Group A) had undetectable values suggestive of hyperthyroidism. Figure 2 and Table 1 show details of the diagnostic and biochemical values, by group (A-G), for the subgroup of 329 patients who received clinical and biochemical follow-up evaluation(s). Surprisingly, thyroid disease did not appear to be the principal cause of the abnormal TSH values in these patients. Indeed, only 24% of Group A patients were classified as having thyroid disease; more of them were being treated with glucocorticoids (36%)-dose 732, SE 316, range 870-2500 mg-equivalents of hydrocortisone daily-or simply had NTI (40%). Note that TSH became detectable (0.39, SE 0.08 milli-int. unitfL) in 9/14 of these glucocorticoid-treated patients upon re-study after glucocorticoid therapy had been discontinued. Glucocorticoid-treated patients (dose 579, SE 176, range 80-1600 mg daily) also accounted for a significant proportion of patients with subnormal values (27% of Group B). 40 35 30 25 (27). Assays The sensitive TSH immunoenzymometric assay used (Abbott, North Chicago, IL) was as previously described (16). The interassay precision (CV) at 0.6, 1.7 and 15.3 milli-int. unitslL was 10.5%, 4.3%, and 4.3%, respectively. Samples with TSH values >60 milli-int. units/L were diluted 10-fold with the zero standard. TSH was measured, in duplicate, in batches of 30 to 40 serum specimens. Some sera with undetectable (<0.1 milli-int. unitlL) TSH values by this assay were re-assayed with a sensitive TSH immunoradiometric assay (Boots Celltech, Hanover, NJ), courtesy of Dr. E. C. Ridgway (University of Colorado, Denver, CO). Concentrations of T4 (detection limit <2.5 g/L) and T3 (detection limit 250 ng/L), the T3UR, and the antithyroglobulin antibody titer were measured in sera as previously described (16, 28). Antimicrosomal antibody titers were measured by a hemagglutination technique (Ames, Elkhart, IN). 1392 CLINICAL CHEMISTRY, Vol. 33, No. 8, 1987 NORMAL REFERENCE POPULATION % HOSPITALIZED PATIENTS 20 TOTAL 5 10 .\ N S I TSI uU/mII I II .1_.351 36-.5.6- GROUPI AIB[C 11.1-1.611.7 -2.52.64. D N 14.2_6.8I6.9_2OI EF G Fig.1. TSH frequency profile in a normal reference population (under solid cuive) and hospitalized patient population (----) TSH groups A to G were constructedwithreferenceto the normal reference population values listed in thetext 100 U HONTHTHOIO*t their initial evaluation, after recovery and discharge from the hospital. TSH values only returned to normal in the two antimicrosomal antibody-negative patients (2.1 and 2.3 milli-int. units/L) but remained high in the remaining five antimicrosomal antibody-positivepatients (9.9, 21.7, 10.1, 11.5, and 28.5 milli-int. units/L). As shown in Table 1, except for thyroid-disease patients in Groups A and G, T4 and FF4! values did not appear to be related to TSH status. In fact, in all but Group C, T4 values in the NT! and the glucocorticoid-treated patients did not differ notably from those of NT! Group D patients. Mean T3 and FF3! values were low or low normal for all seven groups and classifications, although the NT! and glucocorticoid patients in Group F had significantly lower T3 values than did the other NT! patient groups. The prevalence of low T4 in NT! (29) also did not appear to relate to the TSH concentrations; in fact, the prevalence of low T4 in NT! appeared similar for all Groups A-G of the 329-patient subgroup (17%, 14%, 8%, 9%, 13%, 16%, and 27%, respectively) as well as for the 1580 hospitalized patients as a whole, when evaluated according to TSH group (12%, 7%, 10%, 6%, 7%, 10%, and 9% for Groups A to G, respectively). We did find an association between the type of NT! and the TSH concentration. Of the 30 trauma patients 90% had low TSH values that would place them in Group A (n = 7) or B (n = 23). This did not appear to be the result of narcotic administration, because patients receiving similar narcotic treatment frequently presented with normal TSH values. The prevalence of patients with liver disease was equal in most TSH groups (7%, 6%, 3%, 11%, and 3%, in Groups A, B, C, D, and G, respectively), except for Group E, in which 23% of the patients had liver disease. Patients with renal disease also were about equally prevalent in most ILLNESS 80 % #{128}o TOTAL PER GROUP 40 20 0 GROUP n A 45 B C 70 54 0 70 E F S 33 35 22 329) Fig. 2. Percentageof patientsin each TSH group (A-G) classified as having either thyroid disease (S), being treated with glucocorticoid dopamine (s),or havinga nonthyroidalillness(D) Only two patients with normal TSH (Groups C to E) had disease. These patients were receiving thyroxinreplacement therapy. Most of the patients with normal TSH values were #{232}lassified as having NT! (89%, 96%, and 97% of thyroid those in Groups C, D, and E, respectively). Fewer patients were receiving glucocorticoid therapy in these groups (9%, 4%, and 0% respectively) as compared with the low-TSH groups (Groups A and B). NT! also appeared to be a common cause of abnormally high TSH values (71% and 48% of those in Groups F and G, respectively). There was further evidence that the high TSH abnormality in these groups was solely due to NT!: the TSH concentration significantly decreased from 32.4 (SE 3.6) to 12.3 (SE 3.7) milli-int. units/L in seven of 10 Group G patients who were re-studied about 88 (SE 34) days after Table 1. DIagnostic Classltlcatlon vs TSH GFOUPb A (<0.1) n Male/female Age,y T4, .rg/dL FT4I T3, ng/dL FT3I T3UR TSH, milh-int. units/L Antimicrosornal >1:400, % 11 417 49.2 ±5.3 7/9 52.8t3.8 11.1 12.7 160 180 ±1.O#{149} 6.7 ±0.5 ±1.7k 7.2±0.5 73±11 ±24 ±33 76±11 1.11 ±0.04 1.12±07 <0.1#{149}#{149} <0.1” 57*. 6 NIl 18 9/9 NT) T CC Nil CC 19 51 1 0/1 5 48 31/17 3 3/0 14/5 33/18 43.6±3.8 7.8 ±0.7 8.1 ±0.8 6.4 ±0.5 6.8±0.5 94±10 97±10 1.04 ±0.4 <0.1” 17 Age, y T4, rg/dL 50 9.4 8.8 F141 13, ngldL <25 <25 FT3I T3UR TSH, rn,lli-int.units/L Antlmicrosomal >1:400, .94 6.31 % 100 59 7.5 6.6 68±9 69±9 1.07 ±04 0.21 ±.01** 11 254 224 0.88 88±5 1.05 ±02 0.22 ±.01** 0 0.50 0 4/1 54.7±15.5 43.3±2.0 4.5 ±0.4” 4.7±0.6 54±19 55±19 1.03 ±05 0.55 ±08 0 6.7 ±0.3 6.7±0.3 96±8 93±7 1.01 ±02 0.60 ±01” 4 4.8 ±1.6 4.9 ±1.4 7.0 ±0.3 7.0 ±0.3 NT) 32 5 5 25 17/15 1/4 4/1 11/14 ±3.7 49.2 ±10.3 6.0 ±1.1 5.9 ±1.1 94 ±9 89±8 0.99 ±03 5.2 ±0.2** 19 62 ±27 55±20 .98 ±06 12.8 ±0.9’ 80** 48.8 ±8.5 5.4 ±1.5 44.8 ±3.5 6.0 ±0.5 6.3 ±1.8 35 ±10” 6.3 ±0.4 74 ±8’ 78±8 1.10 ±.03’ 9.6 ±0.6” 12 36±11” 1.17 ±04’ 11.9 ±2.1’ 20 88±36 96±6 84±33 1.10 ±0.14 1.26 ±.05** 0 94±5 1.01 ±01 1.9 ±0.1 7 G(>20) CC 6.8 ±0.4 6.5 ±0.3 67 39/28 43.7±2.6 F (6.8-20) 44.5 Nil 33.6±3.7 T 1 0/1 42.5±2.4 6.6 ±0.3 6.8±0.2 89±6 E 14.2-6.7) Male/female 0 (0.6-4.1) OC 44.6±4.2 I n C (0.36-0.59) B (0.1-0.35) OC 16 I T CC 11 5/6 53.3 ±4.4 1 0/1 64 NTI 10 5/5 43.2 ±5.7 1.8 ±0.4” 5.2 6.6 ±0.6 1.8 ±0.5** 4.6 6.3 ±0.6 39±9” 37±8” .97 .08 65.6 ±21.2” 64’ 65 57 0.88 39.4 0 75 ±14 68±11’ 0.96 ±04 35.8 ±3.3” 40” Results are given as mean ± SE where applicable. ‘I, thyroid disease GC,glucocoilicoid treated NTI, nonthyroldal illness. GrOUpS A-G (see text); values In parenthesesare the range of TSH concentrations (milli-int. unlts/L) in each group. * P <0.001 vs NTI patients of “P<005 JGroupD CLINICAL CHEMISTRY, Vol. 33, No. 8, 1987 1393 Groups (7%, 4%, 11%, 4%, 8%, and 6% in Groups A, B, C, D, E, and G, respectively), except for Group F, in which 28% of the patients had renal disease. The prevalence of antimicrosomal antibody positivity (titer >1:400) tended to be higher in the abnormal TSH groups (20%, 23%, and 61%, for Groups A, F, and G, respectively, as compared with 10% for combined Groups C to E). Antimicrosomal positivity was especially prevalent in patients with thyroid disease in Groups A, F, and G (57%, 80%, and 64%, respectively). In the 1580 hospitalized patients, the prevalence of T4 and FF4! abnormality was 12.4% and 7.8%, respectively. Thirteen of these patients had high FF4! values, of whom nine (seven from Group A, two from Group D) had been studied as part of the 329-patient subgroup. None of the other four Group D patients with high FF4! had evidence of thyroid disease, so the probable cause of the high FF4! was familial dysalbuminemic hyperthyroxinemia (30) or NT! (26). Thus, the sensitivity of the TSH test was good for detecting hyperthyroidism. Apparently no hyperthyroid patients were missed when we used TSH as the initial thyroidscreening test. Sensitivity for the detection of hypothyroidism also appeared good, in that all the patients with clinical and biochemical evidence of hypothyroidism had high (>20 milli-int. units/L) TSH values. In contrast, the specificity of the TSH test was clearly compromised by associated NT!. Specificity was especially poor (90.6%) when assessed relative to the mean ±3 SD limits of the normal reference population (as compared with a specificity of 92.3% for the FF4! test). Although the specificity of this TSH test could be improved to 97.0% if the wider cutoff limits (0.1 to 20 milli-int. unitafL) were used, these limits would be considered as pathological if applied to an outpatient population. Discussion We found that both abnormally low (<0.1 milli-int. unitiL) and high (>20 milli-int unitsfL) TSH values, not due to thyroid disease, are frequently encountered in hospitalized patients. In our study, 17.2% of patients had TSH values outside the ±3 SD limits of the range in the normal reference population, and 4.7% had strikingly abnormal TSH values (<0.1 or >20 milli-int. unitsfL). These findings contrasted with the prevalence of abnormal FF4! values, which was 7.8% overall, with 0.8% of patients having high FF4! and 7.0% having low FF4! values. The specificity of the TSH test was inferior to that of FF41 (90.7% vs 92.3%) when the ± 3 SD limits of the normal reference population were used, but could be improved to 97.0% by using wider limits of <0.1 and >20 milli-int. unitsfL. However, these wider limits would be pathologicalwhen applied to outpatients, so it appears appropriate to use different TSH reference limits for hospitalized and nonhospitalized patients. The prevalence of patients with undetectable TSH owing to thyroid disease was double (0.7%) that previously reported for hyperthyroidism in our hospital (31). This was in part because we classified three thyroxin-treated patients, patients in whom TSH is commonly undetectable even with a sensitive assay (32), as having thyroid disease. An additional factor was the improved sensitivity of the assay for detecting hyperthyroidism; two patients with thyroid disease were identified on the basis of TSH being undetectable on initial evaluation when thyroid hormone indices were normal. When these patients were re-studied after recovery 1394 CLINICALCHEMISTRY, Vol. 33, No. 8, 1987 and discharge, their thyroid hormone indices had became high and clinical symptoms of hyperthyroidism had become apparent. These hyperthyroid patients would have been missed if the FF4! value had been used as the initial screening test, because the FF4! value can become inappropriately normal or low when NT! is superimposed on preexisting thyrotoxicosis (33). One striking observation was that only 14% of Group F patients had thyroid disease. All these patients appeared clinicallyeuthyroid, and all but one patient had a normal value for FF4!. Three of these patients had previously documented hypothyroidism and were receiving T4 replacement therapy, whereas the remaining two patients had histories of partial thyroidectomy. Most (7 1%) of these Group F patients had NTI as the sole cause of high TSH, as is consistent with the recovery phase of an NT! (34). One interesting observation was that 28% of these patients had renal disease, suggesting that a high TSH may result from decreased renal clearance of the hormone (35). We were surprised to find that 48% of the patients with strikingly high TSH values (Group G) had NT!. Although this finding is in harmony with another recent report (36), the influence of NT! on TSH has not generally been thought to really limit the clinical usefulness of measuring TSH in serum for diagnosing primary hypothyroidism. In fact, NH measurement currently is widely considered to be the definitive test for distinguishing primary hypothyroidism from low T4 resulting from NTI (26). All the Group G NT! patients’ values for FF4! were within the normal reference interval, suggesting that there was no deficiency of circulating thyroxin. Furthermore, when we re-studied 10 of these patients about three months later, we found that TSH had significantly decreased in five and had returned to normal in the two antimicrosomal antibody-negative patients. The 40% prevalence of antimicrosomal antibody positivity in the Group G NT! patients was higher than that observed for Group D NT! patients (7.0%), who displayed an antimicrosomal antibody prevalence comparable with that in the general population [6.7% (37)]. This suggested that subclinical hypothyroidism (38), possibly owing to chronic thyroiditis, may have been a contributing factor in producing the strikingly high TSH concentration in some patients, although longer follow-up would be needed to definitively resolve this issue. Glucocorticoid therapy was frequently associated with low TSH values, but glucocorticoid-treated patients also appeared in Groups F and G (14% and 4%, respectively). There was no clear relationship between TSH and the glucocorticoid dose or duration of therapy. This observation, together with previous data (39) showing that normal TSH concentrations were usually seen in asthma patients receiving equivalent glucocorticoid doses, suggests that other factors besides glucocorticoid dose may be involved in the genesis of the low TSH in these glucocorticoid-treated patients (40). Trauma appeared to be the major NT! diagnosis associated with low TSH concentrations. Although exogenous narcotic administration did not appear to be the cause of this finding, it is still possible that an increased release of endogenous endorphins may have played some role in this association. We were concerned that the high prevalence (3.1%) of undetectable TSH observed in this study may have reflected method-sensitivity limitations. In an attempt to resolve this issue we remeasured TSH in seven glucocorticoid-treated patients’ sera by a different sensitive TSH technique (Boots Ceiltech) and found that five of the seven sera had comparably low (<0.08 milli-int. unitfL) values for TSH. Thus, this finding, together with that of Dubuis and Burger (22) showing that 3.0% of hospitalized patients had undetectable TSH by the sensitive Serono TSH method, was evidence that our high prevalence of undetectable TSH in hospitalized patients was not method related. Previous reports suggesting that central nervous system impairment of TSH secretion is associated with the low T4 NT! state (41-43) are not supported by our data. There was no relationship between the TSH and the prevalence of low T4 in NT!, in agreement with previous studies in which both the current TSH technique (39) was used as well as studies in which the Boots Celltech method was used (3). In the present study, central nervous system impairment (pituitary or hypothalamic disease, head trauma, stroke, seizure, or brain metastases) was not found to be strongly associated with low TSH (9%, 10%, 15%, 12%, 16%, 0%, and 0% in Groups A to G, respectively, of the 329-patient subgroup). Sensitive TSH measurement had good sensitivity but poor specificity for screening or evaluating hospitalized patients for thyroid disease when the TSH reference limits were based on the ±3 SD limits of a normal population. A high prevalence (3.1%) of undetectable values suggestive of hyperthyroidism was found both in this and in another study in which a differentsensitive TSH assay was used (22). Patients receiving glucocorticoid treatment or having trauma were frequently found to have low or undetectable TSH, a finding that did not appear to be method related. Both high and low TSH values were found to be frequently associated with NT! per se. Thus, although the new sensitive TSH tests appear to offer better sensitivity and specificity for screening for thyroid disease in ambulatory patients (8-16), their cost effectiveness for use with hospitalized patients would be impaired by poor specificity, resulting in expensive and unnecessary follow-up study of patients with abnormal values due to glucocorticoid treatment or NTI. Because the specificity of the TSH test can be improved to 97.0% if wider reference limits (<0.1 and >20 milli-int. unitsfL) are used, it may be appropriate to use different reference intervals for TSH when these tests are used for assessing hospitalized vs nonhospitalized patients. However, at this time, the current specificity and reagent-cost considerations suggest that the FF4! test may be a more cost-effective thyroid screening test for evaluating hospitalized patients than is sensitive TSH. We thank Abbott Laboratories (Chicago, IL) for supplying the for this study, and the Los Angeles County Hospital Chemistry Department, especially Dr. Ed Wong, the associate director of the laboratories, and Eddie Briones and the SMAC Laboratory staff, for their help and support in this study. We also gratefully acknowledge the expertise of Ken Anderson with statistics, Denise Walters in preparing this manuscript, and Anne Santo in preparing the graphics. The computer services for this study were provided by the cuiro Project, funded by the Division of Research Resources of the NIH under Grant no. RR00043. reagents References 1. Woodhead JS, Weeks I. Circulating thyrotropin as an index of function. Ann Clin Biochem 1985;22:455-9. 2. RoesDS. New sensitive iminunoradiometric assays for thyrotropin [Review]. Ann Intern Med 1986;104:718-21. 3. Seth J, Kellett HA, Caidwell G, et al. A sensitive immunoradiometric assay for serum thyroidstimulating hormone: a replacement thyroid for the thyrotropin releasing hormone test. Br Med J 1984;289:1334-6. 4. Martino E, Bambini G, Bartalena L, et al. Human serum thyrotropin measurement by ultrasensitive immunoradiometric assay as a first line test in the evaluation of thyroid function. Clin Endocrinol 5. McBride 1986;24:141-8. JH, Thibeault measured by a sensitive 1985;31:1865-7. RV, Rodgerson immunoradiometric DO. Thyrotropin as assay. Clin Chem 6. Wiersinga WM, Endert E, Trip MD, et al. Immunometric assay of thyrotropin in plasma. Its value in predicting response to thyroliberin stimulation and assessing thyroid function in amiodarone-treated patients. Clin Chem 1986;32:433-6. 7. Pekaiy AE, Hershman JM. Serum thyrotropin as measured with a one-step monoclonal-antibody radloimmunometric assay compared with two commercial radioimmunoassay kits. Clin Chem 1986;32:1007-9. 8. Weeks I, Sturgess M, Siddle K, et al. A high sensitivity immunochemilurninometric assay for human thyrotropin. Clin Endocrinol 1984;20:489-95. 9. Cobb WE, Lamberton RP, Jackson IMD. Use of a rapid, sensitive immunoradiometric assay for thyrotropin to distinguish normal from hyperthyroid subjects. Clin Chem 1984;30:1558-60. 10. Allen KR, Watson D. Thyrotropin as the initial screening test for thyroid disease [Letter]. Clin Chem 1984;30:502-3. 11. Bayer MF, Kriss JP, McDougall IR. Clinical sensitive thyrotropin measurements: diagnostic implications. J Nucl Med 1985;26:1248-56. experience with and therapeutic 12. Roddis MJ, Burrin JM, Johannssen A, et al. Serum thyrotropin: a first-line discriminatory test of thyroid function. Lancet 1985;i:277-8. HL, Iijala K, Viikari J, et al. Determination of in serum by time-resolved fluoroimmunoassay evaluated. Clin Chem 1985;31:1706-9. 14. Bassett F, Eastman CJ, Ma G, et al. Diagnostic value of thyrotropin concentrations in serum as measured by a sensitive immunoradiometric assay. Clin Chem 1986;32:461-4. 15. Kreutzer HJH, Tertoolen JFW, Thijssen JHH, et al. Analytical evaluation of four sensitive assays of thyrotropin, including effect of variations in patient sampling. Clin Chem 1986;32:2085-90. 16. Spencer CA, Lai-Rosenfeld AO, Guttler RB, et al. Thyrotropin secretion in thyrotoxic and thyroxine-treated patients: assessment by a sensitive immunoenzymometric assay. J Clin Endocrinol Metab 1986;63:349-55. 17. Alexander WD, Kerr DJ, Ferguson MM. First-line test of thyroid function. Lancet 1984; ii:647. 18. Caldwell G, Gow SM, Sweeting VM, et al. A new strategy for thyroid function testing. Lancet 1985;i:1117-9. 19. Clark PMS, Price CP. Enzyme-amplified inununoassays: a new ultrasensitive assay of thyrotropin evaluated. Clin Chem 1986;32:88-92. 20. Malter JS, Manotti SE, Knee GR, et al. Identification of hyperthyroid patients by means of a sensitive assay for thyrotropin. Clin Chem 1985;31:633-43. 21. Semple CG, Slater SD, Reid AM, et al. A sensitive immunoradiometric assay for serum thyroid stimulating hormone. Br Med J 985;290:69-70. 22. Dubuis JM, Burger AG. Thyroid-stimulating hormone measurements by immunoradiometric assay in severely ill patients. Lancet 1986;ii:1036-7. 23. Gordin A, Lamberg BA. Natural course of symptomless autoimmune thyroiditis. Lancet 1975;ii:1234-8. 24. Re RN, Kourides IA, Ridgway EC, et al. The effect of glucocorticoid administration on human pituitary secretion of thyrotropin and prolactin. J Clin Endocrinol Metab 1976;43:338-46. 25. Kaptein EM, Kletsky OA, Spencer CA, et al. Effect of prolonged dopamine infusion on anterior pituitary function in normal males. J Clin Endocrinol Metab 1980;51:488-91. 26. Wartofsky L, Burman KD. Alterations in thyroid function in patients with systemic illness: The “euthyroid sick syndrome” [Review]. Endocrine Rev 1982;3: 164-217. 13. Kaihola thyrotropin CLINICALCHEMISTRY, Vol. 33, No. 8, 1987 1395 27. Dixon WJ, Massey FJ (eds). Statistical inference: estimation and tests. Chap 6 in: Introduction to statistical analysis. 3rd ed. New York: McGraw-Hill, 1969:75-94. 28. Spencer CA, Platler BW, Nicoloff JT. The effect of ‘I thyroglobulin tracer heterogeneity on serum Tg RIA measurement. Clin Chim Acts 1985;153:105-15. 29. Kaptein EM, Brief DA, Spencer CA, et al. Thyroxine metabolism in the low thyroxine state of critical nonthyroidal illnesses. J Clin Endocrinol Metab 1981;53:764-71. 30. Stockigt JR. Topliss DJ, Barlow JW, et al. Familial euthyroid thyroxine excess: an appropriate response to abnormal, thyroxine binding associated with albumin. J Clin Endocrinol Metab 1981;53:353-9. 31. Montoro M, Guttler RB, Spencer CA, et al. Adult thyroid screening in hospitalized patients [Abstract]. Clin Res 1981;29:39A. 32. Mardell RJ, Gamlen TR, Winton MEd. High sensitivity assay of thyroid stimulating hormone in patients receiving thyroxine for primary hypothyroidism and thyroid carcinoma. Br Med J 1985;290:355-6. 33. Lum SM, Kaptein EM, Nicoloff JT. Influence of nonthyroidal illness on serum thyroid hormone indices in hyperthyroiclism. West J Med 1983;128:670-5. 34. Hamblin PS, Dyer SA, Mohr VS, et al. Relationship between thyrotropin and thyroxine changes during recovery from severe hypothyroxinemia of critical illness. J Clin Endocrinol Metab 1986;62:717-22. 35. Constant RB, Weintraub BD. Differences in the metabolic 1396 CLINICALCHEMISTRY,Vol. 33, No. 8, 1987 clearance of pituitary and serum thyrotropin (TSH) derived from euthyroidand hypothyroid rats: effects of chemical deglycosylation of pituitary TSH. Endocrinology 1986;1 19:2720-7. 36. Brent GA, Hershman JM, Braunstein GD. Patients with severe nonthyroidal illness and serum thyrotropin concentrations in the hypothyroid range. Am J Med 1986;81:463-6. 37. Hawkins BR, Cheah PS, Dawkins RL, et al. Diagnostic significance of thyroid microsomal antibodies in randomly selected population. Lancet 1980;ii:1057-9. 38. Bastenie PA, Bonnyns M, Vanhaelst L. Grades of subclinical hypothyroidism in asymptomatic autoimmune thyroiditis revealed by the thyrotropin releasing hormone test. J Clin Endocrinol Metab 1980;51:163-6. 39. Spencer CA. Clinical evaluation of free T4 techniques. J Endocrinol Invest 1986;9:57-66. 40. Otsuki M, Dakoda M, Baba S. Influence of glucocorticoids on TRF-induced TSH response in man. J Clin Endocrinol Metab 1973;36:95-102. 41. Heinen E, Herrrnann J, Konigshausen TH, et al. Secondary hypothyroidism in severe non-thyroidal illness. Horm Metab Res 1981;13:284-8. 42. Vierhapper H, Laggner A, Waldhausl W, et al. Impaired secretion of TSH in critically ill patients with “low T4-syndrome.” Acts Endocrinol 1982;101:542-9. 43. 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;312:546-52.