Abstract
Background:
In light of several recent recommendations to use total thyroxine (T4) measurements in the diagnosis of thyroid function in pregnancy (in particular, “Clinical Practice Guidelines for Hypothyroidism in Adults,” cosponsored by the American Thyroid Association and the American Association of Clinical Endocrinologists, which promote the use of T4 over free T4 [FT4 ]), we have examined the implications of employing T4 for diagnostic discrimination in both pregnant and nonpregnant patient panels. Use of T4 assays has significant drawbacks in this regard, and we believe that the suggestion is a retrograde step in thyroid function testing.
Summary and Conclusions:
Analysis of the interplay between the concentrations of T4 and thyroxine-binding globulin (TBG), typifying their respective reference ranges in either the nonpregnant or pregnant euthyroid state, shows that the effective T4 range is widened significantly by the accompanying hidden variation in TBG levels. Accordingly FT4 assays that fully compensate for serum T4-binding protein concentrations should discriminate dysfunctionality from normality more efficiently than total hormone measurements, whether in pregnant or nonpregnant states. The euthyroid FT4 reference ranges typical of late pregnancy should also be more compact than those for the total hormone, because of the increased dominance of higher, though equivalently variable TBG concentrations on T4 levels. Parallel effects on T4 from similarly variable, though lower concentrations of TBG are indicated in the nonpregnant group. While acknowledging the difficulties in FT4 measurement arising from inconsistent calibration of present-day commercial assays, this finding questions the recommendation that total hormone assays should supersede the former in pregnancy.
Introduction
Commentary
From the earliest studies using analog or two-step assays (6,8), FT4 has consistently been shown to give significantly better diagnostic discrimination in routine panels of nonpregnant subjects than either T4 or FTI. The superior correction for variation in TBG concentrations by the former should be even more evident in pregnancy, given the much stronger influence of the elevated TBG levels in this condition and the corresponding diminished influence of the minor T4 binders transthyretin and albumin. In the clinical trial of the first analog assay for FT4 (8), analysis of a panel of routine hypo-, eu-, and hyperthyroid subjects showed significantly improved diagnostic discrimination over either T4 itself or FTI, with the greatest benefit at the euthyroid–hyperthyroid reference borderline. In this study, using mainly mild to moderately hyperthyroid subjects, the improvement in discrimination at the euthyroid–hyperthyroid borderline by FT4 over T4 was approximately 25%, with an equivalent of 15% at the corresponding hypothyroid borderline (8). This implied that the independent variation in TBG and T4 levels in these patients was responsible for the greater spread of T4 values defining the reference range. We have investigated this possibility by a comparative analysis of the differing trends displayed by the parameters in the nonpregnant situation and in the three trimesters of pregnancy. To this end, the spread of the previously reported 95% reference range of each parameter (6,7) was scaled according to its coefficient of variation (CV). The illustration includes a simulation (using the R statistics program; R Foundation for Statistical Computing, Vienna, Austria) of the maximum theoretical spread of the T4/TBG ratio, assuming an uncorrelated relationship of the two parameters (Fig. 1). This implies that the variable TBG levels significantly broaden the T4 reference range, with consequences for diagnostic efficiency at the borderlines. As a corollary, similar variation in TBG levels in hypothyroid and hyperthyroid subjects will also have equivalent spreading effects at the margins into the reference range, further adversely affecting discrimination. Another comprehensive study has also found a more restricted reference range for FT4 in pregnancy compared with T4 (7). Accordingly, FT4 measurement ranks favorably on our scaled comparison (Fig. 1).
Scaled ranges of T4, TBG, T4/TBG, and FT4. The spreads shown rely on 95% reference ranges reported by Ball et al. (4)
The above theoretical analysis is necessarily a crude, imperfectly corrected representation of a valid FT4 assay in that it ignores the minor additional contributions of similarly variable transthyretin and albumin concentrations, but it implies that in any population with similar spreads of TBG concentration values equivalent overlaps in results will occur. Indeed, in pregnant women, in which TBG concentrations are almost twice the norm on average, the effect of such variations will be even greater, especially as modulating albumin concentrations are simultaneously reduced (6).
One- and two-step FT4 assays are designed to fully compensate for all aspects of T4 binding to TBG, transthyretin, and albumin (9,10). Accordingly from this argument, the spread of the T4 reference range defining euthyroid status should be wider than the corresponding range for FT4. The generally accepted 95% range for T4 is approximately 58–160 nmol/L (11). Spreads can be readily displayed as the ratio of the higher 97.5th percentile to the lower 2.5th percentile of the range. For T4 the ratio is 2.8. In comparison, the corresponding range for FT4 is 10–24 pmol/L, giving a ratio of 2.4. The smaller spread for FT4 directly supports the idea that the T4 reference range is compromised by uncorrected TBG variation.
In an early prospective longitudinal study, T4, TBG, T4/TBG ratio, and direct FT4 were measured in pregnant women in the three trimesters of pregnancy (6). The percentile ratios for T4 were 3.3 in the first, 2.4 in the second, and 2.0 in the third trimester. The corresponding ratios for FT4 were 2.14, 1.65, and 1.61. Spread ratios for TBG in all trimesters were similar (∼2.2) to the nonpregnant value, even given the observed elevation in TBG concentration. This implies equivalent effects of TBG variance on results for T4 in both nonpregnant and pregnant scenarios. Hence, as expected the narrowing of the reference range for FT4 compared with T4 was even more pronounced than in the normal nonpregnant population (5). This has significant implications for diagnostic efficiency by the total hormone in these situations, forecasting similar or even greater potential for misdiagnosis.
A criticism could be leveled that the three FT4 assays used were compromised by interference from protein binding effects by the serum T4-binding proteins. This can be discounted in several ways. First, the mean FT4 values in the second and third trimesters were almost identical at approximately 75% of the nonpregnant euthyroid mean, even though the assays used different methodologies and two of the three had no interaction between tracer and serum binding proteins (12,13). Second, this fall in FT4 values was also identical to that found with a two-step assay where by definition such effects are absent (14). Third, the measured mean FT4 values and the percentage reduction in concentration in late trimesters were closely similar to those obtained by the exhaustively investigated and validated reference method of liquid chromatography/mass spectrometry combined with equilibrium dialysis (15). The results implied a smaller spread of values by FT4 in pregnancy compared with the corresponding T4 and forecast better diagnostic definition. This was also found in an earlier study (5). This shows that the assays concerned are not subject to significant distortion from miscalibration or protein effects in these situations.
While the additional manipulation by FTI may partially improve diagnostic efficiency, it is still insufficient to compensate fully for the independent variation in binding proteins. However, there remains an interassay problem of inconsistency in calibration amongst modern commercial FT4 assays, as well as the more minor effects of differences in methodology, dilution of serum by reagents, abstraction of T4 by the antibody probe, and the relative affinities of competitors for antibody binding. Choice of assays therefore rests on whether an elevated T4 range, with a better and more controlled analytic interassay precision but a greater diagnostic imprecision, is preferable to a true lower FT4 range in pregnancy using more diagnostically accurate but, at present, less consistently calibrated methods (16). Additionally, it must always be realized that clinical interpretation does not rest on these parameters alone, but is complemented by the more sensitive thyrotropin (TSH) measurement.
Establishing accurate reference limits is a prerequisite for the successful diagnostic application of any of the methods considered, but it does not completely suffice for prognostic predictions in thyroid testing. Unlike many other laboratory parameters, the interindividual reference range of T4 or TSH is twice as wide when compared with the intra-individual spread of normal values, thereby potentially compromising the valid interpretation of the estimates in a given patient (17). The extra uncertainty engendered by substituting FT4 for T4 can exacerbate this problem further.
Footnotes
Author Disclosure Statement
The authors declare that this article is an original submission and no competing financial interests exist.