Regional adjustment of thyroid hormone reference intervals

Introduction

Regarding the high prevalence of thyroid pathologies in Germany, found in 33.1% of a study population with over 92,000 participants [1], accurate diagnosis of thyroid disease constitutes a relevant public health issue. In vitro testing of thyreotropin (TSH) and the thyroid hormones free triiodthyronine (fT3) and free thyroxine (fT4) plays an important role in thyroid diagnostics.

Due to the great impact of varying regional status of iodine supplementation on basal TSH, fT3, and fT4, we aimed to establish regional reference intervals for TSH, fT3, and fT4 in the greater area of Munich, southern Germany [2, 3]. Our trial is similarly designed to the SHIP study [4] of 2005, performed in Western Pomerania, where regional reference intervals were established. The reference values, presented by Roche Diagnostics for the Elecsys immunoassay systems in the respective method sheets, are based on different studies, of which the biggest one was published in 2005 [5]. The introduction of these reference ranges in our hospital, however, did not correspond well with clinical experiences. In particular, the lower 2.5th limit of the fT3 reference range of 3.9 pmol/L was discussed controversially for a series of patients. In these subjects, the repeatedly measured low fT3 concentrations had no correlation to hypothyroidal symptoms. In all these cases, a low-T3 syndrome could be excluded. Therefore we analyzed healthy subjects in the Munich metropolitan region using a range of exclusion criteria relating to the National Academy of Clinical Biochemistry (NACB) Laboratory practice guidelines [6], based on the NHANES III survey [7], which allowed us to define a regional reference population.

Participants and methods

Results

Of 204 participants included, 93 (45.6%) were male and 111 (54.4%) female. Age range was 20–78 years; median age was 27 years (see Figure 1). Age range in the unselected study population (n=292) was 18–79 years, and median age was 33 years. Regarding the distribution of age within the population of included participants, no significant difference between males and females could be observed (p=0.56).

Figure 1

Distribution of age within the reference population.

Possible outliers for the values of TSH, fT3, and fT4 were detected using a test by Dixon [16]. In accordance with IFCC-CLSI recommendations, data were retained rather than deleted as suspect data were not known to be aberrant [9].

Reference values for TSH and fT4 were reported nonparametrically as medians and empirically determined 2.5th–97.5th percentiles. FT3, however, did adhere to a Gaussian distribution, as indicated by the histogram (see Figure 2A–C). Therefore the reference interval for this measurand was reported parametrically as arithmetic mean values and 2.5th–97.5th percentiles.

Figure 2

(A–C) Distribution of values for TSH, fT3, and fT4, respectively.

The reference intervals established in this study were 0.58–3.49 mIU/L for TSH, 11.58–20.46 pmol/L for fT4, and 3.56–5.88 pmol/L for fT3, which are summarized in Table 1. The 90% CI of the 97.5th percentile of TSH was larger than the recommended value [9], namely, more than 0.2 times the width of the reference interval. A validation check was performed by calculating the upper reference limit and its 90% CI on transformed data showing similar results with a 90% CI less than 0.2 times the width of the reference interval: 3.49 mIU/L (90% CI 3.00–4.20 mIU/L) calculated nonparametrically; 3.44 mIU/L (90% CI 3.20–3.66 mIU/L) calculated parametrically after a Box-Cox transformation [17] of the data. Therefore the 97.5th percentile obtained by the nonparametric approach was considered true.

The median TSH value in male participants (median 1.55 mIU/L) did not differ significantly (p=0.69) from that in female participants (median 1.58 mIU/L). Measured fT3 serum concentrations, however, were significantly higher (p<0.001) in the male study population (mean 5.00 pmol/L) than in the female study population (mean 4.51 pmol/L). FT4 also showed significantly higher (p<0.001) values in the male (median 16.73 pmol/L) than in the female study population (median 15.44 pmol/L).

All autoantibody determinations in the 204 study subjects were below the given serum cut-off levels: anti-TPO <34 IU/mL, anti-Tg <115 IU/mL, and anti-TSHR <1.75 IU/L.

Discussion

Our carefully established reference population of 204 adult Caucasian subjects, based on the correspondent NACB criteria [6], ensured a valid set of reference ranges for TSH, fT3, and fT4. As expected, the data of the study subjects showed a narrower range of reference values of the Elecsys assays than the suggested reference values provided by the manufacturer [19]. This shows that local reference values help to provide a basis for an accurate diagnostic biomarker for a largely local patient population. In particular, the 2.5th percentile for fT3 derived from the Gaussian approach of 3.56 pmol/L is now in better accordance with clinical findings. This value is lower than the originally given expected value of 3.8 pmol/L by the manufacturer and corresponds to the aforementioned findings of low fT3 values in patients without any hypothyroidal symptoms or low-T3 syndrome. Our finding is in contrast to that of Kratzsch et al. [5], who found the fT3 distribution to be non-Gaussian with a 2.5th percentile of 4.02 pmol/L. Interestingly, the immunoassay manufacturer, Roche Diagnostics, changed the data of the “expected values” for their thyroid hormone assay panel. Whereas the older version of the fT3 third-generation method sheet indicated a 2.5th value of 3.85 pmol/L for a study group of 5.366 subjects (study A) and a 2.5th value of 3.08 pmol/L for another smaller German trial with 870 participants (study B), the newer version from 12/2012 now gives only the 2.5th percentile of 3.08 pmol/L and refers this threshold level only to study A. The respective reference intervals for TSH and fT4, however, were not modified. We evaluated the graphical representation of the data [20] and concluded that the distribution adheres well to the normal curve. We therefore assume normality for the fT3 values (see Figure 2B). To further validate this assumption, a Box-Cox transformation to normality of the fT3 reference range data was performed and showed very similar results, which are summarized in Table 2. Transformation of data aims to produce a normal distribution. In the special case of our fT3 data – obtained in a sufficiently large number of cases – this transformation of the observed distribution to Gaussian form showed that the statistical calculations before and after the transformation produced nearly identical results. We are therefore convinced that the values of fT3 do adhere to a Gaussian distribution.

Higher fT3 and fT4 values in men compared to women were also found by Kratzsch et al. [5]. Although the differences were significant, neither Kratzsch et al. [5] nor the assay developers from Roche [19] found these results to be relevant for diagnostic use and did not establish separate reference values for males and females. A unisex set of reference values can most likely be considered accurate enough for clinical purposes because the values for males and females for both fT3 and fT4 largely overlap as they do in this study. We therefore refrain from introducing separate reference values for males and females. A possible influence of the use of contraceptives in women, however, should be taken into account [5]. This was also the conclusion of the study by Grüning et al. [21], who studied the influence of female sex hormones on thyroid parameters.

The question arises why our results differ from the results of the reference studies of Kratzsch et al. [5], which was performed in an area in central Germany in Leipzig. One reason might be the varying iodine supplementation throughout Germany. Hampel et al. [2] showed iodine excretion in urine being significantly higher in central Germany than in southern or northern regions. Also, TSH and fT3 serum concentrations show intra-individual variations depending on the time of the day [22] and on the fasting or nonfasting status of the patient [23]. Differences in blood sampling procedures may have also caused the differing data. Study participants in the Leipzig study were recruited blood donors who are generally recommended to be nonfasting, with blood sampling hours from 8 a.m. to 6 p.m. As both study participants in Munich and Leipzig can be considered to have been nonfasting, fasting status is unlikely to have altered the measured values. Blood drawing took place during the afternoon hours. This may have influenced the measured hormone levels as all three hormone analytes showed a circadian rhythm, peaking during the early morning hours between 2:00 a.m. and 5:00 a.m. [7]. Therefore it can be stated that sampling hours in the morning may have been a factor for the difference in values between the Leipzig study and our trial.

Comparison of the fT3 values of our study to the data of Völzke et al. [4] and Kratzsch et al. [5] revealed that our 2.5th percentile was the lowest (3.56 vs. 3.8 [4] vs. 4.02 [5] pmol/L). These small differences may have been due to methodological differences. Whereas in Kratzsch et al. [5] the Elecsys fT3 assay was used in 2004, Völzke et al. [4] used the LUMItest assay from Brahms (Henningsdorf, Germany). Compared to the first assay version, which was applied by Kratzsch et al. [5] in 2005, the second generation of the fT3 assay used in our trial is characterized by a new ruthenium interference suppression concept, which was introduced in 2006. This explains the marginal discrepancies of the given reference ranges.

In summary, our findings again suggest that local reference values should be established or at least verified for different regions. Even if the iodine supplementation can be considered to be sufficient for most parts of Germany nowadays [2], the presented reference interval is only valid for a southern German region and might be distinct from other areas with different extents of iodine sufficiency. Statistically remarkable is the normal Gaussian distribution of the fT3 serum concentrations, leading to a lower 2.5th percentile value of the reference interval of 3.56 pmol/L compared to the formerly recommended 3.90 pmol/L limit for the Elecsys assay format. This minute difference, however, may be crucial for the in vitro thyroid screening of apparently healthy subjects.

Conflict of interest statement

Author contributions: All the authors have accepted responsibility for the entire content of this submitted manuscript and approved submission.

Research funding: None declared.

Employment or leadership: None declared.

Honorarium: None declared.

Competing interests: The funding organization(s) played no role in the study design; in the collection, analysis, and interpretation of data; in the writing of the report; or in the decision to submit the report for publication.