European Biological Variation Study (EuBIVAS): within- and between-subject biological variation estimates for serum thyroid biomarkers based on weekly samplings from 91 healthy participants

Introduction

Thyroid pathologies represent the most common endocrine disorders after diabetes. Laboratory tests, together with imaging methods, are fundamental for the diagnosis and management of these disorders in clinical practice [1]. Thyroid stimulating hormone (TSH) determination is recommended as a first screening test for the presence of overt and subclinical thyroid disorders, followed by free thyroxine (FT4), and free triiodothyronine (FT3) in patients who are suspected to have hyperthyroidism, as a second step [1], [2]. Human calcitonin (CT) is mainly used as a marker for the presence and monitoring of medullary thyroid carcinoma (MTC), while thyroglobulin (TG) is mostly of value in the management of differentiated thyroid carcinoma (DTC) patients after treatment [1], [2].

Availability of reliable biological variation (BV) data [3] for these central thyroid markers is important to define analytical performance specifications (APS) [4], [5], and necessary to establish the reference change value (RCV), that may be used as a tool for monitoring patients in assessing what changes between two measurements can be explained by biological and analytical variation [6]. Global within-subject (CV _I ) and between-subject (CV _G ) BV estimates are available in the European Federation of Clinical Chemistry and Laboratory Medicine (EFLM) BV Database [7]. To deliver these estimates by meta-analysis, BV studies are appraised using the Biological Variation Data Critical Appraisal Checklist (BIVAC), which is based on 14 Quality Items (QIs), each of which is associated with a score A, B, C, and/or D [8]. BIVAC-compliant studies that satisfy the meta-analysis inclusion criteria (healthy adults, biweekly to monthly sampling, more than two samples per participant) are used to deliver the global CV _I and CV _G estimates reported in the database. For some thyroid biomarkers several studies of varying quality have been published, however, they are small-scale studies unable to assess e.g., differences between genders, and for CT, no BV data have been published previously.

The EFLM Working Group on BV (BV-WG) designed the European Biological Variation Study (EuBIVAS), which is a BIVAC compliant BV study that satisfies all the preanalytical requirements of the critical appraisal checklist for BV studies [9], [10]. The EuBIVAS is a highly powered, large-scale study, involving 91 healthy subjects from five European countries [11], [12]. The aim of the present study was to deliver updated BV data, together with the derived APS and RCVs for TSH, FT4, FT3, TG, and CT.

Materials and methods

Results

Detailed information on participants’ characteristics has been published [18]. Briefly, 20 subjects reported to be smokers, the physical activity and the alcohol intake was low-to-moderate, and the use of medications was limited. The median body mass index (BMI), median age, number, and gender of the included subjects are summarized in Supplemental Table 2. For all the five measurands, two women <50 years (one from Norway, one from Spain) and two women >50 years (both from Turkey) were eliminated from the study because of their TSH results being above the upper reference interval for our assay (>4.8 mIU/L), and they were later diagnosed with a thyroid disease, in line with these subjects having subclinical thyroid disease at the time of sampling for the EuBIVAS.

A woman from The Netherland with a mean TSH value of 5.0 mIU/L throughout the 10 EuBIVAS collections, was not excluded because retesting showed that she presently has not signs of thyroid diseases, with normal thyroid hormones, negative anti-Thyroid peroxidase antibody and anti-TSH receptor.

No differences were found between mean values and BV estimates in the two female subgroups (females below and above 50 years), therefore only results from the overall female group are reported.

In the case of CT, 22 subjects (two men and 20 women) were not included in the data analysis because their results were below the limit of quantitation (LOQ), while for TG, one man was excluded for the same reason. Considering the data for the whole study population after exclusion of the subjects described above, 2.7% of TSH results, 1.9% of FT3, 0.7% of FT4, 7.3% of TG, and 1.8% of CT were excluded to achieve CV _I and CV _A homogeneity. In addition, for TG three subjects were identified as outliers (two men and one woman) by the Dixon-q test, and thus not included in the CV _G estimate. The information about the detection and exclusion of outliers are summarized in the Supplemental Table 3. For all the measurands, the mean concentration of all samples did not change during the 10 weeks of the study (data not shown). Based on the Shapiro–Wilk test, the ln-transformed data for the whole population and for male and female subgroups were normally distributed. The results for each individual for TSH, FT4, FT3, CT, and TG, ordered by sex/age, are shown in Figures 1–5, respectively. Significant differences in concentrations were found among subjects from different countries for TSH (Turkey and the Netherland vs. the other countries) and TG (Turkey vs. the other countries) (Supplemental Figures 1 and 2).

Figure 1:

TSH according to age/sex.

Median values (horizontal bars) and range (minimum-maximum) of thyroid stimulating hormone (TSH) for each individual ordered by sex and age after outliers’ exclusion. Median age and range were 35 (22–59) years and 39 (21–69) years for males and females, respectively. The 95% CI of the mean, the 5th and the 95th percentiles are indicated by continuous lines for both men and women.

Figure 2:

FT4 according to age/sex.

Median values (horizontal bars) and range (minimum-maximum) of free thyroxine (FT4) for each individual ordered by sex and age after outliers’ exclusion. Median age and range were 35 (22–59) years and 39 (21–69) years for males and females, respectively. The 95% CI of the mean, the 5th and the 95th percentiles are indicated by continuous lines for both men and women.

Figure 3:

FT3 according to age/sex.

Median values (horizontal bars) and range (minimum-maximum) of free triiodothyronine (FT3) for each individual ordered by sex and age after outliers’ exclusion. Median age and range were 35 (22–59) years and 39 (21–69) years for males and females, respectively. The 95% CI of the mean, the 5th and the 95th percentiles are indicated by continuous lines for both men and women.

Figure 4:

TG according to age/sex.

Median values (horizontal bars) and range (minimum-maximum) of thyroglobulin (TG) for each individual ordered by sex and age after outliers’ exclusion. Median age and range were 35 (22–59) years and 39 (21–69) years for males and females, respectively. The 95% CI of the mean, the 5th and the 95th percentiles are indicated by continuous lines for both men and women.

Figure 5:

CT according to age/sex.

Median values (horizontal bars) and range (minimum-maximum) of calcitonin (CT) for each individual ordered by sex and age after outliers’ exclusion. Median age and range were 35 (22–59) years and 39 (21–69) years for males and females, respectively. The 95% CI of the mean, the 5th and the 95th percentiles are indicated by continuous lines for both men and women.

The BV estimates of the whole study population and for the male and female subgroups are reported in Table 1. In Table 2, CV_APS and B _APS, RCVs, and NHSPs are presented for each measurand.

The mean serum concentration for TSH was significantly higher in women than in men; while FT3, FT4, and CT serum concentrations were significantly higher in men compared to women (see Table 1). Thus, for these four measurands the lowest CV _G estimate was applied in the APS calculation, while for TG the overall CV _G estimate was used (as highlighted in Table 1). Results for CV _I estimates were similar between men and women, thus the overall CV _I estimates were used for RCV and NHSP calculation (Table 2).

Discussion

The current clinical guidelines for thyroid diseases diagnosis and management [19], [20], [21] outline serum TSH as the most sensitive marker of overt or subclinical thyroid diseases. With abnormal TSH results, FT4 should be examined, especially for hypothyroidism, and/or FT3, in the case of hyperthyroidism and thyrotoxicosis. TSH concentrations are modified even by minimal changes in the thyroid hormones (THs) levels and altered TSH concentrations occur during hyperthyroidism or hypothyroidism onset even before free THs changes [2]. However, serum TSH levels may be influenced by numerous factors, such as ethnicity, gender, age [1], [2], [22] and pregnancy; for the latter specific reference interval related to age of gestation are required [23], [24]. Diurnal and circannual variations in TSH levels have also been reported [22], and smoking, obesity and use of medications or dietary supplements (as a source of excessive iodine intake for some individuals) may affect TSH levels [22], [23], [24]. Therefore, it is not surprising that TSH mean concentrations showed significant differences between EuBIVAS participants from the different countries (see Supplementary Figure 1). We have previously reported lifestyle differences between the country-defined EuBIVAS study cohorts [18]. This includes a high number of smokers and the highest mean BMI in the Turkish cohort [18], which may explain the higher TSH results that were found in this group. Thus, the TSH CV _G estimates and APS for bias based on the EuBIVAS data may not be applicable to all populations and should be used with caution.

The current TSH reference interval, especially the upper limit, is still under debate [22], [25], [26], due to the possible high prevalence of subclinical thyroid diseases that makes it difficult to identify truly healthy subjects. In our study, we excluded four subjects who had a TSH levels above our laboratory’s upper reference limit (>4.8 mIU/L) and who later were diagnosed with thyroid disease, in line with subclinical disease at the time of sampling. 22 subjects had results between 2.5 and 4.8 mIU/L, but all other thyroid biomarkers were normal, and these subjects were not excluded. The lack of harmonization between different analytical assays for TSH, FT4, and FT3 also influences the reference intervals in use in laboratories and limits the clinical applicability of the above cited current guidelines for thyroid diseases diagnosis and management. In this context, the possibility to obtain the personal reference interval, as recently published by Coskun et al. [27], might be a step forward in advancing personalized medicine. The International Federation of Clinical Chemistry and Laboratory Medicine (IFCC) Working Group for Standardization of Thyroid Function Tests is presently working on the standardization and harmonization of TSH, FT3, and FT4 measurements [28], [29].

In the EuBIVAS population, significant differences in mean concentrations of TSH, FT4, and FT3 between men and women were observed (see Table 1), as has previously been described only for TSH [30]. The EuBIVAS-based APS for imprecision (CV_APS) (Table 2) are lower than those as of now reported in the EFLM BV database (10.6, 3.9, and 3.0% for TSH, FT4, and FT3. respectively) [7]. EuBIVAS has delivered high-quality data on BV, using a large cohort of samples collected under optimal conditions, fully BIVAC compliant [8]. As a consequence, the global estimates reported in the EFLM DB [7], will be inevitably influenced by the new meta-analysis calculation after the inclusion of the EuBIVAS based estimates. It should be recalled that the APSs are used for both internal quality control and for external quality assurance (EQA) [4]. Thus, if EQA providers recalculate their APSs based on new and tighter data, it will be more difficult for participating laboratories to satisfy the new quality specification. Generally, EQA providers should calculate their APSs in the same way and use the EFLM BV database as their source for BV data [31], [32]. The CV _A estimates for imprecision obtained in our study are clearly lower than the CV_APS. This indicates that the analytical system applied in our study likely is within these requirements. However, considering the EuBIVAS strict pre-analytical protocol and that the CV _A estimates are based on duplicate analysis of all study samples, a routine setting with long-term analysis may give rise to CV _A estimates exceeding the EuBIVAS APS.

Until now 10, seven, and five papers have been published and included in the EFLM database for TSH, FT4, and FT3, respectively [7]. In addition, another study with BV estimates for TSH, FT4, FT3, and TG has recently been published by Mairesse et al. [33]. The EuBIVAS provides lower BV estimates than those previously published, which delivers not only stricter APS, but also smaller RCV, which would identify smaller changes as being within the expected variation caused by analytical and biological variation. If using the TSH, FT4, and FT3 CV _A and CV _I estimates derived from the EuBIVAS, the RCVs (95% probability for a significant unidirectional change) were 50.7, 12.6, and 13.1%, respectively (see Table 2). In clinical practice, this means that for a subject with 2.5 mIU/L of TSH, an increase to 3.8 mIU/L in consecutive TSH result could be explained simply by biological and analytical variations. This can also explain a subject with 16.7 pmol/L an increase to 18.8 pmol/L, and with 5.4 pmol/L an increase to 6.1 pmol/L, for FT4 and FT3, respectively, in an adult patient. However, it is important to bear in mind that results within the RCV does not rule out that such a change is of clinical importance. It is important to take into account that the EuBIVAS RCVs were obtained using the specific EuBIVAS CV _A based on analysis of duplicate samples, and, for this reason, cannot be considered as universal values thus underlining that each laboratory has to calculate its own RCVs using relevant CV _A estimates.

Another information for the management of thyroid diseases is the calculation of the number of samples required to obtain reliable estimates of the TSH, FT4, and FT3 homeostatic set points. Using the EuBIVAS BV estimates for TSH, 13 samples are required to obtain an approximation of 10%, and six for a 15% approximation, in line with the higher variability observed for TSH, which underlines the challenges in interpreting TSH results. The limit of 10 and 15%, arbitrarily chosen, hint clinician that, considering the cost benefit, depending on the patient needs might decide the number of requested tests. For FT4 and FT3, on the other hand, only one sample is sufficient for the 15% approximation (see Table 2). Robust BV data for TSH may play an important role to improve the TSH measurement specificity and reliability for screening patients with subclinical/overt thyroid diseases, for monitoring patients under therapy and, in some cases, for the identification of patients with an increased risk of bone fractures, thyroid cancer, and cardiovascular diseases. Reliable BV estimates can also be a fundamental tool for clinicians in the correct identification of subjects who require substitutive or suppressive TSH therapy, thus underlining that an adequate training about the usefulness of BV data in the clinical decisions is of importance for endocrinologists, and for giving recommendations on how many repeated samples are required be taken to obtain the patient’s set point.

TG is an important DTC tumor marker that is used to evaluate the effectiveness of treatment and for monitoring patients with recurrent disease after partial or total thyroid ablation. Moreover, increased serum TG concentrations can also be found in subjects with other thyroid diseases, such as thyroiditis, increased thyroid activity, goiter, or Graves’ disease [4], [34]. The EuBIVAS CV _I estimate for TG (all subjects: 10.3% (95% CI; 9.8–10.9)) is lower than the estimates recently reported by Mairesse et al. (15.4%, 95% CI; 13.3–18.3) [33] and by Feldt-Rasmussen, even if for the latter, 12.8%, is not significantly different based on evaluation of the 95% CI; 10.0–17.1 [35]. EuBIVAS, being highly powered and fully BIVAC-compliant, delivers lower RCVs and stricter APS than those based on non–BIVAC-compliant studies with clinical implications both for monitoring patients and ensuring that analytical methods are fit for purpose. However, it must be kept in mind that a high number of results had to be excluded to obtain homogeneity of the data. The EuBIVAS based RCV is 29.0% (95% probability for significant unidirectional change) (see Table 2), this means that for a subject with a TG baseline value of 35.2 μg/L, an increase to 45.4 μg/L could be explained by biological and analytical variations alone. TG measurement is a useful marker for patient monitoring after DTC tumor removal and treatment [34], and the increase of TG in serial consecutive measurements during follow up seems to be more informative than only one TG evaluation after treatment [36], [37]. The NHSP calculated for TG based on the EuBIVAS data is five samples with a 10% approximation and three with 15% approximation. Thus, BV estimates can be influential to improve the TG measurement specificity for monitoring patients with thyroid diseases, potentially also for DTC patients during the follow up. However, it is important to keep in mind that our data have been derived from a healthy population and are based on weekly samplings, whereas for monitoring of thyroid diseases, less frequent samplings intervals are indicated. Moreover, TG measurements post thyroidectomy, may not have a “homeostatic set point” due to expected growth of tumor cells and may have different physiology and so caution should be used in these RCVs.

Even though there is no consensus about the use of CT as a screening method for patients with thyroid nodules for MTC diagnosis or with hereditary MTC forms (about 25% of MTC), some data indicates its potential as a sensitive prognostic and diagnostic marker in both pre- and post-operative phases of thyroid C cells diseases, especially for MTC [36]. The lack of agreement is especially due to the fact that, in case of diagnostic uncertainty, the CT screening results requires the association with CT stimulation assay with pentagastrin or calcium gluconate as confirmatory tests to increase the CT specificity. For adults, values >1,000 ng/L of CT revealed a positive predictive value of 100% for MTC and are thus highly suspicious for MTC presence, while CT values <10 ng/L indicate no C cell diseases, and values >20 ng/L for basal CT are positive predictive values in the diagnosis of MCT [38]. In the EuBIVAS population, circulating levels of CT were significantly lower in women than in men and about 24% of the subjects had CT values lower than the LOQ of our assay, which are both characteristics already described in literature [39]. To the best of our knowledge, no studies have reported BV estimates for CT. The CT RCV value, calculated based on EuBIVAS CV _A and CV _I data, was 36.4% (see Table 2). This indicates that an increase in CT results from 15.0 to 20.5 ng/L could be explained simply by biological and analytical variations, considering an RCV with 95% probability for significant unidirectional change. The CT NHSPs were eight and four with approximation of 10 and 15% respectively, indicating that for CT more than one measurement could be needed for clinical decision if the concentration is near the action limit. Due to the lack of BV studies for CT, the EuBIVAS BV estimates together with RCVs and NHSPs are of fundamental importance for correct interpretation of CT results, and useful to create more specific range of basal CT concentrations for the healthy population. In addition, these results will be useful to create cut-offs when monitoring MTC patients after surgery or in response to therapy.

Conclusions

Our study delivers updated, and for CT not previously available, BV based estimates for thyroid markers based on the large scale EuBIVAS study. No differences in CV _I estimates between men and women or between countries were observed, but differences in mean TSH concentrations between different country-specific subgroups. The EuBIVAS based RCVs, NHSP, and APS may be instrumental in assuring quality of the laboratory analysis of thyroid markers and in the diagnosis and monitoring of overt or subclinical thyroid pathologies, follow-up of thyroid tumor patients.