Determination of reference change values for thyroid-related biomarkers: TSH, fT3, fT4, Tg, Anti-Tg, and Anti-TPO

Introduction

Clinical laboratory test results are frequently evaluated with population-based reference intervals (expected values for healthy individuals) [1]. Individuality index (II) is a parameter that gives information about the suitability of evaluating test results with the reference interval and is determined by the ratio of within-subject biological variation (CV_I) to between-subject biological variation (CV_G). Most of the tests in the laboratory have CV_I considerably lower than CV_G, and II values lower than 0.6 mean the high individuality of the test. When evaluating the serial results of these tests, it is recommended to use the reference change value (RCV) instead of the reference interval, that is, to interpret it according to the previous result [2]. The usefulness of evaluating test results with reference intervals remains limited due to this high individuality [3]. Even if the serial measurement results of many individuals are within the reference interval, significant changes may be observed in the individual, or an abnormal measurement result may not indicate pathology. It is especially valid when the result is close to the critical threshold. Thus, interpreting the results with these traditional intervals can sometimes lead to the wrong assessment. Variations in some pre-analytical, analytical, and post-analytical processes, and most importantly, biological variation (BV) cause this misinterpretation [4].

The RCV, which covers the main natural sources of variation – CV_I and analytical imprecision (CV_A) – is an objective tool that shows whether the change between consecutive measurements is clinically significant. A clinically significant change has occurred if the difference between the individual’s serial results is greater than the RCV [5]. The RCV was formulated by Fraser as follows: 2^1/2×Z×(CV_A ²+CV_I ²)^1/2 [3]. Z–score shows the number of the standard deviation appropriate to the desired confidence interval. When evaluating the test results, an increase or a decrease occurs compared to the previous result. Therefore, one-tailed (unidirectional) Z-scores should be used as there will be no concomitant increase and decrease in clinical decision management (e.g., 1.65 for p<0.05, 2.33 for p<0.01) [6], 7].

In the study of Fokkema et al., it was found that serial results in an individual were not normally distributed. Therefore, different from the traditional Fraser method, an approach involving logarithmic transformation was developed for RCV calculation [8]. Lund et al., in their study with a computer simulation model, made this logarithmic approach more practical for tests where the total variation (CV_T) (CV_T=(CV_A ²+CV_I ²)^1/2) is below 30 % [9]. The CV_I is a random fluctuation that occurs in the individual’s homeostasis process, normally (Gaussian) distributed. However, disease, abnormal metabolic cases, and medication cause abnormal fluctuations of biological variation. In such cases, the relevant analytes should not be expected to show a normal distribution [9], and the RCVs should be determined using the logarithmic transformation method [10], 11].

Thyroid dysfunction is one of the most common clinical presentations in the community and may cause many health problems, including mood changes, hair loss, fatigue, and obesity [12]. To make a clinical decision about a patient’s thyroid functions, it is necessary to evaluate the complaints, physical examination findings, and thyroid function tests together. Thyroid-stimulating hormone (TSH), free triiodothyronine (fT3), and free tetraiodothyronine/thyroxine (fT4) levels are most commonly used in the diagnosis of diseases that cause thyroid dysfunctions [13], 14]. In addition to these thyroid function tests, thyroperoxidase antibodies (Anti-TPO) and thyroglobulin antibodies (Anti-Tg) are helpful in the investigation of autoimmune thyroid diseases. Besides, thyroglobulin (Tg), a tumor marker, is widely used with the anti-Tg test in the follow-up of patients with differentiated thyroid carcinoma [14], [15], [16], [17].

Evaluating the results of thyroid-related biomarkers is essential since thyroid dysfunction and the thyroid diseases that cause it are common [12], [13], [14]. Our hospital is a tertiary care center and serves a large population in the Eastern Black Sea Region, with its location as a regional hospital. Most of those who apply to our hospital are individuals with the disease, medication, or abnormal metabolic conditions. Therefore, their thyroid biomarkers may not show a normal distribution [9]. Our clinical laboratory has three different systems for analyzing these tests. Since analytical imprecision can vary, it may be essential to determine RCVs on a device-specific basis. Consequently, this study aimed to determine instrument-specific RCVs of TSH, fT3, fT4, Tg, Anti-Tg, and Anti-TPO tests, which are routinely examined in our laboratory for the diagnosis, treatment, and follow-up of thyroid diseases, using classical and logarithmic methods.

Materials and methods

Results

The average CV_A values ​​of each test were determined on an instrument-specific basis because the analytical performances of the analyzers differed. While the analytical imprecision of the tests was within the limits recommended by the manufacturers, the CV_A values ​​of fT3, fT4 (for R1 and R2), Anti-Tg, and Anti-TPO tests were higher than the desirable CV. The number of quality-control measurements run for all three instruments, outliers (number of samples rejected according to Westgard rules [18]), the average CV of the two levels, the twelve-month average CV_A values, ​​and the desirable CV data are presented in Table 1. All the II values calculated using BV data of TSH, fT3, fT4, Tg, Anti-Tg, and Anti-TPO parameters were 0.6 and below. The CV_T values were found in the range of 6.9–16.8 %. Lund et al.’s approach was considered appropriate for the logarithmic RCV calculation as all CV_T values were below 30 %. The biological variation (CV_I and CV_G), II, CV_A, and CV_T data of thyroid parameters are presented in Table 2. The instrument-specific RCVs calculated with the conventional and logarithmic approaches using these data are given in Table 3. Since Tg, Anti-Tg, and Anti-TPO parameters were not analyzed on the Roche Cobas e411 instrument, their RCVs were not presented. The upward logarithmic RCVs were higher than the conventional approach, while logarithmic RCVs in the downward change were lower than the conventional RCVs.

Discussion

In this study, the RCVs of thyroid-related biomarkers (TSH, fT3, fT4, Tg, Anti-Tg, and Anti-TPO) were determined by both traditional and logarithmic approaches based on the analytical performance of three different immunochemistry auto-analyzers in our laboratory. Instrument-specific RCVs were calculated since the analytical performances of the instruments differ (Table 1, 3). RCVs were determined for all three instruments’ TSH, fT3, and fT4 tests. However, since Tg, Anti-Tg, and Anti-TPO biomarkers were studied in the routine unit, RCVs were determined on R1 and R2 instruments. It was observed that the II of TSH and fT4, of which RCV was calculated, was 0.6, and the other tests were <0.6 (Table 2). Therefore, since the individuality of the tests was quite high, it was found that it would be more appropriate to evaluate the results of thyroid-related biomarkers with RCV instead of the population-based reference interval [2]. Eventually, if the change between consecutive results for thyroid-related biomarkers is greater than the RCV, it may be a clinically significant change [5]. An illustrative example: in a routine outpatient clinic control of an adult stable patient (Hashimoto thyroiditis) on levothyroxine treatment, serum TSH is measured as 1.6 mIU/L (the previous visit was 1.0 mIU/L). The increase is 0.6 mIU/L (60 %), and this result, when evaluated with RCV, shows a significant change even though the result is within the reference interval (classical and logarithmic RCVs of TSH: 39 and 48 %, respectively). This assessment may indicate that the drug dose needs to be increased.

This study showed that the upward logarithmic RCVs were higher than the conventional approach; the logarithmic RCVs in the downward change were lower than the conventional RCVs (e.g., conventional RCV for TSH 39 %, upward and downward RCVs 48 and 32 %, respectively) (Table 3). The fact that the logarithmic RCV is higher in the upward change compared to the conventional approach can be explained by the Fraser approach’s higher false positive rates, as demonstrated in Lund et al.’s study. In their study with a computer simulation model, the conventional approach’s false positive rate (caused by the confidence interval) increased even more at values where CV_T exceeded 10 %. In contrast, the rate did not change in the logarithmic approach [22]. In our study, the CV_T values of TSH, Tg, Anti-Tg, and Anti-TPO tests were greater than 10 %. Therefore, using logarithmic RCV for these tests may provide a more accurate assessment. In addition, the EFLM BV database and the study by Ricós et al. provide data from studies on healthy individuals [19], 20]. The validity of these data for monitoring disease has been questioned, but in general, it has been shown that CV_I values are indeed similar in health and chronic stable disease. Therefore, the current available BV data are ubiquitously applicable [6]. However, disease, medication, and abnormal metabolic status cause abnormal fluctuations of BV [10], 11]. Therefore, logarithmic RCVs may be more beneficial for evaluating thyroid-related biomarkers of individuals with these conditions.

When instrument-specific RCVs were compared, the RCVs of the TSH test were quite similar, while the RCVs of the fT3 and fT4 parameters in the E instrument were slightly lower than the R1 and R2 instruments (e.g., conventional RCVs of fT4 for R1, R2, and E are 22 , 21, 19 %, respectively) (Table 3). This difference was due to the better analytical performance of fT3 and fT4 hormones in the E analyzer. The RCV results were also quite similar, as the CV_A values of the parameters in the R1 and R2 auto-analyzers were close to each other (Table 2, 3). Since the CV_I and Z values in the RCV formula are independent of the auto-analyzer or the methodology used [23], CV_A is the only variable that affects the calculation when making inter-analyzer RCV comparisons for the same test. Therefore, similar RCVs are determined for auto-analyzers with similar analytical performance. However, since the analytical imprecision of each system is different, it is important to calculate device-specific RCV.

In the study by Erden and colleagues, in which they determined the BV and RCV values ​​of thyroid function tests, RCVs were determined bidirectionally with the classical approach within a 95 % confidence interval. They determined the RCVs as 104 % for TSH, 63.1 % for fT3, and 38.8 % for fT4 [24]. These results were remarkably higher than the RCVs calculated in our study. This difference occurred because the CV_I values ​​determined by Erden and colleagues (37.2 % for TSH, 22.3 % for fT3, and 13.3 % for fT4) were considerably higher than the data in the EFLM database, and they used a two-sided Z-score.

Cinpolat et al. calculated the RCVs of 24 different immunoassay parameters unidirectionally with the classical approach within a 95 % confidence interval. They determined RCVs as 45.2 % for TSH, 19.4 % for fT3, 14.6 % for fT4, 24.6 % for Anti-Tg, and 35.8 % for Anti-TPO [25]. Similar RCV results were obtained for the Anti-Tg and Anti-TPO tests, in which we used the same BV data (Ricós et al. study) in our study. There was a difference in TSH and fT4 tests. The different BV databases used and our possible analytical imprecision differences have caused these results.

Ağca, in his thesis on measurement uncertainty and RCV in biochemistry and hormone tests, RCVs were calculated bidirectionally with the traditional method within a 95 % confidence interval. His results were 53.6 % for TSH and 17 % for fT4 [26]. The difference in the results was thought to be due to the differences in the Z-scores and BV databases we used.

In their study, Walz and Fierz determined the RCVs of many analytes, and they calculated 99 % for the increase and −50 % for the decrease (one-sided with the logarithmic approach) for the TSH test [27]. It was determined that these values, which were considerably higher than our results, were caused by the different databases we used (EFLM/Ricós).

Bottani et al. estimated the BV data of thyroid biomarkers in the European BV study (EuBIVAS). They unidirectionally determined the RCVs of TSH, fT3, fT4, and Tg tests with the logarithmic approach within a 95 % confidence interval. RCVs were calculated as 50.7 , 13.1, 12.6, 29.0 % in the upward change and −33.6 %, −11.6 %, −11.2 %, −22.5 % in the downward change, respectively [28]. Our calculated RCVs for TSH were similar, but our results were higher for other parameters. Our study’s higher CV_A and CV_I values for parameters other than TSH caused these results.

In the meta-analysis study by Fernández-Calle et al., in which the BVs of thyroid-related measurements were estimated, RCVs of TSH, fT3, fT4, Tg, Anti-Tg, and Anti-TPO parameters were also determined. This study determined RCVs unidirectionally with the logarithmic approach within a 95 % confidence interval. RCVs were calculated as 55.1 , 14.9, 16.0, 32.2, 33.3, 43.2 % on the increase and −35.5 %, −13.0 %, −13.8 %, −24.4 %, −25.0 %, −30.2 % on the decrease, respectively [29]. While the RCVs calculated for TSH, Anti-TPO, and Anti-Tg tests were higher than our study, it was observed to be lower for other parameters. These differences were due to different BV data and analytical imprecision.

In the literature, different RCVs have been observed in studies on BV and RCV of thyroid-related markers [24], [25], [26], [27], [28], [29]. The analytical imprecision, BV data, and bi- or unidirectional Z scores resulted in these differences. Even if the BV data used in the studies are taken from the up-to-date database, and the Z values are used the same, the calculated RCV results will be different because the analytical performance of each laboratory will be different. In short, each laboratory needs to determine and use its own RCVs for the analyzed parameters. In addition, reporting the RCV together with the reference interval in the clinical laboratory report is recommended [30], 31]. In this way, biological and analytical variation can be considered and RCV can be beneficial as an objective tool in evaluating patient results [5]. However, reporting RCV results as a percentage exposes clinicians to mathematical calculations and can make it difficult for clinicians to interpret the results. Therefore, RCV can be reported using warning flags for test results that show significant changes [23], 31].

This study has some limitations. First, the lot information of the reagents used throughout 2018 was not recorded. Second, there was no data on the preparation of lyophilized IQC. The other was that since the BV data was obtained from databases simultaneously with the study, it may not overlap with current data, especially the EFLM database. Moreover, the BV data used may only partially reflect the variations in our population. Therefore, further studies may need to be conducted in which preanalytical standardization is provided, and RCVs are calculated by determining BV data. Another critical issue is the use of instrument-specific RCV, which requires that patient results be analyzed on the same instrument. However, patient samples can be analyzed on any instrument in laboratories like ours with multiple instruments. Therefore, in practice, a single RCV can be determined by calculating a pooled CV_A that includes all analyzers [32]. Lastly, the potential disadvantages of RCV are that not every change exceeding the RCV may be clinically significant: a) some biological variations may be affected by health status, b) diurnal variation, and c) differences in the timing of sample collection [6].

Evaluating the results of thyroid-related biomarkers is very important since thyroid dysfunction and the thyroid diseases that cause it are quite common in the community [12], [13], [14]. The II values of all tests in the study were 0.6 and below, indicating a high individuality. Therefore, TSH, fT3, fT4, Tg, Anti-Tg, and Anti-TPO test results should be evaluated using the RCV instead of the population-based reference interval [2]. With the use of RCV in the clinical laboratory, analytical and biological variations will be considered, and it will benefit clinicians as an objective tool in assessing thyroid biomarker results and in clinical decision management of thyroid dysfunction and diseases. Thus, it will accelerate clinical decisions in the treatment processes of thyroid diseases, shorten the discharge times of inpatients, and reduce unnecessary test repetitions, which will be very useful in terms of workload, time, and cost. In addition, a more accurate evaluation will be provided by using RCV to follow up on adverse effects [23], 25], 33], 34]. Significantly, instrument-specific RCVs should be determined because the analytical imprecision of analyzers may vary. Moreover, since most of those who apply to the hospital, like us, have a disease, medication, or abnormal metabolic cases, it may be more accurate to evaluate thyroid test results with logarithmic RCVs. In conclusion, RCVs should be presented with the reference interval when reporting thyroid-related test results in clinical laboratories.