Interpretation of TSH results can be improved by reference values fluctuating in time

Introduction

The determination of TSH in serum or plasma has become one of the most frequently requested laboratory tests [1]. This rise in popularity can be attributed not only to the growing recognition of the wide-ranging health implications of thyroid disease but also to significant advancements in the simplification and cost-effectiveness of TSH analysis techniques.

Serum thyroid stimulating hormone (thyrotropin, TSH) measurement is of primary importance in thyroid hormone disorders but is secreted episodically, a process generated by the suprachiasmatic nuclei (SCN), the central biological clock [1], [2], [3], [4]. TSH concentrations are characterized by a circadian rhythm; a peak can be observed in the morning and a nadir between 5 pm and 8 pm. It also shows secretory pulses every 2–3 h with tonic non-pulsatile periods among them [5]. The amplitude of these pulses is small and only creates insignificant variation because of the long plasma TSH half-life [6]. In function of time, amplitude of TSH variations nearly doubles periodically every 24 h [7]. Rhythmicity of the thyroid hormones triiodothyronine (T3), and thyroxine (T4) is not so clear as TSH. Circulating T4 concentrations are more stable because of the hormone’s long half-life. T3 has a shorter biological half-life, its levels are only following the TSH rhythm to a small extent because only a minor part of T3 is released by the thyroid gland [8].

For any analyte with circadian rhythm, sampling time is very important, ideally requiring standardization for an accurate clinical interpretation. Clinicians mostly use reference intervals to compare the laboratory test results. Reference values are generally calculated out of values of reference individuals/data obtained from the measurements in the morning. As in clinical practice, sampling occurs all day long, the results of samples taken at other times of the day should not be compared to the uniform standard reference intervals. For biochemical analytes characterized by an important biological variation, decision limits need correction for sampling time. Intra-individual variability (CVI) is obtained from the daily samples usually measured in the morning. Sample results taken at different times of day should not be used for comparison with RCVs [9]. Therefore, knowing the effects of sample collection time on TSH values are important when interpreting TSH results. In subclinical hypothyroidism (SCH), where the diagnosis is solely based on increased TSH concentration, sampling time is of utmost importance. According to the Clinical and Laboratory Standards Institute (CLSI), managing pre-analytical factors is crucial to minimize their impact on clinical decision-making. Circadian rhythms of free thyroxine, free triiodothyronine and TSH have been studied in literature; however, there are only a limited number of studies with a focus on this rhythm’s impact on laboratory reference values and its clinical consequences. In this extensive study, based on a large cohort of outpatients, we explored the possibility to construct a mathematical model (cosinor function) based on the evolution of TSH concentrations from 7 am to 5 pm in order to solve the hypothyroidism decision problem in a real world and to support personalized clinical decision-making.

Materials and methods

Results

In the study group of outpatients, gender and age distributions of time groups were found to be homogeneous. Median TSH patient values according to sampling times are presented in Table 1. No significant differences were found in any of the time groups between genders. TSH levels showed a gradual decrease in proceeding sampling times compared to the samples taken at 8–9 am. The distribution of the median, 2.5 and 97.5 percentile of the TSH values in the studied population showed a pattern corresponding with a sine wave (Figure 1).

Figure 1:

Plasma TSH (mIU/L) cosinor function of the hyperthyroidism decision limit (red line), the median (blue line) and the 97.5 percentile (green line) in function of time (h).

In the cosinor model, the TSH follows a sine wave over time which is characterized by a phase shift (location of the peak and trough along the time axis), height (vertical shift of the sine wave), and amplitude (maximum variation of the sine wave from its mean height). Results showed that plasma TSH had a diurnal rhythm with nocturnal peaks. The following equations were obtained for the various percentiles:

In which T represents the time (hour).

For clinical purposes, the P2.5 values are used. Based on the relative variation at the lower TSH concentration range, the TSH cutoff value was modeled as follows: corrected TSH cut-off (mIU/L)=0.40 + 0.24*cos(π/12) *(h − 6), in which h represents the time (hour) and the argument of the cosinus function is expressed in radians.

The implementation of the time-dependent cut-off values has clinical consequence. Tables 2a and b compares the effects of the corrected cut-off values on the diagnosis of hyperthyroidism. As compared to a fixed cut-off value for TSH throughout the day (0.4 mIU/L), the relative number of diagnosed hyperthyroidism cases dropped from 3.78 to 2.58 % in the morning (a relative decrease of 26 %, n.s.). In the afternoon and evening, the decrease was more pronounced: from 4.59 to 2.75 % (a relative decrease of 40.1 %) (p<0.0001).

Discussion

Circadian rhythm of TSH has been repeatedly investigated in medical literature but very few studies focused on its impact on routine laboratory and clinical practice. In our study, TSH levels have been analyzed in detail through the day. In agreement with literature, our study mainly demonstrates a significant difference in serum TSH levels if sampling is carried out at in different moments of the day.

Starting with the early morning, TSH showed a gradual decrease until 1 pm, showing a relative stability afterwards up to 4 pm. TSH deviations of all sampling time groups beginning with 10–11 am hour were greater than desirable/minimum bias values. Besides testing for statistical significance, in this study, we compared deviations with bias values of analytical performance specifications from the EFLM database [14] to assess clinical significance. Biological variability data are reference data providing better applicable performance standards.

The circadian rhythm of TSH secretion has been a research topic in the literature, and a similar curve showing a maximum between 2 and 4 am and the minimum during daytime is observed [6], 15], putting special emphasis on sampling times [16], 17]. Our study has demonstrated this daily rhythm in detail and correlated well with the described TSH release pattern.

The magnitude of the variation in TSH concentration observed due to the circadian rhythm (the zenit values showing an approximatively threefold value of the nadir plasma values) exceeds both the inter-vendor variation and the CV of the TSH assay. Commercial TSH immunoassays show standardization differences with a bias of 1–14 % for the low range [11]. This could lead to a different classification of 1.5 % of all measured TSH concentrations <0.40 mIU/L [10]. As the imprecision of the immunoassays varies from 1.6–5.5 %, this could lead to a similar reclassification as the bias between immunoassays.

Meta-analysis has revealed that the mean analytical variation (CV_A) for TSH is 6.4 %, the CV_I, 17.9 %; and CV_G, the between-subject coefficient of variation (CV_G) 36.2 %; The reference change values (RCV) calculated for TSH was −35.5 to 55.1 % [18].

Food intake has been mentioned as a source of variation in some reports, concluding that it may cause a proceeding decrease of TSH values [19], 20]. The observed trend overlaps with the TSH secretion pattern, which makes it unclear whether nutritional intake has an effect on the TSH secretion rhythm. A controlled small-sized study was carried out in which. TSH values were evaluated in three experimental designs, each controlling one of time of sampling, intraindividual variation and food intake. Differences in TSH values between two sampling times showed no correlation with the fasting state [16]. The unavailability of information regarding the fasting states does not allow to draw conclusions about this issue.

There are limited controversial data sets available for T4 and T3. T4 and T3 show a morning maximum and an evening nadir. However, the T4/T3 fluctuations were considered not important in daily practice [21]. Another study demonstrated synchronous rhythms for TSH and T3 with a very limited change in T3 concentrations, which was attributed to the low fraction of T3 secreted by the thyroid gland [8]. Some studies have failed to find a rhythm in T4 and T3 [22], 23]. The absence of effect on thyroid hormone concentrations by nighttime increase of TSH values has been explained by the secretion of glycosylated TSH isoforms with reduced bioactivity during the night [24].

The observed circadian rhythm is a major contributing factor to the total biological variation of plasma TSH [18], 25], 26]. In relative terms, the effect is more important in the lower TSH concentration range.

In a study involving patients with hypothyroidism, circadian changes in TSH levels reached 73 % in those with sublinical hypothyroidism. In consequence, 50 % of subclinical hypothyroidism cases detected in the morning could not be diagnosed in the afternoon when compared with the same reference intervals [26].

As in our case study, 28 % of blood samples were obtained in the afternoon or evening. This implies that, for many patients, interpretation of TSH values in serum or plasma is not easy. In blood samples collected at different times of the day, we observed statistical and clinical significant variations in TSH values. The circadian rhythm of TSH has various important implications because they allow us to assess the utility of reference values and to determine the significance of differences in serial measurements (i.e., the reference change value, or critical difference). Implementing the cosinor-function derived correction for the TSH cut-off values resulted in a significant reduction of (false-positive) diagnosis of hyperthyroidism, in particular in the samples sampled in the afternoon and evening.

TSH decision limits should be corrected for to the time of the day with at least hourly interval based reference intervals. This will provide accurate interpretation of patients and allow a more appropriate treatment.