https://orcid.org/0000-0002-9047-1556
Abstract
Many individuals with marginally abnormal thyroid function test (TFT) results may be treated and it is unknown if the limits of the thyrotropin (TSH) and free thyroxine (FT4) reference intervals reported alongside the laboratory results are associated with the prevalence of levothyroxine treatment. We obtained information regarding reported TFT reference intervals from UK National Health Service (NHS) laboratories and evaluated its relationship with the prevalence of levothyroxine treatment for corresponding health areas for 2014. The upper limit of serum TSH was significantly, linearly, independently, and negatively associated with prevalent levothyroxine treatment: −0.54% (95% CI, −0.68% to −0.40%). The lower limit of serum FT4 was significantly and independently associated with the prevalence of levothyroxine treatment in a non-linear (J-shaped) manner with an increase being noted from a FT4 level of ≈9.5 pmol/L onwards. We conclude that minor changes in the reference range limits for serum TSH and FT4 are associated with levothyroxine treatment.
Levothyroxine treatment is common and may be influenced by the reference limits of both serum TSH and FT4 as a proportion of individuals with marginally abnormal results may be treated. We analysed if TSH and FT4 reference limits utilised by UK NHS laboratories were associated with the proportion of levothyroxine-treated individuals in the corresponding area. We found that the values of the upper TSH and lower FT4 reference limits significantly and independently predicted the proportion of levothyroxine-treated individuals in that area. These results suggest that minor changes in TSH and FT4 reference range limits may influence levothyroxine treatment and it may be desirable to consider adopting thyroid function reference ranges based on clinical outcomes rather than statistical techniques.
Introduction
Measurement of serum thyrotropin (TSH) level with or without circulating free thyroxine (FT4) concentration is the cornerstone in evaluating a person's thyroid function.1 Hypothyroidism is diagnosed when the serum TSH level is high and circulating FT4 levels are low. Levothyroxine (LT4) therapy is the treatment of choice for hypothyroidism and its use is increasing in both the United Kingdom and the United States of America.2,3 There is large and significant geographical variation in the prevalence of LT4-treated hypothyroidism and various factors such as differences in proportion of older people, females, levels of smoking, prevalence of obesity, and frequency of testing have been implicated as possible explanations.3,4 The decision whether someone's thyroid function test (TFT) result is “normal” depends on where the value lies in relation to its reference range, derived from a healthy population sample. However, serum TSH and FT4 reference ranges are not universally agreed and laboratories adopt different criteria including manufacturers’ recommended range, local data, or data from large population samples.5 Thus, laboratories that utilise a narrower TSH reference range (for example, 0.4-4.0 IU/L) will likely diagnose more individuals with thyroid disease than those that utilise a wider reference range (for example, 0.2-6.0 IU/L). More people with borderline raised serum TSH levels are being treated with LT4 recently.6 However, it is unknown whether the limits of the TSH and FT4 reference ranges have an impact on the prevalence of LT4-treated individuals.
Methods
The aim of this analysis was to study the relationship between the upper limit of the serum TSH and the lower limit of the serum FT4 reference ranges with the prevalence of LT4 treatment in the United Kingdom in 2014. Data held by the NHS and the Office of National Statistics (ONS) were examined. We obtained data relating to TFTs from NHS laboratories and analysed corresponding LT4 treatment prevalence data for each UK health authority area from the Quality Outcomes Framework (QOF) for 2014.7 The QOF is a UK voluntary primary care pay-for-performance program with very high participation with the intention to reward “quality care” and to reduce variation7 (NHS Digital 2015. Quality Outcomes Framework 2014-2015: report for England v1.1. Available at: https://digital.nhs.uk/catalogue/PUB18887. Accessed September 4, 2022). Hypothyroidism was one of several clinical conditions that were included in the QOF until 2015. Points were awarded for maintaining an accurate database of LT4-treated individuals and not for initiating treatment with LT4, which remained a clinical decision. A treated hypothyroidism case was defined by a LT4 prescription for any person of any age. The prevalence of treated hypothyroidism was calculated by dividing the number of LT4-treated patients on the QOF register with the total number of patients registered in each participating practice. Additional data on demographics, tobacco smoking habits, and prevalence of obesity were obtained for each health area from the QOF and/or the ONS database for 2014. No institutional review board approval was required nor gained as all the data collected for this analysis are publicly available (QOF and ONS) or were obtained by contacting publicly funded NHS laboratories.
All NHS laboratories providing a biochemistry service were asked to provide details about the TFTs used in 2014 including the analyser, the reference ranges for TSH and FT4 and how they were derived, and the health area they covered.
The statistical package R (rms package, R Project, Institute for Statistics and Mathematics, R Core Team, version 3.2.2) was used to compute ordinary least squares regression with three restricted cubic splines to assess the relationship between the limits of the TSH and FT4 reference ranges and prevalence of hypothyroidism, after adjusting for the following proportions: individuals aged over 65 years, women, smokers, those with obesity, and White population.
Results
Of the 237 health areas (clinical commissioning groups) that provided healthcare to the UK population, we obtained TSH and FT4 reference range data for 193 areas from 111 laboratories (some laboratories provide a service for more than one area).
Assays and reference ranges
The majority (88.7%) of thyroid function testing was provided by 3 analysers: Roche Cobas e170, Siemens ADVIA Centaur, and Abbott ARCHITECT (Figure 1). The two other analysers that were used to measure thyroid function were Beckman Coulter Access and Siemens IMMULITE (Figure 2). As expected, there were substantial differences in the median and the range of the upper and lower limits of both serum TSH and FT4 reported reference ranges, between the various analysers used to measure thyroid function (Figure 2). Surprisingly, there was a wide variation in the reported values for the upper and lower limits for TSH and FT4 within each analytical method too. For example, the reported values of the lower and upper limits of serum TSH and FT4 for Roche Cobas e170, the most commonly used analytical method, ranged between 0.10-0.40 and 4.00-6.00 mU/L, and 7.5-12.2 and 14.4-28.0 pmol/L, respectively. The reasons quoted by the laboratories underlying the reference ranges also varied: manufacturer's range (63%), historical data carried over from previous assays (22%), obtained from local population sampling (4%), from the literature (4%), agreed as a clinical action limit between biochemists and clinicians (4%), and unknown (3%).
Percentage of various analysers used by laboratories to analyse serum thyroid function.
The reported reference ranges for serum TSH and FT4 across the various assay platforms. (A) Lower (lower section) and upper (upper section) limit of normal for TSH concentrations. (B) Lower (lower section) and upper (upper section) limit of normal for FT4 concentrations. Middle thick horizontal lines depict the median values. The number of laboratories using each analyser was Roche Cobas e170 (n = 40), Siemens ADVIA Centaur (n = 30), Abbott ARCHITECT (n = 28), Beckman Coulter Access (n = 14), and Siemens IMMULITE (n = 4).
Relationship between the TSH and FT4 reference ranges and prevalence of LT4 treatment
There was a strong linear inverse relationship between the upper limit of serum TSH and the prevalence of LT4 treatment: for each unit increase in the upper limit, there was a reduction in the prevalence of LT4 treatment by 0.54% (95% CI, −0.68% to −0.40%) (Figure 3A). Conversely, there was a non-linear (J-shaped) relationship between the lower limit of FT4 level and the prevalence of LT4 treatment: an increase was noted from a FT4 level of ≈9.5 pmol/L onwards (Figure 3B). Among the other independent variables in the model, the presence of obesity (P = .004), proportion of White population (P < .001), and proportion of individuals aged more than 65 years (P < .05) were significantly and positively associated with the prevalence of LT4 treatment. To ensure the robustness of the results obtained, we examined if there was a relationship between the lower limit of serum TSH and the upper limit of FT4 with the prevalence of LT4 treatment: no significant association was noted (data not shown).
Relationship between the upper limit of the serum TSH (A) and FT4 (B) reference range and prevalence of levothyroxine treatment in the various UK health areas. Plot shows the association of levothyroxine treatment prevalence as the predicted mean with 95% CI with the upper limit of serum TSH (A) and lower limit of serum FT4 (B). All analyses are from a multivariate linear regression including the proportion of individuals aged over 65 years, women, smokers, those with obesity, and White population.
Discussion
This analysis demonstrates, for the first time to our knowledge, that the reference ranges of serum TSH and FT4 utilised by laboratories have a significant association with the prevalence of LT4 treatment. In addition, our investigation also reveals that many laboratories use different reference ranges derived from various sources for serum TSH and FT4 even when utilising the same analyser. These results suggest that laboratories need to agree a degree of uniformity when reporting reference ranges (whether this is the central 95% of the healthy population or a clinical action limit) as it may influence clinical decision-making.
The reference ranges for TSH and FT4 are influenced by several factors including the type of analyser used.8 Four analysers that are commonly used to measure thyroid function parameters (Roche Cobas e170, Siemens ADVIA Centaur, Abbott ARCHITECT, and Siemens IMMULITE) have good agreement at low-normal TSH levels but have inter-assay differences of approximately 1 mU/L at higher concentrations.9 Evaluation of serum TSH and FT4 values in younger and middle-aged adults showed marked variation particularly in the calculated upper reference values between various analytical platforms.10 Thus, having similar reference ranges for serum TSH and FT4 by laboratories using the same type of analyser for populations that are broadly similar would be sensible. Our analysis suggests that laboratories have variable criteria in determining the reference ranges for TFTs and this may be impacting clinicians’ prescribing practice.
Whilst our analysis did not directly assess the relationship between the limits of the reference ranges and the clinician's decision to treat individual patients, we hypothesise that clinical practice is influenced by the values of the reference range limits used to signal “normal” thyroid function. The upper (or lower) limit of a reference range indicates the limits of the central 95% of the values derived from a healthy population and is not by itself a threshold for treatment. The limitation of using reference intervals derived from 95% of a healthy population is that 5% of individuals will automatically be classed as having values outside of the range and the range will get narrower with time as more individuals are excluded in repeat exercises. Given the substantial level of uncertainty regarding the effectiveness of LT4 therapy in SCH and the fact that TSH levels change with age and body weight, amongst other factors, more research is required to assess the benefits and risks of utilising different TSH reference ranges, particularly in selected patient groups such as those older than 65 years or those with obesity.11 Most clinical guidelines recommend consideration of treatment when serum TSH is unequivocally greater than 10.0 mU/L12,13 (Thyroid disease: assessment and management. NICE guideline [NG145] Available at: https://www.nice.org.uk/guidance/ng145/chapter/recommendations#investigating-suspected-thyroid-dysfunction-or-thyroid-enlargement. Accessed January 5, 2023). Furthermore, it may be desirable to adopt reference ranges based on clinical outcomes rather than statistical techniques.14 Alternatively, harmonisation of TSH and FT4 assays across the various analytical platforms could allow results to be transferable across health areas.15 This is especially important as neither TSH nor FT4 assays show high degree of agreement across manufacturers although FT4 is difficult to standardise and comparability across manufacturers is poor.10
The strengths of this analysis are the large number of laboratories that provided data on TFTs, the source of the reference ranges used, and the assessment of the relationship between the limits of serum TSH and FT4 with the prevalence of LT4 treatment across multiple areas. The limitations include non-availability of data for all laboratories serving the entire United Kingdom, the fact that data were obtained from 2014 and some assay methods may have changed, and lack of direct evidence that the reference range limits of TSH and FT4 influenced clinical prescribing.
In conclusion, serum TSH and FT4 reference ranges differ across laboratories, even when the same analyser is used, without good reason. Variation in the upper limit of serum TSH and the lower limit of serum FT4 is associated with the prevalence of LT4 treatment. Harmonisation of TFT reference ranges with the aim of developing more uniform reference intervals based on well-defined healthy populations or reference ranges based on clinical outcomes should be considered.
Acknowledgments
We would like to thank all the biochemists and clinicians that responded to our query regarding the analysers used to measure thyroid function in their laboratories and other related questions.
Author contributions
S.R.: conceptualised the project (lead), data collection (equal), data analysis (lead), writing original draft (lead), reviewing, and editing manuscript (lead). A.J.: data collection (equal). J.V., C.A., A.S., H.S., and O.L.: data collection (equal), reviewing, and editing manuscript (equal).
Funding
None declared.
Conflicts of interest: None declared.
