Application of composite reference intervals in the diagnosis of subclinical hypothyroidism in the elderly: a retrospective study

Exploring the effects of gender, age, thyroid antibody, blood lipids, etc. on TSH levels in elderly SCH patients in Chengdu, China

Establishing two-dimensional composite reference intervals for TSH and FT4 in the elderly

The sensitivity and accuracy of the established two-dimensional composite reference intervals were compared with those of the traditional univariate reference intervals in diagnosing SCH

In this study, two reference intervals were used to diagnose SCH. One was the traditional standard, the other was the two-dimensional composite reference interval. The number of SCH patients was obtained by the two reference intervals. In addition, the differences in diagnosing were also compared.

TSH level and its influencing factors in the elderly

Basic information about the subjects

A total of 1,322 self-reported healthy elderly Han Chinese aged ≥60 years in Sichuan province were included. There were 759 males (57.4%) and 563 females (1.3:1). According to age, they were divided into three groups, including 915 (69.2%) in the junior-old group, 309 (23.4%) in the middle-old group, and 98 (7.4%) in the oldest-old group. There were 1,194 patients in TPOAb-negative group (90.3%), and 128 patients in the TPOAb-positive group (9.7%). There were 99 (7.5%) patients in the TgAb-positive group, and 1,220 cases in the TgAb-negative one (92.5%). There were seven patients with subclinical hyperthyroidism (0.5%), and 183 patients with SCH (107 males, 76 females) (13.8%). Seventeen patients with TPOAb-positive in the SCH group, (9.6%). There were 17 patients with SCH in the TPOAb-positive group likewise (13.3%).

The mean age and waist-to-hip ratio were 67.6 and 0.89. The mean body mass index and blood pressure were 24.1 kg/ (m²) and 132/74mmHg. The mean low-density lipoprotein was 3.04 mmol/L. The median of cholesterol and high-density lipoprotein were 4.76 (2.27−9.01) mmol/L and 1.34 (0.67−3.27) mmol/L. The median of triglyceride and glycosylated hemoglobin were 1.1 (0.31–15.18) mmol/L and 5.5 (3.6–13.3) %. The median values of creatinine, urea, uric acid, and TSH were 71.0 (25.0–191.0) umol/L, 4.60 (2.1–18.2) mmol/L, 324 (110-672) umol/L, and 2.35 (0.01–12.74) mU/L, respectively. The median values of T3, FT3, and T4 were 1.76 (0.65−5.06) nmol/L, 5.05 (1.82–18.07) pmol/L, and 97.83 (23.31–227.0) nmol/L, respectively. The median values of FT4, TgAb, and TPOAb were 16.8 (3.32–43.75) pmol/L, 11.9 (10.0–4000.0) IU/ml, and 11.36 (5.0–600.0) IU/ml, respectively.

The differences of TSH, FT3, and FT4 among junior-old, middle-old, and oldest-old group

The analysis found that there were no statistically significant differences in TSH, FT3, and FT4 among the three groups.

Comparison of differences between the SCH group and regular control group

We compared the intergroup differences between the normal thyroid function group(843 cases) and the SCH group (183 cases). We found that there were significant differences in creatinine, uric acid, estimated glomerular filtration rate (eGFR), TSH, FT3, T4, FT4, and TPOAb levels between the two groups (P < 0.05), using the Wilcoxon rank sum test. What’s more, the levels of creatinine, uric acid, eGFR, FT3, T4, and FT4 in the SCH group were lower than those in the regular group. The levels of TSH and TPOAb were higher than those in the regular one (Table 2).

Establishment of two-dimensional composite reference intervals for TSH and FT4

Junior-old group $D^2={\left(\frac{\sqrt{TSH+0.386}-1.595}{0.407}\right)}^2+{\left(\frac{\sqrt{FT4+0.117}-4.098}{0.274}\right)}^2$

Middle-old group $D^2={\left(\frac{\sqrt{TSH+0.403}-1.530}{0.368}\right)}^2+{\left(\frac{\sqrt{FT4+0.006}-4.112}{0.253}\right)}^2$

Oldest-old group $D^2={\left(\frac{\sqrt{TSH+0.057}-1.589}{0.361}\right)}^2+{\left(\frac{\sqrt{FT4+0.643}-4.155}{0.273}\right)}^2$

A unified two-dimensional composite reference interval was constructed by using the whole elderly population as the reference population: ${\displaystyle D^2={\left(\frac{\sqrt{TSH+0.377}-1.581}{0.399}\right)}^2+{\left(\frac{\sqrt{FT4+0.144}-4.104}{0.268}\right)}^2.}$

The constants 0.377 and 0.144 are the standardized skewness coefficients of TSH and FT4, respectively; 1.581 and 4.104 are the standardized mean values of TSH and FT4, respectively; 0.399 and 0.268 are the standard deviations.

Both FT4 and TSH have two corresponding values of TSH and FT4 separately, which are the ranges of TSH and FT4. For example, when FT4 is 12pmol/L, the corresponding range of TSH is 4.0−8.0. When the TSH value is 6.8mIU/L, the range of FT4 values is 15-20. When D² >5.99, hypothyroidism was determined.

Validation of two-dimensional composite reference intervals

183 and 130 SCH patients were diagnosed by using the traditional and the two-dimensional composite reference intervals as the diagnostic standard, respectively. The use of two-dimensional composite reference intervals reduced the number of patients diagnosed with SCH by 4%. All patients were tested for TPOAb, and 128 were positive. Among the TPOAb-positive patients, 17 and 11 SCH patients were diagnosed by using the traditional and the two-dimensional composite reference intervals as the diagnostic standard, respectively.

When we used the two-dimensional composite reference interval formula for the junior, middle, and oldest-old diagnostic criteria accordingly, the number of diagnosing SCH was 75, 37, and nine, respectively, for a total of 121 cases. Compared with the two-dimensional composite reference interval of the whole population, the total number of patients differed by 9. After the chi-square test analysis, there is no statistically significant difference between the formula of the three groups and the whole population. Therefore, the two-dimensional composite reference interval of the whole population can be used as the unified diagnostic standard.

TSH level and its influencing factors in elderly with subclinical thyroid dysfunction in the local area

It was found that the height, weight, and diastolic blood pressure of the elderly showed a downward trend with the increase of age. However, waist circumference, waist-hip ratio, and systolic blood pressure showed an upward trend. It is suggested that the elderly are prone to abdominal obesity and suffer from hypertension with high systolic blood pressure. Xue et al. (2022) included 13,251 elderly population with abdominal obesity of 951(34.34%). The present study did not find TSH levels increased with age, which is inconsistent with other studies. In addition, the sixth round of the census showed that the prevalence of hypertension in people aged 60–69 years increased slowly over time (Zhang et al., 2023). However, Some scholars pointed out TSH is negatively correlated with age. Xue & Feng (2012) included 2,443 patients, dividing the population into junior-old (<60 years old), middle-old (60–79 years old), and oldest-old (≥80 years old) groups. It was found that TSH showed a downward trend with the increase in age, and the decline rate of TSH in the oldest-old group was higher than that in the middle-old group. While some scholars believe that there is little correlation between TSH and age. Li et al. (2011) included 602 healthy physical examination subjects and divided them into non-old (<60 years old), junior-to-middle old (60–79 years old) and oldest-old (≥80 years old) groups. The correlation between TSH and age is low, and TSH increased with age only in the oldest-old group. Multi-center studies at home and abroad have found many factors affecting TSH, and the correlation between age and TSH cannot be considered only. The Third National Survey on Nutrition and Health (NHANES III) in the United States pointed out that the reference range of TSH should be based on age, race, and gender. After 39 years of age, the 97.5th percentile of serum TSH increased by 0.3mIU/L for every ten years of age increase (Boucai, Hollowell & Surks, 2011; Zhai & Shan, 2014). A multi-center study has found that age, gender, and region were all the influencing factors of TSH reference interval in the elderly. In addition, TSH did not increase with age in elderly women. There is a positive correlation between TSH level and age in elderly male patients, and there is a significant difference after the age of 75 (Wang et al., 2023). In our study, the serum TSH level did not increase with age, which may be related to the inclusion of only the elderly but not the non-elderly, as well as different regions and laboratory detection methods.

Our study found that the levels of creatinine, uric acid, eGFR, FT3, T4, and FT4 in the SCH group were lower than those in the normal group. TPOAb-positive was higher than that of normal elderly. While the rest of the clinical indexes with none of significant importance. The phenomenon reminds us that elderly patients with SCH have no obvious clinical manifestations. For instance, the renal metabolic indexes perform better in the SCH group than in the normal elderly. Existing studies are controversial on whether SCH affects the renal system. Gao et al. (2022) retrospectively analyzed 103,513 subjects (18-101 years) and found that serum creatinine gradually increased with aging. In addition to renal function, the influencing factors of serum creatinine include muscle weight, gender, age, and exogenous food intake Meuwese et al. (2019) pointed out that there was no significant difference in renal function decline between patients with normal thyroid function and patients with SCH. However, unlike Meuwese et al. (2019), Kim et al. (2023) showed that SCH is an independent risk factor for chronic kidney disease. The relationship between TSH and serum uric acid is controversial in existing studies. Feng et al. (2021) included 254 type 2 diabetic patients with early renal damage. It was found that TSH is negatively correlated with blood uric acid. Feng et al. (2021) further found that a decrease in TSH is an independent risk factor for subjects with hyperuricemia, speculating that the negative correlation between TSH and serum uric acid was the result of insulin resistance. Yang & Cao (2022) included 19,013 subjects and found that the risk of hyperuricemia was related to elevated TSH; Male patients with elevated TSH values were more likely to suffer from hyperuricemia.

Among the included elderly population in the local area, the median TSH level of SCH elderly was 5.06 (4.22–12.74) mIU/L. The correlation analysis showed that the level of TSH in the elderly increased with the decrease of FT3, FT4, creatinine, and uric acid. Multiple regression analysis showed that serum FT4 and uric acid were the most critical factors affecting TSH in elderly patients with SCH. On the one hand, the secretion of TSH is affected by the promotion of TRH secreted by the hypothalamus; on the other hand, it is also affected by the feedback inhibition of T3 and T4. The two regulate each other, and they form the hypothalamic-adenopituitary-thyroid axis. Under normal circumstances, there is a negative correlation between FT4 and TSH to maintain a relatively stable level of thyroid hormone. Like the present study, Hoermann et al. (2010) included subjects with different thyroid functions and confirmed the negative correlation between TSH and FT4. De Grande, Van Uytfanghe & Thienpont (2015) also observed a negative correlation between TSH and FT4. They included 8152 patients (18–100 years old) and divided the subjects into three groups of low, medium, and high concentration according to TSH levels, and found that there was a negative correlation between TSH and FT4 in all three groups. Benhadi et al. (2010) enrolled 21 healthy volunteers (average age 60 years old), and randomly assigned them to receive an oral placebo, 125 mg T4, and 250 mg T4; Fasting blood samples were collected from volunteers before and after administration. The study found that after administering T4, serum FT4 levels increased while TSH levels decreased, indicating a negative correlation between FT4 and TSH (Benhadi et al., 2010). In clinical practice, patients with elevated TSH should pay attention to screening for any decrease in blood FT4 values and uric acid levels to avoid missed diagnosis.

Establishment and verification of two-dimensional composite reference intervals for TSH and FT4 in the elderly

In this study, 1,194 subjects with normal TPOAb levels were screened out from 1,322 elderly people and included in the study to establish a two-dimensional composite reference intervals Finally, the two-dimensional composite reference interval of the elderly population was constructed.

At present, there are few reports on the establishment of two-dimensional composite reference intervals for TSH and FT4 in the elderly. In this study, the traditional reference interval and two-dimensional composite reference interval were used to diagnose SCH, and the difference in the diagnostic rate of the two methods was compared. The two-dimensional composite reference interval could reduce 4.0% of the patients diagnosed as “SCH” compared with the traditional reference interval. This difference was as high as 4.7% in TPOAb-positive subjects. This is consistent with the demonstration results of the “two-dimensional composite reference interval” established by Ross et al. (2009), as well as our previous research (Liang et al., 2014). The traditional diagnosis of SCH is combined and classified by the cut-off values corresponding to the 95% reference intervals of FT4 and TSH in the healthy reference population. Even if a report with TSH, FT3, and FT4 is obtained, it is interpreted separately, which may increase the probability of misdiagnosis. Several studies have shown that the relationship between TSH and FT4 is not a simple linear correlation. Hoermann et al. (2010) included subjects with different thyroid function statuses (hypothyroidism, normal, and hyperthyroidism), introducing a nonlinear mathematical model. It was found that the relationship between TSH and FT4 is not a simple linear logarithm (Hoermann et al., 2010). Ross et al. (2009) took FT4 and TSH levels of healthy adults as two-dimensional variables at the same time, and conducted two-dimensional probability distribution research after logarithmic normalization, and established the 95% reference interval of FT4 and TSH paired values. Similarly, Hoermann et al. (2016) established a two-dimensional composite reference system based on 271 healthy subjects, validating it in 820 untreated subjects with thyroid disease. The results showed that 26% of the patients diagnosed with “thyroid dysfunction” according to the traditional criteria were “euthyroid” according to the two-dimensional composite reference system (Hoermann et al., 2016). A two-dimensional composite reference interval was used to represent the association between TSH and FT4 in our study. The distribution of FT4 in healthy people was approximately normal, while the distribution of TSH was skewed. After logarithmic transformation, both FT4 and TSH showed normal distribution. The scatter distribution of TSH and FT4 in the traditional reference interval is similar to a duck egg, rather than a regular circle. Thus, the traditional reference interval increases the operability and difficulty of accurately measuring 95% of the data volume, which may increase the accuracy and difficulty of diagnosing diseases in this reference interval. The transformation was fine-tuned by correcting residual skewness, and FT4 was log-transformed and normalized to derive a regular circular two-dimensional composite reference interval, which was convenient for observation and diagnosis of 95% normal population distribution. Compared with the established two-dimensional composite reference interval, the traditional reference interval may be more likely to overestimate the incidence of SCH in the elderly, resulting in overdiagnosis and treatment. We should also notice another possibility that the two-dimensional composite reference interval may lead to the “underdiagnosis” of SCH. It indicates that clinicians should be careful to diagnose SCH with dynamic follow-up of thyroid function. The sample size of this study is relatively small, which affects the accuracy and stability of the research conclusions. In the future, the sample size will be increased to verify the reliability and validity of the two-dimensional composite reference interval.