Sample size verification

Test for sample size (N = 425), to see if it was enough to derive the reference interval (RI), was done based on the data obtained, as cited in CLSI guidelines6 applying the criteria recommended by Harris and Boyd i.e. Range of 90% Confidence interval (CI) at each Reference limit (RI) divided by the central 95th percentile RI should be less than 0.2. As can be seen from Table 5 in results section, sample size taken for this study has been verified to be adequate.

Method of recruitment

This study used direct sampling strategy wherein, healthy reference individuals were selected randomly by questionnaire. Healthy subjects were recruited in the age group 18 to 65 years, after taking informed consent. Thorough physical examination was done and exclusion criteria were applied as mentioned in table below. Blood samples were collected from selected subjects who were fasting for minimum 8 hours and serum separated, aliquots were prepared and kept frozen till study.

Direct selection of reference individuals is the only method that agrees more with the concept of reference values as recommended by the IFCC.3 Care was taken to exclude cases of subclinical thyroid dysfunction by way excluding individuals positive for of anti-TPO test. Generic exclusion criteria were followed for selection of subjects as per CLSI EP23.

The following exclusion criteria were used for serum sample selection:

The information in Figure 1 provides the gender-wise and age-wise distribution of samples selected. Applicability of partitioning was examined gender-wise and age-wise.

The method most often recommended for establishing reference intervals is the non-parametric approach, as mentioned in CLSI guideline EP 28 A3C, For both analytes, a non-parametric method was applied to estimate the indirect reference intervals. The 90% confidence intervals for lower and upper limits were calculated according to the recommendations of the CLSI.6

Analysis of samples was done on VITROS immunoassay based on enhanced chemiluminescence platform. The samples were collected in closed blood collection system using vacuum tubes with gel separators. Before analyzing the data, the calibration accuracies and QC data were examined and accepted. The following analyses were done on the reference interval obtained.

Reference interval

As can be seen from Figure 1, Figure 2, Figure 3 the range of values in the current study after removing outliers as in the Box and whiskers plot differ from the central 95^th percentile method (Table 3 ). The difference between the study and kit insert reference limits were more at the lower range for TSH and at the upper limits for TT4 and TT3.

On examining the outliers in this study for TSH it was seen to up to 12.7 mIU/L of TSH (Figure 2), for TT4 values up to 48 on lower side and 223 on the higher side (Figure 3) and for TT3, values up to 0.80 ng/mL on lower side and up to 3.39 ng/mL on the higher side (Figure 4). It is worth mentioning that the outliers for TSH and TT4 were more at the lower end of the reference interval in males as compared to females, after partitioning for gender. Ioannis Legakis et al 10 postulated that women with either anti-TPO or anti-Tg showed higher levels of TSH. In this study while samples which were anti-TPO positive was excluded, possibility of inclusion of women with anti-Tg cannot be ruled out. E.Arseneau and C.M. Balion have opined that outliers may represent a natural variability within a given group of individuals, more so in the elderly. 11

Table 4 shows the reference limits with their 90% confidence limits. This ensures practical applicability of the given reference range for clinical use, based on statistical probability for trueness.

The test for sample size as per CLSI (Table 5) shows that the sample size (N = 425) in the present study was appropriate. It can also be noticed from Table 5 that there is a wide gap between the passing ratio of lower and upper limits of reference limits for TSH, as compared to TT3 and TT4, which probably indicates that the sample size is critical for ensuring reliable upper limit of TSH in the reference interval.

Need for partitioning gender-wise

The tests for partitioning in current study indicate the need to partition the obtained reference interval gender-wise for TT4 only while there is no need for gender-wise partitioning for TSH and TT3 assays. Sedat et al1 found gender bias in females which was higher at the median for TSH, which was statistically different with p<0.05. Camilla SJ et al in their study, 12 after excluding TPOAb-positive subjects and outliers, in a reference population of 511 subjects, found no statistically significant gender- or age-specific differences in mean TSH and reference intervals. Hubl W et al 13 found no difference in TSH and TT3 between the genders but that significant difference in TT4. These findings 12, 14 agree well with the findings of the current study.

Proportion of population above NACB recommended cut-off

It has been proposed by Jee-young Oh et al15 that population with TSH above 2.5 units should be investigated for metabolic syndrome even in a normal healthy population. Interestingly, as if to support these facts, data indicating that African-Americans with very low incidence of Hashimoto thyroiditis have a mean TSH level of 1.18 mIU/liter4 whereas the mean TSH was 2.35mIU/L (Table 3) in current study.

Jee-young Oh et al.15 observed in their study that the high-TSH group (>2.5 mIU/L) represented 16.4% of the normal female population, whereas in the current study, the incidence was 37%, which is significantly higher. This higher incidence in the current study can have profound significance as Jee-young Oh et al.15 found prevalence of metabolic syndrome significantly higher in the high-TSH group Vs low-TSH group (7.5% vs. 4.8%, p = 0.016) with a 2-fold greater risk of metabolic syndrome than subjects in the low-TSH group after adjusting for age and BMI (odds ratio, 1.9; 95% confidence interval, 1.1 to 3.2).

However, this should be viewed with respect to the observation by Brabant G et al16 that it is still recommended to maintain the TSH reference interval of 0.4–4.0 mU/l. Classifying subjects with a TSH value between 2 and 4 mU/l as abnormal, as well as intervening with thyroxine treatment in such subjects, is probably doing more harm than good. In fact, in the current study, the median itself was 2.29mIU/L, very near the upper limit for TSH suggested by NACB (Table 4).

Waise A and Price HC17 caution that the upper limit of the reference range for thyroid-stimulating hormone should not be confused with a cut-off to define subclinical hypothyroidism, especially in the light of the NACB’s recommendations cited earlier. It is opined that reducing the upper limit of the TSH reference range would result in an increase in the proportion of the population diagnosed with subclinical hypothyroidism, frequently including those with non-specific symptoms probably unrelated to thyroid disease.

Effect of age on reference interval

The magnitude of difference in reference intervals between age groups 18-40 and 41-65 years for TSH was high but not statistically significant to warrant partitioning, though there was slight increase in TSH mean and median as age increased (Table 9 and Table 10). The range of TT3 in age group 18-30 years differed significantly from other slabs of age groups and was higher than the consolidated group of 31 – 65 years (Table 9, Table 10). This may warrant considering a separate reference interval for TT3 in age group 18 – 30 years of age as per this study to be clinically relevant.

Review of literature provides mixed evidence on this. Increasing TSH on ageing was observed for both men and women in study by Salman Rizvi et al 18, wherein they found the lowest TSH levels in middle-aged individuals, with higher values in younger and older age groups. Kratszch et al 19 reported that age was independently and inversely associated with TSH, when considering <40 and more than 40 years. Similarly, Hoogendoorn et al, 20 reported that serum TSH decreases gradually with age, whereas serum free T4 increases with age in individuals older than 60 years. In larger age- groups, Fontes et al 5 found that there was significant difference in TSH between two subset of population, i.e. 20 to 60 years of age 60 to 80 years of age and 80 years or more (0.4 - 6.7 mU/L).

However, studies have demonstrated that age-specific reference intervals for TSH in healthy adults may have minimal impact in the assessment and management of thyroid disorders. 21

Comparison with other published studies

Based on the fact that TSH values vary based on ethnicities, it is important to consider studies within a country. Interestingly, even such studies within a country, show variation in recommended reference intervals as found in a Danish study.4

Two studies in India could be accessed for review of reference intervals of thyroid hormones.8, 9 The findings in current study (Table 11) which includes population in Gujarat, Mumbai and Pune (the western region) match TSH reference interval upper limit only with the two studies. The reference interval for TT4 in the current study is different and higher than seen in Gujarat and Indore (Table 11). However, there is a similarity in the fact that reference interval for TT4 is higher in females than males as seen in this study and Gujarat study8 and hence worthwhile considering establishing separate reference intervals gender-wise for TT4 in western India. In the current study TT3 had narrower range vis-à-vis other studies.

Marwah S et al8 and Bose et al9 found that the value of serum Total T3 & serum Total T4 was higher as compared to the published international values while that of serum TSH was similar. These findings are in agreement with current study, when compared with manufacturer’s (international) reference interval for TT4 and TT3 (Table 4).