Thyroglobulin and thyroglobulin antibodies: assay-dependent management consequences in patients with differentiated thyroid carcinoma

Introduction

Measurement of serum thyroglobulin (Tg) has an essential role in follow-up monitoring of patients with differentiated thyroid cancer (DTC). Tg measurements in combination with a neck ultrasound have a high sensitivity for detecting persistent or recurrent disease in patients previously treated with thyroidectomy and radioactive iodine (RAI) ablation [1].

Currently, most laboratories have adopted an automated immunometric assay for measurement of Tg. In recent years the diagnostic sensitivity has been improved with the introduction of assays with functional sensitivities (FS) ≤0.1 μg/L, also referred to as high-sensitive Tg assays (hs-Tg).

The measurement of Tg is complicated by multiple factors. First, a large between-method variation regarding sensitivity, accuracy and precision prevents comparison of (consecutive) Tg measurements with different assays [2], [3], [4]. Second, Tg assays suffer from interfering endogenous Tg antibodies (TgAb), which are detected in up to 20% of patients with DTC [2]. As for Tg, immunometric assays for TgAb also show large between-method variation, which impedes the correct identification of patients with interfering TgAb [5], [6], [7].

International guidelines state fixed cut-offs for Tg; <0.2 μg/L and <1.0 μg/L for thyrotropin (TSH) suppressed measurements (Tg-on) and <1.0 μg/L and <10.0 μg/L for recombinant human TSH (rh-TSH) stimulated Tg measurements [8, 9]. Prognostic predictions and recommendations on the frequency of follow-up, additional investigations and continuation of TSH suppressive therapy are based on these cut-offs which are derived from studies using a variety of Tg and TgAb assays [10], [11], [12], [13], [14].

This study aimed to evaluate to what extent different methods for Tg and TgAb measurement affect management guidance for individual DTC patients. Therefore, in a large cohort of DTC patients, Tg and TgAb were measured simultaneously with Immulite 2000XPi and Kryptor compact plus, both immunometric assays, and related to clinical disease status.

Materials and methods

Results

Comparison of assays

A total number of 793 samples, derived from 413 DTC patients, were analyzed for Tg and TgAb. The majority of patients (80.6%) had papillary thyroid cancer (PTC). In 184 (44.6%) patients, a single Tg/TgAb measurement was performed, in 137 (33.2%) patients two measurements were performed and in 92 patients (22.3%) three or more measurements. The cohort comprised 702 (88.5%) Tg-on measurements, 71 (9.0%) rh-TSH-Tg measurements and 20 (2.5%) Tg-off measurements (Table 1).

Table 1:

Patient and measurement characteristics.

Patients (n=413)	n (%)
Sex
Female	287 (69.5%)
Male	126 (30.5%)
Age, years
At first measurement	52.4 ± 15.5 (18.5–88.8)
Histology
PTC	333 (80.6%)
FTC	59 (14.3%)
HCC	7 (1.7%)
FVPTC	9 (2.2%)
Other	5 (1.2%)
Metastatic disease	76 (18.4%)
Only localized in the neck	31 (7.5%)
Distant metastases	45 (10.9%)
Number of measurements per individual patient
1	184 (44.6%)
2	137 (33.2%)
≥3 (maximum nine)	92 (22.3%)
Measurements (n=793)
Type of Tg measurement
Thyrotropin suppressed (Tg-on)	702 (88.5%)
Recombinant human TSH stimulated (rh-TSH-Tg)	71 (9.0%)
After withdrawal of levothyroxin (Tg-off)	20 (2.5%)

1. Tg, thyroglobulin; TgAb, Tg antibodies; PTC, papillary thyroid cancer; FTC, follicular thyroid cancer; HCC, hurthle cell carcinoma; FVPTC, follicular variant of papillary thyroid carcinoma.

Tg was measurable in 270 (34.0%) samples with Immulite and in 595 (75.0%) with Kryptor. Mean Tg concentrations were 37.4% lower with Kryptor (mean 39.5 μg/L) compared with Immulite (mean 63.1 μg/L). The between-method correlation coefficient was 0.985 (Figure 1A).

Figure 1:

Bland–Altman plot of agreement between Immulite and Kryptor for Tg (A) en TgAb (B). Outliers (n=3) were excluded.

In 80 (20.2%) and 102 (24.3%) samples, TgAb were tested as positive with Immulite and Kryptor, respectively. These samples were derived from 34 (8.2%) and 56 (13.5%) individual patients, respectively. The between-method correlation coefficient was 0.6155 (Figure 1B).

Median Tg concentrations were significantly higher in TgAb− measurements compared with TgAb+ measurements (Table 2).

Table 2:

Tg (µg/L) in TgAb positive and TgAb negative patients.

	Immulite				Mann–Whitney U (p-value)	Kryptor				Mann–Whitney U (p-value)
	TgAb positive (>9 U/mL)		TgAb negative (<9 U/mL)			TgAb positive (>33 U/mL)		TgAb negative (<33 U/mL)
	n	Median (IQR)	n	Median (IQR)		n	Median (IQR)	n	Median (IQR)
All measurements
Tg Immulite	80	<0.9 (<0.9–<0.9)	713	<0.9 (<0.9–3.65)	0.004	102	<0.9 (<0.9–<0.9)	691	<0.9 (<0.9–3.60)	0.042
Tg Kryptor	80	<0.1 (<0.1–0.47)	713	0.2 (<0.1–2.16)	<0.0001	102	0.1 (<0.1–0.74)	691	0.2 (0.10–2.00)	0.0004
Biochemical diseasea
Tg Immulite	7	<0.9 (<0.9–<0.9)	74	4.6 (1.78–24.88)	0.0002	8	<0.9 (<0.9–0.75)	73	4.6 (1.80–28.25)	<0.0001
Tg Kryptor	7	0.1 (<0.1–0.16)	74	2.8 (1.10–13.63)	0.0002	8	0.1 (<0.1–0.71)	73	2.8 (1.10–14.35)	0.0002
Structural diseasea
Tg Immulite	23	10.0 (<0.9–218.0)	190	22.6 (1.67–90.13)	0.374	30	94.6 (<0.9–605.50)	183	21.1 (1.60–83.10)	0.419
Tg Kryptor	23	6.0 (0.10–130.2)	190	12.2 (0.92–54.00)	0.531	30	63.3 (0.10–485.65)	183	10.8 (0.90–52.10)	0.448
Normal recovery
Tg Immulite	45	<0.9 (<0.9–<0.9)	568	<0.9 (<0.9–2.68)	0.067	71	<0.9 (<0.9–<0.9)	542	<0.9 (<0.9–2.73)	0.173
Tg Kryptor	45	<0.1 (<0.1–0.71)	568	0.2 (0.10–1.40)	0.002	71	0.1 (<0.10–0.80)	542	0.2 (0.10–1.40)	0.031

1. ^aDisease status based on Tg measurements with Immulite. Tg, thyroglobulin; TgAb, Tg antibodies. p-Values < 0.05 are given in bold.

Impact on clinical practice

Applying guideline based cut-off values, 33 (4.7%) samples in 25 (6%) patients had a Tg-on concentration ≥1.0 μg/L with Immulite and <1.0 μg/L with Kryptor (Figure 2). In rh-TSH-Tg measurements, 3 (4.3%) measurements in three individual patients, showed discrepancy with regard to the cut-off of 1.0 μg/L: Tg measured with Immulite was above the cut-off and with Kryptor below the cut-off of 1.0 μg/L. No discrepancy between the assays was observed for the cut-off of 10 μg/L. Of the samples determined as TgAb+ with at least one assay (n=125), 68 (54.4%) samples, derived from 47 (11.4%) individual patients, showed discrepancy in TgAb positivity between either assay.

Figure 2:

Icon array showing the percentage of patients with discordant Tg results when applying a cut-off 1.0 µg/L in Tg-on measurements.

Using either assay, the response to therapy classification was influenced in 94 (12.0%) measurements, derived from 67 (16.2%) individual patients. The Kryptor assay resulted in a larger number of measurements fitting the indeterminate response category, 177 (22.3%) with Kryptor vs. 134 (16.9%) with Immulite. The number of measurements categorized as excellent response decreased from 302 (38.1%) with Immulite to 276 (34.8%) with Kryptor (Figure 3). In 159 (33.8%) measurements with an Immulite Tg <1.0 μg/L the Tg concentration measured with Kryptor was between 0.2 and 1.0 μg/L. In 47 (30%) of these measurements, derived from 39 (9.4%) individual patients, a previous Immulite rh-TSH-Tg had qualified these patients as having an excellent response.

Figure 3:

Discordance in response to therapy classification.

For each Tg/TgAb measurement the response to therapy at time of measurements was assessed according to the classification in the 2015 ATA Management Guidelines for adult patients with thyroid nodules and differentiated thyroid cancer. For the purpose of analysis some adjustments were made as described in the method section. Discordance was observed in 94 measurements (12.0%) derived from 67 (16.5%) individual patients.

Discussion

This study evaluated the impact of guideline recommended Tg cut-offs in clinical management of DTC in a cohort of 413 DTC patients. The data show that, when applying these fixed cut-offs, a substantial proportion of DTC patients are classified differently dependent on which Tg assay is used. This implies that clinical management of these patients can differ dependent on which assay is used.

International guidelines advocate the use of both Tg-on and rh-TSH-Tg measurements for determining the response to therapy throughout follow-up in patients who have initially been treated with thyroidectomy and RAI ablation. Patients are stratified as having an excellent, indeterminate or incomplete response based on fixed Tg cut-offs [8, 9]. The guideline cut-offs for Tg are based on studies measuring Tg with a variety of methods, mostly with a FS of 1.0 μg/L [10], [11], [12], [13], [14]. Although between-method differences for Tg measurement have been clearly recognized, these differences were not addressed in the studies providing the guideline cut-offs [2]. Discrepancies between Tg assays relate, amongst others, to specificity differences in the antibody reagents of the assay, heterogeneity in tumor derived Tg and interference of TgAb [2, 16, 17]. Even after standardization against the same reference standard a two-fold between-method difference in Tg concentrations has been reported [2, 3]. Despite these differences, international guidelines propose fixed cut-offs [8, 9].

When applying the most clinically relevant Tg cut-off of 1.0 μg/L, discrepancy in staging was observed between the Immulite and Kryptor assays in 6.0% and 4.3% of patients for Tg-on and rh-TSH-Tg, respectively. Disagreement on response to therapy was observed in 16.2% of patients. The observed discordance on response to therapy classification is caused by between-method slope bias, differences in FS and disagreement on TgAb positivity. The analytical bias, which can be positive or negative, may add to decreased concordance between-methods as demonstrated in a study of Ross et al. [18] The analytical bias for the studied assays was assessed, and was 35%, with Immulite generating higher results. The negative analytical bias for the Kryptor assay was in line with previously published data from our group [19]. Due to the lower FS of Kryptor (0.15 μg/L compared to 0.9 μg/L for Immulite) more patients will be classified as having an indeterminate response with Kryptor, at the cost of the number of patients with an excellent response. In 39 of these patients an Immulite rh-TSH-Tg was performed before the study period, but a Kryptor rh-TSH-Tg measurement was not available. Based on the institutionally established performance of Kryptor Tg-on as a predictor of Kryptor rh-TSH-Tg (positive predictive value of 68%, see Supplementary Figure S2), one would expect that 32% of these patients will have a Kryptor rh-TSH-Tg below the cut-off of 1.0 μg/L. This implicates that, if a Kryptor rh-TSH-Tg would have been performed, an estimated number of 12 of these 39 patients would have remained in the excellent response category instead of being reclassified from excellent to indeterminate response. Therefore, even after correction for the missing Kryptor rh-TSH-Tg measurements, the use of the Kryptor assay will result in a larger proportion of patients with an indeterminate response to therapy classification. These patients will be exposed to more intensive follow-up, longer TSH suppressive therapy and additional therapies.

The clinical significance of detectable Tg levels in the low range (0.2–1.0 μg/L) is not clear. The origin might be thyroid remnant, (irradiated) tumor tissue or reflect microscopic tumor foci that will eventually result in clinically apparent disease recurrence. To minimize Tg produced by thyroid remnant, measurements taken in the first six months after the first RAI ablation were excluded from the analysis. Assays with lower FS have been reported to have higher sensitivity for predicting recurrent disease compared to assays with higher FS although it is unclear if long term outcomes are influenced [3, 20, 21]. However, the increased sensitivity of these assays is at the cost of specificity. Multiple studies have shown that measurable Tg levels up to 1.0 μg/L spontaneously convert into undetectable Tg in the majority of patients without further treatment, including in patients with high risk for recurrence [22], [23], [24], [25], [26]. Castagna et al. showed that only rh-TSH-Tg measurements could distinguish patients with persistent or biochemical disease from patients free of disease but others have argued that hs-Tg assays can replace rh-TSH-Tg measurements [27, 28]. In our cohort 3 out of 70 rhTSH-Tg measurements showed discrepancy between the assays when the cut-off of 1.0 μg/L was applied. The comparability between Tg measurements is also compromised by interfering TgAb. Immunoassays are not able to detect TgAb-bound Tg which may result in underestimation of Tg concentrations [29], [30], [31]. In our cohort, Tg concentrations were significantly lower in TgAb+ samples compared to TgAb− samples. As the presence of TgAb is associated with a higher risk of recurrence, guidelines recommend to measure TgAb simultaneously with Tg and monitor TgAb+ patients more frequently.

In this study, disagreement on TgAb positivity between the assays was observed in 11.4% of patients. In these patients Tg would be judged as a reliable marker with one assay, but not with the other, which has an obvious impact on the follow-up strategy. TgAb+ patients have often been excluded in the studies providing the cut-offs for Tg concentration included in guidelines [10], [11], [12, 14]. As determination of TgAb positivity is TgAb assay-dependent, patients with present but undetectable TgAb with the specific assays used, have possibly been included in these analyses, potentially influencing the established cut-off for Tg [32].

Between-method discordance for TgAb measurement is caused by heterogeneity of endogenous TgAb, differences in sensitivity of the TgAb method and the (arbitrary) selected cut-off for TgAb positivity [7, 30, 33]. If liquid chromatography-tandem mass spectrometry (LC-MS/MS), which is considered not to be influenced by interfering TgAb, is used as a reference, four tested TgAb immunoassays (including Immulite) missed 16–54% of interfering TgAb [31]. In addition, the concentration of TgAb is not linearly associated with interference and undetectable TgAb does not rule out interference [30, 31, 34]. The clinical impact of interfering antibodies remains unclear. A recent study showed that in patients with evidence of structural disease, TgAb interference rarely resulted in undetectable Tg levels [35]. However, low volume recurrent disease can potentially be missed. As Tg measurement with LC-MS/MS does not suffer from TgAb interference, this method is a promising candidate for reliable Tg measurements. Currently, the inferior FS compared to immunoassays and time consuming procedures hinder implementation of LC-MS/MS in clinical care [28].

TgAb interference can also be qualitatively investigated with a recovery method. In our study a recovery test was performed on the Kryptor platform. Median recovery values were significantly lower in TgAb+ samples. However, also in measurements with a normal recovery value according to the manufacturers’ cut-off, median Tg concentration were lower in TgAb+ measurements. This suggests that a normal recovery does not rule out TgAb interference. When applying the cut-off for disturbed recovery (<80%), only 2.6% of samples were identified as such. In 5.5% of measurements recovery was >120%, most likely caused by interfering heterophilic antibodies (HAbs), which have a reported prevalence of 1.5–3% [36].

Recovery values were investigated in the subcategory of patients with structural disease and undetectable Tg with one or both assays (assuming false negative results due to TgAb interference in at least a proportion of patients). Only one out of 15 samples showed a recovery <80% and a very high concentration of TgAb (>3,000 U/mL) was detected in that specific sample. Although Immulite failed to detect Tg in all these patients, Kryptor Tg was detectable in 12 out of 15 patients. This illustrates the increased diagnostic accuracy of a high-sensitive assay and that a recovery test does not aid in the discrimination of patients with unreliable Tg concentrations due to TgAb interference. Our study confirmed previous studies reporting a limited value of recovery measurement in determining TgAb interference [17, 35, 37] and is in agreement with current guidelines advocating against the use of the recovery method [4, 8, 38].

The results of this study show that comparability between Tg and TgAb assays is affected by the use of fixed cut-offs, which may be ameliorated by the application of assay-specific cut-offs. Trimboli et al. showed that assay-specific cut-offs increases comparability in clinical performance [39]. According to a study of Giovanella et al. these cut-offs should be further tailored based on TgAb status. Lower Tg cut-offs should be applied in TgAb+ patients compared to TgAb− patients [40].

Our study has its strengths and limitations. The strength of this study is the simultaneous measurement of Tg and TgAb with two distinct assays in a large cohort of DTC patients. As all samples were measured directly after the blood was drawn from patients, bias due to freeze-thaw cycles or storage was prevented. Assays were not only compared mutually, but the impact on response to therapy classification was also investigated, contributing to assessing the clinical relevance of the between-method variations observed. We would like to mention a number of limitations. First, although the percentage of patients with positive TgAb and with disease recurrence was as expected and in line with the literature, the total numbers of these patients where relatively small. Second, different methods were used for establishing the cut-off for TgAb positivity for the included TgAb assays. Originally, TgAb were used as a marker for thyroid auto-immunity and the reference values were aimed at distinguishing patients with and without thyroid auto-immunity with adequate specificity [41]. At the time of introduction of the Immulite TgAb assay into our hospital, only the upper reference limit based on detecting auto-immunity was provided by the manufacturer. With the increasing use of Tg as a follow-up marker in DTC patients, it became clear that any level of TgAb can potentially interfere with Tg measurement. Therefore, an institutionally determined upper reference limit for the Immulite TgAb assay was used in the follow-up of DTC patients in our hospital. This cut-off was based on the level at which clinically relevant interference of Tg measurement was observed. For the Kryptor TgAb assay the functional sensitivity of the assay was used as the cut-off, in line with recommendations in current guidelines [4]. Evolving insights into the appropriate cut-off for TgAb positivity limits a full comparison of the studied assays. However, our study represents real-life observational data which may be relevant for daily clinical practice. Third, we only compared two assays, limiting the applicability to other Tg/TgAb assays.

Our study indicates that classification and management of a substantial subset of DTC patients is affected, dependent on which Tg and TgAb assays are used, should fixed Tg cut-offs be applied. Therefore, the recommended Tg cut-offs in guidelines should be used with wisdom rather than as fixed cut-offs.