Thyroglobulin (Tg) Testing Revisited: Tg Assays, TgAb Assays, and Correlation of Results With Clinical Outcomes

Abstract

Context: Measurement of thyroglobulin (Tg) by mass spectrometry (Tg-MS) is emerging as a tool for accurate Tg quantification in patients with anti-Tg autoantibodies (TgAbs).

Objective: The objective of the study was to perform analytical and clinical evaluations of two Tg-MS assays in comparison with immunometric Tg assays (Tg-IAs) and Tg RIAs (Tg-RIAs) in a cohort of thyroid cancer patients.

Methods: A total of 589 samples from 495 patients, 243 TgAb-/252 TgAb+, were tested by Beckman, Roche, Siemens-Immulite, and Thermo-Brahms Tg and TgAb assays, two Tg-RIAs, and two Tg-MS assays.

Results: The frequency of TgAb+ was 58%, 41%, 27%, and 39% for Roche, Beckman, Siemens-Immulite, and Thermo-Brahms, respectively. In TgAb- samples, clinical sensitivities and specificities of 100% and 74%-100%, respectively, were observed across all assays. In TgAb+ samples, all Tg-IAs demonstrated assay-dependent Tg underestimation, ranging from 41% to 86%. In TgAb+ samples, the use of a common cutoff (0.5 ng/mL) for the Tg-MS, three Tg-IAs, and the USC-RIA improved the sensitivity for the Tg-MSs and Tg-RIAs when compared with the Tg-IAs. In up to 20% of TgAb+ cases, Tg-IAs failed to detect Tg that was detectable by Tg-MS. In Tg-RIAs false-high biases were observed in TgAb+ samples containing low Tg concentrations.

Conclusions: Tg-IAs remain the method of choice for Tg quantitation in TgAb- patients. In TgAb+ patients with undetectable Tg by immunometric assay, the Tg-MS will detect Tg in up to 20% additional cases. The Tg-RIA will detect Tg in approximately 35% cases, but a significant proportion of these will be clinical false-positive results. The undetectable Tg-MS seen in approximately 40% of TgAb+ cases in patients with disease need further evaluation.

Figure 1.

Tg-IA bias in TgAb-positive samples. Folded probability (Mountain) plots compare the biases of the Tg-IAs to Tg-MS-1 in TgAb− (black) and TgAb+ (blue) samples with Tg concentrations greater than 0.5 ng/mL by Tg-MS-1 and by the respective Tg-IAs. The number of samples analyzed in each group were as follows: 109 TgAb− and 106 TgAb+ for Roche (panel A); 110 TgAb− and 109 TgAb+ for Beckman (panel B); 108 TgAb− and 93 TgAb+ for Siemens-Immulite (panel C); and 110 TgAb− and 112 TgAb+ for Thermo-Brahms (panel D). The x-axis of each plot shows the percentage difference between Tg-MS-1 and the respective Tg-IA assay. The y-axis depicts the probability of the occurrence of the observed differences. The peak of the mountain is at 50% probability, ie, the median percentage difference to Tg-MS-1. The left slope corresponds to the distribution of less positive, or greater negative, differences, whereas the right slope shows those of less negative, or greater positive, differences, respectively. In each panel, the light gray line denotes a (theoretical) 0% bias to Tg-MS-1; if there was complete concordance between Tg-MS-1 and a Tg-IA, the mountain would be a single line superimposed on this gray line. The black and blue dashed lines correspond to the actual Tg-MS-1 vs Tg-IA median biases in TgAb− and TgAb+ samples, respectively. The difference in percentage bias between TgAb− and TgAb+ samples are indicated by the arrows between the two peaks, and their numerical values are listed.

Figure 2.

ROC curves and clinical sensitivity and specificity analysis for Tg-MS, Tg-IA, and Tg-RIA. TgAb− samples (A) and TgAb+ samples (B) are shown. There are no significant differences between the area under the curves (AUC) of the different assays. However, the smaller number of data points makes AUC assessments of the Tg-RIA ROC curves less reliable than those of the other assays. Sensitivity and specificity were calculated based on the Tg assays' individual FS and a common 0.5 ng/mL cutoff. *, The assay's FS were as follows: 0.1 ng/mL Beckman and Roche; 0.15 ng/mL Thermo-Brahms; 0.9 ng/mL Immulite; 0.5 ng/mL Tg-MS-1, Tg-MS-2, and USC-RIA; and 5.0 ng/mL UK-RIA. ^, Assays with an FS of 0.5 ng/mL or greater are omitted from this portion of the table because the change to a 0.5 ng/mL cutoff does not affect them.