​An endocrinologist contacted our laboratory to report a patient with discordant TSH results: undetectable TSH by our routine assay, the Siemens ADVIA Centaur TSH3-UL, and normal TSH by an alternate TSH assay, the Abbott Architect. The clinical picture of the patient matched the normal results. We ruled out common interferences and ran the sample on a total of eight FDA-approved TSH immunoassays. Four of these assays, all from Siemens, were unable to detect the TSH in the sample while four others, from Siemens, Beckman, Abbott and Roche, all detected TSH at similar concentrations.

Meanwhile, other endocrinologists identified three additional patients who fit this pattern of falsely undetectable TSH. The undetectable results went back over several specimens for two years. Before the two year mark, when our lab switched to the Siemens TSH3-UL assay, the patients had detectable TSH on the Siemens ADVIA Centaur TSH3. Retrospective review determined that three of these first four patients were euthyroid and inappropriately treated with anti-thyroid medication based on the undetectable TSH results (the fourth had a history of hypothyroidism and the synthetic T4 dose was decreased).

While the TSH3-UL assay was still accurate for the vast majority of patients, we realized that a small population was at risk. We therefore set out to identify all other “discordant patients” in our population. Prospectively, we began reflexing all TSH3-UL results <0.03 IU/mL to the Abbott Architect TSH assay to prevent any additional falsely undetectable TSH results from being reported. Retrospectively, we used electronic medical record database mining to analyze thyroid results, diagnoses and medications from 1.6 million individuals to identify patients who might be receiving inappropriate treatment. Ultimately, we have identified 20 discordant patients out of approximately two million patients tested with TSH3-UL over a three-year period, giving an approximate prevalence of 0.001% in our population. All discordant patient samples were tested with additional TSH assays and exhibited the same pattern as the index patient: undetectable results with four of the assays and detectable TSH at similar concentrations in the other four assays. Nine of the 20 patients had been treated; all treatment was halted after we communicated the situation to the treating physician.

All but one of the discordant patients identified as South Asian (usually Indian). This led us to suspect early on that the falsely undetectable results were due to a biologically active TSH variant that is unrecognizable by some of the TSH assays’ monoclonal antibodies. We partnered with scientists from Siemens in our investigation and they performed experiments that supported our hypothesis and identified the antibody in each assay that fails to detect the presumed variant TSH. (A joint paper is currently in press.) Our hypothesis is further supported by DNA sequencing data: we sequenced the gene for TSH in 12 of the discordant patients and found that they were all homozygous for a point mutation that changes a single amino acid in the TSH beta subunit. Four control patients from various ethnicities lacked this mutation.

Low TSH results due to functionally compromising TSH mutations have been reported. In contrast, we have identified a novel phenomenon of biologically active TSH that is undetectable by four commonly utilized FDA-approved TSH immunoassays. We suggest that these assays could be called “hyper-selective” since their monoclonal antibodies seem to be able to detect only the most common functional TSH variant(s). Clinicians and laboratorians should be aware that certain TSH variants may be undetectable in hyper-selective TSH assays, especially since mistaken diagnosis of hyperthyroidism and treatment with anti-thyroid medication can induce iatrogenic hypothyroidism with attendant co-morbidities.