Selective use of Molecular Testing Based on Sonographic Features of Cytologically Indeterminate Thyroid Nodules: A Decision Analysis.

BACKGROUND: Molecular diagnostics can allow some patients with indeterminate thyroid nodule cytopathology to avoid diagnostic hemithyroidectomy; however, the testing is costly. We hypothesized that molecular testing with the intention of preventing unnecessary diagnostic hemithyroidectomy would be cost-effective if this test was applied selectively based on sonographic risk of malignancy. METHODS: A Markov model was constructed depicting a 40-year-old patient with a cytologically indeterminate thyroid nodule. Molecular testing of fine needle aspiration material was compared to a strategy of immediate diagnostic hemithyroidectomy. Data from a single tertiary-referral health system were reviewed to estimate the outcomes of molecular testing of indeterminate nodules stratified by the American Thyroid Association sonographic classification system. Other outcome probabilities and their utilities were derived from literature review. Costs were estimated with Medicare reimbursement data. A $100,000/QALY threshold for cost-effectiveness was applied. Sensitivity analysis was employed to examine uncertainty in the model's assumptions. RESULTS: Of 123 patients who underwent molecular testing for indeterminate cytology, 12 (9.8%) were classified as high sonographic suspicion, 49 (40%) were intermediate suspicion, and 62 (50%) were low or very low suspicion. Molecular testing was only cost-effective when the pretest probability of a negative test was greater than 31%. The model was most sensitive to the cost of molecular testing and the quality adjustment factor for hypothyroidism. CONCLUSIONS: In hypothetical modeling, molecular testing is only cost-effective for cytologically indeterminate thyroid nodules with sonographic features that are intermediate or low suspicion for malignancy. In nodules with high sonographic suspicion, molecular testing is rarely negative and appears to add minimal value.