Table 5.
Comparison of the Theoretical NPV Performance Between mir-THYpe and the Other Commercially Available Molecular Classifier Tests
| Cancer prevalence in FNA validation set cohort | 38.9% [mir-THYpe] | 52.6% [ThyroSeq v3 (19)] | 32.8% [ThyroidPrint (20)] | 32.1% [ThyraMIR/ ThyGenX (21)] | 32.3% [Rosetta GX Reveala (22)] | 23.7% [Afirma GSC (18)] | |
|---|---|---|---|---|---|---|---|
| Theoretical NPV performance (Bayes' theorem) | mir-THYpe | 95.9% | 93.1% | 96.8% | 96.9% | 96.9% | 98.0% |
| ThyroSeq v3 | 98.5% | 97.4%b | 98.8% | 98.9% | 98.8% | 99.2% | |
| ThyroidPrint | 96.8% | 94.5% | 97.5% | 97.6% | 97.6% | 98.4% | |
| ThyraMIR / ThyGenX | 92.1% | 87.0% | 93.8% | 94.0% | 94.0% | 96.0% | |
| Rosetta GX Reveala | 88.4% | 81.4% | 90.9% | 91.1% | 91.1% | 94.0% | |
| Afirma GSC | 92.3% | 87.4% | 94.0% | 94.2% | 94.1% | 96.1% |
The theoretical NPV was calculated based on Bayes' theorem using the sensitivity, specificity, and cancer prevalence in the FNA validation set cohort of each study. Values highlighted in bold correspond to the observed NPV values on the specific cancer prevalence from the respective study.
a
Considering the entire validation set (n = 189).
b
Calculated based on Bayes' theorem, using the cancer prevalence, sensitivity, and specificity published in Nikiforova et al. (19).