99 Diagnostic utility of molecular and imaging biomarkers 2 Molecular biomarkers BRAF point mutation Our systematic literature search identified 52 studies that investigated any BRAF point mutation in a total of 6,382 indeterminate thyroid nodules (Table 3). Thirty-one studies investigated solely BRAF mutations [68, 72-74, 79-82, 84-86, 89-92, 94, 96, 101, 103-106, 110-113, 116, 119-121, 354]; the other 21 combined these with other mutations or included this mutation in a multi-gene panel [60, 67, 69, 75-77, 83, 87, 88, 93, 95, 97-100, 102, 107, 109, 114, 118, 351]. Twenty-six studies had a prospective design [60, 67, 69, 72, 73, 77, 80-82, 85, 87-89, 91, 94, 95, 97, 98, 100, 102, 105, 110, 111, 113, 118, 351]. Twelve studies were conducted in the USA [60, 68, 69, 74, 75, 81, 88, 89, 95, 97, 114, 118]; 14 in Italy [67, 73, 76, 80, 82, 83, 86, 87, 93, 94, 96, 105, 106, 110]; 13 in South Korea [72, 77, 79, 84, 85, 92, 104, 111, 116, 119-121, 354]; 13 in other, mostly Western countries [90, 91, 98-103, 107, 109, 112, 113, 351]. Thirty-nine studies investigated only BRAFV600E [67-69, 72-74, 79, 80, 82-86, 88-92, 94-98, 101-105, 107, 111-114, 116, 119-121, 351, 354], 11 studies both BRAFV600E and the adjacent BRAFK601E [60, 75-77, 87, 93, 99, 106, 109, 110, 118]; two did not specify which variant was investigated [81, 100]. Meta-analysis was performed using summarized patient data from 51 of the 52 included studies, including a total of 6,156 indeterminate nodules. One study (n=226) had to be excluded from meta-analysis because their outcome was defined as PTC versus non-PTC, including both non-PTC malignancies and benign lesions in the latter group [354]. Prevalence of the BRAF mutation was 8.0% (494/6,156): 476 BRAFV600E (7.7%), 16 BRAFK601E (0.3%) mutations and two not specified. Cases with a nondiagnostic test result (n=85) were excluded from the current meta-analysis. We focused on the 64.0% (3,886/6,071) indeterminate thyroid nodules with a conclusive test result and available histopathological follow-up: 86.0% (425/494) of the BRAF mutation-positive nodules (407 BRAFV600E, 16 BRAFK601E, two unknown BRAF type) had histopathological follow-up, opposite 62.1% (3,461/5,577) of the mutation-negative nodules (Pearson χ2, p<0.0001). In the individual studies, the rate of available histopathology ranged from 15% to 100% [77, 88]. Visual inspection of the forest plots suggests presence of between-study heterogeneity: I2 is 91.9% for sensitivity and 92.7% for specificity. Estimated pooled sensitivity, specificity, positive and negative LR are 19.7% (95% CI: 13.3%-28.0%), 99.9% (95% CI: 99.3%-100%), 193.72 (95% CI: 27.09-1385.38) and 0.80 (95% CI: 0.73-0.88), respectively (Table 4, Figure 5). The AUC is 0.92 (95% CI: 0.90-0.94) (Figure 6). For a given prevalence of malignancy of 15%, 25% or 40%, these results correspond to an estimated PPV and NPV of 97.1% (95% CI: 83.5%-99.6%) and 87.6% (95% CI: 86.6%-88.5%), 98.5% (95% CI: 90.6%-99.8%) and 78.9% (95% CI: 77.3%-80.3%), or 99.2% (95% CI: 95.0%-99.9%) and 65.1% (95% CI: 63.0%-67.1%), respectively (Figure 7).
RkJQdWJsaXNoZXIy MTk4NDMw