255 Diagnostic utility of molecular and imaging biomarkers 2 p<0.0001) of the remaining non-oncocytic nodules from these eight studies were PET negative [37, 59, 303, 305, 307-309, 312]. Eight of the 11 PET negative nodules were benign, 3 of 11 were diagnosed as TUMP lesions. Of PET positive nodules, 22 of 38 were malignant or TUMP [37, 59, 305, 307, 308, 312]. This corresponds to an estimated 88% sensitivity (95% CI using Clopper-Pearson: 69%-97%) and 17% specificity (95% CI: 8%-31%). Altogether, even though the series of Hürthle cell nodules in individual studies are relatively small, the overall conclusion is that due to a low benign call rate, [18F]FDG-PET is most likely not effective in differentiating between benign and malignant Hürthle cell nodules. Looking at the performance of [18F]FDG-PET in non-oncocytic indeterminate nodules excluding known Hürthle cell lesions, we observe that the benign call rate increases, the negative test performance of [18F]FDG-PET remains adequate, and the diagnostic accuracy slightly improves. Meta-analysis of the 363 non-oncocytic indeterminate nodules from 10 studies (including the two studies with an unreported number of Hürthle cell neoplasms [58, 306], and two Hürthle cell lesions from one study with unknown correlation to PET result and histopathology [59]) yields an estimated pooled sensitivity, specificity, positive and negative LR of 92.7% (95% CI: 82.8%-97.1%), 55.7% (95% CI: 35.4%-74.3%), 2.09 (95% CI: 1.35-3.25) and 0.13 (95% CI: 0.06-0.30), respectively. The AUC is 0.91 (95% CI: 0.88-0.93) (Table 64, Figure 87 and 88). For a given prevalence of malignancy of 15%, 25% or 40%, these results correspond to an estimated PPV and NPV of 27.0% (95% CI: 19.2%-36.5%) and 97.7% (95% CI: 95.0%-99.0%), 41.1% (95% CI: 31.0%-52.0%) and 95.8% (95% CI: 90.9%-98.1%), or 58.3% (95% CI: 47.3%-68.4%) and 91.9% (95% CI: 83.4%-96.3%), respectively (Figure 89: Bayesian probability plot).
RkJQdWJsaXNoZXIy MTk4NDMw