Abstract
Kim et al. recently evaluated the diagnostic performance of 99mTc-methoxy-isobutyl-isonitrile (MIBI) scintigraphy in differentiating malignant from benign thyroid nodules by a systematic review and meta-analysis (1). Basing on pooled results of 22 studies, the authors question the utility of MIBI due to moderate sensitivity and specificity. We appreciate the statistical efforts of the authors but respectfully disagree with their conclusions. In particular: The authors included manuscripts published from 1996 to 2017. One should note that the pooled results were obtained with different instruments, administered activities, acquisition and elaboration protocols, and displaying techniques. Additionally, enrollment was retrospective in most studies, and both sample size and prevalence of thyroid cancer varied widely between the studies. i) Clinical indications for performing MIBI scans ranged from initial work-up of thyroid nodules (risk of malignancy ∼2–5%) to differential diagnosis of cytologically indeterminate nodules (risk of malignancy ∼30%). In the latter case, that is, the main current indication for MIBI imaging, visual analysis of MIBI scan is suboptimal at best, and semi-quantitative analysis is required to discriminate malignant from benign nodules properly (Table 1) (2).
Diagnostic Performance of MIBI Scan in Thyroid Nodules with Indeterminate Cytology Tests
Not included in the meta-analysis.
PPV, positive predictive value; NPV, negative predictive value; VA, visual analysis; SQA, semi-quantitative analysis; RI, retention index; WOI, washout index.
Statistical heterogeneity is based on the events in the different series and does not consider the nature of the studies. Presumably because of the factors listed above, the authors found high heterogeneity (i.e., the variability rate unrelated to the case) between studies included in their meta-analysis. In such a case, different approaches should be considered: (i) to display the results without pooled findings; (ii) to split the series and perform a meta-analysis of more homogeneous data, or (iii) to perform a meta-regression to evaluate the effect of continuous variables. The latter approach was chosen by Kim et al., and no definite source of heterogeneity was identified. Even if formally correct, however, this approach remains at risk for giving potentially inaccurate estimates.
Notably, positive predictive values (PPV) and negative predictive values (NPV) are much more relevant than sensitivity and specificity to detect (rule-in) or exclude (rule-out) malignant nodules reliably. Kim et al., however, were unable to report PPV/NPVs, as the prevalence of thyroid cancers did not reflect the real-world diagnostic setting due to the design of many studies. Prospective studies in homogeneously selected patients with cytologically indeterminate nodules proved that using a standardized MIBI scan protocol coupled with a semi-quantitative analysis of the tracer kinetic within the nodule, NPVs of 96–100% can be obtained, reaching a comparable and even better performance than molecular cytology (2,3).
In conclusion, in the context of proper patient selection, standardized technical protocols, and appropriate interpretation criteria, the role of MIBI-scan is likely more relevant than what emerges from the interesting paper by Kim et al.