Diagnostic Accuracy of Fine-Needle Biopsy in the Detection of Thyroid Malignancy: A Systematic Review and Meta-analysis.

Importance: Fine-needle biopsy (FNB) became a critical part of thyroid nodule evaluation in the 1970s. It is not clear how diagnostic accuracy of FNB has changed over time. Objective: To conduct a systematic review and meta-analysis estimating the accuracy of thyroid FNB for diagnosis of malignancy in adults with a newly diagnosed thyroid nodule and to characterize changes in accuracy over time. Data Sources: PubMed, SCOPUS, and Cochrane Central Register of Controlled Trials were searched from 1975 to 2020 using search terms related to FNB accuracy in the thyroid. Study Selection: English-language reports of cohort studies or randomized trials of adult patients undergoing thyroid FNB with sample size of 20 or greater and using a reference standard of surgical histopathology or clinical follow-up were included. Articles that examined only patients with known thyroid disease or focused on accuracy of novel adjuncts, such as molecular tests, were excluded. Two investigators screened each article and resolved conflicts by consensus. A total of 36 of 1023 studies met selection criteria. Data Extraction and Synthesis: The MOOSE guidelines were used for data abstraction and assessing data quality and validity. Two investigators abstracted data using a standard form. Studies were grouped into epochs by median data collection year (1975 to 1990, 1990 to 2000, 2000 to 2010, and 2010 to 2020). Data were pooled using a bivariate mixed-effects model. Main Outcomes and Measures: The primary outcome was accuracy of FNB for diagnosis of malignancy. Accuracy was hypothesized to increase in later time periods, a hypothesis formulated prior to data collection. Results: Of 16 597 included patients, 12 974 (79.2%) were female, and the mean (SD) age was 47.3 (12.9) years. The sensitivity of FNB was 85.6% (95% CI, 79.9-89.5), the specificity was 71.4% (95% CI, 61.1-79.8), the positive likelihood ratio was 3.0 (95% CI, 2.3-4.1), and the negative likelihood ratio was 0.2 (95% CI, 0.2-0.3). The area under the receiver operating characteristic curve was 86.1%. Epoch was not significantly associated with accuracy. None of the available covariates could explain observed heterogeneity. Conclusions and Relevance: Accuracy of thyroid FNB has not significantly changed over time. Important developments in technique, preparation, and interpretation may have occurred too heterogeneously to capture a consistent uptrend over time. FNB remains a reliable test for thyroid cancer diagnosis.