Abstract
With the reclassification of non-invasive follicular variant of papillary thyroid carcinoma as "tumors" (NIFTP), the challenge of diagnosing carcinoma lies in recognizing invasion: blood vessel invasion (BVI) or capsular invasion (CI). We aimed to assess reproducibility in diagnosing BVI and CI in follicular-patterned tumors using a web-platform, under conditions as close as possible to routine surgical pathology practice. Six-hundred-fourteen pathologists reviewed 69 thyroid lesions diagnostically challenging for BVI or CI, made available in three rounds: before, immediately after, and three months after a training lecture given by an expert pathologist. Each lesion was associated with a quiz (five queries: certain CI, doubtful CI, certain BVI, doubtful BVI, or no invasion). Intraobserver agreement was fair for five queries and moderate for two categories (i.e., BVI vs. the other four queries). Interobserver overall agreement (OA) of participants was fair for all five queries and moderate for two categories (OA for five queries: 0 at each round; OA [95% CI] for two categories: Round 1 to 3: 8.7 [3.3-18.0], 2.9 [0.4-10.1], 4.3 [0.9-12.2]). Importantly, participants agreement with expert consensus improved following the training lecture and was maintained for three months (OA for five queries: Round 1 to 3: 40.6 [13.0-62.3], 49.3 [21.7-69.6], 49.3 [21.7-78.3]; OA for two categories: Round 1 to 3: 72.5 [60.9-89.9], 76.8 [62.3-88.4], 75.4 [60.9-89.9]). This study reveals the difficulty of applying clearly defined BVI and CI criteria to routine practice, but a specifically designed training lecture improves diagnostic accuracy, and the improvement is retained over months.