Parathyroid weight estimation: beyond ellipsoid volume.

Purpose: Parathyroid weight is a simple, robust functional status indicator and cornerstone of gross/intraoperative assessment, though sometimes unobtainable. We evaluate models to estimate weight from size and create an online calculator.

Methods: Actual weights (AW), cellularity, prosector identity and size were prospectively collected for 124 parathyroids (111 hypercellular, 13 normocellular) in 76 patients (4-6/2023). Simple volumetric weight estimates (VWE) for: ellipsoid, capsular, the novel capsuloid, and box shapes were compared with AW. Multiple linear regression (MLR) was performed with internal (k-fold) validation and external validation on an archival cohort (263 adenomas, 2016-2022). Subsets with surgeon's weight estimates, serum values, and microscopic surface area measured using the QuPath surface area tool were correlated with AW.

Results: The optimal MLR model included capsuloid VWE, cellularity, and prosector identity (R²:0.92, p < 0.0001). A more generalizable model without including prosector identity (R²: 0.89, p < 0.0001) was used for the online calculator weight estimate (CWE). The calculator was a good predictor of AW on the adenoma dataset (R²: 0.86, normalized root mean squared error (nRMSE): 0.058). Interestingly, the surgeon's weight estimate (n = 31) was more favorable (R²: 0.97, nRMSE: 0.087) in this subset. QuPath assisted surface area-based weight estimates showed weaker correlation with AW. Neither AW, CWE, nor size correlated strongly with serum values.

Conclusion: An online calculator modeling capsuloid VWE and cellularity is a good predictor of AW. Variance in prosector measurements is important but impractical to model. Surgeon's estimates were quite accurate, emphasizing the value of this skill.