Computing uveal melanoma basal diameters: a comparative analysis of several novel techniques with improved accuracy

We sought to compare the accuracy of standard and novel echographic methods for computing intraocular tumor largest basal diameter (LBD).

Multicenter, retrospective cohort study.

All patients presenting with new diagnosis of uveal melanoma (UM).

Ultrasounds were obtained for all patients, and axial length (AL) was measured for a subset of patients. LBD was calculated as: (1) a single chord measured on B scan ultrasound (one-chord method [1CM]), or (2) by subdividing the basal diameter into two chords, which were summated (two-chord method [2CM]), or (3) by a mathematically-derived formula (MF) based on geometric relationships. The accuracy of each method was then compared, and sensitivity of each technique to factors such as tumor size and AL were analyzed.

Accuracy, robustness, correctness of predicted plaque size.

116 UMs were analyzed; 1CM-calculated LBD underestimated 2CM-calculated LBD by 7.5% and underestimated LBD by MF by 7.8%; 2CM and MF were tightly correlated (average LBD difference = 0.038%). At larger LBDs, 1CM underestimated 2CM and MF by a much greater percentage (p < 0.001). By linear regression, 1CM underestimated LBD compared to 2CM by 0.8% and underestimated LBD compared to MF by 1.2% for every 1-mm LBD increase (p < 0.001 for each). Increasing the number of ultrasound chords beyond two did not significantly impact LBD calculations. For eyes with AL within two standard deviations of the mean, AL did not impact plaque selection using MF. 1CM would have led to selection of an undersized plaque in 41% of cases compared to 2CM and would have misclassified half of all eyes that actually required enucleation. For tumors with LBD < 12 mm, 1CM does not significantly underestimate LBD.

Tumor LBD by 1CM is an inaccurate means of determining actual LBD, especially for larger tumors. Using either 2CM or MF is much more accurate, especially for tumors > 12 mm, where a single chord on ultrasound is more likely to lead to incorrect, undersized plaque selection. Our MF can be applied with great accuracy even in cases where the AL of the eye is not measured, using the population average AL (23.7 mm), and the formula $$ .