CT-based pleural effusion volume estimation formula demonstrates low accuracy and reproducibility for traumatic...

We aimed to evaluate the accuracy and reproducibility of the CT-based volume estimation formula V = d² * h, where d and h represent the maximum depth and height of the effusion, for acute traumatic hemothorax.

Prospectively identified patients with CT showing acute traumatic hemothorax were considered. Volumes were retrospectively estimated using d² * h, then manually measured on axial images. Subgroup analysis was performed on borderline-sized hemothorax (200-400 mL). Measurements were repeated by three non-radiologists. Bland-Altman analysis was used to assess agreement between the two methods and agreement between raters for each method.

A total of 46 patients (median age 34; 36 men) with hemothorax volume 23-1622 mL (median 191 mL, IQR 99-324 mL) were evaluated. Limits of agreement between estimates and measured volumes were -718 - +842 mL (± 202 mL). Borderline-sized hemothorax (n = 13) limits of agreement were -300 - +121 mL (± 114 mL). Of all hemothorax, 85 % (n = 39/46) were correctly stratified as over or under 300 mL, and of borderline-sized hemothorax, 54 % (n = 7/13). Inter-rater limits of agreement were -251 - +350, -694 - +1019, and -696 - +957 for the estimation formula, respectively, and -124 - +190, -97 - +111, and -96 - +46 for the measured volume.

An estimation formula varies with actual hemothorax volume by hundreds of mL. There is low accuracy in stratifying hemothorax volumes close to 300 mL. Variability between raters was substantially higher with the estimation formula than with manual measurements.