Analysis of factors associated with discrepancies between predicted and observed liver weight in liver transplantation

Background Even using predictive formulas based on anthropometrics in about 30% of subjects, liver weight (LW) cannot be predicted with a <= 20% margin of error. We aimed to identify factors associated with discrepancies between predicted and observed LW. Methods In 500 consecutive liver grafts, we tested LW predictive performance using 17 formulas based on anthropometric characteristics. Hashimoto's formula (961.3 x BSA_D-404.8) was associated with the lowest mean absolute error and used to predict LW for the entire cohort. Clinical factors associated with a >= 20% margin of error were identified in a multivariable analysis after propensity score matching (PSM) of donors with similar anthropometric characteristics. Results The total LW was underestimated with a >= 20% margin of error in 53/500 (10.6%) donors and overestimated in 62/500 (12%) donors. After PSM analysis, ages >= 65, (OR = 3.21; CI95% = 1.63-6.31; P = .0007), age <= 30 years, (OR = 2.92; CI95% = 1.15-7.40; P = .02), and elevated gamma-glutamyltransferase (GGT) levels (OR = 0.98; CI95% = 0.97-0.99; P = .006), influenced the risk of LW overestimation. Age >= 65 years, (OR = 5.98; CI95% = 2.28-15.6; P = .0002), intensive care unit (ICU) stay with ventilation > 7 days, (OR = 0.32; CI95% = 0.12-0.85; P = .02) and waist circumference increase (OR = 1.02; CI95% = 1.00-1.04; P = .04) were factors associated with LW underestimation. Conclusions Increased waist circumference, age, prolonged ICU stay with ventilation, elevated GGT were associated with an increase in the margin of error in LW prediction. These factors and anthropometric characteristics could help transplant surgeons during the donor-recipient matching process.