Hidden blood loss in robot-assisted vs. conventional midline lumbar interbody fusion: a single-center propensity-score-matched cohort study.

BACKGROUND

Hidden blood loss (HBL) is a substantial, often underestimated component of bleeding in spine surgery. Evidence for its reduction with robot-assisted midline lumbar interbody fusion (RMIDLIF) remains limited.

METHODS

We conducted a single-center retrospective cohort comparing RMIDLIF with conventional MIDLIF. After excluding cases outside the common-support region, 1:1 propensity score matching yielded 82 pairs. The primary endpoint was HBL; secondary endpoints included total blood loss (TBL), visible blood loss, operative time, and length of stay (LOS). Within RMIDLIF, multivariable linear regression and receiver operating characteristic (ROC) analyses evaluated factors associated with severe HBL (≥ 300 mL).

RESULTS

Compared with MIDLIF, RMIDLIF showed lower HBL (320.45 ± 97.70 vs. 467.21 ± 146.36 mL; mean difference − 146.76 mL, 95% CI − 185.18 to − 108.35; P < 0.001) and TBL (476.02 ± 159.22 vs. 714.28 ± 211.44 mL; difference − 238.25 mL, 95% CI − 296.01 to − 180.50; P < 0.001). Operative time (− 17.2 min; P < 0.001) and LOS (− 0.79 days; P = 0.01) were also reduced. Within RMIDLIF, higher HBL was independently associated with more fused levels (β = 76.9 mL/level; P = 0.001), longer operative time (β = 2.32 mL/min; P = 0.002), higher ASA class (β = 74.27 mL/class; P = 0.001), and greater preoperative fibrinogen (β = 55.69 mL/g·L⁻¹; P = 0.002). ROC analyses identified clinically useful cutoffs for operative time (128.5 min) and fibrinogen (2.82 g/L) to flag severe HBL.

CONCLUSIONS

RMIDLIF was associated with lower hidden and total blood loss, shorter operative time, and shorter hospitalization than conventional MIDLIF. In RMIDLIF, fusion levels, operative time, ASA class, and preoperative fibrinogen independently predicted higher HBL; the identified thresholds may assist perioperative risk stratification and blood management. CLINICAL TRIAL NUMBER: Not applicable.