Objective: To assess subclinical haemostatic activation and clinical variables to predict bleeding during radical retropubic prostatectomy (RP), as haemostatic activation is common in cancer and might be useful for predicting outcomes, but routine coagulation screening does not correlate with bleeding.
Patients and methods: Clinical data and blood samples were collected from 153 patients (median age 63 years; prostate-specific antigen, PSA, level 5.92 ng/mL) before RP and lymph node dissection. Plasma was assayed for d-dimer and thrombin-antithrombin complex (TAT). Univariable then multivariable analyses were used to identify associations between plasma markers and clinical variables for bleeding and thrombosis.
Results: Most patients (77%) were stage T1c and most (76.5%) had organ-confined cancer (< or =pT2). Pathological Gleason scores were < or =6 in 68 (44.4%) and > or =8 in 14 (9%) of the patients. The median (range) estimated blood loss (EBL) was 400 (50-3000) mL, the median decrease in haemoglobin level 3.5 (-0.1, 6.6) g/dL, and eight men (5.2%) required a transfusion. In the univariable analysis, a lower TAT before RP (P < 0.001) and d-dimer level (P = 0.023) correlated with a greater decline in haemoglobin level. The platelet count, international normalised ratio, and activated partial thromboplastin time (aPTT) did not predict the EBL nor change in haemoglobin level; the eight transfused patients had lower platelet counts before RP (P = 0.004). Higher surgical volume predicted a lower EBL (P < 0.001) and lower decrease in haemoglobin (P < 0.05). Multivariable linear regression showed that TAT remained significant for the decrease in haemoglobin, and surgical volume for EBL and decrease in haemoglobin.
Conclusions: Haemostatic activation before RP was associated with significantly less bleeding when assessed by objective measures, predicting the decrease in haemoglobin level better than prothrombin time, aPTT or platelet counts. Current surgeon volume might also predict both subjective and objective bleeding variables.