Association between blood pressure, ultrafiltration, and hemodialysis graft thrombosis: a multivariable logistic regression analysis.

Although a low blood flow (Q(a)) is the most important cause of graft thrombosis, several studies have shown that Q(a) measurements do not accurately predict thrombosis. This suggests that additional variables may influence thrombosis. Identification of such variables may be essential to designing surveillance protocols that accurately predict thrombosis. In this nested case-control study, we prospectively followed 105 patients for up to 2.5 years in order to test the association of a number of variables with thrombosis. These included Q(a) (monthly by ultrasound dilution), percentage stenosis (quarterly by duplex ultrasound), mean arterial pressure (MAP), percentage ultrafiltration (%UF) during dialysis (%UF = 100[liters]/[kilogram of weight]), and other variables that defined patient and graft characteristics. Patients were divided into patent (n = 53) and thrombosed groups (n = 52), and MAP and %UF from seven consecutive dialysis sessions were analyzed. In the thrombosed group, the last session was the final session before thrombosis. A multivariable logistic regression model showed that Q(a), MAP (the predialysis average of seven sessions), and %UF (from the last session) were independently associated with thrombosis, whereas all other variables were not. The model yielded the following odds ratios for thrombosis: for a single Q(a) value (reduction of 1,000 mL/min), 12.0 (P < 0.01); for %UF (increase of 4%), 5.3 (P < 0.01); for MAP (reduction of 30 mm Hg), 4.1 (P = 0.02); and for percentage decrease in Q(a) (> or =20% versus <20%), 2.4 (P = 0.12). We conclude that in addition to Q(a), both %UF at the last session before thrombosis and average predialysis MAP from seven sessions are independently associated with thrombosis. These results help explain why Q(a) alone does not accurately predict thrombosis. A prospective study is needed to determine whether %UF at each session and a moving average MAP from seven sessions improve the prediction of thrombosis. However, it should be recognized that a large %UF is a preterminal event that likely provides too short a warning for intervention before thrombosis.