Clinical performance characteristics of hemodialysis graft monitoring.

BACKGROUND Regular monitoring of dialysis grafts is recommended, but the value of dialysis graft blood flow monitoring and venous pressures in predicting subsequent outcomes are controversial. METHODS Over a period of one month, we performed simultaneous flow and venous pressure monitoring in 71 dialysis patients with polytetrafluoroethylene (PTFE) grafts. These patients were prospectively followed for one year. Receiver operating characteristic (ROC) curves were constructed to evaluate the performance of the various monitoring techniques. RESULTS During the period of follow-up, there were 71 graft failures (30 angioplasty alone and 41 thrombosis followed by interventional or surgical revisions). Failed grafts had a lower blood flow rate [799 +/- 452 (SD) mL/min] when compared with those without failure (1019 +/- 485 mL/min, P = 0.05) Single static or dynamic venous-pressure monitoring were not predictive of graft failure. ROC analysis showed poor performance of graft flows in predicting graft failures over the short (30 days, AUC = 0.726, 95% CI, 0.509 to 0.942) and long term (one year, AUC = 0.630, 95% CI, 0.499 to 0.761). An adjustment of graft flows for systolic blood pressure or classification of graft based both on flows and venous pressure did not improve test performance. CONCLUSIONS Although dialysis graft blood flow rates are statistically different in patients who have graft failure (graft angioplasty and surgery or thrombosis) versus those who do not, the performance characteristics preclude clinical decision-making from an isolated blood flow or venous pressure study.