American Society of Diagnostic and Interventional Nephrology Section Editor: Stephen Ash: Thresholds for Significant Decrease in Hemodialysis Access Blood Flow

Publisher Website
Google Scholar
Abstract
During hemodialysis access surveillance, referral for evaluation and correction of stenosis is based upon determination that a significant decrease in blood flow (Q) has occurred. However, criteria for determining when a decrease is statistically significant have not yet been established. In this study we established such criteria by analyzing Q variation with the glucose pump test (GPT). We took nine Q measurements in each of 25 patients (18 grafts, 7 fistulas) during three dialysis sessions within a 2‐week period (predialysis and during hours 1 and 3). We determined thresholds that define a significant percentage decrease in Q (ΔQ) for various p values. In order to confirm the general applicability of these thresholds, we computed the average within‐patient Q variation during the three sessions (computed as a coefficient of variation and referred to as short‐term variation). We then determined the relative influences of biological (true) variation and analytical error on short‐term variation. We found that ΔQ must be > 33% to be significant at p < 0.05, whereas the threshold is > 17% for p < 0.20. Measuring Q at uniform versus different times during the sessions did not significantly reduce these thresholds. We also found that biological variation was nearly as large as short‐term Q variation, whereas analytical error contributed minimally to short‐term variation. In conclusion, this study defines thresholds for a significant ΔQ that have wide application in determining access referral for evaluation and correction of stenosis. Selection of a particular threshold should consider the relative importance of avoiding thrombosis versus avoiding unnecessary procedures. If avoiding unnecessary procedures is a priority, then we recommend a threshold of > 33%. These thresholds apply to other methods of measuring Q, provided analytical error is significantly less than biological variation.