Measurement of arteriovenous fistula flow using the temperature gradient method compared with Doppler ultrasound: Predictive capacity for the presence of stenosis

The most common complication of arteriovenous fistula (AVF) is stenosis, which can result in poorer hemodialysis (HD) outcomes and thrombosis. It is important to identify those stenoses at high risk of thrombosis in order to perform elective treatment to restore their function.

1) To assess the degree of agreement between Doppler ultrasound (DU) and the temperature gradient technique (TGT) for blood flow measurement (Q_A). 2) To evaluate the relative importance of different monitoring and/or surveillance methods for detecting significant stenosis using a predictive model.

A cross-sectional, comparative, observational study was conducted, in which the AVFs of 30 consecutive patients were assessed over a 3-month period, with monthly sampling. First-generation methods were applied, and second-generation methods using TGT and DU were used to measure Q_A.

80% were men and 20% women, with a mean age of 67.03 ± 10.25 years. 100% were hypertensive and 63.33% diabetic. 76.67% had a radiocephalic AVF and 23.33% a humeral AVF. The mean Q_A by TGT was 980.79 ± 63.86 ml/min and by DU 1098.43 ml/min. In the first sampling, 4 significant stenoses were detected, none in the second, and 3 in the third. In samplings 1 and 3, statistically significant differences were detected in the measurement of Q_A by TGT and ultrasound. No significant differences were found in any parameter measured by first-generation methods, except for PV in the first sampling. In the Bland-Altman analysis, a 96.7% agreement was observed between Q_A measurements by TGT and DU (p < 0.0001). The dot plot for the difference of each measurement from the mean for AVFs with Q_A > 1000 ml/min also showed agreement between both methods, with a deviation of only 4.34%. When first- and second-generation variables were introduced into the predictive model, four combinations with a significant Wilks' lambda were identified. The combination including Q_A determined by ultrasound was the most significant, with an overall correct classification of 80.5% when first-generation methods were included together with ultrasound-measured Q_A.

In our study, the agreement between Q_A measurement by TGT and Doppler ultrasound is high at all flow levels. Furthermore, the predictive model demonstrates that second-generation methods are superior to first-generation methods for the early detection of clinically significant stenosis.