Clinical validation of glucose pump test (GPT) compared with ultrasound dilution technology in arteriovenous graft surveillance.

BACKGROUND Blood flow (Qa) measurements are an important step in the surveillance protocol of haemodialysis vascular access (VA). The glucose pump test (GPT) is a new test for Qa measurement based on the dilution of a constant glucose infusion. The aim of this study is to verify the clinical accuracy of GPT in a graft surveillance protocol with sequential Qa measurements. METHODS In 30 chronic haemodialysis patients with graft, we compared monthly sequential Qa measurements performed with GPT in pre-dialysis and the ultrasound dilution technique (HD01 device Transonic Systems Inc., USA) during dialysis. The colour Doppler ultrasonography study (CDU) was our reference standard for the diagnosis of stenosis. The endpoints were the graft thrombosis or PTA treatment. RESULTS According to the K/DOQI guidelines we could identify the thrombosis high-risk grafts when Qa was <600 ml/min or <1000 ml/min with a decrease >25% in serial Qa measurements. HD01 yielded 27 of 112 high-risk Qa measurements (21 Qa <600 ml/min; mean 406+/-145 ml/min; 6 deltaQa >25%; mean 43+/-7%). In 12 of 27 cases the CDU control did not show haemodynamically significant stenoses (false positive); 15 of 27 cases were confirmed high-risk accesses by CDU and did PTAs (HD01 specificity 86%). GPT yielded 14 of 112 high-risk Qa measurements (8 Qa <600 ml/min; mean 404+/-135 ml/min; 6 deltaQa >25%; mean 38+/-8%) and all had severe stenoses and underwent PTA treatments showing a GPT specificity of 100%. The CDU study allowed us to correctly assess the Qa negative cases. HD01 method had 10 false negative cases (treated or clotted grafts with a Qa >600 ml/min and deltaQa <25%) with a sensitivity of 60%, while GPT had 11 false negative cases with a sensitivity of 56%. The diagnostic accuracy tested with the ROC curves was similar with both tests (area under the curve was 0.762 and 0.752 with GPT and ultrasound dilution, respectively; P = 0.985). The diagnostic efficiency (percentage of grafts with agreement between test result and factual situation) was 90 and 80% (P = 0.056) for GPT and HD01, respectively. CONCLUSION Compared with HD01, the GPT had a lower false positive rate and similar diagnostic accuracy and efficiency. The clinical implication is a smaller number of unnecessary, invasive procedures (angiographies or PTAs), without increasing the thrombosis risk. This study has shown that GPT is an accurate, quick and economic test for Qa monitoring.