Well-functioning arteriovenous vascular access (VA) is the basis for maintaining successful hemodialysis (HD); however, thrombosis is the leading cause of VA loss in patients with end-stage renal disease (ESRD) undergoing HD [1]. Thus, the latest 2019 kidney disease outcomes quality initiative (K/DOQI) guidelines recommend monitoring, including a detailed physical examination of the arteriovenous VA, by a knowledgeable and experienced health practitioner [2] because such careful physical examination can reveal meaningful VA flow dysfunction. However, such clinical assessments are subjective and require well-trained expert staff who may not be readily available at every HD unit.

One of the indicators of an imminent thrombosis is a reduction in the blood flow rate (Qa) across the VA. Measuring the VA Qa using non-invasive techniques, such as the ultrasound dilution method (UDM), is one of the most commonly used tools to predict subsequent thrombotic events in clinical practice [3-8]. However, data on the effectiveness of this technique in predicting the risk of thrombotic events and improving the longevity of arteriovenous fistulas (AVF) are limited [5-7,9]. Another issue is the insufficient evidence to determine the best possible Qa cutoff value with sufficient sensitivity to detect AVF with a high risk of thrombosis but with good specificity to avoid unnecessary interventional procedures.

Therefore, this study aims to identify VA at a high risk for thrombotic events and consequent loss among patients undergoing HD and investigate whether it can be achieved by monthly measurement of Qa with UDM. In addition, we explored the ideal Qa cutoff value for predicting VA thrombosis occurrence.

The 2019 K/DOQI guidelines recommend regular physical examination of the AVF by a trained dialysis provider instead of routine AVF surveillance by measuring access blood flow as there is inadequate evidence on its measurement [2]. However, not all HD units have experienced individuals perform this examination. Moreover, several studies have reported that subclinical access f low dysfunction may not be detected on physical examination alone [6,8,14]. Thus, the 2019 guidelines commented on the complementary role of surveillance in routine clinical monitoring, as the well-timed identification of VA with flow dysfunction is vital for preserving their function [2].

Among VA surveillance methods, measuring Qa using UDM has been a readily applicable tool owing to its sensitivity and accuracy and can be performed at the bedside more frequently and regularly without a heavy workload on the dialysis staff compared with other methods such as Doppler ultrasonography (DUS) [3,6,15,16]. Since the introduction of UDM in 1995, numerous studies have investigated the usefulness of Qa measurements in detecting access flow dysfunction and predicting future access thrombosis. However, only a few studies have evaluated the impact of Qa surveillance on AVF longevity, and the results have been inconsistent [5-7,9]. A prospective randomized controlled trial (RCT) conducted in Italy reported that the risk of thrombosis was significantly reduced, and the cumulative patency of AVF was improved in the group that was surveyed with trimonthly Qa measurement using UDM compared with in the control group during a follow-up period of 5 years, suggesting that regular Qa measurement can help increase the overall life of VA [5]. In addition, an open-label, prospective, multicenter RCT with a 3-year follow-up in Spain showed that Qa-based surveillance reduced the frequency of thrombosis and improved the thrombosis-free and secondary patency of AVF [6]. Recently, the association between monthly surveillance of AVF using UDM and a reduced number of thrombotic events per patient was demonstrated in the largest RCT in the USA [7]. In contrast, a single-center longitudinal observational study reported that monthly access flow monitoring with increased prophylactic angioplasty (0.67 vs. 0.09 per access-year) did not improve the thrombosis rate or cumulative fistula patency [9]. However, all these studies were limited because they included a small number of patients (n=79) with only AVF with subclinical stenosis and a low Qa of <500 mL/min [5], used historical rather than concurrent controls [9], evaluated patients with a high risk of VA thrombosis identified and hemodynamically proven by UDM and DUS, which can cause an extra burden on HD staff [6], did not separately analyze patients who had AVFs from those who had AVGs, or did not enroll the targeted number of patients, resulting in insufficient power to unveil the impact of access surveillance on the outcomes of VA [7]. In this study, we investigated the clinical usefulness of monthly measuring Qa among patients undergoing HD with functional AVF and reported that patients who had a Qa of <400 mL/min during the follow-up period (median, 3.1 years; IQR, 1.1-5.0) had greater risks of VA thrombosis and loss than the risks of those who had greater Qa. Thus, in agreement with the aforementioned studies [5-7], our findings suggested that regular AVF flow measurements using UDM can identify a subset of patients at an increased risk of future VA thrombosis and loss.

Qa thresholds in AVF for referral to angiography vary according to different guidelines, ranging from 300 to 500 mL/min [13,17-19]. However, such recommendations are mainly based on very few previous studies that used ROC curve analysis. In a prospective controlled open trial of patients with mature forearm AVFs, Tessitore et al. [4] reported that the optimal cutoff value of Qa as the best predictor of incipient thrombosis was <300 mL/min. However, one limitation of their study was that the predictive role of the appropriate threshold for thrombosis was evaluated in a relatively small number of patients (n=65) over a relatively short period (median, 12 months). Moreover, in a long-term (5 years) follow-up study of 79 patients undergoing HD with mature forearm AVF, they reported that the cutoff point for intervention and prevention of the risk of thrombosis was a Qa of 350 mL/min, suggesting that a Qa of >350 mL/min before intervention can lead to a superior outcome with preemptive action in AVFs [5]. The Qa thresholds of 300-350 mL/min are among the lowest reported values. However, selecting such low thresholds may increase the risk of thrombosis in patients undergoing HD, although unnecessary procedures are avoided. In addition, an Australian study confirmed that flow measurement using the online thermodilution method is a valuable tool for predicting AVF failure, and an optimal discriminatory cutoff value of 576 mL/min gives a sensitivity of 89% and a specificity of 81% in ROC curve analyses [20]. As the Qa threshold has been reported to depend on the methods used [16], the disparity in monitoring methods may be associated with differences in the appropriate cutoff values among previous studies. In addition, the AVF function was explored within 6 months, which is not long enough to evaluate the overall outcomes of VA. A recent retrospective study on 302 patients undergoing HD in Taiwan reported that the minimum Qa required to maintain access function was associated with individual ideal body weight (IBW) for radiocephalic AVFs with a Qa of <500 mL/min, and the best cutoff value of Qa was 7.1 times that of IBW [8]. However, they enrolled only a small number of patients (n=52) with AVF with subclinical stenosis and a low Qa (<500 mL/min). In addition, they reported the ideal threshold of Qa as an IBW-based form in just 1 year. Altogether, there is inadequate evidence supported by large RCTs on whether different Qa criteria influence the long-term outcomes of AVF. In the present study, an ROC curve analysis showed that a Qa of 400 mL/min had a high predictive ability for imminent thrombosis, with an AUC of 0.908, indicating that measuring Qa is helpful in clinical practice for identifying patients at a high risk of VA thrombosis. This threshold is similar to that suggested by previously published guidelines as the best predictor of thrombosis in AVFs [13] and slightly higher than those identified in previous studies [4,5]. In their study, all patients had forearm AVFs; the flow rates of which were usually lower than those of upper-arm AVFs. In contrast, more than one-third of the patients (34.1%) had upper-arm AVFs in our study, suggesting that the different AVF locations might explain the discrepancy in the threshold Qa value between that study and ours. Regarding the long-term outcomes of VA, our patients who had Qa of <400 mL/min showed significantly greater AVF loss than those with a Qa of >400 mL/min during a follow-up period of up to 5 years, suggesting that a drop in Qa to <400 mL/min is associated with subsequent VA thrombosis and loss.

The sustained maintenance of VA patency without flow dysfunction is essential and can affect the prognosis of patients undergoing HD [1]. Thus, several studies have assessed the risk factors associated with VA thrombosis and/or loss and revealed that many risk factors such as age [21], sex [22], the presence of DM and hypertension [3,15,21], the use of drugs with antiplatelet and/or vascular antiproliferative activity [23], and AVF location [9] are associated with VA thrombosis. However, no factors were consistently associated with AVF thrombosis. Similar to previous findings, older age, female sex, dialysis vintage, the presence of DM and hypertension, the use of antiplatelet agents, and AVF location were not associated with VA thrombosis in the present study. Only a history of stroke and VA thrombosis, and low Qa were predictors of VA thrombosis, indicating that patients with these risk factors should be surveyed and evaluated more carefully to reduce VA thrombosis.

Our study had several limitations. First, it was a retrospective, single-center, observational study, which might not apply to the general population with ESRD undergoing HD. Second, we did not perform angiography, the gold standard for all patients. Patients who did not undergo angiography were assumed to be negative for thrombosis and were included in the AUC calculation. Third, we did not determine the optimal frequency of Qa measurement because all patients underwent Qa measurement every month, based on previously published clinical practice guidelines [13] to detect access flow dysfunction early and permit sufficient lead time for an intervention. Thus, it remains unclear whether more frequent Qa measurements, such as once a month, would identify AVFs at a risk of thrombosis more timely than less frequent surveillance, such as bi- or trimonthly. Further studies are required to define the optimal interval between measurements for early detection of an increased risk of thrombosis and subsequent AVF loss. Fourth, this study did not assess the economic benefits of this strategy. However, national health insurance covers regular monthly Qa measurements in South Korea. Thus, it is unlikely that this type of surveillance will impose a considerable economic burden on patients.

In conclusion, our study revealed that low Qa values during the follow-up period were strongly associated with the occurrence of thrombosis and subsequent loss of AVFs in prevalent patients undergoing HD, suggesting that the detection of patients at a risk for imminent VA thrombosis and loss could be improved by regularly measuring Qa, and patients with a high thrombosis risk profile, such as those with a prior history of VA thrombosis and low Qa, require a more thorough evaluation.