Doppler ultrasound surveillance of recently formed haemodialysis arteriovenous fistula: the SONAR observational cohort study

Article history

The research reported in this issue of the journal was funded by the HTA programme as award number NIHR135572. The contractual start date was in May 2018. The draft manuscript began editorial review in November 2022 and was accepted for publication in June 2023. The authors have been wholly responsible for all data collection, analysis and interpretation, and for writing up their work. The HTA editors and publisher have tried to ensure the accuracy of the authors’ manuscript and would like to thank the reviewers for their constructive comments on the draft document. However, they do not accept liability for damages or losses arising from material published in this article.

Copyright statement

Copyright © 2024 Richards et al. This work was produced by Richards et al. under the terms of a commissioning contract issued by the Secretary of State for Health and Social Care. This is an Open Access publication distributed under the terms of the Creative Commons Attribution CC BY 4.0 licence, which permits unrestricted use, distribution, reproduction and adaptation in any medium and for any purpose provided that it is properly attributed. See: https://creativecommons.org/licenses/by/4.0/. For attribution the title, original author(s), the publication source – NIHR Journals Library, and the DOI of the publication must be cited.

2024 Richards et al.

Background

The UK Renal Registry data highlight that the incidence of people commencing renal replacement therapy in the UK increased from 109 per million population in 2013 to 115 per million population in 2014 and, most recently, to 151 per million population in 2019. 1 This equates to some 7500 patients each year, of whom approximately 70% receive haemodialysis as the first option, with as many as two-thirds of these initially reliant on a central venous catheter (CVC). Around 70,000 adults currently receive kidney replacement therapy, and notwithstanding numbers remaining static through 2020 (most likely a consequence of the COVID pandemic), it is anticipated that this number will again increase by 2–2.5% annually, as has occurred over the last decade. 1

Arteriovenous fistulas (AVFs) are considered the best modality for providing haemodialysis care in patients with end-stage renal disease (ESRD). Compared to dialysis via an AVF, haemodialysis via a CVC is associated with an increased risk of catheter-related bloodstream infection, with an estimated incidence of 1.2–2.5 per 1000 patient-days. 2–6 This results in increased hospitalisation and additional costs. 7,8 Patient mortality for patients dialysing via a CVC is ~40% higher than for patients dialysing via an AVF. 9 There is therefore a clear incentive to form AVFs in patients requiring haemodialysis, and current UK tariffs for haemodialysis have been deliberately set to incentivise dialysis via an AVF.

However, only 20% of UK dialysis centres currently achieve the 80% Renal Association target for dialysis of their prevalent population via definitive access, and many fall well short. 10,11 The reasons why such a small proportion of the prevalent dialysis population achieve dialysis via an AVF are multifactorial, but once formed, AVFs must ‘mature’ over several weeks before they can be used for dialysis, and the relatively poor maturation rate, with approximately 30% of fistulas failing to do so,12–17 undoubtedly contributes. Early identification of AVFs that are unlikely to mature may allow timely, and more effective, surgical or radiological salvage. The resulting improvements in assisted primary patency would likely increase AVF usage and, by minimising morbidity and mortality related to CVC dialysis, such interventions are likely to be cost-effective. Furthermore, early salvage of AVFs that are still patent but deemed at risk of non-maturation may preserve precious venous ‘capital’ by avoiding subsequent loss of the draining venous segment of the AVF through late thrombosis.

Successful AVF maturation is marked by a massive increase in blood flow through the fistula, and accompanying augmentation in fistula vein diameter. Early ultrasound (US) surveillance of nascent AVFs may therefore identify patterns of fistula blood flow and vein size that are associated with non-maturation. For example, the fistula vein can develop progressive intimal hyperplasia at the site of mobilisation onto the artery, which may compromise fistula maturation and eventually lead to thrombosis. Such juxta-anastomotic (or ‘swing-segment’) stenosis18 may be detectable by early US. However, the literature relating to the use of US surveillance of AVFs, even once mature and being used for dialysis, is conflicting, and a consensus strategy has not been reached. This may relate to differences in the mode of surveillance, the site of the fistula and the scanning method adopted. US can confirm fistula maturation,19,20 although a consensus definition of what constitutes maturity on US has not been agreed. 21 The use of an US scan as a surrogate for fistula maturity enables pre-dialysis patients to be included in dialysis trials.

Only a few studies have used US immediately after fistula creation to assess early maturation, but these report that maturation is characterised by rapid increases in fistula blood flow, as early as the first day after formation. 22–24 The diameter of the fistula vein also increases rapidly. 23 This raises the possibility that US assessment at early time points may predict subsequent fistula maturation. Itoga et al. reported early duplex US on 153 patients with newly formed fistulas (4–8 weeks after creation). 25 US detected a flow-limiting stenosis in 40% of patients, with radiological salvage intervention subsequently performed in 81% of them. Assisted primary patency of the fistulas in this group was 83% and 64% at 6 months and 1 year, respectively, compared to 96% and 89% in the cohort who had no detectable US abnormality. This study did not include a control cohort who did not receive routine surveillance, but the fistula patency rates achieved for the entire study population were better than generally reported in vascular access studies. Routine early US surveillance has been examined in one randomised study to date of 150 patients, with US performed at 2, 4 and 8 weeks after fistula creation. 15 The reported fistula failure/non-maturation rate in the surveillance group was 13.6%, compared to 25.4% in the control group in whom US was performed on the basis of a perceived clinical indication. This difference did not reach statistical significance, but notably, the study was powered for a relatively large (20%) difference in maturation.

Rationale for study

Although US surveillance programmes of mature fistulas are increasingly incorporated into standard practice, controversy persists regarding their effectiveness, with limited high-quality evidence available. 26–28 Design of prospective controlled trials is hampered by the low rates of thrombosis observed once AVFs have matured, particularly for standard radiocephalic fistulas:13,29 audit of the Cambridge prevalent dialysis population (~450 patients) has revealed a thrombosis rate for mature fistulas of 0.18 per patient/year, which matches existing literature. 14,30,31 The relatively poor reported 1-year unassisted (primary) fistula patency rates of around 55% are instead largely due to early failures occurring within the first 3 months after creation, before the fistula has matured. 13,14 Later failures in the first year may also reflect suboptimal maturation. 32,33 Thus, the trial team felt that a surveillance programme that focuses upon identifying potentially rectifiable lesions in the first 3 months after fistula formation has greater potential to deliver an effective and economic approach to improving 1-year AVF patency.

Hypothesis and objectives

The paucity of evidence supporting a role for surveillance in increasing fistula patency rates, coupled with the significant costs involved and the resources required, leads to significant heterogeneity in practice across the UK and internationally. This study aims to fill this gap in knowledge by determining whether US surveillance of newly formed AVFs can improve fistula patency rates and minimise CVC usage. Our overall aim of this 5-year project is to test the hypothesis that:

Doppler US surveillance of AVFs immediately after creation improves longer-term AVF patency, by directing early and effective surgical or radiological salvage of those AVFs at risk of failing or not maturing.

For a trial to demonstrate that US surveillance improves patency rates for newly created fistulas, several conditions must be met:

1. that US can effectively distinguish those newly formed fistulas that are unlikely to mature

2. that maturity failure occurs commonly enough that clinically meaningful improvements in fistula outcomes by early identification of at-risk fistulas are plausible

3. that salvage interventions performed on those ‘at-risk’ fistulas are effective and improve fistula patency.

As discussed above, evidence supporting these conditions is lacking, and given that failure to meet them would be expected to generate a negative trial result, they were addressed separately and sequentially, according to the following objectives:

1. Run an observational cohort study in which consenting participants undergo serial US assessment of their AVF in the first 3 months after its formation (phase 1).

2. Model whether features on early US can reliably identify those fistulas that will not mature or will fail early.

3. Assess from the observational cohort study whether a randomised controlled trial (RCT) evaluating early US-guided salvage intervention is feasible.

4. Run a multicentre RCT in which 1-year fistula patency in a treatment group receiving early US surveillance of their developing fistulas is compared against standard care: monitoring of fistula maturation by clinical assessment only (phase 2).

Progression to the phase 2 study is therefore contingent upon accomplishing the first three of these objectives, and in particular, demonstrating that US surveillance can accurately predict fistula non-maturation. This manuscript will thus focus on the phase 1 study set-up and outcomes.

Study design

A prospective observational cohort study of adult patients undergoing formation of AVF for haemodialysis was performed to test the hypothesis:

Doppler US surveillance early after AVF creation can reliably identify those AVFs that will not mature or will fail early.

Consenting participants underwent serial US scanning at weeks 2, 4, 6 and 10 after fistula formation in addition to standard care (such as regular clinical assessment) as per local centre policy.

Study setting

The study was performed at 17 UK dialysis centres (Table 1).

Selection of participants

Participants were considered eligible for enrolment in this study if they fulfilled all inclusion criteria and none of the exclusion criteria detailed below.

Study outcomes

Assessments and follow-up

The schematic for study participation is shown in Figure 1.

FIGURE 1.

Study schema. (a) The clinical team and the participants will be blinded to the results of the ultrasound scans. Routine clinical assessments may be performed during the trial period, as per standard practice.

Follow-up 12-month study

Upon analysis of the 10-week outcome data for the SONAR study, the Trial Management Group (TMG) and the Trial Steering Committee (TSC) recommended that the study be extended to examine whether early US surveillance could identify fistulas that were likely to have failed by 6 and 12 months. Thus, approvals were gained for a second observational cohort study, the SONAR-12M study.

Statistical analysis plan

All statistical analyses were carried out using Base SAS version 9.4 (SAS Institute Inc., Cary, NC, USA).

Progression to proposed randomised controlled trial

As discussed above, this study was performed to assess the practicalities of introducing a multicentre randomised trial in which early surveillance US would be used to identify patients with newly formed fistulas that are at risk of not maturing or failing, and these patients would then be randomised to standard care (continued observation) or to radiological or surgical salvage.

The key condition set for proceeding to the second SONAR phase was that early US surveillance could identify fistula non-maturation with a PPV of at least 70% (i.e. the 95% CI for PPV includes 70%). This target was chosen because, although the impact of early US surveillance on improving fistula patency through informing radiological or surgical salvage is unknown, pragmatically we felt that surveillance programmes were only warranted if relatively sizeable improvements in fistula patency were achieved. In this regard, the study team felt that a minimal improvement of 10% in primary assisted patency rates (from an estimated 55% to 65%) was reasonable, and that the benefits of surveillance would be questionable if the difference were less marked. Power calculations indicated that to detect this difference would require 1224 fistulas at 90% power, 5% significance and allowing for 20% dropout, which, although a large trial, would be attainable within a multicentre UK setting (see Table 9).

Ethical and regulatory considerations

Patient characteristics

Between 1 September 2018 and 11 November 2019, 682 assessments for eligibility for inclusion into the SONAR study were carried out at 17 UK haemodialysis sites, with 347 consents (see Table 1). Patient demographics are detailed in Table 3. Patients were generally elderly and male, with ~25% having cardiovascular, cerebrovascular or peripheral vascular disease, and with > 40% diabetic. The study cohort mirrored coincidental UK experience. 36 At enrolment, 191 (55.0%) patients were still pre-dialysis, and a further 8 (2.3%) had a previous renal transplant that was now failing.

Patient enrolment, fistula creation and adherence to trial protocol

Participant enrolment is shown in Figure 2. Fourteen patients re-enrolled in the study, following failure of their original SONAR AVF. Fourteen patients did not undergo AVF creation, mostly because of ill health. Operative details for the 333 AVFs created are provided in Table 4: the majority (75.1%) of AVFs were created on the left side, reflecting the non-dominant arm, with slightly more (52.3%) AVFs created at the elbow than at the wrist. Twenty-seven participants withdrew from the study; 9 of these withdrew before an AVF was created and 18 afterwards, mainly because of either failing health or death.

FIGURE 2.

Study CONSORT diagram.

Participants were scheduled to undergo US scanning of their AVF at weeks 2, 4, 6 and 10 after AVF creation. Despite concerns that this regimen would prove too onerous, particularly for the frailer, older participants, attendance for the scans was good: 126 (37.8% of those with an AVF created) cases missed at least one of their expected scans, but at any particular time point approximately three-quarters of patients attended for the scheduled scan (Table 5; note that further scans were not performed once an AVF was confirmed to have thrombosed).

Fistula patency and maturation rates

As detailed in Table 6, by week 10 after AVF creation, four outcomes are possible: the AVF has reached maturation; the AVF is patent but non-mature; the AVF has failed (thrombosed or abandoned); and unknown. By 10 weeks, 219 of 333 (65.8%) of created AVFs had reached maturity according to predefined US criteria (see Primary outcome measures), with 67.2% of elbow and 60.4% of wrist AVFs maturing. Conversely, by week 10, 57 of 333 (17.1%) of AVFs had failed. However, 37 of the fistula failures (64.9%) had occurred by the week 2 scan, and thus too early after fistula creation for salvage procedures to be practical (Table 7). In this regard, it is notable that if these early failures within the first 2 weeks are excluded, fistula maturation of 74.0% at 10 weeks was achieved (see Table 7). A relatively small number of AVFs (n = 29; 8.7%) remained patent, but not mature, on the week 10 scan (see Table 6).

Univariate analysis was performed to identify patient and pre-operative anatomical factors associated with fistula non-maturation at week 10. Candidate factors included: pre-operative vein diameter; quality of artery at the time of surgery; quality of vein at the time of surgery; clinical prediction of fistula maturity; patient sex; patient age; and history of diabetes. Pre-operative vein diameter was excluded in the wrist model due to missing values. Out of those factors known at baseline, only sex was univariately significant at the 5% level, and for wrist fistulas only. For elbow fistulas, no baseline factor was significant at the 5% level; pre-operative vein diameter, identified previously as an important predictive factor in AVF maturation,37–39 was not statistically significant.

Early surveillance ultrasound

The most immediate and surprising finding from analysis of the US data was how rapidly fistula flow rates and venous diameters increased after AVF creation (Figure 3a and b). Fistula flow rates were calculated by the study vascular scientists by measuring brachial artery flow, on the assumption that blood flow through a brachial artery prior to AVF creation is minimal (< 100 ml/minute) and that increased flow thereafter reflects additional flow through the AVF. Thus, for example, at the first scan at 2 weeks, a median blood flow of 770 ml/minute and a median venous diameter of 5.2 mm were achieved, excluding those AVFs that had already thrombosed. Consequently, at week 2, 50.3% of wrist and 48.9% of elbow AVFs had reached maturation (Figure 3c and d), according to the US-determined surrogates (see Primary outcome measures), with 16.4% of wrist and 20.1% of elbow AVFs still patent but not yet mature. Interestingly, the proportion of AVFs that were classed as mature did not change dramatically in the subsequent 4-, 6- and 10-week scans (Figure 3c and d), largely because once matured, AVFs tended to remain so, although small numbers did either regress to an immature state or had thrombosed on the subsequent scan (Figure 3e). Conversely, the proportion of AVFs that were immature gradually fell with each successive scan (Figure 3f), either because they had matured or because they had thrombosed or had been abandoned.

FIGURE 3.

Ultrasound assessment of fistula maturation. Following AVF creation, participants underwent US assessment of their fistula 2, 4 and 6 weeks later, with fistula maturity assessed at week 10. Box and whisker plot depiction of (a) fistula flow rates and (b) representative fistula venous diameter for all fistulas at each time point, showing minimum value (after excluding outliers), 25th centile, median, 75th centile and maximum value (after excluding outliers) from bottom to top. Stacked 100% bar charts showing the proportion of (c) elbow and (d) wrist fistulas, with the following outcomes at each of weeks 2, 4, 6 and 10: died; withdrawn; abandoned; thrombosed; mature by US parameters (at that scan); not mature by US parameters (at that scan); unknown (did not attend scan, or missing data from the scan prevented determination of maturity). (e and f): as for (c) and (d) but for all fistulas, presented as numbers and including arrows depicting status at next scan of those fistulas (e) mature or (f) immature at previous scan.

Comparison of the early scan data in those AVFs that had matured by the week 10 scan against those AVFs that were still immature revealed marked differences in the recorded fistula vein diameter and in particular, in fistula flow rates, for both wrist and elbow AVFs (Figure 4). For example, the median blood flow at 2 weeks was 1135.5 and 691.0 ml/minute in those elbow and wrist fistulas, respectively, that reached maturation by week 10, whereas flows of 349.0 and 395.5 ml/minute were recorded in those elbow and wrist fistulas, respectively, that did not reach maturation at week 10. Similarly, the median venous diameter at 2 weeks was 6.3 and 5.0 mm in those elbow and wrist fistulas, respectively, that reached maturation by week 10, whereas the corresponding values in those elbow and wrist fistulas that did not mature were 4.6 and 4.3 mm, respectively. Scatterplot diagrams of average fistula flow against fistula vein diameter for the week 2, 4 and 6 scan data (Figure 5) highlight the different patterns of fistula development in those that will mature by week 10 compared to those that remain immature.

FIGURE 4.

Ultrasound scan parameters at weeks 2, 4, 6 and 10, according to maturation status at week 10. Representative (a and b) fistula venous diameter and (c and d) fistula volume flow rate for elbow (a and c) and wrist (b and d) according to maturation status at week 10. Box and whisker plot shows minimum value (after excluding outliers), 25th centile, median, 75th centile and maximum value (after excluding outliers) without imputation of primary outcome. Fistulas that failed before week 10 (thrombosis or abandonment after a failure) were excluded from the analysis.

FIGURE 5.

Scatterplot of representative venous diameter by average volume flow at each scan time point, according to fistula site and maturation status at week 10. Scatterplot of representative venous diameter by average volume flow at (a) 2, (b) 4 and (c) 6 weeks with different symbols for mature/non-mature fistulas at week 10 (as per primary outcome with no imputation).

Predictive modelling for fistula non-maturation

The data presented in the preceding section suggest it may be possible to identify from the early US findings those AVFs that are unlikely to mature. If so, this raises the potential of performing early radiological or surgical salvage of these AVFs in anticipation that this improves fistula assisted patency rates, and would merit progression to the proposed phase 2 RCT. A binary-data population-average model with exchangeable correlation structure was therefore developed to calculate the probability of fistula non-maturation by week 10, as detailed in the Methods section. Fistula maturity was known for 75% of fistulas and imputed (using a prespecified method based on previous scan or clinical data; see Handling of missing data) for 17% of fistulas (see Table 7). The maturity status for 8% of fistulas could not be imputed, and these fistulas were excluded from the modelling.

The optimum models identified considered elbow and wrist AVFs separately, and included week 4 scan data only; including data from more than one scan time point did not improve performance (Table 8 and Figure 6). Thus, for elbow AVFs, an algorithm that included the week 4 average resistance index and fistula blood flow predicted fistula non-maturation at week 10 in 27 cases, and correctly so in 18 of these (true positives), giving a PPV of 66.7% (95% CI 48.9% to 84.4%). A similar model could be developed for wrist fistulas, incorporating week 4 fistula venous diameter and fistula blood flow, and predicted fistula non-maturation in 33 cases, with 20 of these true positives (PPV 60.6%, 95% CI 43.9% to 77.3%). Diagnostic tests for model fit and influential observations were run on the optimum models for wrist and elbow AVFs, with both performing well, with area under the curve (AUC) values of at least 0.90 (see Figure 6). Interestingly, although the focus was on identifying on early surveillance US those AVFs that were not going to mature, the NPV – that is, the correct identification of those fistulas that were going to be mature at week 10 – was extremely high for both models (wrist: 95.4%, 95% CI 91.0% to 99.8%; and elbow: 95.6%, 95% CI 91.8% to 99.4%).

FIGURE 6.

Receiver operating characteristic curve analysis of optimum models for predicting fistula non-maturation at week 10. Receiver operating characteristic curves for optimum models for predicting week 10 fistula non-maturation from week 4 US findings for: (a) elbow, (b) wrist and (c) all fistulas.

A composite, all-fistula model was then built using a similar approach to that followed for the elbow and wrist-specific models, incorporating a covariate indicating fistula location (elbow or wrist; see Table 8 and Figure 6c). This performed less well than the separate wrist and elbow models, in that it predicted fistula non-maturation in 68 cases, with 33 of these true positives (PPV 48.5%, 95% CI 36.7% to 60.4%). Unlike the individual models for wrist and elbow non-maturation, which relied entirely on parameters of the week 4 US scan, the surgeon’s prediction at creation as to whether the fistula would mature was an independent, albeit weak, factor for fistula non-maturation in the composite model (see Table 8).

Finally, the above modelling was repeated but including only reported outcome data; these models performed very similarly to those incorporating imputed outcome data (not shown).

Twelve-month fistula outcomes

As discussed above, the phase 1 SONAR study was designed as a feasibility study to determine, firstly, whether early surveillance could detect those fistulas that would fail or not mature, but also, critically, whether it would be practical to then perform a multicentre RCT examining whether early surveillance-directed salvage intervention improves fistula outcomes. In this regard, Figure 7 details the potential numbers of wrist and elbow AVFs that could be rescued by the week 4 surveillance US; it is notable (see Table 7) that the 10-week maturation rates for 293 of the original 333 (88.0%) study fistulas that were still patent at 4 weeks were much higher (74.7%; 95% CI 69.4% to 79.6%) than anticipated in the original trial design. Thus week 4 US surveillance could alter maturation in a maximum of 22.0% of AVFs, and factoring in an overall sensitivity of 80.5% (data not shown) for detecting those fistulas that will not mature, this figure is further reduced to 17.8%. Although this is acknowledged to be an estimate, the initial power calculations for the phase 2 RCT (see Table 9) had proposed a sample size of 1224 fistulas, based on detecting a 10% increase in 1-year assisted primary patency. However, from the SONAR phase I study, it seems unlikely that this difference would be achieved, because the intervention either could not be arranged in time or was ineffective. Instead, if one assumes that 50% of the at-risk fistulas identified on the week 4 scan are successfully salvaged, this would provide an effect of 8.9% in the intervention arm.

FIGURE 7.

Summary of week 4 US modelling on identifying 10-week fistula status. FN, false negative; FP, false positive; TN, true negative; TP, true positive. (a) Observed or imputed as per determination of fistula maturity/non-maturity by week 10 criteria.

The power calculations for the phase 2 RCT based on the phase 1 study findings are detailed in Table 9 and confirm that to show an intervention effect of 8% with early US surveillance, very large numbers of fistulas would be required, raising doubts as to whether this would be achievable. The recommendation from the TMG, endorsed by the TSC, was therefore to assess longer-term outcomes (6- and 12-month patency) in the original study group, on the basis that longer-term patency is more relevant clinically than early maturation and thus more pertinent to the design of the proposed RCT. Additionally, longer follow-up would provide greater insight into the eventual outcomes of those fistulas that remained patent, but immature, at week 10 (8.7% of the original study group; see Table 6); an important distinction in the latter group is whether the fistulas eventually matured spontaneously, or whether they required further intervention or were abandoned.

The longer follow-up study thus addresses the hypothesis that early Doppler US of AVFs predicts 6-month fistula patency failure. As detailed in Follow-up 12-month study, primary fistula patency was selected as the primary outcome analysis, which differs subtly from the primary analysis performed for the original SONAR study: fistula maturation at 10 weeks. We reasoned that if, by 6 months after creation, fistulas had not matured to a functional state, this would have prompted further salvage interventions or creation of a new fistula; hence patency at 6 months would indicate successful maturation. Thus, further US scanning at 6 and 12 months was not proposed.

Twelve-month follow-up

Patient enrolment for the 12-month follow-up study is shown in Figure 8. Of the original 333 SONAR AVFs, 192 (57.7%) were eligible and provided consent for follow-up. Notably, 56 (16.8%) of the original SONAR cohort had died in the first year after fistula creation, and hence the SONAR 12-month cohort (SONAR-12M) does have a survivorship bias, although the participant demographics still closely match the original SONAR cohort (Table 10).

FIGURE 8.

CONSORT diagram for 12-month follow-up study.

TABLE 10 - Baseline characteristics of the SONAR-12M cohort (total number of enrolment forms = 192)

Note

Data are median (interquartile range) for continuous variables and N (%) for categorical variables.

Summary of missing data: IHD/CVA/PVD is missing one observation.

Patients who re-enrolled contribute as many data points as enrolments to this table.

Age (years)	64.5 (50–73)
Male	124 (64.6)
Cause of renal failure
Glomerulonephritis	11 (5.7)
Polycystic kidney disease	9 (4.7)
Hypertension	24 (12.5)
Diabetes	56 (29.2)
Pyelonephritis	0 (0.0)
Renovascular disease	9 (4.7)
Unknown	30 (15.6)
Other	53 (27.6)
Hypertension	160 (83.3)
Diabetes
No	113 (58.9)
Yes: insulin-dependent	48 (25.0)
Yes: non-insulin-dependent	31 (16.1)
IHD/CVA/PVD
No	150 (78.1)
Yes	41 (21.4)

Primary fistula patency at 6 months for all fistulas was 76.6% (95% CI: 69.9% to 82.4%) and was higher at 6 months for elbow fistula than for wrist fistula (elbow: 83.0%, 95% CI 73.8% to 89.9%; wrist: 70.4%, 95% CI 60.3% to 79.2%). This partly reflects the greater rates of early failure observed for wrist fistulas in the original SONAR study, but Kaplan–Meier analysis shows that fistula failure also occurred beyond the original 10-week SONAR study window (Figure 9). Primary fistula patency at 12 months was 68.3% (95% CI 61.2% to 74.5%). Primary patency rates for the SONAR-12M cohort at 6 months are therefore surprisingly greater than the 10-week maturation rates reported for the original SONAR cohort. This at least partly reflects inherent bias associated with the SONAR-12M cohort. For example, at 6 weeks, 46 of the 333 (13.8%) fistulas of the original SONAR cohort had failed, whereas the corresponding figure for the SONAR-12M cohort was 8.3% (16 failures of 192). Additionally, Figure 10 details that small numbers of fistulas that were immature at 10 weeks can still mature to a functional state by 6 months (conversely, some mature fistulas at 10 weeks fail by 6 months).

FIGURE 9.

Kaplan–Meier curve for AVF primary patency of SONAR-12M cohort. Primary patency to 12 months after AVF creation for (a) elbows (76%, 95% CI 66% to 84%) and (b) wrists (61%, 95% CI 50% to 70%). Assisted primary patency to 12 months after AVF creation for (c) elbows (85%, 95% CI 74% to 90%) and (d) wrists (65%, 95% CI 54% to 73%). Secondary patency to 12 months after AVF creation for (e) elbows (90%, 95% CI 82% to 95%) and (f) wrists (69%, 95% CI 59% to 77%).

FIGURE 10.

Fistula status at key study time points.

Surgical or radiological ‘salvage’ procedures were attempted on 43 occasions in the first year after AVF creation to either maintain or restore fistula patency and were reported as part of the 12-month follow-up. Radiological salvage generally consisted of angioplasty of perianastomotic stenotic venous segments, whereas of the surgical interventions, the majority (14/23; 60.9%) involved formation of a proximal neoanastomosis. These interventions were successful in 79.1% of the procedures. Assisted primary and secondary patency rates at 6 months (80.7% and 83.3%) and 12 months (74.1% and 79.5%) were therefore substantially higher than the primary patency rates (see Figure 9). For those patients established on dialysis, functional patency (rates of fistula usage on three successive occasions) was broadly similar (79.2%) to the calculated assisted primary patency rates, and substantially greater than recorded at 10 weeks for the original SONAR cohort (26.6%). Given the favourable maturation rates in the original SONAR cohort, the low rate of functional patency within 10 weeks presumably reflects either concerns regarding needling a fistula too early after creation or that fistulas continue to enlarge and mature beyond week 10.

Predictive model for fistula non-patency at 6 months

In similar fashion to predicting fistula non-maturation at 10 weeks, a binary-data population-average model with exchangeable correlation structure was built to assess the probability of fistula non-patency at 6 months. As with the SONAR modelling, optimum models could be developed using US data from a single week, but separate models for predicting non-patency at 6 months for wrist and elbow fistulas performed superiorly, and relied on a different week’s scan data (Table 11 and Figure 11). Thus, for elbow AVFs, an algorithm that included pre-operative vein diameter, week 4 average resistance index and fistula blood flow predicted 6-month non-patency in seven cases, and correctly so (true positives) in four of these, giving a PPV of 57.1% (95% CI 20.5% to 93.8%). A similar model could be developed for wrist fistulas, based on the week 6 scan data and incorporating fistula blood flow and average resistance index, with additional main effects and an interaction between sex of the participant and fistula venous diameter (see Table 11).

TABLE 11 - Optimum models for predicting primary fistula non-patency at 6 months

A significant p-value was not the only criterion used to select variables for model building. Other criteria, such as the Hosmer et al. delta-beta-hat-per cent measure,35 as well as clinical relevance, were also used.

Week 4 scan data for elbow; week 6 scan data for wrist.

Main effects odds ratio is not presented for the wrist model due this factor’s involvement in an interaction term.

‘Failure’ is defined as abandonment due to failure to mature or due to thrombosis/occlusion, or had an intervention following a thrombosis/occlusion or failure to mature/provide adequate access.

We are interested in true positives that fail at week 6/week 8 or later because, in a phase 2 SONAR trial, a fistula predicted to fail using week 4/week 6 scan results would be randomised to undergo intervention, or not, and arranging a salvage intervention would generally take at least 2 weeks.

Factors included in modela	Elbow (n = 83) odds ratio (95% CI)	Wrist (n = 80) odds ratio (95% CI)
Pre-operative vein diameter (1 unit change from mean)	1.57 (0.91 to 2.72) p = 0.1030	N/A
Average resistance indexb (0.1 unit change from mean)	1.65 (0.61 to 4.46) p = 0.3146	2.59 (1.46 to 4.58) p = 0.0015
Average volume flowb (100 unit change from mean)	0.93 (0.83 to 1.05) p = 0.2471	1.13 (1.06 to 1.20) p = 0.0003
Sexc	NS	p = 0.0067
Representative venous diameterb,c	NS	p < 0.0001
Interaction between sex and representative venous diameterb	N/A	p = 0.0003
1 unit change of representative diameter from mean for males		0.71 (0.53 to 0.95)
1 unit change of representative diameter from mean for females		0.09 (0.03 to 0.26)
Model performance
AUC value	0.71	0.81
Threshold (Youden index)	0.37	0.32
PPV for threshold (95% CI)	57.1% (20.5% to 93.8%)	72.7% (46.4% to 99.0%)
NPV for threshold (95% CI)	88.2% (80.9% to 95.4%)	91.3% (84.7% to 98.0%)
Number of predicted failures vs. actual failures	7 vs. 13	11 vs. 14
Number of correctly predicted failuresd [true positives that fail at week 6 (elbow)/week 8 (wrist) or later]e	4 (3)	8 (8)

FIGURE 11.

Receiver operating characteristic curve analysis of optimum models for predicting wrist and elbow fistula non-patency at 6 months. Receiver operating characteristic curves for optimum models for predicting 6-month fistula non-patency from (a) week 6 US findings for wrist and (b) week 4 US findings for elbow fistulas.

This predicted non-patency in 11 cases, with 8 true positives, giving a PPV of 72.7% (95% CI 46.4% to 99.0%). Diagnostic tests for model fit show that the models perform moderately well, with AUC values on ROC analysis of 0.71 for elbow and 0.81 for wrist (see Figure 11). These values are less than calculated for predicting non-maturation at 10 weeks. As with the modelling for maturation status at 10 weeks, the models for 6-month patency were remarkably effective at identifying those fistulas that would be patent, with NPV of 88.2% (95% CI 80.9% to 95.4%) and 91.3% (95% CI 84.7% to 98.0%) for elbow and wrist fistulas, respectively. The negative and positive predictive powers for modelling 6-month fistula patency are depicted in Figure 12.

FIGURE 12.

Summary of week 4 and week 6 US modelling on identifying 6-month fistula status. FN, false negative; FP, false positive; TN, true negative; TP, true positive. (a) Observed or imputed as per determination of fistula patency/non-patency by 6M criteria.

Patient and public engagement

An initial survey of 29 Cambridge haemodialysis patients informed development of our proposal. All surveyed patients thought it reasonable to have an US scan to assess development of their fistula, with 96.6% indicating that they ‘agree’ or ‘strongly agree’ with the statement that ‘it was acceptable that the scans would be for research purposes and would not affect their care directly’ (one patient was neutral on this point); 58.6% of patients responded that they would be willing to be scanned every fortnight for the first 6 weeks (a further 20.7% were neutral). The panel were, however, unanimous that participants in the first phase should have the opportunity to review the results of the trial, so that they could be reassured that their participation was worthwhile.

The proposed application was further discussed with a patient and public involvement (PPI) panel, consisting of lay members of the public and retired doctors, and provided by the Cambridge University Hospitals PPI team. This panel recommended that US scans in phase 1 should be performed during dialysis sessions to minimise inconvenience, and agreed that blinding during the first phase of the study was important for the study design (one member thought it ‘morally preferable’). The PPI group unanimously supported the two-stage design, as focusing first on the ability of early US to identify AVFs that were unlikely to mature would, they felt, provide ‘better information’ and ‘cleaner data’, and crucially, would provide critical information for justifying progression, or otherwise, to a much larger and costly RCT, thus rationalising the numbers of participants required.

One member of the PPI panel, Andrew Norton, joined the TSC to provide lived experience input as the trial progressed. Andrew shared his experiences as a dialysis, fistula and then transplant recipient with the wider trial group, providing great insight and suggestions to support the success of the research. Andrew also promoted the trial with interviews and articles in local and national press and successfully sought funding from the Addenbrooke’s Kidney Patient Association, for patient travel expenses related to the trial US visits.

The SONAR trial was featured in the National Kidney Federation magazine, Kidney Life in the Spring 2020 issue. This article raised awareness of the research within the kidney patient population, and shared feedback from a trial participant. A further article was published in the Spring 2023 edition, sharing the results of the research.

Acknowledgements

The research team would like to thank the study participants for their support and contributions. We would like to acknowledge the TSC: Professor Patrick Mark (chair); Dr Kate Steiner; Professor Julie Brittenden; Mr Andrew Norton; Mr Laszlo Szabo, and the Independent Data Monitoring Committee: Dr Sian Griffin (Chair); Ms Kerri Barber; and Ms Helen Dixon. In addition, we would like to acknowledge the Addenbrooke’s Kidney Patient Association for their financial support towards patients’ travel expenses related to the study.

Equality, diversity and inclusion

The SONAR and SONAR-12M inclusion and exclusion criteria were kept as simple and succinct as possible, to facilitate wide participation of the adult ESRD population. CKD stage 3–5 prevalence is higher in women than in men, but men are more likely to reach kidney failure sooner. SONAR recruitment was not limited by sex, and 64.5% of our participants were men. People of black or South Asian origin are more likely to develop CKD. We did not collect ethnicity data from participants, but our centres provide care for diverse populations. We therefore cannot assess the ethnic diversity of our participants, and this is something we will collect prospectively for future AVF research projects, along with measures to support the inclusion of any under-served ethnic minority groups.

The trial participant information sheet was provided in English and participating centres made use of NHS Trust-provided translation and communication services where appropriate, to enable informed conversations about the trial with patients and family members, when verbal explanations or written information in English might not be adequately understood. As the participant information sheet was not translated into any languages other than English, this will inevitably have limited inclusion of patients on occasion when translation services were not available or acceptable. In future studies, budgeting for the translation of written information into several suitable languages could improve inclusion. It was also a requirement of the inclusion criteria that patients could provide informed consent, and therefore we did not include patients lacking capacity.

The SONAR Trial Group identified early on in the study design stage that attending for additional US scans would incur travel costs for some patients. In order to reduce the likelihood of this adversely impacting on the ability of any patient to afford to take part, funding for travel costs was identified as a priority. The Addenbrooke’s Kidney Patient Association generously funded patient travel costs for all participants.

Articles and documents for patients and the public were written with minimal technical or medical terminology. Quotes from patients were used to share real-life experiences.

The SONAR research team includes individuals with a range of experience and expertise, and more junior members of the team were actively encouraged to be involved to gain experience and develop their skills. Vascular scientists and sonographers were provided with training sessions to broaden knowledge and skills for fistula scanning. Research fellows and clinical trainees were encouraged to be co-investigators at the participating centres. An apprentice within the Clinical Trials Unit joined the trial team to develop clinical trial administrator experience. All members of the team are acknowledged in publications, to reflect that this trial relied on the skills of a wide range of clinical trial staff, nurses, vascular scientists, doctors and surgeons. This research involved 16 centres in England and 1 in Scotland. The centres provided good geographical coverage, with a mix of large teaching hospitals in cities (e.g. Manchester, London, Coventry, Edinburgh) and medium-sized hospitals.

Contributions of authors

James Richards (https://orcid.org/0000-0001-9049-5745) (Locum Consultant in NORS, HPB and Transplant Surgery) was involved in the conception and design of the study, patient recruitment and data acquisition and drafting of the report.

Dominic Summers (https://orcid.org/0000-0003-3360-0726) (Consultant Transplant and Vascular Access Surgeon) was involved in the conception and design of the study, data acquisition and drafting of the report.

Anna Sidders (https://orcid.org/0000-0003-2687-4710) (Trial Manager) was involved in the conception and design of the study and drafting of the report.

Elisa Allen (https://orcid.org/0000-0001-5220-9015) (Trial Statistician) was involved in the conception and design of the study, analysis and interpretation of the data and drafting of the report.

Mohammed Ayaz Hossain (https://orcid.org/0000-0002-1408-647X) (Consultant Transplant Surgeon) was involved in the conception and design of the study, data acquisition and drafting of the report.

Subhankar Paul (https://orcid.org/0000-0001-7553-0733) (Senior Clinical Fellow in Transplant Surgery and Organ Retrieval) was involved in the conception and design of the study, data acquisition and drafting of the report.

Matthew Slater (https://orcid.org/0000-0002-5773-3657) (Vascular Scientist) was involved in the conception and design of the study, ultrasound data acquisition and drafting of the report.

Matthew Bartlett (https://orcid.org/0000-0003-4898-7468) (Vascular Scientist) was involved in the conception and design of the study, ultrasound data acquisition and drafting of the report.

Regin Lagaac (https://orcid.org/0000-0003-4964-5101) (Renal and Vascular Access Nurse) was involved in the conception and design of the study, data acquisition and drafting of the report.

Emma Laing (https://orcid.org/0000-0002-8309-0990) (Clinical Operations Manager) was involved in the conception and design of the study and drafting of the report.

Valerie Hopkins (https://orcid.org/0000-0002-2167-1650) (Clinical Trial Coordinator) was involved in the conception and design of the study and drafting of the report.

Chloe Fitzpatrick-Creamer (https://orcid.org/0009-0000-8064-6108) (Clinical Trial Administrator) was involved in the conception and design of the study and drafting of the report.

Cara Hudson (https://orcid.org/0000-0002-5831-5959) (Trial Statistician) was involved in the conception and design of the study, analysis and interpretation of the data and drafting of the report.

Joseph Parsons (https://orcid.org/0000-0001-5542-2460) (Trial Statistician) performed statistical analysis and interpretation of the data and drafting of the report.

Samuel Turner (https://orcid.org/0000-0002-4744-0826) (Consultant Renal Transplant and Vascular Access Surgeon) is a Principal Investigator involved in the conception and design of the study, data acquisition and drafting of the report.

Andrew Tambyraja (https://orcid.org/0000-0002-9393-6291) (Consultant Vascular Surgeon) is a Principal Investigator involved in the conception and design of the study, data acquisition and drafting of the report.

Subash Somalanka (https://orcid.org/0000-0002-1547-5127) (Consultant Nephrologist) is a Principal Investigator involved in the conception and design of the study, data acquisition and drafting of the report.

James Hunter (https://orcid.org/0000-0002-4972-6879) (Consultant Renal Transplant and Vascular Access Surgeon) is a Grant Co-applicant and Principal Investigator involved in the conception and design of the study, data acquisition and drafting of the report.

Sam Dutta (https://orcid.org/0009-0003-0993-8142) (Consultant Transplant and General Surgeon) is a Principal Investigator involved in the conception and design of the study, data acquisition and drafting of the report.

Neil Hoye (https://orcid.org/0000-0002-5379-0523) (Consultant Nephrologist) is a Principal Investigator involved in the conception and design of the study, data acquisition and drafting of the report.

Sarah Lawman (https://orcid.org/0009-0000-9865-8799) (Consultant Nephrologist) is a Principal Investigator involved in the conception and design of the study, data acquisition and drafting of the report.

Tracey Salter (https://orcid.org/0000-0003-3259-917X) (Consultant Nephrologist) is a Principal Investigator involved in the conception and design of the study, data acquisition and drafting of the report.

Mohammed Farid Aslam (https://orcid.org/0000-0002-1930-0498) (Clinical Vascular Scientist) is a Principal Investigator involved in the conception and design of the study, data acquisition and drafting of the report.

Atul Bagul (https://orcid.org/0000-0002-8146-1946) (Consultant Transplant and Endocrine Surgeon) is a Principal Investigator involved in the conception and design of the study, data acquisition and drafting of the report.

Rajesh Sivaprakasam (https://orcid.org/0000-0002-6145-4642) (Consultant Transplant Surgeon) is a Principal Investigator involved in the conception and design of the study, data acquisition and drafting of the report.

George E Smith (https://orcid.org/0000-0002-8085-0886) (Honorary Consultant Vascular Surgeon) is a Principal Investigator involved in the conception and design of the study, data acquisition and drafting of the report.

Helen L Thomas (https://orcid.org/0000-0002-7017-7739) (Head of Clinical Trial Statistics) was involved in the conception and design of the study, analysis and interpretation of the data and drafting of the report.

Zia Moinuddin (https://orcid.org/0000-0002-9125-8116) (Consultant Transplant Surgeon) is a Principal Investigator involved in the conception and design of the study, data acquisition and drafting of the report.

Simon R Knight (https://orcid.org/0000-0003-4837-9446) (Honorary Consultant Transplant and Vascular Access Surgeon) is a Grant Co-applicant and Principal Investigator involved in the conception and design of the study, data acquisition and drafting of the report.

Nicholas Barnett (https://orcid.org/0000-0002-7997-7492) (Consultant Transplant and Vascular Access Surgeon) is a Grant Co-applicant and Principal Investigator involved in the conception and design of the study, data acquisition and drafting of the report.

Reza Motallebzadeh (https://orcid.org/0000-0002-5399-9546) (Consultant Renal Transplant Surgeon) is a Grant Co-applicant and Principal Investigator involved in the conception and design of the study, data acquisition and drafting of the report.

Gavin J Pettigrew (https://orcid.org/0000-0003-3724-9945) (Professor of Clinical and Experimental Transplantation) is the Grant Lead Applicant and Chief Investigator responsible for the conception and design of the study, data acquisition, the analysis and interpretation of the data and writing the report.

Disclosure of interests

Full disclosure of interests: Completed ICMJE forms for all authors, including all related interests, are available in the toolkit on the NIHR Journals Library report publication page at https://doi.org/10.3310/YTBT4172.

Primary conflicts of interest: Mr Simon R Knight was a non-paid member of the Health Technology Assessment (HTA) Interventional Procedures (IP) Panel from 1 May 2017 to 31 May 2018. The authors have no other interests to declare.

Data-sharing statement

All data requests should be submitted to the corresponding author for consideration. Access to available anonymised data may be granted following review.

Ethics statement

The SONAR study received a favourable opinion from the Cambridgeshire and Hertfordshire Research Ethics Committee on 26 July 2018, REC Ref: 18/EE/0234. The SONAR-12M study received a favourable opinion from the West Midlands – Edgbaston Research Ethics Committee on 13 January 2021, REC Ref: 20/WM/0331.

Information governance statement

This work uses data provided by patients and collected by the NHS as part of their care and support. Using patient data is vital to improve health and care for everyone. There is huge potential to make better use of information from people’s patient records, to understand more about disease, develop new treatments, monitor safety and plan NHS services. Patient data for the trial have been stored anonymised and within secure NHS servers.

Cambridge University Hospitals NHS Foundation Trust and the University of Cambridge are committed to handling all personal information in line with the UK Data Protection Act (2018) and the General Data Protection Regulation (EU GDPR) 2016/679.

Under the Data Protection legislation, Cambridge University Hospitals NHS Foundation Trust and the University of Cambridge are the Data Controller, and you can find out more about how we handle personal data, including how to exercise your individual rights and the contact details for our Data Protection Officer here https://www.cuh.nhs.uk/patient-privacy/.

Department of Health and Social Care disclaimer

This publication presents independent research commissioned by the National Institute for Health and Care Research (NIHR). The views and opinions expressed by authors in this publication are those of the authors and do not necessarily reflect those of the NHS, the NIHR, MRC, NIHR Coordinating Centre, the HTA programme or the Department of Health and Social Care.

This monograph was published based on current knowledge at the time and date of publication. NIHR is committed to being inclusive and will continually monitor best practice and guidance in relation to terminology and language to ensure that we remain relevant to our stakeholders.

Disclaimers

This article presents independent research funded by the National Institute for Health and Care Research (NIHR). The views and opinions expressed by authors in this publication are those of the authors and do not necessarily reflect those of the NHS, the NIHR, the HTA programme or the Department of Health and Social Care. If there are verbatim quotations included in this publication the views and opinions expressed by the interviewees are those of the interviewees and do not necessarily reflect those of the authors, those of the NHS, the NIHR, the HTA programme or the Department of Health and Social Care.

Image labelling

Acceptable abbreviations are in brackets.

Axillary artery (AX A, Axillary A) – Only if needed in the event of a very high brachial bifurcation

Brachial artery (BA, Brachial A)

Radial artery (RA, Radial A)

Proximal (Prox, P)

Distal (Dist, D)

It is acceptable to use proximal and distal within the context of arteries as normal (e.g. BA prox) but please do not use these in the context of the AVF.

Anastomosis (Anas)

Cephalic vein lower arm (CV LA)

Cephalic vein upper arm (CV UA)

Basilic vein upper arm (BV UA)

If the name of the vein is unknown, then it is acceptable to use the label AVF. Any branches may be labelled AVF Branch or Cephalic vein branch etc.

Any other pathology can be named by its description (e.g. Pseudoaneurysm, Seroma etc.) and its location

For example: Anterior to Radial artery.

After the scan

Complete a blank SONAR AVF ultrasound proforma.

Log in to the electronic Case Report Form (MACRO) and enter the data. Keep completed proforma in the SONAR Ultrasound file provided by the study team.

Ensure images are stored in case of audit, but ensure this is offline and that the clinical team do not have access to the images to ensure they remain blinded.

DO NOT inform the patients of the results (to ensure they do not bias clinicians by passing on the findings).

However, if the AVF is occluded please inform the clinical team at your centre ASAP.