Volume Blood Flow,Static Pressure Ratio and Venous Conductance in Native Arterio-venous Fistulae: Three Surveillance Methods Compared

Abstract

Purpose

Dialysis venous pressure monitoring has been widely recommended as a surveillance method but has not been shown to improve access outcomes in randomised controlled trials. The method has been impaired by the need to either turn off the blood pump or to derive the static venous pressure from the venous pressure measured with the dialysis pump running. We have developed a unique algorithm which converts Doppler-shifted spectral information derived from unsealed pulsatile blood flow waveforms into an estimate of mean blood pressure (MBP) at the point of ultrasound insonation.

Methods

We have devised the unique expression shown here: MBP = MAP/(1 + Pff/Vff) where MAP is the mean arterial pressure, Pff = (systolic – diastolic)/MAP measured on the contralateral arm and Vff = spectral maximum – minimum/mean. Venous conductance (VC) can be measured by combining this pressure data with Duplex ultrasound blood flow data. A new device BlueDop™ has been used to illustrate the potential clinical value of non-invasive static pressure ratio (SPRn) in a monitoring role. Duplex and BlueDop™ technology were tested in an arterio-venous fistula (AVF) study in which VC, Q and SPRn were compared. Thresholds used for detection of ≥60% venous stenosis were VC <10 mL min⁻¹ mm Hg⁻¹, Q <500 mL min⁻¹, SPRn >0.56.

Results

The following accuracy was achieved: VC = 96%, Q = 92%, SPRn = 76% with similar accuracy in predicting premature thrombosis.

Conclusions

A new algorithm has been described and its in vivo accuracy in estimating mean ‘pressure from flow’ has been confirmed. Two new variables and a new dedicated instrument BlueDop™ have been demonstrated in clinical use.

Keywords

Blood flow Doppler Fistulae Monitoring Pressure Stenosis

Introduction

Too many arterio-venous fistulae (AVFs) fail, often without sufficient warning to permit successful pre-emptive intervention (1). Over the last two decades, there has been little improvement in our ability to identify those fistulae that require intervention and to minimise the incidence of fistula failure by modifying the AVF creation process (2 -11). We feel at the research level there is a need to understand the mechanisms of failure in this creation, which although ‘life saving’ is also, in haemodynamic terms, foreign to the recipient body. A localised thrombotic obstruction may form anywhere between the arterio-venous anastomosis and the subclavian vein in any AVF regardless of flow rate and without any prior warning, and at the everyday clinical level there is the need for reliable early warning of such a lesion. We have devised and tested a new ‘pressure from flow’ algorithm that permits local mean blood pressure (MBP) to be estimated from local Doppler-shifted blood velocity waveforms obtained by ultrasound insonation (12). A proof of theory is shown in Figure 1. This paper reports two distinct issues: (a) Validation of the ‘pressure from flow’ algorithm; (b) demonstration of the potential clinical value of the algorithm by assessing the performance of VC, blood flow Q and static pressure ratio SPRn in predicting two end points: 60% stenosis and thrombosis within 90 days of assessment.

Fig. 1

Simple proof for the ‘pressure from flow’ algorithm which estimates local mean perfusion pressure from central BP and local blood velocity waveforms. a) The vascular system can be simplified into two impedance components, source S and load L representing transit artery and recipient organ, respectively. b) Let resultant velocity waveform be characterised as pulsatile velocity/mean velocity = Vff. c) Let incident pressure waveform be characterised as pulsatile pressure/mean pressure = Pff. d) When S/L = 1.0 hydraulic transmission line theory predicts Pff/Vff = 1.0. This is the matched state demonstrated in claudicants immediately following maximal lower limb exercise (17). e) It can also be shown that Pff/Vff must be numerically equal to S/L when L is infinite or zero. f) By definition mean perfusion pressure MBP = MAP(1 + S/L). g) Therefore by substitution MBP = MAP/(1 + Pff/Vff).

Materials and Methods

Validation of the ‘pressure from flow’ algorithm

Results from three in vivo experiments, two carried out in a canine study and one in a human AVF series, were combined. In the first canine experiment femoral blood pressure was momentarily reduced due to the hyperaemic blood flow increase following the release of an iliac artery total constriction. In the second experiment femoral artery blood pressure was reduced by progressive constriction of the iliac artery (Fig. 2). These two experiments adhered to the requirements of the Animals (Scientific Procedures) Act 1986. In the third experiment (approved by the National Research Ethics Service and with written patient consent), the ‘pressure from flow’ algorithm was used to correlate non-invasive blood pressure with invasive static fistula blood pressure in 14 dialysing patients with pump turned off and after correcting for a nominal right heart mean pressure of 5 mmHg. Informed consent was obtained from all patients prior to study commencement. This last experiment extended the pressure range towards 0 mmHg.

Fig. 2

The ‘pressure from flow’ algorithm was initially tested in vivo by recording direct pressure and Doppler blood flow waveforms in a canine preparation. MAP = central (aortic) mean arterial blood pressure. Pff incident (aortic) pressure form factor = (systolic – diastolic)/mean pressure; Vff resultant (femoral) velocity form factor = (max velocity – min velocity)/mean MBP = mean distal femoral artery blood pressure. Distal femoral artery pressure was progressively reduced (a) during the period of hyperaemia immediately following the release of a short period of total iliac artery constriction and (b) with progressive iliac artery constriction.

The clinical application of the ‘pressure from flow’ algorithm

The study adhered to the provisions of the Declaration of Helsinki and involved native AVFs only. All stenoses have been retrospectively assessed with Duplex Ultrasound using Peak Systolic Velocity Ratio (PSVR). Landwehr et al (13) confirmed the precise relationship between PSVRs and stenosis cross-sectional area, proving area ratio (AR) = 1/PSVR where PSVR = peak stenotic velocity/peak reference velocity. Using an in vitro pulsatile model and precisely machined stenotic inserts the Doppler ultrasound PSVR data agreed precisely with known degrees of stenosis.

The optimum point for accurate Duplex blood flow measurement in brachio-cephalic AVF was fixed at the brachial artery immediately upstream of the antecubital fossa. The same point was used for radio-cephalic AVF because accurate assessment of radio-cephalic blood flow must include the retrograde flow pathway through ulnar artery and palmar arch. In order to assess vascular conductance the non-invasive blood pressure measurement was also located at the brachial artery site. Brachial artery blood flow and MBP measurements were made using either colour Duplex ultrasound incorporating the ‘pressure from flow’ algorithm or Duplex flow and BlueDop™ pressure measurements taken separately. BlueDop™ is a pocket Doppler ultrasound device, wirelessly communicating with a tablet computer running the ‘pressure from flow’ algorithm which has been designed to extract MBP data from raw Doppler-shifted velocity/time spectral waveforms. A reference pressure input from the contralateral arm is required. Instructions on reproduction of these experimental results using standard Duplex ultrasound equipment are included in a downloadable file (Three Surveillance Methods Supplement - available online as supplementary material at www.vascular-access.info). The use of this simple protocol means that forearm arterial conductance is included when assessing radio-cephalic fistulae. A single sphygmomanometer BP measurement was recorded from the contralateral arm. Systemic mean arterial pressure (MAP) was estimated by adding one third of the pulse pressure to the diastolic pressure. Needle-free static pressure ratio SPRn was calculated by dividing MBP by MAP. Blood flow Q was recorded in the brachial artery just above the antecubital fossa with a colour duplex scanner (Zonare Inc, CA, USA). Venous conductance (VC) was calculated by dividing brachial artery blood flow Q by MBP. The ‘gold standard’ test for significant stenosis was provided by Duplex measurements. The axillary and brachial artery and complete superficial venous pathway was scanned either immediately following the test or within 10 days of this measurement by the same operator. The presence of colour ‘aliasing’ with velocity threshold set to maximum was used to locate significant stenoses and these were confirmed with direct measurement of peak stenotic blood velocity normalised to the peak systolic blood velocity immediately upstream. Special attention was paid to obtaining accurate values of PSVR at the arteriovenous anastomosis where there is a problem with rapidly changing blood flow direction from antegrade to retrograde. This can be solved by setting the Duplex angle correct setting to zero degrees, manually adjusting ultrasound probe angulation to maximise peak velocity as shown on the spectral display at the moment that the flow direction coincides with the ultrasound beam axis. This is then divided by peak velocity in the proximal artery, which can be obtained by conventional means; 100 separate determinations of Q, SPR and VC were made. All measurements were recorded in a single database. Scattergrams, receiver operator curves and diagnostic matrices for ‘stenosed’ ≥60% and ‘non-stenosed’ <60% groups were generated for all three parameters (MedCalc, Acacialaan 22, 8400 Ostend, Belgium). The same statistical analysis was carried out on the ‘thrombosed within 90 days’ and ‘patent within 90 days’ group, using identical thresholds chosen for ≥60% stenosis groups.

Results

In vivo validation of the ‘pressure from flow’ algorithm

The algorithm is based upon the ratio between systolic and diastolic values taken from the incident blood pressure waveform and the resultant blood flow waveform. Unlike Computed Flow Dynamics (CFD) technology (14) it is not dependent on vessel dimension or geometry. As such it is theoretically applicable to normal and diseased mammalian arterial networks, even in cases where distal pressure is maintained from a number of collateral pathways. A simple linear relationship is predicted between direct and non-invasive mean pressure with a slope of one and an intercept at the origin such that

M B P = M A P / (1 + P f f / v f f)

Pff measured in the ascending aorta characterises the incident pressure wave. Vff the resultant flow velocity wave was obtained heart beat by heart beat from the instantaneous Doppler-shifted blood velocity recorded from the target blood vessel. The combined canine experiments only dropped peripheral blood pressure to one half central pressure and therefore data from the 14 human AVF experiments have been added in order to extend the pressure range. Figure 3 shows a plot of estimated mean blood pressure (MBP estimated) against catheter mean blood pressure (MBP direct). The best-fit straight line passing through the combined data set yielded a slope close to unity with an intercept passing close to the origin:

Fig. 3

The accuracy of the ‘pressure from flow’ algorithm. A concatenation of three separate ‘in vivo’ experiments shows consistency with theory over almost the complete physiologic pressure range.

MBP estimated = 0.9075 \times MBP direct + 0.9638 with an R^{2} value of 0 . 9202

The clinical application of the ‘pressure from flow’ algorithm

There were 106 native AVFs in our initial patient cohort; 4 AVFs were found to be thrombosed at the initial assessment. Surgical intervention was carried out on one patient before the 90 day study period had elapsed and fistuloplasty was carried out on a second patient prior to study completion, leaving 100 AVFs in the study. Of the 100 access, there were 66 brachio-cephalic and 31 radio-cephalic fistulae, 2 forearm harvested vein grafts and 1 transposed radio-basilic fistula. Thirty-six brachio-cephalic fistulae were virgin AVFs and 20 of these had a sutureless anastomosis in place (15). Fistula creation was carried out primarily by two experienced vascular surgeons EC and YP.

Gender distribution was 66 male, 34 female. Mean age was 63 years (range 18-93). Primary causes of end-stage renal disease were diabetes, renal vascular disease, glomerulonephritis, pyelonephritis, malignancy and nephrotoxicity. Risk factors included hypertension, heart failure, malignancy, myocardial infarct, angina and a previous coronary artery bypass graft.

Detection of significant stenosis within the venous arm of an AVF

Colour Duplex was used retrospectively as an acceptable ‘gold standard’ in order to separate out significant stenosis from non-significant stenosis, permitting statistical analysis of the results for VC, blood flow Q and non-invasive static pressure ratio SPRn. Optimum thresholds were set after analysing the data distribution. These were VC <10 mL min⁻¹ mmHg⁻¹; Q <500 mL min⁻¹ and SPRn >0.56.

Figure 4A-C shows the individual ability of VC, Q and SPRn to identify ‘significant’ and ‘non-significant’ (≥60% and <60%) stenosis. Figure 5A-C shows the associated receiver operator curves and Figure 5D-F shows diagnostic matrices. The ability to discriminate between ‘significant’ and ‘non-significant’ stenosis can be gauged from the proportion of the total area occupied by the area under the ROC (AUC) where 0.5 is agreement based on chance and 1.0 is perfect agreement. The ROC statistics were

Fig. 4

Detecting significant AVF stenosis: three methods. A, B, C are scattergrams illustrating the differences between <60% and ≥60% stenosis for VC, blood flow Q and SPRn.

Fig. 5

A, B, C are receiver operator curves illustrating the power to differentiate between <60% and ≥60% stenosis groups for VC, Blood flow Q and SPRn. D, E, F are diagnostic matrices showing true positive, true negative, false positive and false negative data for VC, blood flow Q and SPRn.

VCauc = 0.993, SE = 0.00647, p<0.0001, Qauc = 0.980, SE = 0.0130, p<0.0001 and SPRauc = 0.872, SE = 0.0538, p<0.0001

The individual sensitivity, specificity and accuracy of all three parameters in the detection of ≥60% stenosis in the venous arm of the fistula were

VC sensitivity, specificity, accuracy = 100%, 96%, 96%

Q sensitivity, specificity, accuracy = 100%, 91%, 92%

SPRn sensitivity, specificity, accuracy = 75%, 76%, 76%

Predicting thrombosis within 90 days of 60% stenosis detection

The status of all 100 AVFs remaining in the cohort was followed for up to 90 days following the initial test assessment. Evidence of fistula failure was usually based on indications obtained by the attending clinician or specialist nursing staff during the dialysis procedure. Thrombosis was then confirmed with a Duplex ultrasound scan. Because there was only a marginal change in optimal threshold values indicated by the ‘thrombosis’ data the same ‘stenosis’ thresholds were used to predict AVF thrombosis within 90 days.

Figure 6A-C shows the ability of VC, Q and SPRn to predict AVF thrombosis within 90 days. Figure 7A-C shows the associated ROC and Figure 7D-F shows diagnostic matrices. The ROC statistics for thrombosis prediction were

Fig. 6

AVF thrombosis within a 90-day window. A, B, C are scattergrams illustrating the degree of separation between non-thrombosed and thrombosed for VC, blood flow Q and SPRn.

Fig. 7

A, B, C are receiver operator curves (ROC) illustrating the power to differentiate between AVF that fail within 90 days of the measurement and those that remain patent at 90 days ROC for VC, blood flow Q and SPRn. D, E, F are diagnostic matrices showing true positive, true negative, false positive and false negative data for VC, blood flow Q and SPRn.

VCauc = 0.967, SE = 0.0292, p<0.0001, Qauc = 0.959, SE = 0.0216, p<0.0001 and SPRauc = 0.874, SE = 0.0427, p<0.0001

The diagnostic matrices generated from the raw data summarise the performance of all three parameters in predicting AVF thrombosis within a period of 90 days following the initial breach of the pre-set threshold levels established for ≥60% stenosis. The results were

VC sensitivity, specificity, accuracy = 90%, 97%, 96%

Q sensitivity, specificity, accuracy = 90%, 92%, 92%

SPRn sensitivity, specificity, accuracy = 80%, 78%, 78%

Thrombosis incidence

A total of 10 out of the 100 AVFs (10%) in the study cohort were patent on initial examination but thrombosed during the 90 day follow-up period. Average time to thrombose following breach of any one of the pre-determined thresholds was 37 days (range 7-85 days).

Discussion

In terms of the quality and quantity of information supplied VC is clearly superior as a research tool, not only reliably detecting stenoses but also permitting maturation or deterioration of artery and vein to be studied in isolation. Neither VC nor Q is ideal for regular clinical use on the total patient throughput. SPRn used alone, although less accurate, can be obtained with BlueDop™ in a rapid ‘point and click’ measurement routine. Unlike conventional SPR it can detect any stenoses existent within the arterio-venous anastomosis or situated between the arterio-venous anastomosis and either venous needle without turning the pump off and with the patient on- or offline. Standard clinical indications for intervention or surgical repair were applied in this study and there were only two interventions made, one a surgical repair and the other an angioplasty which then went on to thrombose, both of which were excluded from the study. Apart from the requirement for a functioning fistula no other selection criteria were applied. Time to thrombose after the critical stenosis stage is reached is variable. Haemodynamic models indicating artery/vein diameter ratio is a key factor in determining the rate of AVF deterioration (16). BlueDop™ has been developed to allow SPRn to be estimated at any chosen position within an AVF. A hierarchy of options is now apparent. Colour Duplex ultrasound is an excellent investigational tool because of its ability to locate and grade fistula stenosis and measure blood flow. Use of the ‘pressure from flow’ algorithm in conjunction with colour Duplex data permits measurement of venous and arterial conductance either ‘on-’ or ‘off’ line. This yields valuable research data on fistula maturation rates and can identify the dominant ‘flow-limiting’ component of the AVF, be it arterial or venous. Such assessments are the domain of the vascular technologists and are not necessarily applicable to everyday nursing care whereas SPRn, measured quickly and simply with a hand-held device, may prove to be the more practical solution to improving AVF longevity.

Footnotes

Acknowledgements

The authors thank Professor Berthold Lausen, Department of Mathematical Sciences, University of Essex, for helpful statistical advice provided. Advice and permissions given by the Essex 2 Research Ethics Committee were vital in obtaining direct needle pressure correlations in the venous pressure range. Thanks to Mike Cater, Pump Technician, Renal Unit, Broomfield Hospital, for devising an ingenious correction of needle height used in our in vivo static pressure series. The willing cooperation of the patients in both renal centres, especially those who volunteered to take part in the static pressure series at Broomfield Hospital, was invaluable. Specialist access staff in both centres helped in acquiring full demographic and event timing data from the access databases. A special thanks to Becki Culpin at Broomfield Hospital and Liz Anderson and Hazel Malete at St Georges Hospital.

Conflict of interest: DHK and MAQ hold a provisional patent on the ‘pressure from flow’ algorithm used by BlueDop™ (Haemodynamic Data Estimation: Patent Application WO 2010/103277; International filing date 11 March 2010; published online at ). DHK and MGT designed the BlueDop™ device for Lateral Services Ltd of which DHK is a Director.

References

Kumbar

, Karim

, Besarab

Surveillance and monitoring of dialysis access. Int J Nephrol. 2012; 2012: Article id 649735.

Dember

L.M.

, Holmberg

E.F.

, Kaufman

J.S.

Value of static venous pressure for predicting arteriovenous graft thrombosis. Kidney Int. 2002; 61 (5): 1899–1904.

McCarley

, Wingard

R.L.

, Shyr

, Pettus

, Hakim

R.M.

, Ikizler

T.A.

Vascular access blood flow monitoring reduces access morbidity and costs. Kidney Int. 2001; 60 (3): 1164–1172.

Schwab

S.J.

, Oliver

M.J.

, Suhocki

, McCann

Hemodialysis arteriovenous access: detection of stenosis and response to treatment by vascular access blood flow. Kidney Int. 2001; 59 (1): 358–362.

Besarab

, Sullivan

K.L.

, Ross

R.P.

, Moritz

M.J.

Utility of intra-access pressure monitoring in detecting and correcting venous outlet stenoses prior to thrombosis. Kidney Int. 1995; 47 (5): 1364–1373.

Tessitore

, Bedogna

, Gammaro

Diagnostic accuracy of ultrasound dilution access blood flow measurement in detecting stenosis and predicting thrombosis in native forearm arteriovenous fistulae for hemodialysis. Am J Kidney Dis. 2003; 42 (2): 331–341.

Sullivan

K.L.

, Besarab

, Bonn

, Shapiro

M.J.

, Gardiner

G.A.

Jr. , Moritz

M.J.

Hemodynamics of failing dialysis grafts. Radiology. 1993; 186 (3): 867–872.

May

R.E.

, Himmelfarb

, Yenicesu

Predictive measures of vascular access thrombosis: a prospective study. Kidney Int. 1997; 52 (6): 1656–1662.

Paulson

, Moist

, Lok

Vascular access surveillance: a case study of a false paradigm. Semin Dial. 2013; 26 (3): 281–286.

10.

Moist

L.M.

, Churchill

D.N.

, House

A.A.

Regular monitoring of access flow compared with monitoring of venous pressure fails to improve graft survival. J Am Soc Nephrol. 2003; 14 (10): 2645–2653.

11.

Asif

, Besarab

, Gadalean

Utility of static pressure ratio recording during angioplasty of arteriovenous graft stenosis. Semin Dial. 2006; 19 (6): 551–556.

12.

King

, Taylor

, Al-Qaisi

As easy as A, B, C: pressure gradients in native arterio-venous fistulae: The Society for Vascular Technology of Great Britain & Ireland, 20th Annual Scientific and General Meeting, Edinburgh, United Kingdom, 23 Nov 2011.

13.

Landwehr

, Schindler

, Heinrich

, Dölken

, Krahe

, Lackner

Quantification of vascular stenosis with color Doppler flow imaging: in vitro investigations. Radiology. 1991; 178 (3): 701–704.

14.

Mehrabi

, Setayeshi

Computational fluid dynamics analysis of pulsatile blood flow behaviour in modelled stenosed vessels with different severities. Math Probl Eng. 2012; 2012: Article id 804765.

15.

Roberto

, Ebner

, Gallo

Arteriovenous fistula creation using the Optiflow vascular anastomosis device: a first in man pilot study. Semin Dial. 2013; 26 (1): 97–99.

16.

White

J.J.

, Ram

S.J.

, Jones

S.A.

, Schwab

S.J.

, Paulson

W.D.

Influence of luminal diameters on flow surveillance of hemodialysis grafts: insights from a mathematical model. Clin J Am Soc Nephrol. 2006; 1 (5): 972–978.

17.

King

, Bojanowski

, DiGiovanna

, Kontis

Post exercise Doppler blood velocity patterns in health and disease. In: Jordan

, Perkins

eds. Proceedings of the Biological Engineering Society, 6th Nordic Meeting Aberdeen, 22-25 July 1984.