ORIGINAL ARTICLE
J Vasc Access 2014 ; 15 ( 4): 291-297 DOI: 10.5301/jva.5000210
Accuracy of early postoperative clinical and ultrasound examination of arteriovenous fistulae to predict dialysis use Martin Ferring1, John Henderson2, Teun Wilmink3 Department of Renal Medicine, Worcester Acute Hospitals NHS Trust, Worcester - UK Department of Radiology, Heart of England NHS Foundation Trust, Birmingham - UK 3 Department of Vascular Surgery, Heart of England NHS Foundation Trust, Birmingham - UK 1 2
ABSTRACT Aim: The aim of this article is to assess the accuracy of early clinical and ultrasound (US) examination in terms of predicting arteriovenous fistula (AVF) dialysis use. Methods: Physical and US examination of patent AVF was performed 4 weeks after fistula creation. AVF dialysis use was defined as subsequent use of an AVF for at least six consecutive dialysis sessions with two needles and a blood flow of more than 200 mL/min. Results: Of 119 AVF patent at 4 weeks, 26 (22%) failed. Clinical examination was 96% sensitive for predicting successful dialysis, but only 21% specific for failure. Vein diameter above 5 mm and an arterial end-diastolic velocity above 110 cm/s were the best US predictors for dialysis use. Vein diameter was slightly better than arterial velocity in terms of predicting maturity (sensitivity: 83% vs 67%, specificity: 68% vs 65%). All assessments predicted AVF maturity (positive predictive value: clinical = 81%, US diameter = 90%, US velocity = 87%) much better than AVF failure (negative predictive value: clinical = 63%, US diameter = 53%, US velocity = 37%). Conclusion: One month after surgery, a new AVF with a thrill or a vein diameter >5 mm is likely to be used for dialysis. An AVF not meeting these criteria has an increased risk of failure and further investigations may be required. Key words: Arteriovenous fistulae, Clinical examination, Sensitivity, Specificity, Ultrasound Accepted: December 3, 2013
INTRODUCTION Guidelines on vascular access for long-term haemodialysis agree that arteriovenous fistulae (AVFs) are the preferred form of access (1, 2), yet central venous catheters (CVCs) are still in common use. One reason for more CVC use may be a practice of delayed AVF cannulation. Cannulation within 8 weeks after surgery occurred in only 14% of AVF in the United States, in 45% in the United Kingdom and in over 90% of mainland Europe (3). Provided the AVF is mature, cannulation as early as 2 weeks after surgery appears safe (4). On the other hand, early failure is common so that clinicians often face the task of assessing maturing AVF to determine their future usability for haemodialysis (5). Whilst early AVF thrombosis is usually clinically obvious, identifying AVFs that are patent but fail to mature is more challenging. This distinction is crucial early on because AVF that fail to mature can be salvaged if underlying causes are treated (6). Because blood flow through the ma-
turing AVF increases particularly in the first month, it may be difficult to decide whether low blood flow is due to slow AVF maturation or a result of a flow-limiting stenosis (7, 8). Four months after surgery, clinical and ultrasound (US) assessments predict AVF maturity well. Robbin et al found that physical examination was 80% accurate whilst US had an up to 95% positive predictive value (PPV) for combined measurement of vein diameter and blood flow (9). The accuracy of an earlier assessment is unknown. We therefore assessed patent AVF at 4 weeks after surgery with clinical and US examination, to study how accurate physical and US examinations predict AVF adequacy for future dialysis. METHODS Patients and outcome measures Patients belonged to a cohort recruited for the purpose of a randomised trial conducted at a single centre
© 2014 Wichtig Publishing - ISSN 1129-7298
291
Postoperative examination of arteriovenous fistulae to predict dialysis use
and described in detail elsewhere (10). In brief, patients with end-stage renal disease requiring vascular access were randomly allocated to preoperative assessment with physical examination or routine US. Routine contrast venography was performed additionally in all patients with a CVC in place for 3 months or longer. Patients were included if they required first- or second-time formation of a new AVF, whereas new AVF beyond the second time and AVF salvage surgery were excluded. The postoperative assessment of this patient group consisted of physical and US examination. All patients initially recruited were invited to return at 4 weeks after surgery for postoperative assessment of the AVF. Patients were not further assessed if the AVF was thrombosed, or if patients did not attend (free transport was offered). Definition AVF use for dialysis and AVF failure The study outcome measure was AVF dialysis use, which was defined as AVF that could be used for at least six consecutive haemodialysis sessions, with two needle AVF cannulation and a blood flow rate >200 mL/min. AVF failure was present when haemodialysis use was impossible after appropriate time for maturation or if AVF salvage intervention was deemed necessary, but we excluded early thrombosis that was clinically obvious by 2 weeks. AVF outcome was considered unknown if AVF was patent but the patient had not started haemodialysis during the follow-up period: These AVFs were excluded from the analysis. Clinical AVF examination Experienced haemodialysis nurses with known good AVF cannulation skills were asked to perform a physical examination of the AVF. This included the flow character (categorised as thrill–pulse–audible bruit without a palpable thrill), the vein calibre, the straightness and depth of the vein. The nurses were unaware of the US findings. US scan of AVF A trained nephrologist scanned all the AVFs with a portable US (Sonosite 180Plus, 5-10 MHz linear transducer), using a protocol based on a technique described elsewhere (9). In brief, patients were seated in a chair with the AVF arm supported by a pillow at 60 degrees. With greyscale US, an initial overview of the feeding artery, AVF anastomosis and draining vein was obtained. Radiocephalic (forearm) AVFs were followed from the radial artery some 5 cm proximal to the anastomosis, and then along the course of the vein up to the antecubital fossa. Brachiocephalic or brachiobasilic (upper arm) AVFs were followed from some 5 cm proximal to the anastomosis, 292
and then along the course of the vein up to the shoulder. The entry of the cephalic vein into the subclavian vein, the axillary vein or the brachial veins was not studied. The vasculature was not followed with directional colour duplex because the Sonosite 180Plus machine was only supplied with colour power and spectral duplex. Representative cross-sectional vascular lumen diameters as well as velocity measurements using spectral duplex (peak systolic and end-diastolic velocities) were obtained at the following sites: brachial, radial and ulnar arteries (the AVF feeding artery was scanned within 3 cm distance proximal and distal to the anastomosis); AVF anastomosis (diameter in two planes, parallel and perpendicular to feeding artery); curved vein beyond anastomosis; proximal straight vein at likely future cannulation site. Additionally, focal vessel narrowing by more than one-third was noted in terms of diameter and anatomical site, and further interrogated with spectral duplex. Cross-sectional vascular diameters were measured from the near to the far wall interface between vessel wall and lumen. The average of two diameters was used. To avoid venous compression, at least 3 mm layer of US gel was placed between the skin surface and transducer, whilst the transducer was stabilised by resting part of it on the skin surface area to the side of the vein. The velocities were obtained with spectral duplex, with the vessel in longitudinal view at maximum diameter, the Doppler range gate encompassing the entire lumen and with the angle of insonation set at 60 degrees or less. Pulse repetition gain was adjusted to avoid aliasing. The average of two measurements was used. Velocities in the venous parts of the AVF were often too high to be exactly quantified, but arterial measurements could all be obtained. Statistics Continuous variables were explored and transformed if necessary to obtain an approximately normal distribution (age transformed by square/1,000, arterial diameter and velocities by square root, venous diameters by logarithm*1000). Initial analysis was performed with Chi-square tests and student t-tests; multivariate analysis was carried out using logistic regression. Primary AVF failure was the dependent variable, predicted by patient demographics, timing of assessment and assessment findings (clinical and US). Significant variables from the logistic regression were further examined in terms of sensitivity, specificity and receiver operating characteristic (ROC) curves, to decide on appropriate cut-off points; cross-points between sensitivity and specificity plots were taken as initial cut-point but further cut-points near this value were examined to determine the clinically most useful cut-point (11). Empirical normal ranges were calculated as 2.5th to 97.5th centile around median for skewed data (12). Analyses were carried out with SPSS version 16. A p-value of
J Vasc Access 2014 ; 15 ( 4): 291-297 DOI: 10.5301/jva.5000210
Accuracy of early postoperative clinical and ultrasound examination of arteriovenous fistulae to predict dialysis use Martin Ferring1, John Henderson2, Teun Wilmink3 Department of Renal Medicine, Worcester Acute Hospitals NHS Trust, Worcester - UK Department of Radiology, Heart of England NHS Foundation Trust, Birmingham - UK 3 Department of Vascular Surgery, Heart of England NHS Foundation Trust, Birmingham - UK 1 2
ABSTRACT Aim: The aim of this article is to assess the accuracy of early clinical and ultrasound (US) examination in terms of predicting arteriovenous fistula (AVF) dialysis use. Methods: Physical and US examination of patent AVF was performed 4 weeks after fistula creation. AVF dialysis use was defined as subsequent use of an AVF for at least six consecutive dialysis sessions with two needles and a blood flow of more than 200 mL/min. Results: Of 119 AVF patent at 4 weeks, 26 (22%) failed. Clinical examination was 96% sensitive for predicting successful dialysis, but only 21% specific for failure. Vein diameter above 5 mm and an arterial end-diastolic velocity above 110 cm/s were the best US predictors for dialysis use. Vein diameter was slightly better than arterial velocity in terms of predicting maturity (sensitivity: 83% vs 67%, specificity: 68% vs 65%). All assessments predicted AVF maturity (positive predictive value: clinical = 81%, US diameter = 90%, US velocity = 87%) much better than AVF failure (negative predictive value: clinical = 63%, US diameter = 53%, US velocity = 37%). Conclusion: One month after surgery, a new AVF with a thrill or a vein diameter >5 mm is likely to be used for dialysis. An AVF not meeting these criteria has an increased risk of failure and further investigations may be required. Key words: Arteriovenous fistulae, Clinical examination, Sensitivity, Specificity, Ultrasound Accepted: December 3, 2013
INTRODUCTION Guidelines on vascular access for long-term haemodialysis agree that arteriovenous fistulae (AVFs) are the preferred form of access (1, 2), yet central venous catheters (CVCs) are still in common use. One reason for more CVC use may be a practice of delayed AVF cannulation. Cannulation within 8 weeks after surgery occurred in only 14% of AVF in the United States, in 45% in the United Kingdom and in over 90% of mainland Europe (3). Provided the AVF is mature, cannulation as early as 2 weeks after surgery appears safe (4). On the other hand, early failure is common so that clinicians often face the task of assessing maturing AVF to determine their future usability for haemodialysis (5). Whilst early AVF thrombosis is usually clinically obvious, identifying AVFs that are patent but fail to mature is more challenging. This distinction is crucial early on because AVF that fail to mature can be salvaged if underlying causes are treated (6). Because blood flow through the ma-
turing AVF increases particularly in the first month, it may be difficult to decide whether low blood flow is due to slow AVF maturation or a result of a flow-limiting stenosis (7, 8). Four months after surgery, clinical and ultrasound (US) assessments predict AVF maturity well. Robbin et al found that physical examination was 80% accurate whilst US had an up to 95% positive predictive value (PPV) for combined measurement of vein diameter and blood flow (9). The accuracy of an earlier assessment is unknown. We therefore assessed patent AVF at 4 weeks after surgery with clinical and US examination, to study how accurate physical and US examinations predict AVF adequacy for future dialysis. METHODS Patients and outcome measures Patients belonged to a cohort recruited for the purpose of a randomised trial conducted at a single centre
© 2014 Wichtig Publishing - ISSN 1129-7298
291
Postoperative examination of arteriovenous fistulae to predict dialysis use
and described in detail elsewhere (10). In brief, patients with end-stage renal disease requiring vascular access were randomly allocated to preoperative assessment with physical examination or routine US. Routine contrast venography was performed additionally in all patients with a CVC in place for 3 months or longer. Patients were included if they required first- or second-time formation of a new AVF, whereas new AVF beyond the second time and AVF salvage surgery were excluded. The postoperative assessment of this patient group consisted of physical and US examination. All patients initially recruited were invited to return at 4 weeks after surgery for postoperative assessment of the AVF. Patients were not further assessed if the AVF was thrombosed, or if patients did not attend (free transport was offered). Definition AVF use for dialysis and AVF failure The study outcome measure was AVF dialysis use, which was defined as AVF that could be used for at least six consecutive haemodialysis sessions, with two needle AVF cannulation and a blood flow rate >200 mL/min. AVF failure was present when haemodialysis use was impossible after appropriate time for maturation or if AVF salvage intervention was deemed necessary, but we excluded early thrombosis that was clinically obvious by 2 weeks. AVF outcome was considered unknown if AVF was patent but the patient had not started haemodialysis during the follow-up period: These AVFs were excluded from the analysis. Clinical AVF examination Experienced haemodialysis nurses with known good AVF cannulation skills were asked to perform a physical examination of the AVF. This included the flow character (categorised as thrill–pulse–audible bruit without a palpable thrill), the vein calibre, the straightness and depth of the vein. The nurses were unaware of the US findings. US scan of AVF A trained nephrologist scanned all the AVFs with a portable US (Sonosite 180Plus, 5-10 MHz linear transducer), using a protocol based on a technique described elsewhere (9). In brief, patients were seated in a chair with the AVF arm supported by a pillow at 60 degrees. With greyscale US, an initial overview of the feeding artery, AVF anastomosis and draining vein was obtained. Radiocephalic (forearm) AVFs were followed from the radial artery some 5 cm proximal to the anastomosis, and then along the course of the vein up to the antecubital fossa. Brachiocephalic or brachiobasilic (upper arm) AVFs were followed from some 5 cm proximal to the anastomosis, 292
and then along the course of the vein up to the shoulder. The entry of the cephalic vein into the subclavian vein, the axillary vein or the brachial veins was not studied. The vasculature was not followed with directional colour duplex because the Sonosite 180Plus machine was only supplied with colour power and spectral duplex. Representative cross-sectional vascular lumen diameters as well as velocity measurements using spectral duplex (peak systolic and end-diastolic velocities) were obtained at the following sites: brachial, radial and ulnar arteries (the AVF feeding artery was scanned within 3 cm distance proximal and distal to the anastomosis); AVF anastomosis (diameter in two planes, parallel and perpendicular to feeding artery); curved vein beyond anastomosis; proximal straight vein at likely future cannulation site. Additionally, focal vessel narrowing by more than one-third was noted in terms of diameter and anatomical site, and further interrogated with spectral duplex. Cross-sectional vascular diameters were measured from the near to the far wall interface between vessel wall and lumen. The average of two diameters was used. To avoid venous compression, at least 3 mm layer of US gel was placed between the skin surface and transducer, whilst the transducer was stabilised by resting part of it on the skin surface area to the side of the vein. The velocities were obtained with spectral duplex, with the vessel in longitudinal view at maximum diameter, the Doppler range gate encompassing the entire lumen and with the angle of insonation set at 60 degrees or less. Pulse repetition gain was adjusted to avoid aliasing. The average of two measurements was used. Velocities in the venous parts of the AVF were often too high to be exactly quantified, but arterial measurements could all be obtained. Statistics Continuous variables were explored and transformed if necessary to obtain an approximately normal distribution (age transformed by square/1,000, arterial diameter and velocities by square root, venous diameters by logarithm*1000). Initial analysis was performed with Chi-square tests and student t-tests; multivariate analysis was carried out using logistic regression. Primary AVF failure was the dependent variable, predicted by patient demographics, timing of assessment and assessment findings (clinical and US). Significant variables from the logistic regression were further examined in terms of sensitivity, specificity and receiver operating characteristic (ROC) curves, to decide on appropriate cut-off points; cross-points between sensitivity and specificity plots were taken as initial cut-point but further cut-points near this value were examined to determine the clinically most useful cut-point (11). Empirical normal ranges were calculated as 2.5th to 97.5th centile around median for skewed data (12). Analyses were carried out with SPSS version 16. A p-value of
