Stenosis remains a frequent and troublesome complication even in well-functioning, mature arteriovenous fistulae (AVF) (1). Stenotic lesions can occur anywhere in the access circuit, from the feeding artery down to the central venous system. It has been shown, however, that the site of stenosis varies according to the type of AVF, the most common location being the juxta-anastomotic region in the case of forearm radiocephalic AVFs, and the outflow region for upper-arm brachiocephalic and brachiobasilic AVFs (2–5). It has also been reported that the access blood flow rate (Qa), a measurement widely accepted as a means for detecting access dysfunction, also varies according to the site of the AVF, being lower in the more distal accesses than in the more proximal accesses (6,7). Some researchers have found that the Qa threshold indicative of stenosis is lower in the wrist than in mid-forearm AVFs (6), and others have suggested that the Qa threshold warranting imaging and intervention should depend on access site, a Qa limit of 400 ml/min being acceptable in the forearm as opposed to 600 ml/min in the upper arm AVFs (8). The tools and criteria to use in screening for stenoses may therefore conceivably vary according to the site of the anastomosis, but no studies have systematically addressed this issue to date.
The aim of our study was consequently to establish whether screening for significant (>50%) stenosis can be tailored to the site of the access. We used the database of a prospective blinded trial to seek an optimal bedside screening program for mature AVF stenosis (8a).
Materials and Methods
This prospective blinded study was conducted on an unselected population of 119 consecutive patients with mature AVFs attending the hemodialysis unit at the Ospedale Borgo Roma in Verona, Italy, between December 2006 and May 2010.
On the basis of the finding that among 48 nonstenotic AVFs, the Qa values were significantly lower in the AVFs with an arteriovenous anastomosis located in the lower third of the forearm (n = 18) than in those in the mid-forearm (n = 19) or the elbow/upper-arm region (n = 11) (median Qa 1047 [653–1654] versus 1344 [813–1625] and 1396 [876–2770] ml/min, P < 0.03), whereas there were no significant differences in Qa measured at the latter two sites (P = 0.263), AVFs were arbitrarily collapsed into two groups as distal (anastomoses located in the lower forearm) or proximal (anastomoses located more proximally, in the mid-forearm or elbow/upper-arm region). All of the subjects gave their informed consent to the study protocol, which was conducted according to the principles of the Helsinki Declaration and approved by the local ethical committee (Progretto No. 1330).
All of the AVFs were tested in random order over a 2-week period using the following screening methods, as described elsewhere (8a): physical examination (PE) according to the model described by Beathard (9,10); the dynamic venous pressures measured at a dialysis blood pump flow (Qb) of both 200 ml/min (VP200) and the prescribed Qb of 300 to 350 ml/min (VP300); the derived static venous-pressure ratio (VAPR) computed according to Frinak et al. (11); the ratio of the Qb to the negative dialysis arterial prepump pressure (AP) at the prescribed Qb of 300 to 350 ml/min (Qb300/AP); and the Qa measured using the ultrasound-dilution technique with the HD03 monitor (Transonic Inc., Ithaca, NY).
Venous and arterial pressures and Qa measurements could not be obtained in three patients treated using single-needle dialysis. Qa measurements were also unobtainable in another eight patients because the arterial and venous needles had to be placed in two different, noncommunicating branches because of early venous collaterals immediately beyond the anastomosis. After performing the above-mentioned tests, biplanar fistulography was performed in all AVFs to evaluate the access from the feeding artery to the right atrium, as explained elsewhere (8a). The presence of a significant stenosis (>50% reduction of the vessel's diameter compared with the adjacent segment) was ascertained and quantified by the same radiologist, who was unaware of the results of the other tests.
The data are given as percentages, means ± SD or medians (95% confidence interval [CI]). Receiver operating characteristics (ROC) curve analyses were used to determine each test's overall screening accuracy, as measured by the area under the curve (AUC), and to identify optimal cut-offs for continuous variables (12). Differences in diagnostic performance between the tests and their optimal threshold(s) were assessed by comparing their sensitivity, specificity, accuracy, positive predictive value (PPV), and negative predictive value (NPV), including their 95% CIs (13), and correcting for the fact that not all of the tests could be applied to some AVFs. When a test could not be used, a random allocation was provided by coin tossing. When a test could not be used, an AVF in which imaging had revealed a stenosis was randomly defined as a true positive (TP, i.e. with a positive test result) or a false negative (FN, i.e. with a negative test result). Vice versa, if imaging revealed no stenosis, an AVF was randomly considered as a true negative (TN, i.e. with a negative test result) or a false positive (FP, i.e. with a positive test result). We also calculated the diagnostic performance of two tests in association, according to Macaskill et al. (14). When one of the two tests could not be used, the diagnosis was based on the outcome of the other test alone.
Cohen's κ value was also used to measure the level of agreement beyond chance between the angiographic results and the different tests (15). Differences between groups were tested using t test or the Wilcoxon test, as appropriate. Analyses were carried out with the SPSS rel. 17.0 (SPSS Inc., Chicago, IL). Differences between ROC curves were tested with the DeLong test and differences between binomial data with the McNemar test. The results were considered significant when P was ≤0.05 (two-tailed).
The characteristics of the patients and their AVFs are given in Table 1. The arteriovenous anastomosis was located in the distal third of the forearm (between the radial artery and the cephalic vein) in 43 AVFs (distal AVFs [dAVF]). It was in the mid-forearm, between the radial artery and the cephalic or median vein, in 39 AVFs, and at the elbow in 37 (in the forearm or upper arm, between the radial artery and the cephalic or median vein in 16, between the brachial artery and the cephalic vein in 11, and between the brachial artery and the basilic vein in 10), and these accesses were identified as proximal AVFs (pAVF). Imaging identified 59 AVFs with significant stenoses (49.6%). The stenosis was located on the inflow side (i.e. in the juxta-anastomotic or the cannulation area [STin]) in 43 of 59 AVFs, on the outflow side (i.e. downstream from the cannulation area [STout]) in 12, and at both sites in four. Table 2 shows the relationship between the sites of anastomosis and stenosis, indicating that the prevalence of STin was independent of the anastomotic site, whereas STout was more prevalent in mid-forearm and elbow AVFs (with no difference between the two). Overall, the prevalence of STin was 42% in dAVF and 38% in pAVF (P = NS), whereas the prevalence of STout was 2% in dAVF and 20% in pAVF (P = 0.009).
Diagnostic Performance of the Tests for Stenosis in Distal AVF
Figure 1 and Table 3 show the diagnostic performance of the tests for STin. ROC curve analysis showed that the only tests with a discriminatory capacity were Qa and PE, the AUC being significantly higher for Qa than for PE (P = 0.030).
The best accuracy (88%) was observed for Qa <650 ml/min, which was significantly higher than for PE (P = 0.039) but not significantly different in statistical terms from Qa <900 ml/min (P = 0.070). The sensitivity and NPV of Qa <650 ml/min were also not significantly different from Qa <900 ml/min and PE; its specificity and PPV were significantly higher than Qa <900 ml/min (P = 0.016), however, but no different from PE. The positive PE or Qa <650 ml/min combination led to an insignificant increase in sensitivity and NPV (94%), at the expense of a significant drop in specificity (P = 0.031); the PE and Qa <650 ml/min combination gave rise to an insignificant increase in specificity and PPV and a drop in sensitivity approaching but not reaching statistical significance (P = 0.063).
Diagnostic Performance of the Tests for Stenoses in Proximal AVFs
Figure 2 and Table 4 show the diagnostic performance of the tests for STin. At ROC curve analysis, the only tests with a discriminatory capacity were Qa, PE, and Qb300/AP. Whatever the threshold used, none of the single tests revealed a clinically useful performance, nor did they differ significantly from one another. The positive PE and Qa <900 ml/min combination led to an insignificant increase in accuracy (85%) caused by a better specificity and PPV (which was significantly better than with Qa <900 ml/min [P = 0.016] but not with a positive PE [P = 0.063]), without unduly reducing the sensitivity and NPV. The positive PE and Qa <650 ml/min combination also brought the specificity up to a level similar to the positive PE and Qa <900 ml/min combination (96% versus 91%), but it also reduced their sensitivity by comparison with the tests considered alone (P = 0.022), whereas a positive PE and Qa <900 ml/min did not. Combining a positive PE or Qa <900 ml/min produced an insignificant increase in accuracy (79%), because of a higher sensitivity (which was significant by comparison with a positive PE and Qa <650 ml/min [P = 0.016], but not when compared with Qa <900 ml/min [P = 0.070]), with no change in specificity or PPV by comparison with the separate tests. The Qa <900 ml/min or VAPR >0.5 combination produced an accuracy of 64%, which was significantly lower than the positive PE and Qa <900 ml/min combination (P = 0.001), because of its lower specificity (55% versus 91%, P < 0.001).
Figure 3 and Table 5 show the diagnostic performance of the tests for STout. At ROC curve analysis, all of the tests revealed a discriminatory capacity except for Qa. Among the continuous variables, only a VAPR >0.50 reached a clinically useful accuracy, however, thanks to the combination of an excellent sensitivity and specificity, and its diagnostic performance resembled that of PE. Moreover, associating the two tests did not substantially improve the diagnostic performance of the tests when they were considered separately, except for a significant increase in specificity after combining a positive PE and VAPR >0.50, by comparison with a VAPR >0.50 alone (P = 0.016) but not with a positive PE (P = 0.063). The same applied to the Qa <650 ml/min or VAPR >0.50 combination, which showed a diagnostic performance similar to a positive PE and VAPR >0.5 alone, but it was less accurate when a positive PE and VAPR >0.5 were combined (74% versus 91%, P = 0.001).
Comparisons between Subgroups (Proximal and Distal AVFs) and the Population as a Whole
Tailoring the choice of test to the site of the anastomosis led to an insignificant gain in accuracy for inflow stenosis (88% for Qa <650 ml/min in distal and 85% for the positive PE and Qa <900 ml/min combination in proximal AVFs) by comparison with the best screening test regardless of the site of the anastomosis, i.e. the positive PE or Qa <650 ml/min combination, which showed an accuracy of 81% (73 to 88). The best accuracy for outflow stenosis (91%) was seen for the positive PE and VAPR >0.5 combination, which was identical to the 92% (86 to 96) observed for the same combination in the whole AVF population.
On the other hand, tailoring the screening method to the anastomotic site significantly reduced the need to perform PE (64% [76 of 119 proximal AVFs] versus 100%, P < 0.001), because distal AVFs can be screened effectively by measuring Qa alone. Moreover, if one favored sensitivity, it also significantly reduced the need to measure VAPR (23% [27 of 119 proximal AVFs] versus 55% [66 of 119], P = 0.004). Tailoring screening or not made no significant difference to the need to measure Qa, however, whether one favored sensitivity (59% [70 of 119 AVFs] versus 55% [66 of 119 AVFs], P = NS) or specificity (62% [74 of 19 AVFs] versus 45% [54 of 119 AVFs], P = NS); the same was true of the need to measure VAPR (15% [18 of 119 proximal AVFs] versus 16% [19 of 119 AVFs], P = NS) if one favored specificity.
Our study confirms that the site of stenoses varies with the site of the AVF arteriovenous anastomosis: in distal radiocephalic AVFs, virtually all stenoses were found in the inflow region, whereas outflow lesions were found almost exclusively in mid-forearm and elbow/upper-arm AVFs, as reported elsewhere (2,4). Our data also confirm that Qa (which some consider the best tool for detecting AVF dysfunction) also depends on the site of the anastomosis, higher values being observed in more proximally located accesses (6,7). It would therefore make sense to tailor stenosis screening programs on the basis of an objective and easily-identifiable criterion, i.e. the site of the arteriovenous anastomosis.
In wrist AVFs, the only test substantially consistent with fistulography for inflow stenoses was a Qa <650 ml/min, which proved significantly more accurate than PE and significantly more specific than Qa <900 ml/min (for the same sensitivity). Combining PE and Qa <650 ml/min (in a scenario where PE is used as the initial screening tool and Qa is measured in AVFs with a negative PE) led to an insignificant gain in sensitivity (94% versus 83%) and a significant loss of specificity (68% versus 92%, P = 0.031) by comparison with Qa <650 ml/min. These data suggest that distal forearm AVFs can be screened effectively for inflow stenoses by measuring Qa alone, considering a threshold of Qa <650 ml/min.
The detection of inflow stenoses in proximal AVFs is a different story. For a start, the only tests showing a moderate agreement with fistulography and an adequate accuracy were PE and Qa at thresholds varying from <900 to <650 ml/min; higher Qa thresholds reached a higher sensitivity (76%), whereas lower ones reached a higher specificity (87%). The diagnostic performance can be improved by combining PE and Qa. The scenario of a positive PE and a Qa <900 ml/min achieved a slightly better accuracy (up to 85%), showing equally good NPV and PPV (85%) and a significantly better specificity (91%) than for each test alone, although the sensitivity of this combination remained suboptimal (73%). The alternative scenario of a positive PE or Qa <900 ml/min significantly improved the tests' sensitivity and NPV (97%) by comparison with PE but gave rise to an excessively low specificity (68%) and PPV (65%). These data support the conviction that PE and Qa have a key role in any screening program for inflow stenoses in proximal AVFs (16,17) and suggest that the most effective strategy is to use PE first, followed by Qa measurement in accesses with a negative PE (if priority is given to sensitivity, even if this means that approximately one in three angiograms will be performed unnecessarily) or in accesses with a positive PE (if priority goes to specificity, which carries the risk of approximately 30% of stenotic AVFs being overlooked). The performance of the “or” combinations of a positive PE with Qa <900 ml/min or Qa <650 ml/min were much the same, whereas the combination of a positive PE and Qa <900 ml/min reached a significantly higher sensitivity than the “and” combination with a Qa <650 ml/min, with a comparable specificity. Our results therefore support the notion that the most effective Qa threshold depends on the site of the anastomosis (6) and suggests that lower Qa values should be used for more distally located AVFs, as already suggested by others (8). The Qa thresholds identified in our study are higher than those reported by Lopot et al. (8), however, the difference probably being due largely to the different focus of the two studies (ours concentrated on detecting stenosis, whereas Lopot et al. aimed to identify AVFs at risk of failure and requiring intervention).
Several tests showed a discriminatory power for outflow stenoses in proximal AVFs, ranging from PE to dialysis dynamic and static venous pressures and arterial pressure measurements (but not Qa), but the only tests with a clinically useful cut-off proving more than 80% accurate were PE and VAPR at a threshold of >0.5. This cut-off is similar to the one reported for detecting stenoses in grafts (11), supporting the notion that proximal AVFs with outflow stenoses behave hemodynamically like grafts (1). PE and VAPR >0.5 both had an equally excellent diagnostic profile, although PE can be considered as the screening procedure of choice because it can be used for all AVFs, and it is more straightforward to perform and less variable than VAPR (8a). On the other hand, adding VAPR measurement to PE may be useful because it affords a slight (although not statistically significant) improvement in sensitivity and NPV in the AVFs with a negative PE (93% versus 73% and 98% versus 93%, respectively) and in accuracy, specificity and PPV in the AVFs with a positive PE (92% versus 86%, 97% versus 89%, and 77% versus 63%, respectively) by comparison with PE.
Our study also showed that screening on the basis of Qa and VAPR afforded an inferior diagnostic performance to screening programs that included PE: as an example, in proximal AVFs the Qa <900 ml/min or VAPR >0.5 combination was significantly less accurate and specific than the positive PE and Qa <900 ml/min combination for inflow stenosis (64% versus 85% and 55% versus 91%, respectively, P < 0.002), and the Qa <650 ml/min or VAPR >0.5 combination was significantly less accurate and specific than the positive PE and VAPR >0.5 combination for outflow stenosis (74% versus 91% and 70% versus 97%, respectively, P < 0.001).
Finally, we found that tailoring the choice of test to the site of the anastomosis may have some advantage over screening regardless of the AVF's location in that it contains the screening-associated workload (as shown by the significant reduction in the need to perform PE [64% versus 100%, P < 0.001], and measure VAPR [23% versus 58%, P = 0.004] if priority goes to sensitivity) and improves the diagnostic accuracy for inflow stenosis from an already good 81% to an excellent 88% in distal and 85% in proximal AVFs (even though these differences did not prove statistically significant).
We are aware that our study has some limitations. It is a single-center study on a small group of patients and may be underpowered for the detection of some differences between the various tests. The arbitrary criteria we adopted to group AVFs as distal or proximal may be open to criticism, although our pooling mid-forearm and elbow AVFs in the same group was prompted by their similar hemodynamic and clinical characteristics, such as the Qa values, and the prevalence of the sites of stenosis. Randomly assigning cases in which a test cannot be performed may be a concern because it may skew the results. This concern does not apply to our study, however, because the results obtained in the subgroup of the 108 AVFs in which all tests could be performed did not differ significantly from those obtained in the population as a whole (data not shown). Our approach may also have limitations associated with subgroup analysis, although our study shows that such an analysis (on the basis of easily-identifiable and objective access characteristics) may be useful because it offers insight on how to optimize stenosis screening programs by showing that the wrist and more proximally-located AVFs should be screened using different criteria (Qa <650 ml/min in the former and PE followed by Qa <900 ml/min or VAPR >0.5 in the latter), and tailoring the choice of the screening procedures to the site of the arteriovenous anastomosis leads to a significant reduction in access surveillance-related workload, plus a statistically-insignificant improvement in the diagnostic performance for inflow stenosis by comparison with screening regardless of the anastomotic site.
In conclusion, our study shows that an effective screening program for detecting and locating AVF stenoses with an accuracy of 85% and more can be implemented at the bedside during dialysis by tailoring the choice of screening procedure to the site of the arteriovenous anastomosis, Qa <650 ml/min being the best test in wrist AVFs, and PE (followed by Qa measurement at a threshold of <900 ml/min for inflow stenosis and VAPR at a threshold of >0.5 for outflow stenosis) in more proximally located accesses. Moreover, it also suggests that tailoring the choice of test to the site of the access may reduce the burden of screening by comparison with screening regardless of the access location.
Published online ahead of print. Publication date available at www.cjasn.org.
1. National Kidney Foundation K/DOQI Clinical practice guidelines for vascular access 2006. Am J Kidney Dis 48[suppl S1]: S176–S273, 2006
2. Turmel-Rodrigues L, Pengloan J, Baudin S, Testou D, Abaza M, Dahdah G, Mouton A, Blanchard D: Treatment of stenosis and thrombosis in hemodialysis fistulas and grafts by interventional radiology. Nephrol Dial Transplant 15: 2029–2036, 2000
3. Rajan DK, Clark TWL, Patel NK, Stavropoulos SW, Simons ME: Prevalence and treatment of cephalic arch stenosis in dysfunctional autogenous hemodialysis fistulas. J Vasc Interv Radiol 14: 567–573, 2003
4. Rajan DK, Bunston S, Misra S, Pinto R, Lok CE: Dysfunctional autogenous hemodialysis fistulas: Outcomes after angioplasty: Are there clinical predictors of patency? Radiology 232: 508–515, 2004
5. Asif A, Gadalean FN, Merrill D, Chela G, Cipleu CD, Epstein D, Roth D: Inflow stenosis in arteriovenous fistulas and grafts: A multicenter, prospective study. Kidney Int 67: 1986–1992, 2005
6. Tessitore N, Bedogna V, Gammaro L, Lipari G, Poli A, Baggio E, Firpo M, Morana G, Mansueto G, Maschio G: 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 42: 331–341, 2003
7. Tonelli M, Hirsch DJ, Chan CT, Marryatt J, Mossop P, Wile C, Jindal K: Factors associated with access blood flow in native vessel arteriovenous fistulae. Nephrol Dial Transplant 19: 2559–2563, 2004
8. Lopot F, Nejedly B, Valek M: Vascular access monitoring: methods and procedures: Something to standardize? Blood Purification 23: 36–44, 2005
8a. Tessitore N, Bedogna V, Melilli E, Millardi D, Mansueto G, Lipari G, Mantovani W, Baggio E, Poli A, Lupo A: In search of an optimal bedside screening program for arteriovenous fistula stenosis. Clin J Am Soc Nephrol 2011, in press
9. Beathard GA: Physical examination: The forgotten tool. In: A Multidisciplinary Approach for Hemodialysis Access, edited by Gray R, Sands J New York, Lippincott Williams & Wilkins, 2002, pp 111–118
10. Beathard GA: An algorithm for the physical examination of early fistula failure. Semin Dial 18: 331–335, 2005
11. Frinak S, Zasuwa G, Dunfee T, Besarab A, Yee J: Dynamic venous access pressure ratio test for hemodialysis access monitoring. Am J Kidney Dis 40: 760–768, 2002
12. Tripepi G, Jager KJ, Dekker FW, Zoccali C: Diagnostic methods 2: Receiver operating characteristic (ROC) curve. Kidney Int 76: 252–256, 2009
13. van Stralen KJ, Stel VS, Reitsma JB, Dekker FW, Zoccali C, Jager KJ: Diagnostic methods I: Sensitivity, specificity, and other measures of accuracy. Kidney Int 75: 1257–1263, 2009
14. Macaskill P, Walter SD, Irwig L, Franco EL: Assessing the gain in diagnostic performance when combining two diagnostic tests. Stat Med 21: 2527–2546, 2002
15. McGinn T, Wyers PC, Newman T, Keitz S, Leipzig R, Guyatt G: Tips for learners of evidence-based medicine: measures of observer variability/kappa statistics). CMAJ 171: 1369–1373, 2004
16. Tordoir J, Canaud B, Haage P, Konner K, Basci A, Fouque D, Kooman J, Martin-Malo A, Pedrini L, Pizzarelli F, Tattersall J, Vennegoor M, Wanner C, ter Wee P, Vanholder R: EBPG on vascular access. Nephrol Dial Transplant 22[Suppl 2]: ii88–ii117, 2007
17. Paulson WD, Work J: Controversial vascular access surveillance mandate. Semin Dialysis 23: 92–94, 2010