Randomized Trials in Hemodialysis Patients: Title and subTitle BreakTime to Step Up to the Plate

Hemodialysis requires a reliable conduit to transport blood from the patient to the dialysis apparatus and back again, which is usually termed “vascular access.” Establishing and maintaining vascular access is time-consuming, difficult, and expensive—access creation and complications are the most common causes of hospital admissions in patients with end-stage renal disease,¹ at an estimated annual cost of more than $1 billion in the United States alone.² Accordingly, vascular access has been termed the “Achilles' heel” of hemodialysis.³

Current options for vascular access include central venous catheters, synthetic grafts, and native vessel arteriovenous fistulas. Once established, arteriovenous fistulas are associated with the best clinical outcomes and the lowest costs.^{4 - 10} However, unlike catheters and grafts, a substantial proportion of arteriovenous fistulas never mature sufficiently to be used for hemodialysis treatment, which is a major barrier to increasing the prevalence of their use. Although failure of arteriovenous fistulas to mature for use in dialysis often occurs in association with fistula thrombosis, thrombosis per se is generally asymptomatic and thus is not a clinically relevant outcome. On the other hand, while patency does not ensure that a fistula will be usable for dialysis, a clotted fistula clearly is incompatible with maturation, and therefore thrombosis is an attractive potential surrogate for the clinically relevant outcome of “failure to mature.”

In this issue of JAMA, Dember and colleagues¹¹ from the Dialysis Access Consortium report the results of a multicenter randomized trial comparing clopidogrel with placebo for the risk of early fistula failure. This report is the largest randomized vascular access trial, to my knowledge. The results have implications for clinical practice as well as for the design of future studies of hemodialysis patients.

The investigators enrolled 877 patients with established or incipient kidney failure who were undergoing arteriovenous fistula creation at 9 centers in the United States. Administration of study medication (clopidogrel or placebo) was initiated within 24 hours of fistula creation and was continued for 6 weeks. The primary outcome was thrombosis at 6 weeks after fistula creation, a putative surrogate for the main secondary outcome of suitability failure (another term for failure to mature), defined by the authors as failure to achieve sufficient flow rates to enable satisfactory dialysis. Suitability failure was assessed during a 30-day period commencing at least 120 days postoperatively. Because not all participants had kidney failure at the time of enrollment (and because the timing of dialysis initiation can be unpredictable), some participants did not require dialysis treatment during the study and thus only 758 participants could be assessed for the secondary outcome.

The investigators originally planned to include 1284 participants, but the trial was stopped after only 877 had been enrolled due to the findings of the fourth interim analysis, which showed that clopidogrel significantly reduced the risk of fistula thrombosis (relative risk, 0.63; 95% confidence interval, 0.46-0.97). However, the beneficial effect on the surrogate outcome did not translate into meaningful benefit for patients because the risk of nonmaturation was similar in both treatment groups (relative risk, 1.05; 95% confidence interval, 0.94-1.17 for clopidogrel vs placebo).

The early termination of the trial meant that the effect of therapy on the inability to use the fistula for dialysis (ie, failure to mature)—the clinically relevant outcome—could be assessed in less than 60% of the planned total enrollment. In hindsight, the decision to select a surrogate as the primary outcome (and to recommend stopping the trial based on findings for this surrogate rather than on a clinically meaningful alternative) could be viewed as an error in light of the discordant signals from the primary and secondary analyses. More specifically, it is important to consider whether stopping the trial early for “efficacy” adversely affected statistical power and led to a false-negative result for the effect of clopidogrel on fistula maturation.

Fortunately, it appears that the results of this trial are sufficiently robust to draw conclusions about both the primary and secondary outcomes. The risk of the inability to use the fistula for dialysis (ie, nonmaturation) was substantially higher than expected (60.6%), which increased the statistical power to detect an effect of clopidogrel on this outcome. Conditional power calculations found that completing the trial would not have led to a statistically significant benefit of clopidogrel on fistula nonmaturation. In fact, it was improbable (<3% probability) that even a nonsignificant benefit would have been found if the trial had been completed. Therefore, it seems unlikely that clopidogrel reduces the risk of arteriovenous fistula nonmaturation in patients who may require hemodialysis. Dember et al¹¹ indicate that the data and safety monitoring board for the trial considered these calculations before recommending early termination, suggesting that enrollment would have continued if findings for fistula nonmaturation had been more equivocal.

This important study has immediate implications for clinicians. Nephrologists can be reasonably confident that clopidogrel should not be prescribed for patients undergoing fistula creation because it does not lead to clinically meaningful benefit. In fact, routinely prescribing clopidogrel might actually be counterproductive because it might prolong the survival of immature fistulas that would otherwise have thrombosed, thus delaying referral for a repeat attempt at establishing vascular access. Whether clopidogrel improves patency in fistulas that have already matured is unknown. However, it seems likely that clopidogrel may not prove useful in this setting because thrombosis of an established fistula is almost always a sign of significant underlying stenosis requiring angioplasty.¹² Speculation aside, the findings of Dember et al suggest that clopidogrel cannot currently be recommended to improve maturation or patency of arteriovenous fistulas.

This study¹¹ also has important implications for researchers. First, it emphasizes the importance of studying clinically relevant outcomes (“an outcome that is relevant and noticeable to patients, and is therefore an outcome that they wish to avoid”).¹³ Relying on the analysis based on the primary (surrogate) outcome would have led to the erroneous conclusion that clopidogrel should be used routinely for patients undergoing fistula creation. Although it makes intuitive sense that preventing thrombosis will benefit patients, Dember et al¹¹ have shown that this is incorrect and demonstrated that thrombosis is not a valid surrogate for nonmaturation of the fistula. Rather, avoiding thrombosis is necessary but not sufficient for successful maturation of the fistula. In fact, some episodes of thrombosis may be clues to an underlying process that is responsible for both thrombosis and nonmaturation of the fistula. The higher than expected rate of fistula nonmaturation in the current study warrants further examination, and no doubt future reports from this study are planned. However, the relatively sparse data collected during the randomized trial may be insufficient to identify factors that predict which fistulas will develop successfully. Future studies should collect more detailed structural and functional information at the time of fistula creation and focus on identifying novel clinical and biological characteristics that influence the risk of fistula thrombosis and nonmaturation, as well as the link between these 2 adverse outcomes.

Second, the carefully considered decision to terminate the trial early was made in accordance with sound methodological principles, including correction for multiple comparisons, ensuring that the number of events and information fraction were adequate, and considering the results of conditional power calculations for the secondary outcome. These precautions ensured an informative result despite selection of a surrogate primary outcome. Given that trials terminated early for apparent benefit without due care often lead to biased results,¹⁴ future trialists studying hemodialysis patients would be wise to follow the example set by Dember et al¹¹ when planning how to perform interim analyses.

Third, randomized studies are difficult to conduct in hemodialysis patients, and recommendations for the management of this patient population often have been based on the results of observational studies—leading to several well-publicized examples in which the recommendations led to ineffective or even harmful therapies.^{15 - 17} Despite the critical importance of establishing and maintaining vascular access in hemodialysis patients, current practice in this area is generally based on weak or incomplete evidence. The rigorous design and execution of the large trial by Dember et al represents a key step toward closing this knowledge gap; the Dialysis Access Consortium has recently completed a separate trial of aspirin and extended-release dipyridamole vs placebo for the prevention of stenosis or thrombosis in synthetic dialysis grafts.¹⁸ The substantial investment by the National Institutes of Health to establish the Dialysis Access Consortium is commensurate with the clinical and economic importance of adequate hemodialysis vascular access.

Although important, vascular access is just one of many aspects of hemodialysis care in which there is little evidence to defend contemporary practices. Other recent trials conducted in dialysis patients have also failed to show a benefit of the experimental therapy despite strong a priori rationale,^{15 ,19 - 20} which has been discouraging for clinicians and may have contributed to a sense of therapeutic nihilism.²¹ However, this skepticism must not affect the resolve of the nephrology community to conduct and recruit to future trials (or the willingness of funding agencies to pay for them). “Negative” trials move the field forward by preventing needless use of unhelpful or possibly harmful treatments and by informing future research. Therefore, the real barrier to improving the care of dialysis patients is not the failure of past trials to find a statistically significant result, but rather the striking paucity of randomized studies.²²

The Dialysis Access Consortium investigators and others have demonstrated that it is possible to conduct good-quality randomized trials in hemodialysis populations. Hopefully, this will encourage the design, funding, and implementation of equally rigorous studies relevant for vascular access and other areas of hemodialysis care.