Contextualizing the FHN Nocturnal Trial a Decade Later

The Frequent Hemodialysis Network Nocturnal Trial and Its Pitfalls

In April 2001, the National Institute of Diabetes and Digestive and Kidney Diseases convened a task force in Bethesda, Maryland, out of which the Frequent Hemodialysis Network (FHN) studies group emerged. A statistical model presented at the meeting suggested between 1622 and 3892 patients followed for 3 years would need to be enrolled to demonstrate a mortality difference; enrollment on this scale was not felt to be feasible.

The FHN consortium ultimately chose to study both daily and nocturnal dialysis in parallel but separate randomized controlled trials using a creative coprimary composite end point of mortality plus change in left ventricular mass and mortality plus change in the physical health composite score of the SF-36 (2). The requisite sample size of 250 patients would have had 80% power to detect reasonable changes in left ventricular mass and the physical health composite score after 1 year.

The center-based Short Daily trial achieved the recruitment target and demonstrated a benefit to more frequent HD (3). In contrast, the Nocturnal trial failed to reach statistical significance for either coprimary outcome (4). One possible explanation submitted by the authors for the lack of positive effect could be the greater proportion of participants on nocturnal HD who lost residual kidney function compared with conventional HD (5). Although compelling, especially given the association with residual kidney function and survival, this argument is predicated on accepting the Nocturnal trial as negative and that there are no other reasonable explanations for the loss in function.

With only 87 patients (of 118 ever recruited; 35% of the target sample size) ultimately randomized into the Nocturnal trial—and even then, only 72 had 12-month data for analysis—nocturnal HD would have needed to be nearly twice as effective as was hypothesized in order for the results to have been statistically significant—an optimistic expectation by any standard. This trial was not a negative study the way it is often portrayed; rather, it was an unsuccessful and inconclusive study because it was wholly underpowered for the primary outcome. Often overlooked is that nocturnal HD showed improvements in intermediate biochemical, hemodynamic, and quality of life outcome (6,7); even left ventricular mass was nearly significant with a P value of 0.09—all suggesting the generally positive effect of the modality.

The Nocturnal Trial Extension Study and Its Lessons

Published 4 years after the original investigation, the Nocturnal HD Extension study, with 5 years of follow-up, concluded that patients randomized to nocturnal HD had a higher mortality than those randomized to conventional dialysis (8). Case closed, right?

We suggest there are three critical, yet overlooked, aspects of the FHN Nocturnal trial worth re-examining. First, most patients randomized to conventional dialysis increased dialysis frequency to greater than thrice weekly, and many received >25 hours of dialysis per week after the initial 12 months of the trial. Given that the number of deaths in the extension phase far exceeded the number observed in the 12 months of the original study, it begs extreme caution attributing deaths to any given modality when the dialysis prescriptions of all patients became relatively homogeneous for a much longer period of time.

Second, there seems to be a pronounced difference in outcome depending on where patients received their care, with a mortality rate of 0.128 deaths per patient-year for those treated in the United States and a rate of 0.055 deaths per patient-year in Canada. Although not analyzed formally, it does raise the provocative question of whether the participating Canadian sites systematically conducted nocturnal HD differently than their American counterparts.

Third, centers were categorized as “experienced” with nocturnal HD if they had trained more than five patients for nocturnal HD prior to trial participation. This implies that a center having previously trained just six patients would be considered experienced. However, one could argue that meaningful experience with nocturnal HD might not be achieved until a program had trained 50, 60, or even 70 patients—certainly not five, 10, or 20. Thus, with respect to the Nocturnal trial, six sites (of nine participating centers) had virtually no real experience with nocturnal HD.

Keeping in mind that the logistics of performing home nocturnal HD with conventional dialysis machines (as were used in the FHN trials) involve a hundred or so steps (we have counted) and that there was certainly much variation in ancillary care that would never have been standardized between study sites, one can easily imagine dozens of cointerventions that could cumulatively affect outcomes. This is particularly worrisome considering that the majority of patients were American and that more than half of those were recruited from a single center. Thus, one center could have had a disproportionate influence (good or bad) on the trial, and even a small number of adverse events from inexperienced sites could have profoundly skewed results. For example, returning to the observation of more rapid loss of residual kidney function with nocturnal HD, for believers that the Nocturnal trial suggests increased mortality risk (we contend there is insufficient evidence to support this interpretation), this is often cited as a possible mechanism for that increased risk. However, could an overzealous emphasis on BP reduction have resulted in overly aggressive downward titration of postdialysis target weights and an accompanying loss of residual function, which has more to do with the approach of the prescriber rather than the modality itself?

We speculate that practice patterns and experience with home dialysis among participating centers were important and likely underappreciated determinants of the trial’s outcome. Drawing from the peritoneal dialysis experience, it is recognized that important clinical outcomes, such as peritonitis, can be mitigated with increased patient training time, greater nurse experience with peritoneal dialysis, and periodic patient visits to evaluate dialysis in the home setting. Why would processes of care, such as these and others, have any less bearing for nocturnal HD?

How It Is Done Matters

Is there any evidence that practice patterns in nocturnal HD influence outcomes? To address this question, let us consider nocturnal HD delivery in Canada. Many, if not all, of the early adopters of nocturnal HD in Canada visited Dr. Andreas Pierratos’s units in the late 1990s and early 2000s to learn from, adopt, and adapt his program’s protocols and procedures. Although an outsider may be excused for assuming that Canadian nocturnal HD delivery is homogeneous, there is, in fact, considerable heterogeneity in almost all aspects of home dialysis care provision across the country (9). This led to the hypothesis that home HD disposition is influenced not only by patient mix but also by the treatment center, and a detailed analysis of the seven largest and most established home HD programs in Canada revealed a significant center effect for both technique failure and mortality (10).

This highlights that processes of care between centers are important determinants of that care: where patients are evaluated, trained, and followed up makes a difference to their clinical course. One can view a center effect as the cumulative result of all program policies, procedures, resourcing, and philosophy of care that influence a patient’s experience. Interestingly, four of the seven centers included in this multicenter cohort study were also four of the nine centers involved in the Nocturnal trial, further underscoring the critical influence any given center may have had on the trial’s outcomes.

Disclosures

B.W. Miller reports consultancy agreements with Fresenius Medical Services; receiving honoraria from DaVita, Fresenius Renal Therapies Group, University of Southern California, and UpToDate; and serving as a scientific advisor or member of NxStage Medical Scientific Advisory Board. The remaining author has nothing to disclose.

Funding

None.