The Effects of Frequent Nocturnal Home Hemodialysis

The Frequent Hemodialysis Network Nocturnal Trial

Michael V. Rocco; Robert S. Lockridge Jr; Gerald J. Beck; Paul W. Eggers; Jennifer J. Gassman; Tom Greene; Brett Larive; Christopher T. Chan; Glenn M. Chertow; Michael Copland; Christopher D. Hoy; Robert M. Lindsay; Nathan W. Levin; Daniel B. Ornt; Andreas Pierratos; Mary F. Pipkin; Sanjay Rajagopalan; John B. Stokes; Mark L. Unruh; Robert A. Star; Alan S. Kliger


Kidney Int. 2011;80(10):1080-1091. 

In This Article

Materials and Methods

Study Setting

The FHN Nocturnal Trial was a multicenter, randomized, prospective trial of frequent home nocturnal hemodialysis sponsored by the National Institute of Health, National Institutes Diabetes, Digestive and Kidney Diseases (NIDDK), and the Center for Medicare and Medical Services (CMS). The design of the FHN Nocturnal Trial has been previously described.[19,21] Patients were enrolled between March 2006 and May 2009 and the trial concluded in May 2010. The study was approved by the local institutional review board at each participating site. An independent data safety monitoring board provided oversight.

Study Design

Study Population The inclusion and exclusion criteria are listed in Supplementary Table S1 online. Written informed consent was obtained from all participants.

Randomization Patients were randomized in a 1:1 ratio to either three times per week hemodialysis for <5 h per session (conventional hemodialysis) or six times per week hemodialysis for ≥6 h per session (frequent nocturnal hemodialysis). Randomization was performed centrally using random permuted blocks, stratified by clinical center and by diabetic status.

Interventions and Follow-up The weekly stdKt/Vurea was defined as the ratio of urea generation rate to the average predialysis urea concentration, adjusted for the number of hemodialysis treatments per week.[22] The delivered stdKt/Vurea was calculated by formal urea kinetic modeling as the ratio of the urea generation rate to the averaged predialysis blood urea nitrogen concentration with a correction for residual renal function.[23] Equilibrated eKt/V was computed using a modified Tattersall correction to the single pool Kt/V.[24] The residual renal function exclusion criterion was higher in the FHN Nocturnal Trial than in the FHN Daily Trial (>10 ml/min per 1.73 m2 as calculated as the average of the urea and creatinine clearances versus >3 ml/min per 1.73 m2 of urea clearance, respectively)[21] as it was anticipated that the dose separation would be greater in the Nocturnal Trial.

Patients in the conventional arm remained on their usual three times per week hemodialysis prescription subject to a prescribed eKt/V urea >1.1, a stdKt/V urea of >2.0, and a treatment time ≥2.5 h/session. Patients randomized to the frequent nocturnal arm followed dialysis prescriptions subject to a stdKt/V urea of ≥4.0 and a treatment time of ≥6 h, parameters designed to yield the maximum feasible dose using current dialysis technology.

Treatment parameters were monitored while subjects remained under the care of FHN centers. All treatment parameters were averaged over the first modeled dialysis within each follow-up month after 2 to 3 months. Adherence to the hemodialysis prescription was calculated as the ratio of the number of delivered hemodialysis treatments per month divided by the number of prescribed hemodialysis sessions per month.

All study participants were dialyzed using single-use high-flux dialyzers. A committee on standards of care, blinded to intervention, periodically reviewed and reported to clinical centers results of prespecified measures (phosphate, hemoglobin, bicarbonate, normalized protein nitrogen appearance, and blood pressure relative to achieved target postdialysis weight) that were outside of values recommended in published guidelines. Demographic, clinical, and laboratory data were obtained locally by site investigators and study coordinators. Additional data on missed dialysis sessions were obtained in both study arms on a prospective basis. Detailed information on the delivered dialysis prescription was obtained for all dialysis sessions that took place during 1 week of each follow-up month. Standardized assessments of comorbidity were obtained using a modification of the Charlson Index,[25] supplemented by additional items from the Index of Co-existing Disease Score.[26]


The two coprimary end points were: (1) death or 12-month change in LV mass (death/LV mass), and (2) death or 12-month change in the SF-36 RAND PHC (death/PHC).[27] We stipulated that demonstration of favorable effects on both coprimary outcomes would be interpreted as providing evidence of overall benefit. LV mass was measured by cardiac magnetic resonance imaging and evaluated in a blinded manner by a central reading center.[28] The SF-36 RAND PHC was obtained via blinded telephone interviews conducted by a central quality of life core. Cost and recruitment constraints precluded the possibility of examining survival or hospitalization rates with adequate statistical power. LV mass, however, has been shown to be an independent predictor of survival,[17] and studies in patients with end-stage renal disease[18] have found that a decrease in LV mass over time is associated with lower rates of death and cardiovascular events. In addition, cross-sectional values of self-reported physical health in end-stage renal disease patients correlate with mortality and hospitalization rates.[29]

Nine conceptually distinct therapeutic outcome domains were chosen to reflect the potential impact on multiple aspects of end-stage renal disease. A single outcome measure was considered to be the main secondary outcome for seven of these domains. These included the change from baseline to 12 months in LV mass, PHC score, Beck Depression Inventory,[30] serum albumin, Trail Making B,[31,32] predialysis serum phosphorus level, and erythropoiesis-stimulating agent dose. Other key secondary outcomes included the rate of non-access hospitalization or death, and, for hypertension, predialysis systolic blood pressure and the number of prescribed antihypertensive agents.

Data regarding serious adverse events, including hospitalizations and deaths, were collected prospectively and adjudicated in a blinded manner by an outcomes committee. Vascular access events were defined as access failures, infections requiring a procedure, thrombectomies, angioplasties, catheter replacements, and fibrin stripping of catheters. Hypotensive episodes were defined as the need for a lower ultrafiltration rate, reduced blood flow, or saline administration to ameliorate hypotension. The investigators could not be blinded to the patient's assigned intervention; however, investigators and patients remained blinded to outcome comparisons throughout the trial.

Vanguard Design and Protocol Changes

Because the feasibility of conducting a randomized trial comparing nocturnal home hemodialysis with conventional in-center hemodialysis was not known, the trial was initially conducted using a Vanguard design.[33] Recruitment, retention, and adherence were closely monitored during the initial year of the study to evaluate viability of the sample size targets and delivery of the intervention. Vanguard phase subjects were retained for final analyses.

During and after the Vanguard phase, several study design parameters were changed. The original sample size of 250 patients was estimated to provide 80% power to detect a 12-month change of 11.0 g in LV mass and a 4.2-point change in the SF-36 PHC. The Vanguard phase of the trial revealed that this goal was not feasible, and the target sample size was reduced to 150. The lower target sample size was justified in part by newly published data[15] suggesting that frequent nocturnal hemodialysis might reduce average LV mass by ~15 g, which could be detectable with slightly fewer than 150 patients. However, the sample size was ultimately reduced to 90 patients because of continued difficulties with recruitment. This smaller sample size allowed for the detection of only large effects in the two coprimary outcomes, with 80% power to detect a 19.6-g reduction in LV mass and a 7.4-point improvement in PHC.

In addition, the initial protocol specified that conventional arm patients receive in-center hemodialysis three times a week. Follow-up was 14 months to allow up to 2 months of in-center training for the nocturnal home hemodialysis patients, resulting in at least 12 months of follow-up in the nocturnal arm. Because of the difficulties with recruitment, a revised protocol was adopted in which all of the last 72 participants were first trained in home hemodialysis. Those patients randomized to the conventional arm received hemodialysis at home rather than in-center hemodialysis. Follow-up was also shortened to 12 months for the last 72 patients randomized under the revised protocol; the 14-month follow-up was retained for the first 15 patients.

Statistical Analyses

The two coprimary composite outcomes were analyzed using the Hochberg correction of the Bonferroni procedure,[34] with a studywise two-sided significance level of 0.05. Each of the two coprimary outcomes was analyzed using a rank-based nonparametric procedure. Patients who had died were ranked lowest, with the order of ranking determined by the duration of survival. Those who survived were ranked based on the change in the LV mass (or PHC) from baseline, with the ranking ordered from the most unfavorable change to the most favorable change. Patients were right censored at the time of transplantation or lost to follow-up; hence, patients who survived but did not provide 1-year LV mass or PHC measurements were credited with 1-year survival in the analysis. Ranks between the treatment arms were compared using the log rank test, and Cox regression was used to determine the associated HR and 95% CIs.

The analyses of the prespecified main secondary outcomes were performed on a comparison-wise basis without adjustment for multiple comparisons. Analyses of quantitative secondary outcomes were performed on the observed data without imputation of missing values by applying mixed-effects analyses using an unstructured covariance model to account for correlations in measurements over time,[35] with covariate adjustment for age, diabetic status, baseline level of the glomerular filtration rate, and the baseline variable under analysis and the interactions of these factors with time. These models were used to compare mean changes from baseline to month 12 between the treatment groups while incorporating values at baseline, 4 months (all but LV mass), and 12 months. The 4- and 12-month values were averaged from three monthly assessments (months 3–5 and 10–12) for predialysis levels of albumin, phosphorus, hemoglobin, and average weekly systolic blood pressure. In addition, treatment group comparisons of unadjusted mean changes are provided for patients completing their 12-month assessments for the LV mass and the PHC.

Certain modifications of this strategy were necessary for the analyses of erythropoiesis-stimulating agents, the Trail Making B, and the number of antihypertensive medications. Darbepoetin dose levels were converted to approximate equivalent erythropoietin dose using the expression erythropoietin dose (units)=250 × darbepoetin (mg).[36] The erythropoietin (or equivalent transformed darbepoetin) dose was set to a minimum of 5000 units per 4 weeks for patients using <5000 units, and log transformed before application of the mixed-effects analysis described above. The treatment effect on erythropoiesis-stimulating agents was expressed as the ratio of the geometric mean changes between the frequent nocturnal and conventional groups. Standard errors of the adjusted means for erythropoiesis-stimulating agents were computed using the δ-method. Treatment group comparisons for the number of antihypertensive agents and the Trail Making B (with patients failing to complete the Trail Making B assigned the lowest rank) were obtained using exact Wilcoxon rank-sum tests stratified by quartiles of the corresponding baseline values.

Time to death, first non-access hospitalization/death, and first access intervention were analyzed with Cox regression, controlling for diabetes, age, and baseline glomerular filtration rate. HRs and P-values comparing treatment arm event rates for multiple hospitalizations and vascular access events per patient were calculated using the Andersen–Gill model. Comparison of other adverse events between the two treatment arms was made using Fisher's exact test. All analyses were performed according to intent-to-treat principle and were performed using SAS version 9.2 (SAS Institute, Cary, NC).


Comments on Medscape are moderated and should be professional in tone and on topic. You must declare any conflicts of interest related to your comments and responses. Please see our Commenting Guide for further information. We reserve the right to remove posts at our sole discretion.
Post as: