DESCRIPTION (provided by applicant): The average life-expectancy of end-stage renal disease patients starting maintenance dialysis in the US is about three years. Most clinical trials that tested a variety of potentially promising interventions have been unable to demonstrate a reduction in death risk of dialysis patients. The continued quest to improve outcomes has led to modifications of the conventional hemodialysis prescription to either significantly increase treatment time for each session or the frequency of therapy. The benefits on the co-primary composite of death risk or left ventricular mass increase seen in the recently concluded Frequent Hemodialysis Network (FHN) Daily Trial lends support to these modified prescriptions. However, the therapies being increasingly used in clinical practice differ from the interventions tested in the FHN trial. Nocturnal in-center hemodialysis (NICHD) provides longer treatment times but is generally delivered at a lower frequency (thrice weekly) than nocturnal home hemodialysis (5-6 times/ week) studied by FHN. Similarly, the most popular form of daily hemodialysis is performed at home with a device that is user-friendly but delivers lower solute clearances (short-daily, low-flow, home hemodialysis, SD-LF-HHD) than systems used in the FHN Daily Trial. NICHD or SD-LF-HHD patients cannot be identified in any publicly available data-source, including from the United States Renal Data System (USRDS) but this information is readily available in data from dialysis providers. In this project, we will obtain, refine, and ink data from DaVita, an organization that treats almost one-third of all US dialysis patients across 43 states, with the USRDS to examine outcomes of NICHD and SD-LF-HHD patients (n=2400, and 3500 respectively). The comparisons of these therapies with peritoneal dialysis and/or thrice-weekly conventional hemodialysis will be adequately powered for all-cause mortality, the primary outcome measure. The novel analytic strategy will use marginal structural models, a non-parametric causal model and will adjust for confounding from (1) baseline patient characteristics, (2) time-varying modality change, and (3) censoring for transplantation or drop-out. Confounding from site of care (TWICHD outcomes in facilities with/without NICHD programs) will be examined and to account for difficult-to-measure bias from patients who choose self-care home dialysis SD-LF-HHD outcomes will compared to PD, another home dialysis therapy. The high granularity of the linked data will allow us to study the association of NICHD and SD-LF-HHD with additional outcomes including cause-specific mortality, hospitalizations, solute clearances, hypertension, anemia, mineral metabolism, nutrition, dialysis tolerability, and vascular access morbidity. The DaVita-USRDS data linkage will provide access to Medicare claims data which will be used to calculate incremental societal cost-effectiveness or cost-savings with NICHD and SD-LF-HHD. Thus, this 3-year proposal will efficiently generate a wealth of time-sensitive information about two increasingly popular dialysis therapies that will be of immediate clinical and

public health relevance and inform decision making by physicians, patients, providers, and payers. The average life expectancy of patients with end-stage kidney disease starting maintenance dialysis in the United States is about three years and these patients spend 13 days in the hospital every year. Many more patients are being treated with hemodialysis therapy with either longer treatment times, or performed more frequently than thrice weekly. The overarching goal of this proposal is to determine whether these modifications to conventional hemodialysis treatments lead to improvements in life-expectancy, or reduction in hospitalizations for dialysis patients, and generate cost-savings for the healthcare system. __SpecificAimsTextDelimiter__ SPECIFIC AIMS AND HYPOTHESIS HYPOTHESIS: The overarching hypothesis to be tested during this 3-year project is that extra-corporeal dialysis therapies with either significantly longer treatment times (nocturnal in-center HD, NICHD) or higher frequency (short-daily, low-flow home hemodialysis, SD-LF-HHD) are associated with lower all- cause mortality, and beneficially influence other clinically relevant outcomes compared to conventional thrice- weekly in-center hemodialysis (TWICHD). Additionally, NICHD and SD-LF-HHD generate cost-savings to the healthcare system. SPECIFIC AIMS (Figure 1): Aim 1: Compare outcomes of contemporary HD patients treated with NICHD, a therapy with significantly longer treatment times than prescribed conventionally, with TWICHD in over 1600 DaVita dialysis facilities across the country over 3 years (7/1/08-6/30/11) with follow- up through 6/30/12. DaVita-United States Renal Data System (USRDS) linked data will be used to examine: A. Primary Outcome: All-cause mortality B. Secondary Outcomes: Cause- specific mortality, hospitalization, intermediate measures related to small solute clearances, blood pressure, anemia, mineral metabolism and nutrition, and tolerability of the dialysis procedure, and vascular access morbidity. C. Cost-Related Outcomes: Cost-effectiveness/cost-savings with NICHD compared to TWICHD. Additional analyses will examine the effect of heterogeneity of NICHD prescriptions and differences across racial/ethnic groups and select laboratory/clinical measures. Sub-Aim 1:To compare all-cause mortality of TWICHD patients treated in dialysis facilities with and without a NICHD program and test whether practice patterns account for these outcome differences. This will further elucidate and quantify the degree of facility-level confounding in the NICHD-TWICHD comparison. Aim 2: Compare above-mentioned outcomes in patients treated with SD-LF-HHD, therapy with significantly Aim One Patients Aim Two treated in DaVita 7/1/08 - 6/30/11 ¿ Control Gp followed ¿6/30/12 Thrice- Study Gp Weekly in- Nocturnal Primary Outcome: center HD In-Center All-cause mortality via non- HD parametric MARGNAL STRUCTURAL Study Gp MODEL weighted for censorship Short daily low-flow Control Secondary Outcomes: home HD Gp Thrice- Cause-specific mortality,hospitalizations, Weekly small solute clearances, BP, anemia, in-center mineral metabolism, nutrition, tolerance HD of dialysis procedure, vascular access Control Gp Home PD Incremental Cost-Effectiveness Figure 1: Proposed Study and Specific Aims. higher treatment frequency, with TWICHD; to account for difficult-to-measure bias when one comparator is self-care home dialysis that patients actively choose, comparisons will also be made to home PD. RATIONALE: The average life-expectancy of patients starting maintenance dialysis in the US is about 3 years. Controlled trials have failed to demonstrate a reduction in death risk of dialysis patients using different interventions. The continued quest to improve outcomes has led to modifications of conventional HD prescription to either significantly increase treatment time for each session or the frequency of treatment. The effect of these modified prescriptions was tested in the Frequent Hemodialysis Network (FHN) trials but they were not powered to detect mortality differences. Moreover, the therapies increasingly being used in clinical practice have substantial differences from the ones tested by FHN. Like nocturnal home HD, NICHD provides longer treatment times but is delivered in-center at a lower frequency. SD-LF-HHD is delivered more frequently but is performed at home using devices that have significantly lower solute clearances than tested by the FHN. Whether these therapies have the same effect on outcomes as the ones studied by FHN and if they lower death risk is not known. Individuals treated with NICHD or SD-LF-HHD cannot be identified by the USRDS but we propose to overcome this limitation by linking it with DaVita data from facilities across 43 states, an organization that treats over one-third of all dialysis patients in the country. This will generate highly granular data on 2,400 NICHD and 3,500 SD-LF-HHD patients over a contemporary 3-year period. The analyses will be adequately powered for detect mortality differences and will use robust methods to minimize baseline and time-varying confounding and from patient censoring for transplant/drop-out. This comparative effectiveness research builds upon the TWICHD-PD comparisons supported by a R21 grant undertaken by Drs. Mehrotra and Kalantar, co-PIs for this application. The proposal will efficiently generate a wealth of information about two increasingly popular dialysis therapies that will be of immediate relevance and inform future decision making.