Survival on home dialysis in New Zealand

Abstract

Background: New Zealand (NZ) has a high prevalence of both peritoneal dialysis (PD) and home haemodialysis (HD) relative to other countries, and probably less selection bias. We aimed to determine if home dialysis associates with better survival than facility HD by simultaneous comparisons of the three modalities.

Methods: We analysed survival by time-varying dialysis modality in New Zealanders over a 15-year period to 31-Dec-2011, adjusting for patient co-morbidity by Cox proportional hazards multivariate regression.

Results: We modelled 6,419 patients with 3,254 deaths over 20,042 patient-years of follow-up. Patients treated with PD and facility HD are similar; those on home HD are younger and healthier. Compared to facility HD, home dialysis (as a unified category) associates with an overall 13% lower mortality risk. Home HD associates with a 52% lower mortality risk. PD associates with a 20% lower mortality risk in the early period (<3 years) that is offset by a 33% greater mortality risk in the late period (>3 years), with no overall net effect. There was effect modification and less observable benefit associated with PD in those with diabetes mellitus, co-morbidity, and in NZ Maori and Pacific People. There was no effect modification by age or by era.

Conclusion: Our study supports the culture of home dialysis in NZ, and suggests that the extent and duration of survival benefit associated with early PD may be greater than appreciated. We are planning further analyses to exclude residual confounding from unmeasured co-morbidity and other sociodemographic factors using database linkage to NZ government datasets. Finally, our results suggest further research into the practice of PD in NZ Maori and Pacific People, as well as definitive study to determine the best timing for switching from PD in the late phase.

Figure 1. Kaplan Meier estimates of survival.

Abbreviations: HD, haemodialysis; PD, peritoneal dialysis.

Figure 2. Time dependence of effects Illustrated using -ln [-ln(survival)] versus ln(analysis time) plots (left) and scaled Schoenfeld residuals plots (right).

Plots for facility HD versus home dialysis are in the top panels, and for facility HD versus PD and home HD on the bottom.

Figure 3. Hazard ratios for mortality from the main effects model comparing facility HD versus home dialysis (top panels) and facility HD versus PD and home HD (bottom panels), fully adjusted for the confounders listed in Table 2 (the markers represent point estimates, and the whiskers 95% confidence intervals).

Estimates of effect for follow-up<3years are on the left (“Early period”); corresponding estimates for follow-up>3years are on the right (“Late period”). Abbreviations: BMI, body mass index; eGFR, estimated glomerular filtration rate; HD, haemodialysis; NZ, New Zealand; PD, peritoneal dialysis.