International variation in classification of dialysis withdrawal: a systematic review.

BACKGROUND AND OBJECTIVES In patients with end-stage renal disease (ESRD), the rate of deaths preceded by dialysis withdrawal is high. However, rates vary across studies and national renal registries. This study aimed to (i) determine how dialysis withdrawal mortality is defined in the literature and (ii) whether mortality rates preceded by dialysis withdrawal change over time. METHODS MEDLINE (1946 to March 2012) and EMBASE (1980 to March 2012) databases were searched. We included epidemiological studies that reported data permitting calculation of crude (unadjusted) mortality rates preceded by dialysis withdrawal. Definitions of dialysis withdrawal were also extracted. Crude mortality rates and 95% confidence intervals were calculated using OpenEpi software. Non-English language studies were excluded. RESULTS Twenty-three eligible studies were identified; these included 14 527 885 dialysis patients at risk from six countries. Crude mortality rates preceded by dialysis withdrawal ranged from 3 to 50.2 per 1000 person-years. Seven different definitions of dialysis withdrawal were identified, with no assessment of validity. Crude mortality rates preceded by withdrawal have increased over time across the study period 1966 (3 per 1000 person-years) to 2010 (48.6 per 1000 person-years), although these rates are difficult to interpret because of differences in classification. In the USA crude mortality rates preceded by dialysis withdrawal are higher in the older population and have increased over time in the age group 65+ years. In this age group, the crude mortality rate preceded by dialysis withdrawal was 89.4 per 1000 person-years (2008-10) compared with 26.1 per 1000 person-years in the age group 50-64 years (2008-10). CONCLUSION Mortality rates preceded by dialysis withdrawal over time should be interpreted with caution because of differences in classification. Types of dialysis withdrawal need more careful elucidation, and we propose a unified classification of dialysis withdrawal based on trajectories and causal criteria.