Can Peritoneal Dialysis Be Equivalent to (or Better Than) Optimal Hemodialysis?

tivariate analysis has the missing data bias due to absence of data on comorbidity and dialysis dose. The modality selection bias appears unlikely to be a factor in the three studies (1-3) which showed no difference in patient survival between modalities; 34-56% of patients were prescribed PD compared to only 14% in the report from the USA (4) indicating worse survival on PD. The censoring bias is present and favors PD in 2 of 3 studies reporting PD as equivalent to HD (1-3). It is less evident in the Bloembergen et al. (4) study as an intention to treat analysis ignores transfers and this would decrease the bias favoring PD. On the other hand, the censoring of transplantation would favor the modality with the lower transplant rate. The most important bias is the missing data bias. In no study was residual renal function or adequacy of dialysis included as an independent variable in the multivariate analysis. The importance of dialysis dose in patient survival in HD and PD has been well established (5, 6). Using data from the RKDP Minneapolis database for HD and from the Canada-USA Peritoneal Dialysis Study (6), incident patients were matched by age, diabetic status and by Kt/V according to the peak urea concentration hypothesis (7,s). For a PD weekly Kt/V of 1.7-2.1; the HD equivalent thrice weekly Kt/V was 1.0-1.5. For a PD weekly Kt/V > 2.1, the HD equivalent was > I S . Within each of these Kt/V ranges, patients matched for age and diabetic status had equivalent survival. While these data are also subject to selection bias (perhaps confounded by a country effect), censoring bias and missing data bias, the potential impact of adequacy of dialysis is clear. Until a randomized clinical trial with random allocation to HD and PD is performed, the question of effectiveness will remain unanswered. I believe that such a trial, while difficult, is feasible. Random allocation would remove selection bias. Randomization could be stratified for major comorbidity and residual renal function. The target dialysis dose should be equivalent, perhaps according to the peak urea concentration hypothesis (7). Differential transfer and transplantation rates would occur with an intention to treat analysis being a reasonable approach. Interactions between modality and other important variables [e.g., diabetes, age, compliance] are amenable to analysis in this research design.