The incident patient cohort study design with uncontrolled dose. Substantial over-estimation of mortality as a function of peritoneal dialysis dose?

In the Canada-USA (CANUSA) Study, the dialysis dose was neither randomized nor held constant, was measured at 6 month intervals, and the relative risk of mortality (R) was found to correlate linearly to mean values of weekly peritoneal plus renal urea clearance normalized to volume, (KprT/ V)m, ranging from 1.5 to 2.3. A risk/dose (R/D) function was derived for continuous ambulatory peritoneal dialysis from kinetic criteria for dose equivalency in hemodialysis (HD) and peritoneal dialysis (PD) and the HD R/D function. This PD R/D function was nonlinear with breakpoint from steep to shallow slope at (KprT/V)ud = 2.00, where ud refers to uniform single doses in contrast to mean doses with wide variances on the mean. The predicted decrease in renal urea clearance KrT/V per 6 months of CANUSA follow-up was computed from serial measured KrT/V in the Randomized Dialysis Prescription and Clinical Outcomes Study and showed it to be 0.21 +/- 0.34. The CANUSA (KprT/V)m values were corrected for the distributed values of 3 months decrements in KrT/V, and the population mortality risk at each (KprT/V)m dose level reported in CANUSA was computed from summation of the product of the R/D curve and fractional distribution of (KprT/V)ud values. From these calculations, the authors conclude that maximum (KprT/V)ud level achieved in CANUSA was 2.00, and the study does not define R/D response above this level.