Clinical outcome relative to the dose of dialysis is not what you think: The fallacy of the mean

Several recent retrospective studies of mortality relative to the dose of dialysis have been widely interpreted to indicate that adequate thrice-weekly hemodialysis requires a single pool Kt/V (spKt/V) of at least 1.4 to 1.6 and higher. In these studies, mortality rate has been correlated to the mean delivered spKt/V, (spKt/Vd)m, with coefficient of variation (CV) on the means ranging up to 45%. To evaluate these reported relationships, two large databases were analyzed using population constants to transform urea reduction ratio and spKt/Vd to equilibrated Kt/Vd (eKt/Vd), which expresses dose corrected for treatment time. The eKt/V dose (D) values were correlated to the reported relative risks (RR) of mortality to derive a RR/D function. The RR/D function, derived from these data with stepwise linear regression analysis, is nonlinear, with a steep linear increase in RR for eKt/Vd less than 1.05 and constant RR for eKt/Vd greater than or equal to 1.05. This RR/D function is mathematically expressed as RR = 1.96 - 1.03(eKt/ Vd) (equation 1) when 0.50 less than or equal to eKt/Vd less than or equal to 1.05, and RR = 0.88 (equation 2) when eKt/V greater than or equal to 1.05. We show that regression of RR on (eKt/Vd)m with large CV results in overestimation of RW relative to eKt/Vd for individual patients because of extrapolation of the linear relationship beyond the threshold where the slope becomes zero (see equation 2 above). It is concluded that (1) current clinical data indicate that adequate dialysis is provided with eKt/Vd of 1.0 to 1.1 on a thrice-weekly schedule, (2) it is essential to assure that all patients achieve this level of therapy, which is best accomplished using urea kinetic modeling for both prescription and measurement of delivered eKt/Vd, and (3) the current HEMO study is well designed to determine whether higher levels of eKt/Vd will further improve clinical outcome. (C) 1997 by the National Kidney Foundation, Inc.