Modelling red blood cell survival data
Julia Korell(1), Frederiek Vos(1,2), Carolyn Coulter(1), John Schollum(2), Robert Walker(2), Stephen Duffull(1)
(1) School of Pharmacy, University of Otago, Dunedin, New Zealand. (2) Department of Medicine, Dunedin School of Medicine, University of Otago, Dunedin, New Zealand
Objectives: The survival time of red blood cells (RBCs) is commonly determined based on labelling experiments and an estimate of the mean RBC lifespan is obtained [1]. However, a better insight into the processes of RBC destruction would be desirable, especially in pathological states such as anaemia of chronic kidney disease (CKD), where the lifespan of RBCs is decreased [2] due to either an increase in random destruction or an accelerated senescence.
A previously developed model for RBC survival that accounts for plausible processes of RBC destruction [3,4] was applied to clinical data, and differences in the RBC lifespan in anaemic CKD patients compared to healthy controls were investigated.
Methods: RBC survival data using radioactive chromium as a random labelling method was available from 14 CKD patients receiving haemodialysis and 14 controls. The data were modelled based on two approaches: (1) a two-stage approach in MATLAB using generalized least squares; and (2) a full population approach in MONOLIX.
Two scenarios were considered: (1) estimating the main parameter controlling senescence; and (2) estimating the parameter controlling random destruction. An initial two-stage approach was conducted to assess whether a consistent preference towards one of these scenarios could be found across the individuals within each group. The goodness of fit between the scenarios was compared based on the OFV. Visual predictive checks (VPCs) were plotted for model evaluation.
Results: In the two-stage approach, the mean RBC survival was found to be significantly reduced by about 28% in CKD patients compared to healthy controls. Estimating the random destruction component provided a better fit for the majority of individuals (11 out of 14 in both groups).
In the full population approach, a combined error model described the data best. CKD was included as covariate in the full model, reducing between subject variability by 44% to 27% for the full model. The population approach confirmed the preference for estimating random destruction based on OFV in the whole population and that this mechanism was preferred to describe the decreased lifespan in CKD patients. The population mean RBC survival was 69.4 days for controls and 56.2 days which the covariate effect of CKD included, a reduction of approximately 19%.
Discussion: RBC survival in CKD patients was found to be decreased by approximately 20-30% compared to healthy individuals. The data support an increase in random destruction as the preferred underlying mechanism. Given the known shortfalls associated with the random labelling technique care should be taken when interpreting absolute values of RBC lifespan.
References:
[1.] International Committee for Standardization in Haematology (1980). Br. J. Haematol. 45(4):659-666
[2.] Loge J, Lange R, Moore C (1958). Am. J. Med. 24:4-18
[3.] Korell J, Coulter C, Duffull S (2011). J. Theor. Biol. 268(1):39-49
[4.] Korell J, Coulter C, Duffull S (2010). PAGE 19 Abstr 1701 [www.page-meeting.org/?abstract=1701], Berlin, Germany.
