Abstract Background and Aims Bicarbonate is delivered to hemodialysis (HD) patients via diffusion following a concentration gradient across the dialyzer membrane. However, the concentration of bicarbonate in the dialysis fluid ([HCO3]d) is usually not individualized for the specific patient's conditions, as no clear guidelines are provided, which leads to suboptimal pre- and post-dialytic plasma bicarbonate concentrations ([HCO3]p) during each treatment cycle. We used a mathematical model to predict the outcomes of a week-long HD cycle with time-dependent [HCO3]d in 24 patients, tuning the [HCO3]d prescription to keep the simulated pre-dialytic [HCO3]p in the range 22-24 mEq/L throughout the week. The [HCO3]d profiles thus developed were then tested, and [HCO3]p during HD compared with those using a constant [HCO3]d in the same patients. Method The compartmental model (Pietribiasi et al PLOS ONE 2023) describes the transport of total CO2, O2, and total buffer base in arterial and mixed-venous blood (plasma and red blood cells), tissues (interstitial and intracellular fluid), and lung capillaries and alveoli, calculating derivate quantities such as bicarbonate concentration, pH, partial pressure of CO2. The patients underwent one week of standard HD with constant [HCO3}d (Week 1, [HCO3]d = 33.6 ± 1.8 mEq/L) consisting of 3 sessions (HD1, HD2, HD3) to estimate the baseline model parameters and, successively, an intervention week (Week 2), during which the model-derived [HCO3]d profiles were implemented. Each profile consisted of a single step increase after 2 hours of HD, the two values of [HCO3]d chosen for each session and patient according to the model's simulations. Blood samples for the blood gas analyzer were taken before and after each session, and each hour during HD2. Results [HCO3]p profiles in Week 2 showed a more linear increase compared those of Week 1 (Fig. 1). The efficacy of the response to the model-based treatment was different depending on whether the intervention aimed to decrease or increase pre-dialytic [HCO3]p during Week 2 (Fig. 2). In the group of patients attempting to decrease pre-dialytic [HCO3]p (Decrease group, n = 10, average [HCO3]d = 31.3 ± 1.5 mEq/L), a significant fraction of sessions achieved the target (60% across the whole week). In contrast, among the patients attempting to increase pre-dialytic [HCO3]p (Increase group, n = 6, average [HCO3]d = 35.3 ± 2.0 mEq/L), the model-based prescription was effective only in increasing post-dialytic concentration (20% of sessions achieved pre-dialytic targets). By comparison, during Week 1 only 30% and 10% of sessions were in the target range for Decrease and Increase groups, respectively. Patients for which we didn't want to alter pre-dialytic [HCO3]p were termed the Neutral group (n = 8, average [HCO3]d = 33.3 ± 2.1 mEq/L, 40% and 50% of sessions in the target [HCO3]p range in Week 1 and 2, respectively). Conclusion The model-based profiling intervention was only partially successful in achieving the targets for pre-dialytic [HCO3]p in Week 2. The reasons why pre-dialytic [HCO3]p were resistant to the model-based [HCO3]d in the Increase group are not clear, but it could be that the increase in the average [HCO3]d was too small (6-7% higher than Week 1). However, in the Decrease group a similar difference was sufficient to produce a more marked decrease in pre-dialytic [HCO3]p. The profiling of [HCO3]d was shown to be effective in reducing the time the patients spend at higher [HCO3]p. Nevertheless, our study demonstrated that, upon refinement, mathematical modeling can be used to develop individualized treatments to pilot the markers of the acid-base response to HD towards desired values.
Read full abstract