Patient-specific modelling for hemodialysis therapy setting optimization

Hemodialysis (HD) is the most common therapy to treat the wide range of patients suffering from end stage renal disease. In the most of the cases it is the only applicable solution to keep patients alive, even though the impact on the patient’s cardio-vascular and hydro-electrolytic equilibrium is still nowadays of great relevance. It also has high socio-economic and environmental impact on health systems worldwide. In the last 30 years many technological improvements have been introduced in the hemodialysis treatment process: from enhancing the biocompatibility and purification performances of the filter membrane to on-line treatment monitoring by using biofeedback sensors, passing through the introduction of high-efficiency treatments. What bioengineering can still do to improve patients’ life quality and individual tolerance is working, in close cooperation with clinicians, on HD treatment customization. An example of these efforts is the DialysIS (“Dialysis therapy between Italy and Switzerland”) Project, whose results satisfactorily achieve the goal of treatment tailoring by means of development and sharing of optimised clinical protocols and introduction of decision support tools to improve dialysis prescription. In the framework of the mentioned project, this doctoral thesis proposes a set of integrated math- ematical tools to be introduced in clinical practice in order to better manage the patient-specific hydro-electrolyctic condition and prevent intra-dialysis complications. In addition, an in-vitro simulation platform of the HD patient has been proposed to study the patient-machine interaction. A multi-pool parametric kinetic model has been developed and later vali- dated, thanks to a great amount of clinical data acquired during DialysIS Project. It has been coupled with two different approach-based optimization algorithms, in order to properly identify the parameters defined as patient-specific in the mass and fluid transport phenomena involved in the HD process. The re- sults of the description and prediction capabilities evaluation of the model are very satisfactory. Indeed, the model shows a median error always lower than 4% when plasma electrolytes and catabolites and blood volume trends are de- scribed, while the median error is always lower than 6% when the same vari- ables are predicted, with overall reduced interquartile ranges. Furthermore, the compatibility between the mathematical artifice in identifying the personalised parameters and the physical explanation of the problem has been thoroughly investigated, by means of ad-hoc performed multiple linear regression analysis. Results of this has shown correlations between kinetic parameters identified by the mathematical algorithm and patient’s blood clinical measurements, confirm- ing the expectations. In some other cases, it has been found that the parameter values identified are strongly affected by the mathematical constraints. A brief focus on the most common intra-dialysis complication, namely intra- dialysis hypotension (IDH), has been carried out. The clinical applicability of a set of IDH definition criteria and of a multi-parametric algorithm to prevent IDH has been evaluated. These tools demonstrated to be reliable in managing IDH in advance, giving warnings that could reduce IDH onset of 54% (with a specificity of 99.6%), thanks to preventive, instead of corrective, clinical funda- mental interventions. Finally, an in-vitro simulation platform of the HD patient has been proposed to better investigate the patient-machine interaction. On one hand, the optimiza- tion and preliminary tests of a two-pool virtual patient simulator encourage its use to study the different patients’ vascular compensations mechanisms in re- sponse to HD treatment. On the other hand, the in-vitro simulation platform has allowed to perform some preliminary tests to develop a protocol for dialyser membrane efficiency evaluation as a function of blood protein content. This protocol could be applied to a wider range of current dialyser tests, in order to study protein adhesion and consequences on the overall mass and fluid transport across the dialyser. All these experimental tools could be then coupled with the mathematical model to achieve the best results from treatment customisation and improve HD therapy clinical prescription, as well as prevent HD related complications.

Modellazione paziente-specifica per ottimizzare la prescrizione della terapia emodialitica.